基於即時影像追蹤之交通監測系統

Abstract

本論文發展出一套可應用於交通監測之即時影像追蹤系統。為了實現影像追蹤的即時性，文中設計一影像前置處理卡及影像擷取卡，可達成每秒60個圖場(frame)之320 240像素CCD影像邊緣抽取。配合主動式輪廓模型及卡曼濾波器技術，成功發展出一套即時影像追蹤系統，達成即時追蹤車輛之目的。本論文已完成即時取得二項交通資訊的研究，一為交通參數如車速、車輛數目、車輛密度…等，二為即時車輛碰撞意外之偵測。我們以播放路況影像驗證追蹤車輛及機車的效果。另一方面，於實驗室中建構一模擬路況利用CCD擷取影像，驗證即時偵測車禍發生的功效。

In this thesis, we developed of a traffic monitoring system based on real-time image tracking. To realize the real-time image tracking system, we successfully developed an image grabbing card and an image processing card. These image card together with an industrial PC (IPC) construct a stand-alone vision system for traffic monitoring. Frame rate edge detection for 320 240 pixels image can be obtained using the image processing card. Adopting active contour models and Kalman filtering techniques, we successfully achieve real-time image tracking of moving vehicles using the vision system. This system provides two pieces of information of the traffic monitoring based on real-time image tracking as follows: Firstly, the traffic parameters like velocity and numbers of vehicles would be thus acquired. Secondly, the accident detection of the real-time image tracking. The system demonstrated that the real-time vehicle and motorcycle tracking of the broadcasted video of traffic. We established an imitative lane to demonstrate the effectiveness of the accident detection.