Research On SLAM Technology Of Stereo Garage Based On Laser And Vision Fusion

With the increase in car ownership,stereo garage has been proposed and used to solve the problem of parking difficulties in cities.The stereo garages can increase the number of parking spaces per unit area,while providing corresponding technology to assist in the transformation and optimization of traditional parking garages.AGV(Automatic Guided Vehicle)has the advantages of high automation and convenient management,and it is widely used in stereo garages to undertake vehicle transfer work.Simultaneous localization and mapping(SLAM)is the core technology for AGV to achieve automatic transfer.Accurate positioning can better assist AGV to achieve fixed point access to vehicles and improve the degree of garage intelligence.Therefore,this article conducts research on the simultaneous positioning and mapping of AGVs for vehicle transfer in stereo garages,develops a stereo garage SLAM system that integrates laser and vision,and promotes technological innovation in intelligent parking and static transportation.Firstly,aiming at the problem of accumulated errors in the use of the detection unit,the main sensor models and calibration methods required for SLAM systems are introduced.Establish a lidar measurement model based on likelihood field,and use hand-eye calibration method to calibrate the external parameters of the lidar;The measurement model of the camera is established based on the principle of pinhole imaging,and Zhang Zhengyou calibration method is used to deal with camera distortion and calibrate camera internal parameters;This paper introduces the capacitive accelerometer and gyroscope models based on the differential principle in the inertial measurement unit(IMU),and uses compensation algorithms to calibrate and compensate for IMU bias and random errors;Establish a vehicle transfer AGV kinematics model,combine the motor encoder to obtain the track calculation model of the wheel odometer,and use the machine adjustment software to calibrate the motor parameters.determine sensor static metrics and lay the foundation for experiments.Secondly,aiming at the problem that a single sensor has poor mapping effect in a stereoscopic garage,SLAM method of laser and vision fusion is established.The IMU and wheel odometer data are filtered by using the improved extended Kalman filter based on the time series as the innovation,and then the estimated position and attitude are obtained through the strapdown inertial navigation operation.Use grid probability map to store map information,select key frames for laser front-end matching and visual front-end matching,use feature fusion method to fuse laser data and visual information,use estimated pose as initial value for back-end optimization of pose,establish an objective function with minimum residual to achieve loopback detection,and use branch and bound to accelerate loopback operation.Reverse perspective and improved anti-noise edge detection based on gray morphology are used to extract parking line information.A SLAM framework that relies on lidar and cameras to obtain information about the garage environment was built.Finally,the software and hardware of the SLAM system are designed,and an experimental prototype vehicle is built.The functions of the laser vision fusion SLAM system are tested through posture estimation experiments,open loop and closed loop SLAM experiments,and multi mode parking line extraction experiments.The experimental results show that the system is stable and reliable,and meets the SLAM requirements of a stereo garage.