-
公开(公告)号：KR101388196B1
公开(公告)日：2014-04-23
申请号：KR1020120127648
申请日：2012-11-12
Applicant: 재단법인대구경북과학기술원
CPC classification number: G06K9/00442 , G06F3/01 , G06K9/325 , G06K9/40
Abstract: A disclosed technology relates to a method and a device for recognizing handwriting based on a camera of a mobile terminal and comprises a step of obtaining images of the handwriting by capturing the handwriting using the camera of the mobile terminal; a step of detecting region of interest (ROI) from the images; a step of removing noise from the ROI; and a step of recognizing text in the ROI and of setting the text as a character input means of the mobile terminal. Therefore, the technology provides an effective method of enabling a person who is not familiar with character input method of the existing mobile device to conveniently use. [Reference numerals] (110) Obtaining handwriting images; (120) Detecting region of interest (ROI); (130) Removing noise; (140) Recognizing text; (150) Setting the text as a character input means; (AA) Start; (BB) End
Abstract translation: 所公开的技术涉及一种用于基于移动终端的照相机识别手写的方法和装置，并且包括通过使用移动终端的相机捕获手写来获取手写图像的步骤; 从图像中检测感兴趣区域（ROI）的步骤; 从ROI去除噪声的步骤; 以及识别ROI中的文本并将文本设置为移动终端的字符输入装置的步骤。 因此，该技术提供了一种使不熟悉现有移动设备的字符输入方法的人能够方便地使用的有效方法。 （附图标记）（110）获取手写图像; （120）检测感兴趣区域（ROI）; （130）消除噪音; （140）确认文本; （150）将文本设置为字符输入装置; （AA）开始; （BB）结束
-
-
公开(公告)号：KR101853276B1
公开(公告)日：2018-04-30
申请号：KR1020160174548
申请日：2016-12-20
Applicant: 재단법인대구경북과학기술원
CPC classification number: G06T7/11 , G06T7/254 , G06T7/593 , G06T2207/10028
Abstract: 본발명은깊이영상에서의손 영역검출방법및 그장치에관한것이다. 본발명에따르면, 깊이영상으로부터사용자의손 영역및 상기손 영역을제외한바디영역을각각추정하는단계와, 상기손 영역이추정된경우에상기손 영역내 화소들의깊이정보로부터최소깊이값 및최대깊이값을탐색하는단계와, 상기손 영역외부의화소들중 상기최소깊이값과최대깊이값 사이에해당하는깊이값을가지는화소들을선택하여상기손 영역에포함시키는단계, 및상기바디영역내 화소들중 상기손 영역의중심으로부터손가락이향하는방향상에위치한화소들을선택하여상기손 영역에포함시켜상기손 영역을최종검출하는단계를포함하는깊이영상에서의손 영역검출방법을제공한다. 본발명에의하면, 깊이영상으로부터추정된사용자의손 영역에대한에러를보정할수 있어손 영역을더욱정확하게검출할수 있으며, 이를통하여손 모양이나제스쳐등의인식률을높이는데기여할수 있는효과가있다.
-
公开(公告)号：KR101757550B1
公开(公告)日：2017-07-13
申请号：KR1020150146902
申请日：2015-10-21
Applicant: 재단법인대구경북과학기술원
Abstract: 기기간 파일송수신시스템및 기기간 파일송수신방법이개시된다. 본발명의일실시예에따른기기간 파일송수신시스템은, 파일교환명령이발생하는수신기와연결되는송신기중, 상기파일교환명령과연관되는수신파일을갖는제1 송신기를식별하는식별부와, 상기제1 송신기각각으로파일교환명령을전송하고, 상기파일교환명령이단독수신과관련되면, 상기제1 송신기중, 어느하나의기기로기기지정명령을더 전송하는전송부, 및상기기기지정명령에응답하는제2 송신기로부터, 상기수신파일을수신하는수신부를포함한다.
-
-
Patent Agency Ranking
公开(公告)号：KR101601660B1
公开(公告)日：2016-03-10
申请号：KR1020140154424
申请日：2014-11-07
Applicant: 재단법인대구경북과학기술원
IPC: G06T7/00
Abstract: 본발명은깊이영상을이용한손 영역분류방법및 그장치에관한것이다. 본발명에따르면, 카메라의깊이영상으로부터손 영역을검출하는단계와, 상기손 영역에포함된각 픽셀의특징값을연산하되, 상기깊이영상내에서상기픽셀로부터설정오프셋만큼떨어진두 지점을기준으로임의크기로생성된두 오프셋패치에대한평균깊이값의차이를이용하여상기각 픽셀의특징값을연산하는단계와, 입력되는특징값에대응하는분류라벨을출력하도록미리학습된랜덤결정트리앙상블(Random Decision Forest) 내에상기각 픽셀의특징값을개별입력하여상기손 영역을다수의파트로분류하는단계와, 상기분류된다수의파트별로중심점을설정하는단계, 및상기설정된중심점들을이용하여상기손의스켈레톤(skeleton)을추출하는단계를포함하는깊이영상을이용한손 영역분류방법을제공한다. 상기깊이영상을이용한손 영역분류방법및 그장치에따르면, 손의깊이영상만으로손 영역을다수의파트로효과적으로분류할수 있으며이를통해손의스켈레톤정보를정확하게추출하고손 동작의인식률을향상시킬수 있는이점이있다.
Abstract translation: 本发明涉及使用深度图像的手部分类方法及其装置。 根据本发明，使用深度图像的手部分类方法可以包括：从相机的深度图像检测手部分; 通过使用生成为具有一定尺寸的两个偏移补丁的平均深度值的差异，基于由两个点间隔开的两个点来计算手部分中包括的每个像素的特征值，具体地计算每个像素的特征值 作为参考，与深度图像内的像素的偏移设置; 通过将每个像素的特征值输入随机决策林中，将手部分分为多个部分，其被训练以预先输出与输入的特征值相对应的分类标签; 为多个分类部分设置中心点; 并使用设定的中心点提取手的骨架。 根据使用深度图像的手部分类方法及其装置，通过仅使用深度图像可以有效地将手部分分为多个部分，从而允许精确提取手的骨架信息和提高手部运动的识别率。
-
公开(公告)号：KR1020150042674A
公开(公告)日：2015-04-21
申请号：KR1020130123325
申请日：2013-10-16
Applicant: 재단법인대구경북과학기술원
Abstract: 본발명은외부환경변화에따른모달리티의결합또는선택을통하여사용자인식성능을향상시키기위한방법을제공한다. 본발명은멀티모달사용자인식방법에관한것으로, 얼굴모델템플릿(face model templates)을이용하여입력영상의얼굴특징벡터에대한얼굴유사도를계산하는단계; 음성모델템플릿(speech model templates)을이용하여입력오디오의음성특징벡터에대한음성유사도를계산하는단계; 상기입력영상과연관된조명의크기를이용하여얻어진제1 가중치를상기얼굴유사도에적용하여제1 값을생성하는단계; 상기입력오디오와연관된잡음의크기를이용하여얻어진제2 가중치를상기음성유사도에적용하여제2 값을생성하는단계; 및상기제1 값및 상기제2 값을결합하여얻어진최종스코어를이용하여, 사용자를인식하는단계를포함하는멀티모달사용자인식방법을제공한다.
Abstract translation: 本发明提供一种根据外部环境的变化通过组合或选择模式来增强用户识别性能的方法。 多模式用户识别方法技术领域本发明涉及一种多模式用户识别方法，其提供多模式用户识别方法，其包括：使用面部模型模板计算输入图像的面部特征向量的面部相似度的步骤; 使用语音模型模板计算输入音频的声乐特征向量的声音相似度的步骤; 通过将与使用与输入图像相关联的照明的大小获得的第一权重应用于面部相似度来创建第一值; 通过将声音相似度应用于使用与输入音频相关联的噪声大小获得的第二权重来产生第二值的步骤; 以及使用从组合第一和第二值获得的最终得分来识别用户的步骤。
-
公开(公告)号：KR101436695B1
公开(公告)日：2014-09-01
申请号：KR1020120128410
申请日：2012-11-13
Applicant: 재단법인대구경북과학기술원
Abstract: 깊이 영상을 이용한 손 검출 장치 및 방법이 개시된다. 깊이 영상을 이용한 손 검출 장치는 카메라에 의해 손을 촬영한 영상을 입력 받고, 상기 카메라와 상기 손 간의 거리에 기초하여, 상기 입력된 영상으로부터 상기 손에 대한 손 영역을 검출하는 검출부와, 상기 손 영역으로부터 2차원의 제1 손 모양 및 3차원의 제2 손 모양을 추정하고, 추정된 상기 제1 손 모양 및 상기 제2 손 모양을 이용하여, 상기 손에 대한 최종 모양을 추정하는 프로세서를 포함한다.
-
公开(公告)号：KR1020140060191A
公开(公告)日：2014-05-19
申请号：KR1020120126989
申请日：2012-11-09
Applicant: 재단법인대구경북과학기술원
CPC classification number: G06F3/017 , G06F3/0304 , G06F3/0346 , G06F3/042 , H05B37/02
Abstract: The present invention relates to a touch-based virtual switch control method and a device for the same. According to the present invention, the touch-based virtual switch control method using a three-dimensional camera which is installed in a target space to acquire the depth data of a subject includes: a step of detecting part of the body of a user from an image taken by the three-dimensional camera; a step of determining if the part of the body touches a virtual switch area predefined in the target space; a step of, if it is determined that the part of the body touches the virtual switch area, recognizing the gesture taken by the part of the body after the touch; a step of extracting a control command corresponding to the recognized gesture from a gesture database in which control commands by gesture are stored; and a step of controlling the operation of at least one electronic device according to the extracted control command. According to the touch-based virtual switch control method and the method thereof, part of the body of a user is traced, and when the part of the body of the user touches the predefined area and takes a certain gesture, a designated operation corresponding to the gesture is applied to an electric switch. By doing so, separate fixed structures such as the existing wires and switches are not necessary. It is also possible to define and use switch actions at a desired position by flexibly changing the virtual switch area.
Abstract translation: 本发明涉及一种基于触摸的虚拟开关控制方法及其装置。 根据本发明，使用安装在目标空间中以获取对象的深度数据的三维相机的基于触摸的虚拟切换控制方法包括：从一个或多个对象的身体检测用户的身体的一部分 由三维摄像机拍摄的图像; 确定身体的一部分是否与目标空间中预定义的虚拟交换区域相接触的步骤; 如果确定身体的一部分接触虚拟开关区域，则识别在触摸之后身体部分拍摄的手势的步骤; 从存储有通过手势的控制命令的手势数据库中提取与所识别的手势相对应的控制命令的步骤; 以及根据所提取的控制命令来控制至少一个电子设备的操作的步骤。 根据基于触摸的虚拟切换控制方法及其方法，跟踪用户身体的一部分，并且当用户的身体的一部分触摸预定区域并采取某种手势时，指定的操作对应于 手势被施加到电开关。 通过这样做，不需要单独的固定结构，例如现有的电线和开关。 还可以通过灵活地改变虚拟交换机区域来在所需位置定义和使用交换机动作。
-
公开(公告)号：KR101382892B1
公开(公告)日：2014-04-08
申请号：KR1020120111255
申请日：2012-10-08
Applicant: 재단법인대구경북과학기술원
IPC: G06T7/00
CPC classification number: G06K9/00228 , G06T3/4053 , G06T5/20 , G06T2207/20028
Abstract: The present invention relates to a face recognition method in a low resolution image and a face recognition device in a low resolution image in which an internal image patch is replaced with a high resolution image patch, which is retained through learning in advance, for a low resolution face image detected from a surveillance camera. As an embodiment, the face recognition method in a low resolution image includes the steps of maintaining a database in which a first image patch is recorded in correspondence to a second image patch having higher resolution than the first image patch; identifying a recognition target region from image information input from the camera; among image patches of the recognition target region, defining an image patch not satisfying a reference resolution as the first image patch; in the database, searching the database for the second image patch recorded in correspondence to the defined first image patch; and recovering the recognition target region by replacing found second image patch with the defined first image patch. [Reference numerals] (110) Database control unit; (115) Database; (120) Region identification unit; (130) Image processing unit; (140) Image recovery unit; (150) Compensation unit; (160) Face recognition unit
Abstract translation: 本发明涉及一种低分辨率图像中的人脸识别方法和低分辨率图像中的面部识别装置，其中内部图像贴片被高分辨率图像贴片替代，该高分辨率图像贴片通过预先学习而保持低电平 从监控摄像机检测到的分辨率脸部图像。 作为实施例，低分辨率图像中的面部识别方法包括以下步骤：对应于具有比第一图像补丁更高分辨率的第二图像补片，维持其中记录第一图像补片的数据库; 从从相机输入的图像信息中识别识别对象区域; 在所述识别目标区域的图像块中，将不满足参考分辨率的图像块定义为所述第一图像块; 在所述数据库中，搜索所述数据库中对应于所述定义的第一图像补丁记录的所述第二图像补丁; 以及通过用所定义的第一图像补片替换所发现的第二图像补片来恢复识别目标区域。 （附图标记）（110）数据库控制单元; （115）数据库; （120）区域识别单位; （130）图像处理单元; （140）图像恢复单元; （150）赔偿单位; （160）脸部识别单元
-
-
-
-
-
-
-
-
-
Search Results
50
records
(took 0.025 seconds)
