Abstract: The present disclosure uses automated, computer-driven analysis of video feeds to extract raw motion metadata, identify moving objects and assess whether or not the moving objects are compliant with certain predefined rules. The motion can be visually plotted on a background in the form of motion trails, and a chronological plot of the motion may be provided as a menu to find anomalies through visual screening. The motion metadata may be filtered through annotations. A highlighted portion of the video feed relating to a specific topic of interest may be replayed.

Abstract: A method, apparatus and computer program product are provided for warping a perspective image into the ground plane using a homography transformation to estimate a bird's eye view in real time. Methods may include: receiving first sensor data from a first vehicle traveling along a road segment in an environment, where the first sensor data includes perspective image data of the environment, and where the first sensor data includes a location and a heading; retrieving a satellite image associated with the location and heading; applying a deep neural network to regress a bird's eye view image from the perspective image data; applying a Generative Adversarial Network (GAN) to the regressed bird's eye view image using the satellite image as a target of the GAN to obtain a stabilized bird's eye view image; and deriving values of a homography matrix between the sensor data and the established bird's eye view image.

Abstract: A sensor calibration system configured to receive a first frame of one or more markers on a repositionable platform at a first location within a space from a sensor. The system is further configured to determine pixel locations in the first frame for a first marker and a second marker from among the one or more markers. The system is further configured to receive distance information that corresponds with a distance between the platform and distance measuring devices. The system is further configured to determine (x,y) coordinates for the first marker and the second marker based on the distance information. The system is further configured to generate a homography based on the (x,y) coordinates and pixel locations of the first marker and the second marker. The homography includes coefficients that translate between pixel locations in the first frame of the sensor and (x,y) coordinates in the global plane.

Abstract: A system is configured to receive an indication that an apparatus in a first assembled state should comprise a component with a first digital fingerprint and a component with a second digital fingerprint. The system is configured to receive video footage of apparatuses in the first assembled state. The system is configured to isolate an image of an apparatus in the first assembled state. The system is configured to split the image into a frame comprising a first component and a frame comprising a second component. The system is configured to generate first and second filtered images. The system is configured to identify feature points in the filtered images. The system is configured to determine that the first set of feature points matches the first digital fingerprint and that the second set of feature points matches the second digital fingerprint. The system is configured to update a component database.

Abstract: A method for manufacturing a semiconductor device includes designing a layout, manufacturing a photomask based on the designed layout, and performing a photolithography process using the photomask to form a pattern on a substrate. The manufacturing of the photomask includes preparing the photomask including first and second chip regions, extracting first and second images from the first and second chip regions, respectively, averaging the first and second images to generate a preliminary standard image including a difference region between the first and second images, inserting a normal image into the difference region based on the layout to generate a standard image, and comparing each of the first and second images with the standard image to detect a defect of the first and/or second chip regions.

Abstract: An electronic device may include a processor configured to: obtain a plurality of images, the plurality of images including a first image captured by a first camera or image sensor, and a second image captured by a second camera or image sensor. The processor may further be configured to: determine whether a first object and a second object respectively included in the first image and the second image are a same object, and based on determining that the first object and the second object are the same object, train a neural network data recognition model to recognize the first object and the second object as the same object in both the first image and the second image by using the first image and the second image as training data of the neural network data recognition model.

Abstract: A method and system of selecting a surface finish for a computer designed component associated with an anatomical area of a specific patient. A computer designed component is created. The component includes at least one surface. An interface of finishing software displays an image of the component and the at least one surface. A plurality of finishes available for the at least one surface is displayed. A selection of one of the plurality of finishes for the at least one surface is input by a user. The component data and finishing data may then be sent to a computer aided manufacturing system to manufacture the computer designed component.

Abstract: Systems and methods for multimodal classification include a plurality of expert modules, each expert module configured to receive data corresponding to one of a plurality of input modalities and extract associated features, a plurality of class prediction modules, each class prediction module configured to receive extracted features from a corresponding one of the expert modules and predict an associated class, a gate expert configured to receive the extracted features from the plurality of expert modules and output a set of weights for the input modalities, and a fusion module configured to generate a weighted prediction based on the class predictions and the set of weights. Various embodiments include one or more of an image expert, a video expert, an audio expert, class prediction modules, a gate expert, and a co-learning framework.

Abstract: A computing system may train an autoencoder to generate a first set of codes from a first set of thermal video images of activities of a user in an environment. The activities may represent routine behaviors of the user in the environment. The computing system may use an unsupervised machine-learning algorithm to categorize the first set of codes into a set of clusters. The computing system may use the autoencoder to determine a code representative of a second set of thermal video images of an activity in the environment. Based on the code not being associated with any cluster in the set of clusters, the computing system may determine that the code is an anomalous code. The computing system may perform an alert action based on the anomalous code.

Abstract: A method of generating a template image includes: receiving an input from a user representing identifications of an object in different respective slices of a volumetric image; using the input to determine a volume-of-interest (VOI) that includes voxels in a subset of the volumetric image; and determining the template image using at least some of the voxels in the VOI, wherein the act of determining the template image comprises performing a forward projection of the at least some of the voxels in the VOI using a processor. An image processing method includes: obtaining a volumetric image; performing forward projection of voxels in the volumetric image from different positions onto a first plane using a processor; and summing projections on the first plane resulted from the forward projection from the different positions to create a first image slice in the first plane.

Abstract: A subject support (14) is configured to dock with a medical imaging device (50) with a fixed spatial relationship between the docked subject support and the medical imaging device. A patient positioning device includes a range camera (10) that acquires a two-dimensional (2D) range image of a human imaging subject (12) disposed on a subject support (14). The range image has pixel values corresponding to distances from the range camera. An electronic processor (16) is programmed to perform a positioning method to determine a reference point on or in the human subject in a frame of reference (FS) of the subject support from the 2D range image. This reference point is translated to a frame of reference (FD) of the imaging device based on a priori known spatial relationship of the medical imaging device and the docked subject support. Using 3D models, the loading process may be simulated.

Abstract: Methods and systems are described. A method includes accessing training data samples that includes a plurality of transaction laundering associated features and generating random samples, training a first plurality of different models to identify transaction laundering merchants based on random samples generated from each of the training data samples, training a second plurality of different models to identify the transaction laundering merchants based on the training data, generating a transaction laundering classification for a merchant from each of the first plurality of models and each of the second plurality of models, generating a first model group classification based on a first majority vote on transaction laundering classifications from the first plurality of models, and generating a second model group classification based on a second majority vote on transaction laundering classifications from the second plurality of models. A likelihood that the merchant is a transaction launderer is determined.

Abstract: A drug inspection apparatus inspects drugs that are prepared based on prescription information and are packaged in a prescription bag. A drug database includes drug images of drugs that can be prepared. A comparison target selection section acquires drug images of drugs, which are prepared according to the prescription, and drugs similar thereto from the drug database. A first drug determination section compares captured images of prepared drugs with the drug images acquired from the drug database, and determines drugs present in the captured images and the number thereof. An inspection result determination section determines whether or not the prepared drugs and the number thereof match the prescription information based on the prescription information.

Abstract: A method of detecting a flame in an image comprising the step of acquiring a series of infrared images from an infrared imager. Image flow vectors are calculated for each of the series of infrared images and stored in a corresponding 2D vector array. At least one 3D image flow data set is generated from the corresponding 2D vector arrays. A 2D pixel intensity array is generated for each of the series of infrared images. A 3D pixel intensity data set is generated from the 2D pixel intensity arrays. A Fourier Transform and infrared metrics are calculated. The infrared metrics are compared to an infrared threshold criteria. A corresponding 2D threshold array for each of the series of infrared images is generated. A 3D threshold data set is generated from the 2D threshold arrays. A flame status is determined based on the 3D threshold data set.

Abstract: Disclosed are a background filtering method and a system based on a roadside solid-state lidar, whereby a spatial dividing method of solid-state lidar based on channel scanning angle is designed, single-channel point cloud is extracted, point cloud of each channel of all background data frames are fused and superimposed, each background space unit and non-background space unit based on channel scanning space are identified based on a given threshold; then real-time data frames are subjected to background difference based on channels to obtain the single-channel road user information data frames; the background filtering is achieved by splicing single-channel road user information data frames into a complete road user information data frame.

Abstract: Fill techniques as implemented by a computing device are described to perform hole filling of a digital image. In one example, deeply learned features of a digital image using machine learning are used by a computing device as a basis to search a digital image repository to locate the guidance digital image. Once located, machine learning techniques are then used to align the guidance digital image with the hole to be filled in the digital image. Once aligned, the guidance digital image is then used to guide generation of fill for the hole in the digital image. Machine learning techniques are used to determine which parts of the guidance digital image are to be blended to fill the hole in the digital image and which parts of the hole are to receive new content that is synthesized by the computing device.

Abstract: A method for detecting a human skeleton is provided. The method includes: receiving a video frame, wherein the video frame comprises a human body; determining whether the video frame comprises prediction information; determining whether a first intra-coded macroblock (IMB) ratio of a target area comprising the human body in the video frame is greater than a first threshold when the video frame comprises the prediction information; and using a motion vector (MV) to estimate skeleton information of the human body when the first IMB ratio of the target area is not greater than the first threshold.

Abstract: Vehicle parking monitoring systems and methods are disclosed herein. An example method can include receiving images from a camera of a parking spot, each of the images being time stamped, determining presence of a vehicle in the images, placing a bounding area around a region of interest of the vehicle, the region of interest including no personally identifiable information, retaining the bounding area and discarding a remainder of the images, and determining when the vehicle is no longer present based on a change in the bounding area of the images.

Abstract: The present application is applicable to the technical field of image processing, and provides a target object tracking method and an apparatus, and a terminal device. The target object tracking method includes: obtaining an image sequence including a target object, wherein the image sequence includes a plurality of image; analyzing the plurality of image, and determining a motion vector and a motion vector residual of the target object in a current image; and inputting the motion vector, the motion vector residual, and a known target region of the target object in a previous image into a first target detection network, and determining a predicted target region of the target object in the current image. The present application can reduce the calculation amount and thus improve the operation speed of target object tracking.

Abstract: The prediction control part 204 predicts a three-dimensional position and a three-dimensional velocity of the world coordinate system of the flying ball at a specific time after the initial time as a predicted position and a predicted velocity based on initial position and initial velocity of flying ball, and an equation indicating a parabolic shape of the flying ball. The conversion control part 205 converts the predicted position into a two-dimensional position of a camera coordinate system as a temporary position. The acquisition control part 206 specifies a flying ball image and acquires a two-dimensional position of the camera coordinate system of the flying ball image as an observation position. The correction control part 207 corrects the predicted position and the predicted velocity as a corrected position and a corrected velocity based on the predicted position and the predicted velocity, the observation position, and a Kalman filter.