目录
摘要 II
关键词 II
ABSTRACT II
KEY WORDS: III
引言
1 材料与方法 1
1.1 试验设计 1
1.2 研究区概况 1
1.3 数据获取 2
1.3.1 稻瘟病发病状况田间调查 2
1.3.2 Planet多光谱影像获取 4
1.4 数据处理 5
1.4.1 光谱数据获取 5
1.4.2 植被指数筛选 5
1.5 发病区与对照区差异性分析及制图 6
1.6 病情指数监测模型构建及评价 6
2 结果与分析 8
2.1 发病区与对照区的分类识别 8
2.1.1 发病区和对照区各生育时期植被指数差异性分析 8
2.1.2 发病区与对照区的决策树分类结果 8
2.1.3 研究区内的发病识别空间分布图 10
2.2 发病区病情指数监测 11
2.2.1 病情指数与植被指数的估算模型建模及验证精度 11
2.2.2 发病区水稻病情指数空间分布 13
3 讨论 15
3.1 传感器分辨率对发病状况的监测能力的影响 15
3.2 不同植被指数对发病状况的监测能力的差异 15
3.3 不同病情指数预测模型预测能力的差异 16
3.4 复杂的气象及大田环境对病情监测效果的影响 16
4 总结与展望 17
致谢 18
参考文献 19
基于Planet卫星影像的稻瘟病发病区识别及病情指数监测研究
摘要
稻瘟病（Magnaporthe oryzae）是一种危害严重的真菌性病害，在流行年份会导致作物显著减产。应用卫星遥感技术可以在区域尺度探究作物冠层光谱信息，有助于及时定位发病源并进行精确防控。本研究基于多时相Planet卫星影像和发病区田间调查数据，利用植被指数进行决策树分类和病情指数线性回归，探索了卫星影像在稻瘟病监测中的应用潜力。从发病区与对照区决策树分类 *51今日免费论文网|www.51jrft.com +Q: &351916072& 
结果来看，本研究选取的植被指数于各时期均具有识别能力，识别能力最佳的植被指数是GNDVI（Green Normalized Difference Vegetation Index），采样点分类结果总体精度为100%。在病情指数监测上，使用植被指数建立一元和多元线性回归模型，TVI（Triangular Vegetation Index）模型是表现最佳的一元线性回归模型，R2为0.32。使用多元线性回归有助于进一步提高病情指数监测能力，全株病情指数估算模型R2为0.41，叶瘟病情指数估算模型R2为0.49。本研究表明高时空分辨率Planet影像可用于发病区识别及病情指数监测，对稻瘟病区域尺度快速同步监测具有指导意义。
IDENTIFICATION OF RICE BLAST FIELDS AND MONITORING OF DISEASE INDEX BASED ON PLANET SATELLITE IMAGES
ABSTRACT
Rice blast (Magnaporthe oryzae) is a severe fungal disease that can cause a significant reduction in crop yields in popular years. It can cause significant crop yield reduction in epidemic years. The application of satellite remote sensing can explore crop canopy spectral information covering a regional scale, which can help locate the disease center in time and carry out accurate chemical control. which can provide timely and objective information services for agricultural production. Based on multitemporal Planet satellite images and field disease survey data in the experimental area, we applied the decision tree classifier and linear regression model with vegetation indexes to explore the application potential of satellite images in rice blast monitoring. The tree classification results shows that the all vegetation indexes selected in this study have identification ability at jointing stage. The vegetation index with the best performance is GNDVI (Green Normalized Difference Vegetation Index), and the overall accuracy of the classification results of sampling points is 100%.The present study uses vegetation index to establish ULR (Univariate Linear Regression) and MLR (Multiple Linear Regression) models to estimate the disease index. The result shows that TVI (Triangular Vegetation Index) model is the best linear regression model with R2 of 0.32. Besides, the use of multiple linear regression is helpful to further improve the monitoring ability of disease index, R2 of the plant blast index estimation model is 0.41, and R2 of the leaf blast index estimation model is 0.49. This study shows that Planet images with high spatialtemporal resolution can be used to identify disease areas and estimate disease index, which provides significant guiding for rapid and synchronous monitoring of rice blast at regional scale.

原文链接：http://www.jxszl.com/nongxue/zwbh/606135.html