摘要
手部形状测量是人体测量的重要组成部分,也是身份认证的重要方法之一,基于二维图像的手部测量已经得到了广泛的研究与应用。提出一种新的手部图像测量点自动识别方法。该方法基于移动最小二乘图像配准技术;采用DOS方法提取两幅图像的匹配点;通过建立手部像素点集的8邻域无向带权图,在计算MLS配准权值时采用基于图Dijkstra最短路径距离来代替欧式距离。实验结果表明,与传统的基于特征提取和统计的方法相比,该方法具有统一、简洁、高效的特点,同时在测量点的识别率、鲁棒性以及精度方面也有明显提高。
Hand shape measurement is an important part of anthropometry, and also a way of identity authentication. The non--contact hand measurement based on 2D image is widely studied and applied. This paper proposes a novel automatic method of hand measurement oint recognition. The method is based on the MLS image registration teehnique; padopts the DOS（Different Of Slope） method to get the corresponding points of any two images; an undirected weighted graph based on the hand pixel point set is created, and the Euclidean distance is replaced by the Dijkstra shortest path to calculate the weight function. Compared with the traditional method based on the feature extraction and statistics, the experimental results show that the method features conciseness, uniformity, and efficiency. The robustness, accuracy and recognition rate of the method is also prominently improved.
出处
《计算技术与自动化》
2010年第1期88-94,共7页
Computing Technology and Automation
基金
国家科技支撑计划重点项目资助(2008BAH29B00)
