具有影像特徵辨識之視覺系統於人形機器人的設計與實現

Design and Implementation of Vision System with Feature Recognition for Humanoid Robot

洪志輝(Chih-Hui Hung)

指導教授 ： 翁慶昌

淡江大學/工學院/電機工程學系碩士班/碩士(2010年)

https://doi.org/10.6846/TKU.2010.00960

摘要

針對小型人形機器人，本論文提出一個有效整合機器人系統的架構，此機器人系統擁有共同的資料庫以及統一的機器人變數名稱，使得於機器人各個子系統間溝通能夠更加順利。此外，本論文在此架構下設計實現一個具有影像特徵辨識的視覺系統。本論文使用尺度不變特徵轉換(Scale Invariant Feature Transform, SIFT)作為影像特徵辨識的演算法，以及使用K-d-BBF Tree作為影像特徵比對的演算法。從實驗結果可知，本論文所設計實現的方法確實可以有效的辨識影像特徵，並且能夠準確的辨識出目標物。

關鍵字

視覺系統 ； 尺度不辨特徵轉換 ； 特徵辨識 ； 特徵比對 ； 人形機器人

並列摘要

In this thesis, a system architecture is proposed to effectively integrate a robotic system for small-size humanoid robot. This system has a common database and variable names are unified so that the communication between each subsystems in the robotic system can be more smoothly. In addition, a vision system with the abilities of feature recognition and matching is designed and implemented in this system architecture. Scale Invariant Feature Transform (SIFT) is used as the image feature recognition algorithm and Kd-BBF Tree is used as the image feature matching algorithm. From the experimental results, we can see that the design and implementation of the method can effectively identify image features, and can accurately identify the target.

並列關鍵字

Vision System ； SIFT ； Feature Recognition ； Feature Matching ； Humanoid Robot

參考文獻

[1] D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” International Journal of Computer Vision, vol. 60, no. 2, pp. 91-110, Jan. 2004.

[2] T. Lindeberg, “Scale-space theory: A basic tool for analyzing structures at different scales,” Journal of Applied Statistics, vol. 21, no. 2, pp. 224-270, Feb. 1994.

[3] D. G. Lowe, “Invariant features from interest point groups,” British Machine Vision Conference, pp. 656-665, 2002.

[4] J. S. Beis and D. G. Lowe, “Shape indexing using approximate nearest-neighbour search in high-dimension spaces,” Procedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1000- 1006, Jun. 1997.

[6] 石兆孙，高維度空間中使用多棵K-d Tree 搜尋最近鄰居，淡江大學資訊工程學系碩士論文，民98。