Prediction Model Optimization on Odd-Even License Plates Using YoloV8 Algorithm

Abstract

Traffic congestion in urban areas encourages the implementation of vehicle restriction policies based on license plate numbers, such as the odd-even system. Therefore, to support this policy, an accurate vehicle license plate detection system is needed and can work in real-time. The main challenge faced is how to develop an accurate and efficient detection model in recognizing license plates in various environmental conditions. The research method used is Knowledge Discovery in Databases (KDD) with five main stages, namely: data selection, preprocessing, transformation, data mining, and evaluation. This research method aims to develop and evaluate a vehicle license plate detection model based on the YOLOv8 algorithm, focusing on the classification of license plates into the "Odd" and "Even" categories. However, the dataset used was only obtained from the Roboflow platform and primary data in the parking environment, which was then processed through the cropping, resizing, and labeling stages using a format that was in accordance with YOLOv8's needs. The model was trained for 100 epoches with performance evaluation using precision, recall, F1-score, and Average Precision (mAP) metrics. The training results showed that the model achieved a precision of 0.879, a recall of 0.888, an mAP50 of 0.954, and an mAP50–95 of 0.830, with a fitness value of 0.843. In addition, the image resolution of 640x480 pixels results in the highest detection accuracy, which is 92% for odd plates and 85% for even plates. Tests were carried out on both images and videos, showing that the model was able to work in real-time with stable results. Based on these results, it can be concluded that YOLOv8 is effectively used to detect odd-even license plates with high accuracy. This research contributes to the development of intelligent systems based on computer vision to support efficient traffic management, especially in the implementation of odd-even policies in urban areas.