天津市农作物空间分布遥感调查与质量评定方法研究

发布时间：2018-03-05 01:24

本文选题：GF-1　切入点：WFV　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：农作物主要类型识别是估算农作物面积、及时监测其长势以及预测其产量的基本首要工作,是农情遥感的基础[1]。因此,及时、准确、大范围获取农作物种植类型、长势及其分布情况,可以为粮食生产和农业发展提供相关信息,对于农业生产管理和政府制定粮食政策具有重大的意义。近年来,遥感技术发展日新月异,数据源持续增加,光谱分辨率和空间分辨率也越来越高,遥感技术在农业中的应用也在不断扩展,使得快速、准确获取农作物种植面积和空间分布成为可能[2]。本文以天津市为研究区,以GF-1 WFV为遥感数据源,选取了该区2016年主要农作物生长关键期影像,结合地面实地调查样方数据,综合运用RS、GIS等技术,采用决策树分类法提取了天津市主要作物冬小麦、水稻、玉米的空间分布数据。取得的研究成果主要有以下几个方面:(1)通过对研究区主要作物生长关键期影像的正射纠正、辐射定标、大气校正、镶嵌和裁切等处理,取得了覆盖整个研究区的农作物生长关键期遥感影像数据。(2)提取并合成了各期遥感影像的植被指数(NDVI)和水体指数(NDWI),结合地面调查样方数据,绘制了天津市冬小麦、水稻、玉米等主要农作物NDVI和NDWI的季相变化曲线。(3)结合天津市主要农作物的物候数据,分析了冬小麦、水稻、玉米等主要农作物NDVI和NDWI的季相变化规律,确定了研究区常见地物目标的遥感影像分类指标和分类准则,确定区分不同类型农作物的时间节点,通过反复试验,选择了合适的分类阈值。(4)利用ENVI5.1软件平台,构建了天津市夏收作物——冬小麦以及秋收作物——水稻和玉米的决策树分类模型,取得了天津市主要农作物的空间分布图。(5)根据研究区农业统计数据和农作物空间分布的实地调查样方数据,用混淆矩阵方法统计分析了研究区不同季节收获的农作物空间分布遥感测量结果的精度。统计分析结果表明:夏收作物总体精度为95%,Kappa系数为0.89,冬小麦生产者精度、用户精度分别为95%、91%;秋收作物总体精度为90%,Kappa系数为0.82,其中水稻生产者精度、用户精度分别为86%和85%,玉米生产者精度、用户精度分别为93%、87%。由此可见,本文所述的农作物空间分布遥感调查方法的预测精度满足三农普对精度要求,是一种可行的农作物空间分布遥感测量方法。
[Abstract]:The identification of main types of crops is the basic primary task of estimating crop area, monitoring its growth and predicting its yield in time, which is the basis of remote sensing of agricultural conditions [1] .Therefore, it is timely, accurate and extensive to obtain crop planting types. Growth and its distribution can provide relevant information for food production and agricultural development, and are of great significance for agricultural production management and food policy formulation by governments. In recent years, remote sensing technology has developed rapidly and data sources have continued to increase. The spectral resolution and spatial resolution are higher and higher, and the application of remote sensing technology in agriculture is also expanding, which makes it possible to obtain crop planting area and spatial distribution quickly and accurately [2]. Taking GF-1 WFV as the remote sensing data source, this paper selected the image of the key period of crop growth in 2016 in this area, combined with the ground survey sample square data, synthetically used the technology of GF-1 WFV and so on, used the decision tree classification method to extract the main crops winter wheat and rice in Tianjin. Spatial distribution data of maize. The main research results obtained are as follows: 1) through orthographic correction, radiometric calibration, atmospheric correction, mosaic and cutting of images of key periods of crop growth in the research area, The vegetation index (NDVI) and water body index (NDWI) of each remote sensing image were extracted and synthesized, and the winter wheat and rice in Tianjin were drawn based on the ground survey sample data. Based on the phenological data of main crops in Tianjin, the seasonal variations of NDVI and NDWI, such as winter wheat, rice, maize and other main crops, such as winter wheat, rice and maize, were analyzed. The classification indexes and criteria of remote sensing images of common ground objects in the study area are determined, and the time nodes for distinguishing different types of crops are determined. Through repeated experiments, the appropriate classification threshold is selected. (4) the ENVI5.1 software platform is used. The decision tree classification model of summer harvest crops, winter wheat and autumn harvest crops, rice and maize, was constructed in Tianjin. The spatial distribution map of main crops in Tianjin is obtained.) according to the agricultural statistical data of the study area and the field survey sample data of the spatial distribution of crops, The precision of remote sensing measurement of the spatial distribution of crops harvested in different seasons in the study area was statistically analyzed with the method of confusion matrix. The results showed that the total precision of summer harvest was 0.89 and the precision of winter wheat producer was 0.89. The user's accuracy is 95 / 91s respectively; the overall precision of the autumn harvest crops is 90 and Kappa's coefficient is 0.82, of which the rice producer's accuracy and the user's precision are 86% and 85 respectively, the corn producer's accuracy and the user's precision are 9330.87. thus, we can see that, The prediction accuracy of the remote sensing survey method for the spatial distribution of crops described in this paper meets the requirements of the precision of San Nong Pu, and is a feasible remote sensing method for measuring the spatial distribution of crops.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S127;TP751

【相似文献】