A pedestrian detection system using applied log-Gabor filters

Pedestrian detection is one of the most important research contents of road safety. The crucial idea behind such pedestrian safety systems is to protect the driver and pedestrian from any accident. In this paper, a pedestrian feature extraction based on applied log-Gabor filters is presented. The resulting filtered images show desirable segmentation performance which allows support vector machines to be able to efficiently classify and recognize the pedestrian. The proposed system is capable of detecting multiple pedestrians from complex background and providing size and position information of pedestrians within the image. Pedestrians with different sizes, shapes, postures, and clothes can be detected effectively. Results and discussions are presented.