基于机器视觉的田间小麦开花期判定方法  

Determination Method of Field Wheat Flowering Period Baesd on Machine Vision

作　　者：刘平[1] 刘立鹏 王春颖 朱衍俊 王宏伟 李祥[2] LIU Ping;LIU Lipeng;WANG Chunying;ZHU Yanjun;WANG Hongwei;LI Xiang(College of Mechanical and Electronic Engineering,Shandong Agricultural University,Taian 271018,China;College of Life Sciences,Shandong Agricultural University,Taian 271018,China)

机构地区：[1]山东农业大学机械与电子工程学院,泰安271018 [2]山东农业大学生命科学学院,泰安271018

出　　处：《农业机械学报》2022年第3期251-258,共8页Transactions of the Chinese Society for Agricultural Machinery

基　　金：山东省自然科学基金项目(ZR2020KF002);国家自然科学基金项目(31871543、31700644);山东省农机装备研发创新计划项目(2018YF004)。

摘　　要：针对大量小麦育种材料花期难以精准、快速检测的问题,提出了一种基于综合颜色特征和超像素分割算法的小麦开花期判定方法。首先,根据光照强度及图像清晰度对综合颜色特征的过红颜色分量、HSV颜色空间的S分量和红绿归一化颜色分量自适应调节,增强小花和小穗的差异性。其次,基于中心距离函数和灰度变化函数改进超像素分割算法的聚类规则,获得由同质特征的相邻像素组成的图像区域。随后,优化图像区域路径搜索算法实现各图像区域精确分割,通过灰度和对比度指标完成各图像区域分类,实现小花与小穗的精准、快速分割,并根据小花与小穗的比例完成开花期判定。实验结果表明,本文所提出算法平均计算时间为0.172 s,小花平均识别精度为91%,小穗平均识别精度为90.9%,预测开花率与实际开花率的平均差值仅为1.16%,满足田间小麦开花期判定基本要求。The timing of flowering is one of the important indexes of wheat breeding,but it is difficult to detect the flowering stage from a large number of wheat breeding materials accurately and quickly.A method to determine the flowering date of wheat based on comprehensive color features and super-pixel segmentation algorithm was proposed.Firstly,according to the light intensity and image clarity,the excess red color component of comprehensive color features,the saturation component of HSV color space and the normalized red green color component were adaptively adjusted to enhance the difference between florets and spikelets.Secondly,the clustering rules of the super-pixel segmentation algorithm were improved based on the center distance function and the gray change function to obtain the image region composed of adjacent pixels with homogeneous features.Then the image area path search algorithm was optimized to achieve accurate segmentation of each image area,and the classification of each image area was completed through grayscale and contrast indicators to achieve accurate and rapid segmentation of florets and spikelets,and the flowering period was determined according to the proportion of floret and spikelet.The experimental results showed that the average computing time of the proposed algorithm was 0.172 s,the average recognition accuracy of floret was 91%,the average recognition accuracy of spikelet was 90.9%,the average error between the predicted flowering rate and the actual was only 1.16%,which met the basic requirements of determining the flowering date of wheat in the field.

关 键 词：田间环境 小麦 花期判定 图像识别 综合颜色特征 超像素分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]