The Research Of Fast Panoramic Image Mosaic For Parking Assistant System

With the increasing of the car ownership for urban residents,the contradiction between the more and more congested road traffic environment with the safe-driving becomes more and more prominent.The surround view parking assistant system is one of the most pop advanced driving assistant systems,which aims to provide the driver with the real-time surround top-view of the car to solve the problem of blind spots when parking the car,and thus enhancing the active safety of parking.However,the performance of real-time is of vital importance for surround view parking assistant system,and this paper researched the related panorama mosaic algorithms based on surround view parking assistant system,and proposed and selected the fast panoramic image mosaic scheme which satisfied the real-time performance of surround view parking assistant system.This paper mainly analyzed and studied the fast panoramic image mosaic technology of surround view parking assistant system,and mainly studied the following aspects:1.Studied the fish-eye camera imaging model,detailedly analyzed and compared nonlinear model camera several calibration algorithms,then selected Zhang zhengyou calibration algorithm which was of less low experimental requirement and high precision to calibrate the fish-eye camera.According to the distortion principle and characteristic of fish-eye camera,detailedly analyzed and compared several common distortion correction algorithms,finally selected the polynomial model correction algorithm to realize fish-eye images distortion correction;2.Analyzed two common top-view projection transformation algorithms,and selected the direct linear transformation algorithm,meanwhile,in order to satisfy the real-time of system,it realized the fast top-view projection transformation based on look-up mapping table.Then chose the image interpolation technology to correct the decimals coordinates pixels of geometric transformation image,and selected the quadratic interpolation method to complete the image interpolation correction;3.Analyzed several classic corner detection algorithms based on image feature registration algorithm,and it had been optimized for the typical Harris corner detection algorithm,which effectively reduced the time consumption of corner detection and extraction.Then the normalized cross correlation(NCC)algorithm was used to match the feature corners,and an improved RANSAC algorithm is proposed to eliminate the wrong matching corners,finally realized the fast registration of images;4.Studied several typical algorithms of pixel-level image fusion,and considered the computing ability of embedded platform,and chose the weighed image fusion algorithm based on gradually fading out to make image fusion.Finally used the gain compensation algorithm to balance the brightness of panoramic image;5.Proposed and introduced the system overall design that this paper adopt,according to the analysis of the key algorithms of the surround view parking system,this paper proposed the software algorithm overall design scheme,and completed the fast panoramic image mosaic based on look-up table method.Finally the performance was validated on the car.The experimental results show that this paper proposed a fast panoramic mosaic scheme for parking assistant system,which could meet the real-time performance of the system and the effect of stitching was good,and it was suitable for the practical application in engineering.