在本論文中,我們提出一套以形態學及分水嶺為基礎的影像處理,搭配色調(hue)、飽和度(saturation)、強度(intensity)分布做路面區域的偵測,以及立體視覺和人工智慧做自動車導航,以期可行走在校園內外之人行道上。本系統主架構分成區域導航模式及全域導航模式。在導航開始前,攝影機校正是必要的,我們利用特定8點的3D點及投射到左右攝影機中之影像點以最小平方法來求得左右攝影機校正參數,之後我們可利用這些參數及左右影像點進行影像中景物之3D重建。 在區域模式中,路面萃取及子目標特徵搜尋為主要研究重點。由於室外環境存在著許多複雜的景物,因此需要一個強健的分割方法來為我們找出路面的候選區域,所以利用分水嶺演算法提供這樣的需求,但分水嶺演算法也必須配合一個完善的梯度影像,因此我們利用形態學中的膨脹、收縮、斷開、閉合這些函數處理出符合我們要求的梯度影像以及消除一些太過細微資訊,經由分水嶺分割出數個路面的候選區域。首先由於一般的路面具有較一致的紋理及低飽和度的特性,因此我們應用此特性做我門路面候選區域的特徵化。並且我們利用立體視覺來取得環境的3D資訊,因此影像點對應是必要的,在此我們使用感測點理論來得到影像對應點並得到對應點的3D資訊。在子目標特徵搜尋方面我們利用物形比對的方法,依據景物的長度、角度、面積及景物之間的間隔距離來萃取出特定景物,如軌道線及圍牆上的交叉線當作地圖,並依此做為是否到達子目標或目標位置的依據並往下一個子目標或目標前進或是下達停車指示。 在全域模式中,依據路面資訊以及車子與子目標或目標之間的方向角規劃出最佳的導航路徑為主要研究重點,其中車子與子目標或目標之間的方向角是由電子羅盤來提供。在此我們利用一套基於人工智慧的導航方法來完成,使車子可以安全地繞過障礙物並在最佳路徑前提下朝向目標前進。透過實驗已可證明本系統可以實際在校園內外的人行道上行走,證明所提方法之可行性。
In this thesis, we have developed an implement image process method using watershed transformation and mathematical morphology to detect road by hue, saturation and intensity information. Then we use binocular stereovision system and artificial intelligence (AI) policy to the autonomous land vehicle (ALV) can be navigated at the campus and the pavement. The system is divided into local navigation and global navigation. Before the navigation, the camera calibration is necessary. We employ the linear least-square method to obtain calibration parameters of the left and the right cameras using eight known 3D points and image points projected from real world into cameras. Then we can reconstruct the 3D information from the image points of two cameras by using the calibrated parameters. Road detection, sub-goal and goal searching are the focuses of the study in navigation. Because the sceneries in the outdoor environment are very complicated, so we need a powerful image segmentation method to find the road candidates. To find the candidates, in our study we use watershed transformation to conform our needs. However, the transformation needs an obvious gradient in the image. Dilation, erosion, opening, and closing are the morphological operations. We use these operations to implement the gradient and reduce the too small information that we do not need in the image. For this reason we can segment several regions of road by watershed transformation. Firstly, since the general road has consistent hue and lower saturation, we employ these features to characterize the candidate region. Meanwhile, we need to use stereovision to obtain 3D information of the environment. So the stereo correspondence is necessary. In the study of the sensor-like points approach [1] is employed to obtain corresponding points and use them to reconstruct the 3D information. In the sub-goal and the goal searching, the string match approach is employed to find out them. In the navigation and path planning, we define some features such as length, angle, area and distance of the objects as a map. Following the map, we find the sub-goal and the goal. In here the sub-goal is a pair of rail lines and the goal is the cross lines on the wall beside the road. When they are found out, it indicates that the ALV has arrived at the position of the sub-goal or the goal and runs toward the goal or stop. In global navigation, the optimal path planning based on road information and the direction between ALV and the sub-goal or the goal is the study focus. The direction between ALV and the sub-goal or the goal is obtained by an E-compass. We also employ an AI-based navigation method to obtain the angle where ALV has to turn and make ALV avoid the obstacle safely and run toward the sub-goal or the goal in an appropriate path. The ALV system has been performed in the campus and the pavement to demonstrate the effectiveness of the presented method.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。