摘要：This paper proposes a method for road extraction from high resolution imagery based on Mathematical Morphology theory. The proposed method consisted on the development of mathematical morphology and structuring elements based routine using the SDC Morphology Toolbox for MATLAB. The routine uses operators as opening, closing, skeletonisation, erosion, dilation, morphologic gradient, among others. Prior to processing, the whole study area was subset into 11 sub-images due to the limitation of MATLAB and available hardware on computing large raster data. Thus, the developed morphologic routine was capable to be replicable for all 11 sub-images by considering only changes for the thresholds values, which differ for each image. The 11 sub-images were mosaicked later than the road extraction to recompose the overall test area. Next, the road vectors were generated to compute the quality analysis of the result obtained with the extraction. The assessment of the result was based on linear correspondence analysis. The high degree of the completeness metric indicated quite satisfactory results. It proves the effectiveness of the Mathematical Morphology theory as part of the features extraction process.
