基於環車鳥瞰影像之自動停車系統
No Thumbnail Available
Date
2010
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
車輛駕駛輔助系統的主要目的為提升行車駕駛者的安全、效率與舒適。本研究在車輛的周圍裝設攝影機，建立一環車全景鳥瞰影像，如此可以解決車輛四周的盲點區域問題，而本研究利用此種影像從事自動停車的目的。
本研究使用四個魚眼攝影機，將其架設在車型機器人AmigoBot四周，擷取其四周的影像，並分別對其影像各自執行攝影機的暗角效應消除(vignetting elimination)、魚眼影像的反扭曲(distortion recovery)、俯視影像轉換(top-view image transformation)，最後再將俯視影像做影像接合(image registration)與色彩混合(color blending)，建立車輛周圍的全景鳥瞰影像。本研究並紀錄了在各影像轉換階段中影像像素位置的對應關係，將其整合成一轉換表，如此將來可以利用查表法(look-up table)來加速鳥瞰影像建立的速度，使之能到達即時轉換的目的。當鳥瞰影像建立出來後，本實驗假設已偵測到車輛周圍停車格位的位置，再實地考察真實車輛的運動行為，使AmigoBot模擬一般車輛的運動行為。再依據台灣現有的三種停車格位(直角停車格位、斜角停車格位、路邊停車格位)，各自設計停車的策略，使車型機器人能依策略達到自動停車的目的。實驗顯示，本研究可達成即時建立車輛周圍的鳥瞰影像與各種不同的自動停車策略。
This article presents an automatic parking system based on a bird`s-eye view image around a vehicle. There are four fisheye cameras mounted around the vehicle for capturing its surrounding images. Cameras are calibrated beforehand. Thereafter, for each input image we apply vignetting elimination, distortion recovery, and top-view transformation to the image. The resultant four images are then registered and blended to arrive at what we call the bird`s-eye view image around the vehicle. The above process is transformed into a table that records all stages of the process. With this table, real-time bird`s-eye view image generation can be achieved simply through looking-up table. We assume here parking spaces have been located based on the bird`s-eye view images. Three kinds of parking spaces, parallel, diagonal, and vertical, are then considered when planning paths for automatic parking. A car-like robot, AmigoBot, has been used to extensively test on bird`s-eye view image generation and automatic parking.
This article presents an automatic parking system based on a bird`s-eye view image around a vehicle. There are four fisheye cameras mounted around the vehicle for capturing its surrounding images. Cameras are calibrated beforehand. Thereafter, for each input image we apply vignetting elimination, distortion recovery, and top-view transformation to the image. The resultant four images are then registered and blended to arrive at what we call the bird`s-eye view image around the vehicle. The above process is transformed into a table that records all stages of the process. With this table, real-time bird`s-eye view image generation can be achieved simply through looking-up table. We assume here parking spaces have been located based on the bird`s-eye view images. Three kinds of parking spaces, parallel, diagonal, and vertical, are then considered when planning paths for automatic parking. A car-like robot, AmigoBot, has been used to extensively test on bird`s-eye view image generation and automatic parking.
Description
Keywords
鳥瞰影像, 俯視影像轉換, 影像接合, 色彩混合, 曝光補償, 扭曲消除, 攝影機校正, 暗角效應, 路徑規劃, 自動停車, 非完整系統, Bird`s-eye view image, Top-view transformation, Image registration, Color Blending, Exposure compensation, Radial distortion recovery, Camera calibration, Vignetting effect, Path planning, Automatic parking, Nonholonomic system