Improved Yolov5 and Image Morphology Processing Based on UAV Platform for Dike Health Inspection

Dike health inspection is crucial in river channel regulating. The conventional manual collapse inefficient and costly so that the unmanned aerial vehicle (UAV)-based has been widely applied. However, existing vision-based defect detection methods face challenges, such as lack of sample data closed specified sets. To address them, a method based on improved YOLOv5 recognition combined with image morphology processing proposed for dike zero samples. Specifically, coordinate attention mechanism introduced model to improve capability dikes. Also, rotating bounding box target designed arbitrary orientation dikes under UAV view, due ineffective horizontal detection. Furthermore, suspected locating efficiency promotion, specific recognized area isolated process. results show outperforms traditional Yolov5 algorithm recall rate, F1, mAP.