A calibration method of 2D LIDAR-Visual sensors embedded on an agricultural robot

Abstract

In the robotics field, mobile platforms typically combine data from multiple sensors and have been extensively researched. Estimating the homogeneous transformation between sensors is a critical step in associating their measurements data. In this paper, we present a calibration method of 2D LIDAR-Visual sensors embedded on an agricultural robot intended for strawberry greenhouse applications. The method is based on plan-to-plan homography approach and employs a set of point correspondences. The correspondences points are the measured intersection points of the laser range finder plane with the edges of the calibration pattern. The extracted points are used to estimate the homography matrix. The RANSAC algorithm is used with more points obtained from pairs scans/images by sensors to generate a robust homography. The estimated homography matrix is identified as the rigid transformation between the 2D LIDAR and the monocular camera. The results of the proposed method show that we can achieve accurate extrinsic parameters, avoiding the need for intrinsic parameters in advance or complex calibration objects. The system is composed of the Hokuyo URG-04LX-UG01 sensor and a monocular camera with an infrared filter. All of the software modules were created using Robot Operating System (ROS) and Python.