Landscape modeling from aerial imagery digital surface models generated with the use of the dense matching algorithm

الملخص EN

DSM generation through exploitation of different digital image matching techniques has been a hot research area for the last two decades where different image matching algorithms have been developed and become available under different image processing software.

this motivated to increasing in researches for investigation of the qualities of those DSMs.

however, many questions regarding the qualities of the image matching DSMs and their capabilities in representing different types of landscapes have not yet been completely answered.

this research aimed at undertaking qualitative and quantitative investigations of the DSMs generated from aerial imagery stereo pairs with the use of the dense matching algorithm in addition to examining the efficiency of those DSMs in modelling of rural landscapes.

aerial imagery DSMs generated with the use of dense matching algorithm at two different resolutions for Vaihingen, Germany have been used in the study.

visual analysis shows clear representation of features in both DSMs of 0.09 and 0.14315 meter ground resolutions.

However, the differences between the two DSMs are hardly noticeable visually.

on the other hand, statistical analysis shows higher standard deviation of the DSM of 0.09 meters compared to that of the DSM of 0.14315 meter ground resolution which refers to better detailing of the rural landscape in the DSM of 0.09 meter ground resolution.

also, a standard deviation of the residual DSM of about 0.15 meters has been obtained.

additionally, the accuracy analysis with external ground truth checkout points shows that the residuals of the elevations extracted from 0.09 meter DSM are more accurate than those extracted from the 0.14315 meter DSM.

moreover, feature extraction from both DSMs shows that fine geometrical shapes of the multiple bay business building have been extracted, however, editing for removal of spikes from the raw DSMs could lead to extraction of better geometrical shapes that could be close to the real world.

finally, it can be recommended that the dense matching algorithm works well at ground elevation features such as roads compared to its performance at buildings and complicated structures.