基于卷积神经网络的绝缘子目标检测方法研究

Research on Insulator Target Detection Method Based on Convolutional Neural Network

绝缘子是高压输电线路中的必备装置,为实现航拍图像中绝缘子的快速准确检测,提出基于卷积神经网络的绝缘子目标检测方法。首先介绍了基于区域候选的卷积神经网络Faster R-CNN;然后基于本实验构建了相应的数据集,描述了Faster R-CNN模型的训练过程,提出了以DarkNet作为Faster R-CNN的特征提取网络,并与ResNet对比;最后根据实验结果验证了Faster R-CNN用于绝缘子检测的可行性。ResNet和DarkNet在检测效果上都取得了不错的表现,但DarkNet作为特征提取网络在检测效率上要高于ResNet,是电力巡检在绝缘子目标检测中的有效探索。

Insulators are a necessary device in high-voltage transmission lines.In order to achieve rapid and accurate detection of insulators in aerial images,a target detection method for insulators based on convolutional neural networks is proposed.This paper introduces the convolutional neural network Faster R-CNN based on region candidates.Then built a corresponding data set based on this experiment,described the training process of the Faster R-CNN model,and proposes DarkNet as the feature extraction of Faster R-CNN Network,and compared with ResNet.Finally,according to the experimental results,the feasibility of Faster R-CNN for insulator detection is verified.Both ResNet and DarkNet have achieved good performance in detection effects,but DarkNet as a feature extraction network requires detection efficiency.