Adaptive Detection and Classification of Micro-Terrain Features from LiDAR Data using Sparse Decomposition Based Analytics

Important information lies at the processing resolution limits of current LiDAR systems. Subtle discontinuities such as runoff channels, fault lines, and ridges can reveal important local terrain modeling data as well as provide an indication of factors affecting small unit movement. Unpaved roads, trails, or wheel-ruts can provide critical information about preferred movement paths. Man-made features and disturbed earth can indicate human occupation and activity. Regional textures such as boulder fields, talus slopes, areas of downed trees, or understory foliage are important to efficient route planning, to OACOK analysis, and to tasks such as landing zone (LZ) planning. The challenge has two components: (1) Providing means to automatically classify local terrain regions by texture and other user-selected characteristics of interest; and (2) providing means to find and map connected elevation discontinuities within the data. We will demonstrate highly effective algorithms based on Adaptive Sparse Decomposition techniques to accomplish each of these tasks. In Phase 2 we will enhance these tools, and integrated them with GIS software to afford analysts new and powerful capabilities for exploiting terrain information at the resolution limits of LiDAR point clouds. Our multi-expert team has prior experience working with Army ERDC TEC data, including Buckeye.