영상에서 물체 검출 장치 및 방법
    1.
    发明公开
    영상에서 물체 검출 장치 및 방법 有权
    用于检测图像中对象的装置和方法

    公开(公告)号:KR1020130036514A

    公开(公告)日:2013-04-12

    申请号:KR1020110100635

    申请日:2011-10-04

    Abstract: PURPOSE: An object detection device in an image and a method thereof are provided to divide an object area in an original image by using a thermal image for the image, thereby detecting an object. CONSTITUTION: An object detection unit(102) detects an object from an input image by using a thermal image. A restoration unit(106) restores an in-painting area by using surroundings information of the in-painting area set in an in-painting area setting unit(104) in the input image. A similarity comparison unit(108) determines whether or not the object is existed in the input image through similarity comparison of a restoration image and the input image. When the object is existed, an object area division unit(110) extracts an area of the object. [Reference numerals] (102) Object detection unit; (104) In-painting area setting unit; (106) Restoration unit; (108) Similarity comparison unit; (110) Object area division unit; (AA) Obtained image; (BB) Division result of an object area;

    Abstract translation: 目的:提供图像中的物体检测装置及其方法,通过使用图像的热图像来分割原始图像中的对象区域,从而检测对象。 构成:物体检测单元(102)通过使用热图像从输入图像检测物体。 恢复单元(106)通过使用设置在输入图像中的在画区域设置单元(104)中的绘画区域的环境信息来恢复绘画区域。 相似度比较单元(108)通过恢复图像和输入图像的相似性比较来确定输入图像中是否存在对象。 当对象存在时,对象区域分割单元(110)提取对象的区域。 (附图标记)(102)对象检测单元; (104)绘画区设定单位; (106)修复单位; (108)相似性比较单位; (110)对象区划单位; (AA)获取图像; (BB)对象区域的分割结果;

    로봇의 위치 인식 방법 및 장치
    2.
    发明公开
    로봇의 위치 인식 방법 및 장치 无效
    提供机器人位置的方法和装置

    公开(公告)号:KR1020110070004A

    公开(公告)日:2011-06-24

    申请号:KR1020090126634

    申请日:2009-12-18

    CPC classification number: B25J9/1653 B25J9/1664

    Abstract: PURPOSE: A self positioning method and apparatus of a robot are provided to enable more accurate self positioning of a robot by recognizing pillars of a structure as feature points, not obstacles. CONSTITUTION: A self positioning method of a robot is as follows. Data about the coordinates and radius of pillars based on a global coordinate system are set as input(S200). LRF(Laser Range Finder) data based on a robot coordinate system are set as input(S202). Segmentation is implemented using the distance and angle between two or more LRF data. Segment filtering is implemented. Pillars are detected through circle fitting. The pillars are recognized from the distance and angle between the pillars calculated through circle fitting. Transformation matrices are estimated and the position and heading direction of the robot are computed(S204).

    Abstract translation: 目的:提供机器人的自定位方法和装置,以通过将结构的支柱识别为特征点而不是障碍物来实现机器人的更精确的自定位。 构成:机器人的自定位方法如下。 基于全局坐标系的坐标和半径的数据被设置为输入(S200)。 基于机器人坐标系的LRF(Laser Range Finder)数据被设置为输入(S202)。 使用两个或更多个LRF数据之间的距离和角度来实现分段。 实现段过滤。 支柱通过圆圈拟合检测。 从通过圆形拟合计算出的支柱之间的距离和角度识别支柱。 估计变换矩阵,并计算机器人的位置和行进方向(S204)。

    비디오 정보를 통한 인간 행동 예측 방법
    3.
    发明公开
    비디오 정보를 통한 인간 행동 예측 방법 无效
    人类活动预测形式流行视频的方法

    公开(公告)号:KR1020130091596A

    公开(公告)日:2013-08-19

    申请号:KR1020120013000

    申请日:2012-02-08

    CPC classification number: G06K9/00771 G06K9/4642

    Abstract: PURPOSE: A human activity prediction method using video information is provided to predict a human activity by detecting a clue to start an incident or an action as early as possible with not enough video information, thereby early recognizing a human activity. CONSTITUTION: A system extracts spatial and temporal region features from a video stream including video information about human activities. The system groups the extracted spatial and temporal region features into a plurality of visual languages based on appearances. The system calculates an activity likelihood value by modeling each human activity into an integral histogram of the visual languages. The system predicts a human activity based on the calculated activity likelihood value.

    Abstract translation: 目的:提供一种使用视频信息的人类活动预测方法,通过检测不到足够的视频信息尽可能早地发现事件或动作的线索来预测人类的活动,从而早日识别人类活动。 构成:系统从视频流中提取空间和时间区域特征,包括有关人类活动的视频信息。 该系统基于外观将提取的空间和时间区域特征组合成多种视觉语言。 系统通过将每个人类活动建模成视觉语言的整体直方图来计算活动似然值。 系统基于计算的活动可能性值预测人类活动。

    합성 영상을 이용한 예제 영상 생성방법과 상황 인식 방법 및 그 장치
    4.
    发明公开
    합성 영상을 이용한 예제 영상 생성방법과 상황 인식 방법 및 그 장치 无效
    使用组合视频生成训练视频和识别情况的方法及其装置

    公开(公告)号:KR1020120060599A

    公开(公告)日:2012-06-12

    申请号:KR1020100122188

    申请日:2010-12-02

    Inventor: 유상원 유원필

    CPC classification number: G06K9/00718 G06K9/00335

    Abstract: PURPOSE: An example video generating method, a condition recognizing method thereof, and an apparatus thereof are provided to reduce effort, time, and cost to obtain a real photographing image in order to generate an example image. CONSTITUTION: A synthesis image generating unit(502) generates a synthesis image based on configuration information of a source image. A synthesis image selecting unit(503) selects a synthesis image which satisfies a restriction condition of the generated synthesis image. An example image configuring unit(504) configures the example image. The example image includes the selected synthesis image.

    Abstract translation: 目的:提供一种示例视频产生方法,其条件识别方法及其装置,以减少获得实际拍摄图像的努力,时间和成本,以便生成示例图像。 构成:合成图像生成部(502)基于源图像的配置信息生成合成图像。 合成图像选择单元(503)选择满足所生成的合成图像的限制条件的合成图像。 示例图像配置单元(504)配置示例图像。 示例图像包括所选择的合成图像。

    영상분석장치 및 그 방법
    6.
    发明公开
    영상분석장치 및 그 방법 无效
    视频分析装置及其方法

    公开(公告)号:KR1020120067890A

    公开(公告)日:2012-06-26

    申请号:KR1020100129531

    申请日:2010-12-16

    CPC classification number: G06K9/00805 G07C5/0866

    Abstract: PURPOSE: A video analyzing apparatus and a method thereof are provided to summarize and record a photographed image according to an event. CONSTITUTION: A path analyzing unit(110) analyzes an object which is expressed on a first-person image. The path analyzing unit generates path information of the object. An event analyzing unit(120) classifies the first-person image through the path information. A sub-event is classified according to a predetermined standard of the object and the subject.

    Abstract translation: 目的:提供一种视频分析装置及其方法,用于根据事件对拍摄的图像进行总结和记录。 构成:路径分析单元(110)分析在第一人称图像上表达的对象。 路径分析单元生成对象的路径信息。 事件分析单元(120)通过路径信息对第一人称图像进行分类。 子事件根据对象和被摄体的预定标准进行分类。

Patent Agency Ranking