Comparison of automatic target recognition system performance with full- and reduced-resolution correlators.

Area of Science:

• Optical Engineering
• Computer Vision
• Signal Processing

Background:

• Automatic Target Recognition (ATR) systems are crucial for identifying targets in various conditions.
• The efficiency of ATR systems can be influenced by the resolution of optical correlators and filter design methods.
• Evaluating the impact of reduced-resolution techniques is essential for optimizing ATR performance.

Purpose of the Study:

• To compare the performance of an ATR system using full- and reduced-resolution correlators.
• To evaluate reduced-resolution filter sets designed via multiresolution analysis (MA) and downsampling (DS) techniques.
• To determine the system-level applicability of subsystem-level findings regarding MA and DS filter designs.

Main Methods:

• Performance comparison of ATR systems with full-resolution and reduced-resolution correlators.
• Evaluation of reduced-resolution filter sets generated using multiresolution analysis (MA) and downsampling (DS).
• Analysis of receiver-operating-characteristic (ROC) curves to quantify performance differences.

Main Results:

• Subsystem-level findings on MA versus DS techniques did not directly translate to system-level ATR performance.
• Differences in target signatures between training and testing imagery had a greater impact on system performance than filter design.
• Reduced-resolution correlators exhibited minor, practically insignificant degradation in ROC curves compared to full-resolution correlators.

Conclusions:

• Target signature variability is a critical factor in ATR system performance, often overshadowing filter design choices.
• Reduced-resolution optical correlators are a viable option for ATR systems, offering potential benefits with minimal practical drawbacks.
• Further consideration of reduced-resolution optical correlators is warranted for advancing ATR technology.