公开(公告)号：US09633449B2

公开(公告)日：2017-04-25

申请号：US14699532

申请日：2015-04-29

Applicant: FUJITSU LIMITED

Inventor： Song Wang ,  Akihiro Minagawa ,  Wei Fan ,  Jun Sun

IPC: G06K9/00 ,  G06T7/40 ,  G06T5/00 ,  G06K9/46 ,  G06T7/73 ,  G06T7/11 ,  G06T7/168 ,  G06T7/90

CPC classification number: G06T7/168 ,  G06K9/4604 ,  G06K9/4652 ,  G06K9/4671 ,  G06K2009/4657 ,  G06T5/007 ,  G06T7/11 ,  G06T7/73 ,  G06T7/90 ,  G06T2200/12 ,  G06T2207/10004 ,  G06T2207/10024 ,  G06T2207/20048 ,  G06T2207/30168

Abstract: Embodiments provide an apparatus and method for detecting a color checker in an image including: an extracting unit to extract sliding window features of sliding windows at multiple positions from an input image by using a predefined sliding window; a selecting unit to select, by using multiple predefined sliding window templates, a sliding window having a minimum distance from each sliding window template from the sliding windows at the multiple positions as a matched sliding window of the sliding window template; and a determining unit to determine segmentation lines of colored squares of a color checker in the input image according to a matching result of each sliding window template.

公开(公告)号：US09292931B2

公开(公告)日：2016-03-22

申请号：US13912805

申请日：2013-06-07

Applicant: FUJITSU LIMITED

Inventor： Pan Pan ,  Yuan He ,  Jun Sun ,  Satoshi Naoi

IPC: G06T7/00

CPC classification number: G06T7/0085 ,  G06T7/12

Abstract: The invention provides an apparatus and method for extracting a boundary of an object in an image and an electronic device. The apparatus includes: a position determining unit, configured to determine a start point and an end point of a boundary of an object in an image and to determine a position of a reference point relevant to the start point and the end point; a first direction determining unit, configured to determine a first direction of the boundary; a gradient map obtaining unit, configured to obtain a gradient map of a first region; a gradient attenuating unit, configured to attenuate in the gradient map the gradients of a second region; and an extracting unit, configured to extract a boundary of an object. The technology of the invention can improve the accuracy of boundary extracting, and can be applied in the field of image processing.

Abstract translation: 本发明提供一种用于提取图像和电子设备中的对象的边界的装置和方法。 该装置包括：位置确定单元，被配置为确定图像中的对象的边界的起始点和终点，并确定与开始点和终点相关的参考点的位置; 第一方向确定单元，被配置为确定所述边界的第一方向; 梯度图获取单元，被配置为获得第一区域的梯度图; 梯度衰减单元，被配置为在梯度图中衰减映射第二区域的梯度; 以及提取单元，被配置为提取对象的边界。 本发明的技术可以提高边界提取的准确性，可以应用于图像处理领域。

公开(公告)号：US09070035B2

公开(公告)日：2015-06-30

申请号：US13904477

申请日：2013-05-29

Applicant: FUJITSU LIMITED

Inventor： Yifeng Pan ,  Jun Sun ,  Yuan He ,  Satoshi Naoi

IPC: G06K9/00 ,  G06K9/18 ,  G06K9/32

CPC classification number: G06K9/18 ,  G06K9/00442 ,  G06K9/3208 ,  G06K2209/01

Abstract: The disclosure provides a document processing apparatus, method and a scanner. The document processing apparatus includes: a text line extraction unit extracting a text line from an input document; a language classification unit determining whether an OCR process is necessary for a language of the input document; an OCR unit determining, by performing the OCR process, an OCR confidence in the case that it is determined that the OCR process is necessary; an graphic feature recognition unit determining an graphic feature recognition confidence; and a determination unit determining a combination confidence based on at least one of the determined graphic feature recognition confidences and the determined OCR confidences, and determining an orientation of the input document based on the combination confidences. This technical solution can determine better an orientation of the document, and is especially applicable when the quality of the image of the document is deteriorated.

Abstract translation: 本公开提供了一种文件处理装置，方法和扫描仪。 文件处理装置包括：文本行提取单元，从输入文档中提取文本行; 语言分类单元，确定输入文档的语言是否需要OCR处理; OCR单元通过执行OCR处理确定在确定OCR处理是必要的情况下的OCR置信度; 确定图形特征识别置信度的图形特征识别单元; 以及确定单元，基于所确定的图形特征识别信息和所确定的OCR信心中的至少一个确定组合置信度，以及基于所述组合机密来确定所述输入文档的取向。 该技术解决方案可以更好地确定文档的方向，并且当文档的图像的质量劣化时特别适用。

公开(公告)号：US20150036920A1

公开(公告)日：2015-02-05

申请号：US14448049

申请日：2014-07-31

Applicant: FUJITSU LIMITED

Inventor： Chunpeng WU ,  Wei Fan ,  Yuan He ,  Jun Sun

IPC: G06K9/62

CPC classification number: G06K9/6272 ,  G06K9/4628 ,  G06K9/6267 ,  G06N3/0454 ,  G06N3/0481

Abstract: The present invention relates to a convolutional-neural-network-based classifier, a classifying method by using a convolutional-neural-network-based classifier and a method for training the convolutional-neural-network-based classifier. The convolutional-neural-network-based classifier comprises: a plurality of feature map layers, at least one feature map in at least one of the plurality of feature map layers being divided into a plurality of regions; and a plurality of convolutional templates corresponding to the plurality of regions respectively, each of the convolutional templates being used for obtaining a response value of a neuron in the corresponding region.

Abstract translation: 本发明涉及基于卷积神经网络的分类器，通过使用基于卷积神经网络的分类器的分类方法和用于训练基于卷积神经网络的分类器的方法。 基于卷积神经网络的分类器包括：多个特征图层，所述多个特征图层中的至少一个特征图层中的至少一个特征图被划分为多个区域; 以及分别对应于多个区域的多个卷积模板，每个卷积模板用于获得相应区域中的神经元的响应值。

公开(公告)号：US12210826B2

公开(公告)日：2025-01-28

申请号：US17695857

申请日：2022-03-16

Applicant: FUJITSU LIMITED

Inventor： Yiling Cao ,  Zhongguang Zheng ,  Jun Sun

IPC: G06F40/253 ,  G06F40/169 ,  G06F40/279 ,  G06F40/30 ,  G06N3/04 ,  G06N3/045

Abstract: A method of presenting prompt information by utilizing a neural network which includes a BERT model and a graph convolutional neural network (GCN), comprising: generating a first vector based on a combination of an entity, a context of the entity, a type of the entity and a part of speech of the context by using BERT model; generating a second vector based on each of predefined concepts by using BERT model; generating a third vector based on a graph which is generated based on the concepts and relationships thereamong, by using GCN; generating a fourth vector by concatenating the second and third vectors; calculating semantic similarity between the entity and each concept based on the first and fourth vectors; determining, based on the first vector and the semantic similarity, that the entity corresponds to one of the concepts; and generating the prompt information based on the determined concept.

公开(公告)号：US11386248B2

公开(公告)日：2022-07-12

申请号：US16695834

申请日：2019-11-26

Applicant: FUJITSU LIMITED

Inventor： Liuan Wang ,  Jun Sun

IPC: G06F30/20 ,  G06F17/18 ,  G06N3/04

Abstract: A method and a device for simulating atomic dynamics includes setting initial positions for multiple specific atoms in a specific scene; calculating, based on the initial positions, positions of the multiple specific atoms at each time in a first time series by utilizing a Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) configured with respect to the specific scene, as real positions; calculating, based on the initial positions, positions of the multiple specific atoms at the same time in the first time series by utilizing a generative adversarial network (GAN), as predicted positions; improving a configuration of the GAN based on the real positions and the predicted positions at a same time. Initial positions are settable for multiple atoms to be simulated in a scene; positions of the multiple atoms to be simulated are calculated at each time in a second time series in the scene by utilizing the improved GAN.

公开(公告)号：US11227213B2

公开(公告)日：2022-01-18

申请号：US15799151

申请日：2017-10-31

Applicant: FUJITSU LIMITED

Inventor： Liuan Wang ,  Wei Fan ,  Jun Sun

IPC: G06N3/08 ,  G06N3/04

Abstract: A device and a method for improving a processing speed of a neural network and applications thereof in the neural network where the device includes a processor configured to perform: determining, according to a predetermined processing speed improvement target, a dimension reduction amount of each of one or more parameter matrixes in the neural network obtained through training; preprocessing each parameter matrix based on the dimension reduction amount of the parameter matrix; and retraining the neural network based on a result of the preprocessing to obtain one or more dimension reduced parameter matrixes so as to ensure performance of the neural network meets a predetermined requirement. According to the embodiments of the present disclosure, it is possible to significantly improve the processing speed of the neural network while ensuring the performance of the neural network meets the predetermined requirement.

公开(公告)号：US11049007B2

公开(公告)日：2021-06-29

申请号：US15587803

申请日：2017-05-05

Applicant: Fujitsu Limited

Inventor： Song Wang ,  Wei Fan ,  Jun Sun

IPC: G06N3/08 ,  G06N3/04

Abstract: A recognition apparatus based on a deep neural network, a training apparatus and methods thereof. The deep neural network is obtained by inputting training samples comprising positive samples and negative samples into an input layer of the deep neural network and training. The apparatus includes: a judging unit configured to judge that a sample to be recognized is a suspected abnormal sample when confidences of positive sample classes in a classification result outputted by an output layer of the deep neural network are all less than a predefined threshold value. Hence, reliability of a confidence of a classification result outputted by the deep neural network may be efficiently improved.

公开(公告)号：US10187546B2

公开(公告)日：2019-01-22

申请号：US15410040

申请日：2017-01-19

Applicant: FUJITSU LIMITED

Inventor： Wei Liu ,  Wei Fan ,  Jun Sun

IPC: H04N1/387 ,  G06K9/46 ,  H04N1/393 ,  G06K9/00 ,  G06K9/32 ,  G06K9/36

Abstract: The present application relates to a method and a device for correcting a document image captured by an image pick-up device. The method includes: determining world coordinates of four vertices of the document image; calculating an original aspect ratio of the document image based on a correspondence between the world coordinates of the four vertices and projective coordinates of the four vertices in a projective space, and an intrinsic matrix and characteristics of an extrinsic matrix of the image pick-up device; determining a projective transformation matrix based on the world coordinates of the four vertices and the aspect ratio; and obtaining a corrected document image based on the determined projective transformation matrix and the document image. According to the application, perspective transformation can be corrected by using only one captured image and an original image can be recovered based on an original aspect ratio.

公开(公告)号：US10074163B2

公开(公告)日：2018-09-11

申请号：US15334437

申请日：2016-10-26

Applicant: FUJITSU LIMITED

Inventor： Jile Jiao ,  Wei Fan ,  Jun Sun

IPC: G06K9/00 ,  G06T5/20 ,  G06T5/00

CPC classification number: G06T5/20 ,  G06T5/009 ,  G06T2207/30176

Abstract: An image correction method and an image correction apparatus when the image correction method includes: an identifying step of identifying each pixel in an image as a foreground pixel or a background pixel; a background filling step of estimating brightness of a background corresponding to a foreground pixel based on brightness and gradient of the brightness of background pixels adjacent to the foreground pixel to fill the background located in a position of the foreground pixel, to obtain a background illumination map of the image according to filled backgrounds along with background pixels; and a correcting step of correcting the image based on the brightness of each pixel in the image and the background illumination map. A non-uniform illumination image can be corrected effectively.