Pedestrian detection based on an assembly of sub-image detectors

A method is put forward for pedestrian detection in single-frame which has a robust performance in the presence of clutter and occlusion. The search window is divided into 20 blocks in which areas are named sub-image. Each block is detected by sub-image detector which is learnt from the relevant block of training images using Support Vector Machine (SVM). We assemble all the sub-image detectors with SVM to approach the detection decision. The system has been tested on different databases and on real scenes. The experimental results show that the proposed algorithm has a strong ability to detect the pedestrians, even under a condition of occlusion.