首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于NDVI加权指数的冬小麦种植面积遥感监测
引用本文:王利民,刘佳,杨玲波,杨福刚,滕飞,王小龙.基于NDVI加权指数的冬小麦种植面积遥感监测[J].农业工程学报,2016,32(17):127-135.
作者姓名:王利民  刘佳  杨玲波  杨福刚  滕飞  王小龙
作者单位:中国农业科学院农业资源与农业区划研究所,北京,100081
基金项目:高分辨率对地观测系统重大专项"高分农业遥感监测与评估示范系统(一期)"。
摘    要:该文针对农业信息服务中冬小麦种植面积调查业务的现状与需求,提出了一种基于NDVI(normal difference vegetation index)时间序列的冬小麦NDVI加权指数(WNDVI,weighted NDVI index)影像算法,可在训练样本、验证样本选择的基础上实现冬小麦面积的自动提取,并以河北省安平县及周边地区2013-2014年度冬小麦面积提取为例,采用GF-1/WFV(wide field view)数据进行了算法实现。算法的主要思路是在时序影像基础上,通过冬小麦NDVI加权指数影像的构建,扩大冬小麦地类与其他地类的差异,结合自适应的阈值获取方法,区分冬小麦地类,获取冬小麦作物面积。算法包括冬小麦时间序列影像的获取、基于网格的样本点设置、构建冬小麦 NDVI 加权指数影像、迭代确定冬小麦NDVI加权指数提取阈值、精度验证这5个部分。影像的获取根据冬小麦的生长时间确定,保证每月1景GF-1/WFV无云影像,并进行预处理及NDVI计算;同时将研究区划分为一定数量的网格,每个网格再等分为2×2个子网格,根据目视解译、专家知识、实地调查等方法,确定左上网格中心点及右下网格中心点的地物类型。统计该期所有左上网格点冬小麦及其他地物的NDVI均值,冬小麦NDVI大于其他地物的将该期影像的权值设置为1,否则设置为?1,将所有时相NDVI影像进行加权平均,即可获取冬小麦NDVI加权指数影像。获取冬小麦NDVI加权指数影像后,还需设置合适的阈值提取冬小麦。该文选用右下网格点目视解译分类结果作为阈值提取依据,具体方法是将冬小麦指数从小到大按照一定间隔划分,作为冬小麦 NDVI 加权指数提取阈值,将各阈值二值法运用,与右下网格点的冬小麦提取的目视解译结果对比,精度最高的就是最优冬小麦 NDVI 加权指数分割阈值。在所有网格中,以初始识别获取的冬小麦面积为准,等概率选择10个样方作为精度验证样方进行验证。精度验证结果表明分类总体精度达到94.4%,Kappa系数达0.88。该文通过构建冬小麦NDVI加权指数,将比较复杂的多个参数转换为一个参数,并且农学意义明确,相比传统的NDVI时序影像进行冬小麦面积的提取,具有自动化程度高、面积提取精度高、分类结果稳定的特点,已经在全国农作物面积遥感监测业务中进行了应用。

关 键 词:遥感  作物  监测  种植面积  冬小麦  GF-1  NDVI  多时相  NDVI加权指数
收稿时间:2015/12/30 0:00:00
修稿时间:2016/6/14 0:00:00

Remote sensing monitoring winter wheat area base on weighted NDVI index
Wang Limin,Liu Ji,Yang Lingbo,Yang Fugang,Teng Fei and Wang Xiaolong.Remote sensing monitoring winter wheat area base on weighted NDVI index[J].Transactions of the Chinese Society of Agricultural Engineering,2016,32(17):127-135.
Authors:Wang Limin  Liu Ji  Yang Lingbo  Yang Fugang  Teng Fei and Wang Xiaolong
Institution:Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China,Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China,Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China,Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China,Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China and Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China
Abstract:Abstract: Remote sensing technology is a major method to obtain spatial distribution and quantity of winter wheat area, and classification method suitable for business operation is a key technology target of annual winter wheat remote sensing monitoring. Aimed at the conditions and demands of winter wheat background survey business operation in agriculture information service, this paper has proposed a weighted NDVI index (WWAI) based on normal difference vegetation index (NDVI) time sequence. By taking the extraction of 2013-2014 winter wheat area of Anping County, Hebei Province as an example, the algorithm is realized by using GF-1/WFV (wide field view) data. The main idea of the algorithm is to amplify the difference between winter wheat land type and other ground object types by establishing a winter wheat area index based on time sequence images, and to differentiate winter wheat land type from the others and thus to obtain the crop area of winter wheat by automated threshold value setting method. The algorithm includes the following 5 parts: acquisition of winter wheat time sequence images, sample points setting based on grid, establishment of winter wheat area index, identifying winter wheat area index estimation threshold value by iteration, and accuracy validation. Acquisition of images is based on the identification of growth time of winter wheat, and the principle is to ensure to get one GF-1/WFV cloudless image each month. Growth period of winter wheat in Anping County is from October 1st to June 30th of the next year, including 9 growing stages, i.e. seeding, germinating, tillering, overwintering, reviving, jointing, head sprouting, milking maturity and maturity. One GF-1/WFV cloudless image is selected in the middle 10 days of each month, and a total of 9 images are selected for pre-processing and NDVI calculation. Meanwhile, the study area is divided into a certain number of grids, and each grid is further divided into 2×2 sub-grids. The ground object types of central points in upper left and lower right grid are identified by visual interpretation, expert knowledge and field investigation. In this paper, a total of 10×10 equal interval grids with the average grid size of 4.1 km × 4.0 km, as well as 400 sub-grids with the size of 2.05 km × 2.0 km are obtained. The average NDVI values of winter wheat and other ground objects on all upper left centers of this period are calculated. If the winter wheat NDVI is higher than that of other ground objects, the weight of the images of the period is set to 1, and otherwise, set to -1. The winter wheat area index images can be obtained by using the weighted average of NDVI images of all time phases. After obtaining winter wheat area index, it is also necessary to set appropriate threshold value for winter wheat area extraction. The paper takes the visual interpretation classification results of lower right grid points as the basis for threshold value extraction. The specific method is to divide winter wheat area index from small to large with certain intervals, and then to make dimidiate extraction of winter wheat area indices of the lower right centers by taking each divided value as the extraction threshold value. By comparing with the visual interpretation result, the result with the highest accuracy is taken as the optimal winter wheat area index extraction threshold value, which is identified to be approximately 1 600 with self-adaptation approach finally. In all grids, accuracy validation is conducted by taking the 10 plots with equal probability. Accuracy validation results show that the overall classification accuracy has reached 94.4%, with Kappa coefficient of 0.88. The area extraction accuracy of this method is about 1.7% higher than that of conventional method based on NDVI time sequence images. By establishing winter wheat area index, this paper turns a complicated multiple-parameter problem into a single-parameter problem with clearly defined agricultural significance. This method is featured with high automatic degree and stable classification results, and it has been widely applied in the crop area remote sensing monitoring practices in China.
Keywords:remote sensing  crops  monitoring  planting area  winter wheat  GF-1  NDVI  multi-temporal phases  weighted NDVI index
本文献已被 万方数据 等数据库收录!
点击此处可从《农业工程学报》浏览原始摘要信息
点击此处可从《农业工程学报》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号