摘要
[目的]小麦穗部小穗数及籽粒数能够直接反映小麦产量,也是小麦穗表型研究中2个非常重要的参数。[方法]为了快速测量这2个参数,针对小麦穗部正视图像,提出了一种基于图像处理技术的小麦小穗抛物线分割方法,并实现了小穗数及籽粒数的同步识别计数。首先采用图像预处理算法获得麦穗二值图像,然后将二值图像沿穗轴方向的像素按行求和,根据行像素求和曲线图中波峰波谷确定所需要的小穗3个位置点,由小穗3个位置点在二值图像上确定三点拟合抛物线,最后运用抛物线位置分割出各小穗,同时通过阈值法确定各小穗面积值与其籽粒数之间的关系。[结果]使用3个品种小麦穗图像对小穗数及籽粒数识别方法进行验证,结果表明:采用该方法 3个品种小麦穗部小穗数识别的平均零误差率为68.16%,平均绝对误差为0.46,平均相对误差为2.99%,对比已有文献识别小穗数方法,识别精度显著提高;3个品种小麦穗部籽粒数识别的平均绝对误差为2.11,平均相对误差为5.62%;3个品种单株麦穗平均测量时间为7.99 s。[结论]运用本方法可以快速高精度地对小麦穗部小穗数及籽粒数进行自动计数。
[Objectives]Wheatear spikelet number and grain number directly reflect the yield of wheat,which are two very important parameters in the research of wheat phenotype.This measuring accuracy about the number of spikelet is very low at present,which can not meet the actual needs of breeding and estimating yield.The literature on measuring grain number has not yet appeared.[Methods]In order to measure these two parameters quickly,a parabolic segmentation method based on image processing technique for wheatear spikelet was presented in this paper,and the synchronously counting spikelet number and grain number were realized by using the wheatear front view image.Firstly,image preprocessing algorithm was used in order to obtain adjusted binary image of wheatear.Then the pixels were summed according to rows of binary image,and the curve of the row pixels sum was plotted.Then the smoothing curve was obtained by using smoothing filtering algorithm and the peaks and troughs of the smoothing curve were obtained by using extracting extremum.The stalk of the wheat was separated by setting a pixels threshold depending on the width difference between the stalk and the ear head.The number of peaks at the ear head in the smoothing curve indicated the number of spikelet.Then the abscissa values of the peaks and troughs were obtained,which represented the ordinal number Y of rows in binary image.A set of intersection points between the rows ordinal number Y and the boundary of the wheatear in binary image could be obtained.Then the midpoint coordinate of the wheatear boundary was obtained at the trough.And the maximum and minimum point coordinates of the wheatear boundary were obtained at the peak.Then three position points of fitting the parabolic curve for spikelet were ascertained according to the coordinate of peaks and troughs of the smoothing curve at the binary image.Finally,each spikelet in wheatear was segmented using the fitted parabolic curve.The area of each spikelet was extracted.Meanwhile,the grain number of each spike
作者
杜世伟
李毅念
姚敏
李玲
丁启朔
何瑞银
DU Shiwei;LI Yinian;YAO Min;LI Ling;DING Qishuo;HE Ruiyin(College of Engineering,Nanjing Agricultural University,Nanjing 210031,China)
出处
《南京农业大学学报》
CAS
CSCD
北大核心
2018年第4期742-751,共10页
Journal of Nanjing Agricultural University
基金
江苏省政策引导类计划(产学研合作)项目(BY2016060-01)
国家重点研发计划项目(2016YFD0300908)
关键词
小麦穗
小穗数
籽粒数
图像处理
抛物线分割
wheatear
spikelet number
grain number
image processing
parabolic segmentation
