联合影像纹理分类与地形骨架线优化DEM的方法

The method of DEM refinement by combing image texture classification and terrain skeleton lines

针对丘陵区山谷地形起伏小且易受植被遮挡的影响,1∶50000数字高程模型(DEM)精细化生产中地形特征线采集效率和可靠性难以兼顾的问题,该文结合面向对象的图像分类和骨架提取等空间分析方法,提出了适合丘陵区高分辨率数字表面模型(DSM)的地形特征线半自动提取方法,优化了DEM生产流程和精度。对测区不同类型的1∶50000图幅开展生产试验,并对结果进行定量和定性评价。试验结果表明,所有图幅的特征线几何位置精度均高于77.8%,提取效率是人工采集的2~4倍。利用特征线精编后的DEM成果高程中误差小于3.5 m,最大误差不超过6.0 m,满足丘陵地高程精度要求,地貌形态表达合理,与立体模型套合良好。

Aiming at the problem that acquisition of terrain feature lines is difficult to balance efficiency and reliability in 1∶50000 DEM refinement production,due to the terrain occlusion caused by vegetation cover in hilly region of relatively gentle undulating terrain,a semi-automatic extraction method of terrain feature lines for high-resolution DSM in hilly region combining with remote sensing object-oriented classification and spatial analysis methods such as skeleton extraction was proposed in this paper to optimize DEM production process and accuracy.With the quantitative and qualitative evaluation,experiments and production work on four types of 1∶50000 DSM were carried on.Experimental results showed that the geometrical position accuracy of terrain feature lines of eight test data were higher than 77.8%,and the efficiency was 2~4 times of manual acquisition.The mean square error of optimized DEM was less than 3.5 m,while the maximum error were less than 6.0 m,which could meet the elevation accuracy requirement in hilly region.Moreover,the DEM products were capable of describing topographies reasonably and fitted well with stereoscopic model.