利用參考模型校正車用鏡頭之影像轉換方法與相關裝置之應用

Abstract

本文主要提出一種車用镜頭的自動化校正與應用其的影像轉換方法與装置。該方法包括：校正参考圖像的影像提取。辨識該校正参考圖像的多個第一與第二特徵圖像。求出这些第一與第二特徵圖像在該第一視角座標系的各别座標，以及這些第一與第二特徵圖像在該第二視角座標系的各别座標，以求出该第一與该第二視角座標系间的座標對應關係。根據該座標對應關係，將输入影像轉换成輸出影像。 現已有許多技術來輔助行車安全。一般來說，這些車輛安全輔助系統(全周監視系統、前方防碰撞警示系統、後方安全輔助系統、側方盲點警示系統等)通常都具有車用鏡頭，其用以提取車外影像。在出廠之前，安裝于車輛的車用鏡頭得要先經過校正。 以目前來看，傳統的車用鏡頭人機校正方法需仰賴校正人員手動微調，使得影像中的特徵圖像能符合預設條件，以通過人機操作將影像調整至正確位置。之後，系統計算鏡頭外部參數(extrinsic parameters)與內部參數(intrinsic parameters)以進行座標轉換。本論文提出一種車用鏡頭校正參考圖像、車用鏡頭校正參考圖像的設置方法、及應用其的影像轉換方法與裝置，自動化地校正車用鏡頭。

The purpose of this thesis is to apply the method that can automatically correct the car camera by using reference image. The method includes : extract image by correcting reference image . Identify multiple first and second characteristic image in reference image, and get the coordinates that is first perspective and secondperspective, and analysis the relationship between them . According to the relationship , transform the input image to output image.

In today , have a lot of technology to support car security , in general , these car secure support system (multi view camera system , front crash alarm system , back secure support system , side blind alarm system …) should have a car camera , use the camera to get the car outside image, before the car sale , the camera installed in car should be correct first .

At present , traditional man-machine car camera correct method should depends on manual adjustment , ensure the image characteristic image can fit the default condition by using the man-machine operate to adjust to the right position. And the system compute the camera extrinsic parameters and intrinsic parameters to transform the coordinates. This thesis apply the method that automatically correct car camera by using reference image.