-
公开(公告)号：KR101268520B1
公开(公告)日：2013-06-04
申请号：KR1020090123943
申请日：2009-12-14
Applicant: 한국전자통신연구원
CPC classification number: G06K9/4614
Abstract: 본 발명에 따르면, 입력 영상의 x축, y축 다차원 그래디언트를 이용하여 하 라이크 필터로부터 보다 다양한 특징을 추출할 수 있으며, 추출된 입력 영상의 특징과 True 영상에 대한 다중 임계값 및 False 영상에 대한 다중 임계값을 단계별로 이용하여 입력 영상을 정확하게 True/False로 분류할 수 있다. 따라서, 연산량은 적으면서도 인식률은 높아지므로 빠르면서도 정확하게 물체를 인식할 수 있어 실시간 영상 인식이 가능하다는 효과가 있다.
하 라이크(Haar like), 특징, 그래디언트, 절대값, 인식률, 연산량-
公开(公告)号：KR1020130042783A
公开(公告)日：2013-04-29
申请号：KR1020110106863
申请日：2011-10-19
Applicant: 한국전자통신연구원
CPC classification number: G06K9/6228 , G06K9/6269 , G06N99/005 , G06F17/00 , G06F17/18
Abstract: PURPOSE: A characteristic vector classification device and a method thereof are provided to reduce calculation amount by determining recalculation by comparing an error value and an error threshold value. CONSTITUTION: A variable calculation unit(120) calculates a weighted value and errors through a value set in an initial condition setting unit. A variable calculation unit calculates a training vector. A loop determination unit(130) determines recalculation by comparing an error value and an error threshold value. A boundary surface generation unit(140) generates a boundary surface by receiving a termination signal from the loop determination unit. An error value calculated by the variable calculation unit is a standardized average square error value. [Reference numerals] (110) Initial condition setting unit; (120) Variable calculation unit; (130) Loop determination unit; (140) Boundary surface generation unit;
Abstract translation: 目的：提供特征矢量分类装置及其方法，通过比较误差值和误差阈值来确定重新计算来减少计算量。 构成：可变计算单元（120）通过在初始条件设置单元中设置的值来计算加权值和误差。 可变计算单元计算训练向量。 循环确定单元（130）通过比较错误值和错误阈值来确定重新计算。 边界面生成单元（140）通过从循环判定单元接收终止信号来生成边界面。 由变量计算单元计算的误差值是标准化的平均平方误差值。 （附图标记）（110）初始条件设定单元; （120）变量计算单位; （130）循环确定单元; （140）边界面生成单元;
-
公开(公告)号：KR1020120066999A
公开(公告)日：2012-06-25
申请号：KR1020100128379
申请日：2010-12-15
Applicant: 한국전자통신연구원
CPC classification number: G06F13/28 , G06F13/1668 , G06F13/24
Abstract: PURPOSE: A direct memory access(DMA) controller and a method for operating the same are provided to reduce the generated frequency of interrupts representing the completion of DMA transmission. CONSTITUTION: A bus interface(610) receives the DMA controlling data of a processor. A DMA transmission controlling logic(640) implements a DMA transmission operation by referring the DMA controlling data which is set in a register set unit(620). After the completion of the DMA transmission operation, the DMA transmission controlling logic generates interrupts. A DMA master interface(650) implements a physical data transmission operation. An FIFO buffer memory device(660) temporarily saves transmission data and related controlling signals in the DMA transmission operation.
Abstract translation: 目的：提供直接存储器访问（DMA）控制器及其操作方法，以减少表示DMA传输完成的中断的产生频率。 构成：总线接口（610）接收处理器的DMA控制数据。 DMA传输控制逻辑（640）通过参考设置在寄存器组单元（620）中的DMA控制数据来实现DMA传输操作。 在DMA传输操作完成之后，DMA传输控制逻辑产生中断。 DMA主接口（650）实现物理数据传输操作。 FIFO缓冲存储器件（660）在DMA传输操作中临时保存传输数据和相关控制信号。
-
公开(公告)号：KR1020120041002A
公开(公告)日：2012-04-30
申请号：KR1020100102547
申请日：2010-10-20
Applicant: 한국전자통신연구원 , 고려대학교 산학협력단
IPC: G06T7/00
CPC classification number: G06K9/4642 , G06K9/3241
Abstract: PURPOSE: An object detecting device and a system thereof are provided to form a feature vector by using HOG(Histogram of Oriented Gradients) and pixel coordinate information . CONSTITUTION: An outline image extracting unit(120) extracts an outline image from an input image. A feature vector calculating unit(130) calculates a feature vector from the outline image by using HOG and pixel coordinate information. The HOG displays frequency distribution of gradient vectors for pixels in the outline image. The pixel coordinate information is changed by spatial distribution of the gradient vectors. An object determining unit(150) determines a target object corresponding to the feature vector referring to pre-learned data.
Abstract translation: 目的：提供一种物体检测装置及其系统，通过使用HOG（定向梯度直方图）和像素坐标信息形成特征向量。 构成：轮廓图像提取单元（120）从输入图像中提取轮廓图像。 特征向量计算单元（130）通过使用HOG和像素坐标信息从轮廓图像计算特征向量。 HOG显示轮廓图像中像素的梯度向量的频率分布。 通过梯度矢量的空间分布来改变像素坐标信息。 对象确定单元（150）参考预先学习的数据来确定与特征向量相对应的目标对象。
-
公开(公告)号：KR1020120038764A
公开(公告)日：2012-04-24
申请号：KR1020100100399
申请日：2010-10-14
Applicant: 한국전자통신연구원
CPC classification number: H04N7/183 , G06K9/6203
Abstract: PURPOSE: An image recognition method and an image recognition device thereof are provided to offer additional memory to store a standard image and a reduced image. CONSTITUTION: An image recognition device selects a part of data from standard image data(S110). The image recognition device recognizes an image based on selected data(S120). The image recognition device reduces the selected data(S130). The image recognition device recognizes the image based on the reduced data(S140).
Abstract translation: 目的：提供图像识别方法及其图像识别装置，以提供用于存储标准图像和缩小图像的附加存储器。 构成：图像识别装置从标准图像数据中选择一部分数据（S110）。 图像识别装置基于所选择的数据识别图像（S120）。 图像识别装置减少所选择的数据（S130）。 图像识别装置基于减少的数据识别图像（S140）。
-
Patent Agency Ranking
公开(公告)号：KR1020130071782A
公开(公告)日：2013-07-01
申请号：KR1020110139202
申请日：2011-12-21
Applicant: 한국전자통신연구원
CPC classification number: G06F13/4059 , G06F13/1605 , G06F13/1668 , G06F13/4031 , Y02D10/14
Abstract: PURPOSE: A bus bridge apparatus is provided to maximize data transmission performance among interconnections, by transmitting and receiving data by considering characteristics among different interconnections. CONSTITUTION: A slave port (210) performs interface with a master device of a bus based interconnection (110), and receives read and write transmission command, address data and write data from the master device, and transmits read data to the master device. A command controller (220) receives the transmission command, and an address buffer (230) stores the address data. A write data buffer (240) stores the write data, and a read data buffer (270) stores the read data. A protocol converter (260) outputs the write data of the master device to the slave device, by using the address and write data in case of the write transmission command. [Reference numerals] (200) Bus bridge apparatus; (210) Slave port; (220) Bus based interconnection; (230) Address buffer; (240) Write data buffer; (250) Transmission mode controller; (260) Protocol converter; (270) Read data buffer; (AA) Bus based interconnection; (BB) Network based interconnection
Abstract translation: 目的：提供一种总线桥接装置，通过考虑不同互连之间的特性，通过发送和接收数据来最大化互连之间的数据传输性能。 构成：从端口（210）与基于总线的互连（110）的主设备进行接口，并且从主设备接收读写传输命令，地址数据和写数据，并将读数据发送到主设备。 命令控制器（220）接收发送命令，地址缓冲器（230）存储地址数据。 写数据缓冲器（240）存储写数据，读数据缓冲器（270）存储读数据。 在写入发送命令的情况下，协议转换器（260）通过使用地址和写入数据将主设备的写入数据输出到从设备。 （附图标记）（200）总线桥装置; （210）从站端口; （220）总线互连; （230）地址缓冲区; （240）写入数据缓冲区; （250）传输模式控制器; （260）协议转换器; （270）读数据缓冲; （AA）总线互连; （BB）基于网络的互连
-
公开(公告)号：KR1020130058286A
公开(公告)日：2013-06-04
申请号：KR1020110124209
申请日：2011-11-25
Applicant: 한국전자통신연구원 , 고려대학교 산학협력단
IPC: G06T7/00
CPC classification number: G06K9/00369 , G06K9/6269
Abstract: PURPOSE: A pedestrian detection method of a pedestrian detection device is provided to perform pedestrian detection on a search window using a second classifier with high accuracy after reducing the number of search windows, thereby performing highly accurate detection of an object while reducing the complexity of detection procedure and power consumption. CONSTITUTION: A pedestrian detection device obtains an image from a digital image device and performs blocking of search windows(210,220). The pedestrian detection device selects a specific block from blocks determined by a pre-learned classifier(230) and produces a specific vector of HOG features from the selected block(240,250). The pedestrian detection device calculates a SVM(Support Vector Machine) response value using the produced feature vector, and performs a first object detection by applying the response value to an AdaBoost Classifier(260,270). If a pedestrian is detected at the first object detection, the pedestrian detection device performs a second object detection to the search window(280). [Reference numerals] (210) Obtain an image; (220) Performs blocking of search windows; (230) Pre-learned classifier; (240) Select a specific block; (250) Produce a specific vector; (260) Calculates a SVM response value; (270) Perform a first object detection; (280) Perform a second object detection; (290) Output decision; (AA) No; (BB) Yes
Abstract translation: 目的：提供一种行人检测装置的行人检测方法，在减少搜索窗口数量之后，使用第二分类器在搜索窗口上进行行人检测，从而对目标进行高精度的检测，同时降低检测的复杂度 程序和功耗。 规定：行人检测装置从数字图像装置获取图像并执行搜索窗口的封锁（210,220）。 行人检测装置从预先学习的分类器（230）确定的块中选择特定的块，并从所选择的块（240,250）生成特定的HOG特征向量。 行人检测装置使用所产生的特征向量来计算SVM（Support Vector Machine）响应值，并通过将响应值应用于AdaBoost分类器（260,270）来执行第一对象检测。 如果在第一物体检测中检测到行人，则行人检测装置对搜索窗口进行第二物体检测（280）。 （附图标记）（210）获取图像; （220）执行搜索窗口的阻止; （230）预先学习分类器; （240）选择一个特定块; （250）生成特定载体; （260）计算SVM响应值; （270）执行第一个对象检测; （280）执行第二对象检测; （290）产出决定; （AA）否 （BB）是的
-
公开(公告)号：KR1020110073173A
公开(公告)日：2011-06-29
申请号：KR1020100028448
申请日：2010-03-30
Applicant: 한국전자통신연구원
Abstract: PURPOSE: A memory system comprising a plurality of DMA channels and an integrating management method for a plurality of DMA channels are provided to improve data transmission efficiency of a memory controller by the integrated management of multichannel memory controller and connected multiple DMA channels. CONSTITUTION: A memory controller(200) performs data transceiving operation with a memory(100). The memory controller comprises multiple channels which are physically separated each other. A DMA controller (300) is connected to the multiple channels of the memory controller and includes multiple DMA channels which are physically separated each other. The DMA controller performs data transceiving operation with the memory through the multiple DMA channels and the memory controller. An access module(400) connects the channels of the memory controller with the DMA channels each other.
Abstract translation: 目的：提供包括多个DMA通道的存储系统和用于多个DMA通道的集成管理方法，以通过多通道存储器控制器和连接的多个DMA通道的集成管理来提高存储器控制器的数据传输效率。 构成：存储器控制器（200）用存储器（100）执行数据收发操作。 存储器控制器包括物理上彼此分离的多个通道。 DMA控制器（300）连接到存储器控制器的多个通道，并且包括物理上彼此分离的多个DMA通道。 DMA控制器通过多个DMA通道和存储器控制器与存储器执行数据收发操作。 访问模块（400）将存储器控制器的通道与DMA通道相互连接。
-
公开(公告)号：KR101729976B1
公开(公告)日：2017-04-25
申请号：KR1020110055762
申请日：2011-06-09
Applicant: 한국전자통신연구원
CPC classification number: G06K9/6239 , G06K9/6269
Abstract: 본발명에따른영상인식장치는입력영상으로부터입력벡터를추출하는입력벡터추출부, 입력벡터를프로젝션벡터를사용하여압축벡터로변환하는압축벡터변환부, 훈련벡터를입력받고, 훈련벡터의폴딩연산을통해획득한프로젝션벡터를사용하여훈련파라미터를생성하는훈련파라미터생성부, 및압축벡터를훈련파라미터를사용한영상분류를통해영상인식데이터를출력하는영상분류부를포함한다.
-
公开(公告)号：KR101725121B1
公开(公告)日：2017-04-12
申请号：KR1020110106863
申请日：2011-10-19
Applicant: 한국전자통신연구원
CPC classification number: G06K9/6228 , G06K9/6269 , G06N99/005
Abstract: 본발명은특징벡터분류장치및 방법에관한것이다. 본발명의실시예에의한특징벡터분류장치는초기조건설정부, 훈련벡터를입력받아상기초기조건설정부에서설정된값을통해오차및 가중치를계산하는변수계산부, 상기계산된오차및 가중치를입력받고, 상기계산된오차와오차문턱을비교하여재계산여부를결정하는루프결정부및 상기루프결정부로부터종료신호를입력받으면경계면을생성하는경계면생성부를포함한다. 따라서본 발명은높은분류성능을가지면서도적은연산만을요구하는특징벡터분류장치및 방법을제공한다.
-
-
-
-
-
-
-
-
-
Search Results
26
records
(took 0.023 seconds)
