Research on image processing algorithm for unmanned aerial vehicle inspection of damaged power lines in debris flow disasters

With the continuous advancement of computer vision and deep learning technology, unmanned aerial vehicle inspection is widely used in disaster assessment due to its advantages of high efficiency, safety, and accuracy. A drone inspection image processing algorithm for damaged wires in debris flow disasters is proposed to address the issues of complex background interference, variable target scales, and occlusion in the detection of damaged wires after debris flow disasters. With accuracy, recall, and F1 score of 95.3%, 94.2%, and 94.7%, respectively, the enhanced "You Only Look Once version 5" (YOLOv5) algorithm outperforms previous approaches in object recognition tasks of unmanned aerial vehicle inspection photos, according to the experimental findings. The improved YOLOv5 algorithm has the best target detection performance for unmanned aerial vehicle inspection images, with a missed detection rate of less than 5%. When the number of images is 100, the maximum missed detection rate is 4.8%. Verify the reliability of the improved YOLOv5 algorithm in object detection of unmanned aerial vehicle inspection images. The method proposed in the study has significant performance advantages in detecting damaged power lines in debris flow disasters, and can provide strong technical support for disaster response and power restoration.