Research On Panoramic Imaging Of 360 Degree Parking Assist System

360 degree panoramic vehicle-parking is the key technology in Advanced Driver Assitance Systems.Compared with the monocular dashcam,360 degree panoramic vehicle-parking can get wider range fields,competitive in no dead corner view,which has been widely concerned and studied.Currently,360 degree panoramic parking system is still in the exploration stage,and most of the research is limited to two-dimensional panoramic view.It has not yet formed a universal algorithm for two-dimensional and three-dimensional panoramic parking.In term of these two cases,a panoramic view method has been studied and improved,including fisheye camera calibration for internal and external parameters,spherical surface reprojection and four fisheye camera curved surface reprojection and imaging methods.Based on the common character in panoramic imaging,fisheye camera model and calibration algorithms are studied.A fisheye camera model and calibration method was adopted for the both cases.The calibrated internal and external parameters can be applied in two-dimensional and three-dimensional panoramic parking,which achieved a unified method for the both cases.A method for sphere reprojection algorithm after calibration is proposed and verified the effectiveness by Open GL.By moving point of view far away from the centre after sphere texture mapping,it is approximately regarded that lines of sight parallel with each other when the distance is far greater than the radius of the spherical.The rendering effects verifiy the correctness of the calibration and feasibility of surface projection algorithm.The transformation relationship between multi coordinate system is studied,so as to generate unified panoramic image in the car coordinate system.The mapping from panoramic image to the car coordinate system according to the proportion is devided by four sub region corresponding to four fisheye cameras.After coordinate transformation and pixel searching within four fisheye cameras,the final index to the pixel information of the original fisheye image is obtained.On the base of two-dimensional panoramic imaging,a virtual spherical surface combined with plane is constructed.It is assumed that all the lines of light emit from the surface.The mapping from panoramic image to intercept surface indexs the 3D coordinate value,and the final index to the pixel information is obtained after coordinate transformation and pixel searching within four fisheye cameras afterwards.In this paper,the 360 degree panoramic parking system is modular,and the algorithm is validated by the calibration experiment and the panoramic synthesis experiment.Experiments show that calibration algorithm adopted can be used for two-dimensional and three-dimensional panoramic imaging.Panoramic imaging algorithm in this paper can be used in both two-dimensional and three-dimensional panorama,and the 3D imaging algorithm has the advantage of undistortion imaging in the near distance and wide view in the far distance,meeting the needs of 360 degree panoramic parking.