Multiple Targets Tracking Using Attention GA

There have been many attempts to realize human-like visual function by image processing. Methods for recognition and tracking of targets are expected to be applied in security systems and in the field of ITS (intelligent transport systems). This study was performed to construct a detection system capable of recognizing multiple areas of the targets in real time. We employed a attention-GA (genetic algorithm) based on selective attention, which is the human visual function used to reduce processing load, to search for the position of a face in input images. The attention-GA consisted of random searching as a rough search and an improved GA that performed a more careful search. These methods were combined by grouping to allow simultaneous detection of multiple targets in input images in real time. We confirmed the effectiveness of our proposed detection system by experiments involving detection of multiple targets.