基于近红外光谱技术鉴别植食性金龟子的初步研究
Identification of chafers using near-infrared spectra
吕欣航1** 高传部1 田有文2 王小奇1*** 方 红1
点击：1431次 下载：22次
DOI：10.7679/j.issn.2095-1353.2015.155
作者单位：1. 沈阳农业大学植物保护学院，沈阳 110161；2. 沈阳农业大学信息电气与工程学院，沈阳 110161
中文关键词：金龟子，快速鉴别，近红外光谱，支持向量机，模型
英文关键词：chafer, rapid identification, NIRS, SVM, models
中文摘要： 【目的】 植食性金龟子是我国的重要农林害虫，探索一种快速而准确地鉴别植食性金龟子的新方法，为将此法推及至其他鞘翅目昆虫的识别来建立研究基础。【方法】 利用近红外光谱法对金龟子进行鉴别，提出了用支持向量机（Support vector machine，SVM）算法对15种植食性金龟子近红外光谱图（数据）进行分析，经过噪声波段去除后，用平滑求导与标准化法对的光谱进行预处理，选取金龟子标本150个，针对不同分类阶元和分类单元将66%样本谱图作为校正集，用SVM建立鉴别模型并对模型进行自身检验，用剩余样本图谱作为预测集对这些模型进行验证。【结果】 模型的自身检验显示在金龟科4个亚科的鉴别模型中，鳃金龟亚科正确识别率为86%，其他样本的识别准确率均大于95%，在亚科不同属和属下不同种的鉴别模型中，除疏纹星花金Protaetia cathaica (Bates)外，其他样本的识别准确率均为100%；模型的预测集验证结果显示，在不同分类阶元和分类单元的鉴别模型中，由于云斑鳃金龟Polyphylla laticollis Lewis样本较少未能正确识别，其他样本的识别准确率均为100%。整体试验结果较为理想，说明模型性能较好。【结论】 基于已定金龟子建立的模型能够很好地鉴别大部分样本，采用近红外光谱扫描技术结合支持向量机得到的植食性金龟子鉴别模型具有很强的推广能力。
英文摘要： [Objectives] To explore a new method for the rapid and accurate identification of chafers, which are important agricultural and forestry pests in China, and extend this to other beetles. [Methods] Near-infrared reflectance spectroscopy (NIRS) was used to analyze near-infrared spectrum data from 15 chafers with the support vector machine (SVM) algorithm. The spectra were processed using the methods described by Savitzky-Golay for smoothing and autoscaling, after removal of the noise region. Using the spectra of 150 specimens, 66% of the NIR spectra of chafer samples were used to establish SVM-based discrimination models, and the remaining 34% of NIRS spectra were used to validate these models. The models were self-validated. [Results] The self-validated results of calibration samples are shown below. In the identification model of 4 subfamilies of the Scarabaeidae family, the accuracy rate for identifying the Melolonthinae was 86% and that for other subfamilies 95%. Identification models for species and genera all had accuracy rates of 100%, except that for Protaetia cathaica. (Bates). The self-validated results of prediction samples showed that all accuracy rates reached 100%, except that for Polyphylla laticollis Lewis which was because of the low number of specimens available. These results are consistent with those obtained by morphological identification. [Conclusion] This novel method has a considerable potential for application in the future due to its speed, reliability and easy of use, as well as its veracity and credibility in pest control and inspection and quarantine.