Silvilaser 2019 - Poster Presentations »
ALTIMETRIC EVALUATION OF GROUND MODELS OF GROUND MODELS FROM ALS OVER LOBLOLLY PINE STANDS
Filtering ground points is an important step for Airborne Laser Scanner (ALS) datasets. The accuracy of the DTM generated influences the precision of information extraction and is strongly linked to the filtering method used. This work aimed to compare different filtering methods in relation to a parametric model obtained with a total station, in 16 years old loblolly pine plantation with density of 1,275 trees.ha-1, in order to check the accuracy of the generated models. A total of 200 points were collected with total station for generating the parametric model by means of the open polygonal supported by two control points method. We selected three ground filtering methods were evaluated over the ALS dataset (Kraus & Pfeifer, Adaptive Triangular Irregular Network (ATIN) and Chen’s Morphological). The terrain points were modeled as a continuous surface through the Inverse Distance Weighted (IDW) interpolator and the evaluation was performed by visual analysis contour lines of 0.50 m equidistant and statistical analysis of error. Our results showed that all selected filters underestimated the ground height values. values of the heights. The differences of maximum and minimum height in the parametric model reached 14.96 m. The ATIN and Kraus & Pfeifer filters had a similar performance as the minimum and maximum heights with a difference of 14,02 m and the Morphological filter had a difference of 13,31 m. By analyzing the contour lines, the Kraus & Pfeifer and Morphological filter presented many noises, meaning presence of remaining vegetation in the DTM, in which the ATIN filter was the most efficient, with less noise of remaining objects. All the filtering methods homogenized areas with lowest elevations which were not detected during the filtering process. The errors for the highest height were 0.19% for the Kraus & Pfeifer and ATIN filter and 0.25% for Morphological filter. The lowest height was 0.09% for all tested methods. Based on the evaluation methods of this study, the ATIN filter obtained the best performance, mainly by providing the smaller noises. The existence of sub-forest as well the lack of silvicultural treatments hampered the penetration of many laser pulses into the ground, increasing the errors when compared to the parametric model. This study was sponsored by CAPES, CNPq and FAPESC. The ALS data were provided by the Sustainable Landscape Project.