综合图谱信息的茶树炭疽病遥感监测机理与模型

As an important foliar disease in tea plant, tea plant anthracnose endangers both quality and yield of tea seriously. Currently, given the lack of fast and effective ways in monitoring of tea plant anthracnose, it might cause the failure prevention of the disease or pesticide contamination. To solve this problem, this project introduces both spectral and imaging information in developing the monitoring method for tea plant anthracnose. The main contents of this study includes:(1) to investigate the symptoms and physiological changes of tea plant anthracnose and their temporal patterns through a controlled experiment, and study the corresponding spectral response and image characteristics that were caused by the disease; (2) to study the methods of feature selection and construction, and form the optimal feature set for disease monitoring by combining spectral, image and temporal analysis, as well as the sensitivity analysis of features; (3) to develop a robust model for tea plant anthracnose monitoring based on the integration of both the spectral and image features from hyperspectral imaging data. The model should be able to not only recognize the disease under complicated circumstances with other stressors existed, but also provide the estimation of disease severity. The outcomes of the study would facilitate the precision management and prevention of tea plant anthracnose, which would also benefit the sustainable production of tea. Meanwhile, the corresponding methods and models may provide references for further studies on remotely sensed monitoring of diseases on commercial crops.

茶树炭疽病是我国茶树中普遍发生的一种重要叶部病害，对茶叶的品质和产量构成严重威胁。目前茶树植保工作中缺乏对该病快速、有效的监测方法，容易造成漏防或农药污染等问题。针对这一问题，本项目将光谱和图像分析引入茶树炭疽病的监测中，拟开展下述研究：（1）通过开展系统的小区控制实验，考察茶树炭疽病的症状和生理改变及其在不同时相的发展，通过分析光谱和图像数据掌握该病的光谱响应和图像特征变化规律；（2）结合光谱、图像、时序分析及特征敏感性分析，研究针对该病监测的专属性特征选择和构建方法，形成茶树炭疽病识别监测的优选特征集；（3）基于病害优选特征，结合成像光谱数据研究光谱和图像特征的融合策略，构建能够在易混胁迫存在的条件下高专一性地识别茶树炭疽病，并监测其程度的模型。项目研究成果将有助于茶树炭疽病的精准防控，为茶叶绿色可持续生产提供重要支撑。同时，该研究模式将有利于为其它经济作物病害遥感监测研究提供参考。

茶树炭疽病是我国茶树中普遍发生的一种重要叶部病害，对茶叶的品质和产量构成严重威胁。.目前茶树炭疽病的监测和防控主要依靠传统人工调查采样和实验室分析进行诊断和决策，不仅效率较低，而且容易发生漏检误判的情况。本项目针对茶树炭疽病的无损探测问题，将光谱和图像信息引入病害检测中，研究了病害高光谱探测和OCT探测方法，构建了茶树炭疽病专属性光谱特征，并提出了叶片和冠层尺度茶树炭疽病识别和区分模型，并在此基础上基于彩色数字图像对研究成果进行推广应用，取得下述研究成果：（1）2017-2019年开展了茶树炭疽病成像高光谱测试实验、病害侵染过程连续时相动态监测实验、茶树不同病虫害区分实验及炭疽病病叶OCT观测实验，为开展茶树炭疽病探测特征与模型研究提供数据基础；（2）对茶树炭疽病发生演变过程中成像光谱探测和OCT探测的响应变化和特征进行分析，详细对比了叶片正常、潜伏、病斑、菌丝部位的光谱响应变化，在此基础上结合SPA分析进行波段优选并构建COBA指数；基于OCT影像开展了边缘检测、深度信息对比、LBP特征提取和图像间隙值计算等分析，明确了茶树炭疽病OCT观测响应及特征；（3）基于自然发病条件下茶树炭疽病叶片成像高光谱数据，构建了TARI和TANI指数，在此基础上，提出了结合IsoData和二维光谱特征空间分析的茶树炭疽病病斑分级识别方法，经检验能够准确专一性识别茶树炭疽病病斑（OA=94%）；（4）结合成像光谱数据和连续小波分析提出一种能够区分和识别炭疽病、茶小绿叶蝉和灼伤等几种茶园常发易混胁迫的方法，确定了5个能够有效进行胁迫区分的小波特征，结合KNN，RF，SVM算法，在叶片和冠层尺度上建立了病害区分识别模型，区分精度超过88%。在上述研究基础上，从应用角度结合数码相片和自适应模糊分析提出了茶树炭疽病识别方法，为病害快速检测提供一种低成本方法。