光学学报, 2019, 39 (5): 0515003, 网络出版: 2019-05-10
基于深度信息的大豆株高计算方法 下载: 1524次
Calculation Method of Soybean Plant Height Based on Depth Information
机器视觉 大豆冠层 深度信息 Kinect 2.0 三维重建 表型参数 株高 machine vision soybean canopy depth information Kinect 2.0 three-dimensional reconstruction phenotypic parameters plant height
摘要
为高通量地计算农作物株高,克服传统测量方法低效、耗时耗力等不足, 以抗线9号、13号和富豆6号寒地大豆为研究对象,构建了基于Kinect 2.0的大豆冠层图像同步采集平台,并在三维重建大豆冠层结构形态的基础上,提出了基于深度信息的个体和群体大豆株高计算方法。实验结果表明,与实测值相比,计算得到的个体和群体大豆株高的平均误差分别为0.14 cm和0.54 cm,抗线9号、13号和富豆6号株高计算值与实测值之间的决定系数依次为0.9717,0.9730,0.9697。所提方法能够较为精确地计算大豆植株的株高特征。
Abstract
In this study, three cold soybean varieties, namely Kangxian-9, Kangxian-13, and Fudou-6, were used for the high-throughput calculation of soybean plant heights. Three varieties were used to overcome the disadvantages of the traditional measurement methods, which were inefficient and laborious. The method for calculating the plant heights of individual and grouped soybean plants was proposed using the depth information acquired from the three-dimensional reconstruction of soybean canopies obtained using the Kinect V2.0 synchronous image acquisition platform. The experimental results show that compared with the measured value, the average errors of the proposed calculation method for the plant heights of individual and grouped soybean plants are 0.14 cm and 0.54 cm, respectively. The determination coefficients between calculated and measured values for Kangxian-9, Kangxian-13, and Fudou-6 are 0.9717, 0.9730, and 0.9697, respectively. Thus, the proposed method can accurately calculate the heights of soybean plants.
冯佳睿, 马晓丹, 关海鸥, 朱可心, 于菘. 基于深度信息的大豆株高计算方法[J]. 光学学报, 2019, 39(5): 0515003. Jiarui Feng, Xiaodan Ma, Haiou Guan, Kexin Zhu, Song Yu. Calculation Method of Soybean Plant Height Based on Depth Information[J]. Acta Optica Sinica, 2019, 39(5): 0515003.