多源信息融合的褐飞虱发生早期混沌特性及智能监测机理研究

Brown Plant Hopper (BPH) (Nilaparvata lugens) is one of the most serious disasters in rice production in China. To implement timely targeted pesticide applications, reduce input costs and benefit the environment, an accurate early detection and quantification of damage caused by BPHs in rice plants is required. Early detection and accurate monitoring of its development trend is regarded as the hottest topic by the experts in this field, but it has not yet been resolved satisfactorily. It was founded in the preliminary study there are several problems, such as instability and limited scope, by using a single or fewer parameter monitoring methods. A new method based on multi-source information fusion and its chaotic behaviors in the process of BPHs spreading is proposed. The multi-source information includes the BPH history amount of data, environment meteorological data, bionic olfaction data, visible image and the multi-spectral image of the rice plant. In order to construct a fault tolerance intelligence monitoring model, relevant experiments and research will be proposed in this project, include the construction of the integrated platform about multi-source information acquisition, the research of the development law and its chaotic characters based on heterogeneous information in the time series during the BPHs early occurrence, and then combine with the intelligent theories, such as fuzzy, artificial neural network, wavelet etc., construct a fault tolerance monitoring model for the early detection of the stress in rice caused by BPH infestation. The proposed project will form a new method for the early detection of the BPH stress, and provide theoretical and practical basis to improve the timeliness and the monitor accuracy for the BPH stress early control.

褐飞虱灾害是我国水稻生产中最严重的生物灾害之一，在其发生早期实时、快速、准确地掌握虫情，是有效防治褐飞虱的前提和关键。在早期发现并准确监测其发展态势一直是国内外专家致力研究的问题，至今尚未圆满解决。项目组前期研究发现采用单一或较少参数监测时存在稳定性不高、适用范围有限等问题，据此提出采用基于多源信息融合的方法对褐飞虱发生早期虫害蔓延过程的混沌行为进行研究，结合智能化理论和技术构造具有较强分类与容错能力的褐飞虱虫害发生早期智能监测模型。项目拟构建褐飞虱虫害发生早期相关信息的快速获取平台，重点研究褐飞虱发生早期多种异质信息在时间序列上的发生、发展规律，揭示褐飞虱发生早期多种异质信息在时间序列上的混沌特性，寻找出适合于描述褐飞虱虫害早期蔓延混沌行为的解析模型，形成褐飞虱虫害发生早期智能监测的新方法，为提高褐飞虱监测的时效性及其早期防治提供理论和实践上的依据。

褐飞虱灾害是我国水稻生产中最严重的生物灾害之一，在其发生早期实时、快速、准确地掌握虫情，是有效防治褐飞虱的前提和关键。针对已有研究中采用单一或较少参数监测时存在稳定性不高、适用范围有限等问题，本课题采用基于多源信息融合的方法对褐飞虱发生早期虫害蔓延过程进行研究，结合智能化理论和技术构造具有较强分类与容错能力的褐飞虱虫害发生早期智能监测模型，实现了以下研究目标：构建了褐飞虱虫害发生早期相关信息的快速获取平台，可同时进行水稻冠层多光谱、水稻中下部可见光图像、水稻冠层气体挥发物电子鼻气味等信息的获取，重点研究了褐飞虱发生早期多种异质信息在时间序列上的发生、发展规律，揭示了褐飞虱发生早期多种异质信息在时间序列上的变化规律，给出适合于进行褐飞虱虫害早期发生的虫害检测和分级识别的解析模型，提出了褐飞虱虫害发生早期智能监测的新方法，为提高褐飞虱监测的时效性及其早期防治提供了理论和实践上的依据。课题执行期间，共撰写相关学术论文9篇（其中SCI/EI收录5篇）；申请发明专利3件，实用新型专利1件；申请计算机软件著作权登记1件；正在整理撰写著作《农作物虫害的机器监测技术》1部；培养博士后1名，研究生8人（其中，博士研究生2人，硕士研究生6人）。