【作者】 韦冬冬； 赵豫红；

【Author】 WEI Dong-dong, ZHAO Yu-hong (Department of control science and engineering, Zhejiang University, Hangzhou, 310027, Zhejiang, China)

【机构】 浙江大学控制科学与工程系；

【摘要】 针对图像处理中多个颗粒重叠的问题,提出了一种利用重叠区域边界寻找凹点来分割重叠图像的算法。算法利用canny边缘检测获得重叠区域的边界轮廓,对于边缘检测所得边界存在断裂的情况,首先进行连续边界轮廓的恢复;然后在获得的连续边界轮廓上寻找凹点,将相邻的凹点视为一个凹点群,取其中凹陷最显著的点作为待匹配凹点,并赋以一定的权值;再根据凹点匹配条件对待匹配凹点进行匹配,将成功匹配的凹点对作为分割点对。对多幅模拟图像和PTA晶体图像的分割结果表明,算法不仅适用于颗粒形状规则、大小差异较小颗粒的重叠情况,对于狭长颗粒和大小差异明显颗粒的重叠情况也有良好的分割效果。更多还原

【Abstract】 To deal with overlapped particles in image processing, a segment algorithm is proposed by matching concave points that are obtained from the edge of the overlapped area, which is obtained by canny edge detection. In the case that there is crack during the edge from the edge detection, the continuous edge should be recovered. Then the concave points are searched by judging the edge points and the sequential concave points are regarded as a cluster during which the most concave point is to be matched. Each concave point is given a particular weight and matching criteria is applied to these concave points and the successfully matched concave points are used to segment the overlapped particles. The segment results for simulation images and real images of PTA particles demonstrated that the algorithm is suitable for the situation that regular-shaped particles or overlapped particles with similar size. What’s more, the algorithm is suitable for the situation that long-narrow particles or particles with wide-range-square overlap with each other.更多还原