Research On Fast Pedestrian Detection Algorithm Based On Deep Learning

Pedestrian detection is one of the most important branches in the field of computer vision.It has great development potential and application prospects.In recent years,the application of intelligent monitoring,unmanned driving and intelligent robots has rapidly promoted the research and development of pedestrian detection.In the dense crowd,the body of a small number of pedestrians cannot be fully displayed,which makes the detection of pedestrians missing.In the re-identification research of specific pedestrian,it is difficult to extract effective pedestrian features due to occlusion,posture and shooting angle.Therefore,in view of the above problems,this paper studies pedestrian detection algorithm to achieve fast and accurate pedestrian positioning,the main research is as follows:(1)In the crowd with high density,the head is the most prominent part of the human body,and it is difficult to occlude and deform.Detecting the head to determine whether there are pedestrians can effectively reduce the rate of pedestrian detection missed.This paper takes human head as the research object,adds batch normalization on the basis of MTCNN cascade network according to the difference of human head,and establishes BN-MTCNN human head detection network.The training and testing are carried out in the self-built human head data set.The test set is divided into Easy and Hard subsets according to the degree of intensity,illumination change and occlusion.For video sequences,the tracking algorithm KCF is used to track the human head steadily in the detection stage,and a detection-tracking-detection(DTD)cyclic detection model is constructed,which can detect the human head quickly and steadily.The experimental results show that the detection accuracy of Easy subset is 95.01% and that of Hard subset is 86.35%.For indoor corridor video sequences,the detection accuracy of GTX 1080 hardware platform with GPU is94.85%,and the detection speed is 38 frames per second in 1920 *1080 video sequences.(2)In order to further improve the speed and stability of non-significant human head detection in the actual environment,combined with the YOLOv3 network and KCF algorithm with higher accuracy for non-significant object detection,the DTD cycle detection model is adopted to achieve rapid stabilization of non-significant head targets.The experimental results show that the detection accuracy of the Easy subset is 97.62%,the detection accuracy of the Hard subset is 89.53%,the detection accuracy in the video sequence is 97.43%,and the speed is 62 frames/second.(3)In order to improve the accuracy of target recognition of specific pedestrians,based on pedestrian detection,a pedestrian re-identification method based on improved AlignedReID network is proposed.The method combines the DenseNet121 network with less parameters and good generalization performance,and integrates the global and local features of pedestrians to achieve accurate identification of specific pedestrian targets.The experimental results show that the recognition rates of Rank-1 and Rank-5 are 68.2% and 80.2%,respectively,and the mean Average Precision(mAP)is 71.6% on the CUHK03(detected)data set with occlusion between pedestrians,which are 0.6%,0.7% and 0.9% higher than that of the original AlignedReID network;on the pedestrian unobstructed Market1501 data set,the Rank-1 and Rank-5 recognition rates were 93.8% and 97.3%,respectively,and the mAP was 90.5%,which are 1.8%,0.9% and 2.0% higher than that of the original AlignedReID network.