基于多时相环境星NDVI时间序列的农作物分类研究

时相和归一化植被指数(NDVI)时间序列特征在农作物分类提取方面具有重要的应用价值。以黑龙江红星农场为研究区,利用多时相环境星HJ-1A/B CCD数据及其多期平滑重构后的NDVI时间序列曲线特征,在对象尺度上采用决策树算法开展了农作物分类研究,通过与单独利用多时相遥感数据分类结果的对比分析,研究了增加NDVI时序曲线特征对分类精度的影响。结果表明:面向对象分类方法得到的地块较为规则,平滑了地块内部同种作物间的噪声,避免了"椒盐现象",适合于我国东北地区农作物分类识别;利用NDVI时序曲线特征参与分类,增强了不同作物之间的光谱差异,提高了作物分类精度,比仅使用3个多时相HJ-1A/B CCD数据分类精度提高了5.45%,Kappa系数提高了0.09。通过该研究探讨了NDVI时序曲线特征在作物分类中的应用,拓展了遥感数据在农业领域的应用范围,具有推广价值。

Research on Crop Identification Using Multi-temporal NDVI HJ Images

Research on crop identification has an important significance for temporal,spectral and the time\|series of Normalized Difference Vegetation Index(NDVI) profiles features.In the paper,a series of NDVI dataset based on the Savitaky\|Golay filtering smoothing method was generated and the classification method based on object\|oriented was used in the study area of Hongxing Farm by HJ images.Then this paper discussed the effects on the multi\|temporal data classification accuracy by time\|series NDVI database.The results indicated that the object\|oriented classification method is suitable and feasible for the northeast area in our country,and compared to only using 3 temporal HJ images by the NDVI time\|series curves in classification which can improve crop classification accuracy,the overall accuracy improved by 5.45%,and the Kappa increased by 0.09.NDVI profiles features associating with multi\|temporal data is used in crop identification to enlarge the application of remote sensing data and have the potential of spread in the domain of agriculture.