隨著科技進步與社福照護意識的提升,與視障者行走相關的技術研究已成為重要的研究課題,近年由於影像處理技術日益提升,已有許多以影像為基礎的視障輔具研究,傳統可見光影像會受色彩與亮度影響,降低路面區域的偵測效能,因此本論文探討以紅外線深度影像為基礎的視障者行走導引系統,提出以形態分割法將路面區域進行定位,並指引視障者行進方向。 深度影像經由紅外線感測器取得,論文中使用梯度運算器進行前置處理,計算路面特徵參數,再據以進行形態學的斷開閉合等程序,可判定路面位置。實驗以多個不同環境之路面影像為例,除了平坦路面的偵測結果能導引正確行進方向外,也針對路面有障礙物的情形進行分析探討,提出偵測路中障礙物的方法,實驗結果證明可正確導引至避開障礙物的方向,驗證本論文所提行走導引系統的有效性。
As the modern technology advances, there are many researches on creating better navigation aid for visually impaired or blind people. In recent years, many image based way finding systems using RGB cameras have been proposed. Because the RGB images are highly correlated with the environment, the road finding results are unstable when the color and intensity of images changed. Infrared rays camera can provide stable depth images in spite of the color and intensity changes in the environment. This thesis studied the performance of a guidance system for the blind by using depth images grabbed by infrared rays camera. A morphology segmentation algorithm including opening and closing procedures is proposed to segment the road boundaries out of an image. Character parameters of road are calculated by using gradient operators. A series of road images are used to evaluate the performance of the guidance system. Experiment results demonstrate the excellence of the proposed guidance system.