基于混合光谱理论的农田NDVI土壤背景影响分析与去除方法

发布时间：2018-06-08 17:51

本文选题：NDVI + LAI　；参考：《南京大学》2017年硕士论文

【摘要】：归一化植被指数(NDVI)是遥感领域中应用最为广泛的植被指数之一,用于作物长势监测、农业估产、旱情预测、精准农业等方面的应用。NDVI能够在一定程度上消除大气、阴影、传感器定标、观测角度等方面的影响,但是土壤背景的混入干扰使NDVI产生较大误差。生长于不同土壤类型背景条件下的相同长势冬小麦农田NDVI有很大差异,也一直影响着利用NDVI进行小麦长势有效监测和精确评价。现有NDVI 土壤影响的去除方法,多为发展和改进现有的NDVI模型,但这些新植被指数多依赖研究区土壤线特征,无法像NDVI形成产品进而广泛应用于大尺度、宽覆盖地区的作物长势监测。因此研究NDVI的土壤背景影响去除方法依然是当今的热点和难点。论文在以典型土壤类型为农田背景对不同植被覆盖度冬小麦NDVI的影响模拟分析的基础上,基于混合光谱理论,提出基于混合光谱理论的两种土壤背景影响去除模型(NDVIT)。以安徽省滁州地区的冬小麦农田为研究区,以水稻土和黄褐土等农田土壤背景的拔节前期冬小麦为研究对象,采用实测冬小麦冠层光谱及叶面积指数(LAI)数据,利用传统的相片估算法求算植被覆盖度,研究分析两种模型的土壤背景影响去除能力。同时以山东省济宁地区冬小麦农田为研究区,以褐土、潮土、水稻土和砂姜黑土等农田土壤背景的拔节后期冬小麦为研究对象,结合Landsat-8 OLI卫星多光谱遥感数据,验证与评价土壤背景影响去除模型适用性和有效性。主要研究内容与结论如下:(1)结合典型土壤类型和植被覆盖度,研究土壤背景对冬小麦农田NDVI信息的影响,结果表明,不同类型土壤背景对冬小麦农田NDVI造成很大差异,且造成冬小麦农田NDVI对植被覆盖度的敏感性也存在明显差异,为不同类型土壤背景的各小麦生长期遥感NDVI信息估算频次选择提供依据。(2)基于线性混合光谱理论,构建基于NDVI的土壤背景影响去除模型NDVI1T,以及对基于像元二分理论对NDVI1T进行简化得到简约模型NDVI2T。验证了模型的适用性和有效性,同时采用信噪比的分析方法定量研究两种模型抵抗土壤噪声影响的能力,分析发现NDVI1T提取植被信息抵抗土壤噪声能力更佳。在中低LAI(LAI=3)环境条件下,两种模型更适用于植被叶片覆盖程度较为均匀,或植被类型单一的情况;两种土壤背景影响去除模型和NDVI的拟合关系良好,相关关系R2均达到0.9以上,表明了利用NDVIT模型可实现修正地面实验计算的NDVI 土壤背景影响。(3)基于Landsat-8 OLI卫星遥感影像对土壤背景影响去除模型(NDVIT)进行分析验证,研究结果表明,在对应OLI影像的研究区具有4种土壤背景类型的情形下,通过土壤背景影响去除模型(NDVIT)和基于影像计算NDVI拟合公式,依然可实现修正大尺度NDVI产品的土壤背景影响。
[Abstract]:The normalized vegetation index (NDVI) is one of the most widely used vegetation indices in remote sensing. It can be used in crop growth monitoring, agricultural yield estimation, drought forecasting, precision agriculture, and so on. NDVI can eliminate the atmosphere and shadow to a certain extent. Sensor calibration, observation angle and other aspects of the impact, but the mixing of soil background interference caused a large error in NDVI. The NDVI of the same growing winter wheat field under different soil type background conditions is very different and has been affecting the effective monitoring and accurate evaluation of wheat growth using NDVI. Most of the existing NDVI soil impact removal methods are to develop and improve the existing NDVI models. However, these new vegetation indices depend on the characteristics of soil lines in the study area, so they can not be used in large scale as NDVI products. Crop growth monitoring in wide coverage areas. Therefore, it is still a hot and difficult point to study the soil background removal method of NDVI. On the basis of simulating and analyzing the effects of typical soil types on NDVI of winter wheat with different vegetation cover, two soil background removal models based on mixed spectral theory are proposed in this paper. The winter wheat field in Chuzhou area of Anhui Province was used as the research area, and the winter wheat in the early jointing stage of paddy soil and yellow cinnamon soil was studied. The data of canopy spectrum and leaf area index (Lai) of winter wheat were measured. The traditional photo estimation method was used to calculate the vegetation coverage and the soil background removal ability of the two models was studied and analyzed. At the same time, the winter wheat field in Jining area of Shandong Province was used as the research area, and the winter wheat in the late jointing stage was studied in the soil background of cinnamon soil, tidal soil, paddy soil and shajiang black soil, and Landsat-8 OLI satellite multispectral remote sensing data were used. Validation and evaluation of the applicability and effectiveness of the soil background impact removal model. The main contents and conclusions are as follows: (1) combined with typical soil types and vegetation coverage, the effects of soil background on NDVI information of winter wheat farmland were studied. The sensitivity of NDVI to vegetation coverage in winter wheat farmland is also different, which provides a basis for the frequency selection of NDVI information estimation in different types of soil background, which is based on the theory of linear mixed spectrum. The NDVI1T model based on NDVI and the simplified NDVI1T model based on pixel dichotomy were constructed. The applicability and validity of the model were verified, and the ability of two models to resist soil noise was quantitatively studied by using the method of SNR analysis. It was found that NDVI1T extraction of vegetation information was better than that of NDVI1T in resisting soil noise. The two models are more suitable for the condition that the vegetation leaf cover degree is more uniform or the vegetation type is single, and the fitting relationship between the two soil background impact removal models and NDVI is good, and the correlation between the two models is more than 0.9. The results show that the NDVI soil background effect calculated by the NDVIT model can be modified and verified based on Landsat-8 OLI satellite remote sensing image. The research results show that the NDVI soil background impact removal model is based on Landsat-8 OLI satellite remote sensing image. In the case of four soil background types in the study area corresponding to Oli images, the soil background effect of large-scale NDVI products can still be corrected by removing the soil background effect model and calculating the NDVI fitting formula based on the image.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S127

【相似文献】