基于GF-1遥感影像的农作物面积测量方法研究

发布时间：2018-07-07 10:51

本文选题：农作物面积 + 遥感测量　；参考：《吉林大学》2017年硕士论文

【摘要】：在第三次全国农业普查的背景下,针对本次普查的主要工作,即对农作物种植面积以及空间分布进行全面准确的调查,结合目前自动化提取农作物面积中存在的问题,本文试图寻找一种有效地快速获取大区域农作物总面积测量的方法,为农作物面积测算提供基础数据支撑,从而为制定科学的农业发展计划提供科学依据。根据遥感测量的难易程度,本文设定了三个等级的测量区,选取宁夏平原区、甘肃梯田区和贵州破碎区作为三个级别的代表区域。结合研究区物候数据以及高分一号(GF-1)WFV 16米卫星影像数据,选择性获取时相处于2016年3月-4月的国产卫星影像作为农作物面积遥感测量基础数据,数据源以GF-1为主,天绘一号卫星(TH-1)为辅。为了提高遥感影像数据处理效率,针对GF-1和TH-1全色与多光谱数据同步获取但匹配精度较差的特点,选择“先配准融合、后正射纠正”的处理流程进行遥感数据的批量、快速处理。以研究区为单位,分别制作正射影像成果。分别采用面向对象的计算机自动分类法对4波段正射影像数据、人工目视解译法对真彩色合成后的3波段正射影像数据进行农作物遥感测量。最后以外业抽样调查数据作为真值,从目视效果、测量精度以及测量时间三个方面,对两种方法测得的农作物面积测量结果进行对比分析。通过以上研究,本文主要取得了以下成果:(1)通过对研究区作物播种前、生长旺盛期、收获后等多个时相的GF-1WFV 16米卫星影像数据的充分对比、分析,并结合物候数据,在单期高分辨率影像数据的基础上,完成了农作物遥感测量。由此,提出了一种充分利用GF-1WFV16米数据宽幅大、回访周期短的特点,结合GF-1 2米融合数据的高分辨率特点,进行快速、大区域农作物面积测量的调查方法。(2)针对GF-1、TH-1数据特点,选择“先配准融合、后正射纠正”的处理方式,充分结合各遥感数据处理软件优势,对DOM制作中的配准、波段组合、融合、正射纠正、镶嵌、色彩调整、裁切等各主要环节均总结、研制了一系列批量、快速处理方法,为今后进行大规模农作物遥感测量提供了海量DOM快速制作解决方案。(3)本文选取了面向对象分类和人工目视解译两种方法,分别对三个研究区进行了农作物遥感测量,并以外业抽样调查数据作为真值,对两种测量结果进行了对比分析。研究表明:采用面向对象分类法获得的农作物面积测量结果在空间分布上与人工目视解译法基本一致;三类研究区的整体精度均能达到90%以上,能够满足应用需求;而在处理速度上,面向对象法相比人工目视解译法,可提高两倍左右,且随着测量面积的增加,其测量速度优势越明显。因此,当需要快速获取大范围农作物遥感测量结果时,采用面向对象的计算机自动分类法是一种比较好的选择方式。
[Abstract]:In the context of the third National Agricultural Census, in view of the main work of this census, that is, to carry out a comprehensive and accurate survey of crop planting area and spatial distribution, combined with the problems existing in automatic extraction of crop area at present, This paper attempts to find an effective and rapid method for measuring the total area of crops in a large area, which provides the basic data support for the calculation of crop area and provides scientific basis for making scientific agricultural development plan. According to the degree of difficulty and ease of remote sensing measurement, this paper sets up three grades of measuring areas, including Ningxia Plain, Gansu terraced area and Guizhou broken area as the representative regions of the three levels. Combined with phenological data of the study area and Gaof-1 (GF-1) WFV 16m satellite image data, the domestic satellite images from March to April 2016 were used as the basic data of crop area remote sensing measurement. GF-1 was the main data source. Tianyi-1 satellite (TH-1) is auxiliary. In order to improve the efficiency of remote sensing image data processing, aiming at the feature that GF-1 and TH-1 panchromatic data are acquired synchronously with multi-spectral data, but the matching accuracy is poor, the batch of remote sensing data is selected as "registration fusion first, then forward correction". Quick processing. Taking the research area as the unit, the orthophoto image results were made respectively. The orthophoto data of 4 bands were classified by object oriented automatic classification method, and the 3 band orthophoto image data of true color were measured by artificial visual interpretation method. Finally, as the true value, the results of crop area measurement obtained by two methods are compared and analyzed from three aspects: visual effect, measuring precision and measuring time. Through the above research, this paper mainly obtained the following achievements: (1) by comparing and analyzing the GF-1WFV16m satellite image data of the crops in the study area before sowing, growing vigorous period and after harvest, and combining phenological data, Based on single phase high resolution image data, crop remote sensing measurement is completed. Therefore, this paper puts forward an investigation method to make full use of the wide width of GF-1WFV16m data and the short period of return visit, combined with the high resolution characteristics of GF-1 / 2m fusion data, to measure the area of crops in a fast and large area. (2) aiming at the characteristics of GF-1WFV16m data, Selecting the processing method of "first registration fusion, then forward correction", fully combining the advantages of each remote sensing data processing software, making registration, band combination, fusion, orthographic correction, mosaic, color adjustment in Dom production, Cutting and other major links are summarized, developed a series of batch, rapid processing methods, This paper provides a solution for large-scale crop remote sensing measurement in the future. (3) in this paper, two methods, object oriented classification and artificial visual interpretation, are selected to measure crops in three research areas. The data of field sampling survey were used as true value to compare and analyze the two kinds of measurement results. The results show that the measured results of crop area obtained by the object-oriented classification method are basically consistent with the artificial visual interpretation method in spatial distribution, and the overall accuracy of the three study areas can reach more than 90%, which can meet the needs of application. In terms of processing speed, the object-oriented method can increase the speed of measurement by about twice as much as the artificial visual interpretation method, and with the increase of the measurement area, the advantage of the measurement speed is more obvious. Therefore, when it is necessary to quickly obtain the results of crop remote sensing measurement on a large scale, it is a better choice to adopt the object-oriented computer automatic classification method.
【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】