Complex urban background anomaly target self-learning detection method and device

Abstract

The application provides a complex urban background anomaly target self-learning detection method, and relates to the field of information technology, wherein the method comprises the following steps: collecting urban scene data through a multi-modal visual sensor, and performing preliminary anomaly detection on the collected multi-modal data to obtain a preliminary anomaly detection result; adding uncertainty knowledge to the preliminary anomaly detection result by using prior knowledge and data tuning to construct an uncertainty model; calculating and comparing the relative entropy between different modal data based on the uncertainty model, and determining the resampling magnitude according to the relative entropy difference to adjust the weight of different modal data; and performing multi-factor coupling reasoning on the collected multi-modal data based on the adjusted weight to obtain an anomaly detection result. The application adopting the above scheme can reliably monitor the anomaly target in a complex urban background environment in an unsupervised environment.

Images