Research On Pedestrian Recognition Algorithm Based On Video Frame Image

In recent years,with the continuous development of intelligent video surveillance systems,we are increasingly using high-definition cameras in surveillance.At the same time,video analysis technology is constantly improving,and the application scenarios of intelligent video surveillance systems are more popular than in the past.And refinement,all walks of life put forward higher requirements for intelligent video surveillance systems,allowing users to monitor in real time according to their needs,can identify different objects in the video screen,discover abnormal conditions in video surveillance,and transmit them in real time.Outgoing these dangerous information to the relevant security staff,informing them to deal with these emergencies,thus improving the intelligent video surveillance system,making it more efficient,smarter and better to be effective in responding to attacks and dealing with emergencies.Auxiliary tools.This paper focuses on the video image sequence frame image obtained by video surveillance.Under the scene of sudden change of illumination and heavy pedestrian shadow problem,the task of extracting the foreground of pedestrian motion and then performing fast pedestrian recognition based on the foreground image is completed.The main research object of this paper is the video frame single-frame image containing pedestrian motion image obtained from the intelligent video surveillance system.Firstly,the current research background,research status and common research methods in the field of pedestrian recognition are introduced.Next,this paper The algorithm of the two stages of pedestrian foreground extraction and pedestrian recognition is taken as the research focus,mainly in two aspects:In this paper,based on the research and analysis of the traditional pedestrian motion foreground extraction algorithm,a certain degree of improvement and optimization is carried out.The lab color difference formula is used to calculate the Euclidean distance between two pixels,and the chromaticity and brightness are segmented to better avoid the influence of illumination on the foreground of pedestrians.The spatial consistency of the image is used to remove the holes on the foreground image.The update of the background model improves theinfluence of shadow on the foreground extraction of pedestrians;extracting key frames according to the pedestrian motion cycle greatly reduces the amount of redundancy calculation,which lays a foundation for the feature extraction and matching of the target pedestrians.Based on the theoretical research of traditional pedestrian recognition algorithm,this paper proposes a pedestrian recognition algorithm based on gait feature and color feature fusion.The algorithm uses the statistical method to express the pedestrian's color features and gait features separately,adds the spatial information of the color to the statistical features of the color,and extracts the centroid contour moment as the gait feature from the acquired pedestrian motion foreground image.The two features are combined to determine the final decision result.Compared with the learning-based pedestrian recognition algorithm,it has significant advantages in time.Compared with the single-feature pedestrian recognition algorithm,the recognition algorithm proposed in this paper has a higher recognition rate.