Research on UAV image recognition and fault classification for intelligent power inspection system using deep learning

With the advancements in artificial intelligence (AI) and unmanned aerial vehicle (UAV) inspection technologies, enhancing the automated accuracy of UAV inspection images and ensuring stable power grid operation has become a crucial task. Applying AI algorithms to detect targets and identify defects in UAV inspection images is an inevitable trend towards the construction of smart grids. This study proposes an improved image recognition algorithm based on deep learning, specifically an enhanced Faster-RCNN model combined with HSV color feature extraction technology. This integrated approach enables real-time processing of UAV images and recognition of various common faults in the power grid system. By converting UAV-transmitted images into HSV models and comparing pixel values with color features of grid equipment in a database, the system can accurately determine the presence of faults. Furthermore, leveraging deep learning algorithms enhances the precision of target defect recognition in UAV inspection line image datasets. Experimental simulations demonstrate that the proposed model achieves a recognition and fault classification accuracy of 92.81%, outperforming conventional methods. In conclusion, the intelligent power inspection system's image detection and fault classification model based on deep learning AI algorithms exhibits precise computational performance, fostering the development of smart grids.