以雙鏡面環場攝影機及自動車作人行道上導盲犬之研究

Abstract

本研究提出了一個應用於機器導盲犬在人行道上行走的自動車系統，利用一部搭載雙鏡面環場攝影機的自動車當實驗平台，在環場影像中直接求出實際物體的立體資訊，並引領盲人沿著規劃的路徑行走。首先，我們利用建立的環境學習系統來記錄導航地圖，此地圖包含自動車在戶外環境中行走的路徑、路徑中附近路標的位置，以及相關的導航參數。接著，利用人行道上常出現的路標(人行道路緣、樹幹、草地角點、交通錐、道路白線和招牌)作自動車定位，以輔助導航。 此外，我們提出空間映射的方法來發展新的直線偵測技術，在環場影像上直接偵測出直線特徵，並計算出人行道上與自動車相互平行或垂直的直線位置，藉以偵測前述自然特徵物及人工特徵物之位置，作機械誤差之校正，並計算出正確的自動車位置。 最後，本研究也提出動態調整門檻值的技術，讓系統適應戶外環境的各種光影變化所造成的影響。實驗結果顯示本研究所提方法完整可行。

Various techniques for construction of a machine guide dog using a two-mirror omni-camera and an autonomous vehicle for navigations on sidewalks are proposed. The autonomous vehicle can compute 3D information from acquired omni-images to localize itself using pre-selected landmarks, and guide a blind person to follow a planned path to a destination on a sidewalk. Firstly, a method for learning the sidewalk environment is proposed to construct a navigation map, including a navigation path, along-path landmark locations, and relevant vehicle guidance parameters. Next, a navigation system with self-localization and automatic guidance capabilities using landmarks including curb lines, tree trunks, stop lines on roads, lawn corners, traffic cones, and signboards is proposed. By the use of a space-mapping technique, three space line detection techniques for use directly on the omni-image are proposed, which can be used to compute the 3D position of a specific space line in the shape of a sidewalk landmark. Moreover, based on the techniques for detecting three space lines, techniques for detections and localizations of the above-mentioned natural and artificial landmarks are proposed. Using these vehicle self-localization techniques, imprecise vehicle positions due to incremental mechanic errors can be adjusted. In addition, for the purpose of continuous navigation, a curb line following technique is proposed as well to guide the vehicle along a sidewalk when landmarks are not available during the navigation process. To detect landmarks in the outdoor environment, techniques for dynamic threshold adjustments are also proposed for adapting the system’s capability to varying lighting conditions in navigation environments. Good experimental results showing the flexibility and feasibility of the proposed methods for real applications are also included.