浙江农业学报 ›› 2017, Vol. 29 ›› Issue (8): 1375-1383.DOI: 10.3969/j.issn.1004-1524.2017.08.20

• 生物系统工程 • 上一篇    下一篇

基于主成分分析及LVQ神经网络的番茄种子品种识别

赵学观1, 2, 王秀1, 2, *, 李翠玲1, 2, 高原源1, 2, 3, 王松林1, 2, 冯青春1, 2   

  1. 1.北京农业智能装备技术研究中心,北京 100097;
    2.国家农业智能装备技术研究中心,北京 100097;
    3.中国农业大学 信电学院,北京 100083
  • 收稿日期:2016-12-12 出版日期:2017-08-20 发布日期:2017-09-06
  • 通讯作者: 王秀,E-mail: wangx1@nercita.org.cn
  • 作者简介:赵学观(1988-),男,山东聊城人,博士,助理研究员,主要从事农业智能装备的研究。E-mail: zhaoxg@nercita.org.cn
  • 基金资助:
    国家高技术研究发展计划(2013AA102406); 北京市农林科学院青年基金项目(QNJJ2017)

Tomato seed varieties recognition based on principal component analysis and LVQ neural network

ZHAO Xueguan1, 2, WANG Xiu1, 2, *, LI Cuiling1, 2, GAO Yuanyuan1, 2, 3, WANG Songlin1, 2, FENG Qingchun1, 2   

  1. 1. Beijing Research Center of Intelligent Equipment for Agriculture, Beijing 100097, China;
    2. National Research Center of Intelligent Equipment for Agriculture, Beijing Academy of Agriculture and Forestry Science, Beijing 100097, China;
    3.College of Electrical and Electronic Engineering, China Agricultural University, Beijing 100083, China
  • Received:2016-12-12 Online:2017-08-20 Published:2017-09-06

摘要: 提出了一种基于主成分分析优化(PCA)及竞争性神经网络(LVQ)的番茄种子品种识别方法,对番茄品种识别技术与算法进行了研究,提取了番茄种子的几何特征、纹理特征和7个不变矩特征,通过主成分分析选取了4个主成分作为人工神经网络的输入,对黑迪、红迪、佳粉十八、金迪、丘比特5个品种进行了LVQ神经网络品种识别试验。试验结果表明,竞争层节点数目为20,训练次数为96时每粒种子识别的平均耗时最短,识别准确率最高,分别为0.2 s、90.5%,基于机器视觉的番茄种子品种识别与检测方法是可行的。

关键词: 番茄种子, 品种识别, 计算机视觉, 神经网络

Abstract: In order to realize the real-time, accurate and no-damage mechanization identification of tomato seed varieties, according to the characteristics of tomato seeds and its image, the tomato varieties identification technology and algorithm were studied. This paper proposed a tomato seed varieties identification method, which is a kind of optimization by LVQ neural network based on principal components analysis, extracting the shape characteristics, texture feature and seven moment invariants of the tomato seeds. Four principal components as the input of artificial neural network were chosen through the principal components analysis. The identification test was conducted on five varieties of Heidi, Hongdi, Jiafen18, Jindi and Cupid. The number of competitive layer neurons and training trials were determined according to the test, which were 20 and 96. Under the condition, the average time of each seed identification was the shortest, and the recognition accuracy was the highest, which were 0.2 s and 90.5% respectively. The research showed that the method of identification and detection of tomato seed varieties based on machine vision is feasible.

Key words: tomato seed, variety recognition, computer vision, neural networks

中图分类号: