Recognition Method of Abnormal Behavior in Electric Power Violation Monitoring Video Based on Computer Vision

In order to improve the accuracy of abnormal behavior recognition in electric power illegal behavior monitoring, a method of abnormal behavior recognition in electric power illegal video based on computer vision is proposed. The monitoring image of electric power violations is collected by sensors, and the monitoring video image is preprocessed based on mathematical morphology and neighborhood average filtering; The static target detection method and background difference method in computer vision technology are used to separate the background and moving foreground in the video frame sequence; Locate the staff in the video image and track their movement track; Fusing FAST corner and SIFT algorithm to extract corner features and texture features of staff action behavior in the monitoring image; The above features are input into the long and short memory recurrent neural network to realize the recognition of abnormal behavior in the electric power illegal monitoring video. The results show that the Kappa coefficient between the method and the measured results remains above 80%, which proves that the recognition method improves the accuracy of abnormal behavior recognition.