›› 2015, Vol. 28 ›› Issue (3): 33-.

• 论文 • 上一篇    下一篇

基于ELM理论的昆虫分类

徐源浩,齐焕芳   

  1. (西安电子科技大学 数学与统计学院,陕西 西安 710126)
  • 出版日期:2015-03-15 发布日期:2015-03-12
  • 作者简介:徐源浩(1989—),男,硕士研究生。研究方向:图像处理。E-mail:1032395666@qq.com。齐焕芳(1988—),女,硕士研究生。研究方向:压缩感知重建方法。

Insect Classification Based on the Theory of the ELM

XU Yuanhao,QI Huanfang   

  1. (School of Mathematics and Statistics,Xidian University,Xi'an 710126,China)
  • Online:2015-03-15 Published:2015-03-12

摘要:

机器视觉技术应用在昆虫分类领域,取代传统人眼观察识别过程、提高了工作效率。自动识别技术包含昆虫特征提取和分类器设计两个主要步骤。根据整个识别过程,文中提出了一种基于混合特征的ELM理论昆虫识别方法。在特征提取阶段,提取混合特征包括颜色特征、形态特征、空域纹理特征和频谱纹理特征。在分类器设计阶段采用具有学习速度快且泛化性能好的极限学习机。实验结果表明,该方法使昆虫识别的正确率达到97%,且分类器训练时间短,优于传统的自动识别方法。

关键词: 特征提取, 颜色特征, 形态特征, 空域纹理特征, 频谱纹理特征, 极限学习机器

Abstract:

Applied in the field of insect taxonomy,machine vision technology displaces the human eye observation identification process and improves working efficiency.The automatic identification technology has two main steps:insects feature extraction and classifier design.Based on the entire identification process,this paper puts forward an insect identification method based on hybrid features of ELM theory.In the feature extraction stage,the hybrid features will be extracted including the color,morphological characteristics,spatial texture,and spectral texture;while in the classifier design stage,high learning speed and good generalization performance ELM extreme learning machine are taken.The experimental results show that this method offers an insect identification accuracy up to 97% with a short classifier training time,superior to the traditional automatic identification method.

Key words: feature extraction;color features;morphological characteristics;spatial texture features;spectral texture feature;extreme learning machine

中图分类号: 

  • TP393