基于电子鼻及环境因子融合信息的花铃期棉花虫害快速诊断研究

Crop insect pests monitoring requires the development of modern detection technology. Because there are many factors involved, pest damage mechanisms is still not clear. It is difficult to accurately diagnose the cotton pests due to its characteristics of concealment, migration and sudden. Research results on herbivore-induced volatiles provide a new way of thinking for cotton pests diagnosis. According to the present researches, electronic nose(E-nose) can recognize herbivore-induced plant volatiles which contains integrated pests information. Thus, it has a good promotion prospects in pest diagnosis. In this scientific research item, an applied background research on the principles and methods of cotton pests rapid diagnosis during the flower-bolling stage using E-nose. The research content involves: analyzing effective extraction method for pest information using E-nose; exploring the relationship between sensors signal response and volatiles components binding GC-MS; investigating the influence of various uncertain factors and lucubrating the association of electronic nose and environmental factors fusion information with cotton pest information by pattern recognition methods, such as artificial neural network, probabilistic fuzzy etc., and establishing cotton insect pests diagnosis model. This project aims at revealing the E-nose diagnosis mechanism for cotton pests through techniques such as analyzing, monitoring and modeling. The research results achieved in this project could provide effective reference information for crop insect pests diagnosis.

农作物虫害监测需要发展现代检测技术。棉花虫害具有隐蔽性、迁飞性和突发性特点，并且影响因素众多，虫害作用的机理不明确，棉花虫害准确地诊断是该领域的难点问题。虫害诱导挥发物的研究为棉花虫害诊断提供了一种新的思路，已有研究表明电子鼻可以识别虫害诱导的挥发物，获得虫害综合信息，在虫害诊断方面具有很大的应用前景。本项目以受到虫害的花铃期棉花为研究对象，开展基于电子鼻的棉花田间虫害诊断原理和方法研究。具体内容包括：研究电子鼻虫害信息的有效提取方法；结合棉花挥发物GC-MS分析结果，深入探讨电子鼻传感器响应信号与挥发物成分的关联关系；综合考虑各种不确定因素，采用人工神经网络、概率模糊等模式识别方法研究电子鼻及环境因子融合信息与棉花虫害的关联关系，建立棉花虫害诊断模型。本项目旨在通过检测、监测、建模等多种技术手段，揭示电子鼻花铃期棉花虫害诊断的机理，为解决农作物田间虫害诊断难题提供理论和技术支持。

本项目以受到虫害的花铃期棉花为研究对象，针对田间棉花虫害信息获取困难，虫害诊断影响因素多，虫害预测模型精度低、适用范围窄等科学问题，研究基于电子鼻的棉花虫害信息的有效提取方法，探索电子鼻传感器响应信号、挥发物成分与棉花虫害程度的关联关系，采用合适的模式识别方法建立棉花虫害关联诊断模型，探索电子鼻棉花虫害诊断的机理。研究结果表明：健康棉花和受虫害棉花的挥发物共有约40种，包括萜烯类16种，芳香族化合物16种，烷烃、酸类、酯类和醇类7种。健康棉花和受虫害棉花组的挥发物在成分和含量上均存在差异，而受不同数量棉铃虫危害的棉花的挥发物在成分组成上差异较小，但在含量上有一定差异，受虫害棉花释放的挥发物总量普遍显著高于健康棉花释放的挥发物总量。棉铃虫危害棉花的电子鼻传感器响应曲线显示，健康棉花与受虫害棉花对于电子鼻传感器的响应不同，受不同数量棉铃虫危害的棉花的电子鼻传感器响应也存在差异，电子鼻结果与气质联用仪检测的结果一致。健康棉花释放的挥发物随着时间的变化具有明显的节律性，虫害诱导棉花的特异性挥发物释放是持续的，同时也有具有昼夜循环节律性，但是虫害棉花的规律明显与健康棉花组不同，且与健康的棉花差异显著。当长时间对棉花虫害进行电子鼻检测，电子鼻检测分析结果虽然对虫害程度的预测还不完全准确，但是建立的模型对健康棉花组的预测正确率均达到了100%。采用特征提取和优化算法可以去除棉花损伤电子鼻响应的冗余信息，提取有效的特征参数，优化排序结果与各传感器性能特点能很好的对应，结合神经网络模型提高了棉花损伤分类的准确性。说明电子鼻可以作为棉花虫害发生的有效监测手段，为电子鼻农作物虫害检测提供了理论和技术支持。