首页> 外国专利> Method for feature extraction using locally linear transformations, and method and apparatus for image recognition employing the same

Method for feature extraction using locally linear transformations, and method and apparatus for image recognition employing the same

机译:使用局部线性变换的特征提取方法以及使用该方法的图像识别方法和装置

摘要

Characterized by a plurality of local linear transform function vector extraction method and apparatus and image recognition method and apparatus using the same are disclosed. Local linear way feature vector extraction using a conversion function of (a) class learning image of a second predetermined number of separated into local groups and the average vector of each local group that is separate from the local linear conversion function consisting of a first predetermined number generating and stored in the database; (B) by comparing the input image vector and the average vector of each of the local group assigning the input image to the local group; And (c) a step of extracting a feature vector by the vector projection locally linear transformation function of the local group assigned to the input image. Accordingly, it classified as a local group of the predetermined number of the degree of change is large for a number of data structures having parent dollar tea distribution according to the pose and lighting, obtained by learning the local linear transformation function for each local group, the local linear by extracting a feature vector of the registration image and the recognized image by using the conversion function it can recognize the image to greater accuracy.
机译:公开了以多个局部线性变换函数为特征的矢量提取方法和装置以及使用该方法的图像识别方法和装置。使用(a)被划分为局部组的第二预定数量的类学习图像的转换函数以及与由第一预定数量组成的局部线性转换函数分开的每个局部组的平均矢量的转换函数来提取局部线性方式特征向量生成并存储在数据库中; (B)通过比较输入图像矢量和将输入图像分配给本地组的每个本地组的平均矢量; (c)通过分配给输入图像的局部群的向量投影局部线性变换函数提取特征向量的步骤。因此,对于通过学习每个局部组的局部线性变换函数而获得的,根据姿势和照明度具有父元茶分布的多个数据结构,将其分类为具有预定数量的变化程度的局部组是较大的,通过使用转换函数提取配准图像和识别图像的特征向量,局部线性可以更准确地识别图像。

著录项

  • 公开/公告号KR100571813B1

    专利类型

  • 公开/公告日2006-04-17

    原文格式PDF

  • 申请/专利权人

    申请/专利号KR20030052131

  • 发明设计人 김태균;

    申请日2003-07-28

  • 分类号G06K9/52;

  • 国家 KR

  • 入库时间 2022-08-21 21:23:55

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号