結合邊緣特徵資訊之半全域式匹配法

Semi-Global Matching by Integrating Edge Feature Information

李欣錡(Hsin-Chi Lee)

指導教授 ： 趙鍵哲

國立臺灣大學/工學院/土木工程學研究所/碩士(2016年)

https://doi.org/10.6342/NTU201602993

摘要

半全域式匹配法，已廣泛施用於利用拍攝影像重建真實場景之三維空間重建的任務。然而，半全域式匹配法對懲罰參數的高敏感性，使得在幾何邊緣處容易出現帶狀的匹配錯誤，且其錯誤量級也較其他區域大。為了解決上述問題，本研究透過Edge Drawing偵測邊緣特徵，以及改良廣義霍夫轉換獲得共軛邊緣特徵，並將該資訊引入半全域式匹配法作業。共軛之邊緣特徵資訊除了作為約制條件之外，也用於決定懲罰參數。於實驗成果證實本研究方法可以修正位於物體邊緣處的匹配錯誤，有效提高視差成果品質，以支援後續三維點雲產製。

關鍵字

邊緣偵測 ； 邊緣匹配 ； 改良廣義霍夫轉換 ； 懲罰參數 ； 半全域式匹配法

並列摘要

Semi-global matching has been widely applied to 3D space reconstruction of real scenes. However, due to high sensitivity to penalty parameters, semi-global matching easily incurs significant amount of matching errors on geometric than other regions. To solve the problems mentioned above, the author detected edges by Edge Drawing, matched edges by modified Generalized Hough Transform, and added matched edge features into semi-global matching optimization. The edge features do not only serve as discontinuity constraints, but also as bases to decide penalty parameters. Through the experiments, it was proved that the disparity errors on discontinuity were finely refined, and the disparity map quality was also effectively improved to support 3D point clouds generation.

並列關鍵字

Edge Detection ； Edge Matching ； Modified Generalized Hough Transform ； Penalty Parameters ； Semi-Global Matching

參考文獻

丁皓偉，2014。結合十字區塊匹配之半全域式匹配法優化作業，臺灣大學土木工程學研究所學位論文，1-162。

Ballard, D. H., 1981. Generalizing the Hough transform to detect arbitrary shapes. Pattern recognition, 13(2), 111-122.

Bay, H., Ess, A., Tuytelaars, T., and Gool, L.V., 2008. SURF: Speeded Up Robust Features. Computer Vision and Image Understanding, 110(3), pp. 346-359.

Boykov, Y., Veksler, O., and Zabih, R., 2001. Fast approximate energy minimization via graph cuts, IEEE Transactions on Pattern Analysis and Machine Intelligence, 23(11): 1222-1239.

Canny, J., 1986. A computational approach to edge detection, IEEE Transactions on Pattern Analysis and Machine Intelligence, (6): 679-698.