An Imperious Verdict for The Recognition of Vehicles Number-Plate Using an Innovative Methodology

Abstract

Non-standardized number plates are prevalent in India’s present traffic patterns. Vehicle number fixed-plates should be perceived and systematised for a great number of reasons. Authorities have a tough time identifying and tracking down a specific car. In a growing nation like India, setting higher restraints on the effectiveness of any licence plate identification and recognition algorithm is impossible. The major goal of this study is to develop a method for detecting and identifying India’s transitional vehicle licence plates. Using single or multiple state-of-the-art automation, including machine-learning models, the character identification efficiency of drawn and printed plates in diverse styles and fonts improved dramatically. From a range of licence plate data, the proposed technique can develop rich feature representations. To find the licence plate in the appropriate location, the input image is first preprocessed to decrease noise and increase clarity, then separated into appropriate-sized grid cells. After the YOLOV5 has been trained, the licence plate characters should be appropriately divided. The three-character recognition systems OCR, LSTM, and STR are compared and contrasted in this study, with the conclusion that STR is the most accurate of the three. Finally, the data is post-processed, and the proposed model’s accuracy is tested against industry benchmarks. Vehicle monitoring, parking fee collection, detection of automobiles violating speed limits, reducing traffic accidents, and identifying unregistered vehicles are all expected to benefit from the proposed system. The results reveal that the suggested method achieves plate detection and character recognition accuracy levels above 90 percentage.