Insulator Defect Detection Based on Improved Faster R-CNN

In recent years, deep learning has been widely used to identify defects of insulators. This paper proposed an improved Faster R-CNN model based on deep learning to improve the accuracy of fault detection. This method is based on the original Faster R-CNN detection framework to make three improvements: First, ResNet50 is selected to replace VGGNet16 as the feature extraction network. Secondly, the feature pyramid network is used for feature fusion. Thirdly, RoIAlign is used to replace RoIPooling network to reduce the impact of quantization. The dataset in the experiment is 720 marked UAV aerial insulator images, which were divided into training set and test set according to the ratio of 8:2. The mAP of the improved network model reached 84.37%. Compared with the original framework, mAP increased by 7.52%. The results show that the improved network reduced the missed detection rate and false detection rate. On the basis of improving the recognition accuracy, it can better meet the needs of high accuracy in actual scenarios.

Automated summary

This paper proposes an improved Faster R-CNN model based on deep learning for the detection of insulator defects. By selecting a better feature extraction network, utilizing feature fusion, and reducing the impact of quantization, the proposed method significantly improves the accuracy of fault detection and achieves good results in the experiments.