首页> 外国专利> METHOD AND DEVICE FOR PATTERN LEARNING AND PATTERN RECOGNITION USING ATTRACTION AND REPULSION, AND PATTERN LEARNING PROGRAM, PATTERN RECOGNITION PROGRAM AND RECORDING MEDIUM WITH THE PROGRAMS RECORDED

METHOD AND DEVICE FOR PATTERN LEARNING AND PATTERN RECOGNITION USING ATTRACTION AND REPULSION, AND PATTERN LEARNING PROGRAM, PATTERN RECOGNITION PROGRAM AND RECORDING MEDIUM WITH THE PROGRAMS RECORDED

机译:利用吸引力和排斥力进行模式学习和模式识别的方法和装置,以及记录模式的模式学习程序,模式识别程序和记录介质

摘要

PROBLEM TO BE SOLVED: To learn/recognize the variation among images without intervention of a human or teacher signals. SOLUTION: A resultant force calculating means 2 calculates the mutual reaction of the attraction and the repulsion defined as factors of the distance or similarity in a plurality of image arrays. A merging means 3 executes the merging process for each image array without previous information on the category based on the dynamic stability of the cluster calculated with the two kinds of forces calculated by the resultant force calculating means 2. A splitting means 4 executes the splitting process for each merged image array based on the dynamic stability of the cluster calculated with the two kinds of forces calculated by the resultant force calculating means 2 to obtain new image arrays. A repeating means 5 repeats the processes of the merging means 3 and the splitting means 4 for each new image array to obtain a set of new images. A clustering means 6 automatically clusters the set of images obtained by the repeating means 5 by category.
机译:要解决的问题:学习/识别图像之间的差异,而无需人工或教师信号的干预。解决方案:合力计算装置2计算被定义为多个图像阵列中的距离或相似性的因素的吸引力和排斥力的相互反应。合并装置3基于由合力计算装置2计算出的两种力计算出的聚类的动态稳定性,对每个图像阵列执行合并处理,而没有关于类别的先前信息。分裂装置4执行分裂处理。对于每个合并图像阵列,基于由合力计算装置2计算出的两种力计算出的簇的动态稳定性,从而获得新的图像阵列。重复装置5针对每个新图像阵列重复合并装置3和分割装置4的处理,以获得一组新图像。聚类装置6按类别自动对由重复装置5获得的图像集合进行聚类。

著录项

  • 公开/公告号JP2003085557A

    专利类型

  • 公开/公告日2003-03-20

    原文格式PDF

  • 申请/专利权人 NIPPON TELEGR & TELEPH CORP NTT;

    申请/专利号JP20010277563

  • 发明设计人 RAYCHEV BISSEL;MURASE HIROSHI;

    申请日2001-09-13

  • 分类号G06T7/00;G06N3/00;

  • 国家 JP

  • 入库时间 2022-08-22 00:14:44

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