適應性背景重建技術用於多目標追蹤系統與其在交通參數擷取之應用

Abstract

在最近幾年，隨著車輛與其他交通工具的快速發展與普及化，各地的交通狀況也日趨繁忙與混亂。因此，為了改善或有效控管交通狀況，智慧型運輸系統（Intelligent Transportation Systems ,ITS）變成一個在學術研究和業界開發中十分重要的領域。而在傳統的交通監測系統，其監測方法在資料的擷取或系統擴展空間上均表現不佳。而監測系統如同整個系統的眼睛，必須扮演自動檢測出車輛及行人同時持續地追蹤他們的角色。藉由追蹤系統獲得的資訊，ITS可以執行進一步的分析來使得系統有效率地進行有效決策以改善交通狀況。 在這篇論文中，我們提出了一個多目標的即時追蹤系統，它可以在固定式攝影機獲得的交通監視影像中，進行移動物體的檢測。配合適應性背景重建技術，系統可以有效地處理外界光線或其他環境的變化來達到良好的物體擷取結果。另外，我們將擷取出來的物體特徵搭配以區塊為基礎的追蹤演算法來達到物體持續追蹤的目的，也可以正確地追蹤發生重合或分離情形的物體。得到追蹤的結果後，我們可以進一步分析物體的特性及行動來產出有用的交通參數。針對我們提出的系統架構，我們實作了一個監測系統並包括移動物體分類及事故預測的功能。我們以實際的路口交通影像與其他監測影像樣本來進行實驗，實驗結果證明我們所提出的演算法與系統架構達成了強健性的物體擷取結果並在重合及分離的情況下成功地追蹤物體。同時也正確地擷取了有用的交通參數。

In recent years, the traffic situation is busy and chaotic increasingly everywhere because of high growth and popularization of vehicles. Therefore, intelligent transportation systems (ITS) become an important scope of research and industrial development in order to improve and control the traffic condition. In traditional traffic surveillance systems, their monitoring methods are inefficient in information extraction and short of improving. A tracking system is like ITS’ eyes and it plays the role of detecting vehicles or pedestrians automatically in the traffic scene and tracking them continuously. According to results of tracking systems, ITS can do further analysis and then perform efficient and effective actions to make the traffic condition better. In this thesis, we present a real-time multi-objects tracking system and it detects various types of moving objects in image sequences of traffic video obtained from stationary video cameras. Using the adaptive background reconstruction technique we proposed can effectively handle with environmental changes and obtain good results of objects extraction. Besides, we introduce a robust region- and feature-based tracking algorithm with plentiful features to track correct objects continuously and it can deal with multi-objects occlusion or split events well. After tracking objects successfully, we can analyze the tracked objects’ properties and recognize their behavior for extracting some useful traffic parameters. According to the structure of our proposed algorithms, we implemented a tracking system including the functions of objects classification and accident prediction. Experiments were conducted on real-life traffic video of some intersection and testing datasets of other surveillance research. The results proved the algorithms we proposed achieved robust segmentation of moving objects and successful tracking with objects occlusion or splitting events. The implemented system also extracted useful traffic parameters.