傅楸善臺灣大學:資訊工程學研究所陳信銘Chen, Shin-MingShin-MingChen2010-05-182018-07-052010-05-182018-07-052008U0001-1406200815110800http://ntur.lib.ntu.edu.tw//handle/246246/183680基於視覺的交通監控系統是電腦視覺的一項重要的研究，一般的交通資訊如車流量、車流速、壅塞狀況偵測、及壅塞長度預估等，有了正確的交通量分析才能做正確的交通指揮，都市交通規劃。在一般的天橋下裝上全方位照相機可以得到一個較為廣域的道路影像，將所得影像中的連續影格中利用各種電腦視覺的技術找到背景和前景，再分別將前景做車流量計算和對象追蹤，本文所提出的多輛車跟蹤的架構可以找出車輛，並且修正車子行經的軌道。根據我們的實驗性結果, 我們的方法可有效率地分析十字路口中行經公車、汽車及摩托車。A vision-based system for vehicle detection and tracking in video streams is an important research in computer vision. It also plays an important role in ITS (Intelligent Transportation Systems). Mounting high omni-directional camera allows the omni-directional camera to cover a wider area. Motion detection aims to capture the changed region from the omni-directional video image sequences and build a background image. In this paper, a multiple-vehicle counting and tracking framework for intersection traffic surveillance is proposed. In order to accurately locate vehicles, we use Kalman filter to correct vehicle trajectories. According to our experimental results, our approaches can analyze the vehicles at intersection efficiently.誌 謝 i 要 iiibstract ivigure Content viable Content ixhapter 1 Introduction 1.1 Calibration and Dewarping 3.2 Background Model 5.3 Temporal Differencing 6.4 Combination Method 7hapter 2 Vehicle Detection 10.1 Shadow Removal 10.2 Motion Segmentation 11hapter 3 Vehicle Tracking 13.1 Kalman Filter 14.2 Tracking Model 16.3 Multiple Vehicle Tracking 18.4 Multiple-Vehicle Counting 21hapter 4 Experimental Results 23.1 Vehicle Classification 23.2 Precision and Recall 33hapter 5 Conclusion and Future Work 37.1 Conclusion 37.2 Future Work 40eference 41application/pdf1419033 bytesapplication/pdfen-US電腦視覺交通監控車輛追蹤computer visiontraffic surveillancevehicle trackingthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/183680/1/ntu-97-R95922159-1.pdf