摘要
刺吸电位(Electrical Penetration Graph,EPG)仪是研究蚜虫取食行为、传毒机制等的有力工具,然而EPG波形的分类识别一直是靠人工进行,迫切需要波形自动识别来提高分析效率。采用了小波变换、希尔伯特-黄变换和极限学习机等算法对蚜虫EPG信号中7种波形的特征提取和分类识别进行了研究。实验中对不同特征向量的决策树分类性能进行了对比,发现分形盒维数、Hurst指数、HHT前2层谱质心、第2~3层低频小波能量组成的6维特征向量识别效果最好,平均识别率可达91.61%。采用该特征向量进入极限学习机时可以获得更好的分类性能,平均识别率为93.57%,相比前期研究提高了2.14%。实验结果表明本文提出的基于极限学习机的EPG波形分类识别方法具有较高的识别性能,为研发EPG波形自动识别分析系统奠定了理论基础。
Electrical penetration graph(EPG)instrument is a powerful tool to study the feeding behavior and transmission mechanism of aphids.However,the identification of EPG waveform has been carried out manually.It is urgent to identify EPG waveform automatically to improve the analysis efficiency.Wavelet transform,Hilbert-Huang transform and extreme learning machine were used to study the feature extraction and classification of seven waveforms in aphid EPG signals.The performance of decision tree classification with different feature vectors was compared in the experiment,and it was found that the six-dimensional feature vector composed of fractal box dimension,Hurst index,spectral centroid of the first two layers of HHT and low-frequency wavelet energy of the second and third layers had the best recognition effect,the average identification rate was 91.61%.When the six-dimensional feature vector was used as input vector of the extreme learning machine,better classification performance could be obtained,and the average recognition rate was 93.57%,which was 2.14% higher than the previous study.The experimental results showed that the proposed classification and recognition method of EPG waveform based on extreme learning machine has high identification performance,which laid a theoretical foundation for the research and development of EPG waveform automatic identification and analysis system.
作者
吴莉莉
邢玉清
林爱英
郑宝周
潘建斌
闫凤鸣
WU Lili;XING Yuqing;LIN Aiying;ZHENG Baozhou;PAN Jianbin;YAN Fengming(College of Sciences,Henan Agricultural University,Zhengzhou 450002,China;College of Plant Protection,Henan Agricultural University,Zhengzhou 450002,China)
出处
《传感技术学报》
CAS
CSCD
北大核心
2019年第10期1535-1540,共6页
Chinese Journal of Sensors and Actuators
基金
河南省科技攻关计划项目(172102210044,182102110334)
河南省高等学校重点科研项目(17A520036,18A510012)
河南农业大学自然科学类青年创新基金项目(KJCX2018A20)
关键词
极限学习机
小波变换
刺吸电位波形
特征提取
分类
extreme learning machine
wavelet transform
electrical penetration graph waveform
feature extraction
classification
