Inspection of Rail Surface Defects Image Based on Histogram Processing by the Judgment Threshold

It is a challenge to accurately detect the defects under the influence of light, reflection, shadow and rust. This paper optimizes the background difference method as preprocessing by reducing the template so that the influence of shadow can be reduced, and presents a histogram processing algorithm based on the judgment threshold to detect defects of the rail images. The judgment threshold is obtained by Otsu method. As we all know, the applicable condition of the Otsu method is the proportion of the target and background close to each other, and in order to achieve this kind of condition, the histogram processing is to remove partial histogram which is obtained by the judgment threshold. The histogram processing is performed cyclically and a new threshold for judging is obtained by Otsu method on the remaining histogram after the every loop. The constraint formula is introduced to make the threshold converge to the fixed value. Finally, the image is segmented by this threshold. The experimental results show the benefits of the proposed algorithm comparing to the existing algorithms.