Research on Rail Surface Defect Detection Method Based on UAV Images

Abstract

In view of low efficiency and high cost of rail surface defect detection, an unmanned aerial vehicle (UAV) inspection scheme is presented in this paper. UAV digital images are often badly degraded by noise during dynamic acquisition and transmission process. Wavelet threshold and median filtering are the main denoising methods for noisy image, however, the wavelet threshold denoising method is insufficient for UAV digital image denoising. Thereby the method that can reduce the noise of image by using wavelet transform combined with median filtering (WTCMF) is proposed in the paper. And then a new method named Hough-based pixel column cumulation gray (HPCG) for extracting rail regions is proposed in this paper. Finally, maximum entropy (ME) algorithm is used to detect rail surface defects. The Peak Signal to Noise Ratio (PSNR) experiment is carried out based on the different denoising methods. And the proposed method is used to perform the experiment of the rail defect extraction. Experiments show that this method can effectively eliminate the influence of noise and have better edge detection ability, which can effectively detect rail surface defects based on UAV images.