即時影像視覺車道線偵測之研究

Development of A Vision-Based Real Time Lane Detection

廖永盛(Yung-Sheng Liao)

指導教授 ： 曾百由

國立臺北科技大學/機電學院/機電整合研究所/碩士(2007年)

摘要

本論文在開發一個以影像視覺為基礎的高效率車道線辨識演算法，其演算法能夠於嵌入式平台上進行驗證。受限於嵌入式平台的運算與記憶能量，為了增加車道線辨識效率，並維持運算可靠度，演算法使用兩個特別的濾波器(快速中值濾波器、特殊邊緣濾波器)。其中快速中值濾波器最多僅需二十三次比對，即可得到兩個濾波後的中值；另外透過特殊邊緣濾波器將特定區域的車輛或建築物等背景雜訊濾除，僅留下具有斜線特徵的車道線，得以正確辨識車道線。利用上述演算法實際驗證，當車輛不正常偏離車道線時進行警示動作，提醒駕駛者控制車輛回至車道內。
實驗結果證明此系統能夠達到每秒鐘處理二十張影像資料量，如此的效率已可有效應用於車道偏移警示系統。此系統已經在快速道路與高速公路成功地完成實車驗證，並在不同的外界環境，如晴天、雨天、隧道、夜間，均能夠進行辨識。

關鍵字

車道線辨識 ； 影像感興趣範圍 ； 快速中值濾波器 ； 特殊邊緣濾波器

並列摘要

This thesis has developed a vision-based lane recognition method. A dual core DSP embedded system is implemented to verify the functionality. Because the embedded system has limited computing power and memory, the algorithm has to be as efficient as possible while maintaining the system reliability. The algorithm uses two special image filters: the fact median filter and the special edge filter. The fast median filter needs only 23 comparisons for two medians in the worst case, and the special edge filter which makes road marks more obvious than background noises such as vehicles, buildings etc.. This algorithm is used form to construct a lane departure warning system, which will remind the driver to control the vehicle safety when the vehicle abnormally departs the lane.
The real-world road tests have been applied to the developed lane departure warning system. The system performance can process 20 frames per second and can be used on highway or freeway in different weather and environment, including clear day, raining day, and night.

並列關鍵字

Lane Recognition ； Region of Interest ； Fast Median Filter ； Special Edge Filter

參考文獻

1.National Highway Traffic Safety Administration, http://www.nhtsa.dot.gov/

5.J. Lee, “A machine vision system for lane-departure detection,” Proceedings of Computer Vision and Image Understanding, vol.86, pp. 52-78, April 2002.

6.J.P. Gonzalez and U. Ozquner, “Lane detection using histogram-based segmentation and decision trees,” Proceedings of IEEE Intelligent Transportation Systems, pp.346-351, August 2002.

7.M.A. Sotelo, F.J. Rodriguez, L.M. Bergasa, and L. Boquete, “A Color Vision-Based Lane Tracking System for Autonomous Driving on Unmarked Roads,” Proceedings of Autonomous Robots, vol.16, pp. 95-116, November 2004.

8.M. Bertozzi and A. Broggi, “GOLD: a parallel real-time stereo vision system for generic obstacle and lane detection,” Proceedings of IEEE Transactions on Image Processing, vol. 7, no. 1, pp. 62-81, Jan. 1998.