Automatic Compilation of 3D Road Features Using LIDAR and Multi-spectral Source Data

This Phase II proposal will address source data preparation for modeling and simulation and cartographic feature applications using semi-automated road network extraction technology. The primary goal is to use airborne LIDAR data to generate 3D road network geometry augmented with physical attribution. We will utilize RoadMAP, our road network extraction system, which currently performs local-area processing to extrapolate from a road starting point to delineate complete road features. Under this SBIR, we will incorporate both co-registered airborne LIDAR data to directly determine road height, pitch, and slope, and co-registered multi-spectral source (MSS) data to estimate surface material properties. By pre-processing LIDAR/MSS imagery to create a "traffickability map", automated road starting points will be detected that contain likely avenues for roads. Our Phase I results demonstrate that creating the traffickability map successfully classifies road-like features regardless of whether they are in flat, sloped, or undulating terrain. The map is created in world coordinates so that other geo-referenced data can easily be correlated and incorporated into RoadMAP's decision making modules. Additionally, a machine learning module will be integrated into RoadMAP to improve the level of automation within the interactive road extraction process.