Automatic Vehicle Identification and Recognition using CNN Implemented on PYNQ Board

With the development of technology, the need for security is also increasing rapidly. One of the essential considerations is vehicle security in restricted areas. The need for detecting objects, people, or any suspicious vehicle has become crucial. Though there are ample methods through which this is possible, the need for an efficient design with low cost and high precision has always been the highest priority. Hence, the focus of this paper especially lies in detecting and extracting the vehicle number plate information thereby obtaining the vehicle owner's details. Such kind of security is essential for Gated Communities, Defense areas, and private security purposes. Most security systems are not automated. Therefore, to have a better automation security system, the proposed work is done with advanced technology. This research work presents an efficient system for license plate detection and extraction of characters of the vehicle number plate. It is implemented with a Convolutional Neural Network model. The logic is developed with python and prototype verification is done on the PYNQ-Z2 FPGA board. The results obtained using the CNN model perform well with an accuracy of 95%. In addition to this, the developed system is linked up with a web app that automatically recognizes the number plate and extracts the vehicle information from a localized database. As the proposed system is developed with a PYNQ-Z2 FPGA board it is power efficient.