Research And Development Of Gait Feature Extraction System In Video Surveillance

Today,the video surveillance industry is growing rapidly.A large number of surveillance cameras have brought about an increase in the amount of data,resulting in problems such as difficulty in retrieval and poor real-time response.It is more and more important to carry out intelligent analysis of video surveillance content.Among them,pedestrian target identification in video plays a very important role in criminal investigation and self-service.Gait recognition realizes identity recognition through walking posture,without human coordination and adapting to a long distance.To apply gait recognition technology to video surveillance,the primary task is to obtain pedestrian gait feature expression,and it is necessary to solve the problem of obtaining pedestrian effective gait characteristics in complex surveillance scenarios.Therefore,for the gait feature extraction under video surveillance,two problems need to be solved: feature expression selection and acquisition and gait feature extraction in complex background environment.Firstly,in order to avoid the influence of environmental color,illumination and shooting angle on the recognition,combined with the extensiveness of the research objects used in the field of gait recognition,the gait contour map or gait template is selected as the gait feature expression,and the feature extraction is transformed to the problem distinguish between pedestrian outlook and background.Then it is proposed to use traditional background modeling method for processing.According to its existing problems,the deep learning semantic segmentation combined with detection and humanoid snapshot matching is proposed to segment the humanoid image to extract gait features.Finally,the effectiveness of the extracted gait features is verified by integrating the gait recognition algorithm,and the effectiveness of the extraction method is verified by experiments.Experimental results show that the gait recognition algorithm can accurately identify the recurring targets by using the extracted gait features,and the system can effectively extract pedestrian gait features for surveillance video in different environments.As the system running speed,the module communication mode and batch processing adopted have a certain improvement on the running speed.