A novel framework for intelligent surveillance system based on abnormal human activity detection in academic...

Area of Science:

• Computer Vision
• Artificial Intelligence
• Security Systems

Background:

• Robust surveillance is critical in academic settings.
• Existing systems often lack real-time, integrated functionality for complex environments.

Purpose of the Study:

• To develop a novel, automatic, real-time video-based surveillance framework.
• To enable simultaneous tracking, semantic scene learning, and abnormality detection.
• To specifically address the needs of academic environments.

Main Methods:

• Utilized temporal differencing and Gaussian functions for motion object detection.
• Employed an OMEGA equation-based shape model for object filtering (human/non-human).
• Classified human activities into normal and abnormal using a support vector machine (SVM).
• Integrated content-based image retrieval (CBIR) for object recognition and identification.

Main Results:

• The system successfully performed real-time tracking, semantic scene learning, and abnormality detection.
• Demonstrated robust performance in an academic environment without human intervention.
• Achieved accurate classification of human activities as normal or abnormal.

Conclusions:

• The proposed framework offers an excellent, automated solution for academic surveillance.
• The integration of multiple AI techniques enhances the system's effectiveness and efficiency.
• This system provides a foundation for advanced, autonomous security in specialized environments.