Demonstration of Westinghouse Automatic Cueing Techniques Using NVL Imagery
Abstract
This report describes the work and results of a study to establish the performance of existing digital image processing techniques on FLIR imagery supplied by NVL. The image processor would form the basis for an automatic target cueing system to assist the human operator of a sensor system. The study consisted of a statistical test, performed by computer simulation, including training and test phases. The target classes included truck, tank, and APC. Initial detection of targets scored in the 90% range. Depending upon image quality, the classification performance was in the 60% to 80% range. Using the same sensitivity setting, the false alarm rate was 20%. The exact setting, trading false alarm rate for detection rate would depend upon the mission requirements. It was noted that the number of samples was very limited, in view of the number of features used. Future efforts might include a larger data base. It was also suggested that the design of a compact automatic cueing system breadboard be started to keep pace with sensor hardware development.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 23, 1976
- Accession Number
- ADA036538
Entities
People
- G. E. Tisdale
- T. P. Truitt