Home About Login Current Archives Announcements Editorial Board
Submit Now For Authors Call for Submissions Statistics Contact
Home > Archives > Volume 20, No 11 (2022) > Article

DOI: 10.14704/nq.2022.20.11.NQ66108

Random selected leader based optimization with Deep Learning Enabled Object Detection on Video Surveillance System

S. PRABU J. M. GNANASEKAR

Abstract

Video surveillance systems are installed at various places namely malls,airports, and train stationsfor security and monitoring purposes. But it becomes laborious to search and retrieve individuals in multicamera surveillance systems, particularly with appearance changes and cluttered backgrounds among many cameras. Object detection can be defined as computer technology based on image processing and computer vision (CV) focuses on determining and detectingseveral target objects from video data or still images. Thetracking and identification of objects and the analysis of behaviour in intelligent video surveillance were till now affected by practical complexities. New advancements in CV, and specifically with deep learning (DL)approachespresent novel outlooks for this system, rising its abilities and originating novel directions of research in this domain. This study develops a Random selected leader based optimization with Deep Learning Enabled Object Detection (RSLO-DLOD) method for Video Surveillance systems. The presented RSLO-DLOD algorithm concentrates on the accurate and rapid identification of objects in surveillance videos. In the presented RSLO-DLOD technique, YOLO-GD (Ghost Net and Depthwise convolution) model is applied for object detection process with EfficientNet as baseline model. Next, the RSLO-DLOD technique employed fuzzy neural network for object classification process. For enhancing the object detection efficiency of the RSLO-DLOD algorithm, the RSLO algorithm is utilized for the hyperparameter tuning procedure. For demonstrating the enhanced object detection outcomes of the RSLO-DLOD technique, a comprehensive set of simulations were take place on open access dataset. The simulation analysis emphasized the enhancements of the RSLO-DLOD approach over other recent approaches.

Keywords

Object detection; Video surveillance; Deep learning; YOLO network; Hyperparameter tuning

Full Text

PDF

References

?>