基于国产GF-2遥感影像的大麻地块提取方法研究——以安徽省六安市苏埠镇为例

针对传统的基于像元的分类方法提取大麻地块结果存在较为破碎、精度较低的问题,以国产"高分二号"(GF-2)4m的多光谱遥感影像为数据源,在安徽省六安市苏埠镇选取了一个研究区,使用基于规则集的面向对象的方法实现了大麻地块的精确提取。首先,对研究区预处理过的GF-2遥感影像进行多尺度分割,在多尺度分割结果的基础上,确定提取大麻地块的最优分割尺度。其次,针对不同地物类型选取样本对象生成光谱曲线,分析大麻地块与其他地物类型的异同点,并基于光谱分析结果构建规则集最终实现大麻地块的提取。最后,将基于规则集的面向对象分类结果和基于像元分类(监督分类)的结果进行对比分析。结果表明,基于规则集的面向对象方法可以有效的提取出研究区内的大麻地块,精度可以达到91.09%,解决了传统基于像元分类方法提取大麻地块结果较为破碎的问题。

Study on the extraction method of cannabis plot based on China-made GF-2 remote sensing image-taking Subu town in Anhui Province as an example

In this study, the object-oriented classification method was used to extract cannabis plot so as to resolve the broken problem of the traditional classification method based on pixel. China-made GF-2 remote sensing images with a 4×4 m resolution were obtained in Subu town of Anhui Province. Firstly, multi-scale segmentation of GF-2 remote sensing image in the study area was made, and the optimal segmentation scale was determined on the basis of multi-scale segmentation results. Secondly, a spectral curve according to the different feature type samples was generated. The similarities and differences between cannabis plot and other types of surface features were analyzed, and a rule set based on the results of spectral analysis was established to achieve the final extraction. Finally, the classification results of object-oriented classification method based on rule set and traditional method based on pixel were compared. The experimental results showed that the object-oriented classification method based on rule set can extract cannabis plots effectively and solve the broken problem of traditional classification method based on pixel. The accuracy can be reached to 91.09%.