Patents Examined by Qian Yang
  • Patent number: 10354170
    Abstract: Embodiments of the present invention disclose a method and an apparatus of establishing an image search relevance prediction model, and an image search method and apparatus. The method of establishing an image search relevance prediction model comprises: training a pre-constructed original deep neural network by using a training sample, wherein the training sample comprises: a query and image data, and the original deep neural network comprises: a representation vector generation network and a relevance calculation network; and using the trained original deep neural network as the image search relevance prediction model. The technical solution of the present invention optimizes the existing image search technology, and has stronger capabilities than the prior art as well as various integrations and variations in terms of semantic matching between a query and an image text, semantic matching between a query and image content, click generalization and the like.
    Type: Grant
    Filed: September 30, 2016
    Date of Patent: July 16, 2019
    Assignee: BAIDU ONLINE NETWORK TECHNOLOGY (BEIJING) CO., LTD.
    Inventors: Libo Fu, Heng Luo, Gaolin Fang, Wei Xu
  • Patent number: 10354172
    Abstract: A verification device includes an acquisition unit that acquires image data which is an object to be verified, a classification unit that includes a plurality of classifiers caused to learn the image data acquired by the acquisition unit, using a plurality of different pieces of learning data, with respect to a target event to obtain values indicating a possibility corresponding to the event, and a specification unit that specifies a state of the object to be verified which is the image acquired by the acquisition unit from the values obtained by the plurality of classifiers.
    Type: Grant
    Filed: August 21, 2017
    Date of Patent: July 16, 2019
    Assignee: Cresco Ltd.
    Inventor: Yuji Ayatsuka
  • Patent number: 10339669
    Abstract: An approach is provided for a vertex-based evaluation of polygon similarity. The approach, for instance, involves processing, by a computer vision system, an image to generate a first set of vertices of a first polygon representing an object depicted in the image. The approach also involves for each vertex in the first set of vertices, determining a closest vertex in a second set of vertices of a second polygon, and determining a distance between said each vertex in the first set of vertices and the closest vertex in the second set of vertices. The approach further involves calculating a polygon similarity of the first polygon with respect to the second polygon based on a total of the distance determined for said each vertex in the first set of vertices normalized to a number of vertices in the first set of vertices.
    Type: Grant
    Filed: August 22, 2017
    Date of Patent: July 2, 2019
    Assignee: HERE Global B.V.
    Inventors: Richard Kwant, Anish Mittal, David Lawlor
  • Patent number: 10332001
    Abstract: A deep learning based compression (DLBC) system generates a progressive representation of the encoded input image such that a client device that requires the encoded input image at a particular target bitrate can readily be transmitted the appropriately encoded data. More specifically, the DLBC system computes a representation that includes channels and bitplanes that are ordered based on importance. For a given target rate, the DLBC system truncates the representation according to a trained zero mask to generate the progressive representation. Transmitting a first portion of the progressive representation enables a client device with the lowest target bitrate to appropriately playback the content. Each subsequent portion of the progressive representation allows the client device to playback the content with improved quality.
    Type: Grant
    Filed: February 22, 2017
    Date of Patent: June 25, 2019
    Assignee: WaveOne Inc.
    Inventors: Oren Rippel, Lubomir Bourdev
  • Patent number: 10325364
    Abstract: A metrology system for obtaining a measurement representative of a thickness of a layer on a substrate includes a camera positioned to capture a color image of at least a portion of the substrate. A controller is configured to receive the color image from the camera, store a predetermined path in a coordinate space of at least two dimension including a first color channel and a second color channel, store a function that provides a value representative of a thickness as a function of a position on the predetermined path, determine a coordinate of a pixel in the coordinate space from color data in the color image for the pixel, determine a position of a point on the predetermined path that is closest to the coordinate of the pixel, and calculate a value representative of a thickness from the function and the position of the point on the predetermined path.
    Type: Grant
    Filed: August 25, 2017
    Date of Patent: June 18, 2019
    Assignee: Applied Materials, Inc.
    Inventor: Dominic J. Benvegnu
  • Patent number: 10319170
    Abstract: A folded bill recognizing method and a folded bill recognizing device are provided. The folded bill recognizing device includes: a bill input port configured to receive a to-be-recognized bill or a sample bill; a signal collecting module configured to collect a CIS image of the bill, to obtain an infrared transmission image T and an infrared reflection image F; a signal recognizing module configured to recognize whether the to-be-recognized bill has a fold; and a receiving/rejecting module configured to perform a receiving or rejecting operation on the to-be-recognized bill. The device can effectively recognize a folded bill.
    Type: Grant
    Filed: July 13, 2015
    Date of Patent: June 11, 2019
    Assignee: GRG Banking Equipment Co., Ltd.
    Inventors: Guanglu Liu, Jian Chen, Zhuming Xiao, Weirui Zheng
  • Patent number: 10311337
    Abstract: A method for providing an integrated feature map by using an ensemble of a plurality of outputs from a convolutional neural network (CNN) is provided. The method includes steps of: a CNN device (a) receiving an input image and applying a plurality of modification functions to the input image to thereby generate a plurality of modified input images; (b) applying convolution operations to each of the modified input images to thereby obtain each of modified feature maps corresponding to each of the modified input images; (c) applying each of reverse transform functions, corresponding to each of the modification functions, to each of the corresponding modified feature maps, to thereby generate each of reverse transform feature maps corresponding to each of the modified feature maps; and (d) integrating at least part of the reverse transform feature maps to thereby obtain an integrated feature map.
    Type: Grant
    Filed: September 4, 2018
    Date of Patent: June 4, 2019
    Assignee: STRADVISION, INC.
    Inventors: Kye-Hyeon Kim, Yongjoong Kim, Insu Kim, Hak-Kyoung Kim, Woonhyun Nam, SukHoon Boo, Myungchul Sung, Donghun Yeo, Wooju Ryu, Taewoong Jang, Kyungjoong Jeong, Hongmo Je, Hojin Cho
  • Patent number: 10311606
    Abstract: Provided are improved referenceless multi-material beam hardening correction methods, with an emphasis on maintaining data quality for real-world imaging of geologic materials with a view towards automation. A referenceless post reconstruction (RPC) correction technique is provided that applies the corrections in integrated attenuation space. A container-only pre-correction technique also is provided to allow automation of the segmentation process required for beam hardening correction methods.
    Type: Grant
    Filed: August 15, 2015
    Date of Patent: June 4, 2019
    Inventors: Jeremy James Holt, Andrew Maurice Kingston, Adrian Paul Sheppard
  • Patent number: 10311576
    Abstract: A computer extracts a plurality of candidate lines that are observed from a position of an imaging device that captures an image of an object from among a plurality of candidate lines included in shape information of the object. The computer generates plural pieces of association information indicating a prescribed number of combinations obtained by respectively associating the prescribed number of candidate lines of the observed plurality of candidate lines with the prescribed number of feature lines of a plurality of feature lines detected from the image. The computer determines an association result according to respective errors of the plural pieces of association information.
    Type: Grant
    Filed: August 18, 2017
    Date of Patent: June 4, 2019
    Assignee: FUJITSU LIMITED
    Inventors: Nobuyasu Yamaguchi, Yojiro Numata, Hiroshi Kuwabara
  • Patent number: 10304196
    Abstract: The present invention relates to a method for converting an image file or document file into a drawing file including the steps of extracting an outline from an image of a subject file to be drafted; detecting corner feature points from the image of the subject file to be drafted based on the extracted outline to coordinate the corner feature points into outer edge coordinate information and connecting the detected corner feature points with a line to generate a first image; removing the outline from the image of the subject file to be drafted and detecting one or more elements to coordinate the elements into element coordinate information and generating a second image including the one or more detected elements; template matching by superimposing the generated first image and second image; and generating a drawing file for the subject file to be drafted by using the outer edge coordinate information and element coordinate information.
    Type: Grant
    Filed: January 8, 2016
    Date of Patent: May 28, 2019
    Inventors: Ju Sung Lee, Jong Sun Hong
  • Patent number: 10290101
    Abstract: A device to provide a heat map based medical image diagnostic mechanism is described. An image analysis application executed by the device receives a medical image from a medical image provider. A region of interest (ROI) is determined or provided by a user. A disease state score including a malignancy score is calculated for the ROI. Next, the ROI is partitioned into sub-regions. Impact values associated with the sub-regions are also determined. The impact values indicate the influence of a sub-region on the disease state score. Furthermore, annotations are determined based on pixel values associated with the sub-region. A heat map of the sub-regions is also generated based on the impact values. The heat map is labeled with the annotations. Next, the heat map is overlaid on the ROI. The medical image is provided with the heat map to the user.
    Type: Grant
    Filed: December 7, 2018
    Date of Patent: May 14, 2019
    Assignee: SONAVISTA, INC.
    Inventors: Christine I. Podilchuk, Richard Mammone
  • Patent number: 10289680
    Abstract: Systems and methods of natural language processing in an environment with no existing corpus are disclosed. The method includes defining an input grammar specific to a chosen domain, the input grammar having a domain specific knowledge and general grammatical knowledge. Groups of tokens are identified within the input grammar having syntactic and semantic equivalence. The identified groups are assembled into hypernyms, wherein the hypernyms include a semantic output for each token in the hypernyms. A list of fields is then combined with the hypernyms for combination with the hypernyms. A corpus of possible combinations of hypernyms and fields is created. A data structure mapping each possible combination to a partial semantic output is generated and the data structure is saved for use in later processing.
    Type: Grant
    Filed: May 31, 2016
    Date of Patent: May 14, 2019
    Assignee: Oath Inc.
    Inventors: Jonathan R. Scally, Richard Caneba, Nick Cassimatis
  • Patent number: 10289381
    Abstract: An electronic device includes an audio capture device, one or more physical sensors or context sensors, and one or more processors. A method of operating the electronic device includes receiving audio input, identifying a device command from the audio input, and detecting one or more multi-modal social cues from an environment about the electronic device. Where the one or more multi-modal social cues match one or more predefined criteria, the method executes a control operation in response to the device command.
    Type: Grant
    Filed: December 7, 2015
    Date of Patent: May 14, 2019
    Assignee: Motorola Mobility LLC
    Inventors: Kevin J McDunn, Balaji Thiagarajan
  • Patent number: 10282603
    Abstract: Embodiments of the present invention provide techniques for analyzing technical documents against known art. One or more machine learning classifiers are trained by a training set. References are ingested and assembled into a corpus. An input query is processed by natural language processing, and the classifiers are used to retrieve relevant documents and references, providing solutions to the input query. Anti-patterns are identified and provided to a user in the form of a report to alert the user to potential problems with the input query.
    Type: Grant
    Filed: January 9, 2017
    Date of Patent: May 7, 2019
    Assignee: International Business Machines Corporation
    Inventors: Michael Bender, Joseph Celi, Gordan G. Greenlee, David Jaramillo
  • Patent number: 10282153
    Abstract: A function executing device is configured to send a wireless identifier to a communication device via a first type of wireless communication interface for setting up a wireless connection. The function executing device is configured to determine whether the function executing device is in an error state and, if not, to establish the wireless connection via a second type of wireless communication interface. The function executing device operates in a parent state in the wireless connection. Additionally, print data may be received from the communication device. However, if the function executing device is in an error state the wireless connection is not established via the second type of wireless communication interface.
    Type: Grant
    Filed: August 30, 2017
    Date of Patent: May 7, 2019
    Assignee: Brother Kogyo Kabushiki Kaisha
    Inventors: Takanobu Suzuki, Hirotaka Asakura, Munehisa Matsuda, Satoshi Tanaka
  • Patent number: 10275691
    Abstract: An adaptive real-time detection and examination network that employs deep learning to detect and recognize objects in a stream of pixilated two-dimensional digital images. The network provides the images from an image source as pixilated image frames to a CNN having an input layer and output layer, where the CNN identifies and classifies the objects in the image. The network also provides metadata relating to the image source and its location, and provides the object classification data and the metadata to an RNN that identifies motion and relative velocity of the classified objects in the images. The network combines the object classification data from the CNN and the motion data from the RNN, and correlates the combined data to define boundary boxes around each of the classified objects and an indicator of relative velocity and direction of movement of the classified objects, which can be displayed on the display device.
    Type: Grant
    Filed: August 22, 2017
    Date of Patent: April 30, 2019
    Assignee: Northrop Grumman Systems Corporation
    Inventors: Victor Y. Wang, Kevin A. Calcote
  • Patent number: 10275865
    Abstract: There is provided an image processing device comprising: a display panel; and a controller configured to acquire a peak luminance value of a high dynamic range (HDR) image, and adjust a luminance of the high dynamic range image to be within an available output luminance range of the display panel, based on the acquired peak luminance value.
    Type: Grant
    Filed: October 14, 2016
    Date of Patent: April 30, 2019
    Assignee: LG ELECTRONICS INC.
    Inventor: Sangchurl Nam
  • Patent number: 10271908
    Abstract: An optical tracking system includes a reference marker part stationarily disposed relative to a first region; a sticker marker part attached in a sticker form to a second region capable of being rigidly registered with the first region; a shape measurement part measuring three-dimensional shapes of the first region and the second region; a tracking sensor part sensing the reference marker part and the shape measurement part; and a processing part acquiring a first coordinate transformation relationship among the reference marker part, the tracking sensor part, the shape measurement part, and the first region of the patent and a second coordinate transformation relationship among the reference marker part, the tracking sensor part, the shape measurement part, and the second region of the patient and tracking the first region by extracting a third coordinate transformation relationship between the first region and the second region.
    Type: Grant
    Filed: December 18, 2015
    Date of Patent: April 30, 2019
    Assignee: KOH YOUNG TECHNOLOGY INC.
    Inventor: Sang Yong Kim
  • Patent number: 10267626
    Abstract: A phase of a specific frequency component of an image of a grid image on the surface of an object taken by a camera is found, and the location of the surface of the object is found from the phase. As a result, measurement that is strong against noise can be carried out without projecting a grid having a brightness distribution of precise cosine waves. In addition, the process is simple and can be used for the surface of a moving object. The number of pixels used for the measurement is smaller than that in the sampling moire method.
    Type: Grant
    Filed: June 30, 2014
    Date of Patent: April 23, 2019
    Assignee: 4D SENSOR INC.
    Inventors: Yoshiharu Morimoto, Akihiro Masaya, Akifumi Takagi
  • Patent number: 10262224
    Abstract: A camera is connected to a trained neural network. The camera takes an image of a scene and transmits the image to the neural network. A processor connected to the neural network has a localization filter and a robot model implemented therein. A global positioning system (GPS) receiver and inertial measurement unit (IMU) transmit GPS information and IMU information, respectively, to the processor. The localization filter filters the received GPS and IMU information and inputs the filtered information into the robot model. The robot model outputs current position information corresponding to the current image and previous position information corresponding to the respective one or more previous images. The neural network uses the current image and associated current position information and the one or more previous images and respective associated previous position information to generate an estimated optical flow image, which is transmitted to an object detection system.
    Type: Grant
    Filed: July 19, 2017
    Date of Patent: April 16, 2019
    Assignee: The United States of America as represented by Secretary of the Navy
    Inventors: Joshua D. Harguess, Christopher M. Barngrover, Amin M. Rahimi