• 欢迎访问速搜资源吧,如果在网站上找不到你需要的资源,可以在留言板上留言,管理员会尽量满足你!

【速搜问答】面部跟踪是什么

问答 admin 9个月前 (04-13) 170次浏览 已收录 0个评论

汉英对照:
Chinese-English Translation:

面部跟踪,就是要在检测到人脸的前提下,在后续帧中继续捕获人脸的位置及其大小等信息,包括人脸的识别和人脸的跟踪技术。

Face tracking is to continue to capture the position and size of the face in the subsequent frame on the premise of detecting the face, including face recognition and face tracking technology.

面部跟踪,就是要在检测到人脸的前提下,在后续帧中继续捕获人脸的位置及其大小等信息,包括人脸的识别和人脸的跟踪技术。

Face tracking is to continue to capture the position and size of the face in the subsequent frame on the premise of detecting the face, including face recognition and face tracking technology.

介绍

introduce

在海关、机场、银行、电视电话会议等场合,都需要对特定人脸目标进行跟踪。显然,要跟踪图象中的人脸。首先要识别人脸。人脸识别就是利用计算机分析静态图片或视频序列。从中找出人脸并输出人脸的数目、位置及其大小等有效信息。其次就是跟踪人脸。就是要在检测到人脸的前提下。在后续帧中继续捕获人脸的位置及其大小等信息。人脸跟踪技术涉及到模式识别、图象处理、计算机视觉、生理学、心理学及形态学等诸多学科。并与基于其它生物特征的身份鉴别方法以及计算机人机感知交互的研究领域密切相关。与基于指纹、视网膜、虹膜、DNA 等其它人体生物特征识别系统相比。人脸跟踪技术更为直接、友好。不会对用户造成心理障碍。此外,人脸跟踪技术研究与相关学科的发展及对人脑的认识程度紧密相关。这诸多因素使人脸跟踪研究成为一项既困难又极富挑战性的课题。

In customs, airport, bank, teleconference and other occasions, it is necessary to track specific face targets. Obviously, we need to track the face in the image. Face recognition is the first step. Face recognition is the use of computer analysis of static pictures or video sequences. The number, location and size of faces are output. The second is face tracking. It’s on the premise of face detection. The position and size of the face are captured in subsequent frames. Face tracking technology involves many subjects such as pattern recognition, image processing, computer vision, physiology, psychology and morphology. And it is closely related to other biometric identification methods and the research field of computer human-computer interaction. Compared with other biometric recognition systems based on fingerprint, retina, iris, DNA and so on. Face tracking technology is more direct and friendly. It will not cause psychological barriers to users. In addition, the research of face tracking technology is closely related to the development of related disciplines and the understanding of human brain. These factors make face tracking a difficult and challenging task.

目前常见的跟踪技术大致可分为 4 大类:基于模型跟踪;基于运动信息跟踪;基于人脸局部特征跟踪和基于神经网络跟踪等方法。

At present, the common tracking technologies can be roughly divided into four categories: model-based tracking, motion information based tracking, face local feature based tracking and neural network based tracking.

方法

method

基于模型跟踪

Model based tracking

基于模型跟踪的方法就是获取目标的先验知识,建立低价参数模型,对输入的每一帧图象通过滑动窗口进行模型匹配,实现人脸识别与跟踪。常见的跟踪模型有:肤色模型、椭圆模型、纹理模型及双眼模板等。

The method of model-based tracking is to acquire the prior knowledge of the target, establish the low-cost parameter model, and match the model of each input image through the sliding window to realize face recognition and tracking. The common tracking models include skin color model, ellipse model, texture model and binocular template.

基于肤色模型的跟踪方法就是利用恰当的表色系统,把肤色作为实现人脸跟踪的关键信息。由于肤色信息具有对放大和缩小以及对微小变形不敏感的优点,加上人脸相对镜头的变化对肤色信息本身的影响不大,该类方法很容易在前一帧图象分析结果的基础上跟踪到后一帧图象的人脸区域,因此具有速度快、姿态不变性等特点。目前的人脸跟踪技术大都采用基于肤色模型的方法。

The tracking method based on skin color model takes skin color as the key information to realize face tracking by using the appropriate color system. Because the skin color information is insensitive to magnification and miniaturization, and the change of face relative to the lens has little effect on the skin color information itself, this kind of method is easy to track the face area of the next frame based on the analysis results of the previous frame, so it has the characteristics of high speed and pose invariance. Most of the current face tracking techniques are based on skin color model.

基于运动信息跟踪

Tracking based on motion information

基于运动信息跟踪法主要是充分利用图象连续帧间目标运动的连续性规律,进行人脸区域的预测以达到快速跟踪的目的。通常采用运动分割、光流、立体视觉等方法。利用时空梯度,卡尔曼滤波器进行跟踪。

Based on motion information tracking method is mainly to make full use of the continuity law of target motion between consecutive frames of image to predict the face region, so as to achieve the purpose of fast tracking. Motion segmentation, optical flow and stereo vision are usually used. Kalman filter is used to track the target.

光流是空间运动物体被观测面上的像素点运动产生的瞬时速度场,包含了物体 3D 表面结构和动态行为的重要信息。一般情况下,光流由相机运动、场景中目标运动、或两者的运动产生。光流分析经常被用于目标运动估计。当场景中有独立的运动目标时,通过光流分析可以确定运动目标的数目、运动速度、目标距离和目标的表面结构。光流分析可以分为连续光流法和特征光流法,特征光流法是通过特征匹配求得特征点处的光流。

Optical flow is the instantaneous velocity field generated by the motion of pixels on the observed surface of a space moving object, which contains important information about the 3D surface structure and dynamic behavior of the object. Generally, optical flow is generated by camera motion, target motion in the scene, or both. Optical flow analysis is often used for target motion estimation. When there are independent moving targets in the scene, the number, speed, distance and surface structure of moving targets can be determined by optical flow analysis. Optical flow analysis can be divided into continuous optical flow method and feature optical flow method. Feature optical flow method is to obtain the optical flow at feature points through feature matching.

基于人脸局部特征跟踪

Face tracking based on local features

基于人脸局部特征跟踪法的主要思想是根据不同的人脸器官特征信息进行器官跟踪。这类方法经常利用眼睛、嘴和鼻子等器官特征信息进行跟踪定位。传统的人脸特征点跟踪方法通常是在人面部画上标识点进行跟踪,如 Kouadio 等提出了一种通过加标识点来跟踪视频中人脸特征点的方法,采用了一种分析人面部曲线的方法来跟踪人脸。

The main idea of face local feature tracking method is to track organs according to different facial organ feature information. This kind of methods often use the feature information of eyes, mouth, nose and other organs to track and locate. Traditional facial feature point tracking methods usually track the mark points on the face painting, such as kouadio et al. Proposed a method to track the face feature points in video by adding mark points, and used a method to track the face by analyzing the face curve.

基于神经网络跟踪

Tracking based on Neural Network

神经网络是近年来发展较快的一个交叉研究学科。人工神经网络模型具备人脑思维的一些典型特征,如自组织、联想记忆、非线性、大规模并行连接等。并且具有强大的学习能力!将神经网络用于人脸跟踪具有一定的优势。因为要显性地对人脸识别特征进行描述相当困难,而神经网络则可以通过学习,自动地获得识别规律的隐性表达。

Neural network is an interdisciplinary subject which develops rapidly in recent years. Artificial neural network model has some typical characteristics of human brain thinking, such as self-organization, associative memory, nonlinear, large-scale parallel connection and so on. And has a strong learning ability! Neural network for face tracking has certain advantages. Because it is very difficult to describe the features of face recognition explicitly, neural network can automatically obtain the implicit expression of recognition rules through learning.

目前神经网络方法也是人脸识别与跟踪技术中的研究热点。Valentin 提出一种方法,首先提取人脸的 50 个主元,然后用自相关神经网络将它映射到 5 维空间中,再用一个普通的多层感知器进行判别。对一些简单的测试图象效果较好;Intrator 等提出了一种混合型神经网络来进行人脸跟踪,其中非监督神经网络用于特征提取,而监督神经网络用于分类。

At present, neural network method is also a research hotspot in face recognition and tracking technology. Valentin proposes a method, which first extracts 50 principal components of human face, then maps them into 5-Dimensional space by using autocorrelation neural network, and then uses a common multi-layer perceptron to discriminate. For some simple test images, the effect is better; intrator proposed a hybrid neural network for face tracking, in which unsupervised neural network is used for feature extraction and supervised neural network is used for classification.


速搜资源网 , 版权所有丨如未注明 , 均为原创丨转载请注明原文链接:【速搜问答】面部跟踪是什么
喜欢 (0)
[361009623@qq.com]
分享 (0)
发表我的评论
取消评论
表情 贴图 加粗 删除线 居中 斜体 签到

Hi,您需要填写昵称和邮箱!

  • 昵称 (必填)
  • 邮箱 (必填)
  • 网址