Extrinsic Calibration of LiDAR and Camera using Multiple Traffic Sings

Abstract

For fusing accurate multimodal measurements, an external calibration of various sensors is required. Extrinsic transformations, on the other hand, can drift slowly when driving or maintaining, reducing the precision of the sensory system. We propose an extrinsic calibration tool for a camera and a LiDAR sensor using multiple traffic signs on a moving vehicle. We demonstrate that traffic signs can be detected accurately using a camera and a LiDAR, thus contributing to the extrinsic calibration. By using a normal vector of segmented signs from the image and point cloud, we are able to estimate an accurate extrinsic matrix of two sensors. To demonstrate the result of our method, we use our own experimental dataset. This work can provide the essential performance of using existing known objects on the street in order to get a precise measurement of extrinsic calibration of LiDAR and camera sensors.