Zernike moment invariants based photo-identification using Fisher discriminant model

Area of Science:

• Computer Vision
• Marine Biology
• Biometrics

Background:

• Individual identification of marine mammals is crucial for population studies.
• Traditional methods rely on manual analysis of unique physical features, which is time-consuming.
• Gray whale flukes possess unique white patches (blotches) suitable for identification.

Purpose of the Study:

• To develop and evaluate an automated photo-identification algorithm for gray whales.
• To leverage Zernike moment invariants for robust feature extraction from fluke images.
• To reduce the manual workload in gray whale population monitoring.

Main Methods:

• Utilized Zernike moment invariants for pattern recognition.
• Employed Fisher discriminants to project invariants onto an optimal subspace.
• Applied live-wire edge detection and thresholding to isolate fluke blotches.
• Developed a feature vector based on projected invariants for image matching.

Main Results:

• The algorithm successfully identified individual gray whales using fluke blotch patterns.
• Feature vectors derived from Zernike moment invariants effectively represented unique whale identities.
• Demonstrated a significant reduction in manual search effort compared to traditional methods.

Conclusions:

• The proposed photo-identification algorithm is effective for gray whale individual identification.
• Zernike moment invariants and Fisher discriminants provide a powerful approach for biometric analysis in wildlife.
• This automated system offers a valuable tool for marine biologists in conservation and research efforts.