標題: 利用權重式特徵點比對的車輛周遭監控系統
Vehicle surrounding monitoring using weighted feature point matching
作者: 方奎理
林進燈
電控工程研究所
關鍵字: 車輛環景影像;障礙物偵測;Vehicle surroundimg image;Obstalce detection
公開日期: 2011
摘要: 以研究數字來看,每年的交通事故數量仍然居高不下,其中交通事故中又以車輛與其他障礙物碰撞事故發生最為頻繁。因此各種提升安全性的高效能車輛與日俱增,諸多研究人員也提出不同的預防車輛碰撞的安全系統。然而太多的車載系統反而造成使用者在操作上感到麻煩以及切換不易。因此在本篇論文中,我們提出一個具有整合系統概念的防碰撞系統,透過四支攝影機來建立車輛周遭影像及整合障礙物偵測技術。為了減低地面標線對障礙物偵測造成的影響,我們提出一個創新的地面移動量估測技術來降低障礙物偵測的誤報率。我們手動建立了一個正確的地面移動量表來計算估測誤差藉此來驗證估測結果。與其他論文的方式做比較,我們所提出的方法提高了地面移動量估測的準確度並且進一步的減少地面標線所造成的誤報率。本篇論文透過整合兩個系統來減低硬體成本及使用上的不方便以及提出一個創新的地面移動量估測方式來有效的降低誤報率,並透過偵測阻礙行進路線上或突出地面具有高度的物體來實現車輛防碰撞警示系統。
The number of traffic accidents is growing up quickly according to the research numbers in every year. Among of all the traffic accidents between vehicle and other generalized obstacles occur most frequently. Therefore, the number of high safety vehicles is increasing over the world. Many researchers proposed various pre-warning collision systems. However, the multi-system will confuse drivers on switching different systems. Hence, this study proposes a pre-warning system and integrates different system. We construct a vehicle surrounding monitoring system through four cameras and integrate an obstacle detection system on the vehicle surrounding monitoring system. In order to decrease the influence caused by ground texture. We also propose a new ground estimation technique to reduce the false alarm rate of obstacle detection. To evaluate the estimation result, we generate a ground truth table by manual to calculate the estimation error. Compared to other paper, our approach promotes the accuracy of ground movement estimation further to reduce the false alarm caused by ground texture. In this thesis, we integrate two systems to decrease the hardware cost and propose a novel ground movement estimation method to reduce the false alarm effectively. By detecting the object significantly arise the road plane, we realize vehicle pre-collision warning system.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079812614
http://hdl.handle.net/11536/46969
Appears in Collections:Thesis


Files in This Item:

  1. 261401.pdf