Design And Implementation Of Vehicle-mounted Panoramic Image System Based On Rapid Calibration

With the increasing popularity of automobiles,problems such as road congestion and insufficient parking spaces have become increasingly prominent.Driving scratches and parking accidents occur frequently.In order to solve this problem,scientific and technical workers in related industries have conducted in-depth and extensive research on the vehicle-mounted panoramic imaging system.The vehicle-mounted panoramic image system can present a complete visual image of the surrounding environment of the vehicle body,effectively eliminating the blind zone of the field of view during parking,and is of great significance to parking safety.This paper studies the key algorithms of the vehicle-mounted panoramic image system,completes the algorithm design and implementation of the system,and realizes the rapid automatic calibration of the system and the generation of panoramic images.This article focuses on the key algorithms in the system,study Zhang Zhengyou camera calibration method and distortion correction algorithm based on polynomial model,realize the calibration of the fisheye camera and correct the collected fisheye images;The corner detection algorithms is studied,a corner positioning method based on straight line detection is proposed,and designed a method for automatically extracting the corner points that are approaching gradually,to automatically extract the coordinates of the corner points required in the image during the calibration process,and improve the efficiency of corner point extraction;studied the DLT algorithm,designed a fast registration method for panoramic images based on marked points,directly transformed the perspective view into a top view and performed registration,using the corner points of multiple regions for direct linear transformation to avoid distortion of the stitched image;the weighted average image fusion algorithm was used to eliminate the stitching lines of the panoramic image,and the weighted panoramic image look-up table was designed to generate the real-time panoramic image.Based on the algorithm,this paper designs the software framework of the system,divided the system into two parts,calibration and real-time display.The hardware environment of the system consists of fisheye camera,video capture card and PC,and the software development environment integrating Open CV,CUDA and QT is built on PC.In the software implementation,the GPU is used to accelerate the process of fisheye image correction and overhead transformation,and the panoramic image is divided into multiple areas,and the multi-threading is used to quickly generate the panoramic image.This article also uses Qt to design the system's graphical user interface,which realizes the functions of one-key calibration and real-time image display;the system is tested under the construction of an experimental model.The results show that the system can quickly complete the calibration process and control the calibration time-consuming within 4s,the image quality and real-time performance meet the system requirements,and the practicability is strong.