Power insulator defect detection method based on improved YOLOX-s

An improved YOLOX- algorithm is proposed to solve the problem of missing and misdetecting insulator defects caused by small target, many kinds and large scale difference during UAV inspection. Firstly, the coordinate attention mechanism is integrated into the backbone network, so that the network can locate the insulator and its defect more accurately. Secondly, the CBS stack structure in the original network is replaced by Shufie unit in the feature fusion network to achieve model compression. Experiments show that the FPS of the improved model is 26.4 frames $/ \mathrm{s}$, and the mAP value reaches $\mathbf{9 5. 0 \%}$. Under the premise of not increasing the number of model parameters and computational complexity, the multi-type and multi-scale insulator defect detection has achieved better results. This method has practical significance to improve the operation and maintenance efficiency of power inspection service.