专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明专利

JP2012243197A Image identification device, image identification method, image identification program and recording medium 有权
标题翻译：图像识别装置，图像识别方法，图像识别程序和记录介质
公开(公告)号：JP2012243197A
公开(公告)日：2012-12-10
申请号：JP2011114665
申请日：2011-05-23
申请人： Morpho Inc , 株式会社モルフォ , Ntt Docomo Inc , 株式会社エヌ・ティ・ティ・ドコモ
发明人： TANASE YASUSHI , MIURA TAKESHI , TODOROKI YASUHIRO , SHIINO HIROKI , HABUKA KANESUKE , HAMADA TETSUYA , YASUDA YASUYO
IPC分类号： G06T7/00 , H04N5/232 , H04N101/00
CPC分类号： G06K9/00684 , G06F17/3025 , G06K9/4642
摘要： PROBLEM TO BE SOLVED: To provide an image identification device capable of improving accuracy in categorizing a block image.SOLUTION: An image identification device 1, which uses a separate plane to categorize a block image, comprises: a target image input unit 10 to which a target image is inputted; a block image generation unit 11 for generating a block image; a feature amount operation unit 12 for operating a feature amount of the block image; and a category determination unit 13 for determining whether or not the block image is categorized. The feature amount operation unit 12, by not only using a local feature amount of the block image and the whole feature amount of the whole target image but also using a coordinate position of a feature amount vector, which is formed by combining multiple local feature amounts in a second feature amount space with multiple features of the block image set as coordinate axes, and one or more arbitrary areas in the second feature amount space, includes the number of block images, which is obtained through area-by-area counting of the block images with the feature amount vector belonging to the area, in the whole feature amount.
摘要翻译：要解决的问题：提供能够提高对块图像进行分类的精度的图像识别装置。解决方案：使用单独的平面对块图像进行分类的图像识别装置1包括：输入目标图像的目标图像输入单元10; 块图像生成单元11，用于生成块图像; 用于操作块图像的特征量的特征量操作单元12; 以及用于确定块图像是否被分类的类别确定单元13。特征量操作单元12通过不仅使用块图像的局部特征量和整个目标图像的整体特征量，而且使用通过组合多个局部特征量形成的特征量向量的坐标位置在具有设置为坐标轴的块图像的多个特征的第二特征量空间和第二特征量空间中的一个或多个任意区域中，包括通过逐区域计数获得的块图像的数量在特征量的整个特征量中具有属于该区域的特征量矢量的图像。版权所有（C）2013，JPO＆INPIT

2. 发明专利

JP2010045613A Image identifying method and imaging device 有权
标题翻译：图像识别方法和成像装置
公开(公告)号：JP2010045613A
公开(公告)日：2010-02-25
申请号：JP2008208407
申请日：2008-08-13
申请人： Morpho Inc , Ntt Docomo Inc , 株式会社エヌ・ティ・ティ・ドコモ , 株式会社モルフォ
发明人： YASUDA YASUYO , HAMADA TETSUYA , SUZUKI TAKASHI , HIRAGA TADAMOTO , TANASE YASUSHI , SHIINO HIROKI
IPC分类号： H04N5/243 , G06T7/00 , H04N5/232 , H04N101/00
CPC分类号： G06K9/00624 , G06K9/4642 , H04N5/232
摘要： PROBLEM TO BE SOLVED: To suppress a processing cost of a CPU when identifying a target by use of an amount of characteristics of a block image in an image photographed with an imaging device, to identify the target accurately.
SOLUTION: Image data are divided into a plurality of blocks to generate the block images, and characteristic numbers of the respective block images are operated by color spacial information and frequency components of the block image. Further, as teacher data, the characteristic numbers of the image for each category have in advance been operated, and a separate hyperplane serving as a boundary for identifying the category is first computed by use of these characteristic numbers. Then, the characteristic numbers of the image in the block image is similarly computed about the newly obtained image also, and the category to which the block image concerned belongs is determined by a distance from the separate hyperplane of each category.
COPYRIGHT: (C)2010,JPO&INPIT
摘要翻译：要解决的问题：为了在通过使用成像装置拍摄的图像中的块图像的特征量来识别目标时抑制CPU的处理成本，以准确地识别目标。解决方案：图像数据被分成多个块以产生块图像，并且通过色彩空间信息和块图像的频率分量来操作各个块图像的特征号。此外，作为教师数据，预先操作每个类别的图像的特征数，并且首先通过使用这些特征数来计算用作识别类别的边界的单独的超平面。然后，对于新获得的图像也类似地计算块图像中的图像的特征数，并且通过距离每个类别的单独超平面的距离来确定所关注的块图像所属的类别。版权所有（C）2010，JPO＆INPIT

3. 发明专利

JP2012243196A Image identification device, image identification method, image identification program and recording medium 有权
标题翻译：图像识别装置，图像识别方法，图像识别程序和记录介质
公开(公告)号：JP2012243196A
公开(公告)日：2012-12-10
申请号：JP2011114660
申请日：2011-05-23
申请人： Morpho Inc , 株式会社モルフォ
发明人： TANASE YASUSHI , MIURA TAKESHI , TODOROKI YASUHIRO , SHIINO HIROKI , HABUKA KANESUKE
IPC分类号： G06T7/00 , H04N5/232 , H04N101/00
CPC分类号： G06K9/4642 , G06K9/00684
摘要： PROBLEM TO BE SOLVED: To provide an image identification device capable of improving accuracy in classifying a block image that is obtained by dividing a target image.SOLUTION: An image identification device 1, which uses an already-learned separate plane to categorize a block image, according to each preset category, that is obtained by dividing a target image, comprises: a target image input unit 10 to which the target image is inputted; a block image generation unit 11 for dividing the target image into multiple blocks and generating multiple block images; a feature amount operation unit 12 for operating a feature amount of each block image; and a category determination unit 13 for using a coordinate position corresponding to a size of the feature amount of each block image in a feature amount space and the separate plane to determine whether or not each block image is categorized. The feature amount operation unit 12 uses a local feature amount calculated from image information for the block image and the whole feature amount calculated from the image information for the whole target image, as the feature amount of the block image.
摘要翻译：要解决的问题：提供能够提高通过划分目标图像而获得的块图像的分类精度的图像识别装置。解决方案：根据通过划分目标图像获得的每个预设类别，使用已经学习的单独平面来对块图像进行分类的图像识别装置1包括：目标图像输入单元10，目标图像输入单元10 输入目标图像; 块图像生成单元11，用于将目标图像划分为多个块并生成多个块图像; 用于操作每个块图像的特征量的特征量操作单元12; 以及类别确定单元13，用于使用与特征量空间中的每个块图像的特征量的大小相对应的坐标位置和分离平面来确定每个块图像是否被分类。特征量操作单元12使用从块图像的图像信息计算的局部特征量和从整个目标图像的图像信息计算的整体特征量作为块图像的特征量。版权所有（C）2013，JPO＆INPIT

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式