基于H-Dome重构的大豆图像分割  

Image Segmentation of Soybean Based on H-Dome

作　　者：孙晓婷[1] 陈江红[1] 陈庆周[1] 

机构地区：[1]福建工程学院信息科学与工程学院,福建福州350118

出　　处：《大豆科学》2013年第6期821-824,共4页Soybean Science

基　　金：福建省"十二五"规划重点课题(FJCGZZ12-013)

摘　　要：在大豆品质检测过程中,大豆图像籽粒粘连给后续检测、分析等处理带来较大的困难。为此,提出了一种基于H-Dome重构的分水岭算法,先对二值图像进行H-Dome灰度形态重构合并极值,再对重建后的图像用距离变换得到距离图像,然后利用分水岭算法进行图像分割,最后提取分割线得到最终分离图像。该算法对籽粒的分割正确率为96.6%,分割效果较好。During quality inspection of soybean, the adhesion of soybean seeds in image brings a great difficulty to the posterior detection and analysis. Accordingly, a watershed algorithm based on H-Dome was put forward. After the local minimums were merged by the grayscale morphological reconstruction of H-Dome, the distance transform of reconstructed image was obtained. And then the watershed algorithm was used to segment the image. At last, the watershed line was extracted to obtain the final segmentation image. The segmentation ratio of proposed algorithm reached 96.6%.

关 键 词：大豆籽粒 图像分割 分水岭 H—Dome 

分 类 号：TS214.2[轻工技术与工程—粮食、油脂及植物蛋白工程]