Method of visual and infrared fusion for moving object detection

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Moving object detection is crucial for surveillance and autonomous systems.
• Fusing visual and infrared data offers complementary information for improved detection.
• Existing fusion methods often require prior strategies, limiting flexibility.

Purpose of the Study:

• To propose a novel method for moving object detection using joint low-rank and sparse decomposition.
• To develop a flexible framework for fusing visual and infrared video data without prior fusion strategies.
• To effectively utilize complementary information from both visual and infrared modalities for object detection.

Main Methods:

• A method based on low-rank and sparse decomposition is proposed.
• Visual and infrared image sequences are decomposed into background, non-object, and object terms.
• A joint minimization cost function involving nuclear norm, F norm, and l(1) norm is utilized.

Main Results:

• The proposed method successfully fuses information from visual and infrared video.
• Complementary information is naturally integrated during the object detection process.
• Experimental results demonstrate the effectiveness of the algorithm.

Conclusions:

• The developed method provides a flexible and effective framework for moving object detection.
• The joint decomposition approach naturally fuses multimodal information.
• The algorithm shows significant potential for applications requiring robust object detection.