透過您的圖書館登入
IP:3.138.114.38
  • 學位論文

結合基於電腦視覺之擴增實境與頭部追蹤實現於汽車的偏光銀漆材質擋風玻璃之抬頭顯示器系統

A Silver-Painted Windshield-Based Head-Up Display System Implemented by the Combination of Vision-Based Augmented Reality and Head Tracking

指導教授 : 駱榮欽

摘要


行車時清楚地看見前方道路景像,對於避免發生車禍是很重要的議題。一般的抬頭顯示器之行車輔助系統,僅單純地將行車相關之輔助資訊投影至車子的擋風玻璃上,當駕駛員的頭部位置發生位移時,被投影在擋風玻璃上的行車相關之輔助資訊在駕駛員的視覺上便會產生與實景不密合的現象。因此本論文提出一套基於電腦視覺與頭部追蹤技術的抬頭顯示器系統,即使駕駛員的頭部產生位移時,也能有效的讓行車相關之輔助資訊影像在駕駛員的視線中與道路景像貼合。本系統不僅具備能強化行車時前方的道路景像以避免駕駛員因視線不良發生車禍,也可藉由頭部追蹤來確保影像的正確位置以提升系統可用性。此外,擋風玻璃的放置角度不與投影機的投射方向垂直,透過影像校正使得投影的行車相關之輔助資訊能確實地與駕駛員的前方視野符合; 且利用配戴在駕駛員眼睛的偏光鏡結合塗抹在擋風玻璃上的偏光銀漆材質以單眼視覺來消除二重像的影像模糊現象 (行車相關之輔助資訊)。實驗結果由5位不同的受測者與客觀統計法進行評估,證明提出的系統在駕駛員頭部水平與垂直位移時,仍能確保被投影的道路景像之邊緣影像與道路實景之貼合度達99.05%。

並列摘要


It is a crucial issue to avoid the car accidents by strengthening the road scenes from the driver’s viewpoint when driving. In general, Head-Up Display system mounted in a vehicle purely projects the driving related information on a windshield. The projected information will not be aligned with the road scenes from the driver’s viewpoint while the driver’s head position moves. In this paper, we propose a Windshield-based Head-Up Display system based on the combination of vision-based driver assistance and head tracking techniques to have the projected images corresponded with the road scenes from the driver’s viewpoint. The proposed system is not only to strengthen the road scenes to avoid the car accidents because of poor visibility at nighttime but also to use the head tracking method for the accuracy of the projected images in order to enhance the usability of our system. Besides, the projected images need to be adjusted via image geometry correction because the windshield placed in the vehicle is not orthogonal to the projective direction of a DLP projector. Furthermore, we combine the polarized silver-painted material with the polarized lenses to eliminate Image Crosstalk on the windshield by monocular vision. The experimental results show that the projected images on the windshield can accurately be corresponded with the road scenes from the driver’s viewpoint when the driver’s head position moves horizontally or vertically, and the goodness of fit of the projected images approximately reaches 99.05% gathered by means of objective statistics.

參考文獻


[19] 洪肇澤,利用調整光流法進行障礙物偵測基於逆透視轉換運用於行車安全,碩士論文,國立臺北科技大學電腦與通訊研究所,臺北,2012。
[20] 邱俊儒,結合視差與光流進行環場鳥瞰影像之障礙物偵測機制,碩士論文,國立臺北科技大學電腦與通訊研究所,臺北,2013。
[18] 雷丰銓,具有影像優化處理的環場鳥瞰監視停車輔助系統,碩士論文,國立臺北科技大學電腦與通訊研究所,臺北,2011。
[7] C. Tran, K. Bark and V. Ng-Thow-Hing, “A left-turn driving aid using projected oncoming vehicle paths with augmented reality,” ACM Proceedings of the 5th International Conference on Automotive User Interfaces and Interactive Vehicular Applications , Oct. , 2013 , pp.300-307.
[23] Richard Hartley and Andrew Zisserman, Multiple View Geometry in Computer Vision, Second Edition , Cambridge University Press , Mar. ,2004.

延伸閱讀