基于侧柏挥发物的电子鼻检测早期钻蛀性害虫的机理与解析

Early symptoms of borer infestation are difficult to be found through the appearance of infected trees, but the mount of volatile organic compounds (VOCs) will start to shift, which can be detected by electronic nose. To reveal the mechanism of borer infestation, the relationship between the E-nose fingerprints of VOCs and pest infestation should be studied. . Platycladus orientalis normally releases a range of VOCs which vary qualitatively and quantitatively according to the levels of Semanotus bifasciatus attacking it. When attacked by pest, tree emit much greater amounts of VOCs than those of unaffected. The object of this project is to investigate the relationship between the major changes of VOCs and the actual seriousness of pest affection, to explore the principles between response characteristics of gas sensors and the levels of pest affection. We focus on sensor sensitivity and optimization of sensors array, and finding the discipline of sensor array signals, and distinguishing methods used to abbreviate redundancy data based on analyzing the variety of volatile organic compounds caused by different insect pests. After pattern recognition methods as well as identification methods of qualitative and quantitative were found. And a new electronic system with sensor array combination mode, signal processing method and pattern recognition method specialized for insect pests detection are presented in this project.. These original approaches and technologies presented in this project will provide theoretical and practical foundation for artificial intelligence, and will provide theory guidance for developing and using of E-nose as a pest detection system for forestry monitoring.

树木钻蛀性害虫早期为害时，外表很难发现、但挥发物将发生变化，这种情况适合采用电子鼻技术进行检测。为探明检测树木虫害的这类方法，首先需要探明“树木挥发物的电子鼻信息与其虫害间的关联性”的这种检测机理。.侧柏受双条杉天牛为害后，其挥发物成分组成明显不同于未受钻蛀性虫害侧柏的。项目针对侧柏钻蛀性虫害发生规律，探明重要挥发物成分与侧柏有无虫害、不同虫害程度之间的关联性；揭示不同虫害侧柏挥发物对气敏传感器响应特性的影响规律，探讨剔除传感器中‘高噪杂’冗余信号的途径；构建虫害侧柏嗅觉信息指纹图库，建立预测模型，提出定性定量检测方法；从而为侧柏钻蛀性害虫的早期检测提出方法。.通过研究，为基于挥发物的森林虫害电子鼻检测方法提出科学论据，并为在林业上开发相关检测系统提供重要借鉴。

侧柏（Platycladus orientalis L.）是我国应用最为广泛的植物物种之一，具有重要的生态价值、经济价值、文化价值以及药用价值。然而，侧柏极易受双条杉天牛和柏肤小蠹等蛀干害虫侵害，进而树势衰弱甚至死亡，造成巨大损失。由于植物受虫害后，其植物挥发物种类和含量变化显著。因此，可通过检测植物挥发物来对侧柏的受害情况进行判断。本研究基于电子鼻以及气质联用（GC-MS）技术对受不同虫害数量、不同感染时长和不同虫害类型危害的侧柏挥发物进行了检测，建立了对受害侧柏的分类和预测模型。为基于电子鼻技术的侧柏蛀干害虫危害检测提供了理论基础和指导方案。主要研究内容和结论如下： .（1）利用GC-MS对受不同数量双条杉天牛危害的侧柏释放的挥发物进行了检测，结果表明不同数量害虫感染的侧柏挥发物之间存在明显区别，与对照组相比，感染组的α-蒎烯、α-葑烯、β-水芹烯、3-蒈烯、对伞花烃、D-柠檬烯和长叶烯等成分的释放量明显改变。.（2）从电子鼻信号中提取了特征值，使用反向传播神经网络、概率神经网络和支持向量机建立分类模型。结果表明，支持向量机在训练集和测试集均取得了较佳的分类效果。利用偏最小二乘法（PLSR）对侧柏感染害虫数量进行预测，结果表明具有良好的回归性能。.（3）GC-MS分析出17种随双条杉天牛危害时长变化引起明显侧柏挥发物成分，其中α-蒎烯、松油醇、D-柠檬烯和长叶烯等成分与电子鼻响应信号相关性较强。利用支持向量回归模型基于小波熵特征值对双条杉天牛危害时长进行预测，较高的R2值和较低的RMSE值。.（4）根据α-蒎烯、β-水芹烯、3-蒈烯和罗汉柏烯等4种特征成分，分别制备了基于分子印迹聚合物的4种QCM传感器阵列、并研制成电子鼻。利用此电子鼻对感染不同类型虫害的侧柏样本进行检测。结果表明结合SVM的分类算法效果更优，训练集和测试集的分类正确率分别为97.62%和93.75%。