基于径向基支持向量机的棉花虫害识别

Translated title of the contribution: Recognition of pest damage for cotton leaf based on RBF-SVM algorithm

张建华, 冀荣华, 袁雪, 李慧, 祁力钧*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

22 Citations (Scopus)

Abstract

针对棉花受棉蚜、棉叶螨、棉盲蝽、斜纹夜蛾和烟粉虱等害虫为害后叶片表面出现不同症状,利用计算机视觉技术识别棉花虫害。通过获取受害棉花叶片图像,预处理后转换至2G-RB空间,结合Otsu算法实现色斑分割,提取色斑图像R变量、(R+G+B)/3变量的一阶矩、二阶矩和三阶矩为颜色特征,提取非色斑图像拓扑描述子和Hu不变矩为形状特征,提取2层双树复小波变换的细节图像均值和方差为纹理特征,并应用径向基支持向量机识别棉花棉蚜、棉叶螨、棉盲蝽、斜纹夜蛾、烟粉虱等虫害和正常叶片。试验结果表明,当径向基参数σ为3时,棉花虫害识别正确率达88.1%。

Based on different symptoms on pest damaged cotton leaf including cotton aphid, cotton spider mites, cotton plant bugs, cotton leafworm and whitefly, the recognition system of pest damage for cotton leaf was presented. After collecting cotton images, the mottling areas with cotton spider mites, cotton plant bugs and whitefly were segmented by Otsu method in 2G-R-B color space. The mean value, variance value and skewness value of mottling areas were extracted on the R and (R+G+B)/3 bands as color features if mottling areas appear, and topological descriptors and Hu invariant moments were extracted as shape features. Two layers dual-tree complex wavelet was used to evaluate the texture features of cotton leaf. A support vector machine (SVM) classifier with radial basis function were employed to classify cotton aphid, cotton spider mites, cotton plant bugs, cotton leafworm, whitefly and normal cotton leaf. Experiment results showed that the classification accuracy was 88.1% when σ was 3.

Translated title of the contributionRecognition of pest damage for cotton leaf based on RBF-SVM algorithm
Original languageChinese (Simplified)
Pages (from-to)178-183
Number of pages6
Journal农业机械学报
Volume42
Issue number8
Publication statusPublished - Aug 2011

Scopus Subject Areas

  • Agricultural and Biological Sciences(all)
  • Mechanical Engineering

User-Defined Keywords

  • Computer vision
  • Cotton
  • Pest
  • SVM

Fingerprint

Dive into the research topics of 'Recognition of pest damage for cotton leaf based on RBF-SVM algorithm'. Together they form a unique fingerprint.

Cite this