基于时序Landsat遥感影像枣树种植面积提取  

Extraction of Jujube Planting Area Based on Time Series Landsat Remote Sensing Images

作　　者：安霞 孙占海 张学东[1] AN Xia(School of Information Engineering,Tarim University,Alar 843300,China)

机构地区：[1]塔里木大学信息工程学院,新疆阿拉尔843300

出　　处：《安徽农学通报》2021年第16期135-137,153,共4页Anhui Agricultural Science Bulletin

基　　金：塔里木大学研究生科研创新项目(TDGRI201930)。

摘　　要：基于不同特征时间序列数据集,使用动态时间规整(dynamic timewarping,DTW)方法对时间序列遥感影像进行分类。基于时间序列Landsat 8影像数据,使用NDVI、EVI、第一主成分(principal componentanalysis 1,PCA1)3种特征数据集结合DTW算法,对比分析不同特征量对枣树的识别精度。结果表明:基于DTW(ND⁃VI)的时间序列特征数据集结合DTW算法能够得到较好的分类结果,基于时序DTW(EVI)特征量的方法次之,基于时序DTW(PCA1)特征量的方法的分类精度最低,总体精度分别为95.23%、93.73%、83.84%,Kappa系数分别为0.858、0.824、0.738。This paper uses dynamic time warping(DTW)method to classify time series remote sensing images based on different characteristic time series data sets.Based on the time series Landsat 8 image data,using NDVI,EVI,and principal component analysis(PCA1)three feature data sets combined with DTW algorithm to compare and analyze the recognition accuracy of jujube trees with different feature quantities.The results show that the time series feature data set based on DTW(NDVI)combined with DTW algorithm can get better classification results,the method based on time series DTW(EVI)feature quantity is second,and the classification based on time series DTW(PCA1)feature quantity method The accuracy is the lowest.The overall accuracy is 95.23%,93.73%,83.84%,and the Kappa coefficient is 0.858,0.824,0.738.

关 键 词：遥感 动态时间规整算法 枣树 时间序列 

分 类 号：S127[农业科学—农业基础科学]