叶菜根系图像的变分水平集分割算法  

Segmentation of Leafy Vegetable Root Images Based on Variational Level Set Method

作　　者：冯太锐 苗玉彬[1] 朱云开[2] 赵爽[1] FENG Tairui MIAO Yubin ZHU Yunkai ZHAO Shuang(School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China Xin Hua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai 200092, China)

机构地区：[1]上海交通大学机械与动力工程学院,上海200240 [2]上海交通大学医学院附属新华医院,上海200092

出　　处：《东华大学学报（自然科学版）》2017年第4期552-558,570,共8页Journal of Donghua University(Natural Science)

基　　金：上海市科研计划资助项目(16391901700);上海市卫生局青年课题资助项目(20134y038);上海交通大学医工交叉课题资助项目(YG2013MS15)

摘　　要：植物根系图像分割是根系构型特征提取和分析的前提.针对传统图像分割方法在处理叶菜根系弱边缘图像中存在分割精度和稳定性较差的问题,提出了一种基于改进C-V(Chan-Vese)模型的变分水平集分割算法.该算法不仅保留了C-V模型对于处理弱边缘图像的适用性,并针对叶菜根系图像局部灰度不均的特点引入了图像梯度信息,改进了原C-V模型.通过对小白菜根系样本图像的分割处理试验,证明了变分水平集分割算法的有效性.研究结果表明,相比传统的阈值处理、边缘检测及区域生长等算法,本文算法能更加精细地解决叶菜根系图像弱边缘和局部灰度不均的问题,并在分割精度和算法稳定性上具有明显的优势.变分水平集算法应用于叶菜根系构型观测系统中,可以有效地提高观测精度.The segmentation of plant root image is the precondition of root feature extractiort and analysis. In order to improve segmentation accuracy and stability in leafy vegetable root images with weak contour, a variational level set segmentation algorithm based on improved C-V（Chan-Vese） model was presented. The algorithm remained the applicability of C-V model to segmentation of images with weak contour, and moreover, it had been improved by introducing the image gradient information to solve the problem of grayscale non-uniformity. The validity of the algorithm was proved by segmenting root images of pakchoi cabbage. The results show that comparing with the traditional segmentation algorithms such as threshold method, Gaussian-Laplacian edge detector and region growing method, the presented algorithm can better solve the problems of weak contour and grayscale non-uniformity and achieve better segmentation precision and robustness. Variational level set method can be applied in leafy vegetable root observing and measuring system to effectively improve the system precision.

关 键 词：叶菜根系 图像分割 变分水平集方法 C-V模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]