Hybrid Camera-LiDAR Trilateration with Lens Distortion Correction

GNSS denied navigation is notoriously difficult for uncrewed vehicles in urban canyons, caves, or forested areas. Current leading methods for navigating in a constrained environment often require additional specific sensing hardware for a localization solution or only provide local frame navigation. Autonomous systems often include LiDAR and RGB cameras for mapping, sensing, or obstacle avoidance that can be leveraged for navigation by providing only or complimentary localization solutions in a global or local frame. Information from scanning LiDAR has been correlated with camera pixel coordinates and used to range unique visual landmarks that have known locations, using a pinhole camera. The present work included image distortion correction to improve landmark ranging error and expand visual field of view to utilize wide-angle lenses with known lens distortion. The expanded camera-LiDAR trilateration process was evaluated using colored spheres with known positions, 128 laser scanning LiDAR, USB camera, and Vicon motion capture for known ranges and positions. Landmark ranging error was evaluated with and without distortion correction at varying locations within an image while projecting LiDAR to camera points and identifying the point that maps nearest to a single landmark’s centroid within a frame. Position error at static locations were computed to identify the impact on localization accuracy when improving ranging error from lens correction. Ranging and localization errors decreased resulting in an average percent increase of accuracy of 90% for the tested configurations. The impact is that camera-LiDAR trilateration can be extended to more visual sensors with larger fields of view for landmark identification.