Ανάπτυξη ενός ολοκληρωμένου συστήματος αυτοματοποιημένης αναγνώρισης πινακίδων κυκλοφορίας οχημάτων ως ενσωματωμένη λύση μηχανικής μάθησης σε εφαρμογή Android
Automatic license plate recognition system development as an Android on-device ML application
KeywordsALPR ; Java ; Kotlin ; On-device ML ; Google maps API ; RESTfull Web Services ; APIs ; NoSQL ; MongoDB ; Java Spring Boot Framework ; Android Studio ; IntelliJ ; TensorFlow ; DBSCAN ; QGIS ; Transfer learning
In the context of this master's thesis, an integrated system of automated license plate recognition (ALPR) has been developed and integrated into an application for the Android operating system. Through the application, users can declare the registration numbers of vehicles they are interested in and are looking for any reason (theft, hit and run after car accident, etc.), to be notified if any of them are detected, to scan in real-time the license plates of passing vehicles, as well as refer back to their scan logs, using spatio-temporal criteria. In addition, capability of processing and analyzing the collected data is foreseen, aiming drawing conclusions, which will allow the more efficient allocation of available resources (human resources and devices), and the prediction of more likely positions of future events. The application was developed in Java and Kotlin programming languages, while at the same time, technologies such as Google's on-device ML (machine learning) were used to detect and recognize license plates. Finally, the application is supported by a NoSQL database (MongoDB) and appropriate Web Services, developed in Java Spring Boot environment.