利用鬆弛相關性改進3D 對應點匹配與重建應用於室外自動導航車

Using Relaxed Correlation to Improve Corresponding of 3D Reconstruction for Outdoor Guidance of Autonomous Land Vehicle

葉哲昌(Che-Chang Ye)

指導教授 ： 駱榮欽

國立臺北科技大學/機電學院/自動化科技研究所/碩士(2006年)

摘要

在本論文中，提出利用立體極線限制和鬆弛相關性改進左右影像對應點匹配的演算法重建3D道路資訊，並應用於室外自動車導航。影像由架設在室外自動導航車上的兩台攝影機所擷取。由於室外環境的複雜，光線強度不易控制，在室外影像中找到正確的對應點匹配是困難且花費時間的。克服此困難，論文中採用改進的特徵點對應匹配，以及立體視覺和人工智慧搭配模糊控制做自動車導航，以期可行走在室外道路上。在導航開始前，攝影機校正是必要的，我們考慮了透鏡失真，並利用特定8點的3D位置及投射到左右攝影機中之影像點以求得左右攝影機參數，之後我們可以利用這些參數及左右影像點進行3D重建。
在特徵點對應匹配方面，首先，要先將兩張左右影像的特徵點找出來，再從左影像的特徵點我們可以在右影像上畫出一條極線(epipolar line)，極線附近的特徵點極有可能是右影像對應到左影像上相同位置的特徵點，用以減少不必要的搜尋，降低運算時間。第二、利用特徵點週遭的灰階強度相關性(correlation)來找出候選點。第三、將特徵點與週遭特徵點之間的鬆弛相關性做比較，找出最相近的特徵點。得到對應點匹配以及左右攝影機參數之後，我們便可得到該對應點的3D資訊，利用所得到的3D資訊進行道路擴張，來求得目前道路區域，並區分出障礙物等非道路資訊。
得到道路區域之後，我們利用電子羅盤規劃出目標方向並搭配人工智慧做出路徑規劃，配合模糊控制做車輛轉向，使車子可以安全繞過障礙物並在最佳路徑前提下朝目標前進。透過實驗已可證明本系統可以實際在室外道路上行走，證明所提方法之可行性。

關鍵字

立體視覺 ； 攝影機校正 ； 特徵點 ； 極線 ； 相關性 ； 鬆弛相關性 ； 對應點匹配 ； 自動車 ； 人工智慧

並列摘要

In this paper, using epipolar constraint and relaxed correlation we propose an improved algorithm of two images corresponding. Two images are captured from two cameras, which are put on an outdoor Autonomous Land Vehicle (ALV). Since environment of outdoor is complex, and light is not easy to control it is difficult and time-consuming to get correct corresponding points. In order to surmount this problem, we improve the corresponding algorithm with feature points. Then we use binocular stereovision system and artificial intelligence (AI) with fuzzy control policy to navigate the ALV at the road of outdoor. Before the navigation, the camera calibration is necessary. We consider the lens distortion and using eight known 3D points and image points projected from real world into cameras to obtain calibration parameters of the left and the right cameras. Then we can reconstruct the 3D information from the image points of two cameras by using the calibrated parameters.
In the stereo corresponding, first we should find out the feature point from the left and right images buffers. And then, a feature point in left image, we can draw the epipolar line in right image. The right feature points which adjacent to the epipolar line are most be the corresponding point with left feature point, that reduce searching time and computing time. Second, perform the correlation with the intensity of the neighboring pixels. Third, compare with the relaxed correlation result, and find out the optimal corresponding point candidates. After stereo corresponding and camera calibration, we can get the 3D information to exact the desired road region by road expanded algorithm.
After road searching, the direction of navigation is obtained by an digital compass. We employ an AI-based navigation with fuzzy control 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 outdoor to demonstrate the effectiveness of the presented method.

並列關鍵字

Stereo Vision ； Camera Calibration ； Feature point ； Epipolar Line ； Correlation ； Relaxed Correlation ； Stereo Corresponding ； Autonomous Land Vehicle ； Artificial Intelligence

參考文獻

[4] 陳冠州，以型態學搭配分水嶺為基礎偵測室外路面及障礙物作自動車導航，碩士論文，國立台北科技大學機電整合研究所，台北，2005。

[8] Roger Y. Tsai, “A Versatile Camera Calibration Technique for High-Accuracy 3D Machine Vision Metrology Using Off-the-shelf TV Cameras and Lenses”, IEEE Journal of Robotics and Automation, vol. ra-3, no. 4 August 1987, pp 323-344.

[14] Chris Harris and Mike Stephens, “A Combed Corner And Edge Detector”, The Plessey Computer plc, 1988, pp 147-151.

[15] Zhengyou Zhang, “A Robust Technique for Matching Two Uncalibrated Images Through the Recovery of the Unknown Epipolar Geometry”, Institut National De Recherche En Informatique Et En Automatique, 1994, pp 87-119.

[17] Kevin M. Passino and Stephen Yurkovich, Fuzzy Control, Addison Wesley Longman, 1998, pp 21-169.