| abstract |
An insulator defect detection method based on semantic segmentation and target detection is disclosed. In the method, an insulator image is collected, an insulator body is marked based on the insulator image, and a semantic segmentation data set is constructed; a semantic segmentation model is built; and a mask image of the insulator is generated based on the semantic segmentation. , and annotate the mask map to obtain a defect detection data set; put the defect detection data set into the yolov3 target detection network for training, and the trained semantic segmentation model detects the defect position in the mask map , and map the bounding box of the defect position to the defect detection data set to realize defect detection. |