A Multi-pass Generation of DEM for Urban Planning

High quality DEM (Digital Elevation Model) is indispensable for a smart city nowadays. A large quantity of useful spatial information can be extracted from the big data acquired by surveying, imaging, and even mobile device with GPS system. This information is critical to many applications such as urban management, emergency event planning, and social location services. The extraction of non-ground features such as buildings, constructions and vegetation can be used to generate accurate boundary shape and three dimensional models. These outcomes can be utilized to provide useful information for the location service for mobile social networks. The DEM can be generated by surveying the area. LIDAR (Light Detection and Ranging) is a widely used remote sensing technology to survey the urban area. High resolution three dimensional data can be acquired from the LIDAR survey. How to separate the terrain surface and non-terrain features from three dimensional spatial data is the most challenging task for generating high quality DEM. In this paper, a multi-pass morphological filter is proposed to generate better DEMs on complex urban areas that have difficulties to use large filtering window sizes. Multi-pass filtering will use relatively small filtering window sizes in multiple rounds to avoid the cut-off problem commonly present when large window sizes are used in filtering. A priority boundary interpolation method is proposed and dedicated to interpolate the intermediate filtering result between each pass. A second pass of the morphological filter will use filtering window sizes in descending order to remove the remaining areas from large non-ground objects that cannot be removed with small window sizes. The filtering results demonstrate that the proposed method is able to effectively use relatively small filtering windows to filter large non-ground objects on complex terrains and avoid the cut-off problem.