Research On Algorithm And Implementation Of The Vehicle Panoramic View System

In recent years,with the improvement of people's living standards,cars are gradually entering thousands of families,and the harm caused by traffic accidents has also gradually aroused people's attention.Visual blind area has always been an important factor in traffic accidents,especially in children's injuries.To solve the problem of large amount of computation for the existing splicing slot in the vehicle panoramic system,In this paper,the fish eye image correction algorithm is optimized based on the actual camera parameters,and the matching time of the system features is less by the specific calibration site.In this paper,the whole system is implemented in the embedded development board and tested under different lighting and scenes.In this paper,the development process of the vehicle panoramic view system is described in detail.The main work and innovation point are as follows:(1)The imaging features of the fish eye camera are analyzed and the acquisition and distortion correction of the fish eye image are completed.Using the fish eye camera,the video acquisition card completes the analog conversion,and the fish eye image is obtained.By analyzing the imaging characteristics of fisheye camera,the correspondence between ideal pixel and actual pixel is obtained.By comparing with the traditional correction algorithm,the spherical projection algorithm is applied to complete the distortion correction of fisheye image.(2)The acquisition of image overlook.According to the general formula of perspective transformation,we deduce the transformation matrix of perspective transformation.We know that only by extracting coordinates from four pairs of corresponding points,we can achieve perspective transformation.According to the characteristics of the calibration site of the system,4 groups of coordinate points which are easy to mark and extract are selected to realize the perspective transformation of the image.(3)The realization of image stitching.Two step preprocessing operations of median filtering and gray level correction based on Gamma are performed on the acquired overlook view,The system uses the template matching algorithm to register the image in the image registration stage,and the parameters of the whole memory registration are stored in the system.Image fusion uses an improved gradual approach,which can enable the image to simultaneously achieve seamless splicing in the direction of the X and Y axes.(4)Selecteing Tiny210 hardware platform to realize.Through the communication between the virtual machine,PC and the development board,the software developed inthe virtual machine QT is transplanted into the development board.The test and time-consuming statistics of the system are carried out under the conditions of strong illumination and weak illumination.After testing,this system can meet the needs of real time and can work in different conditions and situations.The design of this system can meet the actual requirements and is easy to operate,stable and real-time,and can be applied well.