Abstract: With the increase in the service life of masonry hydraulic culvert, the existing culvert structure needs to be quickly and accurately detected and evaluated. The detection method based on digital image can significantly improve the detection efficiency, but the application of this method is faced with many challenges due to the poor lighting conditions and narrow space in the detection of hydraulic culvert diseases.In view of the above problems, a hydraulic culvert disease detection model based on YOLOv5 network is established. Through the target detection of the collected culvert images and optimization of the recognition results, the rapid identification of hydraulic culvert diseases is realized. Mosaic data enhancement, multi-scale training and Adam optimizer are used to optimize and analyze the performance of the model to improve the reconstruction effect of the model. The application results of field examples show that the YOLOv5 network can quickly and efficiently identify hydraulic culvert diseases. This method can overcome the appearance interference characteristics of hydraulic culvert, and its recognition accuracy is consistent with the actual situation. It can accurately identify the disease grade of hydraulic culvert, and has a broad application prospect.