GILV: an integrated GNSS/IMU positioning method enhanced by lane line and template matching VO

Ensuring precise and flexible positioning is a fundamental prerequisite for autonomous mobility of unmanned ground vehicles. In the context of Global Navigation Satellite System (GNSS) signals facing suppression, the traditional joint GNSS and IMU positioning is subject to positioning drift due to the cumulative error of the IMU. In this paper, we propose a low-cost integrated localization system called GILV that fuses inertial and visual data through an error state Kalman filter (ESKF) framework to achieve continuous localization under GNSS rejection. GILV utilizes a downward-facing camera to provide vehicle position constraints through a template matching algorithm that is used to smooth out IMU predictions and mitigate dispersion of localization errors. At the same time, the forward-facing camera detects lane lines and combines this information with high-precision maps to provide lateral position constraints without cumulative errors, ensuring that the Integrated Positioning System (IPS) accuracy is stabilized at the decimeter level without degradation due to cumulative IMU errors. The Integrated Positioning System (IPS) is realized on the ROS platform and has been verified through outdoor experiments. The results show that GILV consistently limits the position error to within 0.5 meters and maintains an average position error within 10 centimeters during a 100-second GNSS signal outage. This resilience greatly reduces the dependence of the positioning system on the quality of GNSS signals.