|
The rapid development of rail transit has put forward higher requirements for the reliability testing of all components of rail lines. There are some drawbacks in relying on manual handheld device detection, so it is urgent to develop an intelligent and convenient rail fastener detection system. This paper introduces a rail fastener detection method based on YOLOV5, which can identify fasteners by training existing classified pictures, and continuously improve the accuracy of rail fasteners by modifying training documents and capturing frameworks. In this experiment, through training and training of the classified and labeled pictures, take 500 or 600 pictures as a group and 200 rounds as an experiment to improve the recognition accuracy of the experiment, so as to identify the position of the rail fastener with a suitable frame, and classify the identified fastener into its category. Through the analysis of the pr value of the experimental results, the experimental parameters are continuously modified to achieve the above effect. Finally, the pr value obtained from the experiment is about 0.6.
|
