CN101593022B - Method for quick-speed human-computer interaction based on finger tip tracking - Google Patents

Method for quick-speed human-computer interaction based on finger tip tracking Download PDF

Info

Publication number
CN101593022B
CN101593022B CN 200910040699 CN200910040699A CN101593022B CN 101593022 B CN101593022 B CN 101593022B CN 200910040699 CN200910040699 CN 200910040699 CN 200910040699 A CN200910040699 A CN 200910040699A CN 101593022 B CN101593022 B CN 101593022B
Authority
CN
China
Prior art keywords
image
hand
color
fingertip
obtained
Prior art date
Application number
CN 200910040699
Other languages
Chinese (zh)
Other versions
CN101593022A (en
Inventor
徐向民
Original Assignee
华南理工大学
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 filed Critical 华南理工大学
Priority to CN 200910040699 priority Critical patent/CN101593022B/en
Publication of CN101593022A publication Critical patent/CN101593022A/en
Application granted granted Critical
Publication of CN101593022B publication Critical patent/CN101593022B/en

Links

Abstract

The invention discloses a method for quick human-computer interaction based on finger tip tracking, which comprises the following steps: firstly, preprocessing an image, adopting a wide-angle high-definition camera to perform high-resolution image pick-up on an indoor area, and performing image geometric distortion correction on the obtained images; secondly, extracting hand images, processing the obtained corrected images by a skin color filter, a motion filter and a color splitter, merging the results, and splitting the hand images out of the obtained corrected images; and finally, performing finger tip positioning: applying a histogram to perform rough finger tip positioning, using the rough position of a finger tip as a center to construct a searching window, and performing precise finger tip positioning through template matching. The method effectively improves the perceiving efficiency of scene images, achieves finger detection and positioning in a large range, judges the image distance between two eyes as a basis to perform the finger tip positioning without prior knowledge of the environment, and has outstanding distance robustness.

Description

一种基于指端跟踪的快速人机交互方法技术领域[0001] 本发明涉及基于指端跟踪的的人机交互方法,属于计算机视觉和视频跟踪领域, 适用于虚拟现实系统中的人机交互环节。 BACKGROUND rapid human-computer interaction based finger tracking [0001] The present invention relates to human-computer interaction method based finger tracking, which belongs to the field of computer vision and video tracking for virtual reality system in human-computer interaction links . 背景技术[0002] 人机交互技术是虚拟现实系统的关键技术之一,它实现了人与计算机、现实世界与虚拟世界的交互。 [0002] human-computer interaction technology is a key technology of virtual reality system, which implements human interaction with the computer, the real world and the virtual world. 手势是一种自然而直观的人际交流方式。 Gestures are a natural and intuitive way of interpersonal communication. 基于视觉的人手跟踪及手势识别是实现新一代人机交互必不可少的一项关键技术。 Based on visual hand tracking and gesture recognition is a key technology for a new generation of human-computer interaction essential. [0003] 在人机交互中,手部的跟踪主要有数据手套和视觉识别等方法。 [0003] In human-computer interaction, the tracking of the hand glove with a primary data and visual recognition method. [0004] 数据手套,即人可以戴上一个类似于手套的传感器,计算机通过它可以获取手的位置和手指的伸展状况等丰富信息。 [0004] The data glove, i.e., one can put a sensor similar glove, by which the computer can access to a wealth of information and the extended position of the finger of the hand condition and the like. 如1993年.BThmaas等人做的自由手控制目标的系统是凭借数据手套作为输入的媒介,但这需要实验者带上一个专用设备。 The system in 1993 .BThmaas, who do free hand control targets by virtue of the data input glove as a medium, but it requires the experimenter to bring a special equipment. 这不便于使用和推广。 It's not easy to use and promotion. 随着计算机硬件的发展,计算机视觉已逐渐应用于手部定位与跟踪。 With the development of computer hardware, computer vision has been gradually applied to hand positioning and tracking. 基于视觉的手部跟踪方法,首先要考虑的就是环境因素,特别是背景。 Vision-based hand tracking method, the first factor to consider is the environment, especially the background. 为了减少背景的影响,往往采取限制背景,如背景全为黑色或白色等。 To reduce the influence of the background, often taken to limit the background, such as background and so on are all black or white. 还有就是采用手指标记的方法,但对使用者造成不便,因此手部跟踪的注意力便转移到自然手的跟踪。 There is a method of finger marks, but the user inconvenience, therefore hand tracking of the attention shifted to natural hand tracking. [0005] 基于视觉的手部跟踪的方法已经越来越广泛,特别在手写识别的领域。 [0005] The method of vision-based tracking hand has become increasingly widespread, especially in the field of handwriting recognition. 由于面向个人的智能手机等智能设备的迅速发展,目前大部分的日常应用所采取的手部跟踪都是局限在小范围,即采集的视频中都不包含除了手部之外的其他部分,这限制了使用者及其手部的活动范围,只能在摄像系统附近操作计算机,缺乏灵活性,也不太适合边操作边交流的多人场合。 Due to the rapid development of personal-oriented smart phones and other smart devices, the current hand tracking most everyday applications are taken confined in a small area, that is not included in the captured video in addition to other parts of the hand, which It limits the scope of activities of users and their hands, and only in the vicinity of operating a computer camera systems, lack of flexibility, it is not suitable side edge exchange operations multiplayer occasion. 发明内容[0006] 本发明的目的在于,提供一种基于计算机视觉的可大范围活动的使用手指控制的人机交互方法。 SUMMARY OF THE INVENTION [0006] The object of the present invention is to provide an interactive method of using a large range of activities may be based on computer vision finger control. [0007] 本发明的目的通过如下技术方案实现:[0008] 一种基于指端跟踪的快速人机交互方法,包括以下步骤:[0009] (1)图像预处理:采用一个60至120度的广角高清摄像头对一室内区域进行高分辨率摄像,对所得图像进行图像几何畸变纠正;[0010] (2)手部图像提取:对步骤(1)所得的校正图像应用肤色滤波器、运动滤波器和色彩分割器进行处理,然后把结果融合,把手部图像从步骤(1)所得的校正图像中分割出来, 包括如下步骤:[0011] a、进行肤色滤波处理,采用TSL色彩模型将与皮肤颜色接近图像区域选择出来, 对肤色滤波后得到的二值图像,其中肤色区域为1,其他区域为0,再进行膨胀运算,减少肤色滤波导致的空洞;[0012] b、采用图像帧间差分的算法进行运动滤波处理,将运动区域从静态背景中分离出来;[0013] C、对步骤a膨胀运算后的二值图像进行色彩 [0007] The object of the present invention is achieved by the following technical solutions: [0008] A fast interactive tracking finger based method, comprising the steps of: [0009] (1) image preprocessing: using a 60 to 120 degrees HD camera of a wide indoor area high-resolution image, the resulting image is a geometric distortion correction image; [0010] (2) extracts a hand image: step (1) is obtained corrected image color filter applications, the motion filter and a color divider for processing, then the results fusion, the grip portion of the image (1) corrected image obtained in the dividing step out, comprising the steps of: [0011] a, for color filtering, using TSL color model with a skin color close to the selected image region, the binary image obtained by the color filter, which is a skin color region, the other region is 0, then the expansion operation, reducing voids caused by the color filter; [0012] b, using frame difference image motion filtering algorithm, movement to separate from the static background region; [0013] C, after the binary image dilation operation for a color step 割,得到的图像为包含脸部和手部肤色区域的二值图像,提取手部的完整图像;[0014] d、在步骤a、b、c处理的基础上,进行图像融合,将手部的完整图像与运动滤波得到的图像进行“与”运算,并对结果进行膨胀运算,得到主要包含手部区域的二值图像;对主要包含手部区域的二值图像,如果判断该图像为新使用者手部区域的二值图像,按从上到下、从左到右的优先顺序搜索人脸,并计算其双眼的图像距离,第一个符合双眼图像距离条件的人取得控制权;如果判断该图像为已知使用者手部区域的二值图像,则跟踪其脸部位置,重新计算双眼图像距离,若符合双眼图像距离条件,则根据手部在脸部右下方而预先设定的搜索区域中跟踪手部,具体步骤在图像的对应区域中进行,并在图像中进一步消除非手部的图像,得到仅包含手部的二值图像;若双眼 Cut, the resulting image is a binary image containing the face and hands of the skin color region, extracting a complete hand image; [0014] d, based on the step a, b, c of the process, image fusion, the hand portion filtered complete image obtained motion image is expanded operation "and" operation, and the results, obtained binary image of the hand region mainly containing; primary binary image comprising a hand area, if it is determined that the image is a new binary image area of ​​a user's hand, top to bottom, left to right search priority face, eyes and calculate the distance image, the first matching image distance condition human eyes take control; if determining whether the image is a binary image area of ​​a user's hand is known, the tracking its face position, eye image distance is recalculated, if they meet the condition of the eyes from the image, according to the lower right hand portion of the face is set in advance track search region hand, in the step corresponding to the specific region of the image, and to further eliminate the non-image portion of the hand in the image, to obtain a binary image containing only the hand; if binocular 图像距离不符合图像中能清楚地分辨出来的要求,放弃控制权,不再进行指尖定位,转而搜索新的使用者;[0015] (3)指尖定位:对步骤⑵中得到的仅含手部的图像进行指尖定位,首先,应用直方图进行指尖的粗略定位,规定使用者的指尖向上;对O)中得到的仅包含手部的二值化图像,通过边缘检测进行手部轮廓提取,并把轮廓点进行横纵坐标投影,从上至下、从左至右,搜索到投影值明显变化的地方,作为手指尖的粗略位置,以此位置为中心,构造一个搜索窗口;然后,通过模板匹配进行指尖的精确定位。 Image distance images can not meet the requirements clearly resolved, and relinquishes control, no longer fingertip location, instead of searching for a new user; [0015] (3) fingertip location: only obtained in step ⑵ containing the image of the hand fingertip position, first, coarse positioning application histogram fingertip, the user's fingertip predetermined direction; binarized image contains only the hand O) are obtained by edge detection contour extraction hand, and the horizontal and vertical coordinates contour points projected, from top to bottom, from left to right, where the projection values ​​searched significant change, as a rough position of the fingertip, this position as a center, a construct search window; then, precise positioning of the fingertip by template matching. [0016] 所述的对所得图像进行图像几何畸变纠正是指采用三次多项式变形技术和双线性插值法消除从广角摄像头采集到的图像的几何畸变失真。 The resulting image is an image of a geometric distortion correction means cubic polynomial modification techniques and bilinear interpolation to eliminate from the wide-angle camera to capture an image of the geometric distortion is [0016]. [0017] 所述的三次多项式变形技术和双线性插值法,是通过选定一个自定义的基准图像及其畸变图像,联立方程组,并通过最小二乘法求解,确定理想图像与畸变图像的具体变换关系。 [0017] The modification of the cubic polynomial and bilinear interpolation technique is, simultaneous equations, solved by the least square method and by selecting the reference image and a distorted image of the custom, is determined over the image and the distorted image the specific conversion relations. [0018] 4所述的“与”运算是指色彩分割后得到的二值图像与运动滤波得到的二值图像之间的,逐点进行的二进制逻辑与运算。 "And" operation means, point by point between the binary image obtained by dividing the color image and the binary motion filtering is performed to obtain a binary AND operation [0018] 4. [0019] 所述的膨胀运算是指数学形态学中用已定义的模板子图像对原图像进行的能实现平滑或减少图像空洞运算。 Dilation [0019] The template is a sub-picture index with a defined morphology learn the original image can reduce the image or voids smooth operation. [0020] 所述步骤(3)的搜索窗口是指以指尖粗定位位置为中心的,且其边长由统计设定的、大小为粗定位位置与实际精确位置的最大误差的两倍的矩形窗口。 [0020] The step (3) refers to a search window centered on the position fingertip coarse positioning, and the side length is set by the statistics, the coarse position location is twice the actual size and precise location of the maximum error rectangular window. [0021] 所述步骤(3)的模板匹配是指用已定义的若干个手指指端模板去匹配得到的指端图像,找到最佳匹配模板与指端图像的最佳匹配位置,该匹配位置即为指尖的精确位置。 [0021] The step (3) refers to the use of template matching of a plurality of fingers have been defined to match the template finger finger image obtained by finding the best match position of the best matching template finger of the image, the matching position is the exact position of the fingertip. [0022] 所述步骤d符合双眼图像距离条件是指该图像距离大到保证使用者是符合正对摄像头的使用规则,同时也保证了手指在图像中是能够分辨出来的。 [0022] Step d meet the conditions of eyes from the image means that the image distance is large enough to ensure that users comply with rules being used camera, but also ensure that the finger is able to distinguish the image out. [0023] 与现有技术相比,本发明具有以下优点:[0024] (1)本发明采用广角镜头(60至120° )及高分辨率的摄像头摄像,允许使用者在比较大的区域内(摄像头视角范围内)活动,并使用图像变形修正算法,有效提高场景图像感知效率,实现较大范围的手指检测和定位;[0025] (2)按区域划分的优先级对多使用者情况进行管理,同时缩小了计算量,极大提高速度和效率;[0026] (3)对自然手进行实时的跟踪定位,无需任何指端标记,具有更高实用性;[0027] (4)使用肤色滤波与运动检测相结合的方式对手部进行定位跟踪,不仅提高了手部定位的准确性,而且增加了对环境(特别是背景)的适应性,能够适应一般的室内应用, 并能够在比较复杂的背景情况下使用;[0028] (4)把双眼的图像距离判断作为能够进行指端定位的依据(图像距离过大表明手指在图像的分辨率太低 [0023] Compared with the prior art, the present invention has the following advantages: [0024] (1) the present invention uses a wide-angle lens (60 to 120 °) and high-resolution camera imaging, allow a user over a relatively large area ( the camera viewing angle) activity, and using an image distortion correction algorithm to improve the image of the scene perceived efficiency, and positioning a large finger detection range; [0025] (2) priority of the divided areas to manage multiple user situation while reducing the calculation amount greatly improve the speed and efficiency; [0026] (3) real-time tracking of natural hand positioning, without any finger marks, having a higher practicability; [0027] (4) using the color filter opponents of the motion detection mode and the combination of positioning tracking not only improves the accuracy of the positioning of the hand, and increased adaptability to the environment (especially the background), and able to adapt to the general indoor use, and can be more complex using the background; [0028] (4) determines the image distance of both eyes as a basis can be positioned fingertip (finger image distance is too large too low to show the resolution of the image 而导致无法精确定位),无需环境的任何先验知识,故距离鲁棒性突出,优于现有技术。 Result can not pinpoint), without any prior knowledge of the environment, so the distance robustness outstanding, over the prior art. 附图说明[0029] 图1为基于指端跟踪的快速人机交互系统结构示意图,示出了本发明的第一实施例的基于指端跟踪的快速人机交互系统的系统结构。 BRIEF DESCRIPTION [0029] FIG. 1 is a schematic system configuration of the interactive man-machine fast tracking finger based, shows a system configuration of finger based on a fast interactive system to track a first embodiment of the present invention. [0030] 图2为基于指端跟踪的快速人机交互方法流程框图,示出了基于指端跟踪的快速人机交互系统的具体实现方法的步骤。 [0030] FIG. 2 is a block flow diagram of a method interactive fast tracking finger based, shows a specific implementation of steps Fast interactive system to track the finger. 具体实施方式[0031] 下面结合实施例对本发明作进一步的描述,但需要说明的是,实施例并不构成对本发明要求保护范围的限制。 DETAILED DESCRIPTION [0031] The following embodiments in conjunction with embodiments of the present invention will be further described, it should be noted that the embodiments do not limit the scope of the claimed invention. [0032] 如图1所示,基于指端跟踪的快速人机交互系统包括广角高清摄像头101、DSP(数字信号处理器)设备102和计算机106。 [0032] 1, the interactive system fast tracking finger based on a wide-definition camera 101, DSP (digital signal processor) device 102 and the computer 106. DSP设备102包括图像采集部103、信号转换部104 和图像预处理部105 ;广角高清摄像头101与DSP设备102的图像采集部103信号连接,图像采集部103与信号转换部104和图像预处理部105依次信号连接;图像预处理部105与计算机106信号连接。 DSP device 102 comprising an image acquisition unit 103, a signal conversion unit 104 and the image pre-processing unit 105; wide HD camera DSP device 101 and the image acquisition unit 102 is connected to signal 103, the image acquisition unit 103 and the signal converting unit 104 and an image pre-processing unit signal 105 are sequentially connected; pre-processing unit 105 is connected to the image signal from the computer 106. 采用广角高清摄像头101进行检测,由DSP设备102的图像采集部103进行图像采集,通过信号转换部104将视频模拟输入信号转换为图像数字信号,并通过图像预处理部105进行图像预处理,然后由计算机106完成手部提取和指尖的识别定位。 High-definition wide-angle camera 101 is detected, performed by the image acquisition unit 103 of DSP image pickup device 102, via the signal conversion unit 104 converts the analog video image signal into a digital signal input, and an image pre-processing by the image pre-processing unit 105, and then complete identification and positioning the hand fingertip is extracted by the computer 106. [0033] 广角摄像头101负责采集大范围的高清晰度图像;DSP设备102设备负责将模拟图像信号转换成数字图像信号,并进行图像几何畸变纠正的图像与处理。 [0033] The wide-angle camera 101 is responsible for collecting high-resolution images of a wide range; the DSP device 102 device is responsible for converting the analog image signal into digital image signals, image processing and image geometric distortion correction. 计算机106负责完成通过肤色滤波、运动滤波和色彩分割来提取使用者的手部,然后进一步识别指尖的位置,最后把指尖的图像坐标转换成实际的坐标,实现控制输出。 The computer 106 is responsible for the completion of the hand by the user to extract color filtering, color filtering and motion segmentation, and further identify the position of the fingertip, and finally converting the image coordinates into actual coordinates of the fingertip, by controlling both the. [0034] 广角高清摄像头101可选美国微软公司的LifeCam NX-6000广角高清摄像头;DSP 设备102具体可选用TI公司的TMS320系列处理器。 [0034] Alternatively wide definition camera 101 US Microsoft LifeCam NX-6000 high-definition wide-angle camera; the DSP device 102 may specifically use TI TMS320 family of processors. [0035] 如图2所示,基于指端跟踪的快速人机交互方法具体包括如下步骤:[0036] (1)图像预处理:采用一个60至120度的广角高清摄像头对一室内区域进行高分辨率摄像,对所得图像进行图像几何畸变纠正。 [0035] 2, rapid interactive tracking finger based method includes the following steps: [0036] (1) Image preprocessing: a wide-angle high-definition camera 60 to 120 degrees to a high indoor area resolution imaging, the resultant image of the image geometric distortion correction. 从广角摄像头采集到的图像存在比较严重的几何畸变,因此在后续图像处理之前,必须进行图像几何畸变纠正。 From the wide-angle camera to capture the image exists serious geometric distortion, and therefore before a subsequent image processing, image geometric distortion must be corrected. 为运算量简化下,获得尽可能好的修正效果,可采用三次多项式变形技术和双线性插值法消除从广角摄像头采集到的图像的几何畸变失真。 To simplify the computation, to obtain the best possible correction effect, cubic polynomials can be used and modified bilinear interpolation technique to eliminate from the wide-angle camera to capture the geometric distortion of the image distortion. 三次多项式变形技术和双线性插值法具体如下:[0037] 设理想图像g的像素坐标为(u,ν),畸变图像f对应像素坐标为(X,y),则三次多项式坐标变换关系为:[0038] Cubic polynomial deformation and bilinear interpolation technique, as follows: the pixel coordinates [0037] is provided over the image g (u, ν), the distorted image corresponding to the pixel coordinates as f (X, y), the coordinate transformation cubic polynomial relationship : [0038]

Figure CN101593022BD00071

其中aij和(i,j = 0,1,2,3)为待定的多项式系数;式①中的多项式系数和bu(i,j =0,1,2,3)是仅与摄像头相关的参数,可通-个自定义的基准图像及其畸变图像,联立方程组,并通过最小二乘法求解,得到%[0039][0040]过选定,HlZ的值,从而确定理想图像与畸变图像的具体变换关系。 And where aij (i, j = 0,1,2,3) for the polynomial coefficients to be determined; polynomial coefficients in the formula ① and bu (i, j = 0,1,2,3) associated with only the camera parameters , can pass - custom reference image and image distortion, simultaneous equations and solved by the least squares method to obtain% [0039] [0040] through the selected HlZ value, thereby determining the ideal image and the distorted image the specific conversion relations. 由于式①计算得到的x,y不一定为整数,所以不能直接使用g(u,ν) = f(x, y),而必须进行灰度差值运算。 ① Since x calculated by the formula, y is not necessarily an integer, it can not be used directly g (u, ν) = f (x, y), but must gradation difference calculation. 因此,采用双线性插值法,即式②:[0041] g(u, ν) = (1-α ) (l-β )f (χ0, y0) + α (1-β )f (χ0+1, y0) + (la ) β f (χ0,Yo+D + α β f(x0+l,Y0+D[0042] 其中α = χ-χ0, β = y-y0 ;f, g分别为对应的畸变图像和理想图像;;x,y可由将U,ν代入式①得到,且h,ytl分别为不大于x,y的最大整数;对所有的像素点(U,ν)进行式②的运算,最终得到理想图g ;即实现图像及和畸变纠正。[0043] (2)手部图像提取:对步骤(1)所得的校正图像应用肤色滤波器、运动滤波器和色彩分割器进行处理,然后把结果融合,把手部图像从步骤(1)所得的校正图像中分割出来, 具体步骤如下:[0044] a、进行肤色滤波处理,采用TSL色彩模型将与皮肤颜色接近图像区域选择出来。 TSL色彩模型进行肤色滤波,比RGB、HIS、YIQ与CIELUV模型过滤出来的肤色区域准确。其中,TSL色彩模型与RGB模型的转换见式③;TSL色彩 Thus, using the bilinear interpolation method, i.e. the formula ②: [0041] g (u, ν) = (1-α) (l-β) f (χ0, y0) + α (1-β) f (χ0 + 1, y0) + (la) β f (χ0, Yo + D + α β f (x0 + l, Y0 + D [0042] where α = χ-χ0, β = y-y0; f, g respectively correspond the distorted image and the ideal image ;; x, y may be U, ν substituted into equation ① obtained, and h, ytl are not greater than x, the largest integer of y; for formula ② all the pixels (U, ν) of operation, FIG finally get over G; i.e., distortion correction, and to achieve image and [0043] (2) extracts a hand image: step (1) is obtained corrected image color filter applications, motion segmentation and color filters for processing. and the fusion result, the grip portion of the image (1) corrected image obtained in the dividing step out the following steps: [0044] a, for color filtering, using TSL color model will be selected with a skin color close to the image area. TSL color model color filter, than RGB, HIS, YIQ model filtered and CIELUV color region wherein accuracy, the RGB color model TSL model into see formula ③;. TSL color 空间把亮度及色度分开处理,RGB模型为原始图像的色彩模型,转换为TSL模型有助于肤色丛集化。 [0045] The spatial luminance and chrominance separately, the RGB color model for the model original image, color conversion to a clustered models help TSL. [0045]

Figure CN101593022BD00072

[0046]其中 [0046] in which

Figure CN101593022BD00073

[0047] R、G、B分别为RGB色彩模型下的RGB分量;Τ、S、L分别为TSL色彩模型下的Τ、S、L分量。 [0047] R, G, B are the RGB components of the RGB color model; Τ, S, L are Τ under TSL color model, S, L component. 由于彩色图像一般采用RGB模型,所以在使用TSL模型时,必须用式③进行转换。 Since a color image is generally the RGB model, so the model TSL in use, must be converted by the formula ③. [0048] 通过对500张包含肤色区域的图像的脸部和手部区域进行采样,估计TSL模型下肤色0~和幻的(二维高斯分布)概率分布参数(均值矩阵E和协方差矩阵Σ );并且采用马氏距离进行肤色的判别,即对每个像素进行检测,若一个像素的T和S分量组成的C = (Τ,S)向量与均值向量E的马氏距离低于某个阈值Threshold,则认为该像素属于肤色区域。 [0048] By the face and hand region of the image containing color areas 500 is sampled, the color model is estimated Σ TSL 0 and phantom (two-dimensional Gaussian) probability distribution parameters (mean and covariance matrix E matrix ); Mahalanobis distance and color discrimination, i.e., each pixel is detected, if a pixel is composed of components S and T C = (Τ, S) vector and mean vector E Mahalanobis distance is below a certain threshold threshold, it is considered that the pixel belongs to the skin color region. 具体如下:[0049]马氏距离 d = (CE)τ Σ — 1(CE)[0050] 若d < Threshold,像素属于肤色区域[0051] 若d > Threshold,像素不属于肤色区域[0052] 其中Threshold为阈值,在得到肤色区域统计数据(均值矩阵E和协方差矩阵Σ )后,估计正常肤色(T,S)与均值E的距离,得到初始值,再经过实验调整,可确定阈值为0. 99。 As follows: [0049] Mahalanobis distance d = (CE) τ Σ - 1 (CE) [0050] Threshold, the pixel does not belong to the skin color region [0052] where if d <If d Threshold, pixels belonging to color region [0051]> threshold is a threshold value, after obtaining the skin color region statistics (mean matrix E and covariance matrix [Sigma), estimated from the normal color (T, S) and the mean value E to give an initial value, and then after experimental adjustment, may determine the threshold value is 0 99. [0053] 对肤色滤波后得到的二值图像(肤色区域为1,其他区域为0)进行膨胀运算(模板子图像使用3X3模板子图像,且模板中每个像素均为1),减少肤色滤波导致的空洞。 [0053] The binary image obtained color filter (a color region, the other region 0) for a dilation operation (3X3 template using the template for the sub-picture image, and each pixel in the template are 1) to reduce the color filter cause cavities. [0054] b、在进行上述a处理的同时,并行地进行运动滤波处理,将运动区域从静态背景中分离出来:具体采取图像帧间差分的算法来检测出运动区域,为了防止把由于摄像系统等原因导致的偶发性变化判为运动,进一步加强条件,即在连续5帧图像里至少有3帧都变化的像素才视为运动像素。 [0054] b, a parallel with the above processing, filter processing is performed in parallel motion, the motion region is separated from the static background out: taking particular image frame difference algorithm to detect motion regions, since the image pickup system in order to prevent sporadic changes and other causes sentenced to sports, to further strengthen the conditions, that there are at least three changes are only considered as a pixel motion image pixels in five consecutive frames inside. 根据各像素运动状态将原始图像二值化,得到的二值图像记为B,对B的每个像素B (i,j):[0055]{1- (/,y)为运动像素(/,y)为静止像素[0056] 对差分后得到的二值图像B进行腐蚀运算(使用3X3模板子图像,且模板中每个像素均为1),更清晰的把运动区域和背景区分开来。 The state of motion of each pixel of the original image is binarized to obtain a binary image referred to as the B, B for each pixel B (i, j): [0055] {1- (/, y) is the motion pixels (/ , y) [0056] B binary images obtained after etching difference calculation (using the template sub-image is a still 3X3 pixels, each pixel in the template and are 1), the clearer the motion region and the background to distinguish . [0057] C、在步骤a处理的基础上,进行色彩分割,提取手部的完整图像:[0058] 由于亮度的影响,肤色滤波器有可能把属于肤色的像素或区域错判为非肤色。 [0057] C, at a processing step on the basis of, for color segmentation, the complete extraction of the image of the hand portion: [0058] due to the influence of the brightness, color filter having color pixel may belong to miscarriage or non-skin color region. 为了尽量排除误判的影响,加入色彩分割器,定义色彩的相似度量,把原始图像分成几块区域,这样,整个手部形成了一个连通域。 In order to eliminate the influence of possible false, adding color segmentation, a similarity measure defined color, the original image is divided into a few regions, so that the entire hand is formed a communication domain. 使用RGB色彩空间下的色彩相似度量,具体如下(对相邻像素a和b):[0059] Use of color in the RGB color space similarity measure, as follows (adjacent pixels a and b): [0059]

Figure CN101593022BD00081

[0060] 其中,Ra,Ga和Ba分别为a的RGB分量;Rb,Gb和Bb分别为b的RGB分量[0061] 若ρ (a, b) < threshold则a和b属于同一个色彩区域,否则属于不同的色彩区域;thresholds是估计不同区域像素点的RGB向量距离,得到初始值,再经过实验调整确定;本发明确定thresholds为12。 [0060] wherein, Ra, Ga and Ba are RGB components a,; Rb, Gb and Bb are b RGB components [0061] If ρ (a, b) <threshold then a and b belong to the same color region, otherwise to the different color regions; thresholds are estimated from the different regions of the vector of RGB pixel, to obtain an initial value, and then determine experimentally adjusted; determining thresholds according to the present invention is 12. [0062] 考虑误判被排除的肤色点与肤色滤波得到的肤色区域,在共同组成完整的肤色区域中应是连通的,因此采用区域生长的色彩分割方法,即以肤色滤波得到的肤色区域中所有点为种子点,进行区域生长,最后得到包含完整肤色区域的二值图像C。 [0062] consideration of the skin color region color point and color filter to be false positives are excluded in the skin color region together form a complete communication should be, so a color region growing segmentation, i.e. to the skin color region of the color filter obtained All points seed point, region growing, and finally obtain a complete binary image comprising a skin color area C. [0063] d、在上述步骤a、b、c处理的基础上,进行图像融合。 [0063] d, in the above steps a, b, c based on the processing, image fusion. 首先,色彩分割后得到的图像C为包含脸部和手部等肤色区域的二值图像:考虑到手部的运动,将图像C与运动滤波最终得到的图像进行“与”运算,并对结果进行膨胀运算(模板子图像使用3X3模板子图像,且模板中每个像素均为1),就得到主要包含手部区域的二值图像D(与图像C相比,仅滤除了非运动区域,其中可能包含脸部)。 First, the color image C obtained by dividing binary image containing color areas like the face and hands: hand movement this part, and the image C is finally obtained motion filtering "and" operation, and the results were dilation (the template for the template image using the 3X3 sub-images, each pixel in the template and are 1), to obtain a binary image region D mainly comprises a hand portion (as compared with the image C, filtering off only the non-motion area, wherein It may contain face). [0064] 最后,在二值图像C中,如果判断该图像为新使用者手部区域的二值图像,就按从上到下、从左到右的优先顺序搜索人脸,并计算其双眼的图像距离,第一个符合双眼图像距离条件的人取得控制权,取得控制权的用户即为被跟踪的用户,系统继续进行步骤(3)指尖定位步骤;符合双眼图像距离条件是指该图像距离应该足够大,使得保证使用者是符合正对摄像头的使用规则,同时也保证手指在图像中有足够分辨率,以便于指尖的定位,如可设双眼图像距离为大于10个像素距离。 [0064] Finally, the binary image C, if it is determined that the new image is a binary image area of ​​a user's hand, on the top to bottom, left to right of priority to search for faces, and calculating the binocular the image distance, first to meet the user of the user image from the eyes, of control conditions, achieving control shall be tracked, the system proceeds to step (3) the step of positioning the fingertip; eyes meet the condition means that the image distance image distance should be large enough to ensure that the user is in line with the rules being used for cameras, but also to ensure that there is sufficient resolution in the finger image, to facilitate positioning of the fingertip, as may be provided for the eyes image from pixel distance is greater than 10 . 相应地,不符合双眼图像距离要求是指图像距离过小,小于预设值,则意味着脸部没有正对着摄像头,不符合使用规则或者人离开摄像头太远,导致无法进行跟踪;如果判断该图像为已知使用者手部区域的二值图像,则跟踪其脸部位置,重新计算双眼图像距离,若符合双眼图像距离条件,则根据手部在脸部右下方而预先设定的搜索区域中跟踪手部,在图像D的对应区域中进行,并在图像D中进一步消除非手部的图像,最终得到的二值图像记为H ;若双眼图像距离不符合双眼图像距离条件,放弃控制权,不再进行指尖定位,转而在图像C中继续搜索新的使用者(人脸),若整个图像C中均找不到符合双眼图像距离图像的人脸,则返回步骤(1),重新开始。 Accordingly, the eyes do not meet the requirements from the image refers to an image distance is too small, less than the predetermined value, it means that the face is not facing the camera, does not meet the rules or to leave the camera is too far, resulting in not track located; the image is a binary image area of ​​a user's hand is known, the tracking its face position, eye image distance is recalculated, if they meet the condition of the eyes from the image, the search and the lower right hand portion of the face is set in advance in accordance with area tracking hand, in the corresponding area of ​​the image D, and to further eliminate the image non-hand portion of the image D, the resulting binary image referred to as H; if the eyes image distance does not meet the eyes image distance condition, to give control, no longer fingertip positioning, in turn, continue to search for new users (face) in image C, C if the entire image in both eyes to find people in line image from the image of the face, return to step (1 ),Restart. [0065] (3)指尖定位:对步骤(2)中得到的仅含手部的图像H进行指尖定位。 [0065] (3) fingertip location: in step (2) of the hand image H obtained containing only the fingertip is positioned. [0066] 首先,应用直方图进行指尖的粗略定位:对二值图像H进行轮廓提取(提取轮廓图像设为Hl),并进行网格大小为2X2像素的网格采样(网格中存在轮廓点,在采样后的图像中对应的点仍为轮廓点),以保证轮廓的连续性,网格采样后的图像设为H2。 [0066] First, the application coarse positioning fingertip histogram: H binary image contour extraction (extracting a contour image is set on Hl), and the sampling grid mesh size of 2X2 pixels (present contour grid point, the corresponding image points in the sampled contour points still), to ensure the continuity of the outline, the image is set to the sampling grid H2. 由于指尖的粗略位置一般是轮廓在四个方向的顶点之一,手指可以近似看成由矩形和一个半圆组成,因此在H2中找出4个候选点G个方向的顶点)后,对每个候选点,分别从逆时针和顺时针选择第2、3、4共3个轮廓点,构成3个像素对,由于手指的宽度近似是不变的,所以计算每个候选点的邻近的3对像素点的距离的方差,方差最小的候选点就是最佳的候选点,根据该候选点在H2中的位置,在图像Hl (或H)中找到对应的候选点,作为手指尖的粗略位置。 Since the rough position of the fingertip is generally one of the four vertices of the contour in the direction of the finger can be approximated by a rectangle and a semicircle as composition, thus find four vertices candidate points in the direction G H2), each candidate points, were selected from a total of counterclockwise and clockwise 2,3,4 three contour points constituting the three pixels, the width of the finger is approximately constant, so the calculation for each candidate point adjacent to the three pairs variance, the minimum variance is the best candidate point from the candidate point of the pixel, in accordance with the position of the candidate points in H2, find the corresponding candidate points in an image on Hl (or H), as the rough position of the fingertip. [0067] 在轮廓图像Hl中,以手指尖的粗略位置为中心,构造一个的搜索窗口,该搜索窗口中的轮廓点都有可能是指尖的精确位置,对指尖的精确定位便在该窗口进行。 [0067] In the contour image Hl, the rough position of a finger tip to the center, a construct of the search window, the search window outline points are likely to be the exact position of the fingertip, then precise positioning of the fingertip window. 搜索窗口以能包括指尖的所有可能的精确位置为准,一般可以设置为9X9像素。 In all possible search window can include a precise position of the fingertip subject, generally can be set 9X9 pixels. 然后,对搜索窗口的所有轮廓点,进行模板匹配,找出指尖的精确位置:模板匹配是目前手指检测中常用的方法,模板匹配是指用预定义好的若干个手指指端模板去匹配得到的指端图像,找到最佳匹配模板与指尖图像的最佳匹配位置,以使得绝对距离测度最小,该匹配位置即为指尖的精确位置。 Then, for all contour points of the search window, template matching, to identify the exact position of the fingertip: template matching finger detection is commonly used method, template matching refers to the use of several pre-defined templates to match the finger tip of finger obtained finger image, finding the best match position of the best matching template image of the fingertip, so that the absolute minimum distance measure, which is the exact position matching the position of the fingertip. 常用的距离测度有欧式距离、相关距离等。 Commonly used distance measure from Continental, the relevant distance. 本发明采用绝对值距离测度。 The present invention employs the absolute value of the distance measure. 模板匹配方法可用下式来描述:[0069] 式中ρ是搜索窗口中的待匹配子图像,tk代表第k个模板,模板大小为MXN。 Template matching method will be described by the following formula: [0069] where ρ is the sub-image to be matched in the search window, TK represents the k-th template, size MXN. (i,j) 表示搜索窗口中的任意轮廓点,(im,jm)表示最终检测到的手指尖精确位置的坐标,m,η在上式中表示用来求和迭代过程中的临时变量,P (i+m,j+n)表示的是图像在坐标(i+m,j+n) 上的值;考虑到手指指向一般不会向下,选用5个大小为25X25像素的包括0°、45°、 90°、135°和180°的手指指向的指端模板。 (I, j) represents any contour point in the search window, (im, jm) represent the coordinates of the detected fingertip final precise location, m, η represents a temporary variable used in the iterative process summing the above formula, P (i + m, j + n) represents the value of the image at the coordinates (i + m, j + n); a finger pointing generally not taken into account downward, five selected size 25X25 pixels includes 0 ° , 45 °, 90 °, 135 ° and 180 ° finger pointing finger template. 最后把图像中的指尖位置映射到显示器屏幕坐标等控制坐标,作为最终的坐标输出。 Finally, the fingertip position image in the screen coordinates is mapped to the display coordinate control, as the final output coordinates. [0070] 本方法在人数不多的室内环境下基本能够跟踪使用者的手指指尖。 [0070] In this method small number of basic indoor environment can track the user's finger tips. 本方法采用的算法比较简单,容易实现,而且运算复杂度不高。 This method uses the algorithm is relatively simple, easy to implement, and the computational complexity is not high. 通过人眼距离检测的输出量的距离标度, 该方法对于一个距离区间内的使用者均可具有相近的操作体现和输出控制精确度,比现有技术对操作距离有要求的方法具有更高实用性。 Scale by the human eye from a distance of the detected output, the method for the user within a distance of a zone may have the same operating method as embodied and output control accuracy than the prior art have required higher operating distance practicality. [0068] [0068]

Figure CN101593022BD00091

Claims (8)

1. 一种基于指端跟踪的快速人机交互方法,其特征在于包括以下步骤:(1)图像预处理:采用一个60至120度的广角高清摄像头对一室内区域进行高分辨率摄像,对所得图像进行图像几何畸变纠正;(2)手部图像提取:对步骤(1)所得的校正图像应用肤色滤波器、运动滤波器和色彩分割器进行处理,然后把结果融合,把手部图像从步骤(1)所得的校正图像中分割出来,包括如下步骤:a、进行肤色滤波处理,采用TSL色彩模型将与皮肤颜色接近图像区域选择出来,对肤色滤波后得到的二值图像,其中肤色区域为1,其他区域为0,再进行膨胀运算,减少肤色滤波导致的空洞;b、采用图像帧间差分的算法进行运动滤波处理,将运动区域从静态背景中分离出来;c、提取手部的完整图像:对步骤a膨胀运算后的二值图像进行色彩分割,得到的图像为包含脸部和手部肤色 An interactive method for rapid tracking finger based, comprising the steps of: (1) image preprocessing: a wide-angle high-definition camera 60 to 120 degrees to a high-resolution image indoor area, for the resulting image is an image to correct a geometric distortion; (2) extracts a hand image: step (1) application of the resulting color corrected image filter, a color filter and motion segmentation processing, and the integration result, the grip portion image from step (1) corrected image obtained carved out, comprising the steps of: a, for color filtering, using TSL color model will be selected with a skin color close to the image region, the binary image color filter obtained after which the skin color region 1, the other region 0, then the expansion operation, reducing voids caused by color filter; B, inter-frame difference using image filtering algorithms for motion, movement to separate from the static background region; C, complete extraction of the hand images: binary image after a dilation step performs color segmentation, image comprising the face and hands color 域的二值图像;d、在步骤a、b、c处理的基础上,进行图像融合,将手部的完整图像与运动滤波得到的图像进行“与”运算,并对结果进行膨胀运算,得到主要包含手部区域的二值图像;对包含脸部和手部肤色区域的二值图像,如果判断该图像为新使用者手部区域的二值图像,按从上到下、从左到右的优先顺序搜索人脸,并计算其双眼的图像距离,第一个符合双眼图像距离条件的人取得控制权;如果判断该图像为已知使用者手部区域的二值图像,则跟踪其脸部位置,重新计算双眼图像距离,若符合双眼图像距离条件,则根据手部在脸部右下方而预先设定的搜索区域中跟踪手部,具体步骤在主要包含手部区域的二值图像的对应区域中进行,并在主要包含手部区域的二值图像中进一步消除非手部的图像,得到仅包含手部的二值图像;若双眼图像距离不 Binary image field; D, step a, b, c based on the processing, image fusion, the complete image and the image of the hand obtained motion filtering is performed "and" operation, a dilation operation and the results, obtained binary image of the hand region mainly comprises; binary image containing the face and hands of the skin color region, if it is determined that the new image is a binary image area of ​​a user's hand, from top to bottom, left to right search priority face, eyes and calculate the distance image, the first matching image distance condition human eyes take control; if it is determined that the image is a binary image area of ​​a user's hand is known, which face the track position, binocular image distance is recalculated, if they meet the condition of the eyes from the image, then the search area and the lower right hand portion of the face tracking preset hand, mainly comprising specific steps in the binary image of the hand region performed corresponding region and mainly containing the binary image of the hand region to further eliminate the non-image portion of the hand, to obtain a binary image containing only the hand portion; binocular image distance if not 合图像中能清楚地分辨出来的要求,放弃控制权,不再进行指尖定位,转而搜索新的使用者;(3)指尖定位:对步骤O)中得到的仅包含手部的二值图像进行指尖定位,首先,应用直方图进行指尖的粗略定位,规定使用者的指尖向上;对O)中得到的仅包含手部的二值化图像,通过边缘检测进行手部轮廓提取,并把轮廓点进行横纵坐标投影,从上至下、从左至右,搜索到投影值明显变化的地方,作为手指尖的粗略位置,以此位置为中心,构造一个搜索窗口;然后,通过模板匹配进行指尖的精确定位。 Combined images can clearly resolved request to relinquish control, no longer fingertip location, instead of searching for a new user; (3) positioning the fingertip: step O) obtained in the second-hand portion only comprises value image fingertip positioning, first, coarse positioning application histogram fingertip, the user's fingertip predetermined direction; binarized image contains only the hand O) is obtained, performed by the hand contour edge detection extraction, and the horizontal and vertical coordinates contour points projected, from top to bottom, from left to right, where the projection values ​​searched significant change, as a rough position of the hand fingertip position as a center in order to construct a search window; and for precise positioning of the fingertip by template matching.
2.根据权利要求1所述的基于指端跟踪的快速人机交互方法,其特征在于所述的对所得图像进行图像几何畸变纠正是指采用三次多项式变形技术和双线性插值法消除从广角摄像头采集到的图像的几何畸变失真。 The interactive method of fast tracking finger based, wherein said image is an image obtained using the geometric distortion correction means cubic polynomial bilinear interpolation and morphing according to claim 1 elimination from the wide captured by the camera to the geometric distortion of the image distortion.
3.根据权利要求2所述的基于指端跟踪的快速人机交互方法,其特征在于所述的三次多项式变形技术和双线性插值法,是通过选定一个自定义的基准图像及其畸变图像,联立方程组,并通过最小二乘法求解,确定理想图像与畸变图像的具体变换关系。 3. Fast interactive tracking finger based method, wherein according to claim 2 cubic polynomial and morphing according to bilinear interpolation, and a reference image distortion by selecting a custom image, simultaneous equations and solved by the least squares method, to determine the desired image and image distortion specific transformation relationship.
4.根据权利要求1所述的基于指端跟踪的快速人机交互方法,其特征在于所述的“与” 运算是指色彩分割后得到的手部完整图像与运动滤波得到的二值图像之间的,逐点进行的二进制逻辑与运算。 The interactive method of fast tracking finger based, characterized in that the "and" operation refers to the complete image and the binary image of hand motion filtering the obtained color segmentation of the obtained according to claim 1 , binary logic aND operation between the point by point.
5.根据权利要求1所述的基于指端跟踪的快速人机交互方法,其特征在于所述的膨胀运算是指数学形态学中用已定义的模板子图像对原图像进行的能实现平滑或减少图像空洞运算。 The interactive method of fast tracking finger based on 1 wherein said dilation is morphological template sub-image learning index defined by the original image can be smooth or claim reduce image voids operation.
6.根据权利要求1所述的基于指端跟踪的快速人机交互方法,其特征在于所述步骤(3)的搜索窗口是指以指尖粗定位位置为中心的,且其边长由统计设定的、大小为粗定位位置与实际精确位置的最大误差的两倍的矩形窗口。 The interactive method of claim fast tracking finger based, wherein said step (3) of said search window refers to a coarse positioning fingertip position as the center, and the side length of the Statistics set the coarse position location is twice the actual size and precise location of the maximum error of a rectangular window.
7.根据权利要求1所述的基于指端跟踪的快速人机交互方法,其特征在于所述步骤(3)的模板匹配是指用已定义的若干个手指指端模板去匹配得到的指端图像,找到最佳匹配模板与指端图像的最佳匹配位置,该匹配位置即为指尖的精确位置。 The interactive method of claim fast tracking finger based, wherein said step (3) refers finger template matching using a plurality of fingers have been defined to match the template finger of the obtained 1 image, find the best position best matching template matching and finger images, the matching position is the exact position of the fingertip.
8.根据权利要求1所述的基于指端跟踪的快速人机交互方法,其特征在于所述步骤d 符合双眼图像距离条件是指该图像距离大到保证使用者是符合正对摄像头的使用规则,同时也保证了手指在图像中是能够分辨出来的。 8. Rapid interactive tracking finger based method, wherein said step of claim 1 in line with the eyes image distance d refers to the condition of the rule is using the image from the camera to ensure that the user is in line with the large and also to ensure that the fingers are able to tell the difference in the image.
CN 200910040699 2009-06-30 2009-06-30 Method for quick-speed human-computer interaction based on finger tip tracking CN101593022B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200910040699 CN101593022B (en) 2009-06-30 2009-06-30 Method for quick-speed human-computer interaction based on finger tip tracking

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200910040699 CN101593022B (en) 2009-06-30 2009-06-30 Method for quick-speed human-computer interaction based on finger tip tracking

Publications (2)

Publication Number Publication Date
CN101593022A CN101593022A (en) 2009-12-02
CN101593022B true CN101593022B (en) 2011-04-27

Family

ID=41407708

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200910040699 CN101593022B (en) 2009-06-30 2009-06-30 Method for quick-speed human-computer interaction based on finger tip tracking

Country Status (1)

Country Link
CN (1) CN101593022B (en)

Families Citing this family (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2378394A3 (en) * 2010-04-15 2015-03-25 Electronics and Telecommunications Research Institute User interface device and method for recognizing user interaction using same
CN102179814A (en) * 2011-03-23 2011-09-14 浙江大学 Method for controlling robot by using user hand commands
CN102520790A (en) * 2011-11-23 2012-06-27 中兴通讯股份有限公司 Character input method based on image sensing module, device and terminal
CN102592115B (en) * 2011-12-26 2014-04-30 Tcl集团股份有限公司 Hand positioning method and system
CN103196550A (en) * 2012-01-09 2013-07-10 西安智意能电子科技有限公司 Method and equipment for screening and processing imaging information of launching light source
CN103208005A (en) * 2012-01-13 2013-07-17 富士通株式会社 Object recognition method and object recognition device
CN102662460B (en) * 2012-03-05 2015-04-15 清华大学 Non-contact control device of mobile terminal and control method thereof
CN102622601A (en) * 2012-03-12 2012-08-01 李博男 Fingertip detection method
US9471763B2 (en) * 2012-05-04 2016-10-18 Sony Interactive Entertainment America Llc User input processing with eye tracking
CN102799855B (en) * 2012-06-14 2016-01-20 华南理工大学 Video stream based on the positioning method of manpower
CN102799875B (en) * 2012-07-25 2016-01-20 华南理工大学 Any hand hand tracking method
CN102868811B (en) * 2012-09-04 2015-05-06 青岛大学 Mobile phone screen control method based on real-time video processing
KR101393570B1 (en) * 2012-12-28 2014-05-27 현대자동차 주식회사 Method and system for recognizing hand gesture using selective illumination
CN103389799B (en) * 2013-07-24 2016-01-20 清华大学深圳研究生院 A method of opponents fingertip trajectory tracking of
CN103529957A (en) * 2013-11-05 2014-01-22 上海电机学院 Position recognizing device and method
CN103593654B (en) * 2013-11-13 2015-11-04 智慧城市系统服务(中国)有限公司 A method and apparatus for locating face
US9436872B2 (en) * 2014-02-24 2016-09-06 Hong Kong Applied Science and Technology Research Institute Company Limited System and method for detecting and tracking multiple parts of an object
CN104050454B (en) * 2014-06-24 2017-12-19 深圳先进技术研究院 A kind of motion gesture track acquisition methods and system
EP2988248B1 (en) * 2014-07-22 2019-09-18 Canon Kabushiki Kaisha Image processing apparatus, image processing method, information processing apparatus, information processing method, and program
CN104217197B (en) * 2014-08-27 2018-04-13 华南理工大学 A kind of reading method and device of view-based access control model gesture
CN104700073A (en) * 2015-02-10 2015-06-10 广东光阵光电科技有限公司 Method and device for identifying planar fingerprints
CN104700075B (en) * 2015-02-10 2018-06-12 广东光阵光电科技有限公司 A kind of design method and its device of finger scan identification device
CN104796750A (en) * 2015-04-20 2015-07-22 京东方科技集团股份有限公司 Remote controller and remote-control display system
CN104951748B (en) * 2015-04-28 2018-12-07 广东光阵光电科技有限公司 A kind of vacation fingerprint identification method and device
CN105912126B (en) * 2016-04-26 2019-05-14 华南理工大学 A kind of gesture motion is mapped to the adaptive adjusting gain method at interface
CN106709479B (en) * 2017-02-24 2019-11-05 深圳英飞拓科技股份有限公司 Method for detecting human face and system based on video image
CN108012140A (en) * 2017-12-06 2018-05-08 长沙远达华信息科技有限公司 Virtual reality system based on 3D interactions

Also Published As

Publication number Publication date
CN101593022A (en) 2009-12-02

Similar Documents

Publication Publication Date Title
JP5887775B2 (en) Human computer interaction system, hand-to-hand pointing point positioning method, and finger gesture determination method
US9075453B2 (en) Human eye controlled computer mouse interface
US9460339B2 (en) Combined color image and depth processing
CN101344816B (en) Human-machine interaction method and device based on sight tracing and gesture discriminating
JP5567853B2 (en) Image recognition apparatus and method
Gorodnichy et al. Nouse ‘use your nose as a mouse’perceptual vision technology for hands-free games and interfaces
CN101853071B (en) Gesture identification method and system based on visual sense
CN103098076B (en) Gesture recognition system for TV control
US8498444B2 (en) Blob representation in video processing
CN102081918B (en) Video image display control method and video image display device
Pan et al. A real-time multi-cue hand tracking algorithm based on computer vision
CN102360421B (en) Face identification method and system based on video streaming
CN101339607A (en) Human face recognition method and system, human face recognition model training method and system
Jain et al. Real-time upper-body human pose estimation using a depth camera
EP1969560A1 (en) Edge comparison in segmentation of video sequences
CN101533512B (en) Method for automatically extracting interesting image regions based on human visual attention system
CN101142584A (en) Method for facial features detection
US20110299774A1 (en) Method and system for detecting and tracking hands in an image
CN101673161B (en) A visual entity operable without a touch screen system
CN1304931C (en) Head carried stereo vision hand gesture identifying device
JP4597391B2 (en) Facial region detection apparatus and method, and computer-readable recording medium
CN101561710B (en) Man-machine interaction method based on estimation of human face posture
WO2007076890A1 (en) Segmentation of video sequences
CN102324025A (en) Human face detection and tracking method based on Gaussian skin color model and feature analysis
Zhou et al. A novel finger and hand pose estimation technique for real-time hand gesture recognition

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance
C14 Granted