Abstract
Detection and recognition of a licence plate is important when automatingweighbridge services. While many large databases are available for Latin andChinese alphanumeric license plates, data for Indian License Plates isinadequate. In particular, databases of Indian commercial truck license platesare inadequate, despite the fact that commercial vehicle license platerecognition plays a profound role in terms of logistics management andweighbridge automation. Moreover, models to recognise license plates are noteffectively able to generalise to such data due to its challenging nature, anddue to the abundant frequency of handwritten license plates, leading to theusage of diverse font styles. Thus, a database and effective models torecognise and detect such license plates are crucial. This paper provides adatabase on commercial truck license plates, and using state-of-the-art modelsin real-time object Detection: You Only Look Once Version 7, and SceneTextRecognition: Permuted Autoregressive Sequence Models, our method outperformsthe other cited references where the maximum accuracy obtained was less than90%, while we have achieved 95.82% accuracy in our algorithm implementation onthe presented challenging license plate dataset. Index Terms- Automatic LicensePlate Recognition, character recognition, license plate detection, visiontransformer.