【作者】 林精敦； 王涌天； 陈靖； 刘越； 郭俊伟； 刘伟； 薛康；

【Author】 Lin Jingdun,Wang Yongtian,Chen Jing,Liu Yue,Guo Junwei,Liu Wei,Xue Kang (School of Optics and Electronics,Beijing Institute of Technology,Beijing 100081)

【机构】 北京理工大学光电学院；

【摘要】 本文提出一种基于区域分割的随机树分类算法。随机树分类算法是一种有监督学习的模式识别分类算法,可有效地应用于增强现实系统中的特征识别与匹配。为克服现有随机树算法中特征点选择的随意性,使其更具自适应性,本文提出利用区域分割算法改进现有随机树的离线训练机制,提高随机树分类器的鲁棒性和自适应性。实验结果表明,将该改进算法应用于增强现实跟踪定位系统中,显著提高了系统的鲁棒性。更多还原

【Abstract】 We present a novel segmentation randomized tree based classification algorithm in this paper.Randomized tree is a supervised classification algorithm for pattern recognition,which can be effectively used in augmented reality feature recognition and matching.In order to overcome the randomness which exists in the randomized tree algorithm,this paper proposed a regional segmentation algorithm to improve the robustness and adaptability of the randomized tree algorithm.Experimental results show that the improved algorithm used in augmented reality tracking system can significantly improve the robustness of the system.更多还原