Accuracy assessment of Digital Surface Models generated by Semiglobal matching algorithm using Lidar data

Abstract

To measure the accuracy of Digital Surface Models (DSMs) generated by high resolution satellite images (HRSI) using semi-global matching algorithm in comparison with LIDAR DSMs, two different test areas with different properties and corresponding attributes and magnitudes of errors are considered. Error characteristics are classified as systematic and gross errors and significance of them to measure the accuracy of DSMs are evaluated. In this manner and to avoid the influence of outliers in accuracy assessment robust statistical methods are proposed. According to final values obtained for two test areas it can be concluded that the performance of DSMs generated by stereo matching for mountainous wooden areas in respect to the accuracy of LIDAR DSM are poor. In contrast, in case of residential urban areas the quality of the DSM generated by HRSI is able to follow the accuracy of LIDAR data.