Research On Techniques Of Pedestrian Detection Based On Vehicle-loaded Video

Traffic accidents occurs more frequently,with the increase of car ownership and the rapid development of urban transportation.As vulnerable groups in the traffic environment,the safety of pedestrians cannot be ignored.The study of effective pedestrian active protection technologies is of great significance to ensure pedestrian safety and prevent vehicle-pedestrian accident.Pedestrian detection technology is also the core technology of ADAS(Advanced Driver Assistance Systems)and unmanned driving system.Therefore,pedestrian detection is the premise of security measures.Techniques of pedestrian detection based on vehicle-loaded video is of great academic significance and application value.The algorithms of pedestrian search space、feature expression and classification based on vehicle-loaded video are deeply studied in this paper,from the perspective of computer vision.And we propose a method to locate moving pedestrian from running vehicle、a wavelet-like algorithm to extract the features of pedestrians based on deep neural networks and a brain-inspired emotional learning algorithm to recognize the pedestrians.The studies on techniques of pedestrian detection based on vehicle-loaded video are as follows:(1)Section II describes a method to locate moving pedestrian from running vehicle.The traditional pedestrian detection system uses the sliding window to search and recognize,causing the useless search problems.In addition,it is difficult to apply to the practical application of automobile environment,for example ADAS(Advanced Driver Assistance Systems)and unmanned system.In order to overcome the problems of useless search and improve the real-time in pedestrian detection,based on the difference of moving pedestrian and background in the optical flow field,this paper propose an optical flow clustering algorithm,which can fast estimate the pedestrian RIO(Regions of Interest)area in the video.First,this method calculates the optical flow field in the vehicle-loaded video.Next,the optical flow field is clustered and the background is eliminated according to the difference of moving pedestrian and background in the optical flow field.And then,based on the graph theory,the image segmentation algorithm is used to realize the segmentation of moving foreground target.Finally,according to the characteristics of the human body,we obtain the pedestrian ROI(Regions of Interest)area by distinguishing the area after division,reducing the search space of pedestrian detection.(2)Section III presents a wavelet-like algorithm to extract the features of pedestrians based on deep neural networks.The target in the ROI area needs to be recognized after the effective pedestrian area is obtained.The effectiveness of pedestrian characteristics affects the accuracy of pedestrian detection directly.In this paper,we analyze various algorithms of pedestrian feature extraction.Considering that the wavelet transform could extract the characteristics of images at different levels,a wavelet-like pedestrian feature generator is established based on deep learning to extract the features of pedestrians effectively.(3)Section IV proposes a brain-inspired emotional learning algorithm to recognize the pedestrians.Brain emotional learning(BEL)model is a novel bio-computation model inspired by physiology,and has strong learning ability and generalization ability.From the perspective of brain-inspired intelligence,the brain emotional learning(BEL)model is improved in this paper.According to the pedestrian features obtained in this paper,we propose a brain-inspired emotional learning algorithm to recognize the pedestrians.The improved model has a certain degree of robustness and improves the accuracy of pedestrian recognition to some extent.In this paper,we establish a pedestrian detection model based on vehicle-loaded video,from the perspective of computer vision.The proposed model can recognize pedestrians in the camera view effectively when the vehicle is moving,providing technical support for the driver assistant system.And it plays an important role in vehicle safety distance keeping and Vehicle-pedestrian Accident prevention.