ABSTRACT

ABSTRACT: Lidar (Light Detection And Ranging) is a novel technology that can be used for fast acquisition of digital surface models and generation of DEMs as well as for building extraction and reconstruction. In this paper, we focus on the application of Lidar data for extracting grid structured streets in a dense urban environment. Due to occlusions and complicated image patterns of the streets, from an image it is a huge obstacle to detect and extract streets reliably and accurately, while from segmented Lidar data, the streets are demonstrated as grid structures clearly. First, we classify the study area by using ranging and intensity information obtained from Lidar data. A binary image is formed by the segmented Lidar data. The ribbon open areas are candidate street segments. A two step procedure is applied to find the grid streets automatically. The first step is to detect all possible straight streets using the Hough transform, by which the street grid and street width can be detected and estimated roughly. In the second step, verification processing is deployed to verify each street and form the final grid structuralized streets. We tested our algorithm on a dataset of the Toronto downtown area. The extraction result demonstrates the potential power of using Lidar data for road extraction in dense urban areas.