-
公开(公告)号:US09152857B2
公开(公告)日:2015-10-06
申请号:US14326812
申请日:2014-07-09
发明人: Hyung II Choi , Sung II Joo , Sun Hee Weon
CPC分类号: G06K9/00389 , G06K9/00201 , G06K9/6257
摘要: A system for detecting an object is provided. The system includes a depth image receiver that receives a depth image from a depth camera; a strong classifier that classifies an object region and a non-object region in the depth image based on a characteristic of an object; and an object detector that detects the classified object region, wherein the strong classifier comprises a plurality of weak classifiers which are cascade connected to each other and classifies the object region and the non-object region by passing the depth image through the weak classifiers, the characteristic of the object is extracted based on a center depth value of the depth image, and the plurality of the weak classifiers are generated through a training process for classifying positive training images among a multiple number of positive training images and a multiple number of negative training images.
摘要翻译: 提供了一种用于检测物体的系统。 该系统包括从深度摄像机接收深度图像的深度图像接收器; 一种强分类器,其基于对象的特征对深度图像中的对象区域和非对象区域进行分类; 以及检测分类对象区域的对象检测器,其中所述强分类器包括彼此级联的多个弱分类器,并且通过使深度图像通过弱分类器来对对象区域和非对象区域进行分类, 基于深度图像的中心深度值提取对象的特征,并且通过训练处理生成多个弱分类器,用于对多个正训练图像和多个负训练中的正训练图像进行分类 图片。
-
公开(公告)号:US20150016676A1
公开(公告)日:2015-01-15
申请号:US14326812
申请日:2014-07-09
发明人: Hyung Il Choi , Sung Il Joo , Sun Hee Weon
CPC分类号: G06K9/00389 , G06K9/00201 , G06K9/6257
摘要: A system for detecting an object is provided. The system includes a depth image receiver that receives a depth image from a depth camera; a strong classifier that classifies an object region and a non-object region in the depth image based on a characteristic of an object; and an object detector that detects the classified object region, wherein the strong classifier comprises a plurality of weak classifiers which are cascade connected to each other and classifies the object region and the non-object region by passing the depth image through the weak classifiers, the characteristic of the object is extracted based on a center depth value of the depth image, and the plurality of the weak classifiers are generated through a training process for classifying positive training images among a multiple number of positive training images and a multiple number of negative training images.
摘要翻译: 提供了一种用于检测物体的系统。 该系统包括从深度摄像机接收深度图像的深度图像接收器; 一种强分类器,其基于对象的特征对深度图像中的对象区域和非对象区域进行分类; 以及检测分类对象区域的对象检测器,其中所述强分类器包括彼此级联的多个弱分类器,并且通过使深度图像通过弱分类器来对对象区域和非对象区域进行分类, 基于深度图像的中心深度值提取对象的特征,并且通过训练处理生成多个弱分类器,用于对多个正训练图像和多个负训练中的正训练图像进行分类 图片。
-