白丝美女被狂躁免费视频网站,500av导航大全精品,yw.193.cnc爆乳尤物未满,97se亚洲综合色区,аⅴ天堂中文在线网官网

Rapid point cloud alignment and classification with basis set learning

專利號(hào)
US11176693B1
公開日期
2021-11-16
申請(qǐng)人
Amazon Technologies, Inc.(US WA Seattle)
發(fā)明人
Javier Romero Gonzalez-Nicolas; Sergey Prokudin; Christoph Lassner
IPC分類
G06T7/50; G06T17/00; G06K9/62
技術(shù)領(lǐng)域
cloud,data,point,may,mesh,points,basis,or,e.g,generate
地域: WA WA Seattle

摘要

A system configured to process an input point cloud, which represents an object using unstructured data points, to generate a feature vector that has an ordered structure and a fixed length. The system may process the input point cloud using a basis point set to generate the feature vector. For example, for each basis point in the basis point set, the system may identify a closest data point in the point cloud data and store a distance value or other information associated with the closest data point in the feature vector. The system may process the feature vector using a trained model to generate output data, such as performing point cloud registration to generate mesh data, point cloud classification to generate classification data, and/or the like.

說明書

In some examples, the system 100 may perform point cloud registration (e.g., the point cloud 112 is registered to a common template) and generate output data such as mesh data 118, which may correspond to an output mesh, 3D scan, 3D model, deformable model, reposable human avatar, and/or the like. Thus, the mesh data 118 may include detailed geometry and/or color information representing an appearance of the user 5 and may be used to represent the user 5 using 3D surfaces. However, while these terms are closely related, there are slight differences between them and therefore these terms may have a specific meaning. For example, the system 100 may generate depth information corresponding to the point cloud 112 (or a structured data representation of the point cloud 112), which may be referred to as the scan. The system 100 may then perform an alignment process to deform a set of free body vertices (referred to as the mesh) so that they are close to the point cloud 112 (or the structured data representation of the point cloud 112) while also being likely according to a statistical model of human shape and pose. Thus, the scan (e.g., 3D scan) may refer to the depth information, with or without color information, whereas the mesh (e.g., output mesh, 3D model, human avatar) may refer to a deformable model or reposable avatar.

權(quán)利要求

1
微信群二維碼
意見反饋