本篇論文提出一套以Kinect 為基礎的盲人步行警示系統，利用 空間測距的概念將環境中的資訊轉換到距離影像上進行分析。Kinect 具有距離量測的功能，利用其量測到的深度資訊，配合路面偵測演算 法，將其資訊轉換到三維空間中，以建立起相對於環境中的空間資訊， 進而偵測各種障礙物、凹洞、坡度等情形。 本系統進一步開發障礙物偵測、凹陷偵測、坡度偵測等演算邏輯， 希望能讓使用者在行走時更能瞭解周遭的環境，有別於其它被動提供 行走指示的系統，本系統提供環境資訊讓盲人能夠判斷行走的路徑。 使用的偵測概念包括，在凹陷偵測的想法上是利用凹陷的最大面積與 最大深度最為考量依據；而坡度偵測則是利用空間向量的概念進行角 度計算，作為坡度的考量。系統的研究成果已經證明能夠進行坡度與 障礙物的即時偵測，未來研究將讓使用者能更自主性的選擇移動的方 向，改善盲人對於周遭環境的判斷情形。 This thesis proposed a Kinect-based surveillance system to help blind people in walking environments. With the salient feature of Kinect in distance measurements, the depth information can be converted into three -dimensional space by means of a road surface detection algorithm. After establishing the 3D spatial information of ramp and obstacles relative to the environment, walking environmental obstacles, such as caves, ramps and other circumstances, can be detected. The present system developed detection algorithms for obstacle, cave, and ramp to assist users better understanding of the environmental situation. Different from other system which provides users walking instructions passively, our system provides environmental obstacle information for blind people to plan optimal walking routes under variant circumstances. The cave detection takes both the maximum area and maximum depth of the cave the ramp detection is based on calculating the angle between normal vectors of ramp and reference planes. The study results have demonstrated the effectiveness of the proposed system. Future studies will apply this system to help blind people for improving judgment of the surrounding environment.
