Active learning anomaly detection method and system based on graph propagation

The invention discloses an active learning anomaly detection method and system based on graph propagation, and relates to the technical field of image anomaly detection.A product image is collected and then preprocessed, a pre-training automatic encoder is constructed and trained, an image code without low-level features is obtained through the trained automatic encoder, and further reconstruction is performed, so that the detection accuracy is improved; more continuous and smooth image low-dimensional representation can be obtained, and better anomaly detection performance can also be obtained. The image annotation information is updated based on the graph propagation of the k-affinity propagation matrix, so that the time consumption of updating the image annotation information is lower, the uncertainty and representativeness of the image sample are considered at the same time in the active learning stage, the most representative image sample is selected, the image sample is fully explored in space division,