Next Article in Journal
Automated Detection of Glacier Surges from Sentinel-1 Surface Velocity Time Series—An Example from Svalbard
Next Article in Special Issue
A New Method for Continuous Track Monitoring in Regions of Differential Land Subsidence Rate Using the Integration of PS-InSAR and SBAS-InSAR
Previous Article in Journal
A ViSAR Shadow-Detection Algorithm Based on LRSD Combined Trajectory Region Extraction
Previous Article in Special Issue
Detection and Evaluation of Flood Inundation Using CYGNSS Data during Extreme Precipitation in 2022 in Guangdong Province, China
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 15
Issue 6
10.3390/rs15061543
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

How to Systematically Evaluate the Greenspace Exposure of Residential Communities? A 3-D Novel Perspective Using UAV Photogrammetry

by
Tianyu Xia
,
Bing Zhao
,
Zheng Xian
and
Jinguang Zhang
*
The College of Landscape Architecture, Nanjing Forestry University, Nanjing 210037, China
*
Author to whom correspondence should be addressed.
Remote Sens. 2023, 15(6), 1543; https://doi.org/10.3390/rs15061543
Submission received: 7 January 2023 / Revised: 8 March 2023 / Accepted: 10 March 2023 / Published: 11 March 2023
(This article belongs to the Special Issue Intelligent Perception in Urban Spaces from Photogrammetry and Remote Sensing)
Browse Figures
Versions Notes

Abstract

:
The quantity and quality of green space (GS) exposure play an important role in urban residents’ physical and psychological health. However, the current framework for assessing GS quality is primarily based on 2-D remote sensing data and 2.5-D street-view images. Few studies have comprehensively evaluated residential community GSs from an overall 3-D perspective. This study proposes a novel systematic framework for evaluating the quantity and quality of residential GSs based on the generation of a high-resolution 3-D point cloud using Unmanned Aerial Vehicle (UAV)-digital aerial photogrammetry (DAP). Nine indices were proposed: green volume ratio, floor green volume index, green groups diversity index, vegetation diversity index, greenspace fragmentation, average vegetation colour distance, vegetation colour diversity, activity areas ratio, and green cohesion index of activity site. These metrics were calculated using the classified point clouds from four typical Chinese residential communities with different residential greenery types and population densities. The results showed that our method could quantitatively identify the differences in residential GS exposure within urban residential communities. For example, a residential community with a large plant distribution and rich greenery variations had higher greenspace volume ratio and vegetation diversity index values. Our findings suggest that this novel framework, employing cost-effective UAV-DAP, can clearly describe different GS attributes and characteristics, aiding decision-makers and urban planners in comprehensively implementing GS interventions to improve the residents’ quality of life.

1. Introduction

Urban green spaces (UGS) are virtual natural environments in cities [1]. Furthermore, UGS can provide many ecosystem services [2,3] and have a series of positive effects on the health of residents exposed to this greenery [4]. Green spaces (GS) can reduce exposure to air pollution, noise, and thermal pressure from environmental pressure sources [5], resulting in a lower incidence of cardiovascular disease [6] and a lower prevalence of chronic illness [7] among urban residents. In cities, green spaces are usually classified as structured (such as urban and forest parks) or unstructured (such as street-side green plants and backyards) [8]. Community green spaces are the most accessible green land for urban residents, especially those living in highly urbanised areas. Compared with large-scale urban parks and forests, community green spaces are usually small, with high pollution and noise levels, large population density, and low upper limit for green growth [9]. Therefore, increasing the physical quantity and quality of GS in communities has become an important challenge for urban planning. In community-oriented environmental planning promoting health and well-being, we should seek to increase the available exposure quantity and improve the exposure quality in all respects [10].
Current research defines green space exposure as the exposure of individuals (or people) to the natural environment [11]. There are subjective and objective evaluation methods for greenspace exposure [12]. Specifically, subjective exposure refers to the perceived greenness of residential communities obtained by observers through investigation and interviews [13,14]. The evaluation methods can be divided into three main categories. First, there is on-site investigation. Key indicators can be obtained through interviews for small- and medium-sized refined research designs [15]. Second, the Internet society census can directionally obtain evaluation data for medium- and large-scale research [16]. Third, extensive data crawling based on targeted evaluation crawling under Scrapy, PySider, Crawler, and other frameworks is generally sourced from social software and review apps within the research region [17]. This method is suitable for large-scale evaluations. However, network evaluation involves many uncertain factors, such as resource acquisition and technical facilities [18]. Objective exposure refers to the physical characteristics and visual exposure indicators of different dimensions of a GS [19]. Compared to subjective exposure, it can provide corresponding perceptual experiences according to different scales and needs. In most studies, objective exposure only measures GS from the overhead and eye level angles while capturing quality information on the green tricolour space [20]. First, the normalised difference vegetation index (NDVI) is a standard indicator for measuring the exposure of trees and plants [21,22]. The NDVI is a green timing index derived from Landsat satellite data. However, according to the literature, the NDVI poorly describes the human experience of contacting trees and plants because scenes with the same NDVI vary based on human observations [23]. As a result, many scholars suggest that previous research has not fully integrated measurement methods for open green spaces from a human-centred perspective.
Recently, the increasing abundance of street-level city images, such as those from Google Street View, Baidu Street View, Tencent Street View, Mapillary, and related street view evaluations, has gradually become a meaningful means of performing green exposure quality evaluations [24,25]. Combining the entire convolution network (FCN) visual image semantic segmentation and human-computer confrontation image scoring [26], the evaluation model combining street scene images and depth convolution network has achieved good results. Some studies have found that evaluation results based on street scene images are consistent with actual scene evaluation results [27]. However, using street views to evaluate the quality of community GS exposure still has limitations. First, street-view images are mostly captured along urban streets rather than communities. Additionally, many closed and semi-closed communities are in urban built-up areas [28]. Second, the time and space intervals between consecutive images often differ within this street region image system [29]. Additionally, regardless of the 2-D ordinary view or 2.5-D panoramic image, there is a lack of quantisation information in the vertical direction [30], thus ignoring the benefits of green exposure aboveground, which result in errors in the results.
Recently, with the development of light detection and laser (LiDAR) technology, we can more finely capture urban medium- and high-resolution 3-D environmental data [31]. Positioning and mobile LiDAR can provide larger data structures in spatial data collections and are increasingly used to describe and evaluate the configuration and composition of urban spaces [32,33]. However, site surveys are often limited by scale and region due to the high cost and limitations associated with location surveys, which hinders their wide application [34,35]. Therefore, a low-cost, efficient, and high-precision measurement system is necessary to evaluate the GS of urban communities.
With the development of computer vision and Unmanned Aerial Vehicle (UAV) technologies, digital aerial photogrammetry (DAP) is a cost-effective alternative for urban space measurement [36,37,38,39]. It can generate high-resolution 3-D true-colour point-cloud data based on digital images and aerial triangulation. Some research has shown that, under certain conditions, the UAV-DAP method is equivalent to or even better than UAV LiDAR [40,41] when measuring forest structures. However, as the UAV-DAP method is sensitive to light intensity and does not have the penetrability of airborne LiDAR [42], the digital terrain model (DTM), as generated under a dense forest canopy or near building surfaces, is not accurate. To overcome this problem, researchers have proposed a multi-angle DAP measurement method for UAVs to obtain ground information under a forest structure [43,44]. Additionally, for urban communities with high non-ground coverage, researchers have found that the measurement accuracy of UAV-DAP is higher than that of UAV LiDAR (R2 = 0.86–0.94, RMSE = 1.56–2.93 m) [45] in areas with either flat or slopping ground coverage (>0.95), whether flat or sloping.
Compared with 2-D and 2.5-D GS quality evaluation indicators, the 3-D perspective-based evaluation system can better reflect the characteristics of the site perceived by users [46]. Environmental quality must be quantified based on the user’s feelings and experiences. Some researchers [47] have used LiDAR point cloud data to derive 14 morphological factors by quantifying the shape of open spaces. They discussed the relationship between the morphological characteristics of open landscape spaces and subjective aesthetic preference. Wu [33] used mobile laser scanning to calculate seven famous key design elements from a point cloud, evaluated the internal quality of urban streets, and verified it using a random forest model trained by perceptual samples. According to the theory of landscape visualisation and LiDAR point cloud, Qi et al. [48] described three-dimensional (3-D) attributes of street landscapes based on the volume of eight different classifications (high vegetation, low vegetation, sky, building, street pole, wall, ground, and vehicle). These existing studies mainly use LiDAR to evaluate some visual factors of local areas to map the overall environment for landscape quality assessments. However, owing to limited equipment and environmental conditions, these methods cannot fully reveal the overall quality of GS in a large urban area (i.e., the visual quality of all positions in the area) [32]. Additionally, UAV LiDAR in high-density urban areas is often limited due to related legal flight issues and capital costs [49]. Therefore, the accuracy, efficiency, and economy of UAV-DAP render it a reasonable choice for measuring the green data of urban communities.
To compensate for this absence and promote the development and management of residential community GS, the main objective of this study is to build an assessment system for residential community GS exposure quantity and quality (GSQ_Q) based on the overall UAV-DAP scale. The objectives of this study were as follows: (1) to develop novel 3-D urban residential greenspace quantity and quality assessment metrics and (2) to use the UAV-DAP method to collect 3-D information on four residential communities in the Nanjing metropolitan area with different characteristics for indicator verification.

2. Methods

This study’s research framework consisted of three steps. First, we developed a set of 3-D space metrics based on UAV-DAP to quantify the spatial characteristics of urban communities and measure the space, structure, colour, and function of the GSQ_Q (Figure 1 shows the conceptual framework). According to current 2-D to 3-D research (Figure 2), nine evaluation indicators for GSQ_Q were proposed in four dimensions. Figure 3 shows the specific structure of the 3-D spatial measurement system. Second, we introduced the preparation and processing of UAV-DAP data, including the UAV parameter settings, as well as the registration, classification, and voxelization of the generated point cloud. Third, to verify the accuracy of the indicators at different community levels, we referred to the urban community classification standard in existing urban design literature [50], which adopts a classification system based on the quality of the building environment of the community (e.g., open communities with high building density and low vegetation covering and gated communities with low building density and good greenery and sports facilities). Based on this standard, we selected four typical urban communities located in the same city at different levels as the research area (Table 1), which cover almost all types of residential communities in China, extracted the regional GS characteristics using the UAV-DAP method, and calculated the GSQ_Q of the four communities.

2.1. 3-D Metrics

2.1.1. Spatiality

The volume ratio between the natural and artificial landscapes quantifies the spatial attributes of green exposure. In this study, a natural landscape was classified as having high and low vegetation; an artificial landscape was defined as buildings, roads, activity areas, and parking areas. The green group diversity index and average floor greening index were then introduced to represent the green exposure benefits of the ground and non-ground, respectively. The greenspace volume ratio [Equation (1)] reveals the relationship between high and low vegetation and the landscape volume. The product of the voxel size and number can be used to calculate the volume of the corresponding landscape classification. The sub-dimension of the green spatial clustering index generally includes four indicators: cluster size, cluster density, cluster number, and cluster distance.
In this study, we mainly discuss the diversity of clustering [Equation (2)] because it can better reflect the richness of green spaces at the site and thus map the quality of GS. Owing to multi-dimensional GS, the green exposure obtained by residents through windows was generally included in the calculation of non-ground GS at the site. We selected the total floor greening ratio to quantify the quality of non-ground GS. Considering the overlap of the green amount in the multi-angle calculation, we did not include the view angle and building interference in the calculation scope:
G V R = V h i g h − v e g   +   V l o w − v e g T V × 100 % ,
where G V R   is the greening volume ratio ( V h i g h − v e g and V l o w − v e g respectively represent the total volume of high and low vegetation at the site) and   T v is the sum of the volumes of all landscape elements at the site. At the spatial level, the representation of green diversity is reflected in the magnitude of differences in the size of community volumes.
σ 2 = ∑ i = 1 N ( x i − μ ) 2 N ,
where σ 2 is the diversity index of green groups, x i is the volume of the ith group of green groups, μ   is the average volume of green clusters in all groups, and N is the number of groups. The green exposure of the residents in the flats is captured by calculating the proportion of the volume of plants parallel to the window section of the block building.
F G V R = ∑ i = 1 t V i − v e g T i ,
where F G V R is the greening proportion of the total floor, V i − v e g is the greening volume of the i th floor, and T i is the total volume of the i th floor in the community. According to the Code for the Design of Windows in Residential Buildings, the sight range of the observer is limited; therefore, the calculated height for each floor was set as 1.5 m.

2.1.2. Structure

The landscape form structure reflects the ecological benefit and visual beauty of the landscape itself and indirectly reflects the user preferences of site residents. Two indices were used to quantify the shape index of GS: the vegetation diversity index (VDI) and the greenspace fragmentation index. First, we established a VDI regarding the landscape ecology indicators. The VDI reflects the heterogeneity of the vegetation by calculating the relationship between the actual 3-D surface area and the projected vegetation area [51,52]. Compared with the shape index based on the edge length and 2-D area, we used the 3-D object surface area and volume. When the landscape becomes more complex, the 3-D diversity should increase. When the surface tends to be flat, the index tends to be 2-D.
V D I = − ∑ k = 1 n P k ln ( P k ) ,
where P k is the 3-D projected area of the k-type plant community divided by the 3-D curved surface area.
In contrast, we used fragmentation to measure the isolation of plant groups. According to previous studies, dense foliage can reduce air pollution, increase the negative oxygen ion index, and reduce noise transmission [53,54]. Therefore, in this study, we used Lidar360 to classify and count specific data for the site’s plant crown width, diameter at breast height (DBH), and height. The number of tree and shrub groups at the site and the area of the 3-D surface were counted via cluster calculations. The fragmentation degree of the plant group distribution within the site was verified by calculating the ratio of the number of plant groups to the site’s surface area.
C i = N i A i
where C i is the degree of fragmentation of the GS,   N i is the number of all plant groups at the site, and A i represents the 3-D GS area at the site.

2.1.3. Colour

Colour is everywhere in environmental space; the enormous visual response caused by people acquiring external information is colour. Some studies have shown that a colour change is positively correlated with psychological perceptions. The degree of colour change in plant space also partially reflects the physiological and psychological benefits the site provides residents. In this study, we used colour changes between voxels to quantify the colour differences between plants. First, the colour distance between the voxels and baseline was calculated; the baseline was defined as black, RGB (0, 0, 0). The greening colour diversity index (GCDI) [48] was used to calculate the colour change of the vegetation. A high colour change indicates a significant difference in landscape colour, whereas a low change indicates a similar landscape colour.
C D i = Δ R i 2 + Δ G i 2 + Δ B i 2 ;
G C D I = ∑ i = 1 N p ( C D i − D a v e ) 2 N p ,
where C D i is the colour distance between the i th voxel and specific colour defined as RGB (0, 0, 0); Δ R i , Δ G i , Δ B i are the differences between the red, green, and blue values of the ith voxel and specific colour, respectively; G C D I is the greening colour change level; and D a v e is the average colour distance.

2.1.4. Function

There is a strong correlation between human activities and landscape perception; different people produce significantly different landscape evaluations for different forms and green quantities. From a biological perspective, perception reflects the landscape quality of human needs for survival and prosperity as a species (Tveit, 2006). Therefore, in urban communities, the activity areas set for leisure and sports activities are the main locations for daily outdoor activities and are an essential source of outdoor green exposure. Therefore, the green exposure level of the activity site significantly impacts the quality of life. In the point cloud classification step, we divided the residential activity area according to whether there were leisure seats and sports facilities in the classified site and counted the size and edge length of the site, as follows:
S R a = ∑ i = 1 n ( S i + S i + 1 + … + S n ) S t ,
where S R a refers to the total area ratio of the activity site, n refers to the number of activity sites within the scope of the measurement activities, and S t   represents the 3-D area of the study area. The greening cohesion index (GCI) [55] reflects the aggregation and dispersion of patches in a landscape, reflected in this study as the amount of green exposure that participants in the community activity space could contact:
G C I = [ 1 − ∑ j = 1 m P i j ∑ j = 1 m P i j a i j ] [ 1 − 1 A ] − 1 × 100 ,
where a i j is the surface area of the j th patch in the i -type landscape (m2), P i j is the edge perimeter of the j th patch in the i -type landscape (m), A is the total area of the landscape (hm2).

2.2. UAV-DAP Data Acquisition and Processing

Figure 4 illustrates the UAV data analysis process. First, the SFM algorithm in DJI Terra 3.5.5 generated point clouds based on dense images to obtain digital surface models (DSM) and orthogonal images. CloudCompare was then used to denoise and filter the point cloud generated by the DAP data from the UAV for registration. Lidar360 was used to segment the registered point cloud to obtain six categories (buildings, high plants, low plants, roads, activity areas, and ground parking areas). The ground activity area was divided manually in the later stage owing to problems with point cloud accuracy and definition. Third, we used the voxelization function in the PCL library to extract the spatial Cartesian coordinates and RGB values in the voxels generated by the classified point cloud using the editing code. Finally, we integrated the site information measured by the UAV DAP to calculate the difference between three-dimensional indicators in the four levels of communities.

2.2.1. UAV Data Collection

This study used a UAV remote sensing multi-angle measurement method to obtain site information. The UAV digital aerial photography images were processed using computer vision technology into high-resolution 3-D point cloud data. As the UAV-DAP method is more sensitive to lighting conditions, it does not have the penetration capability of laser radar. Therefore, we selected sunny and high-visibility weather conditions and adopted the multi-angle tilt photography method to more accurately measure the surface under the forest canopy or near buildings: the flight method for airborne five-direction inclinometer photogrammetry. The relevant flight parameters varied according to the survey location, among which the overlap of heading, sideways, and tilt photography were 70, 80, and 70%, respectively. The camera direction was parallel to the residential direction to avoid shooting dead angles. The GPS mode was switched to RTK mode to improve the accuracy of the shooting coordinates. Simultaneously, to further improve the accuracy, we arranged several image control points internally according to different sites. After the preparation was complete, the UAV automatically launched and performed the task and automatically landed after completing the task.
Finally, the UAV data were imported into DJI Terra for modelling and point cloud generation. The experiment obtained UAV data using the DJI Phantom4 RTK quadrotor UAV, equipped with an airborne RGB camera (20 million effective pixels), which meets the accuracy requirements for the GB/T7930-2008 1:500 topographic map aero-photogrammetry office specification. The diagonal length was 350 mm (13.8 in), and the take-off weight was 1391 g. This UAV can operate in environments between 0 and 40 °C. The maximum flight time of the Phantom4 RTK is 30 min. The entire operational area of a single flight is approximately 1 km2. Complete image data for the study area were obtained during multiple flights via route planning.

2.2.2. Point Cloud Data Processing

Figure 4 illustrates the processing flow of the point cloud data. The accuracy of the 3-D reconstruction model significantly affected the accuracy of the scene factor analysis. Therefore, in the first step of this study, we used the accuracy evaluation function in the CloudCompare open-source software to compare the point cloud model with ground truth. The four-point method was used to adjust the model size to a similar size for size registration; it sets the corresponding alignment points according to different model parameters. Furthermore, it uses Align to align the model while obtaining the offset matrix and root mean square between models. Finally, the distance between the point and face (moment to mesh) was calculated. The maximum offset distance was set not to exceed 1.0 m.
Second, to divide different landscape elements at the site, we assigned a semantic marker to each point in the initial point cloud and assigned similar or identical attributes (such as buildings, vehicles, roads, high vegetation, and low vegetation) to different point clusters. We divided the process into two steps: manual segmentation of the training dataset and machine classification based on the dataset.
Finally, to calculate the space and colour indicators of the point cloud, we used PCL point cloud down-sampling to achieve data voxelization. Voxelization simulates the geometric shape of point clouds through a voxel grid of uniform size. There are many methods to achieve model voxelization, such as the voxelization of a 3-D network model based on an octree. To obtain more accurate sampling points for the representation of the surface, we used the VoxelGrid filter to realize down-sampling, reduce the number of point clouds while maintaining the shape characteristics of the point cloud, divide the point cloud into multiple cubes according to different coordinates, and set the voxel size to 0.5 m × 0.5 m × 0.5 m.

2.3. Validation

This study used nine 3-D spatial indicators as examples to represent the GSQ_Q. The indices included the green volume ratio, plant clustering diversity index, and total green quantity index of floors. The VDI and green community fragmentation expressed the morphological structure characteristics. The green colour distance and colour diversity represented colour characteristics. The functional elements were defined by the area of residential activity and the green cohesion of the activity area. In Nanjing, China, we selected four residential communities at different levels with six types of common landscape elements (architecture, high plants, low plants, activity areas, roads, and parking areas) as community representatives. This study adopted the UAV-DAP method to collect 3-D point clouds containing site information and flexibly create 3-D models to adjust and verify indicators.

3. Experiments and Results

3.1. Study Area

In this study, we established four research areas (a, b, c, and d) in Nanjing, China, (located between 31°14′–32°37′N and 188°22′–199°14′E), a rapidly developing city in the Yangtze River Delta metropolitan area of China, as shown in Figure 5. These areas are located in the Xuanwu District of Nanjing, which is the core urban development area. The four areas are residential districts with relatively low-row buildings (non-high-rise buildings). The study area is approximately 115,274.26 m2, the population density is high, and the community environmental quality is widely affected by adjacent working and living groups. Oblique photography data from the four study areas were collected from 8:00 to 10:00 a.m. on 11–12 November 2022 (flight altitude: 100 m, GCPs:5). After shooting, the point cloud model was generated after DJI Terra calculated the aerial triangle data. Figure 5a–d shows the geographic reference point clouds of the four regions. The point clouds generated in study areas I, II, III, and IV were around 3.29 million, 8.44 million, 8.06 million, and 9.02 million, respectively. The average point density was 250 points/m2. The collected point cloud noise and outliers were removed using the CloudCompare V2.12.4 open-source software.

3.2. Results of Point Cloud Data Processing

Table 2 lists the registration results for the four study areas. The average offset distance of the point clouds in the four study areas compared to the ground truth model was 0.602 m (standard deviation = 0.513 m). The average migration distances of study areas I and II were low (0.469 and 0.450, respectively). The average offset distances of study areas III and IV were relatively high (0.576 and 0.786, respectively). The main error points were distributed in regions with significant changes in non-ground coverage elevation and insufficient DAP data coverage, such as the area adjacent to high-rise buildings at the edge of the survey site. After extracting the contour of the study area from the primary data, the current model offset distance was less than 0.1 m to meet the accuracy requirements. The current classification accuracy for the four dimensions of the evaluation indicators met our green exposure quality analysis requirements. Additionally, we evaluated and visualised a voxelized point cloud (Octrees = 11). The average offset distance between the PCL-filtered point cloud data and original data was 0.077 m (0.070, 0.078, 0.077, and 0.082 m, respectively, in the four study areas). The standard deviation was 0.054 m, similar to the initial point cloud results generated directly based on DAP.

3.3. Validation of 3-D Metrics

Figure 6 shows comparable 3-D spatial metrics for the four study sites. For non-ground green elements, the GS volumes of the four study areas were 34,537.36, 218,526.07, 449,956.7, and 176,489.62 m3, respectively. In terms of the scale measurement, owing to the different community sizes, we formulated thresholds (30 and 50%) for the GS proportion according to local policies to define the green quantity grades of the communities. In Baseline Study Area I, 19.84% of the total volume was covered by green (i.e., trees and shrubs). In contrast, although study area II had a higher green volume than study area IV, i.e., the larger site area and higher proportion of non-natural elements, the GS proportion was only 27.86% (the proportion of I in the study area was 67.86%).
For spatial characteristics, the green space volume ratio (GVR) considers the space volume of lower-level plants to a greater extent than the GS coverage. The results showed that the amount of green vegetation in study areas I and II (19.85 and 27.86%, respectively) was significantly lower than in study areas III and IV (66.32 and 67.86%, respectively). The overall green amount conformed to the configuration at the community level (Table 1), from low to high. The comprehensive ratio of floor green volume (FGVR) considers different indoor green volume acquisitions at the community level. Different physical attributes of vegetation in the study area affected the green exposure of floor residents.
The results showed that study areas III and IV had similar GVR values. However, the overall green amount level was high owing to differences in the physical attributes of the plants (e.g., tree age, tree species, crown width, branch point, and other factors). The indoor green exposure of the residents was significantly different (59.93 and 71.51%, respectively). In contrast, in study areas I and II (where the GVR values were different while the overall greenness level was low owing to the simplicity of green physical attributes and the small amount of vegetation), the difference in the amplitude of the floor greenness level performance was roughly similar to the overall GVR (16.39 and 22.89%, respectively). The diversity index of a green community is the difference in the scale of vegetation, which reflects the fairness associated with green access for residents near residential areas. The results showed that the diversity of communities in study areas I and II (0.791 and 0.693, respectively) was significantly higher than that in study areas III and IV (0.046 and 0.064, respectively). In study areas III and IV, owing to the different proportions of the plant space density in community parks (large plant communities), there was a large difference in the proportion of the total green amount. This was reflected in the lower difference index value of study area III compared with that of study area IV. This factor is difficult to account for in traditional landscape pattern index statistics.
Owing to the difference between the 3-D and 2-D indicators, the traditional indicator threshold based on remote sensing data measurement was no longer applicable to the actual structural characteristics. This study did not discuss the temporary setting of a threshold. The VDI reflects the richness of GS morphology and the diversity of vegetation species in the region. The results showed that the VDI of the four study areas increased, consistent with the grading setting of the study area. However, GS fragmentation showed the complexity of the green groups in the regional community dimension. As many artificially dominated non-natural landscape elements at the site, study area IV had a high degree of GS fragmentation despite a high abundance of plant groups.
For colour features, the colour distance of the colour indicators was mainly manifested in the lightness tendency of the colour. The larger the colour distance, the higher the brightness of the element and the more types of visible light in the included spectrum. The results showed that in voxels with green as the main tone, the average colour distance was affected more by the light transmittance at the site. This also reflected the outdoor permeability of the site from the side. In study areas I and II, with similar green community types, the average colour distance ( C D i ) of study area II, with a higher average building height, was higher than that of Study Area I owing to the higher light-blocking rate. Additionally, vegetation morphology and canopy density also affected the light transmittance of the site. There was a high plant canopy density in study area IV; therefore, the brightness of the space below was relatively low. The colour diversity shows the change in the colour range of a GS. Among the four study areas, study area IV had the highest greening colour diversity index with the highest plant diversity. The overall trend was consistent with the cascade division of the study area.
For the functional characteristics, the scale of the activity site was measured according to the proportion of each study area owing to the different sizes of the four areas. Except for the absence of activity areas in study area I, the other three activity areas were all set with outdoor activity areas. The highest proportion of active areas in study area III was 8.59%, while the lowest proportion in study area IV was 2.16%. Additionally, to measure the ability of residents in the activity area to obtain green exposure, we used green cohesion to measure the aggregation and dispersion of green exposure in the activity area. The results showed that the green exposure levels of the activity sites in the study areas were generally high. The GS cohesion of the three study areas with activity areas was roughly similar after standardisation (0.494, 0.503, and 0.502). We achieved residential community GSQ_Q level measurements based on the UAV-DAP 3-D space indicators. The results were consistent with the assumed direction, which showed the effectiveness and feasibility of the 3-D space indicators.

4. Discussion

In urban green exposure quality assessments, we resort to methods that quantify physical information, such as the spatial structure, in three dimensions, at a site that requires a comprehensive understanding. Therefore, in this study, we developed a set of UAV-DAP-based 3-D metrics to assess the quality of green exposure in urban communities. Unlike previous 2-D and 2.5-D approaches, we captured the overall 3-D spatial information of the site based on a high-resolution point cloud from the UAV view and divided the geometric structure of the GS to obtain a more comprehensive observation. In the following section, we compare the 3-D metrics of this study with those of existing studies and discuss our study’s practical implications and limitations.

4.1. Advantages of 3-D Metrics in Assessing GS Exposure Quality

Our metrics have several advantages over existing GS exposure assessment methods. In contrast to previous methods, which assess the green quality of a community from 2-D satellite [56] and 2.5-D street views [57], our method focused on the 3-D physical characteristics of the site’s GS. It analysed general information on site vegetation from multiple perspectives. Most previous 3-D metrics were calculated using a single viewpoint or side of the scanned data. For example, the method reported by Susaki [58] for calculating the vegetation greenspace ratio (GVR) from a single viewpoint and that reported by Wu [33] for the 3-D street visual quality based on human viewpoints.
Nevertheless, the outer environment of the natural community often includes observational data from multiple viewpoints. Especially for community residents, access to green exposure not only derives from outdoor activity areas and roads, but windows are another important means of exposure [59,60]. However, most contemporary studies are based on scene experiments and remote sensing images [61,62], which require a description of the physical characteristics of the green area outside the windows of natural communities. Acquisition of window images from community households is also problematic. In contrast, our floor green volume index calculates the unobserved part that is easily ignored in the scenario experiment and the neglected cut data in the remote sensing data from the UAV viewpoint to ensure the exclusive representation of green exposure from the window view while accurately calculating the specific proportion of GS to landscape space (Figure 7).
Second, at the level of greenspace representation, traditional 2-D analyses often ignore the vertical structure of plants and the roughness of surfaces [63,64]. We referred to the binary-style spatial segmentation method, as reported by Petras et al. [30], and the 3-D landscape pattern analysis method, as reported by Wu et al. [51], which aims to measure the diversity of the 3-D morphology of GS within the site. Furthermore, based on the structural level values obtained for the study area, the complexity of the 3-D surface area was significantly higher than that of the 2-D projected area. For example, the range of values for the VDI in the remote sensing calculations was often between zero and one. However, the indices in this study were significantly larger than those in that region. This was because the structural index in this study is a new metric based on the surface area of the entire landscape rather than the projected area.
Additionally, no studies have indicated community residents’ propensity for green spatial colour variables. However, some studies have demonstrated the spatial navigation effect of indoor colour, i.e., cool and warm environments, on the visual space of the elderly through virtual environment tests [65]. Santiago et al. [66] also verified a linear relationship between conspicuousness and colour differences in a fish test. In this study, we referred to the RGB colour metrics of Qi et al. [48]: mean chromatic aberration and colour diversity. The farther the mean colour difference of voxels in the GS from the baseline point (0, 0, 0), the higher the conspicuousness of the GS. Regarding colour diversity, Hoyle et al. [67] demonstrated that plant floral diversity positively enhances human aesthetics and invertebrate diversity. However, whether GS colour differences and colour diversity have physiological or psychological cues and feedback on humans must be verified by further studies.
Different functional sites provide different possibilities for residents to engage in various activities in the community. We referred to the NGST Neighborhood Green Space Tool [68] to categorise non-natural landscape areas as access functions (roads or parking) and recreation functions (open activity sites). Previous studies have shown that increased physical activity and social participation in urban GS is one means of promoting health [69]. Open activity spaces in urban communities are often the primary means of access to residents’ exposure. Furthermore, most forms of acquiring green exposure within communities use subjective exposure [13,14,15], in which feedback from residents on GS is obtained through interviews and questionnaires. Therefore, we must implement more realistic assessments of the objective exposure of activity spaces within communities. Thus, we obtained accurate spatial data for activity sites within the community by calculating the classified point clouds and clustering and dispersion of green patches in the activity area by cohesion (GCI) to assess the green exposure level of community activity spaces at an objective level.

4.2. Implications for Environmental Planning and Policies

In this paper, a systematic 3-D index framework was proposed to assess the quality of residential exposure to community GSs; this was verified by realistic scenarios and showed the validity and applicability of the index. We quantified landscape attributes and detailed features of the 3-D perspective using digital aerial surveys, proposing a theoretical, dimensional, and metric hierarchy framework to quantify GS physical characteristic representations as specific values, such as individual volume, surface area, and colour. This framework also allows users to target different exposure indices in the overall environment for queries from different evaluation dimensions. Additionally, we encourage the combination of these measured values with more landscape space evaluation metrics, such as the open space assessor tool (PSQUAT) and the community park assessor tool (CPAT) [70,71], to obtain more comprehensive green exposure assessments. In this manner, higher-quality urban community upgrades and GS renovations can occur.
We should also consider subjective exposure indices based on feedback from household perceptions as quantitative evaluation methods. We emphasise that the objective index should not replace mainstream subjective evaluation methods; both must be integrated and quantified in the implementation. Specifically, users sequentially collect site information, process point cloud data according to our design, and follow the site index to calculate the exposure quality. The final solution obtained was the ideal result. Meanwhile, the subjective evaluation of owners should be included in studies, produces an ideal result.
In addition, the 3-D GSQ_Q evaluation metrics serve as a tool that combines landscape with other multidisciplinary disciplines, aiming to achieve GSQ_Q feedback and links to people. The spatial attributes, structure, shape, and function of physical entities are mapped to 3-D point clouds to form high-resolution multi-physical volume models. However, the compatibility of the data in different dimensions still requires determination. For example, in the multisensory domain, mapping acoustic environmental data generated by vegetation communities in 3-D space is still under investigation [72]. Therefore, in this study, we integrated multiple metrics to create composite metrics, e.g., G V R and C i , which can only represent the metric level of their respective domains.
Finally, our approach significantly enhances urban planning and green space management. For instance, the green volume index for window views in urban communities can be easily obtained using our model, which is vital for calculating the green exposure of indoor occupants, as in the well-known exposure framework ‘3-30-300’ rule [73]. Our framework proposes a convenient way to quickly obtain the ‘3’ and ‘30’ indicators, which has implications for health-oriented GS exposure assessment.

4.3. Limitations and Future Studies

This study had some limitations. First, drone acquisition of point clouds in areas with high population density is demanding and risky. Owing to the requirement of site modelling resolution, there are specific requirements for the UAV camera image overlap rate and flight height. However, owing to the limitation of the observation angle, the shade remains challenging when capturing object features close to the tree canopy and under a building. According to Zhou [45], when the canopy or building coverage reaches a certain level (>60%), it is still difficult to generate accurate DTMs using DAP, regardless of the change in the observation angle. Therefore, in this study, some ground cover plants and shrubs should still be addressed when measuring GS exposure. To address this situation, we may need to combine ground-based mobile LiDAR and UVA-LiDAR for multiview data capture, select multiple angles for observation collection, and calculate more comprehensive measurements. Second, the design of some metrics (e.g., FGVR and GCI) must be further refined according to different scenarios. For example, in FGVR design, the design view of the floor’s green volume is based on the human viewpoint parallel to the ground. Considering that the distance a person is from a window affects the sight range, we further need to explore the indoor activity trajectory of community residents to obtain more accurate green exposure acquisition. Third, our study used cross-sectional data with specific spatial-temporal characteristics. For example, vegetation has a more pronounced colour tendency in autumn than in other seasons. Therefore, calculating the attribute weights of sites using tracking studies is necessary. Fourth, the nine essential elements do not represent additional effects other than the metric level that expresses the respective dimensions. Finally, we will use less costly data sources (e.g., city information models) and more efficient evaluation processes (e.g., machine learning) in future study.

5. Conclusions

This study is unique in that it adopts a systematic framework to comprehensively estimate residential GSQ_Q, introducing nine metrics to quantify GSs spatial, structural, colour, and functional characteristics. In addition, the UAV-DAP point cloud is introduced as novel data for quantity and quality evaluation, taking advantage of the fact that 3-D data can change any angle and position in the virtual environment, surpassing traditional fixed-view data limitations. Finally, according to the calculation results of four distinct representative communities, we verified that the index can be applied to a wide variety of urban communities based on satisfying the accuracy of 3-D point clouds or models. The high-resolution green spaces depicted in our study using UAV-DAP provide a relatively objective measurement tool for urban community planning and management. This has substantial implications for studying the effects of environmental exposure on residents’ physiological and psychological health.

Author Contributions

T.X.: conceptualisation, methodology, data collection, data curation, figure production, writing—original draft, writing—review, and editing. B.Z.: project administration, supervision, funding acquisition. Z.X.: data curation. J.Z.: Conceptualisation, methodology, writing—review, and editing. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Natural Science Foundation of Jiangsu Province (No. BK20220410); Humanity and Social Science Youth foundation of Ministry of Education of China (No. 22YJCZH237); Natural Science Research of Jiangsu Higher Education Institutions of China (No. 1020221108); Philosophical and Social Science Foundation of Jiangsu Universities (No. 2022SJYB0162); and Priority Academic Program Development of Jiangsu Higher Educations Institutions (No. 164120230).

Data Availability Statement

The data are not publicly available due to the Regulation on the Administration of the Public Use of Remote Sensing Images.

Acknowledgments

We are grateful to the Landscape Architecture Department of Nanjing Forestry University for supporting our research.

Conflicts of Interest

The authors declare no conflict of interest.

Appendix A

Remotesensing 15 01543 g0a1 550
Figure A1. Validation result of the FGVR (lateral view).
Figure A1. Validation result of the FGVR (lateral view).
Remotesensing 15 01543 g0a1

References

  1. Wolch, J.R.; Byrne, J.; Newell, J.P. Urban green space, public health, and environmental justice: The challenge of making cities ‘just green enough’. Landsc. Urban Plan 2014, 125, 234–244. [Google Scholar] [CrossRef] [Green Version]
  2. Guilland, C.; Maron, P.A.; Damas, O.; Ranjard, L. Biodiversity of urban soils for sustainable cities. Environ. Chem. Lett. 2018, 16, 1267–1282. [Google Scholar] [CrossRef]
  3. Matos, P.; Vieira, J.; Rocha, B.; Branquinho, C.; Pinho, P. Modeling the provision of air-quality regulation ecosystem service provided by urban green spaces using lichens as ecological indicators. Sci. Total Environ. 2019, 665, 521–530. [Google Scholar] [CrossRef] [PubMed]
  4. Handy, S.L.; Boarnet, M.G.; Ewing, R.; Killingsworth, R.E. How the built environment affects physical activity: Views from urban planning. Am. J. Prev. Med. 2002, 23, 64–73. [Google Scholar] [CrossRef] [PubMed]
  5. Dadvand, P.; Rivas, I.; Basagaña, X.; Alvarez-Pedrerol, M.; Su, J.; De Castro Pascual, M.; Amato, F.; Jerret, M.; Querol, X.; Sunyer, J.; et al. The association between greenness and traffic-related air pollution at schools. Sci. Total Environ. 2015, 523, 59–63. [Google Scholar] [CrossRef]
  6. Yao, Y.; Xu, C.; Yin, H.; Shao, L.; Wang, R. More visible greenspace, stronger heart? Evidence from ischaemic heart disease emergency department visits by middle-aged and older adults in Hubei, China. Landsc. Urban Plan 2022, 224, 104444. [Google Scholar] [CrossRef]
  7. Gascon, M.; Zijlema, W.; Vert, C.; White, M.P.; Nieuwenhuijsen, M.J. Outdoor blue spaces, human health and well-being: A systematic review of quantitative studies. Int. J. Hyg. Environ. Health 2017, 220, 1207–1221. [Google Scholar] [CrossRef]
  8. Markevych, I.; Schoierer, J.; Hartig, T.; Chudnovsky, A.; Hystad, P.; Dzhambov, A.M.; de Vries, S.; Triguero-Mas, M.; Brauer, M.; Nieuwenhuijsen, M.J.; et al. Exploring pathways linking greenspace to health: Theoretical and methodological guidance. Environ. Res. 2017, 158, 301–317. [Google Scholar] [CrossRef]
  9. Lo, A.Y.H.; Jim, C.Y. Differential community effects on perception and use of urban greenspaces. Cities 2010, 27, 430–442. [Google Scholar] [CrossRef] [Green Version]
  10. Wang, R.; Feng, Z.; Pearce, J.; Liu, Y.; Dong, G. Are greenspace quantity and quality associated with mental health through different mechanisms in Guangzhou, China: A comparison study using street view data. Environ. Pollut. 2021, 290, 117976. [Google Scholar] [CrossRef]
  11. Bratman, G.N.; Anderson, C.B.; Berman, M.G.; Cochran, B.; de Vries, S.; Flanders, J.; Folke, C.; Frumkin, H.; Gross, J.J.; Hartig, T.; et al. Nature and mental health: An ecosystem service perspective. Sci. Adv. 2019, 5, eaax0903. [Google Scholar] [CrossRef] [Green Version]
  12. James, P.; Banay, R.F.; Hart, J.E.; Laden, F. A Review of the Health Benefits of Greenness. Curr. Epidemiol. Rep. 2015, 2, 131–142. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  13. Groenewegen, P.P.; van den Berg, A.E.; Maas, J.; Verheij, R.A.; de Vries, S. Is a Green Residential Environment Better for Health? If So, Why? Ann. Assoc. Am. Geogr. 2012, 102, 996–1003. [Google Scholar] [CrossRef]
  14. White, M.P.; Pahl, S.; Ashbullby, K.; Herbert, S.; Depledge, M.H. Feelings of restoration from recent nature visits. J. Environ. Psychol. 2013, 35, 40–51. [Google Scholar] [CrossRef]
  15. Hipp, J.A.; Gulwadi, G.B.; Alves, S.; Sequeira, S. The Relationship between Perceived Greenness and Perceived Restorativeness of University Campuses and Student-Reported Quality of Life. Environ. Behav. 2015, 48, 1292–1308. [Google Scholar] [CrossRef]
  16. Feng, Z.; Li, Q.; Zhou, L.; Chen, Z.; Yin, W. The relationship between depressive symptoms and activity of daily living disability among the elderly: Results from the China Health and Retirement Longitudinal Study (CHARLS). Public Health 2021, 198, 75–81. [Google Scholar] [CrossRef]
  17. Wang, Z.; Zhu, Z.; Xu, M.; Qureshi, S. Fine-grained assessment of greenspace satisfaction at regional scale using content analysis of social media and machine learning. Sci. Total Environ. 2021, 776, 145908. [Google Scholar] [CrossRef]
  18. Manley, K.; Nyelele, C.; Egoh, B.N. A review of machine learning and big data applications in addressing ecosystem service research gaps. Ecosyst. Serv. 2022, 57, 101478. [Google Scholar] [CrossRef]
  19. Zhang, J.; Cheng, Y.; Li, H.; Wan, Y.; Zhao, B. Deciphering the changes in residential exposure to green spaces: The case of a rapidly urbanizing metropolitan region. Build. Environ. 2021, 188, 107508. [Google Scholar] [CrossRef]
  20. Zeng, P.; Sun, F.; Liu, Y.; Tian, T.; Wu, J.; Dong, Q.; Peng, S.; Che, Y. The influence of the landscape pattern on the urban land surface temperature varies with the ratio of land components: Insights from 2D/3D building/vegetation metrics. Sustain. Cities Soc. 2022, 78, 103599. [Google Scholar] [CrossRef]
  21. Twohig-Bennett, C.; Jones, A. The health benefits of the great outdoors: A systematic review and meta-analysis of greenspace exposure and health outcomes. Environ. Res. 2018, 166, 628–637. [Google Scholar] [CrossRef]
  22. Madzia, J.; Ryan, P.; Yolton, K.; Percy, Z.; Newman, N.; LeMasters, G.; Brokamp, C. Residential Greenspace Association with Childhood Behavioral Outcomes. J. Pediatr. 2019, 207, 233–240. [Google Scholar] [CrossRef] [PubMed]
  23. Donovan, G.H.; Gatziolis, D.; Derrien, M.; Michael, Y.L.; Prestemon, J.P.; Douwes, J. Shortcomings of the normalized difference vegetation index as an exposure metric. Nat. Plants 2022, 8, 617–622. [Google Scholar] [CrossRef]
  24. Li, X.; Zhang, C.; Li, W.; Ricard, R.; Meng, Q.; Zhang, W. Assessing street-level urban greenery using Google Street View and a modified green view index. Urban For. Urban Green. 2015, 14, 675–685. [Google Scholar] [CrossRef]
  25. Long, Y.; Liu, L. How green are the streets? An analysis for central areas of Chinese cities using Tencent Street View. PLoS ONE 2017, 12, e0171110. [Google Scholar] [CrossRef] [Green Version]
  26. Yao, Y.; Liang, Z.; Yuan, Z.; Liu, P.; Bie, Y.; Zhang, J.; Wang, R.; Wang, J.; Guan, Q. A human-machine adversarial scoring framework for urban perception assessment using street-view images. Int. J. Geogr. Inf. Sci. 2019, 33, 2363–2384. [Google Scholar] [CrossRef]
  27. Liu, L.; Silva, E.A.; Wu, C.; Wang, H. A machine learning-based method for the large-scale evaluation of the qualities of the urban environment. Comput. Environ. Urban Syst. 2017, 65, 113–125. [Google Scholar] [CrossRef]
  28. Hebbalaguppe, R.; Garg, G.; Hassan, E.; Ghosh, H.; Verma, A. Telecom Inventory Management via Object Recognition and Localisation on Google Street View Images. In Proceedings of the 2017 IEEE Winter Conference on Applications of Computer Vision (WACV), Santa Rosa, CA, USA, 24–31 March 2017; pp. 725–733. [Google Scholar]
  29. Liu, D.; Jiang, Y.; Wang, R.; Lu, Y. Establishing a citywide street tree inventory with street view images and computer vision techniques. Comput. Environ. Urban Syst. 2023, 100, 101924. [Google Scholar] [CrossRef]
  30. Petras, V.; Newcomb, D.J.; Mitasova, H. Generalized 3D fragmentation index derived from lidar point clouds. Open Geospat. Data Softw. Stand. 2017, 2, 9. [Google Scholar] [CrossRef] [Green Version]
  31. Underwood, J.P.; Hung, C.; Whelan, B.; Sukkarieh, S. Mapping almond orchard canopy volume, flowers, fruit and yield using lidar and vision sensors. Comput. Electron. Agric. 2016, 130, 83–96. [Google Scholar] [CrossRef]
  32. Hu, T.; Wei, D.; Su, Y.; Wang, X.; Zhang, J.; Sun, X.; Liu, Y.; Guo, Q. Quantifying the shape of urban street trees and evaluating its influence on their aesthetic functions based on mobile lidar data. ISPRS J. Photogramm. Remote Sens. 2022, 184, 203–214. [Google Scholar] [CrossRef]
  33. Wu, B.; Yu, B.; Shu, S.; Liang, H.; Zhao, Y.; Wu, J. Mapping fine-scale visual quality distribution inside urban streets using mobile LiDAR data. Build. Environ. 2021, 206, 108323. [Google Scholar] [CrossRef]
  34. Ren, Z.; Zheng, H.; He, X.; Zhang, D.; Yu, X.; Shen, G. Spatial estimation of urban forest structures with Landsat TM data and field measurements. Urban For. Urban Green. 2015, 14, 336–344. [Google Scholar] [CrossRef]
  35. Isibue, E.W.; Pingel, T.J. Unmanned aerial vehicle based measurement of urban forests. Urban For. Urban Green. 2020, 48, 126574. [Google Scholar] [CrossRef]
  36. Dandois, J.P.; Ellis, E.C. High spatial resolution three-dimensional mapping of vegetation spectral dynamics using computer vision. Remote Sens. Environ. 2013, 136, 259–276. [Google Scholar] [CrossRef] [Green Version]
  37. Panagiotidis, D.; Abdollahnejad, A.; Surový, P.; Chiteculo, V. Determining tree height and crown diameter from high-resolution UAV imagery. Int. J. Remote Sens. 2017, 38, 2392–2410. [Google Scholar] [CrossRef]
  38. Cucchiaro, S.; Fallu, D.J.; Zhang, H.; Walsh, K.; Van Oost, K.; Brown, A.G.; Tarolli, P. Multiplatform-SfM and TLS data fusion for monitoring agricultural terraces in complex topographic and landcover conditions. Remote Sens. 2020, 12, 1946. [Google Scholar] [CrossRef]
  39. Ganz, S.; Käber, Y.; Adler, P. Measuring tree height with remote sensing—A comparison of photogrammetric and LiDAR data with different field measurements. Forests 2019, 10, 694. [Google Scholar] [CrossRef] [Green Version]
  40. Diaz-Varela, R.A.; Zarco-Tejada, P.J.; Angileri, V.; Loudjani, P. Automatic identification of agricultural terraces through object-oriented analysis of very high resolution DSMs and multispectral imagery obtained from an unmanned aerial vehicle. J. Environ. Manag. 2014, 134, 117–126. [Google Scholar] [CrossRef]
  41. Surový, P.; Almeida Ribeiro, N.; Panagiotidis, D. Estimation of positions and heights from UAV-sensed imagery in tree plantations in agrosilvopastoral systems. Int. J. Remote Sens. 2018, 39, 4786–4800. [Google Scholar] [CrossRef]
  42. Alonzo, M.; Andersen, H.-E.; Morton, D.C.; Cook, B.D. Quantifying boreal forest structure and composition using UAV structure from motion. Forests 2018, 9, 119. [Google Scholar] [CrossRef] [Green Version]
  43. Kimes, D.; Ranson, K.; Sun, G.; Blair, J. Predicting lidar measured forest vertical structure from multi-angle spectral data. Remote Sens. Environ. 2006, 100, 503–511. [Google Scholar] [CrossRef]
  44. Wang, Q.; Ni-Meister, W. Forest canopy height and gaps using BRDF index assessed with airborne lidar data. Remote Sens. 2019, 11, 2566. [Google Scholar] [CrossRef] [Green Version]
  45. Zhou, L.; Meng, R.; Tan, Y.; Lv, Z.; Zhao, Y.; Xu, B.; Zhao, F. Comparison of UAV-based LiDAR and digital aerial photogrammetry for measuring crown-level canopy height in the urban environment. Urban For. Urban Green. 2022, 69, 127489. [Google Scholar] [CrossRef]
  46. Hoechstetter, S.; Thinh, N.X.; Walz, U. 3D-indices for the analysis of spatial patterns of landscape structure. Proc. InterCarto–InterGIS 2006, 12, 108–118. [Google Scholar]
  47. Wang, Y.; Zlatanova, S.; Yan, J.; Huang, Z.; Cheng, Y. Exploring the relationship between spatial morphology characteristics and scenic beauty preference of landscape open space unit by using point cloud data. Environ. Plan. B Urban Anal. City Sci. 2020, 48, 1822–1840. [Google Scholar] [CrossRef]
  48. Qi, J.; Lin, E.S.; Yok Tan, P.; Chun Man Ho, R.; Sia, A.; Olszewska-Guizzo, A.; Zhang, X.; Waykool, R. Development and application of 3D spatial metrics using point clouds for landscape visual quality assessment. Landsc. Urban Plan 2022, 228, 104585. [Google Scholar] [CrossRef]
  49. Stöcker, C.; Bennett, R.; Nex, F.; Gerke, M.; Zevenbergen, J. Review of the Current State of UAV Regulations. Remote Sens. 2017, 9, 459. [Google Scholar] [CrossRef] [Green Version]
  50. Zhang, J.; Yu, Z.; Cheng, Y.; Chen, C.; Wan, Y.; Zhao, B.; Vejre, H. Evaluating the disparities in urban green space provision in communities with diverse built environments: The case of a rapidly urbanizing Chinese city. Build. Environ. 2020, 183, 107170. [Google Scholar] [CrossRef]
  51. Wu, Q.; Guo, F.; Li, H.; Kang, J. Measuring landscape pattern in three dimensional space. Landsc. Urban Plan 2017, 167, 49–59. [Google Scholar] [CrossRef]
  52. Sapena, M.; Wurm, M.; Taubenböck, H.; Tuia, D.; Ruiz, L.A. Estimating quality of life dimensions from urban spatial pattern metrics. Comput. Environ. Urban Syst. 2021, 85, 101549. [Google Scholar] [CrossRef]
  53. Tikhonov, V.P.; Tsvetkov, V.D.; Litvinova, E.G.; Sirota, T.V.; Kondrashova, M.N. Generation of Negative Air Ions by Plants upon Pulsed Electrical Stimulation Applied to Soil. Russ. J. Plant Physiol. 2004, 51, 414–419. [Google Scholar] [CrossRef]
  54. Ye, T.; Guo, Y.; Abramson, M.J.; Li, T.; Li, S. Greenspace and children’s lung function in China: A cross-sectional study between 2013 and 2015. Sci. Total Environ. 2023, 858, 159952. [Google Scholar] [CrossRef] [PubMed]
  55. Yu, S.; Chen, Z.; Yu, B.; Wang, L.; Wu, B.; Wu, J.; Zhao, F. Exploring the relationship between 2D/3D landscape pattern and land surface temperature based on explainable eXtreme Gradient Boosting tree: A case study of Shanghai, China. Sci. Total Environ. 2020, 725, 138229. [Google Scholar] [CrossRef]
  56. Fung, T.; Siu, W.-L. A Study of Green Space and its Changes in Hong Kong Using NDVI. Geogr. Environ. Model. 2001, 5, 111–122. [Google Scholar] [CrossRef]
  57. Chen, J.; Zhou, C.; Li, F. Quantifying the green view indicator for assessing urban greening quality: An analysis based on Internet-crawling street view data. Ecol. Indic. 2020, 113, 106192. [Google Scholar] [CrossRef]
  58. Susaki, J.; Kubota, S. Automatic Assessment of Green Space Ratio in Urban Areas from Mobile Scanning Data. Remote Sens. 2017, 9, 215. [Google Scholar] [CrossRef] [Green Version]
  59. Elsadek, M.; Liu, B.; Xie, J. Window view and relaxation: Viewing green space from a high-rise estate improves urban dwellers’ wellbeing. Urban For. Urban Green. 2020, 55, 126846. [Google Scholar] [CrossRef]
  60. Sharam, L.A.; Mayer, K.M.; Baumann, O. Design by nature: The influence of windows on cognitive performance and affect. J. Environ. Psychol. 2023, 85, 101923. [Google Scholar] [CrossRef]
  61. Shin, S.; Browning, M.H.; Dzhambov, A.M. Window Access to Nature Restores: A Virtual Reality Experiment with Greenspace Views, Sounds, and Smells. Ecopsychology 2022, 14, 253–265. [Google Scholar] [CrossRef]
  62. Yu, S.; Yu, B.; Song, W.; Wu, B.; Zhou, J.; Huang, Y.; Wu, J.; Zhao, F.; Mao, W. View-based greenery: A three-dimensional assessment of city buildings’ green visibility using Floor Green View Index. Landsc. Urban Plan 2016, 152, 13–26. [Google Scholar] [CrossRef]
  63. Casalegno, S.; Anderson, K.; Cox, D.T.; Hancock, S.; Gaston, K.J. Ecological connectivity in the three-dimensional urban green volume using waveform airborne lidar. Sci. Rep. 2017, 7, 45571. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  64. Austin, M.P. Spatial prediction of species distribution: An interface between ecological theory and statistical modelling. Ecol. Model. 2002, 157, 101–118. [Google Scholar] [CrossRef] [Green Version]
  65. Süzer, Ö.K.; Olguntürk, N. The aid of colour on visuospatial navigation of elderly people in a virtual polyclinic environment. Color Res. Appl. 2018, 43, 872–884. [Google Scholar] [CrossRef]
  66. Santiago, C.; Green, N.F.; Hamilton, N.; Endler, J.A.; Osorio, D.C.; Marshall, N.J.; Cheney, K.L. Does conspicuousness scale linearly with colour distance? A test using reef fish. Proc. R. Soc. B 2020, 287, 20201456. [Google Scholar] [CrossRef]
  67. Hoyle, H.; Norton, B.; Dunnett, N.; Richards, J.P.; Russell, J.M.; Warren, P. Plant species or flower colour diversity? Identifying the drivers of public and invertebrate response to designed annual meadows. Landsc. Urban Plan 2018, 180, 103–113. [Google Scholar] [CrossRef]
  68. Gidlow, C.J.; Ellis, N.J.; Bostock, S. Development of the Neighbourhood Green Space Tool (NGST). Landsc. Urban Plan 2012, 106, 347–358. [Google Scholar] [CrossRef]
  69. Jennings, V.; Bamkole, O. The Relationship between Social Cohesion and Urban Green Space: An Avenue for Health Promotion. Int. J. Environ. Res. Public Health 2019, 16, 452. [Google Scholar] [CrossRef] [Green Version]
  70. Jenkins, G.R.; Yuen, H.K.; Rose, E.J.; Maher, A.I.; Gregory, K.C.; Cotton, M.E. Disparities in quality of park play spaces between two cities with diverse income and race/ethnicity composition: A pilot study. Int. J. Environ. Res. Public Health 2015, 12, 8009–8022. [Google Scholar] [CrossRef] [Green Version]
  71. Kaczynski, A.T.; Wilhelm Stanis, S.A.; Besenyi, G.M. Development and Testing of a Community Stakeholder Park Audit Tool. Am. J. Prev. Med. 2012, 42, 242–249. [Google Scholar] [CrossRef]
  72. Bunting, O.; Chesmore, D. Time frequency source separation and direction of arrival estimation in a 3D soundscape environment. Appl. Acoust. 2013, 74, 264–268. [Google Scholar] [CrossRef]
  73. Nieuwenhuijsen, M.J.; Dadvand, P.; Márquez, S.; Bartoll, X.; Barboza, E.P.; Cirach, M.; Borrell, C.; Zijlema, W.L. The evaluation of the 3-30-300 green space rule and mental health. Environ. Res. 2022, 215, 114387. [Google Scholar] [CrossRef] [PubMed]
Remotesensing 15 01543 g001 550
Figure 1. A conceptual framework for dimension generation.
Figure 1. A conceptual framework for dimension generation.
Remotesensing 15 01543 g001
Remotesensing 15 01543 g002 550
Figure 2. Comparison of the 2-D, 2.5-D, and 3-D observation areas.
Figure 2. Comparison of the 2-D, 2.5-D, and 3-D observation areas.
Remotesensing 15 01543 g002
Remotesensing 15 01543 g003 550
Figure 3. 3-D spatial metrics representing the greenspace exposure quality.
Figure 3. 3-D spatial metrics representing the greenspace exposure quality.
Remotesensing 15 01543 g003
Remotesensing 15 01543 g004 550
Figure 4. Workflow of the proposed method.
Figure 4. Workflow of the proposed method.
Remotesensing 15 01543 g004
Remotesensing 15 01543 g005 550
Figure 5. Schematic overview of the site (A) and plot location (B), as well as the experimental set-up (C). Residential communities included four area types. According to the satellite map, there are four communities (a–d) with green levels ranging from low to high. The satellite map of Nanjing city was downloaded from Bigmap GIS Office (industry version).
Figure 5. Schematic overview of the site (A) and plot location (B), as well as the experimental set-up (C). Residential communities included four area types. According to the satellite map, there are four communities (a–d) with green levels ranging from low to high. The satellite map of Nanjing city was downloaded from Bigmap GIS Office (industry version).
Remotesensing 15 01543 g005
Remotesensing 15 01543 g006 550
Figure 6. Validation results for the 3-D metrics in the four study areas. The highlighted green (red) areas in the picture are the vegetation (ground) areas for the indicator analysis. Metrics values are shown below the images based on the calculations (Table 3). A lateral view of the FGVR result has been added in the Appendix A for improved clarity.
Figure 6. Validation results for the 3-D metrics in the four study areas. The highlighted green (red) areas in the picture are the vegetation (ground) areas for the indicator analysis. Metrics values are shown below the images based on the calculations (Table 3). A lateral view of the FGVR result has been added in the Appendix A for improved clarity.
Remotesensing 15 01543 g006
Remotesensing 15 01543 g007 550
Figure 7. The visual scale between the two observation models.
Figure 7. The visual scale between the two observation models.
Remotesensing 15 01543 g007
Table
Table 1. Conceptual classification of four typical communities in China based on spatial features.
Table 1. Conceptual classification of four typical communities in China based on spatial features.
Area LevelAreaClassification Criteria
IRemotesensing 15 01543 i001Open communities in urban areas are mainly located in some “old city” areas, with high building density, low green coverage, low landscape diversity, and no separate residential activity space.
IIRemotesensing 15 01543 i002Open communities in urban areas are mainly located in the “old city” area. However, as surrounding public facilities and urban renewal projects are prioritized, the building density is high, green coverage is low, and there is an independent community activity space.
IIIRemotesensing 15 01543 i003Semi-open community in an urban area has complete supporting public facilities, low building density, high green coverage, high landscape diversity, and an independent community garden.
IVRemotesensing 15 01543 i004Gated communities in the city’s central area are equipped with independent gardens and multiple landscaping spaces to meet residents’ daily activities and entertainment needs, with a high green coverage and rich vegetation diversity.
Table
Table 2. Point to mesh registration results in the study area (cm).
Table 2. Point to mesh registration results in the study area (cm).
3-D-MetricsStudy Area IStudy Area IIStudy Area IIIStudy Area IV
RMS2.391.592.542.61
Mean-distance0.4690.4500.5970.892
Std deviation0.3430.3480.5760.785
Table
Table 3. Performance of the 3-D assessment metrics in the study area.
Table 3. Performance of the 3-D assessment metrics in the study area.
Dimensional3-D MetricsStudy Area IStudy Area IIStudy Area IIIStudy Area IV
Spatiality G V R 0.1980.2790.6630.679
σ 2 0.7910.6930.460.64
F G V R 0.1690.2290.5990.715
Structure V D I 13.7214.7619.7126.74
C i 0.0060.0040.0020.003
Colour C D i 132.66101.22138.13123.4833
G C D I 1190.061658.901756.222464.62
Function S R a 00.040.090.05
G C I 00.4940.5030.502
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Xia, T.; Zhao, B.; Xian, Z.; Zhang, J. How to Systematically Evaluate the Greenspace Exposure of Residential Communities? A 3-D Novel Perspective Using UAV Photogrammetry. Remote Sens. 2023, 15, 1543. https://doi.org/10.3390/rs15061543

AMA Style

Xia T, Zhao B, Xian Z, Zhang J. How to Systematically Evaluate the Greenspace Exposure of Residential Communities? A 3-D Novel Perspective Using UAV Photogrammetry. Remote Sensing. 2023; 15(6):1543. https://doi.org/10.3390/rs15061543

Chicago/Turabian Style

Xia, Tianyu, Bing Zhao, Zheng Xian, and Jinguang Zhang. 2023. "How to Systematically Evaluate the Greenspace Exposure of Residential Communities? A 3-D Novel Perspective Using UAV Photogrammetry" Remote Sensing 15, no. 6: 1543. https://doi.org/10.3390/rs15061543

APA Style

Xia, T., Zhao, B., Xian, Z., & Zhang, J. (2023). How to Systematically Evaluate the Greenspace Exposure of Residential Communities? A 3-D Novel Perspective Using UAV Photogrammetry. Remote Sensing, 15(6), 1543. https://doi.org/10.3390/rs15061543

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop