Target contour recovering for tracking people in complex environments

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Partial occlusion poses a significant challenge in visual tracking systems, leading to inaccurate contour changes of target objects.
• Existing occlusion detection methods often lack flexibility and are not solely based on contour information.

Purpose of the Study:

• To develop a novel method for detecting partial occlusions specifically on people's contours.
• To enable the recovery of occluded portions of people's contours within a visual tracking system.

Main Methods:

• A novel method is proposed that relies exclusively on contour information for occlusion detection and recovery.
• The method is designed for flexibility and extensibility to various applications.

Main Results:

• Experiments with synthetic images confirmed the method's accuracy in detecting partial occlusions.
• Real-world video sequence experiments demonstrated the method's effectiveness in recovering target contours.

Conclusions:

• The proposed contour-based method offers an accurate and flexible solution for detecting and recovering partially occluded people.
• This approach enhances the robustness of visual tracking systems dealing with occlusion challenges.