基于二值形态学算子的轨道图像分割新算法  

A New Algorithm for Orbital Image Segmentation Based on Binary Morphological Operator

作　　者：张雯柏 柴晓冬[1] 郑树彬[1] 彭乐乐 李立明[1] ZHANG Wen-bai;CHAI Xiao-dong;ZHENG Shu-bin;PENG Le-le;LI Li-ming(School of Urban Rail Transportation,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学城市轨道交通学院,上海201620

出　　处：《测控技术》2018年第10期10-13,21,共5页Measurement & Control Technology

基　　金：国家自然科学基金项目(51405287);上海工程技术大学研究生科研创新项目(E3-0903-16-01258)

摘　　要：为了高效地提取轨道图像特征空间,采用改进的Sobel算子对轨道结构中的扣件和钢轨轮廓进行边缘检测,结合二值形态学理论处理检测结果,使用结构元素尺度变换条件下的开闭序列级联滤波方法,得到特征空间像素面积相互独立的轨道图像,并使用统计分析的方法计算出特征空间像素面积大小,确定出扣件和钢轨的特征区域通过在不同天气状况下的多组对比实验,验证了本方法能够在不同的温度和光照环境中,对轨道图像中的特征区域进行稳定、准确、高效的提取。In order to efficiently extract the orbital image feature space, the improved Sobel operator is used to detect the edge of the fastener and rail profile in the track structure, combining the binary morphology theory to deal with the test results, then the orbital images with independent featme space pixel area are obtained by u- sing the opening and closing sequence cascaded filter method under the condition of the scale transformation of structural elements, and the statistical analysis method is used to calculate the pixel size of feature space, the characteristic legions of the rail and fastener are determined. Through multiple sets of contrast experiments under different weather conditions, it is proved that the method can extract the feature regions of orbital images stably, accurately and efficiently under different temperature and illumination environments.

关 键 词：轨道图像 SOBEL算子 二值形态学 结构元素 

分 类 号：TN911.7[电子电信—通信与信息系统]