Target Recognition And Analysis Based On Video Structural Model

With the rapid construction of smart city and safe city system,it is difficult to extract the required information from the massive surveillance video.Therefore,the demand for video analysis system is urgent.Video structured analysis system is a basic platform for the construction of smart city,safe city and other projects.Video structural analysis system usually includes image processing,feature extraction,object detection,object recognition,video standardized description and other steps.After the processing methods,the video information can be transformed into text information which accurately describes the content of video.That is actually available in public security system.Currently,with the development of deep learning technology,deep learning technology plays an increasingly important role in the realization of road monitoring video information.In this paper,using the object detection technology in deep learning,such as Yolo,SSD,the car face recognition technology,such as the metric learning technology based on facenet,the original video can be intelligently analyzed to extract the key information.For a road monitoring video,there are mainly three tasks:the first task is the detection of the target,the detection of moving or stationary vehicles and pedestrians in the picture;the second task is the recognition of the target features,that means what features the target in the picture has,for example,the description information of the vehicle includes:license plate,vehicle color,vehicle type,brand,sub brand,car sticker,car accessories information and other vehicle description information,The description information of pedestrians includes:wearing,hat wearing and other description information;the third category is the trajectory analysis of the target,that is,the description of the behavior of the target in the screen,including:area,wandering,gathering and other description information.Finally,it provides semantic retrieval,image search and other retrieval methods,which can quickly and accurately find the target from a large number of surveillance video.Through the above analysis:1:Improve search and troubleshooting efficiency,2:Can reduce storage share,3:The processed video data can be used for deep data mining and other applications.Compared with traditional algorisms,structural description model based on deep learning has obvious advantages in both accuracy and anti-interference ability,which can greatly improve the utilization efficiency of road traffic videos,reduce unnecessary repetitive labor of public security system.Therefore,it has a broad application prospect.