Research On SLAM Technology Based On The Fusion Of Lidar And Depth Camera

Automobile plays an irreplaceable role in the production and life of today's society,but at the same time,the safety problems caused by it are also paid more and more attention by the society.The development of unmanned vehicle has become the trend of the times,and SLAM plays an indispensable role in the unmanned vehicle system.When SLAM uses low-cost 2D lidar or visual camera as sensor,there are some defects,such as single scanning range,large influence by light etc.,which lead to the environmental map constructed has some problems such as incomplete information and poor accuracy.The accuracy and integrity of environmental map will directly affect the unmanned vehicles subsequent navigation and obstacle avoidance and so on a series of work.In order to get the environment map with rich information and high accuracy at a low cost,this paper proposes a SLAM method which integrates 2D lidar data with depth camera data.The main contents of this paper are as follows:First,this paper analyzes the research status of SLAM and data fusion at home and abroad.The motion model,coordinate transformation and trajectory calculation of the mobile robot system are modeled and derived.In order to reduce the error caused by the internal parameters of the camera,the depth camera was calibrated.Analyzed the reason of motion distortion of lidar in motion,and the quadratic interpolation method is used to eliminate the motion distortion,and the simulation experiment is carried out.Secondly,based on laser SLAM,the algorithm flow of the SLAM system and the function of each link in the system are researched.In order to ensure the accuracy of front-end matching,the ICP algorithm is contrasted with its variant algorithm.In this paper IMSL-ICP algorithm is used for feature matching to ensure the accuracy of laser radar SLAM,branch and bound algorithm is used for loopback acceleration,and covering grid algorithm is used for map construction;The above algorithm is deduced in detail and the simulation experiments of front end matching and map building algorithm are carried out.Then,for the disadvantages of using a single sensor to build a map,this paper designs a scheme for multi-sensor data fusion,using Kalman filter and Bayesian estimation methods to fuse the laser data and depth data in the data layer and decision-making layer respectively,to obtain more information rich environment map.Based on the analysis of the application scope,advantages and disadvantages of the common fusion algorithms,the application process of Kalman filter and Bayesian estimation in this paper is deduced,and the simulation experiment of Kalman filter data fusion algorithm is carried out.Finally,the experimental platform is built based on the above multi-sensor data fusion scheme,the remote connection,coordinate conversion,sensor installation,multi sensor time synchronization and other contents are tested before experiment;In the same experimental environment,the established experimental platform is used to carry out many SLAM experiments,and the experimental results verify the feasibility of the data fusion scheme.