Abnormal Event Detection And Recognition In Tourism Video

With the booming development of tourism,travel security problems are becoming more and more prominent.Congestion,stampedes,fights and other tourism emergency events occurred frequently,which should be a wake-up call for tourism security.Therefore,it is of great research value and application prospect to real-time monitor tourists and detect and recognize abnormal events in tourism surveillance video by using computer vision and video intelligent processing technology,which can realize the timely forecast and early warning of tourism emergencies.The main work done in this thesis includes the following four aspects:(1)Aiming at the problem that the existing event representation methods do not fully consider the inter-frame temporal and spatial correlation of video,which is difficult to apply to complex motion scenarios such as background confounding and the mutual occlusion between targets,a salient spatio-temporal feature extraction method of tourism video is proposed.The background modeling method based on the Gaussian mixture model is used to realize the background modeling and foreground detection of tourism video,and the foreground target in tourism video is extracted,which solves the problem of background information interference in tourism video.On this basis,the feature extraction method based on spatio-temporal gradient model is used to extract the salient spatio-temporal features on the foreground target of the input video,which realizes the well representation of the local activities under the crowded scene while the spatio-temporal correlation between video frames is fully considered.(2)Aiming at the problem that the existing abnormal event detection methods have low robustness and timeliness in complex motion scenarios,and are unable to adapt to the real-time abnormal event detection in practical applications,an abnormal event detection method in tourism video based on sparse combination learning is proposed.Combined with the extracted salient spatio-temporal features of tourism video,an abnormal event detection model of tourism video based on sparse combination learning is constructed,which solves the problem that the existing abnormal event detection methods based on sparse representation take a long time in the detection phase.The experimental results show that the proposed method has good robustness and timeliness in complex motion scenarios.The recall rate on ScenicSpot dataset,Avenue standard dataset and UCSD Pedl dataset has been improved to 95.9%,95.1%and 92.4%,F1 value has been improved to 0.880,0.870 and 0.868,and the frames per second has been improved to 402.55,558.52 and 461.94,respectively.(3)Aiming at the problem that most of the existing abnormal event recognition methods only learn video features in the spatial domain and lose the time information of the input video,an abnormal event recognition method in tourism video based on spatio-temporal perception deep network is proposed.By combining the spatio-temporal deep convolutional neural network and the spatio-temporal pyramid pooling,the high-level semantic features of tourism video are learned in the time and space domain,and an abnormal event recognition model of tourism video based on spatio-temporal perception deep network is established,which solves the problem that the existing abnormal event recognition methods based on deep learning cannot model time signals well and the deep network limits the size and length of input video.The experimental results show that the proposed method significantly improves the accuracy of video abnormal event recognition,and the accuracy on the scenic spot dataset and Avenue dataset has increased to 97.7%and 93.3%,respectively.(4)Combined with the salient spatio-temporal features extraction of tourism video,abnormal event detection in tourism video based on sparse combination learning,and abnormal event recognition in tourism video based on spatio-temporal perception deep network,an abnormal event detection and recognition system of tourism video is designed and implemented.The three functional modules of the abnormal event detection and recognition system of tourism video are realized:video feature extraction module,abnormal event detection module and abnormal event recognition module.The video feature extraction module realizes the foreground detection and the salient spatio-temporal feature extraction of tourism video.The function of the abnormal event detection module for real-time detection of abnormal events in tourism video is realized.The function of the abnormal event recognition module to automatically recognize the type of abnormal events in tourism video is realized.This thesis realizes the salient spatio-temporal features extraction of tourism video,abnormal event detection in tourism video based on sparse combination learning,and abnormal event recognition in tourism video based on spatio-temporal perception deep network,and an abnormal event detection and recognition system of tourism video is designed and implemented.The experimental and test results show that the system can realize the automatic detection and recognition of tourists' abnormal behavior,and timely discover and monitor tourism emergencies.