Detecção de pessoas com febre por termografia infravermelha

Abstract

The identification of patterns and faces in images are some branches of Computer Vision and can serve as a tool to automate a process that demands the manual work of an observer. Among the types of existing images, the infrared image or thermographic image is used in areas of medicine and engineering. In medicine, infrared images can be used to detect areas of the human anatomy with irregular situation. With this work, it is shown a tool that will locate and classify the maximum temperature value of the face in an infrared image. In order to do this, a machine learning training was performed using the AdaBoost algorithm, with face samples in infrared images, and in the face detection the Viola and Jones algorithm was used. The operation of the tool was by the detection of the face of the individual followed by the location and classification of the value of the higher temperature of the face. Comparisons were made with the results of detections obtained with different training sessions. With this data, it was noticed that more rigid classifiers discard more areas of the image, while less rigorous classifiers result in a greater amount of false positives. The final results of the development brought a methodology of detection of fever in individuals without need of direct contact, showing little difference between the axial temperature and the temperature in the infrared image. It was also possible to create a face detection file in infrared images, which correctly detected 79.51% of the faces in the images provided.