Rail Surface Defects Detection Based on Faster R-CNN

Rail surface defects detection is very important for improving railway safety. Therefore, we use Faster R-CNN to conduct research on the rail surface defects detection. Firstly, a data set of rail surface defect images is established, and then the training set of random segmentation is used for effective training to a certain extent, and the corresponding verification set is verified. After undergoing corresponding effective training, the model is tested accordingly. From the detection results we can see that AP is 97.8%. Compared with other algorithms, the recognition effect of Faster R-CNN is higher, therefore this method can efficiently identify the rail surface defects.