Related Experiment Video
Updated: Sep 21, 2025
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
9.1K
Real-Time Abnormal Object Detection for Video Surveillance in Smart Cities.
Palash Yuvraj Ingle1, Young-Gab Kim1
1Department of Computer and Information Security, and Convergence Engineering for Intelligent Drone, Sejong University, Seoul 05006, Korea.
Sensors (Basel, Switzerland)
|May 28, 2022
Summary
This study introduces a lightweight convolutional neural network for real-time detection of weapons like guns and knives in video surveillance. The method efficiently classifies and locates these objects on resource-constrained devices, improving security monitoring.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Machine Learning
Background:
- Video surveillance systems face challenges in real-time abnormal behavior monitoring, especially detecting weapons like guns and knives across multiple cameras.
- Resource-constrained devices limit the effectiveness of current object detection methods in surveillance.
- Manual monitoring of multiple camera feeds for threats is tedious and prone to human error.
Purpose of the Study:
- To develop a resource-constrained, lightweight subclass detection method for accurately identifying and classifying various types of guns and knives in real-time video surveillance.
- To improve the efficiency and effectiveness of object detection on devices with limited computational power.
- To enhance the capabilities of automated security systems in detecting dangerous objects.
Main Methods:
- A multiclass subclass detection convolutional neural network (CNN) was designed for classifying object frames into abnormal and normal categories.
- The proposed CNN method focuses on classifying, locating, and detecting different types of guns and knives.
- The model was evaluated on diverse datasets including ImageNet, IMFDB, Open Images, and Olmos, as well as multiview camera setups.
Main Results:
- The proposed method achieved high precision rates: 97.50% on ImageNet/IMFDB, 90.50% on Open Images, 93% on Olmos, and 90.7% on multiview cameras.
- Compared to state-of-the-art methods (84.21% for handguns, 90.20% for knives), the proposed approach demonstrates superior performance.
- The system achieved a satisfactory precision score of 85.5% for detection in multiview cameras, even on resource-constrained devices.
Conclusions:
- The developed lightweight CNN method effectively detects and classifies guns and knives in real-time surveillance scenarios.
- This approach offers an efficient solution for object detection on resource-constrained devices, enhancing security monitoring capabilities.
- The method shows significant promise for improving automated threat detection in complex, multiview video environments.
