摘要:以研究区2012年的HJ卫星CCD影像为数据源，通过物候历和主要农作物的光谱特征分析，确定棉花识别最佳时相。采用分类回归树分析(CART)的决策树方法提取棉花种植面积信息，并以农田实地调查样点和统计数据为参考对提取的棉花种植面积结果进行评价。结果表明，基于HJ-CCD数据，使用CART算法的决策树可以较好地提取棉花覆盖信息，最终提取的棉花种植面积总量精度为94.29%，位置精度为88.57%；本研究采用的决策树方法，操作方便、容易实 现，分类结果较为实际，基本满足棉花种植面积遥感监测的需求，可对棉花种植面积估算和种植结构分析提供一定的参考。

关键词:棉花  面积提取  决策树分类  CART算法

AbstractIn this paper, the HJ satellite CCD images in research region of 2012 were selected as the source data, through the analysis of phenological calendar and spectral characteristics of major crops, the cotton identifiable best phase was determined. The cotton planted areas information were extracted by using the decision tree method based on CART analysis, also the extracted areas result have been evaluated by the field real survey samples and the statistical data as a reference. The results showed that: Based on the HJ satellite CCD images data, using the decision tree of CART algorithm can be well applied to extract the cotton covering information. The final total accuracy for extracted cotton planted areas was 94.29%, and the position precision was 88.57%. In this research, adopted the decision tree method, has the advantages of convenient operation, easy to implement and more realistic classification results. It can be basically met the needs of remote sensing monitoring for cotton planted areas, and provided a certain reference to estimate the cotton planted areas and analyze the planting structures.

Key wordscotton  areas extraction  decision tree classification  CART algorithm

文章编号： 0258_7106 (2016) 01_0018_15 中图分类号： P618.41 文献标志码：A

改回日期：2015_07_11

基金项目

**通讯作者耿新霞， 女， 1979年生， 助理研究员， 成矿规律研究方向。 Email： gen gxinxia@cags.ac.cn