DeepGuard: real-time threat recognition using Golden Jackal optimization with deep learning model.

Area of Science:

• Computer Science
• Artificial Intelligence
• Security Systems

Background:

• Violence recognition in surveillance is crucial for public safety and critical infrastructure security.
• Deep learning (DL) offers advanced capabilities for analyzing complex visual patterns in video footage.
• Traditional security measures can be enhanced with intelligent, real-time threat detection systems.

Purpose of the Study:

• To develop and evaluate the DeepGuard model for real-time violence and non-violence event recognition in surveillance videos.
• To leverage deep learning and optimization techniques for improved threat detection accuracy.
• To enhance proactive security measures in public spaces and critical infrastructures.

Main Methods:

• Utilized an improved ShuffleNetv2 model for extracting intrinsic features from surveillance images.
• Applied Golden Jackal Optimization (GJO) for optimal hyperparameter tuning of the ShuffleNetv2 model.
• Employed long short-term memory neural networks (LSTM-NNs) for the final violence detection classification.

Main Results:

• The DeepGuard model demonstrated optimal performance in violence detection tasks.
• Achieved high accuracy rates of 99.00% and 98.63% on benchmark datasets.
• Outperformed other existing techniques in distinct performance measures for threat recognition.

Conclusions:

• The DeepGuard model effectively identifies violence in surveillance videos using a combination of DL and GJO.
• The proposed system offers a sophisticated and intelligent approach to enhance classical security measures.
• Real-time threat recognition capabilities contribute to proactive safeguarding of public safety and critical infrastructures.