农机化研究Issue(12)：222-226,231,6.
基于计算机视觉的玉米穗粒重无损测量方法
Methods of Nondestructive Measurement for Kernel Weight of Maize Based on Multi-source Image Fusion
摘要
Abstract
At present , traditional methods such as eye-measurement and dipstick metering are generally used in maize varieties test interiorly , which is a waste of time and energy , and always account to a large calculation error .Besides , it has to be threshed before kernel weight testing , which means the tested materials could not be traced back , that is not suitable for extensive and meticulous breeding .This paper presents methods for extraction of 3 D phenotypic characteris-tics ( row number , bald length , grain plumpness and so on ) and internal structure features ( ear diameter , grain depth and so on ) of maize respectively based on visible light imaging technology and X optical transmission imaging technology . Furthermore , according to the parameters measured by these methods , three calculation models were built to calculate grain weight per ear: model based on single kernel weight , model based on volume-weight and model based on axis weight .The results show that error fluctuations of both models based on single kernel weight and based on volume -weight are relatively large and greater dependence on varieties , while the former is better than the latter .By comparison , the model based on single kernel weight is better .Nevertheless , the model based on cob weight has a high degree of accura-cy, whose average error precision can reach 1.30%, which has little data fluctuation , and the measurement results are stable , which meets the needs of extensive maize testing , and provides the basic data and model for the goal of high throughput and meticulous breeding .关键词
穗粒重/无损测量/计算机视觉/可见光成像/X光成像Key words
grain weight/nondestructive measurement/computer vision/visible light imaging/X-ray lig分类
农业科技引用本文复制引用
周金辉,马钦,朱德海,王越,张晓东,刘哲,李绍明..基于计算机视觉的玉米穗粒重无损测量方法[J].农机化研究,2015,(12):222-226,231,6.基金项目
中央高校基本科研业务费专项（2014 JD042）；公益行业（农业）科研专项 （201203026）；教育部博士点基金项目 （）
