基于GF-1与Landsat8 OLI影像的作物种植结构与产量分析

发布时间：2018-04-28 15:53

本文选题：遥感 + 作物　；参考：《农业工程学报》2017年11期

【摘要】：作物种植结构监测和估产是精准农业遥感的重点领域,其研究对于指导作物种植结构和制定农业政策具有重要意义。该文以黑龙江省北安市为研究区,以2015年的Landsat8 OLI和多时相GF-1为遥感数据源,基于物候信息和光谱特征确定的农作物识别关键时期和特征参数,构建面向对象的决策树分类模型,开展作物种植结构监测研究;综合植被光谱指数和地面采样数据,采用逐步回归方法建立产量遥感估算模型。结果表明:多源与多时相的遥感数据可以反映不同农作物的季相特征,应用本文所构建的决策树分类模型,作物分类效果较好,总体精度达87.54%,Kappa系数为0.8115;2015年,北安市的主要作物类型为大豆、玉米、水稻和小麦,面积分别为2204、1955、122和19 km~2,其中大豆的种植面积最大,占作物种植面积的51.24%。基于NDVI、EVI和GNDVI构建的多元回归模型为北安市大豆和玉米产量估算最优模型(R~2=0.823 7,均方根误差135.45 g/m~2,精度80.55%);北安市玉米高产区集中分布在西部,大豆的高产区主要分布在东部;2015年北安市玉米和大豆的单产分别为8 659、2 846 kg/hm~2,总产量分别为16.93×10~8、6.27×10~8 kg。利用作物关键物候期的多源多时相遥感数据能够精确高效地提取作物种植结构,构建的产量估算多元回归模型,为精准农业科学发展提供参考。
[Abstract]:Crop planting structure monitoring and yield estimation are the key fields of remote sensing in precision agriculture, and its research is of great significance in guiding crop planting structure and formulating agricultural policies. Taking Bei'an city of Heilongjiang Province as the research area and Landsat8 OLI and multitemporal GF-1 of 2015 as remote sensing data sources, this paper constructs an object-oriented decision tree classification model based on the key periods and feature parameters of crop identification determined by phenological information and spectral features. The monitoring of crop planting structure was carried out, and the remote sensing estimation model of yield was established by using stepwise regression method based on vegetation spectral index and ground sampling data. The results show that the multi-source and multi-temporal remote sensing data can reflect the seasonal characteristics of different crops. Using the decision tree classification model constructed in this paper, the effect of crop classification is better, the overall precision is 87.54 and the Kappa coefficient is 0.8115. The main crop types in Bei'an city are soybean, corn, rice and wheat, with an area of 2204An 1955122 and 19kmm2, respectively. Among them, soybean has the largest planting area, accounting for 51.24g of the crop area. The multivariate regression model based on NDVIVI and GNDVI is the best model for estimating soybean and maize yield in Bei'an city, the root mean square error is 135.45 g / m ~ 2, and the precision is 80.550.The high yield area of corn in Bei'an City is concentrated in the west of China, and the root mean square error is 135.45 g / m ~ (-2). In 2015, the yield of maize and soybean in Bei'an was 8 659 kg / hm ~ (-2) and 16.93 脳 10 ~ (8) ~ 8 ~ (-8) kg ~ (-1) 路hm ~ (-2) 路hm ~ (-1) 路hm ~ (-1), respectively. Using the multi-source and multi-temporal remote sensing data of crop key phenology, the crop planting structure can be extracted accurately and efficiently, and the multiple regression model of yield estimation can be constructed, which provides a reference for the scientific development of precision agriculture.
【作者单位】：赤峰学院资源与环境科学学院;中国科学院东北地理与农业生态研究所中国科学院湿地生态与环境重点实验室;中国科学院大学;吉林大学地球科学学院;中国科学院遥感与数字地球研究所;
【基金】：中国科学院野外站联盟项目(KFJ-SW-YW026) 国家重点研发计划子课题(2016YFC0500201-03)
【分类号】：S127;TP79

【相似文献】