基于机器视觉的玉米三维表型参数提取算法

Extracting algorithm of maize three-dimensional phenotypic parameters algorithm base on machine vision

针对目前玉米植株表型参数提取存在精度低、测量时间长和需要人工辅助等问题,提出了基于机器视觉的三维表型参数提取方法。采用瞬时成像设备,获取玉米植株64个角度无背景的图像;基于机器视觉获取三维点云,采用基于拉普拉斯的骨架提取算法提取玉米骨架,对叶片骨架和茎秆骨架优化后,提取叶长、叶最大宽度、叶基部高度、叶夹角、株高和植株最小包围盒体积。比较两个玉米自交系参数提取结果,取误差上限,骨架优化前后叶长、叶最大宽度和叶基部高度平均绝对百分比误差分别下降5.89%、0.04%和0.86%;与人工测量值相比,测得的叶长、叶最大宽度和叶基部高度的平均绝对百分比误差分别为6.66%、6.45%和5.43%。本研究提供了一种高精度、高通量、自动化的玉米生长动态定量化测量方法。

Aiming at the problems of low precision,long measurement time,artificial assistance in extracting maize plant phenotypic parameters,a three-dimensional phenotypic parameter extraction method based on machine vision was proposed.Instantaneous imaging equipment was used to obtain images of maize plant at 64 angles without background.Based on machine vision,the three-dimensional point cloud was obtained.The skeleton extraction algorithm based on Laplace was used to extract maize skeleton.After optimizing the leaf skeleton and stem skeleton,the leaf length,maximum leaf width,leaf base height,leaf angle,plant height and plant minimum bounding box volume were extracted.Comparing the extraction results of two maize inbred lines,taking the upper limit of error,the root mean square errors of leaf length,maximum leaf width and leaf base height decreased by 5.89%,0.04% and 0.86%,respectively before and after skeleton optimization.Compared with the manually measured values,the mean absolute percentage errors of leaf length,maximum leaf width and leaf base height were 6.66%,6.45% and 5.43%,respectively.This study provided a high precision,highthroughput,automated quantitative measurement method of maize growth dynamics.