Vision-based methods, systems, and media for container anomaly detection

Abstract

This invention provides a vision-based method, system, and medium for container anomaly detection, applicable to the field of intelligent quality inspection technology. The invention acquires operational status change data, then obtains first multi-source image data of the area where the target container is located. Preprocessing the first multi-source image data yields second multi-source image data, effectively improving image clarity. Next, multi-object recognition is performed on the area of ​​the target container based on the second multi-source image data to obtain multi-object recognition results. Anomaly judgment is then performed based on the multi-object recognition results to obtain anomaly judgment results. Simultaneously, multi-modal fusion is performed based on the second multi-source image data to obtain multi-modal fusion results. Consistency verification is then performed on the data in the multi-modal fusion results to obtain anomaly verification results. Finally, anomaly linkage early warning data is generated based on the anomaly judgment results or anomaly verification results, thereby effectively improving the accuracy and timeliness of container anomaly detection.

Images