-
公开(公告)号：KR1020120008211A
公开(公告)日：2012-01-30
申请号：KR1020100068944
申请日：2010-07-16
Applicant: 성균관대학교산학협력단
Abstract: PURPOSE: An object recognition and posture assumption method of a robot is provided to efficiently discover the postures possible to occur around three dimensional lines by using three dimensional lines based on direction and relationship. CONSTITUTION: The object recognition and posture assuming method of robot is as follows. A prior first interpretation is propagated to a state space by using information about robot motions(S210). A probability distribution function of each posture candidate of an object is obtained based on the first evidence selected among obtained evidences(S220). The probability value [probability value] toward each posture candidate of the object material calculates based on 'the second evidence among the obtained evidence' (S230). The probability value for each posture candidates of the object is computed based on the second evidence among the obtained evidences(S230). The second interpretation is created based on probability distribution function and probability value (S240). Current first interpretation is created based on the propagated first and second interpretations(S250) and returns to the very stage for propagating to sate space of the first interpretation.
Abstract translation: 目的：提供机器人的物体识别和姿态假设方法，通过使用基于方向和关系的三维线，有效发现三维线周围可能出现的姿势。 构成：机器人的物体识别和姿态假设方法如下。 先前的解释通过使用关于机器人运动的信息传播到状态空间（S210）。 基于从获得的证据中选择的第一证据，获得对象的每个姿势候选者的概率分布函数（S220）。 基于“获得的证据中的第二个证据”，对象物的每个姿势候选者的概率值[概率值]计算（S230）。 根据获得的证据中的第二个证据来计算对象的每个姿势候选者的概率值（S230）。 第二个解释是基于概率分布函数和概率值创建的（S240）。 基于传播的第一和第二解释创建当前的第一解释（S250），并且返回到传播到第一解释的状态空间的阶段。
-
公开(公告)号：KR101225644B1
公开(公告)日：2013-01-24
申请号：KR1020100068944
申请日：2010-07-16
Applicant: 성균관대학교산학협력단
Abstract: 로봇의 물체 인식 및 자세 추정 기술이 개시된다. 개시된 기술 중 일실시예에 따른 방법은 (a) 로봇의 움직임에 대한 정보를 이용하여 이전의 제1 인터프리테이션들을 상태 공간에 전파하는 단계; (b) 현재 획득된 증거들 중 선택된 제1 증거를 기초로 대상 물체의 자세 후보들 각각에 대한 확률 분포 함수를 획득하는 단계; (c) 상기 현재 획득된 증거들 중 제2 증거를 기초로 상기 대상 물체의 자세 후보들 각각에 대한 확률값을 산출하는 단계; (d) 상기 획득된 확률 분포 함수 및 상기 산출된 확률값을 기초로, 제2 인터프리테이션들을 생성하는 단계; 및 (e) 상기 전파된 제1 인터프리테이션들 및 상기 생성된 제2 인터프리테이션들을 기초로 현재의 제1 인터프리테이션들을 생성하고, 상기 (a) 단계로 복귀하는 단계를 포함한다.
-
公开(公告)号：KR101054736B1
公开(公告)日：2011-08-05
申请号：KR1020100041642
申请日：2010-05-04
Applicant: 성균관대학교산학협력단
IPC: H04N13/00
CPC classification number: G06K9/6212 , G06K9/6215 , G06K2209/40 , G06T5/40 , H04N13/204
Abstract: PURPOSE: A 3D object recognition and pose measuring method are provided to measure similarity with a target model with a D image by using a mult-part HSV color histogram. CONSTITUTION: A 2D image is obtained through a stereo camera(110). A 3D point cloud is generated from the 2D image(115). Similarity between a color histogram of image patches in the 2D image and a color histogram of a target model is measured(135). Based on the similarity, a plurality of image patches among the image patches are selected(160). A pose hypothesis corresponding to the selected image patches is predicted(165).
Abstract translation: 目的：提供3D对象识别和姿态测量方法，通过使用多部分HSV颜色直方图来测量与D图像的目标模型的相似度。 构成：通过立体相机（110）获得2D图像。 从2D图像生成3D点云（115）。 测量2D图像中图像斑块的颜色直方图与目标模型的颜色直方图之间的相似性（135）。 基于相似度，选择图像块中的多个图像块（160）。 预测与所选择的图像块相对应的姿势假设（165）。
-
-
Search Results
3
records
(took 0.016 seconds)
