Machine Vision to Lane Departure Sensing System for Urban Traffic

機械視覺於主動式車輛偏航偵測系統之研究

冷玉琦(Yu-Chi Leng) ； 陳介力(Chieh-Li Chen) ； 戴崇禮(Chung-Li Tai)

《中國機械工程學刊》 37卷3期 (2016/06) Pp. 201-211

摘要

一般市區道路常附加各種道路指標與資訊，加上各類車種互動，因此比高速公路具有更高的環境複雜度。本文提出非對稱二維高斯分佈矩陣與輔助運算指標，針對車道標線特性估算像素符合率，藉此降低雜訊及車道標誌對車道偵測所造成的影響。並根據車道標線具有連續性之特徵進行道路邊線辨識與配對。演算法具強健穩定性，可應用於道路駕駛偏航感測系統的開發。由實驗結果顯示本系統可有效於環境複雜之市區道路使用，有助於道路駕駛安全之提升。

並列摘要

A lane departure sensing system with uncalibrated camera is proposed in this paper. The study focuses on urban traffic scenes with varieties of lane marks instead of simple highway scenes. A distance map and non-symmetrical 2-D Gaussian distribution matrix are used to generate various voting value for Hough transform, so that the lane marking and lane signs are excluded and the accuracy of lane boundary detection is improved. Lane width and lane boundary continuity are obtained for better reliability and quality of lane detection for urban traffic scenes. The result of lane boundary detection is then used to establish a lane departure sensing system, which can be used to enhance driving safety. Experimental results of urban road scenarios have demonstrated feasibility and robustness of the proposed method.

並列關鍵字

Lane departure sensing system ； Lane mark detection ； Driving assistance ； Machine vision ； Edge detection