【摘要】 针对雾霾天气对输电线路航拍图像成像质量的负面影响会降低输电线路典型金具缺陷检测识别算法的精确度问题，论文提出了一种基于改进AOD-Net网络模型和ReVGG识别算法的输电线路金具缺陷识别算法。改进的AOD-Net网络模型中引入特征融合模块，采用人类感知视觉MS-SSIM损失函数，提升去雾效果，修复图像色彩失真，提高图像质量。金具故障缺陷识别基于ReVGG识别算法，嵌入轻量型注意力模块SimAM提升网络的特征提取能力。实验结果表明：改进的AOD-Net算法的信息熵高于原算法4.25%，提高了雾气消除效果；与ReVGG算法相比，嵌入SimAM模块的算法识别准确率提高了2.1%；总体算法使用的模型占用空间小，实时性好，利于边缘部署。更多还原

【Abstract】 In view of the negative impact of hazy weather on the imaging quality of transmission line aerial images can reduce the accuracy of the typical metallic tool defect detection and recognition algorithm for transmission lines,this paper proposes a metallic tool defect recognition algorithm for transmission lines based on an improved AOD-Net network model and the ReVGG recognition algorithm. The improved AOD-Net network model introduces a feature fusion module and uses the human perceptual vision MS-SSIM loss function to improve the defogging effect,repair image colour distortion and improve image quality. Defect recognition is based on the ReVGG recognition algorithm,and the lightweight SimAM attention module is embedded to enhance the feature extraction capability of the network. Experimental results show that the improved AOD-Net algorithm has 4.25% higher information entropy than the original algorithm,which improves the fog removal effect,the recognition effect improves the accuracy by 2.1% compared with the ReVGG algorithm. The overall algorithm uses a model with small space occupation,good real-time performance and facilitates edge deployment.更多还原