基于多时相影像植被指数变化特征的作物遥感分类

高分一号GF1/WFV遥感影像具有较高的时间和空间分辨率,利用多时相影像开展农作物分类调查具有明显优势。以安徽省颍上县为研究区域,利用2017年5月至9月共6景多时相GF-1/WFV卫星遥感影像数据对主要农作物的分类识别提取。首先,通过分析研究区主要农作物的典型植被指数NDVI、EVI和WDRVI时序变化特征,明析了不同作物在各时相对不同VI的响应特征;其次,基于作物在不同时相的敏感VI变化响应,构建了决策树分层分类模型,成功提取了研究区玉米、水稻、大豆和甘薯四种主要作物种植空间分布情况。结果表明:总体精度达到90.9%,Kappa系数为0.895。同时,采用最大似然法、支持向量机对研究区作物进行分类,通过分类效果对比发现,最大似然法最差,支持向量机次之,决策树分类方法最佳。研究表明:利用多时相时间序列的遥感影像数据,结合作物植被指数特征,采用决策树分类方法可以有效提高作物分类的精度。

Remote Sensing Classification of Crops Based on the Change Characteristics of Multi-phase Vegetation Index

GF-1/WFV remote sensing image has higher temporal and spatial resolution. It has obvious advantages to carry out crop classification survey using multiphase images. This paper took Yingshang county of Anhui Province as the study area and extracted main crop classification based on the GF-1/WFV satellite remote sensing image data from May to September 2017. Firstly, through analyzing the sequential variation characteristics of major crops typical vegetation index NDVI, EVI and WDRVI in the study area, made clear the response characteristics of different crops in each time. Then, based on sensitive VI changes of crops in different time phases, the paper constructed decision tree classification model, and successfully extracted the spatial distribution of 4 major crops: corn, rice, soybean and sweet potato in the study area. The results showed that the overall accuracy was 90.9%, and Kappa coefficient was 0.895. At the same time, this paper used the maximum likelihood method, supported vector machine (SVM) for crop classification. The comparison results showed that the maximum likelihood method was the worst, and the support vector machine was second, and the decision tree classification method was the best. Above results indicated that using the remote sensing image data of time series of multi-phase combined with the feature of crop vegetation index, the classification method of decision tree could effectively improve the accuracy of crop classification.