Automatic Mobile Warning System against People with Elevated Body Temperature

Abstract

The paper proposes a system that allows for the automatic detection of people with elevated body temperature and estimates distance from them using a smartphone-type device and a single mobile thermal camera. The algorithm automatically finds and selects humans with the highest temperature, and tracks changes in their position in an image sequence. On the basis of the change in the position of the human head in the image, in subsequent frames, the algorithm estimates the distance between camera and human. Owing to the use of fast machine-learning methods, the proposed system can immediately alert the user about the presence of a people with an elevated temperature at a distance of 1–3 m as soon as it appears in the field of view of the camera. The effectiveness of the algorithm was assessed as the ratio of correct distance classifications in the test image set to the total number of test images. Values ranging from 73% to 100% were obtained for over 4000 images of humans at different distances. The proposed method allows for the quick and completely automatic warning aboutt people with elevated temperature, and can be used in popular Android mobile devices.

Keywords

• thermovision image segmentation
• object detection and localization
• machine learning
• computer vision