CN101021948A - Automatic identifying device and method for joint in human body symmetric motion image - Google Patents

Automatic identifying device and method for joint in human body symmetric motion image Download PDF

Info

Publication number
CN101021948A
CN101021948A CN 200710027147 CN200710027147A CN101021948A CN 101021948 A CN101021948 A CN 101021948A CN 200710027147 CN200710027147 CN 200710027147 CN 200710027147 A CN200710027147 A CN 200710027147A CN 101021948 A CN101021948 A CN 101021948A
Authority
CN
China
Prior art keywords
image
module
joint
frame
position
Prior art date
Application number
CN 200710027147
Other languages
Chinese (zh)
Inventor
范毅方
聂文良
吕长生
Original Assignee
华南理工大学
广州体育学院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华南理工大学, 广州体育学院 filed Critical 华南理工大学
Priority to CN 200710027147 priority Critical patent/CN101021948A/en
Publication of CN101021948A publication Critical patent/CN101021948A/en

Links

Abstract

This invention provides an automatic identifying device and a method for arthrosis in human body symmetrical moving images, in which, the device includes a high speed vidicon, a file converter, an image processor and a computer, in which, the high speed vidicon, the file converter and the image processor are connected with the computer, and the image processor includes orderly connected image segment module, an automatic identification module and an automatic tracking module, the file converter is connected with the image segment module, the converter and the automatic tracking module are connected with the storage of the computer separately, which can realize automatic identification of arthrosis point of human bodies.

Description

人体对称性运动图像中关节的自动识别装置及方法 Automatic recognition apparatus and method for a moving image symmetry of the joint body

技术领域 FIELD

本发明涉及人体运动图像处理技术,具体是指人体对称性运动图像中关节的自动识别装置及方法。 The present invention relates to moving image processing technology body, specifically refers to automatic recognition of human joints moving image symmetry apparatus and method.

背景技术 Background technique

分析人体运动首先要要实现人体运动数据的采集,人体运动数据主要包括了运动学量和动力学量,直接获得动力学量的方法主要是测力台,间接动力学量是在运动学量的基础上结合人体的惯性参数计算得到。 First of all to analyze human motion capture human motion data to realize the human body motion data includes the amount of kinematics and dynamics of volume, direct access to the amount of kinetic method is to force platform, the amount of indirect dynamics in the amount of kinematics human binding parameters calculated on the basis of inertia. 运动学量通过运动捕捉获得,人体运动捕捉可以分为电磁式运动捕捉、机电式运动捕捉和光学运动捕捉,光学运动捕捉系统以其快速、安全、准确、方便等优点,被广泛应用。 Kinematics amount obtained by motion capture, motion capture human motion capture can be divided into electromagnetic, electromechanical and optical motion capture motion capture, optical motion capture system with its fast, safe, accurate, convenient, etc., are widely used.

光学运动捕获是较为精确的一种捕捉方式。 The optical motion capture is a more accurate capture mode. 常用的光学运动捕捉方法是在人体关节点上的标记点是一种涂有特殊反光材料的球状体,利用摄像机从不同角度拍摄,然后利用软件分析图像上标记点的图像坐标,利用计算机视觉原理进行三维重建,得出标记点的运动数据。 Commonly used methods are optical motion capture markers in the human body joints are coated with a reflective material special spherical body, taken from different angles by the camera, and the image coordinates using the image analysis software on the marked points, computer vision principle three-dimensional reconstruction, the motion data derived marker points. 光学运动捕获主要指高速摄影图像的处理与分析。 The optical motion capture mainly refers to high-speed processing and analysis of the captured image.

对于人体运动高速摄像机图像的处理与分析主要包括图像分割、标记点识别和标记点跟踪、标记点运动学动力学计算和运动再现等。 For body-movement high-speed camera image processing including image segmentation and analysis, marker identification and tracking markers, markers kinematics and dynamics calculations motion reproduction or the like.

图像的边缘是图像分割依赖的重要特征,图像边缘处理主要有微分法(如LOG算子和DOG算子方法)、拟合法(Prewitt拟合法、Haralick拟合法和Huechel拟合法)、神经网络分析方法(如Poggio法和Hopfield法)和变尺度人体边缘检测(人工设置尺度法、基于知识的计算尺度法和自动确定滤波尺度法)。 Edge image is an important feature of image segmentation dependent image edge processing mainly differentiation (e.g. LOG operator and the DOG operator method), a fitting (the Prewitt fitting, Haralick fitting and Huechel fitting), Neural Network Analysis (e.g., Poggio method and Hopfield law) and human variable scale edge detection (manual setting scales, based on knowledge of the calculated scales and scales to automatically determine the filter). 对于对称性人体运动中关节的图像边缘检测,现有的方法存在效率较低和实时性较差,难以实现对称性人体运动中人体关节自动测试的要求。 For symmetry in human motion detection image edge joint, there is a conventional method is less efficient and less real-time, it is difficult to achieve the required movement of the human body joints symmetry automated testing.

人体运动的识别即标记点的识别,识别算法有基于统计的方法(HMM模型法、动态贝叶斯法)、模板匹配方法(DTW法)和基于语法的方法(FSM法、STS法)。 Recognition of human motion that is the point of identification marks, recognition algorithms statistical methods (HMM models, dynamic Bayesian method) based on template matching method (DTW method) and method syntax (FSM law, STS method) based. 粘贴在关节处的标记点对于对称的人体平面运动,有其相邻关节点之间的距离不变这一特征,现有的方法没有涉及人体对称平面运动这一特征,这给识别效果带来了不利的影响。 Attached joints marked point symmetry plane of the body motion, there is a distance between the adjacent articulation point of the invariant features, the conventional method does not relate to the characteristics of human movement plane of symmetry, which brings effect to the identification had a negative impact.

现有的人体关节标记点跟踪方法有基于运动场估计的方法、矩形区域预测跟踪方法、人体关节运动自由度限制跟踪方法和一阶预测局部搜索方法。 Conventional human joint marker tracking method is a method based on the estimated motion field, tracking rectangular region prediction method, the articulation body to limit the degree of freedom tracking method and a local search order prediction method. 在无约束或保守力作用下,人体关节只能做绕近端环节转动,在根据区域上应该是与人体关节轴线对称的区间,现有的跟踪方法没有完全遵循人体关节的转动特点和人体结构限制下解剖学特点。 Unconstrained or conservative forces, the joint body can be rotated about the proximal end of the link, in accordance with rotation of the characteristics and structure of the human body with the joint axis should be symmetrical section, a conventional tracking method is not fully comply with the human joint area anatomical features under restrictions.

发明内容 SUMMARY

本发明的目的在于克服上述现有技术的缺点和不足,提供一种结合人体结构特点和运动规律,高效率的人体对称性运动图像中关节的自动识别装置。 Object of the present invention is to overcome the above disadvantages and drawbacks of the prior art, to provide human motion symmetry efficient automatic image recognition apparatus of one binding joints and structural characteristics of human motion law.

本发明的目的还在于提供采用上述人体对称性运动图像中关节的自动识别装置的自动识别方法。 Object of the present invention is to provide a method for automatic recognition of the human bodies moving image symmetry automatic recognition apparatus joints.

本发明的目的通过下述技术方案实现:本人体对称性运动图像中关节的自动识别装置,包括高速摄像机、文件转换器、图像处理器、计算机,所述高速摄像机、文件转换器、图像处理器分别与计算机连接,所述图像处理器包括依次连接的图像分割模块、自动识别模块和自动跟踪模块,所述文件转换器与图像分割模块连接,且文件转换器、自动跟踪模块分别与计算机的存储器连接。 Object of the present invention is achieved by the following technical scheme: the present body motion image symmetry joint automatic recognition apparatus, comprising a high-speed camera, the file conversion, an image processor, a computer, a high-speed camera, file converter, an image processor are respectively connected to the computer, the image processor includes an image segmentation module sequentially connected, automatic recognition module, and automatic tracking module, a document converter is connected with the image segmentation module, and a file converter to automatically track a computer memory module, respectively connection.

所述图像分割模块包括依次连接的归一化图像处理模块、分层处理模块、阈值化处理模块、边缘拟合处理模块,归一化图像处理模块与所述文件转换器连接;所述自动识别模块包括依次连接的第一帧人工设置处理模块、标志点边缘坐标处理模块、标志点位置计算模块,第一帧人工设置处理模块与所述边缘拟合处理模块连接,标志点边缘坐标处理模块、标志点位置计算模块分别与所述自动跟踪模块连接;所述自动跟踪模块包括依次连接的关节标志点搜索模块、关节位置及位移计算模块,所述关节位置及位移计算模块与计算机的存储器连接。 The image segmentation module comprises normalizing the image processing module which are sequentially connected, hierarchical processing module, the processing module thresholding, edge fitting processing module, normalizing the image processing module and the file converter; said automatic recognition a first frame module comprises a manual setting processing module connected successively, the coordinates of the edge processing module mark point, mark position calculation module, the first manual setting frame processing module processing module is connected to the edge of the fitting, the edge mark point coordinate processing module, mark position calculation module is connected to the automatic tracking module; said automatic tracking flag articulation point module comprises a search module connected successively, the joint position and displacement calculating module, a memory of the joint position, and movement calculating module connected to the computer.

采用上述人体对称性运动图像中关节的自动识别装置的自动识别方法,其步骤包括:(1)建立包括高速摄影机标定和测量标志标定的人体运动测试场,并保证人体运动在所述人体运动测试场中完成,人体关节位置贴标志点,其具体要求按照人体测量中的约定进行;在所述人体运动测试场中,人体运动通过高速摄影机记录,记录的模拟图像信号通过高速摄影机自带的A/D器转换为数字图像信号传送给计算机; Automatic recognition method for automatically recognizing apparatus of the above-described moving image symmetry in human joints, comprising the steps of: (1) establishment of high-speed video camera comprising a calibration and calibration measurement flag field of human movement, and to ensure that the body motion in human movement field is completed, the position of the joint body affixed mark point, according to the specific requirements of the human measurement convention; human movement in the field, by a high speed camera body motion recording, recording by the analog image signal carrying the high-speed camera a / D converted into a digital image signal transferred to the computer;

(2)所述数字图像信号由文件转换器先转换为视频格式文件,再转换为静态帧图像,视频格式文件与静态帧图像文件通过计算机的存储器存储后,计算机把静态帧图像文件以帧的形式传送给图像处理器;(3)所述静态帧图像文件经过图像处理器的图像分割模块、自动识别模块、自动跟踪模块作相应处理后,所得的数据通过自动跟踪模块输送到计算机的存储器存储为流式文件,从而实现人体对称性运动图像中关节的自动识别。 (2) the digital image file by a signal converter to convert video format, and then converted to a static frame image, and video format still-frame image file via the memory storing computer files, the computer files the still frame image of a frame transferred to the image processor form; (3) the static frame image file after the image processor image segmentation module, automatic recognition module, for automatically tracking module corresponding processing, resulting in data supplied to the memory storing the computer by automatically tracking module a streaming file, the moving image in order to achieve symmetry of the body automatically recognize the joint.

所述图像分割模块的处理,其步骤包括:(1)通过归一化图像处理模块对所述静态帧图像文件进行归一化处理,归一化处理首先根据设定摄影测量标志标定来确定“运动”的不变性,也就高速摄影机坐标的固定性,同时为了加快图像的处理速度,归一化图像处理模块对每帧图像的几何尺寸进行统一;(2)通过分层处理模块对上述归一化处理后的静态帧图像进行分层处理,得到图像层R层、G层、B层,各图像层上,灰度图像的色点只有彩色图像的1/65535的信息量;(3)为了保证标志点的准确分割,通过阈值化处理模块分别对图像层R层、G层、B层进行阈值化处理,阈值化处理模块在系统初始化值的基础上,根据图像的总色值对控制参数进行一次性调整,这个过程只做一次,原因是我们假设拍摄现成的环境是相对恒定的(一般要分析的动作时间在1秒钟之内);(4)为了 Processing the image segmentation module, comprising the steps of: (a) normalizing the image processing module of the static frame image files normalized, the normalization processing is first determined according to the calibration measurement flag is set by the photographic " movement "invariance, also high-speed camera coordinate fixability, and in order to speed up processing of the image, the normalized image processing module for each image geometry unified; (2) by layering the above-described normalization processing module a static frame image of the slicing process, the image layer to obtain the R layer, G layer, B layer, on each image layer, the color point gray image only color image information of 1/65535; (3) to ensure accurate segmentation of landmarks, the image layer is R layer, G layer, B layer thresholding respectively thresholding module, the threshold value processing module on the basis of the initialization value of the system, based on the total color values ​​of the image control one-time parameter adjustment, this process only once, because we assume that the environment is ready captured relatively constant (typically operating time to be analyzed within 1 second); (4) to 确保标志点的自动识别与自动跟踪,上述阈值化处理模块分割后的图像经极值判断后,由边缘拟合处理模块进行一次边缘拟合处理,得到静态标志点图像。 Landmark ensure automatic tracking and automatic identification, the image thresholding module after the extreme value is determined by dividing, by the edge of a fitting edge processing module fitting process, to obtain still image mark point.

所述自动识别模块的处理,其步骤包括:(1)获取所述图像分割处理模块处理后的静态标志点图像,通过所述第一帧人工设置处理模块对首帧图像进行人工设置,所述人工设置是指给出标志点的形心名称与位置,用直线确定两标志点之间的环节,同时给首帧图像标上序号为1;(2)根据首帧图像标志点的位置,以及根据所述自动跟踪模块的控制参数得到的其余序号的帧图像标志点的位置,由标志点边缘坐标处理模块对标志点的边缘进行轮廓处理;(3)所述标志点位置计算模块根据所述标志点边缘连续轮廓点的平面坐标,结合合力矩原理获取标志点的位置与在同一环节上的标志点距离,把包括图像帧序号、标志点名称、标志点的位置和同一环节上的标志点距离的数据传送给所述自动跟踪模块的关节位置及位移计算模块。 The automatic recognition processing module, comprising the steps of: (1) obtain a static image of the mark point image segmentation processing module, manually set by the first frame image the first frame processing module manually set, the refers to manually set the position of the centroid given names of landmarks to determine the link between the two marking points with a straight line, while a first frame image of a superscript number; (2) the position of the first frame image of the landmark, and the position of the frame image of the remaining number of marker points of the control parameter automatic tracking module obtained by the processing contour edge point coordinates of the edge processing module mark point flag; (3) the marker according to the position calculation module plane coordinate marker point edge successive contour points, binding moment principle engagement acquired mark point position marking points on the same link distance, the flag on the position and the same part of an image including the frame number, landmark name, landmark points transmitting data from and to the joint position displacement calculation module automatically tracking module.

所述自动跟踪模块的处理,其步骤包括:(1)通过所述关节标志点搜索模块接受所述自动识别模块处理得到的数据,根据第n帧标志点的位置,以及第n-1帧的位置,预测第n+1帧的图像中标志点的位置,结合远端环节绕近端环节转动的原理,确定第n+1帧中标志点的搜索区域,把标志点和搜索区域的预测值作为控制参数传送到自动识别模块;(2)关节位置及位移计算模块根据上述控制参数,结合人体惯性参数,并以髋关节为坐标原点,根据人体环节长度不变的特性,获得关节位置和角度;关节位置及位移计算模块根据线速度、线加速度、角速度、角加速的定义,对所述关节位置和角度以及由所述标志点位置计算模块传送的包括图像帧序号、标志点名称、标志点的位置和同一环节上的标志点距离的数据进行处理,得到相应的关节参数,所有关节参数输送到所述 The automatic tracking processing module, comprising the steps of: (1) the automatic identification module accepts data obtained by processing the articulation point mark search module, according to the position of the n-th frame mark point, and the n-1 frame, position, the position of the image points mark the n + 1 frame prediction, the proximal end of distal link part rotatable about the principles, the search area is determined the n + 1 frame mark point, the predicted value of the flag point and the search area transmitted as a control parameter to automatically identify the module; (2) the joint position, and movement according to the control parameter calculation module, combined with human inertial parameters, and as the coordinate origin in the hip joint, according to the characteristics of the body part of the same length, the position and angle of the joint is obtained ; joint position and linear velocity displacement calculating module, linear acceleration, angular velocity, angular acceleration definition, position and angle of the joint and the marker point position calculated by the module includes an image transfer frame number, landmark name, landmark the data point location and distance markers on the same link, to give the corresponding joint parameters, all the parameters supplied to the joint 存储器以流式文件存储。 The memory stored streaming file.

所述关节参数的数据结构包括:帧图像序号、标志点名称、每个标志点的位置、每个标志点的线速度、每个标志点的线加速度、环节的角度(包括髋角度、膝角度、踝角度)、环节的角速度、环节的角加速度。 The joint parameter data structure comprising: a position of the frame image number, landmark name, each marker point, the linear velocity of each point mark, each mark point linear acceleration, angular part (including the angle of the hip, knee angle , ankle angle), part of the angular velocity, angular acceleration links.

本发明与现有技术相比,具有如下优点和有益效果:本发明根据人体对称性运动的特点,采用了基于层的图像分割方法,使标志点能够完全从运动背景和人体中提取出来,标志点自动识别采用连续边缘轮廓坐标的位置来计算,在确保准确性的同时加快了识别速度,自动跟踪方法在传统运动估计的基础上结合了人体的解剖学特点和人体环节运动形式的特点,预测帧中标志点的运动趋势,保证了自动识别的高效与准确,根据人体运动分析的需要计算了人体对称性运动中所有的运动学量。 Compared with the prior art the present invention has the following advantages and beneficial effects: the present invention according to the characteristics of symmetry of the body motion, using image segmentation method based layer, so that the marker point can be completely extracted from the body and the moving background, logo automatic identification point using continuous edge outline coordinate position is calculated, while ensuring the accuracy of accelerating the speed of recognition, automatic tracking method combines the anatomy and motion of the body forms part of a human body on the basis of characteristics of the conventional motion estimation, prediction movement trend frame mark point, to ensure the efficient and accurate automatic recognition, according to the needs of the body motion analysis calculated amount of all the kinematic motion of the symmetry of the human body. 本发明可实现人体关节点的自动识别,对伤病研究和康复、分析诊断运动技术和人体工效学等有着重要的意义。 The invention can automatically identify the key points of the body, research and rehabilitation for injuries, diagnostic analysis is significant movement of technology and ergonomics and so on.

附图说明 BRIEF DESCRIPTION

图1是本发明人体对称性运动图像中关节的自动识别装置的工作原理图;图2是图1所示图像分割模块的工作原理图;图3是图1所示自动识别模块的工作原理图;图4是图1所示自动跟踪模块的工作原理图。 FIG 1 is a principle diagram of a human symmetry moving image recognition apparatus of the present invention automatically joints; FIG. 2 is an image segmentation module works shown in FIG. 1; FIG. 3 is a diagram illustrating the working principle of automatic identification module shown in FIG. 1 ; FIG. 4 is a principle view of an automatic tracking module shown in Fig.

具体实施方式 Detailed ways

下面结合实施例及附图,对本发明作进一步地详细说明,但本发明的实施方式不限于此。 The following Examples and the accompanying drawings in conjunction with embodiments of the present invention will be further described in detail, but the embodiment of the present invention is not limited thereto.

实施例一如图1所示,本人体对称性运动图像中关节的自动识别装置,包括高速摄像机、文件转换器、图像处理器、计算机,所述高速摄像机、文件转换器、图像处理器分别与计算机连接,所述图像处理器包括依次连接的图像分割模块、自动识别模块和自动跟踪模块,所述文件转换器与图像分割模块连接,且文件转换器、自动跟踪模块分别与计算机的存储器连接。 , The present body motion image symmetry joint automatic recognition apparatus, comprising a high-speed camera, the file conversion, an image processor, a computer, a high-speed camera, file converter, an image processor as shown in Example 1, respectively, and the computer is connected to the image processor includes an image segmentation module sequentially connected, automatic recognition module, and automatic tracking module, a document converter is connected with the image segmentation module, and a file converter, automatic tracking module is connected to the memory of the computer.

如图2所示,所述图像分割模块包括依次连接的归一化图像处理模块、分层处理模块、阈值化处理模块、边缘拟合处理模块,归一化图像处理模块与所述文件转换器连接;如图3所示,所述自动识别模块包括依次连接的第一帧人工设置处理模块、标志点边缘坐标处理模块、标志点位置计算模块,第一帧人工设置处理模块与所述边缘拟合处理模块连接,标志点边缘坐标处理模块、标志点位置计算模块分别与所述自动跟踪模块连接;如图4所示,所述自动跟踪模块包括依次连接的关节标志点搜索模块、关节位置及位移计算模块,所述关节位置及位移计算模块与计算机的存储器连接。 As shown, the image segmentation module 2 comprises normalizing the image processing module which are sequentially connected, hierarchical processing module, the processing module thresholding, edge fitting processing module, normalizing the image processing module and the file converter connection; As shown, a first frame 3 of the automatic identification module processing module comprises a manual setting are sequentially connected, mark points edge coordinates processing module mark position calculation module, a first frame processing module provided with the artificial edge Quasi bonding processing module is connected, the processing module mark point edge coordinates, marker point position calculating module is connected to the automatic tracking module; 4, said automatic tracking flag articulation point module comprises a search module connected successively, the joint position and displacement calculating module, and a displacement calculating the joint position memory module from the computer.

如图1~4所示,本人体对称性运动图像中关节的自动识别装置是这样进行人体对称性运动图像中关节的自动识别的:(1)建立包括高速摄影机标定和测量标志标定的人体运动测试场,并保证人体运动在所述人体运动测试场中完成,人体关节位置贴标志点,其具体要求按照人体测量中的约定进行;在所述人体运动测试场中,人体运动通过高速摄影机记录,记录的模拟图像信号通过高速摄影机自带的A/D器转换为数字图像信号传送给计算机;(2)所述数字图像信号由文件转换器先转换为视频格式文件,再转换为静态帧图像,视频格式文件与静态帧图像文件通过计算机的存储器存储后,计算机把静态帧图像文件以帧的形式传送给图像处理器;(3)所述静态帧图像文件经过图像处理器的图像分割模块、自动识别模块、自动跟踪模块作相应处理后,所得的数据通过自 As shown in FIG. 1 to 4, the body moving image symmetry automatic recognition apparatus joints is carried out moving image symmetry body joint automatic recognition: (1) establishment of high-speed video camera comprising a body movement measurement flag calibration and calibration the test field, and to ensure completion of said body movement human movement field, the position of the joint body affixed mark point, according to the specific requirements of the human measurement convention; human movement in the field, by a high speed video camera recording moving body , the analog image signal recorded by the camera built-in high-speed a / D converted into a digital image signal transferred to the computer; (2) the digital image file by a signal converter to convert video format, and then converted to a static frame image , video format and the still frame image files stored in the memory of a computer, the computer still frame image of a frame in the form of file transfer to the image processor; (3) the static frame image file after the image processor image segmentation module, after the automatic recognition module, for automatically tracking module corresponding processing, data resulting from the 跟踪模块输送到计算机的存储器存储为流式文件,从而实现人体对称性运动图像中关节的自动识别。 Tracking module delivered to the memory of a computer stores a streaming file, the moving image in order to achieve symmetry of the body automatically recognize the joint.

如图2所示,步骤(3)所述图像分割模块的处理过程是:1、通过归一化图像处理模块对所述静态帧图像文件进行归一化处理,归一化处理首先根据设定摄影测量标志标定来确定“运动”的不变性,也就高速摄影机坐标的固定性,同时为了加快图像的处理速度,归一化图像处理模块对每帧图像的几何尺寸进行统一;2、通过分层处理模块对上述归一化处理后的静态帧图像进行分层处理,得到图像层R层、G层、B层,各图像层上,灰度图像的色点只有彩色图像的1/65535的信息量;3、为了保证标志点的准确分割,通过阈值化处理模块分别对图像层R层、G层、B层进行阈值化处理,阈值化处理模块在系统初始化值的基础上,根据图像的总色值对控制参数进行一次性调整,这个过程只做一次,原因是我们假设拍摄现成的环境是相对恒定的(一般要分析的动作时间在1秒钟 As shown in FIG 2 in step (3) processing of the image segmentation module are: 1, for the normalization of the static frame image file by normalizing the image processing module, according to the normalization processing is first set photogrammetric calibration flag to determine invariant "movement", it fixability speed camera coordinates, and in order to speed up the processing speed of the image, the normalized image processing module for each image geometry unified; 2, by fractional layer processing module static frame image after the normalizing process slicing, image layer to obtain the R layer, G layer, B layer, on each image layer, the color point gray image only the color image 1/65535 information; 3, in order to ensure accurate division mark point, respectively, the image layer R layer, G layer, B layer thresholding by thresholding module, the threshold value processing module on the basis of the initialization value of the system, according to the image the total color value of a one-time adjustment of control parameters, this process only once, because we assume that the environment is ready captured relatively constant operating time (typically 1 second to be analyzed 之内);4、为了确保标志点的自动识别与自动跟踪,上述阈值化处理模块分割后的图像经极值判断后,由边缘拟合处理模块进行一次边缘拟合处理,得到静态标志点图像。 Within); 4, in order to ensure the automatic recognition and landmark tracking automatically, after dividing the image thresholding module after the extreme value is determined by fitting the edge of the edge processing module a fitting process, to obtain still image mark point .

如图3所示,步骤(3)所述自动识别模块的处理过程是:1、获取所述图像分割处理模块处理后的静态标志点图像,通过所述第一帧人工设置处理模块对首帧图像进行人工设置,所述人工设置是指给出标志点的形心名称与位置,用直线确定两标志点之间的环节,同时给首帧图像标上序号为1;2、根据首帧图像标志点的位置,以及根据所述自动跟踪模块的控制参数得到的其余序号的帧图像标志点的位置,由标志点边缘坐标处理模块对标志点的边缘进行轮廓处理;3、所述标志点位置计算模块根据所述标志点边缘连续轮廓点的平面坐标,结合合力矩原理获取标志点的位置与在同一环节上的标志点距离,把包括图像帧序号、标志点名称、标志点的位置和同一环节上的标志点距离的数据传送给所述自动跟踪模块的关节位置及位移计算模块。 As shown, in step (3) 3 The process is automatic identification module: 1, still obtaining the mark point image processing module processes image segmentation, manually set by the first frame of the first frame processing module image manual setting, the manual setting means gives the name and location of the centroid of landmarks to determine the link between the two marking points with a straight line, while a first frame image superscript number is 1; 2, the first frame image in accordance with positions of landmarks, and the frame image according to the remaining number of marker points of the control parameter automatic tracking module obtained position, the edge module marker point mark point coordinates edge contour processing treatment; 3, the marker position the module calculates the coordinates of said mark point plane continuous edge contour points, in conjunction with the torque engagement acquisition flag principle point position of the landmarks on the same link distance, an image including the frame number, landmark name, the landmark position and the same flag data point distances to the transmitting part of the joint position and displacement calculating module automatically tracking module.

如图4所示,步骤(3)所述自动跟踪模块的处理过程是:1、通过所述关节标志点搜索模块接受所述自动识别模块处理得到的数据,根据第n帧标志点的位置,以及第n-1帧的位置,预测第n+1帧的图像中标志点的位置,结合远端环节绕近端环节转动的原理,确定第n+1帧中标志点的搜索区域,把标志点和搜索区域的预测值作为控制参数传送到自动识别模块;2、关节位置及位移计算模块根据上述控制参数,结合人体惯性参数,并以髋关节为坐标原点,根据人体环节长度不变的特性,获得关节位置和角度;关节位置及位移计算模块根据线速度、线加速度、角速度、角加速的定义,对所述关节位置和角度以及由所述标志点位置计算模块传送的包括图像帧序号、标志点名称、标志点的位置和同一环节上的标志点距离的数据进行处理,得到相应的关节参数,所有关节参 As shown, in step 4 (3) during the automatic tracking processing module are: 1, the automatic recognition module to accept data obtained by processing the articulation point mark search module, according to the position of the n-th frame mark point, and the position of the n-1 frame, the position of the image points mark the n + 1 frame prediction, the proximal end of distal link part rotatable about the principles, the search area is determined the n + 1 frame mark point, the flag point search area and the predicted value is transmitted as a control parameter to automatically identify the module; 2, joint position, and movement according to the control parameter calculation module, combined with human inertial parameters, and as the coordinate origin in the hip joint, according to the characteristics of the body part of constant length obtained joint position and angle; and a joint position displacement calculation module linear velocity, linear acceleration, angular velocity, angular acceleration definition, position and angle of the joint and the marker point position calculated by the transmitting module comprises a number of image frames, landmark name, the data point from the position of the mark and the same part of landmarks, to give the corresponding joint parameters, all parameters joints 数输送到所述存储器以流式文件存储。 The number of delivered to the streaming file is stored in the memory.

所述关节参数的数据结构包括:帧图像序号、标志点名称、每个标志点的位置、每个标志点的线速度、每个标志点的线加速度、环节的角度(包括髋角度、膝角度、踝角度)、环节的角速度、环节的角加速度。 The joint parameter data structure comprising: a position of the frame image number, landmark name, each marker point, the linear velocity of each point mark, each mark point linear acceleration, angular part (including the angle of the hip, knee angle , ankle angle), part of the angular velocity, angular acceleration links. 所述关节参数的数据结构包括:图像帧序号、标志点名称、每个标志点的位置、每个标志点的线速度、每个标志点的线加速度、环节的角度(包括髋角度、膝角度、踝角度)、环节的角速度、环节的角加速度。 The joint parameters in the data structure comprising: a frame number of an image, landmark name, the location of each marker point, the linear velocity of each point mark, each mark point linear acceleration, angular part (including the angle of the hip, knee angle , ankle angle), part of the angular velocity, angular acceleration links.

通过实验和理论分析,本发明根据人体在对称性动作中人体相邻关节在平面运动中的距离保持不变的特征和人体关节运动的规律(关节的解剖学活动范围和运动学特征),提出的通过对高速摄影机图像的人体关节(在关节处贴标志)的阈值化图像处理、关节点自动识别和关节点自动跟踪,从而实现人体对称运动图像中关节的自动识别的方法是完全可行的。 Experiments and theoretical analysis, according to the present invention, the distance joint plane motion law remains unchanged features and articulation of the body (anatomical range of motion and kinematics of the joint) in the human body adjacent to the symmetry of operation, it is proposed by thresholding image processing on human joint high-speed camera image (posted flag in the joints), the articulation point automatic identification and articulation point automatic tracking, thereby realizing the moving image automatic identification joint body symmetry is entirely feasible.

如上所述,便可较好地实现本发明。 As described above, the present invention can be better achieved.

Claims (5)

1.人体对称性运动图像中关节的自动识别装置,其特征在于:包括高速摄像机、文件转换器、图像处理器、计算机,所述高速摄像机、文件转换器、图像处理器分别与计算机连接,所述图像处理器包括依次连接的图像分割模块、自动识别模块和自动跟踪模块,所述文件转换器与图像分割模块连接,且文件转换器、自动跟踪模块分别与计算机的存储器连接。 1. Human symmetry of the joints in the automatic moving image recognition apparatus comprising: a high-speed camera, the file conversion, an image processor, a computer, a high-speed camera, file converter, an image processor connected to the computer, the sequentially connecting said image processor comprises an image segmentation module, an automatic recognition module, and automatic tracking module, a document converter is connected with the image segmentation module, and a file converter, automatic tracking module is connected to the memory of the computer.
2.根据权利要求1所述人体对称性运动图像中关节的自动识别装置,其特征在于:所述图像分割模块包括依次连接的归一化图像处理模块、分层处理模块、阈值化处理模块、边缘拟合处理模块,归一化图像处理模块与所述文件转换器连接;所述自动识别模块包括依次连接的第一帧人工设置处理模块、标志点边缘坐标处理模块、标志点位置计算模块,第一帧人工设置处理模块与所述边缘拟合处理模块连接,标志点边缘坐标处理模块、标志点位置计算模块分别与所述自动跟踪模块连接;所述自动跟踪模块包括依次连接的关节标志点搜索模块、关节位置及位移计算模块,所述关节位置及位移计算模块与计算机的存储器连接。 2. The symmetry of the body 1 automatically moving image recognition apparatus according to claim joint, wherein: the image segmentation module comprises normalizing the image processing module which are sequentially connected, hierarchical processing module, thresholding module, fitting the edge processing module, normalizing the image processing module and the file converter; said automatic recognition module includes a first frame processing module manually set connected successively, the coordinates of the edge processing module mark point, mark position calculation module, artificial setting processing module and the edge of the first frame processing module connected to the fitting, the edge mark point coordinate processing module, mark position calculation module is connected to the automatic tracking module; said automatic tracking module comprises successively connecting joint landmarks search module, the joint position and displacement calculating module, and a joint position displacement calculation module connected to the computer memory.
3.采用权利要求1所述人体对称性运动图像中关节的自动识别装置的人体对称性运动图像中关节的自动识别方法,其特征在于包括以下步骤:(1)建立包括高速摄影机标定和测量标志标定的人体运动测试场,并保证人体运动在所述人体运动测试场中完成,人体关节位置贴标志点,其具体要求按照人体测量中的约定进行;在所述人体运动测试场中,人体运动通过高速摄影机记录,记录的模拟图像信号通过高速摄影机自带的A/D器转换为数字图像信号传送给计算机;(2)所述数字图像信号由文件转换器先转换为视频格式文件,再转换为静态帧图像,视频格式文件与静态帧图像文件通过计算机的存储器存储后,计算机把静态帧图像文件以帧的形式传送给图像处理器;(3)所述静态帧图像文件经过图像处理器的图像分割模块、自动识别模块、自动跟踪模块作相应处理 Human moving image symmetry automatic recognition of human joints symmetry of the moving image of the automatic recognition apparatus using a joint according to claim 3, characterized by comprising the steps of: (1) establishment of high-speed video camera comprising a calibration and measurement flag calibration field human movement, and to ensure completion of said body movement human movement field, the position of the joint body affixed mark point, according to the specific requirements of the human measurement convention; human movement in the field, human motion by recording a high-speed camera, an analog image signal recorded by the camera built-in high-speed a / D converted into a digital image signal transferred to the computer; (2) the digital image signal by the converter to convert the file into a video format, and then converted after a static frame image, video format and files by still picture frame stored in a computer memory, the computer still frame image of a frame in the form of file transfer to the image processor; (3) the static frame of the image file after the image processor The image segmentation module, an automatic recognition module, for automatically tracking module corresponding processing ,所得的数据通过自动跟踪模块输送到计算机的存储器存储为流式文件,从而实现人体对称性运动图像中关节的自动识别。 The resulting data is conveyed through the module to automatically track a computer memory storage for streaming file, the moving image in order to achieve symmetry of the body automatically recognize the joint.
4.根据权利要求3所述人体对称性运动图像中关节的自动识别方法,其特征在于:所述图像分割模块的处理包括以下步骤:(1)通过归一化图像处理模块对所述静态帧图像文件进行归一化处理,归一化处理首先根据设定摄影测量标志标定来确定“运动”的不变性,也就高速摄影机坐标的固定性,同时,归一化图像处理模块对每帧图像的几何尺寸进行统一;(2)通过分层处理模块对上述归一化处理后的静态帧图像进行分层处理,得到图像层R层、G层、B层,各图像层上,灰度图像的色点只有彩色图像的1/65535的信息量;(3)通过阈值化处理模块分别对图像层R层、G层、B层进行阈值化处理,阈值化处理模块在系统初始化值的基础上,根据图像的总色值对控制参数进行一次性调整;(4)上述阈值化处理模块分割后的图像经极值判断后,由边缘拟合处理模块进行一次边缘 The symmetry of the body 3 automatically moving image recognition method as claimed in claim joints, wherein: the image segmentation processing module comprising the steps of: (1) by normalizing the image processing module of the static frame image files normalized, the normalization processing to determine the first calibration invariance "movement" based on the setting flag photogrammetry, it fixability speed camera coordinates, while normalizing the image processing module for each image unified geometry; (2) slicing static frame image after the normalization processing by the hierarchical processing module, an image layer to obtain the R layer, G layer, B layer, on each of the image layers, the image gradation color point only 1/65535 of the amount of information of the color image; (3) the image layer R layer, G layer, B layer thresholding by thresholding module, respectively, based initialization value system thresholding processing module the total value of the color image is one-time adjustment of control parameters; (4) after the image is determined by the extreme values ​​of thresholds dividing processing module, the processing module by the edges of a fitting edge 拟合处理,得到静态标志点图像;所述自动识别模块的处理包括以下步骤:(1)获取所述图像分割处理模块处理后的静态标志点图像,通过所述第一帧人工设置处理模块对首帧图像进行人工设置,所述人工设置是指给出标志点的形心名称与位置,用直线确定两标志点之间的环节,同时给首帧图像标上序号为1;(2)根据首帧图像标志点的位置,以及根据所述自动跟踪模块的控制参数得到的其余序号的帧图像标志点的位置,由标志点边缘坐标处理模块对标志点的边缘进行轮廓处理;(3)所述标志点位置计算模块根据所述标志点边缘连续轮廓点的平面坐标,获取标志点的位置与在同一环节上的标志点距离,把包括图像帧序号、标志点名称、标志点的位置和同一环节上的标志点距离的数据传送给所述自动跟踪模块的关节位置及位移计算模块;所述自动跟踪模块的 Fitting process, to obtain still image mark point; processing the automatic recognition module comprises the steps of: (1) obtain a static image of the mark point image segmentation processing module, manually set by the first frame processing module the first frame image manual setting, the manual setting means gives the name and location of the centroid of landmarks to determine the link between the two marking points with a straight line, while a first frame image of a superscript number; (2) the first position of the frame image landmark and landmark image according to a frame number of the remaining control parameters obtained by automatic tracking of the position of the module, the processing by the edge contour edge mark point coordinate processing module mark point; (3) said mark position calculation module based on the marker coordinate point plane continuous edge contour points, acquiring a position marker point marking points on the same link distance, an image including the frame number, landmark name, the landmark position and the same flag data point distances to the transmitting part of the joint position, and movement tracking module automatically calculating module; said automatic tracking module 处理包括以下步骤:(1)通过所述关节标志点搜索模块接受所述自动识别模块处理得到的数据,根据第n帧标志点的位置,以及第n-1帧的位置,预测第n+1帧的图像中标志点的位置,确定第n+1帧中标志点的搜索区域,把标志点和搜索区域的预测值作为控制参数传送到自动识别模块;(2)关节位置及位移计算模块根据上述控制参数,结合人体惯性参数,并以髋关节为坐标原点,根据人体环节长度不变的特性,获得关节位置和角度;关节位置及位移计算模块对所述关节位置和角度以及由所述标志点位置计算模块传送的包括图像帧序号、标志点名称、标志点的位置和同一环节上的标志点距离的数据进行处理,得到相应的关节参数,所有关节参数输送到所述存储器以流式文件存储。 Process comprising the steps of: (1) the automatic identification module accepts data obtained by processing the articulation point mark search module, according to the position of the n-th frame mark point, and the position of the frame n-1, n + 1 Prediction position in the image marker point in the frame, determining a search area of ​​the n + 1 frame landmarks, the landmark and the predicted value of the search area is transmitted as a control parameter to automatically identify the module; (2) the joint position, and movement calculation module according to said control parameter, in conjunction with the body inertial parameters, and as the coordinate origin in the hip joint, according to the characteristics of the body part of the same length, the position and angle of the joint is obtained; and a joint position calculating module displacement position and angle of the joint and on the mark position includes an image frame number, landmark name, landmark point position calculation module and the data transfer distance marker points on the same link, to give the corresponding joint parameters, all the parameters supplied to the joint in a streaming file memory storage.
5.根据权利要求4所述人体对称性运动图像中关节的自动识别方法,其特征在于所述关节参数的数据结构包括:帧图像序号、标志点名称、每个标志点的位置、每个标志点的线速度、每个标志点的线加速度、环节的角度、环节的角速度、环节的角加速度;所述环节的角度包括髋角度、膝角度、踝角度。 The symmetry of the body 4 automatically moving image recognition method as claimed in claim joint, wherein said joint parameter data structure comprising: a position of the frame image number, marker point name, point of each marker, each marker the linear velocity of the point, the linear acceleration of each marker point, part of the angle, the angular velocity of the link, link angular acceleration; and the hip link angle comprises an angle, the angle of the knee, ankle angle.
CN 200710027147 2007-03-14 2007-03-14 Automatic identifying device and method for joint in human body symmetric motion image CN101021948A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200710027147 CN101021948A (en) 2007-03-14 2007-03-14 Automatic identifying device and method for joint in human body symmetric motion image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200710027147 CN101021948A (en) 2007-03-14 2007-03-14 Automatic identifying device and method for joint in human body symmetric motion image

Publications (1)

Publication Number Publication Date
CN101021948A true CN101021948A (en) 2007-08-22

Family

ID=38709704

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200710027147 CN101021948A (en) 2007-03-14 2007-03-14 Automatic identifying device and method for joint in human body symmetric motion image

Country Status (1)

Country Link
CN (1) CN101021948A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102668572A (en) * 2009-12-22 2012-09-12 汤姆逊许可证公司 Method and apparatus for optimal motion reproduction in stereoscopic digital cinema
CN103533242A (en) * 2013-10-15 2014-01-22 中国科学院深圳先进技术研究院 Method and system for extracting and tracking cursor point in out-of-focus video
CN104887238A (en) * 2015-06-10 2015-09-09 上海大学 Hand rehabilitation training evaluation system and method based on motion capture
CN104898837A (en) * 2015-05-22 2015-09-09 燕山大学 Portable hand virtual rehabilitation experiment box based on gesture interaction and method
CN105832343A (en) * 2016-05-22 2016-08-10 上海大学 Multi-dimensional vision hand function rehabilitation quantitative evaluation system and evaluation method

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102668572A (en) * 2009-12-22 2012-09-12 汤姆逊许可证公司 Method and apparatus for optimal motion reproduction in stereoscopic digital cinema
CN102668572B (en) * 2009-12-22 2015-04-29 汤姆逊许可证公司 Method and apparatus for optimal motion reproduction in stereoscopic digital cinema
US9030525B2 (en) 2009-12-22 2015-05-12 Thomson Licensing Method and apparatus for optimal motion reproduction in stereoscopic digital cinema
CN103533242A (en) * 2013-10-15 2014-01-22 中国科学院深圳先进技术研究院 Method and system for extracting and tracking cursor point in out-of-focus video
CN103533242B (en) * 2013-10-15 2016-08-10 中国科学院深圳先进技术研究院 The method and system with tracking cursor point are extracted in video out of focus
CN104898837A (en) * 2015-05-22 2015-09-09 燕山大学 Portable hand virtual rehabilitation experiment box based on gesture interaction and method
CN104887238A (en) * 2015-06-10 2015-09-09 上海大学 Hand rehabilitation training evaluation system and method based on motion capture
CN105832343A (en) * 2016-05-22 2016-08-10 上海大学 Multi-dimensional vision hand function rehabilitation quantitative evaluation system and evaluation method

Similar Documents

Publication Publication Date Title
Krotkov Active computer vision by cooperative focus and stereo
US8565479B2 (en) Extraction of skeletons from 3D maps
US7889905B2 (en) Fast 3D-2D image registration method with application to continuously guided endoscopy
US8989518B2 (en) Information processing method and apparatus for calculating information regarding measurement target on the basis of captured images
DE112004002219B4 (en) Method for driving a robot
US9510914B2 (en) System and methods for intraoperative guidance feedback
JP4972201B2 (en) Man-machine interaction based on sign
EP1950688B1 (en) Method for automatically following the movements of a hand in a sequence of images
Holte et al. Human pose estimation and activity recognition from multi-view videos: Comparative explorations of recent developments
Ionasec et al. Patient-specific modeling and quantification of the aortic and mitral valves from 4-D cardiac CT and TEE
JP5167248B2 (en) Modeling of humanoid shape by depth map
EP1260179A1 (en) X-ray image registration device with a medical navigation system
US20040109595A1 (en) Method for automated analysis of digital chest radiographs
CN102165492B (en) Tracking of hair follicles
JP2007333690A (en) Motion capture system, method, and program
US7609865B2 (en) 3D fingerprint and palm print data model and capture devices using multi structured lights and cameras
US7804998B2 (en) Markerless motion capture system
US6978040B2 (en) Optical recovery of radiographic geometry
US7340100B2 (en) Posture recognition apparatus and autonomous robot
KR20110139694A (en) Method and system for gesture recognition
JP2012518856A (en) Capturing and recognizing hand postures using internal distance shape related methods
JP4886560B2 (en) Information processing apparatus and information processing method
CN100486476C (en) Method and system for automatic generating shoe sole photopolymer coating track based on linear structure optical sensor
CN101566465B (en) Method for measuring object deformation in real time
CN101561710B (en) Man-machine interaction method based on estimation of human face posture

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance
C02 Deemed withdrawal of patent application after publication (patent law 2001)