透過您的圖書館登入
IP:44.205.5.65
  • 學位論文

影像式自動事件偵測誤報特性分析之研究-Citilog之應用

False-Alarm Analysis of Video-based Automatic Incident Detection System-- Application of Citilog System

指導教授 : 許添本

摘要


近年來,各國交通管理單位為監看交通道路狀況,陸續在高速公路、隧道或市區道路交叉路口架設閉路電視系統(Closed Circuit Television,CCTV)即時掌握異常路況,相關資訊除作為交通管理之依據外,亦可提供用路人參考。但CCTV之影像資料常常無法真正有效使用。因此將影像處理技術與CCTV結合發展影像式自動事件偵測系統(Video based Automatic Incident Detection System),對於事件資訊之掌握將可更具即時性,且大幅提昇監控效率。 影像式自動事件偵測系統於事故發生時,即時發出警告提醒監控人員,並由影像畫面確認事故地點、車輛、傷亡情況與當時交通狀況,以迅速採取相關處理措施,縮短事故處理時間,並避免二次事故的發生。本研究應用影像式自動事件偵測系統,以台北市市區道路路口與路段為主要拍攝地點,選擇適當錄影拍攝範圍,並將影像資料進行事件自動偵測。探討各種事件類型與特性,並進行情境測,藉由使用上之調整與設定,來改善與降低誤報情形,發揮系統之功能。 由事件偵測結果,可將誤報事件區分為偵測錯誤、環境因素、車輛與行人因素等;本研究以偵測區域的畫設與停等時間設定作為主要參數進行事件與情境測試,認為偵測區域之劃設範圍,以能涵蓋整個影像範圍為佳且數量較少為佳;市區道路上常見之設施如人孔蓋,可使用較高之參數值於該設施範圍,以降低誤報情況;而參數設定可依不同需求人員,在不同偵測區域以不同參數進行調整與偵測,掌握其所關心之事件。本研究根據上述結果擬定影像式自動事件偵測系統之原則與方式,可以有效改善誤報之情形,提升系統偵測效率,供後續使用之參考。

並列摘要


Department of Traffic Management in many countries set up Closed Circuit Television (CCTV) at highways and tunnels to control the situation of traffic recent years. Expect for the usage of traffic management, the real-time information also can provide for the users. However, the images of CCTV are seldom used efficiently. The Video-based Automatic Incident Detection Systems integrate CCTV and advanced image processing techniques to detect vehicle and detect incidents. By using Video- based Automatic Incident Detection Systems, operators in Traffic Control Center could be more efficient. When an accident happens, the Video-based Automatic Incident Detection System would alarm in a short time, and warn the operators to pay attention to the images of CCTV to know where the accident happens, the situation of the accident, and the nearby traffic. Furthermore, operators could quickly take some actions to decrease the accident duration time, lower the level of damage, and avoid the second-hand accidents. This study applies the Video-based Automatic Incident Detection System at urban road in Taipei City, and classifies the incident types and the false-alarm incidents. Then, this study takes some scenario into tests to improve false-alarm of Video-based Automatic Incident Detection Systems. As the results of incident detection at urban road, this study classifies false-alarm incidents into three types: error detections, environmental factors, and factors of vehicles and pedestrians. The two important functions of Video-based Automatic Incident Detection Systems are detection areas and time of stop cars. According to the results of incident detection at urban road and the results of scenario tests, this study brings up some principles of using Video-based Automatic Incident Detection Systems, which can improve the false-alarms, shorten the detection time, and detect incidents efficiently.

參考文獻


16. 徐泰詮,「以視覺為基礎之道路交叉路口車輛事故自動偵測系統」,元智大學資訊工程學系碩士班碩士論文,民國95年。
26. 鄭其政,「影像監控系統交通參數擷取技術之研究」,國立臺灣大學土木工程學研究所碩士論文,民國95年。
12. 林銘樂,「夜間即時影像處理擷取交通參數之研究」,國立臺灣大學工學院土木工程學研究所碩士論文,民國96年。
20. 許添本、鄭雅文,「具備預測機制的追撞事件偵測新演算模式之研發」,運輸計畫季刊,第三十卷,第三期,第539~576頁,民國90年9月。
29. 蕭銓宏,「應用自動車輛辨識技術於長隧道自動事件偵測」,國立中央大學土木工程研究所碩士論文,民國95年。

延伸閱讀