AN IMAGE-BASED METHOD FOR FASTENER INSPECTION

For high-speed railway operation security, it is very important to inspect the status of fastener regularly and issue alert information in time. In this paper, an image-based inspection method for the status of fasteners is proposed. Firstly, a fast enhancement algorithm based on the Retinex model is proposed in order to reduce the influence of non-uniform illumination. Then, the sub-images of fasteners are obtained on the basis of locating the rails and crossties with space prior knowledge. Finally, a SVM classifier is trained to identify the status of the fasteners according to the LBP (Local Binary Pattern) features. Performance of the proposed method is tested on the image data of Thong Railway in China and the results show that it is effective and practical with high accuracy and low omission rate.