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

多时相遥感影像检测平乐县晚稻种植面积变化
引用本文:黄维,黄进良,王立辉,胡砚霞,韩鹏鹏,王久玲.多时相遥感影像检测平乐县晚稻种植面积变化[J].农业工程学报,2014,30(21):174-183.
作者姓名:黄维  黄进良  王立辉  胡砚霞  韩鹏鹏  王久玲
作者单位:1. 中国科学院测量与地球物理研究所,武汉 430077; 中国科学院大学,北京 100049
2. 中国科学院测量与地球物理研究所,武汉,430077
基金项目:中科院战略性先导科技专项-应对气候变化的碳收支认证及相关问题(XDA05050107)
摘    要:为检测中国主要的粮食作物水稻的种植面积变化,该文以广西平乐县为例,利用多时相陆地卫星专题成像仪(landsat thematic mapper)影像数据和面向对象的分类方法,提取出晚稻种植的变化区域。该文探索了变化强度计算和阈值确定的方法,并利用晚稻在不同时相的影像光谱特征变化来提取晚稻种植区域。试验结果表明:3种变化强度计算方法中,变化向量分析法对河流、滩地变化的抑制效果优于相关系数方法,而相关系数方法对山体阴影的抑制效果则优于变化向量分析法,向量相似度法对山体阴影非常敏感,对水田变化则敏感度较低;3种阈值确定方法中最小错误率方法比最大类间方差法更为精确,比双窗口变步长阈值搜寻法更为稳定。综合利用3种变化提取方法对平乐县晚稻种植面积变化进行检测,得到变化检测混淆矩阵总正确率为96.8%,稻田面积变化误差为2.85%。该方法可为作物种植面积的变化检测提供参考。

关 键 词:遥感  作物  提取  变化向量  相关系数  面向对象  水稻  变化检测
收稿时间:2014/5/15 0:00:00
修稿时间:2014/10/28 0:00:00

Detection of late rice's planting area change in Pingle County based on multi-temporal remote sensing images
Huang Wei,Huang Jinliang,Wang Lihui,Hu Yanxi,Han Pengpeng and Wang Jiuling.Detection of late rice''s planting area change in Pingle County based on multi-temporal remote sensing images[J].Transactions of the Chinese Society of Agricultural Engineering,2014,30(21):174-183.
Authors:Huang Wei  Huang Jinliang  Wang Lihui  Hu Yanxi  Han Pengpeng and Wang Jiuling
Institution:1. Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China2. University of Chinese Academy of Sciences, Beijing 100049, China;1. Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China;1. Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China;1. Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China2. University of Chinese Academy of Sciences, Beijing 100049, China;1. Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China2. University of Chinese Academy of Sciences, Beijing 100049, China;1. Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, China2. University of Chinese Academy of Sciences, Beijing 100049, China
Abstract:Abstract: Rice is the most important food crop in China and its production ranks first in three major grain crops which are wheat, rice and corn. Therefore, it is very important to understand planting area and growth of rice. Because change detection is time-sensitive, remote sensing data is often used as a data source for it. Most of widely-used change detection methods using TM image are at pixel level, or only use image of single time phase. Based on multi-temporal Landsat TM images, this research extracted change area of rice planting in Pingle County. The process explored the methods of calculating changes in intensity and determining the threshold. First, the paper selected images of heading stage and harvest stage as data sources, and except for crops, there are no significant changes of ground objects in a few months. Since the spectral characteristics of late rice in the two periods are different from those of other crops, the changes of the images at two stages in the same year could be used to extract the change area of rice planting. In the process, change vector analysis method, correlation coefficient method and vector similarity method were used to calculate the change intensity. Otsu method, minimum error rate method and the method based on double window with variable step size were used to determine the threshold of the change intensity map. The change area maps of rice planting extracted by nine combinations of the methods were compared. Comprehensive utilization of three methods for extracting changes could get change area of rice planting with higher accuracy. This paper chose minimum error rate method based on histogram curvature or the method based on double window with variable step size to determine the threshold. The intersection of change area based on change vector analysis method and change area based on correlation coefficient method could inhibit the pseudo change of river beach and mountain shadow. Then the rice area was extracted exactly according to the difference between NDVI values and the object-oriented method. In order to refine the change detection of rice planting areas, the change areas of all ground objects and the change area of rice planting in two years were intersected to get the final variation for rice planting area. The results showed that the inhibitory effect of change vector analysis method for river beach change was better than correlation coefficient method, and yet the inhibition effect of correlation coefficient method for mountain shadow was better than change vector analysis method. Besides, vector similarity method was sensitive to mountain shadow and had low sensitivity to paddy field change. Among the three threshold determination methods, minimum error rate method was more accurate than Otsu method and more stable than the method based on double window with variable step size. Finally, the overall accuracy of change detection for late rice's planting area in Pingle County reached 96.8% in confusion matrix for verification. The error of the change area was 2.85% compared with the statistical data. This method could effectively extract the change of rice planting area in Pingle County.
Keywords:remote sensing  crops  extraction  change vector  the correlation coefficient  object-oriented  rice  change detection
本文献已被 CNKI 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