Improvement of Lane-Level Lateral Position Estimation for Ego Vehicle From Front Camera Data

In the rapidly evolving landscape of automotive technology, the concept of connected car services has emerged as a transformative force, reshaping the way of interaction and experience with vehicles. Since the services utilize vehicle-generated data, to provide effective services, it is important to provide data accurately. To estimate accurate vehicle position, vehicle-captured image data is often used to supplement GNSS data. The position of vehicles in image data is estimated by detecting white lines on the road and predicting the ego lane. However, obstacles around vehicles can cause noise that reduces estimation accuracy. This study proposes a method to improve the accuracy of the estimated ego lane position by using a combination of Hampel filter and low-pass filter. Evaluation results show that the estimation accuracy of the ego car's lateral position was improved by up to 6% compared to the conventional method. Since the algorithm is lightweight, it can be effectively implemented into systems with limited computational resources.