The Method Research On The Detection And Identification Of Specific Targets In Video Surveillance

With the improvement of living standards, people pay more attention to the security increasingly. Building safe city and intelligent surveillance set off boomly. The further development of the security industry is promoted with the rapid development of Internet. The target detection and identification in surveillance video has become an important research direction. In this paper, detection and identification of the specific objectives of surveillance video for the subject, focusing on surveillance video in the target pedestrian and person face detection, tracking and recognition algorithm, aimed at creating a video surveillance alarm system, with the contribution to the reducing workload of monitoring staff, improve work efficiency and building the safe city. In this paper, the core content that the detection and identification of target face in surveillance video is divided into the following:Firstly, we create the development environment based on Eclipse and OpenCV2.4.8, study the particular subject detection methods, complete the pedestrian detection based on HOG feature and SVM with training specific classifier to meet the application, achieved a multi-angle face detection in surveillance video combined CS-Adaboost algorithm.Secondly, we achieve object tracking based on perceptual hashing algorithm. First, it needs to save the tracked target, then calculate its hash value, then scan the entire image when each frame comes and calculate hash code for each scan window,at last we compare the hash code Hamming distance to the target, minimum Hamming distance is closest to the target window, then we can get location of the target. Then, combining the spatial-temporal context fast tracking to optimize, we complete target tracking real-time video.Thirdly, we implement the target recognition in surveillance video combined the advantages of machine learning and computer vision as well as in-depth study of PCA.Fourthly, We designed the target detection and identification alarm system with the combination of the above detection, tracking and recognition method. When monitoring staff find the suspicious target, then they can sign the suspicious targets and then get the target image with matching the image to the database to identify, at last We finish the task of specific target face detection and recognition in surveillance video. Further more, we can avoid the occurrence of risk by the alarming function.