Improved YOLOv5s Method for Nut Detection on Ultra High Voltage Power Towers.

As an important power transmission facility, the operational stability and safety of ultra high voltage (UHV) power transmission towers are crucial to energy supply and social stability. With the continuous development of science and technology and artificial intelligence technology, the research and development of UHV tower maintenance robots has become an inevitable trend. However, the existing vision-based maintenance robot research is not yet mature, and there are problems such as poor real-time performance, low positioning accuracy, large distance measurement error, and poor performance of embedded devices. To solve these problems, this paper proposes a lightweight nut object detection algorithm based on YOLOv5s and MobileNetV3. In addition, experiments were conducted on the generated nut dataset. Compared with YOLOv5s, the proposed method can reduce the model size by 77.78% and increase the detection speed by 4.17%, and improved accuracy by 0.73%. The experimental results show that the improved algorithm greatly reduces the model size and improves the detection speed while maintaining the original accuracy, and effectively solves the problems of poor real-time detection and poor performance of embedded devices in existing methods.