具有樓梯偵測及移動功能之室外型主動式 履帶機器人的開發

Abstract

在本論文當中，我們結合影像(RGB)與深度(Depth)資訊，提出了一種以RGB-D為基礎的自主式戶外履帶機器人，所設計履帶機器人可於日夜間戶外自主式完成偵測、迫近、及上下樓梯的動作。根據室外光線的變化，所開發履帶機器人可自動切換為深度資訊模式、混合影像深度資訊模式、及影像資訊模式。其中，深度資訊可用於環境光線不足時（如夜間或室內）履帶機器人的自主式控制；在室外陽光照射下，紅外線深度感測器資訊將無法正常取得，因此，本論文提出一混合影像與深度資訊的影像處理系統，首先，我們使用Hough建立RGB影像中樓梯邊緣特徵，再配合可取得的深度資訊，建立樓梯深度平面方程式，以完成樓梯距離偵測；在陽光完全直射下，紅外線深度感測器將完全無法取得資訊，履帶機器人將僅使用影像資訊，進行簡單的迫近動作。在RGB影像處理上，我們採用雙邊濾波器來強化樓梯邊緣及平滑影像。最後，經由實驗結果證實所開發之戶外履帶機器人，在戶外陽光及環境的影響下，可自主式地完成戶外連續上下樓梯的任務。In this thesis, we combine RGB and depth information to develop an autonomous tracked robot for detecting stairs and moving on stairs in outdoor environments. This tracked robot can work not only in daytime but also in nighttime. The robot can detect, approach, climb up and down stairs autonomously. According to the variation of sunlight intensity, the robot switches from depth mode to RGB-D fusion mode or RGB mode by itself. Depth information from IR camera can be used in dark and indoor environments. Under sunlight, however, IR camera cannot work properly. Given the limitation, in this paper we propose an algorithm which is based on RGB-D data fusion. First, we use Hough to detect the features of stairs and then combine useful depth information to acquire needed information. Second, we use color image to approach stairs since IR camera can only obtain RGB data under sunlight. As for image processing, we use bilateral filter to enhance stair features and smooth the image. Finally, through the results of the experiment, we can confirm that the self-made tracked robot can detect, approach and climb up and down stairs continuously and autonomously in outdoor environments even under sunlight.