Study On Calibration Algorithms Of 3D Point Cloud Acquisition System For Intelligent Robots

With the continuously increasing applications of intelligent robots in different fields such as environment perception,virtual reality,medical diagnosis,and automatic driving,higher requirements are put forward for the environment perception and adaptability of intelligent robots.Systems with only a single or single-type sensor are difficult to meet the work requirements of robot in complex scenes.With the advantages of high reduction degree and high stability,three-dimensional(3D)point cloud acquisition systems are widely used in the field of intelligent robots.And accurate calibration of the 3D point cloud acquisition system is an important prerequisite to ensure that it can obtain accurate point cloud data.The traditional calibration methods of 3D point cloud acquisition systems are usually sensitive to noise and outliers,and the calibration accuracy and stability need to be further improved.Therefore,in-depth study on the calibration problems of 3D point cloud acquisition systems for intelligent robots,has great theoretical and realistic significance for improving the perception and adaptability of intelligent robots to 3D scenes.This dissertation deeply studies the calibration problems of 3D point cloud acquisition systems for intelligent robots,and the calibration parameters are solved by adopting the success-history based parameter adaptation for differential evolution(SHADE)algorithm.The main contents of the dissertation include the following three aspects:(1)The intelligent optimization algorithm SHADE,adopted in the calibration algorithms of this dissertation,is briefly described,including its basic idea,genetic background,evolution process and mathematical model.And the characteristics and optimization performance of the algorithm are analyzed as well.(2)A calibration algorithm for a 3D point cloud acquisition system composed of high-precision pan/tilt platform and Kinect sensor is presented,to estimate the measurement model parameters of the 3D point cloud acquisition system.By analyzing the measurement model of the 3D point cloud acquisition system,the space sphere is chosen as the calibration object.And the nonlinear objective function of the measurement model parameters of the system is established,according to the geometric constraint that the distance from each scanning point on the surface of the calibration sphere to the center of the sphere is equal toits radius.The objective function is optimized by adopting SHADE algorithm to estimate the measurement model parameters.The experimental results demonstrate that,the proposed algorithm has high accuracy and good consistency for the calibration of model parameter calibration.(3)A calibration algorithm for a 3D point cloud acquisition system composed of two Kinect sensors is given,to estimate the position and orientation relationship between the two Kinect sensors in the 3D point cloud acquisition system.The calibration algorithm chooses the space sphere as the calibration object.By placing the calibration sphere at different locations,the point cloud data of the calibration sphere at different locations is obtained by using the two Kinects,respectively.Based on the joint estimation of the coordinates of the sphere center at different locations in two Kinect local coordinate systems,the nonlinear objective function of the extrinsic parameters between the two Kinects is established.And the objective function is optimized by SHADE algorithm to estimate the extrinsic parameters.The experimental results demonstrate that,the proposed algorithm has high accuracy and good consistency for the estimation of the extrinsic parameters between the two Kinects.