基于国产高时空分辨率卫星影像的作物种植信息提取研究

在面向对象技术支持下,首先利用高空间分辨率ZY-3遥感影像提取农田地块专题信息;然后在地块边界控制下以地块对象为单元融入HJ-1及GF-1中分传感器的多时相光谱信息,获取作物生长关键期内的时间序列光谱特征;最后,结合不同作物的物候差异性规律构建作物种植信息提取模型,对甘蔗和水稻进行识别。结果表明,所有地类的总体分类精度为86.80%,Kappa系数为0.84,总体分类效果良好。甘蔗的制图精度和用户精度分别达到92.11%和90.91%,水稻的制图精度和用户精度分别达到88.89%和90.91%。说明协同利用国产卫星的高空间和高时间分辨率影像数据提取作物种植信息确实可行,可作为作物种植面积和种植结构的精细化、快速调查方法。

Cropping Information Acquisition from National High Temporal-and-spatial Resolution Satellite Imaging System

Supported by the object-oriented technology, information on sugarcane and rice fields were acquired from the high temporal-and-spatial resolution ZY-3 remote sensing images.Using the crop plots as the objects, the multi-temporal spectral signatures from HJ-1 and GF-1 sensors were integrated, and the time series spectrum signatures at critical growth points were captured.Based on them, the cropping models showing phenology differences on selected features of sugarcane and rice fields were constructed.The overall classification accuracy of the models was 86.80%, the Kappa coefficient approximately 0.84, and the overall classification effect desirable.The mapping accuracy and user accuracy for sugarcane reached 92.11% and 90.91%, respectively;and those for rice, 88.89% and 90.91%, respectively.It appeared that the national high temporal-and-spatial resolution satellite imaging system could provide adequate information for proper management on the planting area and structure for the crops.