使用頭戴式顯示裝置之虛擬後視鏡

Abstract

車輛雖有後視鏡,但傳統後視鏡提供的視野會受限於鏡面大小與駕駛人的視 角,產生盲點;且因車體結構,左右側後視鏡距離駕駛人太遠,駕駛人觀看左右 後視鏡的頭部物理旋轉量太大,導致駕駛人無法專注於目前的駕駛線道上。現行 主流的盲點輔助系統是於左右後視鏡上架設攝影機,並於中控臺螢幕提供影像予 駕駛人。 本研究使用現行某車款,此車型於車輛後側及左右兩側後視鏡皆已架設攝影 機,套用此款車型之攝影機參數於本實驗室之駕駛模擬器,取得實驗數據,提出 一虛擬後視鏡顯示系統,此系統分為兩部分: (1)將此三攝影機拍攝之影像接合,提供駕駛一廣域的後視鏡。 (2)使用頭戴式顯示裝置將左右後視鏡攝影機影像完整呈現予駕駛人。 影像接合方面,透過攝影機之內外部參數,將影像投影至一三維模型,以取 得一廣域電子後視鏡影像。為避免兩兩影像重疊區塊鬼影現象的產生,本研究提 出一可依據深度調整的三維模型,並探討此三維模型的精細程度與深度資訊的正 確率所影響的成像效果,並且評估即時運算的可能性。 後視鏡影像顯示方面,本研究提出一使用頭戴式顯示裝置之顯示系統,將後 視鏡影像提供給駕駛人。使用此系統,後視鏡視野將不再受限於鏡面大小與駕駛 人的視角,且可以自定義虛擬後視鏡的位置,以降低駕駛人頭部物理旋轉量,更 可專注於駕駛線道上。

To increase drivers' environment awareness during car driving, current cars usually equip blind spot detection system that provides additional visual hints, helping the drivers perceive environmental conditions. However, previous solutions usually provide limited environmental information and require users to frequently check the information, resulting driving distraction. In this thesis, we proposed a “Virtual Rear Mirror”, allowing users to view wide field-of-view environment image through a head-mounted display (HMD) and efficiently reduce driver’s head movement which caused by accessing the left and right rear mirror. The wide FOV image are stitched from images captured by cameras mounted around a commercially-available car, which can assist drivers to perceive the blind spot situation more easily with less gaze distraction. The main contributions of this thesis are stated as follows: (1) Our system allows users to perceive rich environment conditions with less head movement, improving driving focusness. (2) Our solution can be applied to commercially-available cars, greatly lowering down the costs of development.