以霍夫轉換為基礎之車道線偵測偏移警示系統

Lane Departure Detection based on Hough Transform

鄭羽馨(Yu-Sin Jheng)

指導教授 ： 杜維昌

義守大學/電機資訊學院/資訊工程學系/碩士(2016年)

https://doi.org/10.6343/ISU.2015.00342

摘要

車道線自動偵測在自動駕駛方面佔有重要的角色，然而畫面上經常受非車道線的干擾，在偵測過程中容易會有一些雜訊而產生誤判。本文主要以霍夫轉換為基礎，提出改進的新策略。首先，設計一動態的感興趣區域，將偵測範圍縮小至車輛前方，用以減少非車道線的干擾。其次是時間軸上估測，藉以輔助車道線的偵測校正，用來排除例外狀況，如車道線的消失可使用先前影像來補正。為了強化偵測的穩定性，利用車道線邊緣兩側線段互為平行的特性，設計車道偵測演算法以取代直線偵測，藉以去除路面上其他警示標記的干擾。本研究主要針對國內高速公路為主，其目標是開發具備即時性、抗干擾、高準確的車道線偏離系統，以期能主動提醒駕駛者，維護行車安全。

關鍵字

車道線偵測 ； 霍夫轉換 ； 感興趣區域

並列摘要

Automatic land detection plays an important role in utopilot systems, but it is often interfered with the non-lane area on the screen so that it is easy to make some noise to produce false positives in the detection process. This hesis is mainly to propose novel strategies based on Hough transform. First of all, design an adaptive region of interest, and narrow down the detected range to the front of vehicle. The second step is the timeline estimation to correct the detected lane, which is used to exclude exceptions; for instance, the disappearance of the lane can be corrected by using the previous frames. In order to strengthen the stability of detection, make use of the characteristics, both sides of the lane edge paralleling to each other, to propose the land-based detection instead of the line-based detection for removing the interference of other warning marks on the road. This research is focused on the domestic highway, and its objective is to develop lane departure warning systems with real time, anti-interference and high accuracy such that the system can be initiative to remind drivers to maintain traffic safety.

並列關鍵字

lane detection ； Hough transform ； region of interest

參考文獻

[2]J. C. McCall and M. M. Trivedi, “Video-based Lane Estimation and Tracking for Driver Assistance: Survey, System, and Evaluation,” IEEE Trans. Intelligent Transportation Systems, vol. 7, no. 1, pp. 20-37, 2006.

[3]M. Bertozzi and A. Broggi, “Real-Time Lane and Obstacle Detection on the GOLD System,” Proc. IEEE Intelligent Vehicles Symposium, Tokyo, Japan, Sep. 18-20, pp. 213-218, 1996.

[4]M. Bertozzi and A. Broggi, “GOLD: A Parallel Real-Time Stereo Vision System for Generic Obstacle and Lane Detection,” IEEE Trans. on Image Processing, vol. 7, no. 1, pp. 62-81, Jan. 1998.

[5]M. Bertozzi, L. Bombini, A. Broggi, P. Cerri and P. Grisleri, “GOLD: A Framework for Developing Intelligent-Vehicle Vision Applications,” IEEE Trans. Intelligent Systems, vol. 23, no. 1, pp. 69-71, 2008.

[6]M. Beauvais and S. Lakshmanan, “CLARK: A Heterogeneous Sensor Fusion Method for Finding Lanes and Obstacles,” Image and Vision Computing Special Issue on Intelligent Vehicles, vol. 18, pp. 397-413, 2000.