Anomaly detection method, model training method, device, equipment and storage medium

Abstract

The application discloses an anomaly detection method, a model training method, a device, a device, and a storage medium, and relates to the field of computer technology, in particular to the fields of computer vision and image processing. The anomaly detection method disclosed in the present application includes: by preprocessing the acquired video to be detected, the first detection instances of multiple pipeline categories and the second detection instances of multiple video levels are obtained, and the detection instances of these two categories are input To the anomaly detection model, obtain the detection instance where the target object in the video to be detected is abnormal. Wherein, both the first detection instance and the second detection instance include a plurality of continuous image frames, and the image frames in the two types of detection instances correspond one-to-one, and the size of the image frame in the first detection instance is smaller than that in the second detection instance image frame size. The above anomaly detection method involves anomaly detection on two types of detection instances, and the obtained video anomaly detection results are more accurate.

Description

technical field [0001] The embodiments of the present application relate to the field of computer vision in computer technology, and in particular to an anomaly detection method, a model training method, a device, a device, and a storage medium. Background technique [0002] Anomaly detection of surveillance video is an important research field in computer vision. It is used to intelligently analyze surveillance video and automatically identify abnormal events in the video or abnormal behavior of monitored objects, such as fire accidents, people gathering, and vehicles traveling in reverse. At present, this technology is widely used in intelligent transportation, security and other fields. [0003] In the prior art, surveillance video anomaly detection is mostly based on supervised learning. Supervised learning mainly builds a predictive model by learning a large number of training samples, where each training sample has a label indicating its true value output. Anomaly de...

AI Technical Summary

Problems solved by technology

Anomaly detection methods based on supervised learning can only detect pre-trained abnormal events or behaviors in surveillance videos, and have poor scene adaptability and great limitations

Images