The project is a driver monitor system, and it uses the thermal camera to collect the driver images. The thermal camera can overcome the influence of light and background for face detection, and provide the temperature information. During the driving process, the system can detect the human pose, and by the thermal images of face, it can obtain the physiological parameters of the driver, like temperature, heart rate and respiratory rate. The system can use those information to detect the driver’s fatigue and dangerous behavior, and warn the driver to increase your attention by speech.
In university, I am working on the hot spot effect of photovoltaic power system. So, I have the opportunity to use the thermal camera. The thermal camera is amazing and interesting, and allows you to see the world in a whole new way. Therefore, I decide to DIY something use the thermal camera. After watching a news about car accident caused by driving fatigue, I think it is necessary to try to do something for the drivers. Compared to the RGB visible light image, the thermal image can be not influenced by light, such as through a tunnel, in night and so on. Simultaneously, it is easy to split out the body from the thermal image.
However, the price of thermal camera usually is high. Fortunately, the price of thermal camera FLIR ONE for phone is acceptable. Because of the FLIR ONE for android, it is appropriate to select the Android things OS and Raspberry Pi 3. At the software aspects, I choose the Android Studio to build the project and use the Opencv Library to process the image. Due to short time, I only carry out the face detection by face training used the Terravic Facial IR Database (http://vcipl-okstate.org/pbvs/bench/). The system is shown in the figure.
Next, there is still a lot of work to do. For example, by the thermal images of face, it can monitor the physiological parameters of the driver, like temperature, heart rate and respiratory rate, and those are helpful for analyzing the state of the driver. Others, the system need to detect the driver’s pose using the TensorFlow, it can warm the dangerous action during driving.
Comments