Multi-feature Fusion Recognition Method of Rail Surface State Based on SVM Algorithm

In order to solve the problems that the rail surface status is still judged by manual experience and the recognition efficiency is low, a multi-feature fusion recognition method of rail surface status based on SVM algorithm is proposed. First, the rail surface image database is obtained by preprocessing the collected rail surface images in different states. The preprocessing includes image graying, image denoising and image extraction. Then, the gray mean and variance of the contact surface region are calculated to describe the color features of the rail surface image. And the texture features of the contact surface region are extracted using the gray level co-occurrence matrix. Then the two features are fused as the basis for judging the rail surface state. SVM is used to recognize the rail surface state. Finally, the proposed method is verified and analyzed by experimental simulation, and the results prove the effectiveness of the proposed method.