Power transmission line insulator defect detection method based on computer vision

The invention discloses a power transmission line insulator defect detection method based on computer vision, and the method comprises the steps: S1, constructing an image database of power transmission line insulator defects, and dividing the image database into a training set, a verification set and a test set; s2, constructing an MIDDNet insulator defect detection model, wherein the model comprises an RGNet trunk feature extraction network, an RSNet feature fusion network and a GSNet defect detection network; s3, training an MIDDNet insulator defect detection model; and S4, carrying out power transmission line insulator defect detection through the trained MIDDNet insulator defect detection model. According to the method, the overall parameter quantity of the detection model is reduced, the fusion effect of multi-scale feature fusion is enhanced, and dense context feature information is extracted, so that the detection precision and robust performance of the model are finally enhanced. The method can provi