基于灰度对比图与最大熵的钢轨图像分割

针对CCD相机采集的钢轨表面图像灰度不均、过度曝光、噪声过大等导致的一维最大熵法无法准确分割缺陷的问题,提出一种将灰度对比图和形态学重构相结合,再用最大熵法进行分割的图像分割算法。首先求得钢轨图像的灰度对比图,接着将灰度对比图经过形态学开-闭重构获得钢轨背景图像,然后将背景图像与灰度对比图相减,得到包含缺陷的差分图,最后将差分图用最大熵法进行分割。实验表明,提出的灰度対比图法能够很好地缓解光照不均,过度曝光对检测带来的影响,形态学开-闭重构不仅能够获得所需的背景模型还能抑制一定的噪声,该算法简单、有效、鲁棒性较高,分割精度可以达到90%。

Image Segmentation Based on Gray Contrast and Maximum Entropy

Due to the uneven gray level,overexposure and excessive noise of collected images,the one-dimensional Maximum Entropy can not accurately segment the defects of the rail images. This paper presents an image segmentation algorithm based on gray contrast and morphological reconstruction and the Maximum Entropy to segment images. Firstly,the gray contrast image of rail images is obtained. Then the gray contrast image is reconstructed by morphological reconstruction,and the reconstructed image is subtracted by gray contrast image to get the difference graph containing the defects. Finally, the difference graph is segmented by the Maximum Entropy. The experimental results show that the gray contrast image proposed in this paper can well alleviate the effects on detection caused by uneven illumination and overexposure,and the morphological reconstruction can not only obtain the desired background model but also suppress the noise. This algorithm is simple,effective and robust,and the segmentation accuracy can reach up to 90%.