中国农业科技导报 ›› 2018, Vol. 20 ›› Issue (8): 46-53.DOI: 10.13304/j.nykjdb.2017.0614

• 农业创新论坛 • 上一篇    下一篇

基于X-ray透射成像的稻穗米粒粒型及谷粒饱满度测量

黄成龙1,杨万能1,2,吴迪1,段凌凤1*   

  1. 1.华中农业大学工学院, 武汉 430070; 2.华中农业大学作物遗传改良国家重点实验室, 武汉 430070
  • 收稿日期:2017-09-12 出版日期:2018-08-15 发布日期:2017-12-01
  • 通讯作者: *通信作者:段凌凤,讲师,主要从事植物表型组学研究。E-mail: duanlingfeng@mail.hzau.edu.cn
  • 作者简介:黄成龙,讲师,主要从事作物表型研究。E-mail: hcl@mail.hzau.edu.cn。
  • 基金资助:
    国家自然科学基金项目(31600287);中央高校基本科研业务费项目(2662018JC005);湖北省科研条件与资源研究开发项目(2015BCE044)资助。

Grain Size and Filling Degree Extraction for Rice Panicle Based on X-ray Imaging

HUANG Chenglong1, YANG Wanneng1,2, WU Di1, DUAN Lingfeng1*   

  1. 1.College of Engineering, Huazhong Agricultural University, Wuhan 430070;2.National Key Laboratory of Crop Genetic Improvement, Huazhong Agricultural University, Wuhan 430070, China
  • Received:2017-09-12 Online:2018-08-15 Published:2017-12-01

摘要: 稻穗是水稻产量、干物质积累的最终表现,稻穗米粒粒型和谷粒饱满度对水稻遗传育种及功能基因解析具有重要意义。传统稻穗米粒粒型测量需要经过脱粒、脱壳步骤,不仅过程繁琐而且可能有损伤、可靠性差,而谷粒饱满度测量主要通过水沉法或风选法,无法实现定量测量。为了实现在不脱粒、不脱壳情况下无损观测稻穗内部米粒,提出一种测量稻穗米粒粒型和谷粒饱满度的新方法。采用X-ray透射成像方式获取稻穗射线吸收图,然后采用自动阈值分割、分水岭、图像标记等算法得到单颗米粒标记图像;再采用椭圆检测、边缘提取计算得到米粒的粒长、粒宽、粒面积。对100株稻穗图像做测试,结果表明粒长、粒宽平均测量误差MAPE分别为2.14%、3.45%。同时采用K-means和高斯混合模型聚类(GMM)算法识别出米粒对应的谷粒区域,基于二者面积的比值得到稻穗谷粒饱满度的定量测量结果,基于该结果可以准确识别稻穗中的实粒和瘪粒,为稻穗复杂性状测量研究提供了一种新的有效途径。

关键词: X射线, 稻穗, 米粒粒型, 谷粒饱满度

Abstract: Rice panicles can indicate the final rice yield and dry matter accumulation. Rice grain shape and seed plumpness are of great significance for rice genetic breeding and functional genes analysis. However, the traditional measuring method for grain shape needed threshing and peeling steps, which was inaccurate and time-consuming. And seed plumpness was measured mainly by water-deposition or wind-separation method, which could not achieve the quantitative measurement. Therefore, this paper proposed a new measuring method  to obtain the rice grain traits accurately and fast without threshing and peeling. The imaging modalities was based on X-ray to transmis image of rice panicle. Then the X-ray image was also processed to obtain the single grain image by OTSU, watershed, and labelling algorithms. After that, ellipse detection and edge extraction were applied to extract grain length, width and area. Finally 100 rice panicles were tested, and the results showed that the MAPE was 2.14% and 3.45% for grain length and width, respectively. In addition, the K-means and GMM algorithms were adopted in this study to identify the grain area and husk area. The visual distribution of rice grain filling was obtained based on the ratio of the above areas, which was able to distinguish the filled grain from unfilled grain quantitatively. In conclusion, this study provided a new effective method for measuring the complex traits of rice panicle.

Key words: X-ray, rice panicle, grain shape, filling degree