-
公开(公告)号：KR1020070061289A
公开(公告)日：2007-06-13
申请号：KR1020060084815
申请日：2006-09-04
Applicant: 한국전자통신연구원
CPC classification number: G06K9/00234 , G06K9/00288 , G06K9/00516 , G06K9/6215
Abstract: A method and a system for recognizing a face with a linear projection-based ICA having class-specific information are provided to improve sensitiveness and low recognition performance for big changes such as lighting and facial expression by using the linear projection-based ICA, which has class discrimination ability and is independent of statistics. Compressed sub images are obtained by performing 2D(Dimensional) wavelet decomposition to a raw face image and using a low frequency band filter(S108). Feature vector and eigenvector, which are components of the face image, are found by applying a PCA(Principal Component Analysis) to the sub images(S114). Two systematic structures using a basis image expression method and a factorial code expression method are constructed by applying the ICA to the components(S124). Face recognition is performed by performing FLD(Fisher Linear Discriminant) having class discrimination ability and the class-specific information to two structures(S130).
Abstract translation: 提供了一种利用具有类特定信息的基于线性投影的ICA识别脸部的方法和系统，以通过使用基于线性投影的ICA来改善对诸如照明和面部表情的大变化的敏感性和低识别性能， 阶级歧视能力，独立于统计。 通过对原始脸部图像进行2D（Dimensional）小波分解并使用低频带滤波器来获得压缩子图像（S108）。 通过对子图像应用PCA（主成分分析），找出作为面部图像的成分的特征向量和特征向量（S114）。 通过将ICA应用到组件来构建使用基础图像表达方法和阶乘代码表达方法的两个系统结构（S124）。 通过对两个结构执行具有类别识别能力的FLD（Fisher线性判别）和类别特定信息来执行人脸识别（S130）。
-
公开(公告)号：KR1020070061056A
公开(公告)日：2007-06-13
申请号：KR1020060039741
申请日：2006-05-03
Applicant: 한국전자통신연구원
Abstract: A sound source tracing system under robot environment and a method thereof are provided to control the motion of a robot toward the generation direction of a sound source by continuously tracing the generation direction of a present sound source in consideration of the sound source tracing time. A sound source tracing system under robot environment is composed of a sound source input unit(100) obtaining and transmitting a sound source; a noise removing unit(200) decreasing noise of the input sound source by using a Kalman filter(210); a sound source detector unit(300) detecting the starting and finishing points of the sound source with decreased noise; a sound source direction recognizing unit(400) checking the direction of the input sound source by extracting energy of the sound source and time difference from the input sound source with decreased noise; an azimuth detecting unit(500) calculating an azimuth angle of the recognized direction; and a robot control unit(600) controlling a motion of a robot to make the robot face the generation direction of the sound source.
Abstract translation: 提供了一种在机器人环境下的声源跟踪系统及其方法，其通过考虑到声源跟踪时间连续跟踪当前声源的发生方向，来控制机器人朝向声源的发生方向的运动。 机器人环境下的声源跟踪系统由声源输入单元（100）组成，获取和发送声源; 通过使用卡尔曼滤波器（210）降低输入声源的噪声的噪声去除单元（200）; 声源检测器单元，用于以噪声降低检测声源的起始点和终点; 声源方向识别单元，通过提取声源的能量和噪声降低的输入声源的时间差来检查输入声源的方向; 方位检测单元，计算识别方向的方位角; 以及机器人控制单元（600），其控制机器人的运动以使机器人面对声源的生成方向。
-
公开(公告)号：KR1020060065197A
公开(公告)日：2006-06-14
申请号：KR1020040103979
申请日：2004-12-10
Applicant: 한국전자통신연구원
IPC: G06K9/78
CPC classification number: G06K9/00463 , G06K9/42 , G06K2009/4666 , G06T5/30 , G06T7/13
Abstract: 1. 청구범위에 기재된 발명이 속한 기술분야
본 발명은 휴대형 카메라로 획득한 문자영상의 문자영역선택을 통한 문자인식 장치 및 그 방법에 관한 것임.
2. 발명이 해결하려고 하는 기술적 과제
본 발명은 휴대형 카메라로 획득한 문자영상에서 문자영역을 선택하여 국소 이진화한 후에 문자를 인식함으로써, 문자 추출 및 인식 성능을 향상시킬 수 있는 문자인식 장치 및 그 방법을 제공하는데 그 목적이 있음.
3. 발명의 해결방법의 요지
본 발명은, 휴대형 카메라로 획득한 문자영상의 문자영역선택을 통한 문자인식 장치에 있어서, 휴대형 카메라를 이용하여 획득한 문자영상을 입력받기 위한 문자영상 입력 수단; 상기 문자영상 입력 수단을 통하여 입력받은 카메라 문자영상을 처리하여 문자영역을 추출하기 위한 문자영역 추출 수단; 상기 문자영역 추출 수단에서 추출한 문자영역에 대해 이진화를 수행하기 위한 이진화 수단; 상기 이진화 수단에서 이진화한 문자영역에서 문자를 추출하기 위한 문자 추출 수단; 상기 문자 추출 수단에서 추출한 문자의 특징을 추출하기 위한 특징 추출 수단; 및 상기 특징 추출 수단에서 추출한 문자의 특징을 이용하여 문자를 인식하기 위한 문자 인식 수단을 포함함.
4. 발명의 중요한 용도
본 발명은 영상처리 기술분야 중 패턴인식분야의 문자인식 기술분야 등에 이용됨.
휴대형 카메라, 문자영상 획득, 문자영역 선택, 국소 이진화, 문자인식-
公开(公告)号：KR100513789B1
公开(公告)日：2005-09-09
申请号：KR1020020080336
申请日：2002-12-16
Applicant: 한국전자통신연구원
IPC: H04N5/225
Abstract: 본 발명은 유무선 통신 하부구조 혹은 인터넷을 이용하여 사용자에게 언제 어디서나 영상 정보를 추출하고 가공, 전송, 저장 및 이를 이용한 정보 서비스의 제공을 위해 디지털 카메라를 포함한 각종 영상 장비를 통해 획득된 가공되지 않은 영상으로부터 렌즈 왜곡 보정된 영상을 생성하고, 촬영 당시의 장비의 자세 정보를 실시간으로 획득하여 정사영상을 생성하는 기술에 관한 것이다. 이를 위해 본 발명은 영상 장비에 사용되는 렌즈의 왜곡 성분을 카메라 캘리브레이션을 통해 미리 결정하고, 이를 장비의 내부에 저장하도록 함으로써 사용자는 장비를 사용시에 특별한 캘리브레이션 과정을 수행할 필요가 없을 뿐만 아니라 본 발명에서 고안된 이미지 워핑 방법을 이용하여 실시간으로 왜곡 보정된 고품질의 영상을 획득할 수 있다. 특히, 각종 정보 서비스 및 영상에 기반한 정보 추출의 경우 촬영된 패턴이 영상 장비의 촬영 당시의 자세에 의해 기하학적 성질이 변하는 것을 막고, 최종적으로 신뢰성 높은 정보 추출이 이루어지도록 하기 위해 정사영상을 생성하도록 한다.
-
公开(公告)号：KR1020040053877A
公开(公告)日：2004-06-25
申请号：KR1020020080336
申请日：2002-12-16
Applicant: 한국전자통신연구원
IPC: H04N5/225
CPC classification number: H04N5/23216 , H04N5/23229 , H04N5/238
Abstract: PURPOSE: A method for correcting lens distortion and producing orthoimage and a digital camera by using the same are provided to produce the orthoimage from the planar pattern photographed by the digital camera taking an arbitrary rotational angle after processing the lens distortion correction by using automatically computed camera's relative rotational angle information to a plane. CONSTITUTION: A camera parameter previously obtained through a camera calibration(901) is stored to the inside of a digital camera(910). An image(911) obtained by a user(902) is passed through a lens distortion correction module(912). An image in which a radial lens distortion is corrected is transmitted to an orthoimage generation module(913). An image which is composed again in the orthoimage generation module(913) is inputted to a pattern recognition module(914). Information extraction and pattern recognition processes are performed by the pattern recognition module(914) and then transmitted to a display(903) and a user(902) through a communication module(915).
Abstract translation: 目的：提供一种用于通过使用它来校正透镜畸变并产生正射影像和数字照相机的方法，以便通过使用自动计算的相机的处理透镜畸变校正后的数字照相机拍摄的任意旋转角拍摄的平面图案产生正射影像 相对旋转角度信息到平面。 构成：先前通过摄像机校准获得的摄像机参数（901）被存储到数字照相机（910）的内部。 由用户（902）获得的图像（911）通过透镜畸变校正模块（912）。 校正径向透镜畸变的图像被发送到正射图像生成模块（913）。 在正射图像生成模块（913）中再次组合的图像被输入到图案识别模块（914）。 信息提取和模式识别处理由模式识别模块（914）执行，然后通过通信模块（915）发送到显示器（903）和用户（902）。
-
Patent Agency Ranking
公开(公告)号：KR1020040022348A
公开(公告)日：2004-03-12
申请号：KR1020020053623
申请日：2002-09-05
Applicant: 한국전자통신연구원
IPC: H04N5/225
CPC classification number: H04N5/225 , G03B21/142 , H04N5/205
Abstract: PURPOSE: An image-based lens distortion correcting apparatus and a method therefor are provided to reduce the quantity of calculations in an image analysis procedure. CONSTITUTION: An image is received(401), one-dimensional image pixels are extracted from the image(402), and a coordinate value of each of the pixels is recorded(403). One-dimensional pixel sampling is performed for the image(404), straight lines in a radial direction are obtained on the basis of the center of the image. Pixels intersect the straight lines are stored in an one-dimensional image arrangement(405). A point at which a range of a lens distortion parameter of the image becomes the maximum is found, and the range is divided into intervals to increase the lens distortion parameter. A luminance value of pixel coordinate values after a distortion of the image is corrected by increasing the lens distortion parameter is obtained through inverse spatial mapping. A sampling angle for the entire image is increased. The one-dimensional image arrangement is split into records(407), and the lens distortion parameter having the smallest representative value is set as a lens distortion parameter(409). A distortion-corrected resultant image is obtained from the original image using the lens distortion parameter and provided to a thumbnail image display(410).
Abstract translation: 目的：提供一种基于图像的镜头畸变校正装置及其方法，以减少图像分析程序中的计算量。 构成：接收图像（401），从图像（402）提取一维图像像素，并记录每个像素的坐标值（403）。 对图像（404）执行一维像素采样，基于图像的中心获得径向的直线。 与直线相交的像素被存储在一维图像排列（405）中。 找到图像的透镜失真参数的范围变为最大的点，并且将该范围划分为间隔以增加透镜失真参数。 通过增加透镜失真参数来校正图像失真后的像素坐标值的亮度值，通过逆空间映射获得。 整个图像的采样角度增加。 将一维图像排列分割为记录（407），将具有最小代表值的透镜失真参数设置为透镜失真参数（409）。 使用透镜失真参数从原始图像获得失真校正的合成图像，并提供给缩略图图像显示（410）。
-
公开(公告)号：KR1020010075874A
公开(公告)日：2001-08-11
申请号：KR1020000002772
申请日：2000-01-21
Applicant: 한국전자통신연구원
IPC: G06F17/30
CPC classification number: G06K9/00469
Abstract: PURPOSE: A method for analyzing an image structure of a treatise is provided to detect the locations of title, writer and abstract parts from a thesis document image and to automatically execute contents recognition. CONSTITUTION: If candidate regions are inputted(201), the number of candidate regions is checked(202) and the number of letters for each candidate region is calculated(203). Among the letter number of each candidate region and the letter number of 40 appointed specific words(204), each letter is compared as much as the letter number of the side having a smaller number(206-210). And how many letters in the same location are accorded each other is checked. In case that the proportion obtained by taking the letter number of the side having a larger number as a denominator and the letter number of the smaller side as a numerator is larger than the judgement probability of a specific word, the specific word candidate region is concluded as the specific word region.
Abstract translation: 目的：提供一种用于分析论文的图像结构的方法，以从论文文档图像中检测标题，作者和抽象部分的位置，并自动执行内容识别。 构成：如果输入候选区域（201），则检查候选区域的数目（202），并计算每个候选区域的字母数（203）。 在每个候选区域的字母数字和40个指定的特定单词（204）的字母数字中，每个字母与具有较小数字的一侧的字母数字（206-210）进行比较。 并检查相同位置中的多少个字母。 在将具有较大数字的字母数字作为分母而获得的比例和较小侧的字母数字作为分子获得的比例大于特定词语的判断概率的情况下，结束特定词候选区域 作为具体单词区域。
-
公开(公告)号：KR1020000034645A
公开(公告)日：2000-06-26
申请号：KR1019980052019
申请日：1998-11-30
Applicant: 한국전자통신연구원
IPC: G06K9/46
Abstract: PURPOSE: A feature extraction method for recognition of large scale print character is provided to extract the feature of characters by combination of the accumulated frequency to position shift of mesh values and the accumulated value of hamming distance about all characters such as korean letter, Chinese letter, alphabet letter, number and special symbol. CONSTITUTION: A feature extraction method for recognition of large scale print character comprises steps of: inputting the image of large scale print character using an optical scanner; regulating the size of the input character image into a pattern of constant size; extracting the feature of the regulated characters during a feature extraction process; and reclassifying the extracted characters by their feature.
Abstract translation: 目的：提供识别大尺度打印字符的特征提取方法，通过将累积频率与网格值的位置偏移和汉明距离的累积值相结合，提取汉字的汉字，汉字 ，字母，数字和特殊符号。 构成：用于识别大规模打印字符的特征提取方法包括以下步骤：使用光学扫描器输入大尺寸打印字符的图像; 将输入字符图像的大小调整为恒定大小的图案; 在特征提取过程中提取调节字符的特征; 并通过其特征重新分类提取的字符。
-
-
-
-
-
-
-
-
Search Results
106
records
(took 0.023 seconds)
