A real-time detection and positioning method for small and weak targets using a 1D morphology-based approach in 2D...

Area of Science:

• Computer Vision
• Image Processing
• Real-time Systems

Background:

• Traditional small target detection methods struggle with real-time processing demands.
• Existing approaches often require complex hardware or significant latency.

Purpose of the Study:

• To develop a real-time small and weak target detection method with superior performance.
• To enable hardware implementation for ultra-low latency image processing.

Main Methods:

• Processing two-dimensional (2D) images using one-dimensional (1D) structuring elements in a morphology-based approach.
• Implementing a parallel image readout and processing structure for nanosecond latency.
• Utilizing a row-by-row structure with electrical rolling shutter for hyper-frame resolution.

Main Results:

• Successful detection of targets under various interferences with 0.1-pixel accuracy (1σ) in worst-case conditions.
• Achieved centroiding precision better than 0.03 pixels (1σ) in static tests.
• Demonstrated superior real-time capability compared to all other methods.

Conclusions:

• The proposed method offers high robustness and accuracy for real-time small target detection.
• The technique is suitable for applications requiring ultra-high processing speeds, including medical imaging and surveillance.
• This morphology-based approach enables efficient hardware implementation for real-time image analysis.