The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLII-2/W13
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-2/W13, 67–72, 2019
https://doi.org/10.5194/isprs-archives-XLII-2-W13-67-2019
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-2/W13, 67–72, 2019
https://doi.org/10.5194/isprs-archives-XLII-2-W13-67-2019

  04 Jun 2019

04 Jun 2019

FOSS4G DATE FOR DSM GENERATION: SENSITIVITY ANALYSIS OF THE SEMI-GLOBAL BLOCK MATCHING PARAMETERS

L. Lastilla1,2, R. Ravanelli1, F. Fratarcangeli1, M. Di Rita1, A. Nascetti3, and M. Crespi1,2 L. Lastilla et al.
  • 1Geodesy and Geomatics Division, DICEA - University of Rome “La Sapienza”, Rome, Italy
  • 2Sapienza School for Advanced Studies, Rome, Italy
  • 3Geoinformatics Division, Department of Urban Planning and Environment - KTH Royal Institute of Technology, Stockholm, Sweden

Keywords: DSM generation, High Resolution Optical Satellite Imagery, Semi-Global Matching, Free and Open Source Software for Geospatial

Abstract. DATE (Digital Automatic Terrain Extractor) is a Free and Open Source Software for Geospatial (FOSS4G), which combines photogrammetric and computer vision algorithms in order to automatically generate DSMs from multi-view SAR and optical high resolution satellite imagery, following an iterative and pyramidal workflow in order to refine a coarse DSM used as reference. Consequently, DATE is able to face both the issues of DSM generation and epipolar resampling of satellite imagery. The aim of this work is to evaluate DATE performance, by carrying out a sensitivity analysis based on the dense matching parameters. In particular, DATE implements the Semi-Global Block Matching (SGBM) algorithm, a modified version of Semi-Global Matching method: thus, the sensitivity analysis aims at assessing how SGBM parameters – namely, the difference between maximum and minimum disparity (ndisparities), the minimum disparity value (minimumDisp) and the matched block size (SADWindowSize) – affect the efficiency of the disparity map computation and the final DSM accuracy. The analysis focuses on the case study of Trento and of the Adige Valley, which was chosen due to its geomorphological heterogeneity and complexity, allowing to perform an accuracy evaluation on four tiles, characterized by specific roughness frequencies and morphologies (thus having different effects on disparity variations). Several practical indications on the optimal and critical parameter combinations were retrieved; in addition to this, this work highlighted the most influential parameters both in terms of accuracy (minimumDisp) and computation time (ndisparities), paving the way to further principal component analyses. Finally, the obtained results showed no clear relationship between the area morphology and the solution structure.