Related Experiment Video
Updated: Aug 1, 2026
05:41
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
Published on: February 6, 2020
Enhanced deep learning model for anomaly object detection and tracking from surveillance videos
Baliram Sambhaji Gayal1, Sandip Raosaheb Patil2, Dewanand Atmaram Meshram3
1Department of Electronics & Telecommunication Engineering, RMD Sinhgad School of Engineering, Pune, Maharashtra, 411058, India. ram.gayal@gmail.com.
Scientific Reports
|July 15, 2026
Summary
This study introduces an enhanced wolf optimization deep learning model for efficient video anomaly detection, significantly improving intruder identification in surveillance systems with high accuracy.
Area of Science:
- Computer Science
- Artificial Intelligence
- Machine Learning
Background:
- Video anomaly detection is vital for security but faces challenges like diverse contexts and limited data.
- Conventional deep learning methods often suffer from high false positives and occlusion issues.
Purpose of the Study:
- To propose an efficient anomaly object detection and tracking system to overcome limitations of existing methods.
- To enhance video surveillance security through improved intruder identification.
Main Methods:
- Utilized an enhanced wolf Crocuta optimization-based deep Bidirectional Long Short-Term Memory (EnWC-DBiLSTM) classifier.
- Implemented Timber Prairie Wolf Optimization (TPWO) for optimal keyframe selection.
- Employed enhanced Wolf Crocuta optimization (EnWC) to improve model convergence speed.
Main Results:
- Achieved 98.226% accuracy on the ShanghaiTech campus dataset.
- Obtained an equal error rate of 1.774%, sensitivity of 98.111%, and specificity of 99.551%.
Conclusions:
- The proposed EnWC-DBiLSTM model effectively detects anomalies in video surveillance.
- The system demonstrates superior performance in accuracy and efficiency compared to conventional techniques.
