-
公开(公告)号:KR1020130067758A
公开(公告)日:2013-06-25
申请号:KR1020110134652
申请日:2011-12-14
Applicant: 한국전자통신연구원
IPC: G06T7/00
CPC classification number: G06K9/00362 , G06F17/30247 , G06K9/6269 , G06N99/005
Abstract: PURPOSE: Human detection apparatus using SVM(Support Vector Machine) learning and a method thereof are provided to generate a histogram for a binary characteristic pattern, obtain a characteristic vector, and then detect a person from an input image, thereby obtaining a stable pattern through a cumulative sum-based analysis in spite of noise in the vicinity when obtaining a characteristic pattern for each pixel. CONSTITUTION: A learning image characteristic extraction block(112) extracts a binary learning characteristic pattern from an input learning image(102) and generates a learning characteristic vector. An SVM learning block(114) generates a classifier through the SVM learning of the generated learning characteristic vector. An image characteristic extraction block(132) extracts a binary input characteristic pattern from an input image(104) and generates an input characteristic vector. A human detection judgment block(136) detects a person by using a pre-set search region and the classifier stored in a database(120). [Reference numerals] (112) Learning image characteristic extraction block; (114) SVM learning block; (116) SVM determination weighted value generation block; (132) Input image characteristic extraction block; (134) Search area setting block; (136) Human detection judgment block
Abstract translation: 目的:提供使用SVM(支持向量机)学习的人体检测装置及其方法,以生成二进制特征图案的直方图,获得特征向量,然后从输入图像中检测人,从而获得稳定的图案 当获得每个像素的特征图案时,尽管附近有噪声的累积和分析。 构成:学习图像特征提取块(112)从输入学习图像(102)提取二进制学习特征模式,并生成学习特征向量。 SVM学习块(114)通过生成的学习特征向量的SVM学习生成分类器。 图像特征提取块(132)从输入图像(104)提取二进制输入特征图案并生成输入特征向量。 人检测判断块(136)通过使用存储在数据库(120)中的预设搜索区域和分类器来检测人。 (参考数字)(112)学习图像特征提取块; (114)SVM学习块; (116)SVM确定加权值生成块; (132)输入图像特征提取块; (134)搜索区域设置块; (136)人体检测判断块
-
公开(公告)号:KR1020030056568A
公开(公告)日:2003-07-04
申请号:KR1020010086830
申请日:2001-12-28
Applicant: 한국전자통신연구원
IPC: H04L9/32
CPC classification number: H04L63/0869
Abstract: PURPOSE: An authentication method for protecting an agent and a message is provided, which defends the attack from a malicious agent and message level, and assures authentication at an initial access trial process. CONSTITUTION: According to the authentication method, a transmitter agent(1) request authentication to a facilitator(2) in order to transfer a message to an unknown agent. The facilitator performs mutual authentication with the transmitter agent and then exchanges a session key. The facilitator finds an agent coinciding with its object and performs mutual authentication, and then exchanges the session key. And the transmitter agent constitutes an independent security channel with a receiver agent(3) selected using the session key.
Abstract translation: 目的:提供一种用于保护代理和消息的认证方法,从而防止恶意代理和消息级别的攻击,并在初始访问试用过程中确保认证。 构成:根据认证方法,发送方代理(1)向促进者(2)请求认证,以便将消息传递给未知代理。 协调者与发射机代理执行相互认证,然后交换会话密钥。 协调人找到与其对象一致的代理,并进行相互认证,然后交换会话密钥。 并且发射机代理构成具有使用会话密钥选择的接收方代理(3)的独立安全信道。
-
公开(公告)号:KR1020160116585A
公开(公告)日:2016-10-10
申请号:KR1020150044495
申请日:2015-03-30
Applicant: 한국전자통신연구원
Inventor: 최진우
IPC: H04N21/4545 , G06T7/00 , G06T5/00
Abstract: 동영상유해영역차단장치는입력되는동영상의현재프레임이유해한프레임인지판단하고, 유해한프레임에대하여얼굴및 상기얼굴을제외한사람의적어도일부분을검출한후, 얼굴및 상기사람의적어도일부분중 검출된영역을마스킹하며, 현재프레임의이후프레임에대해순차적으로직전프레임에서검출된얼굴및 사람의적어도일부분에대한정보를초기값으로사용하여얼굴및 사람의적어도일부분을추적하여추적된영역을마스킹한다.
-
公开(公告)号:KR1020160104174A
公开(公告)日:2016-09-05
申请号:KR1020150026657
申请日:2015-02-25
Applicant: 한국전자통신연구원
IPC: H04L29/06
CPC classification number: H04L63/308 , H04L63/306
Abstract: 본발명은유해후보인터넷식별정보의사전검출방법및 그장치에관한것으로, 본발명의일 실시예에따른인터넷유해성판단방법은단말로부터인터넷식별정보를수신하는단계; 상기수신한인터넷식별정보에대한인터넷접속통계정보를이용하여상기인터넷식별정보의통계정보기반유해가능성을계산하는단계; 상기수신한인터넷식별정보에대응하는콘텐츠의내용을이용하여상기인터넷식별정보의콘텐츠내용기반유해가능성을계산하는단계; 상기통계정보기반유해가능성및 상기콘텐츠내용기반유해가능성을이용하여상기인터넷식별정보가유해후보인지판단하는단계; 상기인터넷식별정보가유해후보인경우, 상기인터넷식별정보에대한유해여부를판단하는단계; 및상기판단결과에따라서로다른메시지를전송하는단계를포함할수 있다. 본발명의일 실시예에따르면, 유해후보인터넷식별정보의사전검출방법및 그장치를제공할수 있다.
Abstract translation: 本发明涉及一种用于确定互联网有害性的方法,其可以预先检测与有害候选者相对应的因特网识别信息,并且还涉及一种其设备。 根据本发明的实施例的用于确定因特网有害性的方法可以包括以下步骤:从终端接收因特网识别信息; 通过使用关于所接收的因特网识别信息的因特网访问统计信息来计算基于统计的有害可能性的互联网识别信息; 通过使用与所接收的因特网识别信息相对应的内容来计算因特网识别信息的基于内容的有害可能性; 通过使用基于统计的有害可能性和基于内容的有害可能性来确定互联网识别信息是否对应于有害候选者; 当互联网识别信息对应于有害候选者时,确定互联网识别信息是否对应于有害性; 以及根据确定的结果发送不同的消息。 根据本发明的实施例,提供的可以是用于确定因特网有害性的方法,其可以预先检测与有害候选者相对应的因特网识别信息,并且还可以提供其设备。
-
15.
公开(公告)号:KR1020160063234A
公开(公告)日:2016-06-03
申请号:KR1020150141387
申请日:2015-10-08
Applicant: 한국전자통신연구원
IPC: H04N21/454 , H04N21/442 , G06Q50/10 , H04L29/12
Abstract: 스트리밍 URI를이용한컨텐츠의유해성판정방법및 이를이용한장치가개시된다. 유해성판정방법은 URI(Uniform Resource Identifier)를통해컨텐츠의유해성을판정하는유해성판정장치에서수행되는유해성판정방법에있어서, 사용자의단말로부터상기컨텐츠에상응하는상기 URI를수신하는단계; 상기 URI 및상기컨텐츠에상응하는해시값 중적어도하나를 URI 데이터베이스및 해시데이터베이스중 적어도하나에저장된복수개의리스트와비교하여상기컨텐츠의유해성을판정하는단계; 상기 URI와상기해시값이상기복수개의리스트에존재하지않는경우에, 상기컨텐츠의내용을기반으로상기컨텐츠의유해성을판정하는단계; 및상기컨텐츠의유해성판정결과를상기사용자의단말로전송하는단계를포함한다.
Abstract translation: 公开了使用流式URI确定内容的有害性的方法和使用该URI的装置。 使用流统一资源标识符(URI)来确定内容的有害性的方法由使用流式URI确定内容的有害性的装置来执行。 该方法包括以下步骤:从用户的终端接收对应于内容的URI; 将与内容对应的URI和散列值中的至少一个与存储在URI数据库和散列数据库中的至少一个中的多个列表进行比较,以确定内容的有害性; 当URI和哈希值不存在于列表中时,基于内容的内容来确定内容的有害性; 以及将所述内容的有害性确定结果发送给所述用户的终端。
-
公开(公告)号:KR1020150050224A
公开(公告)日:2015-05-08
申请号:KR1020130131756
申请日:2013-10-31
Applicant: 한국전자통신연구원
CPC classification number: G06K9/00335 , G06K9/00362
Abstract: 비정상배회감지장치는카메라로부터영상을수신하는통신부, 영상에서비정상배회하는객체를감지하도록하는명령어를저장하는메모리및 명령어에따라동작하는프로세서를포함하되, 명령어는영상에서객체를추출하고, 객체의경로에상응하는하나이상의대표점을검출하는단계, 대표점간의벡터를산출하는단계및 벡터간 방향차이가미리지정된임계각도이상인회수가미리지정된수 이상인경우, 객체가비정상배회하고있음으로감지하는단계의수행을명령하는명령어를포함한다.
Abstract translation: 用于检测异常游荡的装置包括:通信部分,从相机接收图像; 存储用于检测来自图像的异常游荡对象的命令的存储器; 以及根据命令操作的处理器,其中所述命令执行以下步骤:从所述图像中提取所述对象; 检测对应于物体的路径的一个或多个代表点; 计算代表点之间的向量; 并且如果矢量之间的方向差异等于或大于预设临界角的情况的数量等于或大于预设值,则感测到物体异常游荡。
-
公开(公告)号:KR1020150033960A
公开(公告)日:2015-04-02
申请号:KR1020130113921
申请日:2013-09-25
Applicant: 한국전자통신연구원
IPC: G06T7/20
CPC classification number: G06T7/277 , G06K9/6223 , G06T7/70
Abstract: 다중파티클군집을이용한비강체객체추적방법이개시된다. 본발명에따른다중파티클군집을이용한비강체객체추적방법은객체의초기위치를설정하는단계; 상기객체의움직임벡터를계산하여움직임벡터값이같은범위내에포함된파티클들로파티클군집을생성하는단계; 상기파티클들의평균값을계산하는단계; 및상기평균값을이용하여상기객체의상태추정치를획득하는단계를포함한다.
Abstract translation: 公开了一种使用多粒子群跟踪非刚性物体的方法。 根据本发明的方法包括以下步骤:设置对象的初始位置; 计算对象的运动矢量以产生包括运动矢量落在相同范围内的粒子的粒子群; 计算颗粒的平均值; 以及使用所述平均值来获得所述对象的状态估计值。
-
公开(公告)号:KR1020140079947A
公开(公告)日:2014-06-30
申请号:KR1020120149178
申请日:2012-12-20
Applicant: 한국전자통신연구원
Abstract: The present invention provides an image recording device for a vehicle. The image recording device includes a camera unit of which the photographing direction is adjustable; a driver view direction detection unit which detects the view direction of a driver; a control unit which controls the camera unit to make the photographing direction correspond to the detected view direction; and a storage unit in which images taken by the camera unit are stored. By acquiring and storing images in the view direction of a driver without installing multiple cameras, the present invention is able to acquire images of an accident happening not only at the front of the vehicle but also at the side of the vehicle or an accident of other vehicles.
Abstract translation: 本发明提供一种车辆用图像记录装置。 图像记录装置包括摄像单元,摄像方向可调整; 驱动器视图方向检测单元,其检测驾驶员的视图方向; 控制单元,其控制相机单元以使拍摄方向对应于检测到的视图方向; 以及存储由相机单元拍摄的图像的存储单元。 通过在不安装多台摄像机的情况下,在驾驶员的视线方向上获取和存储图像,本发明能够获取不仅在车辆前方而且在车辆侧面发生事故或其他事故的图像 汽车。
-
公开(公告)号:KR1020140028755A
公开(公告)日:2014-03-10
申请号:KR1020120095790
申请日:2012-08-30
Applicant: 한국전자통신연구원
IPC: G06T7/00
CPC classification number: G06K9/3233 , G06F17/30781 , G06K9/522
Abstract: The present invention relates to a high speed detection apparatus and a method for target objects having various sizes which detects target object candidates from an area of interest using a plurality of correlation filters according to sizes of the pre-stored target objects, and detects the target objects at a high speed from an image having the target objects with various sizes by determining whether the detected target object candidates are the detection target objects. [Reference numerals] (110) Preprocessing unit; (120) Target object candidates detecting unit; (130) Target object determining unit; (140) Correlation filter generating unit; (150) Filter storing unit; (160) Image input unit
Abstract translation: 高速检测装置和方法技术领域本发明涉及一种高速检测装置和方法,其特征在于,具有各种尺寸的目标对象,其使用多个相关滤波器根据预先存储的目标对象的大小从感兴趣区域检测目标对象候选,并且检测目标 通过确定检测到的目标对象候选是否是检测目标对象,从具有各种大小的目标对象的图像以高速度对象。 (附图标记)(110)预处理单元; (120)目标候选检测单元; (130)目标对象确定单元; (140)相关滤波器生成单元; (150)过滤器存放单元; (160)图像输入单元
-
公开(公告)号:KR1020140012474A
公开(公告)日:2014-02-03
申请号:KR1020120079444
申请日:2012-07-20
Applicant: 한국전자통신연구원
CPC classification number: G06K9/00288 , G06K9/00771 , G08B13/19686
Abstract: A method for protecting personal information based on a face recognition is able to prevent the leakage of personal information by selectively performing a privacy masking or a mosaic process about a specific person and the remaining face images except the specific person after detecting the specific person who is a search object based on a face recognition technology in video image information which is recorded or stored in an image monitoring system including a CCTV and a DVR. Moreover, the automation of a personal information protection about a video image is possible by applying the selective privacy masking or mosaic process which efficiently includes or excludes a specific person by applying the face recognition technology to bulk video image information like a DVR/image storage device. In case personal image information is provided according to regulations of a personal information protection law through the automation of personal information protection, the present invention is able to easily and efficiently perform existing exhausting and difficult works which are necessary for the mosaic process about the image of other people. [Reference numerals] (102) Image frame separating unit; (104) Human detecting unit; (106) Face detecting unit; (108) Face recognizing unit; (110) Privacy processing unit; (112) Privacy level selecting unit; (AA) Search object face information; (BB) Video image; (CC) Personal information protection video image
Abstract translation: 基于面部识别来保护个人信息的方法能够通过在检测到特定人物之后选择性地执行关于特定人物的隐私遮蔽或马赛克过程以及除了特定人物之外的剩余面部图像来防止个人信息泄漏 基于在包括CCTV和DVR的图像监视系统中记录或存储的视频图像信息中的面部识别技术的搜索对象。 此外,通过应用有选择性的隐私掩蔽或马赛克过程,可以对视频图像进行个人信息保护的自动化,该过程通过将人脸识别技术应用于诸如DVR /图像存储设备的批量视频图像信息来有效地包括或排除特定人物 。 在通过个人信息保护的自动化根据个人信息保护法的规定提供个人图像信息的情况下,本发明能够容易且有效地执行马赛克过程中关于图像的图像所必需的现有的排气和困难的作品 其他人。 (102)图像帧分离单元; (104)人体检测单元; (106)面部检测单元; (108)面部识别单元; (110)隐私处理单位; (112)隐私级别选择单位; (AA)搜索对象面信息; (BB)视频图像; (CC)个人信息保护视频图像
-
-
-
-
-
-
-
-
-