基于多时相HJ卫星的渭干河-库车河绿洲主要农作物种植信息提取

CROP EXTRACTION BASED ON MULTI-TEMPORAL HJ SATELLITE IN THE WEIGAN-KUQA RIVER OASIS

关键词:  HJ卫星 决策树分类 CART算法 农作物 渭干河-库车河绿洲

Key words:  HJ satellite  decision tree classification  CART algorithm  crops  Weigan - kuqa river oasis

摘要:

在3S技术的支持下全面、快速、客观地监测农作物种植信息，对于正确的把握该区域的农业结构和布局，进行作物种植空间格局的调整和优化有着十分重要的意义。本文以渭干河-库车河洲绿洲为研究区，根据农作物的物候规律和季相节律的差异性特点，选取2012年的3景不同时相的HJ卫星CCD遥感数据，ENVI下基于CART算法的决策树规则自动提取主要农作物覆盖信息，然后以野外GPS 调查点为依据，对决策树方法预分类结果进行修正，成功提取了研究区的玉米、棉花和小麦的种植面积，总体精度达到了9173%。结果表明HJ卫星CCD影像可以很好地应用于农作物提取，而且CART算法的分类精度较高，能较好地反映作物的分布状况，可为该地区主要作物种植结构调整提供一定的依据。

Abstract:

To comprehensively, rapidly and objectively monitor crop planting information within the support of 3S technology, has a very important significance in understanding region's agricultural crops structure and layout, adjusting and optimizing spatial pattern. This paper, taking Weigan-Kuqa River delta oasis as the study area, according to the crop phonological phase patterns and seasonal differences in rhythm characteristics, selected 3 scene of HJ-CCD satellite remote-sensing images in 2012 as the source data, automatically extracted the main crop coverage information using ENVI Based CART algorithm decision tree rules, then modified the decision tree classification results based on the field GPS survey sites and successfully extracted corn, cotton and wheat acreage in the study area. The overall accuracy reached 9173%. The results showed that HJ satellite CCD image can be well applied to crops extraction, classification accuracy of CART algorithm was higher, which can be used to better reflect the distribution of crops. The conclusion can provide a certain basis for the adjustment of main crop planting structure in the region.