Single Shot Detector Based MobileNet for Automatic Detection and Recognition of License Plates

An Automatic motor vehicle Registration Number Recognition and Monitoring Solution is crucial for the shift towards technological advances in numerous everyday life features such as the management of parking spaces, tollbooth facilities, and roadway control situations. The proposed research concentrates on the databases for automobile registration numbers, regions, and forward images. The suggested approach concentrates on collecting deformed registration number plates caused by angled perspectives. The Single Shot Detector (SSD) is used to recognize as well as distinguish distorted registration plate numbers in frame by frame. Following license plate recognition, distortion on the registration plate is reduced using computerized processing methods. Additionally, it crops the text after retrieving them from the registration plate utilizing a sort outline technique. With the use of a lightweight in nature neural network such as MobileNets, the characters are subsequently turned into electronic characters through Optical Character Recognition (OCR). The suggested model also uses an evaluation measure, like stopping training when the value of the loss does not decrease after five epochs. As a result, it achieves 99.54% overall precision throughout the validation dataset and a precision of 98.26% after 23 training epochs. In order to prevent unauthorized entrance into susceptible sites, the suggested solution uses automated automotive vehicle identification validation through vehicle numbers.