以視覺為基礎之道路交叉路口車輛事故自動偵測系統

A Vision-based System for Car Accidents Detection at Intersections

徐泰詮(Tai-Chuan Hsu)

指導教授 ： 林啟芳

元智大學/資訊學院/資訊工程學系/碩士(2006年)

https://doi.org/10.6838/YZU.2006.00219

摘要

隨著車用市場的平價及普及，路上行駛車輛的密度已經越來越高，不過隨之而來的是交通事故高頻率的發生，特別是交叉路口。因此道路監控系統需求與日俱增，如果可以在第一時間偵測事故的發生，不僅僅對交通安全有大大的幫助，更能迅速的舒緩發生交通意外處的擁塞。本論文的目的是基於研究智慧型運輸系統中，發展一套系統可自動在交叉路口監控每輛車輛的情況，並分析可能發生的事件。我們利用一部個人電腦來擷取交叉路口所拍攝到的畫面，並透過擬提方法來分析畫面，找出畫面中所有的移動車輛，標示出移動車輛，並作追蹤，取得移動車輛的資訊後，最後透過車輛的相對關係以及其行車的軌跡，利用不斷預測的方法，將交通的事件可能性表示出來。
實驗結果顯示，根據擷取一般道路交叉路口的車流影像，並利用本文所提出的方法來偵測，我們的方法是準確且有效的。

關鍵字

影像處理 ； 智慧型運輸系統 ； 移動物體追蹤 ； 事件偵測

並列摘要

As the result of cars is available to all, there are many cars of highly density on road. Unfortunately, conveniences of traffic bring more and more traffic event, especially at intersections. If we could do incident detection immediately, it must be a great help to traffic safety, and release traffic jam around the scene. Our goal is to establish a system that monitoring each car pass through intersection and analyze all possible events which is based on Intelligent Transport System. First of all, we extract a lot of frame at intersection by PC and use our method to analyze the frame. Second, we do Target tracking in frame and mark them. Third, we will track the cars marked and get the information of them. Forth, according to their relationships and trajectories, we use the method of recursive predict to present all the possible traffic events.
According to our method to do Pattern recognition, Experimental results shows that our method is effective and accurate.

並列關鍵字

image processing ； incident detection ； intelligent transport system ； target tracking

參考文獻

[1]R. Gangisetty, “Advanced traffic management system on I-476 in Pennsylvania,” Proc. IEEE ITS Conference, pp. 373–378, 1997.

[2]J. C. Rojas and J. D. Crisman, “Vehicle detection in color images,” Proceedings of IEEE ITS Conference, pp. 403–408, 1997.

[4]A. H. S. Lai and N. H. C. Yung, “A video-based system methodology for detecting red light runners,” Proceedings of IAPR Workshop on MVA, pp. 23–26, 1998.

[6]K. Toyama, J. Krumm, B. Brumitt, and B. Meyers, “Wallflower : principles and practice of background maintenance,” Proceedings of IEEE International Conference on Computer Vision, pp. 255–261, September 1999.

[8]D. Gutchess, M. Trajkovic, E. Cohen-Solal, D. Lyons, and A. K. Jain, “A background model initialization algorithm for video surveillance,” Proceedings Eighth IEEE International Conference, Vol. 1, pp. 733-740, July 2001.