玉米籽粒考种信息获取装置设计与试验

发布时间：2018-05-16 13:15

本文选题：玉米籽粒 + 自动考种　；参考：《农业机械学报》2017年12期

【摘要】：考种是制约育种效率的关键环节。玉米高通量考种过程,存在籽粒堆积和粘连现象,影响籽粒考种参数的提取。本文结合玉米高通量自动考种需求,设计了籽粒考种信息获取装置。通过分析堆积籽粒回旋运动过程的受力情况并根据试验情况确定振动平台回旋速度,实现籽粒的平铺摊种。在此基础上,针对粘连籽粒图像提出了一种先分割后融合的改进分水岭算法,该方法通过比较相邻分割区域极小值与最小分水岭的差值与设定的阈值T,进行邻域融合,对过分割区域进行合并,实现粘连籽粒的准确分割,分割完成后,统计籽粒个数,并基于Graham扫描法建立单个籽粒的最小外接矩形,获取籽粒长宽参数。在构建的玉米籽粒自动考种装置上进行动态试验,结果表明,本文所提出的方法可实现玉米粘连籽粒的准确分割,单穗玉米籽粒计数正确率不低于98.05%,籽粒平均长宽与人工测量结果的决定系数R~2在0.97以上,满足自动考种在线检测的需求。
[Abstract]:Seed examination is the key link to restrict breeding efficiency. In the process of high flux maize seed examination, there exists the phenomenon of grain accumulation and adhesion, which affects the extraction of seed parameters. In this paper, according to the requirement of maize high throughput automatic seed testing, a device for obtaining seed information was designed. Based on the analysis of the stress on the rotation motion of the accumulative grains and the determination of the vibration platform rotation velocity according to the experimental conditions, the seed spreading and spreading were realized. On this basis, an improved watershed algorithm based on segmentation and fusion is proposed for the conglutinated grain image. By comparing the difference between the minimum value of the adjacent segmentation region and the minimum watershed value and the set threshold value T, the proposed algorithm can be used for neighborhood fusion. Combining the over-segmented regions to realize the accurate segmentation of conglutinated grains, the number of grains is counted after the segmentation is completed, and the minimum outer rectangle of a single grain is established based on Graham scanning method, and the parameters of grain length and width are obtained. Dynamic experiments were carried out on the automatic seed testing device for maize grains. The results showed that the method proposed in this paper could achieve the accurate segmentation of corn grains. The accuracy rate of grain counting of single ear maize is not lower than 98.05, and the determination coefficient Rn2 of the average length and width of grain and the result of manual measurement is more than 0.97, which can meet the requirement of automatic seed on-line detection.
【作者单位】：北京农业智能装备技术研究中心;国家农业智能装备工程技术研究中心;农业部农业信息技术重点实验室;
【基金】：北京市科技计划项目(D151100004215002) 国家博士后基金项目(2015M581019)
【分类号】：S513;TP391.41

【相似文献】