Sample confrontation expansion method and system for wire insulation layer damage identification

The invention relates to the technical field of electric power inspection, and provides a sample confrontation expansion method and system for wire insulation layer damage identification, and the method comprises the steps: obtaining an insulated wire image, marking the insulation layer damage type, and constructing a wire insulation layer damage data set; inputting the wire insulation layer damage data set into a condition generation network to obtain an expansion data set; wherein the conditional generative network adopts deformable convolution, an offset is added to the position of each sampling point of a convolution kernel, the offset is learned, and the sampling position is moved to a region of interest by modulating the variation amplitude of the learned offset. The method can adapt to insulating layer damage samples in different states, and the recognition precision of the wire insulating layer damage recognition model obtained based on customized sample data training is improved. 本发明涉及电力巡检技术领域，提供了一种用