Automatyczna lokalizacja twarzy oraz oczu w termogramach dla potrzeb pomiaru temperatury

W artykule omówiono opracowany w Instytucie Automatyki algorytm, który umożliwia automatyczny pomiar temperatury w kącikach ludzkich oczu. Pomiar termowizyjny temperatury wewnętrznej ludzkiego ciała stanowi alternatywę dla tradycyjnych-stykowych metod pomiaru temperatury na czole, czy też pod pachą. Zaprezentowany w pracy algorytm pozwala na dokładną lokalizację twarzy oraz oczu, a następnie wykorzystując metodę segmentacji i morfologii na określenie obszarów kącików oczu i pomiar temperatury.

The algorithm for detection of face and eyes in thermal images developed in the Automatic Control Institute is presented in this paper. There are also presented some major problems connected with face detection in thermal images. Face detection and recognition have been the active research area for more than thirty years. Many studies have proved that face recognition in the thermal spectrum offers a few distinct advantages over the visible spectrum. Thermal imagery is independent of ambient lightning, because IR sensors only measure the heat emitted by the object - human body. Thermovision measurement of the human body's internal temperature is an alternative to traditional methods of the temperature measurement on the forehead or underarm. The algorithm presented in the paper allows the exact localization of the face and eyes, and then using the segmentation method determination of the corners of the eye area and measurement of the temperature in their area. The face detection and recognition algorithms are based on a local feature representation. The proposed algorithm is based on the modified Randomized Hough Transform of the face and eyes verification procedure and, finally, on the seg-mentation growing with morphological operations for localization of the eyes corner area. Then in these two localized areas the temperature is measured. The paper also discusses experimental studies and the results which allowed evaluating the effectiveness of the developed algorithm.