Traffic management is the process of balancing, routing, rerouting vehicles to improve the capacity, safety, security, and reliability of the transport system. This area of research has attracted attention from past authors, however, none of the traditional methods proposed concentrated on the rerouting verification process. Therefore, an effective traffic management system for re-routing verification with traffic congestion prediction by using Mobile Federated Learning-based Gaussian rootsig Ratchet Entropy-Deep Convolutional Neural Network (MFL-GRE-DCNN) and Bernoulli Pufferfish Optimization Algorithm (BPOA) is proposed in this framework. Initially, in the Vehicular Ad Hoc Network (VANET), the vehicles are registered and keys are generated. Then, the registered vehicles are initialized, followed by Rényi Tanimoto-KMeans (RT-KMeans)-based clustering. Afterward, the sensed data are balanced, and optimal routes are selected by using BPOA. Here, the traffic congestion is predicted for the optimal route by using the pre-trained MFL-GRE-DCNN. For training the prediction model, the highway traffic videos are preprocessed, the daytime and nighttime are determined from the preprocessed data. Then, the nighttime data are postprocessed, followed by raindrop detection and removal processes. Next, the vehicles are detected and tracked. From the features of the detected and tracked vehicles, congestion is detected. Based on the congestion results, the vehicles are re-routed, followed by a rerouting verification process. The experimental analysis show that the proposed model outperforms the other state-of-the-art models by attaining 99.78% detection accuracy.
