-
公开(公告)号：KR1020150015733A
公开(公告)日：2015-02-11
申请号：KR1020130091476
申请日：2013-08-01
Applicant: 한국전자통신연구원
CPC classification number: G06K9/00348 , G06K9/3241 , G06T7/66 , G06T2207/30241
Abstract: 본 발명에 의한 걸음걸이 정보를 이용하여 사람을 검출하기 위한 장치 및 그 방법이 개시된다. 본 발명에 따른 걸음걸이 정보를 이용하여 사람을 검출하기 위한 장치는 영상 프레임을 입력 받으면, 입력되는 영상 프레임마다 사람의 영역인 ROI(Region Of Interest)를 검출하는 검출부; 검출된 상기 사람의 영역인 ROI 각각에서 특징점을 추출하는 추출부; 추출된 상기 특징점을 이용하여 영상 프레임들 간 사람의 중심점 궤적을 산출하는 산출부; 및 산출된 상기 사람의 중심점 궤적을 기 설정된 삼각 함수에 매칭시켜 그 매칭시킨 결과에 따라 영상 프레임들 간 사람의 중심점 궤적을 보정하는 보정부를 포함하는 것을 특징으로 한다.
Abstract translation: 本发明公开了一种使用步态信息检测人的装置和方法。 根据本发明的用于检测使用步态信息的人的装置包括：检测单元，如果输入了图像帧，则检测作为每个输入图像帧中的人类区域的感兴趣区域（ROI）; 提取单元，其从检测到的人类区域的每个ROI中提取特征点; 计算单元，其通过使用所提取的特征点来计算人物在图像帧之间的中心轨迹; 以及校正单元，其与计算出的具有预设三角函数的人的中心轨迹相匹配，并且根据匹配结果校正人物的中心轨迹。
-
公开(公告)号：KR1020140076814A
公开(公告)日：2014-06-23
申请号：KR1020120145281
申请日：2012-12-13
Applicant: 한국전자통신연구원
CPC classification number: G06T7/73 , G06T2207/10016 , G06T2207/20024 , G06T2207/30232 , G06T2210/56
Abstract: The present invention relates to a method and an apparatus for automatically tracking an object in a CCTV-based image security apparatus or system. The object tracking method through a particle diffusion according to the present invention includes the following steps of: receiving coordinates about an object in the previous frame of a corresponding frame; diffusing particles by using a first diffusion model for an irregular diffusion of multiple particles for tracking the inputted coordinates and the object and a second diffusion model for a regular diffusion; and tracking the object in the corresponding frame by determining a similarity between a specific value of the diffused particles and a predetermined reference value. According to the present invention, the tracking accuracy of an object can be improved by increasing a probability of positioning particles at an optimal position by adding a deterministic element when diffusing the particles in a particle filter-based object tracking.
Abstract translation: 本发明涉及一种在基于CCTV的图像安全装置或系统中自动跟踪对象的方法和装置。 根据本发明的通过粒子扩散的物体跟踪方法包括以下步骤：接收关于相应帧的先前帧中的对象的坐标; 通过使用用于跟踪所输入的坐标和物体的多个粒子的不规则扩散的第一扩散模型和用于规则扩散的第二扩散模型来扩散粒子; 以及通过确定扩散粒子的特定值和预定参考值之间的相似度来跟踪相应帧中的对象。 根据本发明，通过在将粒子扩散到基于粒子滤波器的物体跟踪中时，通过增加确定性元素来增加将粒子定位在最佳位置的概率，可以提高对象的追踪精度。
-
公开(公告)号：KR1020140035176A
公开(公告)日：2014-03-21
申请号：KR1020120101718
申请日：2012-09-13
Applicant: 한국전자통신연구원
IPC: G06T7/20
CPC classification number: G06T7/20 , H04N19/126
Abstract: The present invention relates to an apparatus and a method for object tracking using an adaptive multiple feature weight decision. One multiple feature which is a weighted sum of feature items of each object is extracted, and accuracy for tracking the multiple objects is improved by estimating a state of the object by determining a weighted sum coefficient regarding each feature item to have maximum differentiation characteristics. [Reference numerals] (110) Multiple features extracting unit; (120) Weight decision unit; (131) Observed value calculator; (132) Particle re-extracting unit; (133) State estimating value calculator; (140) State transition equation calculator; (150) Object detecting unit; (160) Initialization unit
Abstract translation: 本发明涉及使用自适应多重特征加权决策的对象跟踪的装置和方法。 提取作为每个对象的特征项的加权和的一个多重特征，通过对关于每个特征项的加权和系数进行估计以具有最大差异特性来估计对象的状态来提高跟踪多个对象的精度。 （附图标记）（110）多要素提取单元; （120）权重决策单位; （131）观测值计算器; （132）颗粒再提取单元; （133）状态估计值计算器; （140）状态转换方程计算器; （150）对象检测单元; （160）初始化单元
-
公开(公告)号：KR1020130067852A
公开(公告)日：2013-06-25
申请号：KR1020110134843
申请日：2011-12-14
Applicant: 한국전자통신연구원
IPC: G06K9/46
CPC classification number: G07C9/00158 , G06K9/00744 , G06K9/6292
Abstract: PURPOSE: A remote personal authentication system using biometric recognition and a method thereof are provided to perform privacy-enhanced personal authentication by performing personal authentication through the comparison of biometric information received from a security token device. CONSTITUTION: A biometric information sensing unit(310) collects biometric information of a user and image information about an authentication space including the user. A communication unit(320) receives the biometric information and identification information of a security token device received from the security token device. A control unit(340) matches positions of the security token device and the user with each other by using the identification information of the security token device. [Reference numerals] (310) Biometric information sensing unit; (320) Communication unit; (330) Storage unit; (340) Control unit; (350) Power supply unit; (360) Output unit; (370) Security processing unit
Abstract translation: 目的：提供使用生物识别的远程个人认证系统及其方法，以通过通过比较从安全令牌设备接收的生物特征信息进行个人认证来执行隐私增强的个人认证。 构成：生物体信息感知单元（310）收集用户的生物体信息和关于包括用户的认证空间的图像信息。 通信单元（320）接收从安全令牌装置接收到的安全令牌装置的生物体信息和识别信息。 控制单元（340）通过使用安全令牌设备的识别信息来匹配安全令牌设备和用户的位置。 （附图标记）（310）生物信息感知单元; （320）通讯单元; （330）存储单元; （340）控制单元; （350）电源单元; （360）输出单元; （370）安全处理单元
-
公开(公告)号：KR101274324B1
公开(公告)日：2013-06-13
申请号：KR1020100117134
申请日：2010-11-23
Applicant: 한국전자통신연구원
IPC: G06T7/40
Abstract: 본 발명은 시선 트래킹을 위한 눈 검출 장치 및 그 방법에 관한 것으로서, 상세하게는 눈의 단순한 이미지 특성만을 고려한 눈 필터를 이용하는 것이 아니라, 복수의 학습 얼굴 이미지 내의 눈 이미지를 이용하여 다양한 눈의 이미지가 고려된 평균 눈 필터를 생성하고, 입력 얼굴 이미지에 그 생성된 평균 눈 필터를 적용하여 눈을 검출함으로써, 일반적인 눈 필터와 패턴 매칭을 통한 눈 검출에 비해 눈 검출의 성능 및 속도를 향상시킬 수 있다.
-
Patent Agency Ranking
公开(公告)号：KR1020120042101A
公开(公告)日：2012-05-03
申请号：KR1020100103600
申请日：2010-10-22
Applicant: 한국전자통신연구원
CPC classification number: G06K9/00275 , G06T7/11 , G06T7/13
Abstract: PURPOSE: A facial feature extraction apparatus and a method thereof are provided to increase the reliability and stability of face recognition by extracting facial features based on edge information. CONSTITUTION: An extracting unit(100) extracts edge information of a face. A describing unit(200) divides a face image into different areas and describes edge features of the divided areas. An extracting unit(300) extracts facial features of the face through the edge features. The extracting unit extracts edge information of a face image through a photographing unit by using an image filter.
Abstract translation: 目的：提供面部特征提取装置及其方法，通过基于边缘信息提取面部特征来提高面部识别的可靠性和稳定性。 构成：提取单元（100）提取脸部的边缘信息。 描述单元（200）将脸部图像划分成不同的区域并描述分割区域的边缘特征。 提取单元（300）通过边缘特征提取面部的面部特征。 提取单元通过使用图像滤波器通过拍摄单元提取面部图像的边缘信息。
-
公开(公告)号：KR1020150054446A
公开(公告)日：2015-05-20
申请号：KR1020130136891
申请日：2013-11-12
Applicant: 한국전자통신연구원
IPC: G06K9/00
CPC classification number: H04L9/3231 , G06K9/6202
Abstract: 실시예는, 임의의사용자로부터바이오인식정보및 식별자 ID 정보를입력받는단계, 데이터베이스에등록된복수의바이오인식정보각각과상기바이오인식정보를정합하여, 상기복수의바이오인식정보각각에대한유사도를산출하는단계, 상기복수의바이오인식정보를유사도내림순으로정렬된검색순위를기반으로, 상기복수의바이오인식정보중 상기식별자 ID 정보에대응하는입력바이오인식정보가상기검색순위중 미리설정된순위범위에속하는지판단하는단계및 상기입력바이오인식정보가상기순위범위에속하면, 상기사용자에대한개인인증을수행하는단계를포함하는바이오인식시스템인증방법을제공한다.
Abstract translation: 提供一种用于生物识别系统的认证方法，用于搜索先前登记在数据库上的多个生物特征信息，并且通过检查与标识符ID信息相对应的输入的生物特征信息是否属于所搜索的排名来执行个体认证。 生物识别系统的认证方法包括以下步骤：从随机用户接收生物体信息和标识符ID信息的输入; 将登记在数据库上的多个生物体信息与生物体信息进行匹配，并计算多个生物体信息中的每一个的相似度; 从多个生物体信息中的与标识符ID信息相对应的输入的生物体信息是否属于先前设定的等级范围，其中，搜索排名基于多个生物信息按降序对齐的搜索排名 的相似性 并且如果输入的生物特征信息属于分级范围，则为用户执行个体认证。
-
公开(公告)号：KR101472452B1
公开(公告)日：2014-12-17
申请号：KR1020100114368
申请日：2010-11-17
Applicant: 한국전자통신연구원
CPC classification number: G06N99/005 , G06F17/30023
Abstract: 본발명은멀티미디어검색방법, 장치및 패턴인식방법에관한것으로, 본발명의일면에따른멀티미디어검색방법은, 검색대상데이터중에서사용자로부터입력된검색조건데이터에대응하는데이터를검색하는단계와, 상기검색결과에서기계학습(machine learning)을위한학습데이터를선택하는단계와, 상기선택된학습데이터를이용하여기계학습을수행하는단계및 상기기계학습결과를이용하여, 상기검색결과를수정하는단계를포함한다.
-
公开(公告)号：KR1020140075328A
公开(公告)日：2014-06-19
申请号：KR1020120143562
申请日：2012-12-11
Applicant: 한국전자통신연구원
IPC: G06T7/00
CPC classification number: G06K9/00281 , G06K9/00234
Abstract: Disclosed by the present invention are an apparatus for detecting face components and a method thereof. The apparatus for detecting face components according to the present invention includes: a face area detecting unit which detects a face area from an image if the face area detecting unit receives the image; a candidate area detecting unit which detects a candidate area which includes face components from the detected face area; a candidate area dividing unit which divides the detected candidate area into multiple areas; a component clustering unit which clusters the face components in each of the divided multiple areas by using a preset clustering algorithm; and a component determining unit which calculates the coordinates of the cluster of the face components which are divided as the clustered result and determines the face components based on the calculated coordinates.
Abstract translation: 本发明公开了一种面部成分检测装置及其方法。 根据本发明的用于检测面部组件的装置包括：脸部区域检测单元，如果面部区域检测单元接收到图像，则从面部区域检测图像; 候选区域检测单元，其从检测到的面部区域检测包括面部成分的候选区域; 候选区域划分单元，其将检测到的候选区域划分为多个区域; 组件聚类单元，其通过使用预设聚类算法对分割的多个区域中的每一个中的面部分量进行聚类; 以及分量确定单元，其计算被分割为聚集结果的面部分量的群集的坐标，并且基于所计算的坐标来确定面部分量。
-
公开(公告)号：KR1020140060081A
公开(公告)日：2014-05-19
申请号：KR1020120126738
申请日：2012-11-09
Applicant: 한국전자통신연구원
CPC classification number: G06K9/6202 , G06K9/00221 , G06T7/246 , G06T7/33
Abstract: An apparatus for controlling a speed gate based on remote face recognition is disclosed. An apparatus for controlling a speed gate based on remote face recognition according to the present invention includes: an entrance control part for collecting images, and opening and closing the speed gate; an image part which detects motion by obtaining a difference image from the collected image, and collects a face image of a moving Blob; and a security part for judging whether the face image of the Blob is an authenticated user, by comparing the face image of the Blob with a previously registered face image. The entrance control part supplies the collected image to the image part continuously, and the image part collects a face image continuously, from the image information supplied continuously, and the security part classifies each face image, which is continuously collected, as an authenticated user for a person obtaining the most votes in a candidate group classified as similar users, by comparing each collected face image with the previously registered face image.
Abstract translation: 公开了一种用于基于远程面部识别来控制速度门的装置。 根据本发明的用于控制基于远程面部识别的速度门的装置包括：用于收集图像并打开和关闭速度门的入口控制部分; 图像部分，通过从所收集的图像获得差异图像来检测运动，并且收集移动Blob的面部图像; 以及用于通过将Blob的面部图像与先前登记的面部图像进行比较来判断Blob的脸部图像是否是经认证的用户的安全部件。 入口控制部分将收集的图像连续地提供给图像部分，并且图像部分从连续供应的图像信息中连续地收集面部图像，并且安全部件将连续收集的每个面部图像分类为认证用户， 通过将每个收集的面部图像与先前登记的面部图像进行比较，获得在被分类为类似用户的候选组中的投票最多的人。
-
-
-
-
-
-
-
-
-
Search Results
36
records
(took 0.025 seconds)
