A method for face detection, comprising:
capturing a depth map, comprising an array of three-dimensional coordinates, and a two-dimensional image of a scene;
selecting one or more locations in the image to test for presence of human faces;
at each selected location, defining a respective face detection window having a size that is scaled according to an expected face size, in pixels of the image, at a depth of the selected location that is indicated by the depth map; and
processing a part of the image that is contained within each face detection window to determine whether the face detection window contains a human face.
Described herein are various technologies pertaining to utilizing a看戏看戏 XD
commodity RGB camera and depth sensor to generate, in a semi-autonomous
manner, a relatively accurate virtual three-dimensional model of a
head/face of a user. Such three-dimensional model, in an exemplary
embodiment, can be employed as a virtualized representation of the
individual, which can identify such individual in an online/gaming
environment. The three-dimensional model can relatively accurately
represent a shape of a head of the user, a size of the head of the user,
coloring of skin of the user, and facial features of the user,
including eyes, nose, ears, cheek lines, mouth, etc.
To generate a relatively accurate model of a head/face of a user, a
plurality of RGB frames are received from a color camera over a range of
time, and a plurality of depth frames are received from a depth sensor
over the range of time. Both the camera and the depth sensor are
configured to capture images of the head/face of the user. The RGB
frames are subjected to a face tracking algorithm to identify features
in the face captured in the RGB frames. Such features can include, but
are not limited to, centers of the eyes of the user, center of the nose
of the user, and nasal alars of the user. A point cloud can be generated
based upon the depth frames and the features identified in the RGB
frames, wherein the features in the RGB frames are mapped to depth
frames to locate their positions in three-dimensional space, and are
subsequently used to align depth frames. Points in the point cloud are
representative of respective distances of the head/face of the user from
the depth sensor at such points.
1. 一种方便构造用户的头部的计算机实现的三维表示的方法，所述方法包括： 从相机接收所述用户的头部的多个RGB帧，所述多个RGB帧是由所述相机在一时间范围内捕捉到的。 从深度传感器接收多个深度帧，所述深度帧指示所述用户的头部的各个部分离所述深 度传感器的距离，所述深度帧是由所述深度传感器在所述时间范围内生成的； 标识所述多个RGB帧中的所述用户的头部的至少一个特征，所述至少一个特征是以下各项之一：所述用户的眼睛的中心、所述用户的鼻子的中心、所述用户的第一鼻翼或所述用户的第二鼻翼； 至少部分地基于所述多个深度帧以及对所述RGB帧中的所述用户的头部的至少一个特征的标识来生成所述用户的头部的三维网格；以及至少部分地基于所述多个RGB帧来纹理化所述三维网格。我粗粗看了一下戏，权利要求里反复提到的是RGB帧，FaceID有用到这个么？
| 欢迎光临 迅维网 (https://www.chinafix.com/) | Powered by Discuz! X3.4 |
