Approach to Enhancing Target Discrimination via 3D Visualization without 3D Glasses
Agency / Branch:
DOD / MDA
There are many factors that render automatic target recognition challenging. These include cluttered background, adverse weather and acquisition conditions, spatially closed targets, spectrally-matched decoys, and the limits on sensor resolution etc. One solution is to have a 3D display, which will help discriminate the targets. However, a major challenge in rectifying the sensor streams, i.e. to relate one image to the other on a pixel-by-pixel basis, is that the sensors are not calibrated, thus simply using standard stereoscopic vision algorithms will not generate accurate stereo images. Here we propose an algorithm for stereo image creation in real-time, and has very high accuracy due to the use of state-of-the-art algorithms developed recently by this team. The first step is that we apply Gabor filter to systematically select pixel points as features. This feature correspondence and tracking algorithm will minimize the impact of errors caused by uncalibrated cameras and cluttered background. Then an efficient algorithm for tracking the feature points in the subsequent frames is used. Thus potentially we can have a real-time algorithm that can run with stringent constraint of computational resources. Finally, we have an efficient algorithm to create a stereo image.
Small Business Information at Submission:
Director of Research & Development
Contract & Proposal Manager
Research Institution Information:
Intelligent Automation, Inc.
15400 Calhoun Drive, Suite 400 Rockville, MD 20855
Number of Employees:
ARIZONA STATE UNIV.
Depart of Computer Science , Ira A. Fulton School
Tempe, AZ 85287
Nonprofit college or university