Adaptive Algorithm to Improve the Image Quality of Vehicle License Plates based on Lighting Parameters

Abstract

This study aims to develop an Adaptive Algorithm to Improve the Image Quality of Vehicle Number Plates Based on Lighting Parameters. The type of research used by the Author is Research and Development. This research was conducted at the Computer Engineering Laboratory for six months. This research consists of several stages, from the potential and problem stages, needs analysis, literacy studies, building prototypes, system design, and system testing. The collected datasets were taken using smartphone cameras and webcams, with 207 image datasets divided into two categories: training data and validation. The training dataset of 207 objects was 100% successful. System testing was carried out in two conditions, namely during the day and at night, for each two-wheeled and four-wheeled vehicle object. The results of adaptive algorithm testing to improve the quality of vehicle license plate images based on the light parameter experienced a change in the average MSE and PSNR values between the original image and the quality-improved image, although not too much of a difference. Based on this, it can be interpreted that the adaptive algorithm can produce better photos than before.