-
-
公开(公告)号:KR1020170101723A
公开(公告)日:2017-09-06
申请号:KR1020160024672
申请日:2016-02-29
Applicant: 한국항공대학교산학협력단
Abstract: 본발명은항공영상분류장치및 그방법에관한것으로서, 콘볼루션신경망(convolutional neural network) 기반으로영상으로부터특징을추출하는특징추출부, 다층퍼셉트론(Multilayer Perceptron)과로지스틱회귀(Logistic Regression) 알고리즘을이용하여상기추출된특징을분류하는분류부를포함한다.
Abstract translation: 本发明使用的空间像分类装置和涉及一种方法,卷积神经网络(卷积神经网络),其基于提取的特征从图像特征提取单元,多层感知器(多层感知器)配有逻辑回归(逻辑回归)算法 还有一个用于分类提取的特征的分类器。
-
公开(公告)号:KR101320922B1
公开(公告)日:2013-10-21
申请号:KR1020120064331
申请日:2012-06-15
Applicant: 한국항공대학교산학협력단
IPC: G06T7/20
CPC classification number: G06T7/246 , G06K9/3233 , G06T7/277 , G06T7/285 , G06T7/44 , G06T7/90 , G06T2207/10021
Abstract: PURPOSE: Movement tracking and avatar control methods using a weighting detection area are provided to steadily perform tracking without block of a multi-tube seam area. CONSTITUTION: A calculation device sets up two or more regions of interest (ROI) (S520). The calculation device stores the set ROIs in a storage unit. The calculation device extracts feature point values which are pre-set for the ROIs (S530). The calculation device calculates a central point of the feature point values by adding a weighted value to a previous ROI corresponding to the ROIs of a previous frame (S540). The calculation device changes the center of the ROI into the central point of the feature point values (S570). [Reference numerals] (AA) Start; (BB) End; (S510) Input image; (S520) Set region of interest; (S530) Extract feature point value; (S540) Calculate central point; (S5550) Change center of region of interest; (S560) Gather change of center; (S570) Change posture of avatar
Abstract translation: 目的:提供使用加权检测区域的运动跟踪和头像控制方法,以稳定地执行跟踪,而不需要多管接缝区域。 构成:计算装置设置两个或更多个感兴趣区域(ROI)(S520)。 计算装置将设置的ROI存储在存储单元中。 计算装置提取为ROI设定的特征点值(S530)。 计算装置通过将加权值加到与先前帧的ROI相对应的先前ROI中来计算特征点值的中心点(S540)。 计算装置将ROI的中心改变为特征点值的中心点(S570)。 (附图标记)(AA)开始; (BB)结束; (S510)输入图像; (S520)设置感兴趣的区域; (S530)提取特征点值; (S540)计算中心点; (S5550)改变感兴趣区域的中心; (S560)收集中心的变化; (S570)改变头像的姿势
-
公开(公告)号:KR101782914B1
公开(公告)日:2017-09-28
申请号:KR1020160024672
申请日:2016-02-29
Applicant: 한국항공대학교산학협력단
Abstract: 본발명은항공영상분류장치및 그방법에관한것으로서, 콘볼루션신경망(convolutional neural network) 기반으로영상으로부터특징을추출하는특징추출부, 다층퍼셉트론(Multilayer Perceptron)과로지스틱회귀(Logistic Regression) 알고리즘을이용하여상기추출된특징을분류하는분류부를포함한다.
-
-
-