Compact Thermographic Device with Built-in Active Reference Element for Increased Measurement Accuracy ^†

Abstract

Many common thermographic cameras have quite good sensitivity (0.05 °C or better) but limited accuracy, often about ±2 °C. This is not sufficient for quantitative measurements such as human body temperature diagnostics, where an accuracy of 0.3 °C is assumed. A thermographic device with a built-in active reference element for enhanced precision measurement was developed for these purposes. It was adapted for human body temperature diagnostics using face temperature measurements. It is based on a micro-bolometers detector and the reference element is heated to a temperature of 37 °C, which is mostly assumed as decisive for an increased temperature indication. As the reference element is embedded into the device housing, the use of the device can be very flexible. The performed human temperature measurement experiments showed that the accuracy of the introduced device is comparable with the thermographic measurements obtained using an external black body, which is often used for these applications.

1. Introduction

Increased temperature or fever in humans can indicate many infectious diseases and its measurement is thus a widely used diagnostics tool [1]. The core body temperature can vary in relation to many factors [2]; however, it is mostly assumed to be 36–36.9/37.5 °C. An increased temperature is in the range 37–38 °C and a fever is more than 38 °C. A peripheral body temperature can be influenced by location, health or ambient conditions. Contact-based methods [3,4] are the reference methods for human body temperature measurements. Invasive methods are more accurate and difficult to use, while non-invasive methods are less accurate but easier to use. In both cases, these methods are mostly not suitable for screening or for application in public spaces due to their contact nature and long application/measurement time.

Infrared (IR) methods are non-contact, fast, and can be easily used at a safe distance and in public spaces; they have thus been developed and used in medicine [1,5]. These methods are very effective for qualitative measurement (e.g., local inflammations). However, common micro-bolometer-based IR cameras mostly do not have the required measurement accuracy, which is usually assumed to be in the range of 0.1–0.3 °C [3,6] for the core body temperature (it is usually about ±2 °C). Even if a black body reference is used to improve the measurement accuracy, another drawback is that only the skin temperature is measured [7], which can differ from the core body temperature [8,9]. Despite these drawbacks, the applicability of infrared thermography (IRT) in human temperature diagnostics has been widely studied, also in connection with the recent COVID-19 pandemic that started in 2020 [10,11].

This study builds on the previously published research [8] dealing with the statistical evaluation of IRT human temperature measurement. The results presented show the crucial role of a reference black body correction in measurement reliability. The black body improved the results significantly and the IRT-measured temperatures (eyes maximum) were 35.7–36.2 °C, with a standard deviation of 0.4–0.9 °C (based on the IR camera used).

However, the practical implementation of a black body has some disadvantages. It has to be positioned in the field of view of the IR camera near the person being measured. It is not very flexible and it can sometimes be complicated to fix the black body in a proper position. The device IR Face Scanner (IRFS), introduced in this study, overcomes these drawbacks. The device is compact, but it contains an internal reference element, which increases its measurement accuracy. The device is described and the results of its verification, including both laboratory black body measurement and human face comparative measurements, are presented in this contribution.

2. IRFS Thermographic Device

The IRFS device is a device for human face temperature self-measurement. It consists of a Raspberry Pi 4B computer with a 7″ display and a thermal FLIR Lepton 3.5 camera with a resolution of 160 × 120 pixels, a framerate frequency of 8.7 Hz and optics providing a 57° horizontal field of view. The components are assembled in a plastic housing, together with a built-in heated reference element with a thermal sensor and PID controller, which controls its temperature to 37 °C (±0.5 °C). The reference element is positioned 6.5 cm in front of the IR camera in such a way that it is permanently in the field of view of the camera (in the right upper corner of its view). The depth-of-field Lepton detector allows both the measured object and the reference element to be in focus. The measured thermographic data thus can be continuously corrected based on the reference element temperature. The correction function was calculated via a laboratory calibration using blackbodies KLEIBER KBB35/40 at temperatures of 35 °C and 40 °C, respectively, placed at a distance of 36.5 cm in front of the camera. Both the location of the calibration body and its temperature correspond to the object to be measured, i.e., the human face; this, together with the narrow calibration range, increases the accuracy of the measurement for this application. The device is also equipped with software for AI-based automated face and eye-region detection.

3. Experiments and Results

First, a verification experiment with a thermal chamber and an external black body reference was performed. The IRFS device was placed into the thermal chamber with a hole for viewing the black body (Omega BB701, emissivity of 0.95, accuracy of ±0.8 °C). The temperature inside the chamber (ambient) was stepped from 10 °C to 40 °C in 10 °C steps, with a delay of 3 h at each temperature step. The procedure was repeated for black body reference temperatures of 35 °C, 37 °C and 39 °C. The measured temperature, for example, for the ambient temperature of 20 °C and the reference temperature of 37 °C was 36.5 ± 0.2 °C. The mean measured values for the individual reference temperatures (over all ambient temperatures) were 35.0 ± 0.4, 36.8 ± 0.3 and 38.7 ± 0.3 °C, respectively. If the results, on the other hand, are expressed for individual ambient temperatures of 10–40 °C, then the mean differences between the measured values and the reference temperatures were −0.4 ± 0.2, −0.4 ± 0.2, 0.0 ± 0.2 and 0.2 ± 0.2 °C, respectively. Thus, it can be summarized that the average (mean) difference between the measured temperature and the reference temperature, with regard to all reference and ambient temperatures, was −0.2 ± 0.3 °C.

A comparison of the statistical evaluation of human temperature measurement using the IRFS device (263 testers) and a FLIR A315 IR camera (FLIR Systems AB, Täby, Sweden) (105 testers, external black body reference used) is shown in Figure 1. The study was performed in near-laboratory conditions in a similar way to that described in detail in [8]. A group of testers (healthy volunteers), which were thermally stabilized inside the building, repeatedly performed a self-measurement over several months. The results were evaluated via statistical means instead of solving individual testers. The mean values of the maximum eye temperatures were 36.0 ± 0.4 and 36.8 ± 0.4 for the A315 and IRFS devices, respectively. The mean forehead temperatures were 34.5 ± 0.9 and 34.3 ± 0.9, respectively. The comparative measurement thus showed a very good consistency between both devices in terms of mean values and deviations, and even outliers; these are, in Figure 1, visible when performing measurements using both IR devices.

4. Conclusions

The proposed IRFS device is very compact and flexible in use as it does not require an external reference black body, which is mostly necessary for obtaining accurate measurements using standard microbolometer-based IR cameras. The device is also quite cheap in comparison to many commercial high-accuracy IR cameras. However, the thermal chamber verification experiment showed that the device can measure accurately enough in the required range, even when ambient conditions change. The results of the comparative human face temperature measurement, which were also consistent with those previously published in [8], then confirmed that the practical accuracy of the IRFS device was fully comparable with the commercial solution based on the FLIR A315 IR camera. Even if this study does not address other general aspects of the issue of human temperature measurement or screening using thermographic methods, the results achieved and presented by the introduced IRFS device are promising for further development in this area.

5. Patents

M. Honner, V. Lang, M. Švantner, J. Šroub, T. Kohlschütter, and J. Klepáček, “Device for thermographic temperature measurement,” Application No. PCT/CZ2022/050111, WIPO No. WO2023072325A1, 2023.