Automatic classification of criminal activities for security surveillance by keyframes detection and advanced...

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Video surveillance systems are rapidly expanding, increasing the need for automated abnormal behavior detection.
• Traditional human-based surveillance is inefficient and prone to errors.
• Deep Learning (DL) offers advanced capabilities for analyzing complex video data and detecting anomalies.

Purpose of the Study:

• To develop and evaluate an automated abnormal behavior detection system using advanced Deep Learning techniques.
• To address the limitations of traditional surveillance methods.
• To improve the efficiency and accuracy of anomaly detection in video feeds.

Main Methods:

• Utilized the Inceptionv4 model, an enhanced Convolutional Neural Network (CNN), for abnormal behavior detection.
• Implemented a novel key frame extraction method to reduce data size and improve processing efficiency.
• Applied the model to standard datasets for empirical analysis.

Main Results:

• The Inceptionv4 model achieved a high accuracy of 95% in abnormal behavior detection.
• The proposed key frame extraction method contributed to efficient data processing.
• The model demonstrated superior performance compared to other standard deep learning algorithms and pre-trained models.

Conclusions:

• Deep Learning, particularly the Inceptionv4 model with key frame extraction, provides a highly effective solution for automated abnormal behavior detection.
• The approach significantly enhances surveillance system efficiency and accuracy.
• This research contributes to advancing intelligent video surveillance capabilities.