Class-associative multiple target detection by use of fringe-adjusted joint transform correlation

Area of Science:

• Computer Vision
• Pattern Recognition
• Signal Processing

Background:

• Object detection often struggles with classes containing dissimilar patterns.
• Traditional correlation filters may yield inconsistent detection peaks for varied objects within a class.

Purpose of the Study:

• To develop a robust class-associative correlation filter technique for detecting objects with diverse patterns.
• To enhance correlation performance for consistent and strong detection peaks across a class of objects.

Main Methods:

• Utilized a class-associative correlation filter approach.
• Employed the fringe-adjusted joint transform correlation algorithm.
• Incorporated an enhanced fringe-adjusted filter for improved multiple target detection.

Main Results:

• Demonstrated the ability to detect a class of objects with dissimilar patterns.
• Achieved strong and equal correlation peaks for each element within the target class.
• Verified the technique's feasibility through computer simulations.

Conclusions:

• The proposed class-associative correlation filter technique effectively addresses the challenge of detecting dissimilar object classes.
• The enhanced fringe-adjusted joint transform correlation offers superior performance for multi-target detection within a class.