Abstract: The quality of cables is directly related to the safety of the power supply.The recognition of the cross-sectional image of the cable insulation layer is the key technology to realize the automatic and fast measurement of the cable insulation layer parameters.For the problem of many types of cables,small differences within categories,and large differences between categories,a method for classifying cable insulation cross-section images based on the deep residual network was proposed.18 typical cable insulation cross-section images were divided into three categories:circle,quasi-circle,and other shapes for feature analysis.On this basis,a multi-level classification model of the cable insulation layer based on a deep residual network was constructed.The experimental results show that this method fully extracts the deep-level features of the cross-sectional image of the cable insulation layer.The classification accuracy is as high as 99.99%,and the generalization is good.It can meet the needs of automatic detection of cables.