<![CDATA[Effective law enforcement, including the use of the ANPR (Automatic Number Plate Recognition) system, is essential for reducing the number of road traffic accidents. ANPR involves plate localization, character segmentation, and recognition to build a minimum system. This study aims to improve a character segmentation method using a detection approach to address issues like noisy or modified plates. We propose an adaptive improvement on an established sliding window technique, by integrating a CNN (Convolutional Neural Network) for bounding box coordinate adjustment to handle various plate conditions. The proposed method was tested on 280 license plate images and improved the average IoU (Intersection over Union) from 0.4811 to 0.8980. Hence, the recall and precision of the model could be improved to increase any character recognition performance.]]>
Copyrights © 2024
