The Application of Wavelets, Fractal Geometry, and Statistics to Automatic Target Recognition for Ladar

Period of Performance: 02/09/1998 - 02/08/2000


Phase 2 SBIR

Recipient Firm

Orincon Corp.
4770 Eastgate Mall
San Diego, CA 92121
Principal Investigator

Research Topics


Ladar systems with automatic target recognition (ATR) capability will provide the autonomous precision guidance required for "smart" conventional munitions envisioned for the Low Cost Autonomous Attack System (LOCAAS). Current algorithms perform ATR using three-dimensional (3-D) range data and usually ignore texture in ladar images. Since texture is an important feature in human target discrimination, ORINCON undertook a Phase I effort to exploit wavelet, fractal, and statistical measures of texture for ladar ATR and demonstrated that texture and neural networks can be used to detect and segment targets in ladar intensity images.Based on these results, ORINCON proposes a Phase II effort to implement a full-scale, real-time demonstration system for ladar ATR using texture and multiscale edge features as inputs to neural networks. The system will accurately detect, segment, and classify targets in ladar images based on features extracted from both intensity and range data, and enhance the performance of other ATR systems through data fusion. Phase II efforts will focus primarily on feature extraction, neural network design, data fusion, ladar image processing, and real-time implementation.