Presenter : Gaurav Parida

Type: Poster

LiDAR also referred as Light Detection And Ranging, is a method that measures distance to a target by illuminating that target with a laser light. Lidar is popularly used to make high-resolution maps, with applications in geodesy, geomatics, geomorphology, forestry, laser altimetry and etc. With the current methods of object segmentation and extraction and classification being highly manual and tedious task. This work proposes a technique for object segmentation out of LiDAR data using an automatic process using a piecewise scanline based evaluation of the LiDAR data. Major segmentation tasks in the industry are done manually, where different views from different planes are considered to consider the scanline under investigation. On viewing the given scanline from different views, the pattern in the scanline is used to segment different kind of objects. The pattern among the scanline helps to differentiate them among objects of different kind, like the scanline of a tree will be highly fragmented with lack of any pattern and distributed randomly over the space. While the scanline of the edge of the building will show a clear uniform height and set of lidar points which can easily be segmented out as part of building. A geometric rule based approach was used initially to devise a way to segment buildings out of the lidar datasets. Synthetic datasets were created to verify the proof of concept. Initial experiments were done on simple shaped objects like cubes, cuboid. Later the approach was modified to also include the non-collinear objects like roofs which aren’t planar and buildings which don’t have simple geometry. Another method was devised to target the non-planar buildings like cylindrical and spherical buildings which have varying footprints over different axes and non-uniformity in their scanlines. The results show complete segmentation of buildings out from the dataset with promising results for real world data. The current segmentation approach can also be used to get 3D models of the buildings that are segmented out, with further uses in volumetric calculations and footprint extraction. Thus, the method can be used as a tool to get 3D models of the urban landscapes from the LiDAR data and help in urban planning and the smart cities endeavor in the recent times.

Keywords: lidar, building, segmentation, point cloud, 3D modelling