基于改进YOLOX-S的接触网绝缘子缺陷检测

Defect detection of catenary insulator based on improved YOLOX-S

针对接触网绝缘子缺陷检测,提出了一种改进YOLOX-S的接触网绝缘子缺陷检测算法。在Backbone网络CSPNet结构中的残差模块部分引入CA模块,使神经网络能够将位置信息融入通道特征进行注意力调节;为减小正负样本不平衡带来的误差,置信度预测损失采用Focal Loss;同时,将EIoU Loss作为边界框的回归损失,使损失函数对真实框和预测框更加敏感,提升算法的检测精度。实验结果表明,改进的算法在接触网绝缘子缺陷数据集上mAP值达到90.88%,相比YOLOX-S提升了7.47%,证明了所提算法的真实有效性。

For catenary insulator defect detection,an improved YOLOX-S algorithm for catenary insulator defect detection is proposed.CA module is introduced into the residual module in the CSPNet structure of Backbone network,so that the neural network can integrate the position information into the channel features for attention adjustment;In order to reduce the error caused by the imbalance of positive and negative samples,Focal Loss was used to predict confidence loss;At the same time,the EIoU Loss function is used as the regression loss of the boundary box,making the loss function more sensitive to the real box and the prediction box,and improving the detection accuracy of the algorithm.The experimental results show that the mAP value of the improved algorithm on the catenary insulator defect data set reaches 90.88%,which is 7.47%higher than YOLOX-S,which proves the validity of the proposed algorithm.