Active Histogram of Oriented Gradient Based Learning for Free Palm Tracking

Active Histogram of Oriented Gradient Based Learning for Free Palm Tracking
复制标题

用于自由手掌跟踪的基于定向梯度学习的主动直方图

DOI:
10.1007/978-3-642-27552-4_91
复制
发表时间:
2011
期刊:
--
影响因子:
--
通讯作者:
H. Yan
H. Yan
中科院分区:
--
文献类型:
--
作者:
Shuai Zhang;Xiang Chen;Kongqiao Wang;Jiangwei Li;Yanwei Pang;H. Yan

文献摘要

参考文献

被引文献

相似文献

由于手的高度自由度,特别是在低成像条件下,手的检测是计算机视觉中具有挑战性的研究领域。在本文中,我们主要开发了一个新的功能,我们称之为主动直方图的方向梯度(aHOG)的手掌检测在无约束的灰度图像。为了克服HOG的局限性,我们采用局部PCA,这是一个特征合成过程,到原始HOG特征集。因此,输出特征具有更短的描述长度,对光照变化和背景聚类不敏感,而不会有太大的性能损失。然后将这些特征与LBP算法相结合,在线性支持向量机中更好地挖掘手掌信息,进行分类。此外,我们使用一种尺度分割策略来实现快速的手掌跟踪。在我们的实验中,性能被证明是非常有效的红外手掌数据库收集我们自己,其中涉及丰富的面间和面外旋转。
Hand detection is a challenging research field in computer vision due tothe high freedom of hand for discrimination especially under low imaging conditions. In this paper, we mainly develop a novel feature that we called active Histogram of Oriented Gradient (aHOG) for palm detection inunconstrained grey-level images. Toovercome the limitationsof HOG, we apply local PCA, which is a feature synthesisprocedure, to original HOG feature sets. So that the output feature takesshorter description length and are less insensitive to light variations and background clusters, without much performance penalty. Then the features are combined with LBP for better palm information mining in linear SVM for classification. Besides, we use a scale partitionstrategy to achieve fast palm tracking. In our experiments, the performance is demonstrated to be very effective on the infrared palm database collected by ourselves, which involve rich inter-plane and out-of-plane rotations.
DOI: --
发表时间: 2004
期刊: Proc.of 43th Conf.of Japan Society of Medical Electronics and Biological Engineering
影响因子: --
作者:
Y.Hamada;N.Shimada;Y.Shirai
通讯作者: Y.Shirai