In this paper, we propose a novel methodology for vehicular image localization, by incorporating the surveillance image object identification, using a local gradient model, and vehicle localization using the time of action. The aerial images of different traffic densities are obtained using the Histograms of Oriented Gradients (HOG) Descriptor. These features are acquired simply based on locations, angles, positions, and height of cameras set on the junction board. The localization of vehicular image is obtained based on the different times of action of the vehicles under consideration. Support Vector Machines (SVM) classifier, as well as Particle Swarm Optimization (PSO), is also proposed in this work. Different experimental analyses are also performed to calculate the efficiency of optimization methods in the new proposed system. Outcomes from experimentations reveal the effectiveness of the classification precision, recall, and F measure.