You are here

Vehicle Mounted LIDAR Standoff Roadside Hazard Detection

Description:

OBJECTIVE: Design and develop the algorithms needed to detect emplaced roadside hazards under vegetative cover using LIDAR range and intensity data. DESCRIPTION: LIDAR sensors produce an active laser pulse in a linear scanning pattern sweeping across its field of view. These pulses reflect off of the various objects and surfaces present in a given scene. Once processed the LIDAR pulses form a"point cloud"representing the 3D orientations of objects and surfaces within the scene. LIDAR sensors are known to have the capability to detect objects and surfaces partially occluded by vegetation of varying density levels. This SBIR will develop algorithms to detect roadside hazards using this 3D LIDAR point cloud. PHASE I: The Phase I goal is to demonstrate techniques and concepts that could be used to detect targets beneath a layer of light to medium vegetative brush. To support development of these algorithms the contractor must collect LIDAR data on targets of interest under light to medium vegetative cover. The concepts and techniques to be leveraged in Phase II (e.g. range, intensity, texture, spin features, etc.) will be demonstrated to and validated by government Subject Matter Experts (SMEs). Fully autonomous algorithms are not required during this phase. PHASE II: The Phase II goal is to develop fully autonomous algorithms for detecting targets. The algorithms developed in this phase will be based off of the approaches demonstrated in Phase I. In addition to developing the Aided Target Recognition (AiTR) algorithm, a change detection algorithm will also be developed. The change detection algorithm will be developed to discard false-alarms created by objects previously rejected as targets by the operator. The change detection algorithm will be capable of notifying the operator to the presence of a new target. The Phase II final report must include a separate estimated probability of detection for targets with no, light, and medium vegetative cover as well as an overall false-alarm rate. The report must also include an analysis of the probability of detection and false-alarm rates and recommendations for future system and algorithm improvements. PHASE III: Take the algorithms from a TRL 5 to a mature state such that they can be fielded or sold commercially. This includes the system and algorithm improvements described in the Phase II report. These algorithms could then be fielded for use in detecting roadside hazards during military operations or for detecting roadside hazards or other structures of interest under vegetation overgrowth by city planners and utility/highway inspectors. REFERENCES: 1."An approach to target detection in forested scenes,"Christina Gronwall, Tomas Chevalier, Gustav Tolt, and Pierre Andersson, Proc. SPIE 6950, 69500S (2008), DOI:10.1117/12.777042 2."Rapid and scalable 3D object recognition using LIDAR data,"Bogdan C. Matei, Yi Tan, Harpreet S. Sawhney, and Rakesh Kumar, Proc. SPIE 6234, 623401 (2006), DOI:10.1117/12.666235 3."Using non-negative matrix factorization toward finding an informative basis in spin-image data,"Andrew J. Patterson, Nitesh N. Shah, and Donald E. Waagen, Proc. SPIE 6967, 696710 (2008), DOI:10.1117/12.776964
US Flag An Official Website of the United States Government