公路長隧道內交通事件偵測

Traffic Incident Detection in Long Highway Tunnel

柯盟威(M. W. KE)

指導教授 ： 葉榮木 ； 蔡俊明

國立臺灣師範大學/科技學院/機電科技研究所/碩士(2010年)

摘要

隨著工程及科技的發展迅速，開啟了長隧道時代的來臨；然而，隧道內的交通資訊取得、事故處理以及逃生系統，遠比在一般開放空間快速道路搶救，難上幾十甚至幾百倍。因此，一般用路人對於長隧道的特殊交通空間，並未有完善的知識以及交通資訊。若能在事故發生前做事故預防的警示，以及隧道內即時交通資訊的指示，成為交通順暢與否及避免危機的重要課題。
本研究目的，在於以影像處理達成(1)估測隧道內交通資訊(2)車輛變換車道偵測(3)故障停等車輛偵測(4)長隧道內火焰偵測。如果能在事故發生前、後，透過資訊可變標誌，給予適時、適當的預警，不但能降低事故的發生，更能大量節省事故發生後，緊急處理所需的時間，而使行車更加順暢。
本研究以公路長隧道內，雙白線資訊為偵測基礎，並(1)設定行車車輛數目，估測行車流量(2)當偵測出之車輛與雙白線直線方程式產生交集時，偵測出車輛跨越雙白線(3)以時空域動量分析偵測出車輛故障停等(4)基於公路長隧道標準照明下，分析火焰色彩資訊，進行火焰偵測。
實驗結果顯示，在各項交通事件偵測，準確率皆在九成以上；並且，在長隧道行車中，具備最低速限以及標準照明環境之下，本研究提出(1)以時空域動量分析演算法，偵測車輛故障停等，以及(2)利用色彩空間，偵測火焰區域，以八卦山隧道交通事件為例，進行偵測，所偵測之時間，分別較目前既有影像式事件偵測方式，更為提早1秒與2.6秒。如此便能更即時的通報給行控中心，作適當之處置。

關鍵字

長隧道 ； 影像處理 ； 時空域分析 ； 交通資訊 ； 移動物件偵測

並列摘要

The rapid development of engineering and technology creates the possibility of traffic flow through long tunnels. However, to obtain the traffic information and incident handling and escape systems in the tunnel, are more difficult than open space in the general highway, hard on a few times, even several hundred times. Therefore, the common user for the special transport space in long tunnel, there is no perfect knowledge, and traffic information. If accident prevention alerts was done before the accident, and really time of traffic information in the tunnel, has become an important issue for a smooth traffic flow or not.
The purpose of this study is using image processing to achieve (1) estimation the traffic information in tunnel, (2) the vehicle spanning double white line detection, (3) the failure and stopped vehicle detection (4) fire detection in long tunnel. Before the accident occurred and after, the appropriate warning signal through the CMS (Changeable Message Sign)with timely, not only can reduce accidents but saving more times in the emergency treatment after the accidents, to make driving more smoothly.
In this study, we used the Double White Line information in long highway tunnel to (1) set a number of vehicles and estimating the traffic volume, (2) detect the vehicle spanning Double White Line while the vehicle intersects with the Double White Line, (3) detect the vehicle failure by using the Temporal-Spatial momentum algorithm, (4) detect flame by analyzing its color space, based on standard long highway tunnel lighting.
As the experimental results, our accuracy of the detection is upper than ninety percentage in each incident detection. And, the traffic with a minimum speed limit as well as standard lighting brightness of the environment in long tunnel, the proposed of (1) temporal-Spatial momentum algorithm to detect the failure and stopped vehicle, and (2) color space used to detect the flame region, to detect traffic incident in the Pakuashan tunnel, the experimental results is earlier 1 second, and 2.6 seconds than the existing image detection methods.

並列關鍵字

Long tunnel ； Image Processing ； Temporal-Spatial analysis ； Traffic information ； Moving object detection

參考文獻

[15] 邱建中，「利用時空域分析與背景相減法作視訊移動物偵測」，碩士論文， 國立臺灣師範大學機電科技學系， 2009。

[10] N. Otsu, “A Threshold Selection Method from gray-level Histograms”, IEEE International Conference on System, vol. 9, pp. 62–66, 1979.

[12] K. A. Patwardhan, G. Sapiro, and V. Morellas, “Robust foreground detection in video using pixel layer”, IEEE transactions on pattern analysis and machine intelligence, vol. 30, pp. 746-751, April 2008.

[14] E. J. Carmona, J. M. Cantos, and J. Mira, “A new video segmentation method of moving objects based on blob-level knowledge”, Pattern Recognition Letters, vol. 29, pp. 272-285, 2008.

[16] L. W. Tsai, J. W. Hsien, and K. C. Fan, “Vehicle detection using normalized color and edge map”, IEEE transactions on image processing, vol. 16, pp. 850-864, March 2007.