基于GF-2的喀斯特城市生态用地监测研究——以七星关城区为例

掌握城市生态用地结构变化对保护生态用地、建设新型城市具有重要意义。该文选取毕节七星关城区为研究区,以2016年、2021年的GF-2遥感影像为数据源,通过监督分类、非监督分类以及面向对象分类方法进行生态用地遥感监测及变化分析。研究结果表明,面向对象的分类方法能够更好地实现对高分影像分类,其中,监督分类中最大似然法分类结果的总体精度为80.47%,非监督分类中ISODATA分类结果的总体精度为65.56%,而面向对象的分类方法总体精度可达到91.82%;面向对象的分类方法结合了光谱特征、形状特征、纹理特征以及对象之间的拓扑信息,当分割尺度为20时可以将影像对象很好地分割,解译结果更精准,分类效果更好;2016—2021年间,七星关区生态用地范围持续减小,共减少面积19.32 km^(2)。其中生态用地转化为建设用地的面积为17.78 km^(2),占比92.03%;生态用地转化为道路的面积为1.54 km^(2),占比7.97%。

It is of great significance to master the structural change of urban ecological land for protecting ecological land and building a new city.This paper selects Qixingguan urban area of Bijie City as the study area,takes the GF-2 remote sensing images of 2016 and 2021 as the data source,and carries on the remote sensing monitoring and change analysis of ecological land through supervised classification,unsupervised classification and object-oriented classification.The results show that,the object-oriented classification method can better achieve the classification of high-score images,in which the overall accuracy of the maximum likelihood classification results in supervised classification is 80.47%,the overall accuracy of ISODATA classification results in unsupervised classification is 65.56%,and the overall accuracy of object-oriented classification method can reach 91.82%.The object-oriented classification method combines spectral features,shape features,texture features and topological information between objects.When the segmentation scale is 20,the image object can be segmented well,the interpretation result is more accurate,and the classification effect is better.From 2016 to 2021,the scope of ecological land in Qixingguan District continued to decrease,with a total reduction area of 19.32 km^(2).Among them,the area of ecological land converted into construction land is 17.78 km^(2),accounting for 92.03%;and the area of ecological land converted into roads is 1.54 km^(2),accounting for 7.97%.