基于相关特征约束的分水岭路面裂缝图像分割算法  

Watershed pavement crack image segmentation algorithm based on the correlative feature constraint

作　　者：刘丰年[1] 朱平哲[1] 

机构地区：[1]三门峡职业技术学院信息传媒学院,河南三门峡472000

出　　处：《微型机与应用》2016年第3期96-99,共4页Microcomputer & Its Applications

摘　　要：针对路面裂缝图像的分割问题,提出一种相关特征约束的分水岭分割算法。该算法基于形态学分水岭思想,为了解决分水岭过分割问题,分析了裂缝图像狭长山谷的特点,选择从梯度图像的高频成分中提取与裂缝特征信息相关的局部极小值,从而约束伪极小值的出现,再对原始梯度图像强制标记,消除了过分割现象。同时解决了在消除过分割时边缘信息被滤除的问题,保留了细微有用的边缘。实验表明,该算法分割出的路面裂缝图像准确清晰,有效降低了过分割现象。In order to deal with the problem of the pavement cracks segmentation in the images,a related feature constraint watershed segmentation algorithm is suggested in this paper. The algorithm is based on the theory of the morphology,in order to solve the problem of over segmengtation of watershed,it analyes the characteristics of the narrow valley of crack image,selects feature information extraction and crack local minima associated with high-frequency component from the gradient image,and thus bounds appearance of pseudo-pole minimmum values. And then the original image gradient are forced flag,over-segmentation is eliminated. Meanwhile the problem of edge information will be filtered out is solved,and the slight and useful edge information are retained. Experimental results show that the algorithm of pavement crack image segmentation is accurate,clear and effectively reduces the over- segmentation phenomena of the watershed.

关 键 词：路面裂缝 分水岭 过分割 局部极小值 特征约束 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]