Research on Recognition Method of Electrical Components Based on YOLO V3

• Main Authors: Haipeng Chen, Zhentao He, Bowen Shi, Tie Zhong
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Deep Learning; SRCNN; YOLO V3; electrical components; object detection
• Online Access: https://ieeexplore.ieee.org/document/8886369/

The reliability of electrical components affects the stable operation of the power system. Electrical components inspection has long been important issues in the intelligent power system. The main problems of traditional recognition methods of electrical components are low detection accuracy and poor real-time performance, which are challenging to extract necessary features from the inspection images. This paper proposes a way to detect the electrical components in the Unmanned Aerial Vehicle (UAV) inspection image based on You Only Look Once (YOLO) V3 algorithm. Due to some of the inspection images are not clear, which result in the reduction of the available dataset. On this basis, we adopt Super-Resolution Convolutional Neural Network (SRCNN) to realize super-resolution reconstruction on the blurred image, which achieves the expansion of the dataset. We compare the performance of the proposed method with other popular recognition methods. The results of experiment verify the effectiveness of the proposed method, and the technique reaches high recognition accuracy, good robustness, and strong real-time performance for UAV power inspection system.