Video Anomaly Detection Based on Spatial Awareness and Attention Fusion Method
Mengyao Liu, Zhigang Xu · 2023
Anomaly detection in video is one of the challenging problems in the field of computer vision. To address the problem of insufficient prediction capability due to inadequate extraction of spatial features of object appearance in video anomaly detection methods based on predicting future frames, this paper proposes a video anomaly detection method based on spatial perception and attention fusion. The method uses spatially-aware coding network to extract the appearance spatial feature information of objects in video frames, while modeling the global dependencies of deep spatial coding by a non-local module. Then the attention fusion module is used to establish the interaction between the spatial coding and the temporal coding extracted by the traditional coding network to enhance the discriminative representation of the network for video frames. Experimental analysis shows that the method in this paper is able to achieve 84.2% and 95.7% of frame-level AUC in the public benchmark datasets UCSD Ped1 and Ped2, respectively, with good detection performance for video anomalous behavior.