標題: 即時影像處理技術應用於車道與車輛警示系統之研究
The Study of the Real-Time Image Processing Approaches to Lane and Vehicle Warning Systems
作者: 鐘敏偉
Min-Wei Chung
吳炳飛
Bing-Fei Wu
電控工程研究所
關鍵字: 車輛偵測;車道線偵測;Lane detection;Vehicle detection
公開日期: 2007
摘要: 本篇論文利用數位信號處理器去實現行車輔助系統。在本系統有兩個主要主題,分別是車道偏移警示系統,以及車輛偵測系統,同時整合在TI DM642 平台之上。本系統在高速公路、隧道、與市區道路之下均可運作,並且克服各種天候因素,無論是晴天、陰天、夜晚和雨天。各種車道線像是實線、虛線、雙白線,均可偵測,並且能夠偵測前、後方車輛與車道線。在TI DM642平台上經過最佳化後,我們的系統可以同時偵測前後方影像,並且可以達到每秒25 張CIF的影像。當車輛偏移車道線或是前後方車輛過近時,我們的系統會發出警告。 我們的車道線偵測系統在白天與晚上是非常強健與有效率的,利用道路模型去提升偵測準確度。此外,車道寬度與路面坡度均是即時的更新。我們的系統架設並協助在台灣首輛智慧車TAIWAN iTS-1以時速一百公里在東西向快速道路與國道3號高速公路完成自動駕駛實車測試。 車輛偵測演算法克服了許多環境因素的影響,像是強烈陰影、路面上的國字、橋樑接縫、強烈光線、橋墩下的陰影、旁車道車輛車燈影響、路面上的反光、雨滴和雨刷的影響。我們的演算法不僅可以架設在一般小客車,並且在巴士上也能運作。
In this thesis, a multipurpose digital signal processor (DSP) based driving assistant is implemented. There are two research topics in this system, lane departure and vehicle detection warning systems (LVWS) by TI DM642 platform. The system is applied in highways, tunnels and urban roadways, and works well in sunny, cloudy, and rainy conditions for day and night. Different types of lane markings can be detected, such as solid, dashed, double, etc. The algorithms on LVWS functions not only for the front view but also the rear direction. The system performance can reach 25 frames per second with CIF format through the optimization technologies which are specially designed for the DSP platform. When the vehicle departs from the road or is too close with the previous car, the proposed LVWS will generate a warning signal to alert the drivers avoiding the collision. The presented fast adaptive lane detection algorithm is efficient and robust both at day and night. A modified lane model is proposed to increase the accuracy of detection. Moreover, an automatic calibration between the lane width and the road inclination is considered and adjusted simultaneously. The vehicle detection algorithm overcomes several influences, such as the strong shadows, the text on the road, the joint of the bridge, the harsh lighting, the shadow under the bridge, the other vehicle lights at night, reflected light from the road surface, wipers and water-drops on the windshield. Our algorithm works well for the camera not only mounted in the car, but also mounted on the bus. The whole system is equipped on TAIWAN iTS-1 and has been successfully verified for hundreds of kilometers on Highway No.3 and Expressway No.68 in Taiwan.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009412543
http://hdl.handle.net/11536/80674
顯示於類別:畢業論文