Research on Pedestrian Detection Using CNN-Based Faster-RCNN Algorithm

Now the number of vehicles gradually is increasing in the community, so the possibility of a car accident is also increasing. But if the vehicle can detect pedestrians and issue warnings early, the number of accidents will be reduced greatly. With the development of deep learning, more and more people apply CNN (convolution neural network) to pedestrian detection. In this paper, Faster RCNN developed by RCNN (Region-based CNN) and Fast-RCNN algorithm use convolution neural network (CNN) to extract the feature map, and the anchor box is extracted by RPN (Region Proposal Network). In the RPN (Region proposal network) through the NMS (Non-Maximum Suppression) method to extract a number of the most suitable candidate box, and then the classification and recognition are carried out. The result is that the training speed and accuracy is very high.