Research On Pedestrian Detection Algorithm Based On Deep Learning

Today,pedestrian detection based on deep learning technology is one of the hot topics in computer vision.It is widely used in artificial intelligence such as intelligent monitoring,unmanned driving,intelligent robots and so on.In this paper,the pedestrian detection algorithm based on deep learning technology is studied,and the following results are obtained:(1)Aiming at the problem of high complexity and weak robustness of traditional feature extraction,the general object detection algorithm Faster R-CNN in deep learning is applied to pedestrian detection,and getting satisfactory results.The pedestrian detection accuracy rate of this algorithm on Caltech dataset,KITTI dataset and ETC dataset reached 68.57%,67.27% and 81.30%,respectively.(2)Aiming at the problem of multi-scale pedestrian detection,a pedestrian detection algorithm based on fusion Feature Pyramid Networks(FPN)and Faster R-CNN is proposed,which has achieved satisfactory results on multi-scale pedestrian detection.The pedestrian detection accuracy rate of this algorithm on Caltech dataset,KITTI dataset and ETC dataset reached 69.72%,69.76% and 89.74%,respectively.(3)In order to reduce the average error rate of the disparity map obtained by non-local cost aggregation stereo matching algorithm(NL algorithm),the multi-scale is introduced into non-local cost aggregation,and a stereo matching algorithm based on multi-scale non-local cost aggregation is proposed.Compared with the NL algorithm,the average error rate of this algorithm is reduced by 0.51% on the Middlebury dataset.In the follow-up study,this paper will plan to fuse the binocular disparity map with RGB images to realize 3D pedestrian detection based on deep learning,so as to solve the occlusion problem of pedestrian detection.