-
公开(公告)号：KR101289386B1
公开(公告)日：2013-07-29
申请号：KR1020110117821
申请日：2011-11-11
Applicant: 재단법인대구경북과학기술원
Abstract: 시차맵의 값이 낮은 부분에서도 양호한 검출 성능을 나타낼 수 있고 검출된 장애물체를 영역별로 정교하게 분리할 수 있는 스테레오 비전 기반의 장애물체 검출 방법 및 장치로서, 스테레오 비전 기반으로 좌·우 영상을 입력받아 시차맵을 생성하는 단계와, 상기 시차맵에 히스토그램 평활화 과정을 실행하는 단계와, 상기 히스토그램 평활화가 실행된 시차맵을 V-시차맵으로 변환하고, 상기 V-시차맵으로부터 도로 정보를 추출하는 단계와, 상기 히스토그램 평활화가 실행된 시차맵과 상기 추출된 도로 정보를 이용하여 컬럼 검출을 실행함으로써, 장애물체의 영역을 검출한다. 또한, 상기 컬럼 검출에 의해 장애물체의 영역이 검출된 시차맵에 대해 카메라 정보를 이용하여 조감도 맵핑을 실행함으로써, 시차맵을 조감도로 변환하는 단계와, 상기 조감도에서의 각 행에 대해 최대치 구간을 검출하여 행별 클러스터링을 실행함으로써, 장애물체 영역을 개별적으로 분리하는 단계와, 상기 클러스터링된 조감도에 대해, 역조감도 맵핑을 실행하여 시차맵으로 변환하는 단계와, 상기 역조감도 맵핑에 의해 취득된 시차맵으로부터의 시차값을 이용하여, 장애물체의 정보를 추출한다.
-
公开(公告)号：KR1020130052393A
公开(公告)日：2013-05-22
申请号：KR1020110117802
申请日：2011-11-11
Applicant: 재단법인대구경북과학기술원
CPC classification number: G06T2207/20016 , H04N13/271 , H04N13/275 , H04N2013/0081 , H04N2013/0092
Abstract: PURPOSE: An object detecting device using a 3D image and a method thereof are provided to shorten object detection time by detecting an object according to a size change of the object by using distance information of the object. CONSTITUTION: An image processing unit(120) generates an image pyramid about a 2D image of an object obtained from an image obtaining unit(110), divides a 3D image obtained from the image obtaining unit into one or more object candidate areas by using distance information, and estimates an object size of each object candidate area. An object detecting unit(130) selects a 2D image level corresponding to the object size from the image pyramid and detects the object from the 2D image included in the level by using a detection mask. [Reference numerals] (111) 2D image obtaining unit; (112) 3D image obtaining unit; (121) Image pyramid generating unit; (122) Area dividing unit; (123) Object size estimating unit; (131) Image level selecting unit; (132) Object extracting unit
Abstract translation: 目的：提供使用3D图像的对象检测装置及其方法，通过使用对象的距离信息，通过根据对象的尺寸变化检测对象来缩短对象检测时间。 构成：图像处理单元（120）生成关于从图像获取单元（110）获得的对象的2D图像的图像金字塔，通过使用距离将从图像获取单元获得的3D图像分割成一个或多个候选对象区域 信息，并估计每个对象候选区域的对象大小。 物体检测单元（130）从图像金字塔中选择与物体大小相对应的2D图像水平，并通过使用检测掩模从包含在该层次中的2D图像检测对象。 （111）2D图像获取单元; （112）3D图像获取单元; （121）图像金字塔生成单元; （122）区域划分单位; （123）对象尺寸估计单元; （131）图像级选择单元; （132）对象提取单元
-
公开(公告)号：KR101213951B1
公开(公告)日：2012-12-18
申请号：KR1020110094473
申请日：2011-09-20
Applicant: 재단법인대구경북과학기술원
IPC: G01S19/03
Abstract: PURPOSE: A location determination apparatus and method using a vision system are provided to improve continuity and reliability of location determination. CONSTITUTION: A location determination apparatus comprises a satellite navigation system(510), a database(520), a vision system(530), a map matching system(540), and a location calculating part(550). The database stores coordinate data of a plurality of buildings. The map matching system calculates the coordinate data of a second target building recognized through a stereo camera by applying map matching. The vision system extracts coordinate data of a first target building from the database and calculates two first pseudo distances using the coordinate data of the second target building and the coordinate data of the first target building. The location calculating part calculates a second pseudo distance by location data and measures location of a moving body using the two first pseudo distances and the second pseudo distances. [Reference numerals] (510) Satellite navigation system; (520) Database; (530) Vision system; (540) Map matching system; (550) Location calculating part
Abstract translation: 目的：提供一种使用视觉系统的位置确定装置和方法，以改善位置确定的连续性和可靠性。 构成：位置确定装置包括卫星导航系统（510），数据库（520），视觉系统（530），地图匹配系统（540）和位置计算部分（550）。 数据库存储多个建筑物的坐标数据。 地图匹配系统通过应用地图匹配来计算通过立体相机识别的第二目标建筑物的坐标数据。 视觉系统从数据库提取第一目标建筑物的坐标数据，并使用第二目标建筑物的坐标数据和第一目标建筑物的坐标数据计算两个第一伪距离。 位置计算部分通过位置数据计算第二伪距离，并使用两个第一伪距离和第二伪距离来测量移动体的位置。 （附图标记）（510）卫星导航系统; （520）数据库; （530）视觉系统; （540）地图匹配系统; （550）位置计算部
-
公开(公告)号：KR1020120048958A
公开(公告)日：2012-05-16
申请号：KR1020100110466
申请日：2010-11-08
Applicant: 재단법인대구경북과학기술원
Abstract: PURPOSE: An object tracking and location estimating method is provided to increase tracking/estimating performance of a filter due to accurate prediction performance. CONSTITUTION: An object area is measured according to the size of an ROI(Region Of Interest)(510). Feature points of the detected object are extracted. A two-dimensional homography matrix is presumed by using the extracted feature points(520). A location of particles is predicted by using the presumed homography matrix(530). A weighted value of a particle is updated(540). A particle filter is degraded by the updated weighted value(550).
Abstract translation: 目的：提供对象跟踪和位置估计方法，以提高跟踪/估计性能，因为准确的预测性能。 构成：根据ROI（感兴趣区域）的大小（510）测量对象区域。 提取检测对象的特征点。 通过使用提取的特征点（520）推测二维单应性矩阵。 通过使用推定的单应性矩阵（530）预测颗粒的位置。 粒子的加权值被更新（540）。 粒子滤波器被更新的加权值降级（550）。
-
公开(公告)号：KR1020120044699A
公开(公告)日：2012-05-08
申请号：KR1020100106136
申请日：2010-10-28
Applicant: 재단법인대구경북과학기술원
CPC classification number: G06T5/002 , G06T7/70 , G06T2207/20024 , H04N13/271
Abstract: PURPOSE: An obstacle object detecting method and an apparatus thereof on a road are provided to easily detect an obstacle object of desired size by using horizontal side information of an obstacle object. CONSTITUTION: An obstacle detecting apparatus performs stereo matching about a left image and a right image. The apparatus generates a depth map(310). The apparatus accumulates the distribution of the depth value. The apparatus generates a U-variation map according to the obstacle object(320). The apparatus detects the obstacle object(330).
Abstract translation: 目的：提供道路上的障碍物检测方法及其装置，通过使用障碍物的水平侧信息容易地检测出所需尺寸的障碍物。 构成：障碍物检测装置执行关于左图像和右图像的立体匹配。 该装置生成深度图（310）。 该装置累积深度值的分布。 该装置根据障碍物（320）生成U变化图。 该装置检测障碍物（330）。
-
Patent Agency Ranking
公开(公告)号：KR1020110060600A
公开(公告)日：2011-06-08
申请号：KR1020090117223
申请日：2009-11-30
Applicant: 재단법인대구경북과학기술원
CPC classification number: H04N13/257 , G06T7/593 , G06T2207/10012 , H04N13/243 , H04N13/246 , H04N2013/0081
Abstract: PURPOSE: An obstacle location and velocity estimator and method are provided to improve the precision of the location and velocity estimation of an object with the application of an EKF. CONSTITUTION: A stereo matching unit(410) extracts a depth map from right and left images(400). An object detection unit(420) extracts an object. An integer parallax estimation unit(430) extracts integer parallax for a detected ROI(Region Of Interest). A precision parallax estimating unit(450) estimates precision parallax on a two-dimensional parabola. By the application of an EKF(Extended Kalman Filter), an EKF unit(480) estimates the location and velocity of the object on three-dimensional coordinates.
Abstract translation: 目的：提供障碍物位置和速度估计器和方法，以通过应用EKF来提高对象的位置和速度估计的精度。 构成：立体匹配单元（410）从左右图像提取深度图（400）。 物体检测单元（420）提取物体。 整数视差估计单元（430）针对检测到的ROI（感兴趣区域）提取整数视差。 精密视差估计单元（450）估计二维抛物线上的精度视差。 通过应用EKF（扩展卡尔曼滤波器），EKF单元（480）估计对象在三维坐标上的位置和速度。
-
公开(公告)号：KR1020100068751A
公开(公告)日：2010-06-24
申请号：KR1020080127212
申请日：2008-12-15
Applicant: 재단법인대구경북과학기술원
Abstract: PURPOSE: A method and a device for matching image, and an object detecting apparatus using the same are provided to generate a disparity map with high reliability by removing and interpolating the disparity map with low reliability. CONSTITUTION: An initial matching unit(410) matches a stereo image and generates an initial disparity map. A disparity remover(420) calculates the reliable value of the disparity values of the disparity map and compares the reliable value with the reference value. The disparity remover removes the disparity value with the reliable value below the reference value. A disparity interpolator(430) interpolates the disparity value with low reliability, and generates and outputs the interpolated disparity map.
Abstract translation: 目的：提供一种用于匹配图像的方法和装置，以及使用该方法和装置的对象检测装置，通过以低可靠性去除和内插视差图来生成具有高可靠性的视差图。 构成：初始匹配单元（410）匹配立体图像并生成初始视差图。 视差去除器（420）计算视差图的视差值的可靠值，并将可靠值与参考值进行比较。 视差去除器以可靠的值低于参考值去除视差值。 视差内插器（430）以低可靠性内插视差值，并生成并输出内插视差图。
-
公开(公告)号：KR1020100054285A
公开(公告)日：2010-05-25
申请号：KR1020080113132
申请日：2008-11-14
Applicant: 재단법인대구경북과학기술원
CPC classification number: G06T7/97 , G06T2207/20228 , H04N13/271 , H04N2013/0081
Abstract: PURPOSE: An object feature extraction device for accurately extracting the feature of object regardless of the targets of the various shapes is provided to accurately extract the specific information of a road using a v-parallax map. CONSTITUTION: If a time difference map of an image of a target is inputted, a v-time difference map generator(110) creates the v- parallax map through a parallax map. A binarizer(120) binarizes the v- parallax map through the maximum frequency according to a column and a row. A common unit extractor(130) creates the commonly binarized v-time difference map by the extraction of a common unit of a v-time difference map binarized by column and a v-time difference map binarized by row. An interpolator(140) outputs the feature information of the object by the interpolation of the collaboration binarization of the v- parallax map.
Abstract translation: 目的：提供用于精确提取物体特征的物体特征提取装置，而不管各种形状的目标如何，以使用v-视差图精确提取道路的具体信息。 构成：如果输入目标图像的时差图，则v-时差图生成器（110）通过视差映射生成v-视差图。 二值化器（120）根据列和行将通过最大频率的v-视差图二值化。 公共单元提取器（130）通过提取由列二进制化的v时间差异映射图的公共单元和由行二进制化的v-时间差图来创建通常二进制化的v-时差图。 插值器（140）通过对v-视差映射的协作二值化的插值来输出对象的特征信息。
-
公开(公告)号：KR100938194B1
公开(公告)日：2010-01-21
申请号：KR1020080073303
申请日：2008-07-28
Applicant: 재단법인대구경북과학기술원
Inventor： 이충희
IPC: H04N13/00
CPC classification number: H04N13/204 , H04N13/271
Abstract: PURPOSE: A method for detecting an object and an apparatus thereof are provided to perform stereo matching on a photographed image to detect an object included in the image, thereby estimating the accurate location, distance and speed based on the detected object. CONSTITUTION: A matching unit(810) matches photographed stereo images to produce a depth map. A converting unit(820) converts the depth map so that the depth map is placed horizontally along with a vertical line. An object detecting unit(830) maps parallax values of pixels of the converted depth map on the vertical line. The object detecting unit detects an object included in the image according to distribution of the parallax values.
Abstract translation: 目的：提供一种用于检测物体的方法及其装置，以对拍摄的图像执行立体匹配，以检测包括在图像中的对象，从而基于检测到的对象来估计准确的位置，距离和速度。 构成：匹配单元（810）匹配拍摄的立体图像以产生深度图。 A转换单元（820）转换深度图，使得深度图与垂直线一起水平放置。 对象检测单元（830）将经转换的深度图的像素的视差值映射在垂直线上。 物体检测单元根据视差值的分布来检测包含在图像中的物体。
-
公开(公告)号：KR102249818B1
公开(公告)日：2021-05-10
申请号：KR1020170171587
申请日：2017-12-13
Applicant: 재단법인대구경북과학기술원
Inventor： 이충희
Abstract: 일실시예에따른학습데이터를생성하는방법및 장치는학습데이터들의분포특성을기초로, 오버샘플링할학습데이터들을격리학습데이터와노말학습데이터로구분하고, 격리학습데이터와노말학습데이터간의비율에따라격리학습데이터및 노말학습데이터각각으로부터오버샘플링할추가학습데이터들의개수를결정하며, 오버샘플링할학습데이터들각각의개수에기초하여격리학습데이터및 노말학습데이터각각에대응하는추가학습데이터들을생성한다.
-
-
-
-
-
-
-
-
-
Search Results
112
records
(took 0.028 seconds)
