Three-Dimensional Reconstruction for Missile Recognition

Energid Technologies will create a model-based 3D threat-object identification system. The approach uses 3D models of the threat-object families to automatically construct identification algorithms for ladar and intensity sensor data. Single, dual, and multichannel imagery is supported. The system will identify an unknown object as one of up to 150 family types using three spatial stages: segmentation, initial identification, and refined identification. Spatial results will be combined using Multiple Hypothesis Tracking tailored to 3D reasoning. In addition to type, the system will identify the position, orientation, and geometry. It will execute quickly by performing calculations on low-cost graphics hardware. Energid Technologies’ partner, Spectral Sciences, Inc., will develop a high-fidelity plume database to support algorithm development and validate the results. The identification system will be delivered as a reusable C++ software toolkit, including approximately 500 C++ classes that can be subclassed to change behavior. Software modules will be configurable using the Extensible Markup Language (XML) to exchange run time for accuracy.