摘要
基于机器视觉的裂缝病害检测中存在虚假信息和断点的问题,需要有效的阈值分割和断点连接方法.基于边缘作为判别裂缝的关键,将梯度算子与应用广泛的Otsu相结合进行算法改进,并将其应用于采集图像的二值化,取得了理想的效果.裂缝检测中不同程度的断裂现象,对后续目标识别和目标提取等带来极大的困难.综合考虑实际情况(断接部分在整个裂缝中所占比例小,形态学处理已经对裂缝区域有影响),将裂缝断接段以一簇单像素宽的形式表示,确定待连接段的起点终点坐标,辅以直线顺序法和生长操作进行处理实现连接,便于后续的裂缝信息提取.
In the image processing of cracks detection systems based on the machine vision, the problems of false information and breakpoint need more effective threshold segmentation and break-connection algorithms. Taking edge as the key to distinguish cracks, authors combined gradient operators with the widely applied Otsu method to improve the algorithm and got satisfied results. Break phenomena with different degrees brought great difficulties to the subsequent crack recognition and extraction. Considering the actual situation （breakout occupy a fairly small proportion of the whole cracks and the previous processing have had impact on fracture zone）, so the authors first fixed the endpoints by the chain structuring elements with one pixel wide. Then proposes the connection algorithm supplemented by the linear sequence method and region growing so as to lay the foundation for subsequent processing.
出处
《应用基础与工程科学学报》
EI
CSCD
北大核心
2017年第4期742-750,共9页
Journal of Basic Science and Engineering
基金
国家自然科学基金项目(51278423
51478395)
关键词
阈值分割
梯度算子
断点连接
生长操作
threshold segmentation
gradient operator
break-connection algorithm
region growing
