基于机器视觉的水稻秧苗图像分割  

Machine vision based segmentation algorithm for rice seedling

作　　者：袁加红[1] 朱德泉[1,2] 孙丙宇[3] 孙磊[1] 武立权[2,4] 宋宇[1] 蒋锐[1] 

机构地区：[1]安徽农业大学工学院,安徽合肥230036 [2]安徽省粮食作物协同创新中心,安徽合肥230036 [3]中国科学院合肥智能机械研究所,安徽合肥230031 [4]安徽农业大学农学院,安徽合肥230036

出　　处：《浙江农业学报》2016年第6期1069-1075,共7页Acta Agriculturae Zhejiangensis

基　　金：国家自然科学基金项目(51403005);国家农业科技成果转化项目(2014C30000162)

摘　　要：水稻秧苗的识别是水稻插秧机自主导航系统的关键内容之一。针对插秧机机器视觉导航中稻田图像秧苗与背景分割问题,建立了基于RGB(红绿蓝)颜色空间的秧苗表面颜色模型。通过颜色特征对秧苗图像进行处理,使用Photoshop软件获取秧苗部分和背景R,G,B分量值;通过对G-R值与G-B值的分析统计,发现两者之间存在分界关系:各自的权重与各分量的乘积之和为某个定值;为方便分析,选取权值a,b为0.5,即Ex G因子,采用Otsu法获取定值最佳值,最大程度分割出目标和背景。与适合于大多数绿色作物的传统RGB法进行比较,并采用分割质量因子和算法运算时间作为评判标准,分析各算法的综合性能。试验发现,Ex G因子结合Otsu分割法分割效果相对理想、稳定性更高,而且耗时更短。The recognition of rice seedling is one of the significant parts of autonomous guidance for rice transplan-ting. Considering the segmentation of seedlings and remainder based on machine vision system, a simple dichromatic reflection model was established in RGB color space, which represented that the seedling could be recognized by u-sing its color feature. The values of R, G, B components of seedlings and remainder were obtained in Photoshop soft-ware respectively and analyzed statistically in order to get the relation between them. In order to simplify the compu-ting process, the weight values of a and b were set as 0. 5, ExG index and Otsu method （ExG＋Otsu method） which could obtain the optimal threshold were combined to distinguish the seedlings and remainder well. The RGB method and previous ExG＋Otsu method were carried out to compare their performance intuitively. Their comprehensive per-formance was evaluated with segmentation quality factor and time consuming. The results have proved that the latter for segmenting was more efficient, highly stable and timesaving.

关 键 词：水稻秧苗 ExG因子 OTSU法 图像分割 质量因子 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]