A multi features based background modelling approach for moving object detection.

Area of Science:

• Computer Vision
• Image Processing
• Video Analysis

Background:

• Background subtraction is crucial for video analysis but remains challenging.
• Previous work utilized hybrid models with oriented patterns.
• Statistical methods often lack robustness.

Purpose of the Study:

• To propose a novel and efficient background subtraction method.
• To improve upon existing hybrid models by incorporating new features.
• To reduce computational complexity while maintaining accuracy.

Main Methods:

• Eliminated Gray-Level Co-occurrence Matrix (GLCM) features.
• Incorporated improved local Zernike moments and color intensity components.
• Combined these with orientation-based features from Angle Co-occurrence Matrices (ACMs).
• Replaced Mahalanobis distance with Canberra distance for pixel categorization.

Main Results:

• The proposed method effectively models background using combined features.
• Canberra distance significantly reduces computational complexity by avoiding covariance matrix calculation.
• Comparative analyses demonstrate superior performance over competing methods on diverse video sequences.

Conclusions:

• The novel approach offers an effective and computationally efficient solution for background subtraction.
• The integration of Zernike moments, color components, and ACMs enhances feature representation.
• The use of Canberra distance streamlines the process without compromising accuracy.