会员体验
专利管家(专利管理)
工作空间(专利管理)
风险监控(情报监控)
数据分析(专利分析)
侵权分析(诉讼无效)
联系我们
交流群
官方交流:
QQ群: 891211   
微信请扫码    >>>
现在联系顾问~
热词
    • 1. 发明公开
    • 실시간 눈 검출 장치 및 그 방법
    • 用于实时眼睛检测的装置和方法
    • KR1020120113317A
    • 2012-10-15
    • KR1020110030955
    • 2011-04-05
    • 성균관대학교산학협력단
    • 전재욱정준희진승훈김동균뉴엔덩투이김대진김문상
    • G06K9/20
    • G06K9/00617G06K9/3241G06T3/40
    • PURPOSE: A real time eye detecting device and a method thereof are provided to detect an eye area of downscaling images for an eye candidate area in real time by generating the images in parallel. CONSTITUTION: An eye candidate area obtaining unit(110) obtains an eye candidate area from an image including a face area. An eye candidate area resizing unit(120) generates downscaling images for the eye candidate area. An MCT(Modified Census Transform) unit(130) performs MCT computation for the image of the eye candidate area and the images generated by the eye candidate area resizing unit. A learning data comparing unit(140) generates a window of the images and matches reliability values about pixels composing the window. [Reference numerals] (100) Real time eye detecting device; (110) Eye candidate area obtaining unit; (120) Eye candidate area resizing unit; (130) MCT unit; (140) Learning data comparing unit; (150) Eye detecting unit; (160) Memory unit
    • 目的:提供实时眼睛检测装置及其方法,通过并行生成图像来实时检测眼睛候选区域的缩小图像的眼睛区域。 构成:眼睛候补区域获取单元(110)从包括脸部区域的图像中获取眼睛候选区域。 眼睛候选区域调整单元(120)生成眼睛候选区域的缩小图像。 MCT(修改普查变换)单元(130)对眼睛候选区域的图像和眼睛候选区域调整单元生成的图像执行MCT计算。 学习数据比较单元(140)生成图像的窗口并且匹配关于构成窗口的像素的可靠性值。 (附图标记)(100)实时眼检测装置; (110)眼睛候选区域获取单元; (120)眼睛候补面积调整单位; (130)MCT单元; (140)学习数据比较单元; (150)眼睛检测单元; (160)存储单元
    • 4. 发明公开
    • 실시간 얼굴 검출 장치
    • 用于实时识别的设备
    • KR1020100090829A
    • 2010-08-18
    • KR1020090010019
    • 2009-02-09
    • 성균관대학교산학협력단
    • 전재욱진승훈김동균뉴엔덩투이김대진김문상
    • H04N5/262
    • G06K9/00281G06K2009/4666
    • PURPOSE: A device for recognizing a face at real time is provided to access to a plurality pixels at the same time and performs a parallel processing, thereby the device is used generally in application fields using voice processing by higher performance. CONSTITUTION: A down-scaling unit(110) performs down-scaling of inputted image at least one rate. A facial image comparator(130) generates windows per one or more down-scaled images. The facial image comparator obtains facial area reliability of the window image through comparison between an image of the generated window and reference face image. The facial area comparator determines whether the window image is relevant to a facial area.
    • 目的:提供用于实时识别脸部的装置,以同时访问多个像素并执行并行处理,从而通常在使用更高性能的语音处理的应用领域中使用该设备。 构成:缩小单元(110)执行至少一个速率的输入图像的缩小。 面部图像比较器(130)每一个或多个缩小图像生成窗口。 面部图像比较器通过比较生成的窗口的图像和参考面部图像来获得窗口图像的面部区域可靠性。 面部区域比较器确定窗口图像是否与面部区域相关。
    • 5. 发明授权
    • 실시간 다중 물체 추적 방법 및 장치
    • 用于实时多目标跟踪的方法和装置
    • KR100854005B1
    • 2008-08-26
    • KR1020070046151
    • 2007-05-11
    • 성균관대학교산학협력단
    • 조정욱진승훈전재욱변종은강훈김문상
    • H04N5/262H04N5/225
    • H04N5/2621G06K2017/0045G06T7/246
    • A method and an apparatus for tracking multiple objects in real-time are provided to set priority to each particle filter, receive the subtraction image of the same priority to set a tracking target object in each particle filter, and receive a subtraction image, in which an area corresponding to the tracking target object which an upper particle filter is tracking is removed, and set a tracking target object, thereby reducing the increase of calculation caused by the extension of a tracking range as the number of tracking target objects increases and reducing the deterioration of tracking performance. An apparatus(100) for tracking multiple objects includes S particle filters(170a~170s), wherein S is an arbitrary natural number, a camera controller(120), an image storage memory(130), a priority distributor(150), and an image subtraction unit based on priority(140). The camera controller receives an original image from a camera(110) and outputs image information corresponding to the received original image. The image storage memory stores the image information. The priority distributor sets priority by using an object tracking result of particle filters. The image subtraction unit based on priority generates k priority subtracted image by using the image information of continuous frames read from the image storage memory and the area information of objects tracked by the (k-1) particle filters, wherein k is an arbitrary natural number in a range of 1 to S. The particle filters perform object tracking by using the priority subtracted images to output location information.
    • 提供用于实时跟踪多个对象的方法和装置,以设置每个粒子滤波器的优先级,接收相同优先级的减影图像以在每个粒子滤波器中设置跟踪目标对象,并接收减法图像,其中 去除与上位粒子滤波器跟踪的跟踪目标对象相对应的区域,并且设置跟踪目标对象,由此随着跟踪目标对象的数量增加而减少由跟踪范围的扩展引起的计算的增加,并且减少 跟踪性能恶化。 用于跟踪多个物体的装置(100)包括S粒子滤波器(170a〜170s),其中S是任意自然数,相机控制器(120),图像存储存储器(130),优先分配器(150)和 基于优先级的图像减法单元(140)。 相机控制器从相机(110)接收原始图像,并输出与接收的原始图像相对应的图像信息。 图像存储存储器存储图像信息。 优先级分配器通过使用粒子滤波器的对象跟踪结果设置优先级。 基于优先级的图像减除单元通过使用从图像存储器读取的连续帧的图像信息和由(k-1)个粒子滤波器跟踪的对象的区域信息来生成k个优先相减图像,其中k是任意自然数 在1到S的范围内。粒子滤波器通过使用优先级减去的图像来执行对象跟踪以输出位置信息。
    • 6. 发明公开
    • 실시간 확장 가능한 스테레오 매칭 시스템 및 방법
    • 实时立体机的可扩展系统与方法
    • KR1020080001093A
    • 2008-01-03
    • KR1020060059208
    • 2006-06-29
    • 성균관대학교산학협력단
    • 전재욱진승훈조정욱박성기김문상이경무변종은
    • H04N13/00
    • G06K9/32G06T7/593G06T2207/10021
    • A real time extensible stereo matching system and a method thereof are provided to apply census transform to two images obtained by cameras through hardware and perform stereo matching, thereby extracting depth information in real time. A coordinate generating module(110) calculates two dimensional coordinates of a currently-inputted pixel on the basis of a control signal generated in a camera receiving an image. A census transform module(120) stores the pixel, obtained in the camera, in a window having an extensible size and performs census transform of each pixel of two images to generate a census vector. A delay XOR operating module(130) delays the output of the census transform module as much as a disparity as a search object to calculate correlation between two census vectors. A stereo matching module(140) stores the output of the delay XOR operating module in a window of an extensible size, and outputs a relative distance value between an object and a camera by using the stored output. A control module(150) performs synchronization as controlling the coordinate generating module, the census transform module, the delay XOR operating module and the stereo matching module.
    • 提供了一种实时可扩展立体匹配系统及其方法,将普查变换应用于摄像机通过硬件获得的两幅图像,并进行立体匹配,从而实时提取深度信息。 坐标生成模块(110)基于在接收图像的相机中生成的控制信号来计算当前输入的像素的二维坐标。 普查变换模块(120)将在相机中获得的像素存储在具有可扩展大小的窗口中,并对两个图像的每个像素进行普查变换,以生成普查向量。 延迟XOR操作模块(130)将普查变换模块的输出作为搜索对象的差异延迟,以计算两个普查向量之间的相关性。 立体匹配模块(140)将延迟XOR操作模块的输出存储在可扩展尺寸的窗口中,并通过使用所存储的输出来输出对象和相机之间的相对距离值。 控制模块(150)执行同步,作为控制坐标生成模块,普查变换模块,延迟XOR操作模块和立体匹配模块。
    • 9. 发明授权
    • 실시간 확장 가능한 스테레오 매칭 시스템 및 방법
    • 可扩展系统? 实时立体声加工方法
    • KR100813100B1
    • 2008-03-17
    • KR1020060059208
    • 2006-06-29
    • 성균관대학교산학협력단
    • 전재욱진승훈조정욱박성기김문상이경무변종은
    • H04N13/00
    • G06K9/32G06T7/593G06T2207/10021
    • 본 발명은 실시간 확장 가능한 스테레오 매칭 시스템으로서, 상세하게는 두 개의 카메라를 통해서 얻은 영상을 이용하여 유사한 점을 찾아 그 결과로 깊이(Depth) 영상을 계산하는 스테레오 매칭(Stereo Matching)을 실시간으로 수행하는 실시간 확장 가능한 스테레오 매칭 시스템에 관한 것이다.
      본 발명의 실시예에 따른 실시간 확장 가능한 스테레오 매칭 시스템은 도 3에 나타낸 바와 같이, 좌표 생성 모듈부(110), 센서스 변환 모듈부(120), 지연 XOR 연산 모듈부(130), 스테레오 매칭 모듈부(140), 제어 모듈부(150)로 구성된다.
      본 발명에 따른 실시간 스테레오 매칭 시스템을 이용하면, 별도의 컴퓨터 시스템이나 소프트웨어 프로그램을 사용하지 않고 실시간으로 교정된 영상의 깊이 정보를 획득하는 것이 가능하며 또한, 발명된 장치는 하드웨어로 간단하게 구현될 수 있어 로봇 비전(Robot Vision)이나 지능형 카메라(Intelligent Camera), 산업현장 등에 쉽게 응용될 수 있다.
      실시간, 스테레오 매칭, 센서스
    • 公开了一种可扩展的系统和方法,用于实时地执行用于计算深度图像的立体声捕获,其结果是通过使用用两个相机拍摄的图像来搜索相似点。 该系统包括坐标创建模块,普查变换模块,延迟XOR计算模块,立体匹配模块和控制模块。 因此,通过使用可扩展的用于执行立体匹配的系统,可以实时获取校正图像的深度信息,而不用于特殊目的的计算机系统或软件程序。 此外,由于用于执行立体匹配的可扩展的系统可以通过硬件简单地实现,所以本发明的系统和方法可以容易地应用于实际的智能机器人,工业设置等。
    • 10. 发明公开
    • 실시간 스테레오 영상 교정 시스템 및 방법
    • 实时立体图像的整形系统及方法
    • KR1020080001095A
    • 2008-01-03
    • KR1020060059210
    • 2006-06-29
    • 성균관대학교산학협력단
    • 전재욱진승훈조정욱박성기김문상이경무변종은
    • G06T1/00G06T1/20G06T7/00G06T7/60
    • G06K9/209G06K9/03G06T3/0075G06T3/4007G06T2200/28G06T2207/10021H04N13/239H04N13/246
    • A system and a method for real time stereo image rectification are provided to deal with the variation of a camera medium value flexibly through hardware, and rectify images obtained in cameras in real time. A coordinate generating module(110) calculates 2 dimensional coordinates of a currently-inputted pixel based on a control signal generated in cameras receives images. A rectification coordinate generating module(120) generates rectification coordinates based on a calibration matrix inputted from the outside. A dual linear interpolation value generating module(130) generates a corresponding code after calculating a double linear interpolation value for double linear interpolation of a rectification result. A first memory(140) stores the rectification coordinates generated in the rectification coordinate generating module. A second memory(150) stores a double linear interpolation value generated in the dual linear interpolation value generating module. A third memory(160) stores an input image and an output image. A rectification module(170) performs image rectification by applying the stored rectification coordinates and the double linear interpolation value based on the output image, and outputs the rectified coordinate value. A controller(180) performs synchronization as controlling the above components.
    • 提供了一种用于实时立体图像整流的系统和方法,以通过硬件灵活地处理相机介质值的变化,并且实时校正在相机中获得的图像。 坐标生成模块(110)基于在相机中生成的控制信号来计算当前输入的像素的二维坐标,接收图像。 整流坐标生成模块(120)根据从外部输入的校准矩阵生成校正坐标。 双线性内插值生成模块(130)在对整流结果的双线性插值计算出双线性内插值之后,生成对应的代码。 第一存储器(140)存储在整流坐标生成模块中生成的整流坐标。 第二存储器(150)存储在双线性内插值生成模块中生成的双线性内插值。 第三存储器(160)存储输入图像和输出图像。 整流模块(170)通过应用存储的整流坐标和基于输出图像的双线性内插值来执行图像整流,并输出整流的坐标值。 控制器(180)执行同步控制上述部件。