Research on Power Line Insulator Defect Detection Method Based on TR-MobileViTcense

With the rapid development of industrial economy, the supply demand of electricity consumption is becoming more and more prominent, and the safety and sustainability of power transmission in the power industry is increasingly becoming the focus. As an important insulating part of transmission equipment, the insulator of transmission line is prone to defects in complex environment, which poses a great challenge to safe transmission. The existing insulator defect detection methods based on deep learning often rely on drones to complete, but it is difficult to achieve the balance between detection accuracy and model lightweight. Under the strategic background of smart grid, it is an urgent problem to be solved. In this paper, the application of lightweight network to the defect detection of transmission line insulators is deeply studied. For the task of accurate defect detection, the TR-MobileViT algorithm is proposed in this paper. In the stage of feature fusion, the traditional two-feature fusion is replaced by three-scale feature fusion. In the stage of three-scale feature fusion, based on the small scale feature sampling easy to lose semantic information, the traditional sampling method is replaced by transposed convolution and the modified attention mechanism module EDCA is added to enhance the learning ability of the network. By comparison with many mainstream models, the highest accuracy is obtained, and the mAP accuracy reaches 95.3 %. When the number of parameters is 16.1 M, the balance between precision and number of parameters is achieved.