A enhanced vehicle tracking and detection in multi-camera surveillance systems using advanced optical flow and deep learning techniques

Abstract

The increasing density of traffic has necessitated the deployment of advanced surveillance systems for effective traffic management. In this study, we present an enhanced vehicle tracking and detection method that leverages improved optical flow techniques and deep learning models. By employing the Farneb & auml;ck optical flow method, we achieve robust vehicle tracking across multiple surveillance cameras, even in challenging scenarios such as occlusions and illumination changes. Furthermore, we integrate Faster R-CNN, a state-of-the-art deep learning model, to enhance vehicle detection capabilities. Here we show that our proposed approach achieves an average endpoint error of 0.70 and a mean average precision of 88.91% when using the ResNet-101 backbone network. Our results demonstrate the efficacy of combining optical flow and deep learning for precise vehicle tracking and detection in complex traffic environments.