乡镇尺度的玉米种植面积遥感监测

Remote sensing monitoring of maize planting area at town level

以快速、准确提取玉米种植面积为目标,以多时相HJ-1A/1B CCD影像和数字高程模型(DEM)为信息源,选取吉林省长春市为试验区,将试验区种植结构、物候特征、地形特征、光谱特征及植被指数等多元信息引入决策树分类模型,构建基于决策树分层分类的玉米种植面积遥感估算模型,并将空间化的农业普查数据作为参考值,以乡镇为基本评价单元对玉米种植面积遥感测量结果进行精度评价。研究表明:利用该方法可以有效提高玉米识别精度,满足作物种植面积估算大范围、多时相的需求,有助于解决作物种植面积遥感估算业务运行时空分辨率的矛盾,乡镇尺度的玉米种植面积总量提取精度可达92.57%。

In order to extract maize planting acreage rapidly and accurately, taking Changchun city in Jilin province as study area, a multi-layer decision tree classification model was constructed based on multi-temporal HJ-1A/1B CCD images and digital elevation model （DEM）, which introduced with multi-information including planting structure, phenological characteristics, terrain feature of the study area, spectral characteristics and vegetation index. The precision of classification results was evaluated by spatial agricultural census data at town level. The results indicated that the method could improve maize identification accuracy, which reached up to 92.57%. The method can meet the demand for large-scale and multi-temporal agricultural monitoring system and resolve the spatiotemporal contradiction effectively in the large-scale crop acreage monitoring system.