玉米单粒种子发芽潜力无损检测方法研究  

Study on Nondestructive Testing Method for Germination Potential of Maize Single Seed

作　　者：张晗[1] 董宏图 栗彬彬 闫宁 吴旭东 罗斌[1] ZHANG Han;DONG Hongtu;LI Binbin;YAN Ning;WU Xudong;LUO Bin(Beijing Agricultural Information Technology Research Center,Beijing 100094,China;NONGXIN(Nanjing)Smart Agriculture Research Institute,Nanjing 211800,China;Beijing Research Center for Intelligent Agricultural Equipment,Beijing 100094,China;School of Mechanical Engineering,North University of China,Taiyuan Shanxi 030051,China)

机构地区：[1]北京农业信息技术研究中心,北京100094 [2]农芯(南京)智慧农业研究院,南京211800 [3]北京农业智能装备技术研究中心,北京100094 [4]中北大学机械工程学院,山西太原030051

出　　处：《种子》2021年第11期144-148,共5页Seed

基　　金：国家重点研发计划项目(2017 YFD 0701205);北京市农林科学院改革与发展项目-作物信息无损检测关键技术与设备研发。

摘　　要：以郑单958为研究对象,研究玉米单粒种子发芽潜力无损检测方法。使用相机采集单粒种子图像,提取种子24个颜色特征和13个形状特征。将图像特征与种子发芽实验结果组成数据集,通过使用主成分分析(Principal Component Analysis, PCA)提取8个可以代表所有特征的主成分,结合极限学习机算法(Extreme Learning Machine, ELM)建立模型。通过Matlab软件进行仿真,并分析了选择不同隐含层神经元个数和不同隐含层激励函数对模型预测精度的影响。结果表明,使用极限学习机模型ELM选择sigmoid函数建立PCA-ELMs模型对种子发芽预测的准确率为63.33%,查准率为88.51%,查全率为63.11%。研究表明,通过机器视觉技术结合PCA-ELMs算法建立模型,对种子发芽潜力预测具有一定可行性,通过模型分类后可以使剩余样本发芽率获得提升。To detect the germination potential of maize seeds, a non-destructive detection method for single seed germination potential of maize was studied with Zhengdan 958 as the research material. Single seed images were collected by camera, and 24 color features and 13 shape features were extracted. Image features and seed germination experiment results were incorporated into a data set, and 8 principal components representing all features were extracted by principal component analysis(PCA). Extreme learning machine(ELM) algorithm was used to build the model. Simulation was carried out by Matlab software, and the influence of different number of hidden layer neurons and different excitation functions on the prediction accuracy of the model was analyzed. The results showed that the accuracy of PCA-ELMS model based on ELM and sigmoid function was 63.33%, the accuracy was 88.51%, and the recall rate was 63.11%. The results showed that the model established by machine vision technology combined with PCA-ELMS algorithm was feasible to predict the germination potential of seeds, and the germination rate of remaining samples could be improved after model classification.

关 键 词：机器视觉 玉米种子 种子发芽潜力 极限学习机 

分 类 号：S513[农业科学—作物学]