Anomaly recognition in surveillance based on feature optimizer using deep learning.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Surveillance systems are crucial for public safety but face challenges with accuracy and robustness in detecting unusual incidents.
• Existing anomaly recognition methods often lack the precision required for effective real-world applications.

Purpose of the Study:

• To develop and evaluate an advanced deep learning framework for enhanced anomaly recognition in surveillance systems.
• To improve the accuracy and robustness of anomaly detection compared to current techniques.

Main Methods:

• Image preprocessing using histogram equalization.
• Feature extraction via two deep convolutional neural networks (DCNNs): Up-to-the-Minute-Net and Inception-Resnet-v2.
• Feature fusion and optimization using Dragonfly and Genetic Algorithm (GA) with 5- and 10-fold cross-validation.

Main Results:

• Achieved 99.9% accuracy using the GA optimizer with 2500 selected features during 5-fold cross-validation.
• Demonstrated substantial improvements in accuracy over existing anomaly recognition methods.
• Validated the effectiveness of the combined deep learning and feature optimization approach.

Conclusions:

• The proposed framework sets a new benchmark for anomaly recognition in surveillance systems.
• The innovative combination of deep learning models and advanced feature optimization offers significant potential for practical real-world applications.
• This research highlights the efficacy of sophisticated feature selection in enhancing deep learning model performance for critical safety systems.