接触网风偏检测的亮度矩寻优分割算法

An intensity-adaptive color image segmentation based on contrast stretching and its application in visual detection of catenary wind-deviation

介绍电气化铁道接触网风偏视觉检测原理,以视觉检测的图像处理为背景提出一种能够适应不同光照条件的彩色图像分割算法。该算法融合彩色图像的一阶亮度矩和对比度信息,通过线性亮度寻优优化图像的一阶亮度矩,并以图像的初始一阶亮度矩为基准亮度进行对比度拉伸,实现对目标区域颜色特征的放大;同时,以特征点靶面的像素面积为目标值,以不同寻优亮度下图像的合格像素点统计量为寻优变量,对两者的关系曲线和最佳目标逼近方式进行了深入研究,确定合理的初始寻优亮度及寻优方向。最后,以现场采集图像为训练样本对算法进行验证和分析。研究结果表明:该方法能够实现彩色空间下接触网风偏目标特征点的有效、准确分割,对复杂光照条件具有良好的适应性和算法鲁棒性。

Wind deviation detection principle of electrified railway catenary was introduced, and a novel color image segmentation method, which was adaptive to complex illumination condition, was presented for image processing of visual detection. Incorporating both contrast and brightness parameters of color image, dynamic image brightness adjustment was implemented in linear step, and image contrast was then stretched with the initial image brightness defined as the reference brightness, through which color features of the target region was magnified. Defining the pixel area of target surface as the objective value, and total eligible pixel number under different intensity as the optimizing variable, correlation curves and optimal approaching mode were further studied. Reasonable initial searching brightness and searching direction were thus determined. At last, the suggested algorithm was tested with image samples from the spot. The results show that the recommended method is helpful in segmenting catenary targets effectively and accurately, and thus has good adaptability and strong robustness to complex illumination conditions.