A Deep Learning-Based License Plate Recognition for Tricycles in General Santos City

Presented is an automatic license plate recognition (ALPR) system using a deep learning approach specifically tailored for jeep-type tricycles in General Santos City, a widely favored mode of public transportation. The recognition target in this context is the tricycle's body number, which corresponds to either the motorcycle license plate or its motorized vehicle (MV) file number. Our method leverages YOLOv4-darknet for detection and recognition. We utilized datasets extracted from recorded PTZ camera daytime traffic videos of a city's intersection for training, validation, and testing. Our results demonstrate an impressive accuracy of 98% in detecting tricycle body numbers while 86.88% in recognizing characters within these body numbers. Future directions include the extension of the proposed methodology to encompass the challenges posed by adverse weather (like rainy) conditions and/or low-light conditions (like nighttime with the possibility of considering the glare from headlights and taillights).