For the detect recognition of transmission line insulators, there are some problems for the traditional fault recognition algorithms, such as false detection, omission of detection, low and slow recognition rate. Compared with the traditional convolutional neural networks, vector is used by the capsule network as the input. Each sub-structure of the capsule makes the details highly fidelity in the original graph, which can effectively identify the defective insulator image. Therefore, a detect recognition method based on improved capsule network and YOLO-V5 is proposed in this paper. Experimental results demonstrate that the algorithm performs well and meets the requirements for inspecting insulators.