基于改进YOLOv8s的绝缘子水冲洗检测方法OA

In the task of insulator water washing effect detection using UAVs,the influence of variable lighting and weather con-ditions significantly increases background noise complexity.Meanwhile,the insulators occupy a small proportion in UAV-captured images,and their target features are not prominent.These factors result in lower detection accuracy of models in such scenarios,making it difficult to meet practical application requirements.To address this challenge,an improved YOLOv8s-based method for insulator water-washing detection is proposed.Firstly,the Mean Shift algorithm is used for adaptive segmentation of the insulator image feature maps,separating complex backgrounds from target regions to reduce background noise interference.Secondly,a convolutional attention module is introduced to dynamically adjust the weights of channel and spatial features,en-hancing the key features of insulators.Finally,based on the optimized feature maps after segmentation,an improved YOLOv8s is employed for water washing effect detection.To verify the effectiveness of the proposed method,images of water-washed insula-tors are collected by the Zhifei W330 washing device and the DJI FlyCart 30 UAV platform for experimental validation.The ex-perimental results show that the proposed method achieves an average precision of 96.4%,outperforming classic detection mod-els such as Fast R-CNN,SSD,Deformable-DETR,YOLOv5s,YOLOv8s,and YOLOv11s.Compared with traditional detection models,the proposed method improves the accuracy of water washing effect detection by approximately 12.3 percentage points.