In the context of Smart Cities, vehicle tracking plays a crucial role in various applications, including urban mobility optimization, traffic management, and fleet monitoring. Among these, its significance is particularly notable in enhancing public safety and enabling the identification and recovery of stolen or illegally taken vehicles. This paper presents a low-cost multi-tracking approach that integrates vehicle detection using smartphones positioned inside vehicles or at roadside checkpoints. The methodology encompasses vehicle detection and tracking, license plate extraction using YOLO, and character recognition with PaddleOCR. To improve accuracy, the model distinguishes between the Mercosur license plate standard and the discontinued Brazilian plate format, applying specific corrections and adjustments for each. The proposed approach is based on participatory sensing and can be employed both by law enforcement officers and by citizens wishing to contribute to public safety, thereby reducing reliance on state infrastructure such as fixed cameras, which often involve high costs and may be subject to downtime, particularly due to maintenance issues. Experimental results demonstrate an accuracy exceeding 87% in full plate detection—that is, the correct identification of all seven characters in either the Mercosur or the old Brazilian standard—even in scenarios where both the tracked vehicle and the device hosting the camera are in motion. Unlike existing approaches that depend on static cameras in controlled environments, the proposed method delivers reliable performance even under the more challenging condition of dual mobility.