<![CDATA[This research discusses the development of a vehicle detection and tracking model for traffic using the YOLO v8 algorithm integrated with a Kalman Filter. The goal of this study is to improve vehicle tracking accuracy in traffic video recordings, with enhanced robustness against occlusions and disturbances. The methodology involves data collection from a vehicle image dataset available on Roboflow, followed by data processing into training, validation, and testing subsets. The model was trained over 30 epochs using Google Colaboratory, achieving a Mean Average Precision (mAP50) of 93%, a precision of 90%, and a recall of 89%. Testing was conducted on traffic footage from the City of Madiun, obtained from the Madiun City Government's CCTV website, demonstrating high detection and tracking performance. Model evaluation results indicate an accuracy of 93%, precision of 96%, recall of 85%, and an F1 score of 90%. The confusion matrix evaluation shows good performance in detecting vehicles, including cars, motorcycles, and trucks, making it a potentially effective solution for traffic monitoring challenges.]]>
Copyrights © 2025
