Insulator unsupervised anomaly detection and positioning method and system

The invention belongs to the technical field of insulator state monitoring, and particularly relates to an insulator unsupervised anomaly detection and positioning method and system. The method comprises the following steps: extracting an intermediate feature of an insulator image for reconstruction by using a pre-trained convolutional neural network; constructing a transform network according to the intermediate features, and determining the inter-image block correlation and the inter-image block correlation of the insulator image; learning intra-image block correlation and inter-image block correlation of the insulator image, and determining an abnormal mode of the insulator to obtain reconstructed intermediate features; and determining the abnormal state of the insulator according to the error of the reconstructed intermediate feature. According to the method, the inconsistency between the insulator image blocks and the blocks in the image and between the images is used as a complementary factor, and the d