Hybrid deep learning framework for enhanced target tracking in video surveillance using CNN and DRNN-GWO
Thupakula Bhaskar, K. Sathish, D. Rosy Salomi Victoria, Er. Tatiraju. V. Rajani Kanth, Uma Patil, Naveen Mukkapati, Sanjeevkumar Angadi, P. Karthikeyan, R.G. Vidhya · International Journal of Basic and Applied Sciences · 2025
The growing demand for advanced security solutions has driven significant progress in video surveillance technologies in recent years. A critical component of modern surveillance systems is the ability to accurately track and monitor targets in dynamic environments. In this paper, we present a computer vision-based target-tracking system designed to enhance the efficiency of video surveillance operations. The proposed approach employs hybrid deep learning algorithms for the detection and tracking of targets within video frames. Initially, recorded video footage from surveillance cameras is input into the system, where each frame undergoes preprocessing to enhance quality. A Convolutional Neural Network (CNN) is then utilized to extract spatial features from the preprocessed frames, enabling the precise identification and localization of objects. The CNN also detects regions of interest and labels identified objects (e.g., persons, vehicles). We introduce a novel algorithm that combines the strengths of Deep Recurrent Neural Networks (DRNN) and Grey Wolf Optimization (GWO), referred to as DRNN-GWO. The DRNN module captures spatial and temporal dependencies within the frames to predict the future positions of tracked objects, while the GWO algorithm optimizes the hyperparameters of the DRNN to further enhance tracking performance. The proposed framework was implemented in Python. Experimental results demonstrated outstanding performance, achieving a target tracking accuracy of 99.12%, a recall of 98.75%, a precision of 99.27%, and an F-measure of 99%.