Target-background contrast enhancement based on a multi-channel polarization distance model

Area of Science:

• * Biomimetic optics and vision science.
• * Polarization imaging and signal processing.
• * Computational neuroscience and artificial intelligence.

Background:

• * Stomatopods utilize polarized light for crucial behaviors like prey detection, habitat selection, and navigation.
• * Existing polarization models have limitations in target discrimination and noise resilience.
• * Biological polarization vision, particularly in mantis shrimps, offers a blueprint for advanced optical sensing.

Purpose of the Study:

• * To analyze and compare multi-channel polarization distance models for enhanced target detection.
• * To develop and simulate a polarization-vision neural network inspired by biological systems.
• * To evaluate the performance of a four-channel polarization model in contrast enhancement and noise reduction.

Main Methods:

• * Analysis of multi-channel polarization distance models based on the polarization distance concept.
• * Simulation and experimental validation of a four-channel polarization model.
• * Development of a cascaded polarization-vision neural network using four-orientation receptor input.

Main Results:

• * The four-channel polarization distance model significantly enhances target-background contrast compared to other models.
• * The proposed model demonstrates superior scene discrimination capability.
• * The model exhibits excellent robustness to noise, crucial for real-world applications.

Conclusions:

• * A four-channel polarization model, inspired by mantis shrimp vision, offers superior performance in target detection.
• * The developed polarization-vision neural network provides an effective method for contrast enhancement.
• * This approach has significant potential for applications in camouflage removal and target detection systems.