RECOGNITION OF VEHICLES WITH CLONED PLATES USING DEEP LEARNING

The strategy of using cloned plates is used by criminals to avoid the identification of stolen vehicles, where the plate of the stolen vehicle is replaced with that of a purchased vehicle having similar characteristics. This makes traditional methods for identifying stolen vehicles obsolete. This article describes a model based on Deep Learning and a real-time mobile application for recognizing vehicles with cloned plates. The system assessed video frames and was based on the recognition of vehicle plates and characteristics such as model, brand, and color. The model comprises four main processes: data capture, vehicle recognition, cloned-plate vehicle detection, and alert dispatch. Results from experiments conducted on 1,672 frames of moving vehicles on two high-speed roads in Lima, Peru, revealed the high sensitivity, specificity, and accuracy of the system, averaging 98.08%, 99.07%, and 98.62%, respectively. In terms of alert time (total time taken by the system from image capture to receiving an alert on the mobile device), the average for 824 records was 2.7 s, with a maximum and minimum of 1.3 and 4.0 s, respectively. These findings highlight the efficiency and effectiveness of the proposed application in identifying vehicles with cloned plates.