Video anomaly detection method and device, computer equipment and storage medium

The invention relates to the field of video anomaly detection, in particular to a video anomaly detection method, which utilizes a video image prediction model, constructs a sample optical flow reconstruction image through an image reconstruction method to reflect a motion trend in the image, and performs image prediction by combining a sample image and the sample optical flow reconstruction image. A prediction image is constructed, interference of background factors is avoided, a discriminator module is utilized, an adversarial learning method is adopted, and the video image prediction model is trained, so that the training accuracy and efficiency of the video image prediction model are improved, the video image prediction model can output accurate optical flow reconstruction images and prediction images, and the prediction efficiency is improved. The method is used for performing video anomaly detection on an original image, and the accuracy of video anomaly detection is improved. 本发明涉及视频异常检测领域，特别涉及一种视频异常检测