结合空间分析的面向对象无人机影像土地利用分类

Object-oriented land use classification from UAV imagery with spatial analysis

利用高分辨遥感影像进行土地利用分类,为农村土地利用动态监测及土地综合整治快速地提供基础地理空间数据。以高分辨无人机影像为数据源,研究利用面向对象多尺度分割技术结合GIS空间分析对影像进行土地利用分类。根据对象内同质性高、对象间异质性高的准则,引入加权局部方差与空间自相关指数构建全局最优分割非监督评价指数,然后利用最邻近分类器对影像进行分类。实验结果表明,该方法减少人工目视确定最优分割尺度的主观性,能够避免某些地物不能被有效归类的现象,在单一尺度下获得较高的分类精度。

Using high resolution remote sensing imagery to conduct land use classification can provide basic geo-spatial data for rural land use dynamic monitoring and land comprehensive improvement.Taking high resolution UAV imagery as the test data,this paper makes the land use classification by using object-oriented multi-scale segmentation technique combined with GIS spatial analysis.According to the criterion of high intra-segment homogeneity and inter-segment heterogeneity,the weighted local variance and spatial autocorrelation index are introduced to construct the unsupervised evaluation index of global optimal imagery segmentation,and land use classification is then conducted by use of the nearest neighbor classifier.The experimental result shows that this method can reduce the subjectivity of artificial visual determination of the optimal segmentation scale,avoid the phenomenon that some objects can’t be effectively classified,and obtain a higher classification accuracy at a single scale.