公开(公告)号：WO2014189578A3

公开(公告)日：2015-02-26

申请号：PCT/US2014018794

申请日：2014-02-26

Applicant: QUALCOMM INC

Inventor： KIM KIYOUNG

IPC: G06T7/20

CPC classification number: G06T7/2033 ,  G06T7/246 ,  G06T2207/20021 ,  G06T2207/30244

Abstract: Embodiments include selecting edgels for edge based tracking by dividing (210) a reference image frame (RF) into N x M bins of pixels and projecting (230) a subset of the edgels per bin into a current image frame (CF) using an estimated pose to identify (240) valid bins of the RF as bins having their projected one edgel found within the borders of the CF. Then, K edgels of RF with different orientations from each valid bins may be selected (270). Then, the selected RF edgels of bins may be reduced (280) by removing bins randomly, or first removing bins from the center of the RF of the image (then next removing the next further outward bins), until a desirable edgel number is obtained. Edge-based tracking (290) can then be performed using the desirable edgel number, to track edges in current frame that are found in prior frame.

Abstract translation: 实施例包括通过将（210）参考图像帧（RF）划分成N×M个像素块并且使用估计的每个块将每个块的边缘的子集投影（230）到当前图像帧（CF）中（230）来选择用于基于边缘的跟踪的边缘。 确定（240）RF的有效区域作为在CF的边界内发现其预计的一个边缘的垃圾箱。 然后，可以选择来自每个有效箱体的具有不同取向的RF边缘（270）。 然后，可以通过随机去除箱子，或者首先从图像的RF的中心去除箱（然后去除下一个进一步的另外的箱）来减少（280）所选择的垃圾桶，直到获得期望的反胶束数 。 然后可以使用所需的边缘编号来执行基于边缘的跟踪（290），以跟踪在现有帧中发现的当前帧中的边缘。

公开(公告)号：EP3295375A1

公开(公告)日：2018-03-21

申请号：EP16722990

申请日：2016-04-28

Applicant: QUALCOMM INC

Inventor： WAGNER DANIEL ,  KIM KIYOUNG

IPC: G06K9/62

CPC classification number: G06K9/4671 ,  G06K9/52 ,  G06K9/6211 ,  G06K9/6215 ,  G06K2009/3291 ,  G06K2009/4666 ,  G06T3/0093 ,  G06T7/20 ,  G06T2207/20048

Abstract: Disclosed embodiments facilitate keypoint selection in part by assigning a similarity score to each candidate keypoint being considered for selection. The similarity score may be based on the maximum measured similarity of an image patch associated with a keypoint in relation to an image patch in a local image section in a region around the image patch. A subset of the candidate keypoints with the lowest similarity scores may be selected and used to detect and/or track objects in subsequent images and/or to determine camera pose.