Research On Joint Calibration Of Camera And Lidar

With the popularization of multi-source space sensor fusion technology in the field of robot and surveying and mapping,the fusion of camera and lidar has been widely studied.The camera mainly provides rich texture information,but lacks accurate spatial information.However,lidar can provide high precision 3d information,but it faces the problems of low resolution and difficult to extract semantic information.The information of the two sensors is complementary,and their fusion can enhance the overall perception ability.Therefore,how to meet the space requirements of camera and lidar fusion,and how to obtain the external parameters between them in a diversified way has important research value.In order to solve the problems of poor calibration accuracy,high influence of point cloud noise and low degree of automation in the joint calibration of camera and lidar,two mainstream calibration methods based on target and no target are studied respectively in this thesis,aiming at dealing with the diversified selection of different calibration scenes.The main research contents of this thesis are as follows:(1)Aiming at the low accuracy and poor robustness of external parameter calibration of cameras and lidars,an auxiliary calibration device with unique geometric characteristics is designed,and a two-stage calibration method from coarse to fine based on the device is proposed.The method consists of two stages: in the first stage,the 2D-3D corresponding lines distributed at the edge of the calibration device are extracted from the image and the laser point cloud to obtain the unique initial estimation of external parameters;In the second stage,the 2D-3D center point of the sphere is detected and fitted,and the initial external parameters are further optimized using the nonlinear optimization method.The proposed method provides two different features,which increases the stability of the calibration system against noise.Simulation experiments and real experiments show that this method can calculate high precision external parameters without prior initial values.Compared with the existing methods,this method has advantages in accuracy,stability and robustness.(2)To meet the need of more flexible calibration,a calibration method based on motion and edge matching is proposed.The calibration process of this method consists of two stages: in the first stage,according to the results of visual odometer and laser odometer,a hand eye calibration method is used to preliminarily estimate the spatial external parameters of camera and laser radar;In the second stage,using the assumption that the edge of the image corresponds to the discontinuous point cloud in depth,the calibration is converted into a 2D-2D registration problem to optimize the calibration results in the first stage.This calibration process can automatically calibrate in the environment without providing specific structured targets and initial external parameters.In the simulation and real experiments,the effectiveness of this method is proved,and it has a certain flexibility and adaptability in various unknown environments.