Penerapan Deep Learning untuk Deteksi Objek pada Sistem Kendaraan Otonom

Abstract

The advancement of autonomous vehicle technology represents a critical innovation in modern transportation, particularly in enabling accurate object detection within dynamic environments. This study aims to analyze the application of deep learning methods for object detection in autonomous vehicle systems to improve accuracy, efficiency, and driving safety. A convolutional neural network (CNN) approach was employed, utilizing the You Only Look Once (YOLO) algorithm for real-time object identification. The dataset consisted of diverse road images, including vehicles, pedestrians, traffic signs, and other obstacles. The research stages included data collection, image preprocessing, model training, system testing, and performance evaluation using accuracy, precision, recall, and frame per second (FPS) metrics. The results demonstrate that the YOLO-based model achieves high detection accuracy with fast processing speed, making it suitable for real-time autonomous systems. The model attained a precision of 94% and recall of 91%, effectively reducing detection errors that may impact driving safety. Additionally, the system maintained stable performance under varying lighting and environmental conditions. These findings confirm the significant potential of deep learning in enhancing autonomous vehicle capabilities, particularly in adaptive and automated object detection. Future research should incorporate more complex datasets and integrate additional sensors such as LiDAR and radar.