基于水稻冠层结构参数和光谱特性信息融合的褐飞虱危害程度判定新方法研究

Rice is one of the main food crops in China. Brown planthopper is one of the main pests of rice. The control of brown planthopper is essential for ensuring the safety of rice production in China. The brown planthopper monitoring is the basis for the prevention and control of brown planthopper, but the accurate monitoring of brown planthopper has not been well resolved. The damage of brown planthopper to rice is mainly manifested in the changes of physical properties (rice canopy structure parameters) and biochemical characteristics (spectral characteristics, chemical composition, etc.). These changes can be monitored by means of scientific instruments and technology. Aiming at the monitoring problem of brown planthopper, this project proposes a method to judge the occurrence grade of brown planthopper based on the information fusion of rice canopy structure parameters and spectral characteristic information. LiDAR and multi-spectral cameras would be used to monitor the canopy structure parameters and spectral characteristics of rice respectively. And correlation analysis and pattern recognition based on BP neural network would be used to study the relationship between rice canopy structure parameters and brown planthopper occurrence grade, and rice spectral characteristics and brown planthopper occurrence grade. The analytical models between rice canopy structure parameters, and rice spectral characteristics and brown planthopper occurrence grade would be developed. Based on the information fusion of artificial neural network, a new method for judging the occurrence grade of brown planthopper based on canopy structure and spectral characteristics would be established to improve the monitoring accuracy of the brown planthopper occurrence grade. The theoretical basis and scientific guidance can be provided for the precise application of brown planthopper base on the research of this project.

水稻是我国主要粮食作物之一。褐飞虱是水稻的主要虫害之一，褐飞虱防治对保证我国水稻生产安全至关重要。褐飞虱监测是褐飞虱防治的基础，但褐飞虱的准确监测至今仍未得到很好的解决。褐飞虱对水稻的危害主要表现在物理性状（水稻冠层结构参数）和生化特性（光谱特性、化学成份等）特征的变化，这些变化信息可以借助于科学仪器和技术手段监测获得。针对褐飞虱监测难题，本项目提出基于水稻冠层结构参数和光谱特征信息融合的褐飞虱危害程度判断方法的研究。采用LiDAR和多光谱相机分别监测水稻冠层结构参数和光谱特性，基于BP神经网络进行关联分析及模式识别等，研究水稻冠层结构参数和褐飞虱危害程度之间、水稻光谱特性和褐飞虱危害程度之间的关系，寻求褐飞虱危害程度解析模型。基于人工神经网络信息融合，形成基于冠层结构和光谱特性的褐飞虱危害程度判定的新方法，提高褐飞虱危害程度监测的准确性，为褐飞虱防治精准施药提供理论依据和科学指导。

水稻是我国主要粮食作物之一，水稻生产受病虫害影响严重。褐飞虱是危害水稻的主要虫害之一，褐飞虱精准防控对保障我国水稻生产安全至关重要。褐飞虱监测是褐飞虱防治的基础，但褐飞虱的准确监测至今仍未得到很好的解决。褐飞虱对水稻的危害主要表现在水稻物理性状（水稻冠层结构参数）和生化特性（光谱特性、化学成份等）的变化，这些变化信息可以借助科学仪器和技术手段监测获得。针对褐飞虱监测难题，本项目提出基于水稻冠层结构参数和光谱特征信息融合的褐飞虱危害程度判定方法的研究。采用CI-110植物冠层数字图象分析仪监测水稻冠层结构参数，并采用ASD FieldSpec 3高光谱仪监测水稻光谱特性，基于BP神经网络进行关联分析及模式识别等，研究了水稻冠层结构参数和褐飞虱危害程度之间、水稻光谱特性和褐飞虱危害程度之间的相关性，建立了水稻褐飞虱危害等级与水稻冠层结构之间，以及水稻褐飞虱危害等级与水稻光谱特性之间的相关性解析模型。基于深度学习技术，对水稻冠层结构参数和光谱特征信息进行融合，并建立了融合信息与水稻褐飞虱危害等级之间的解析模型。研究结果表明：（1）采用水稻冠层结构参数建立的褐飞虱危害等级解析模型对褐飞虱危害等级的判定精度为81.4%；（2）采用水稻光谱特征信息建立的褐飞虱危害等级解析模型对褐飞虱的判定精度为86.6%；（3）采用水稻冠层结构参数和光谱特征信息融合的褐飞虱危害等级解析模型对褐飞虱危害等级的判定精度为95.8%。由此可见，基于水稻冠层结构参数和光谱特征信息融合的水稻褐飞虱危害等级判定模型显著提高了水稻褐飞虱危害等级判定精度。该项目形成了基于水稻冠层结构和光谱特性的褐飞虱危害等级判定的新方法，提高了褐飞虱危害程度监测的准确性，为褐飞虱防治精准施药提供了理论依据和科学指导。