公开(公告)号：US09710706B2

公开(公告)日：2017-07-18

申请号：US14493951

申请日：2014-09-23

Applicant: GM GLOBAL TECHNOLOGY OPERATIONS LLC

Inventor： Stephane Francois ,  James W. Wells ,  Neil David McKay

IPC: G06K9/00 ,  G06K9/62 ,  G06T17/10 ,  G06T19/20 ,  H04N13/02 ,  G06K9/32 ,  G06T7/73

CPC classification number: G06K9/00536 ,  G06K9/00201 ,  G06K9/00718 ,  G06K9/00771 ,  G06K9/3208 ,  G06K9/6267 ,  G06K9/6269 ,  G06T7/74 ,  G06T17/10 ,  G06T19/20 ,  H04N13/211 ,  H04N13/218 ,  H04N13/275

Abstract: A method for classifying a known object in a field of view of a digital camera includes developing a plurality of classifier feature vectors, each classifier feature vector associated with one of a plurality of facet viewing angles of the known object. The digital camera captures an image in a field of view including the known object and an image feature vector is generated based upon said captured image. The image feature vector is compared with each of the plurality of classifier feature vectors and one of the plurality of classifier feature vectors that most closely corresponds to the image feature vector is selected. A pose of the known object relative to the digital camera is determined based upon the selected classifier feature vector.

公开(公告)号：US20160086028A1

公开(公告)日：2016-03-24

申请号：US14493951

申请日：2014-09-23

Applicant: GM Global Technology Operations LLC

Inventor： Stephane Francois ,  James W. Wells ,  Neil David McKay

IPC: G06K9/00 ,  H04N13/02 ,  G06T19/20 ,  G06K9/62 ,  G06T17/10

CPC classification number: G06K9/00536 ,  G06K9/00201 ,  G06K9/00718 ,  G06K9/00771 ,  G06K9/3208 ,  G06K9/6267 ,  G06K9/6269 ,  G06T7/74 ,  G06T17/10 ,  G06T19/20 ,  H04N13/211 ,  H04N13/218 ,  H04N13/275

Abstract: A method for classifying a known object in a field of view of a digital camera includes developing a plurality of classifier feature vectors, each classifier feature vector associated with one of a plurality of facet viewing angles of the known object. The digital camera captures an image in a field of view including the known object and an image feature vector is generated based upon said captured image. The image feature vector is compared with each of the plurality of classifier feature vectors and one of the plurality of classifier feature vectors that most closely corresponds to the image feature vector is selected. A pose of the known object relative to the digital camera is determined based upon the selected classifier feature vector.

Abstract translation: 用于在数字照相机的视野中对已知对象进行分类的方法包括开发多个分类器特征向量，每个分类器特征向量与已知对象的多个小平面视角之一相关联。 数字照相机在包括已知对象的视野中拍摄图像，并且基于所述捕获的图像生成图像特征向量。 将图像特征向量与多个分类器特征向量中的每一个进行比较，并且选择与图像特征向量最接近的多个分类器特征向量中的一个。 基于所选择的分类器特征向量来确定已知对象相对于数字照相机的姿态。