基于ME-YOLO算法的输电线路绝缘子缺陷检测研究OA

To address the shortcomings of traditional defect detection algorithms in multimodal and multi-type defect recognition on insulators,particularly in the context of small objects and complex backgrounds in aerial images,we design an MGDB module to improve C2f,resulting in C2f_MGDB.This module replaces C2f in the head of the YOLOV10n network,lightweighting the model while improving its feature extraction capabilities.Furthermore,an EMA attention mechanism is implemented in the Neck region to enhance the model's object perception in complex multimodal backgrounds.Meanwhile,an Inner-Wise-MPDIoU loss function is introduced to replace the native YOLOV10n loss function,thereby improving model training performance and detection accuracy.Finally,the original detection head is replaced with a LSCD lightweight detection head,significantly reducing the number of model parameters and computational overhead while maintaining detection accuracy.The resulting Multimodal Enhanced YOLO(ME-YOLO)model is exper-imentally tested on a multimodal insulator dataset.Experimental results show that the improved ME-YOLO model reduces the number of parameters by34.42%and the computational overhead by10.29%,while improving mAP50 by4.6%.It is proved that the proposed method not only significantly improves the detection effect of multi-modal and multi-type insulator defects,but also reduces the complexity of the model.