Detection of Stationary Foreground Objects Using Multiple Nonparametric Background-Foreground Models on a Finite...

Area of Science:

• Computer Vision
• Artificial Intelligence
• Pattern Recognition

Background:

• Surveillance systems increasingly require methods to detect stationary foreground objects like unattended packages or illegally parked vehicles.
• Existing algorithms struggle to reliably detect static objects in dynamic crowd scenarios over extended periods.

Purpose of the Study:

• To develop an efficient and high-quality strategy for detecting stationary foreground objects.
• To enable the detection of both completely and partially static objects in complex environments.

Main Methods:

• Utilized three parallel nonparametric detectors with varying absorption rates.
• Implemented a novel finite state machine for pixel classification.
• Classified pixels into background, moving foreground, stationary foreground, occluded stationary foreground, and uncovered background categories.

Main Results:

• The proposed strategy achieves high detection quality in challenging surveillance situations.
• It successfully detects both completely static and partially static foreground objects.
• The method demonstrates improvement over previous stationary object detection strategies.

Conclusions:

• The novel detection strategy offers an efficient and effective solution for identifying stationary foreground objects.
• This approach enhances the capabilities of surveillance systems in diverse and complex scenarios.
• The finite state machine framework provides robust pixel classification for improved object detection.