Surround-view for Articulated Vehicles: Sensor fusion and image processing for articulated vehicles in a surround-view system

Abstract

Bird’s-eye-view surround-view systems are widely used for driver assistance in rigid vehicles operating on flat terrain but are poorly suited for articulated vehicles and uneven terrain. This thesis investigates the feasibility of extending surround-view systems to articulated vehicles in uneven terrain using sensor fusion and image processing. A surround-view pipeline is developed that combines camera-based bird’s-eye-view generation with inertial measurement data from Inertial Measurement Units (IMUs). A key contribution is the introduction of independently tilted ground planes for each vehicle frame, enabling camera projections to adapt dynamically to vehicle articulation and terrain orientation. Inertial Measurement Unit-camera fusion is applied for orientation estimation, while image undistortion and adaptive stitching are used to improve visual consistency. The approach is evaluated in simulation and through real-world experiments using relatively low-cost cameras and commercially available Inertial Measurement Units. The results demonstrate improved geometric consistency on uneven terrain compared to conventional flat-plane bird’s-eye-view methods and confirm the feasibility of the system as a proof of concept. The work prioritizes computational efficiency and comprehensibility over dense three-dimensional reconstruction, indicating that IMU-camera fusion with per-frame tilted ground planes is a viable and efficient approach for surround-view systems for articulated vehicles.