Research On The Algorithm Of Fast Calibration For Vehicle-mounted Multi-camera Vision System

With the continuous increase of automobile usage,driving safety has received more and more attention from all sectors of society.The vehicle-mounted multi-camera vision system can provide the driver with a real-time image of the surrounding environment,which is widely used in many technologies to improve driving safety.The calibration of multi-camera systems is a key step in the application of vehicle-mounted multi-camera systems.Due to the structural problems of vehicle-mounted multi-camera vision systems,most multi-camera systems cannot guarantee an overlapping field of vision(FOV)between cameras.However,the existing calibration methods for multicamera systems without overlapping FOVs usually have problems such as a complex calibration process,low calibration efficiency,and high calibration costs.Therefore,how to quickly and efficiently calibrate a vehicle-mounted multi-camera vision system without an overlapping FOV becomes an urgent problem to be solved.Simultaneous localization and mapping(SLAM)can construct a three-dimensional(3D)scene map based on the images collected by the cameras during the motion process,which can fully save the feature information in the scene.A calibration method for multi-camera system based on SLAM is proposed in this thesis.The 3D map constructed by SLAM is used to establish the connection between multiple cameras without overlapping FOVs,and the internal and external parameters of the multi-camera system are estimated.The multi-camera system composed of in-vehicle cameras and out-of-vehicle cameras also cannot guarantee the overlapping FOVs between the cameras due to the occlusion problem of the vehicle.In this thesis,an improved auxiliary camera calibration method is proposed,which uses a combination of calibration target and auxiliary camera to quickly calibrate a multi-camera system including in-vehicle cameras and out-of-vehicle cameras.The main research work and innovation of this thesis are as follows:1.In order to improve the accuracy of the calibration method proposed in this thesis,the optimization method with constraints is proposed based on the traditional SLAM.In the process of SLAM back-end optimization,constraint optimization is added to make the 3D map constructed by SLAM more accurate.2.Based on SLAM with constrained optimization,an efficient and fast calibration method of multi-camera system is proposed in this thesis.This method does not require complex operations or expensive measurement tools to establish the connection between the cameras.It can complete the calibration of various distributed multi-camera systems only by the 2D-3D correspondence between the image and the 3D map.Therefore,the calibration method proposed in this thesis has the advantages of convenient operation,wide application scenarios,and high calibration efficiency.3.In order to calibrate the vehicle-mounted multi-camera system including invehicle cameras and out-of-vehicle cameras,an improved auxiliary camera calibration method is proposed in this thesis.This method only needs one image to complete the calibration.This method has high calibration efficiency and does not require complicated operations.