Abstract: A system and method of generating trackable frames from a broadcast video feed are provided herein. A computing system retrieves a broadcast video feed for a sporting event. The broadcast video feed includes a plurality of video frames. The computing system generates a set of frames for classification using a principal component analysis model. The set of frames are a subset of the plurality of video frames. The computing system partitions each frame of the set of frames into a plurality of clusters. The computing system classifies each frame of the plurality of frames as trackable or untrackable. Trackable frames capture a unified view of the sporting event. The computing system compares each cluster to a predetermined threshold to determine whether each cluster comprises at least a threshold number of trackable frames. The computing system classifies each cluster that includes at least the threshold number of trackable frames as trackable.

Abstract: A computer-implemented method for attention generation is provided. In this method, a plurality of image frames can be obtained from a video stream. An original attention for a first image frame of the plurality of image frames can be generated. Then, at least one interested area can be identified in the first image frame. A local attention for each of the at least one interested area can be generated. Moreover, a total attention for the first image frame can be generated based on the original attention of the first image frame and the local attention of each of the at least one interested area.

Abstract: A system including circuitry configured to determine a reflectance feature of a liquid based on reflectance image data generated based on multispectral image data of the liquid; determine a structural feature of the liquid based on image data of the liquid; and to provide quality information of the liquid based on the reflectance feature and the structural feature.

Abstract: The present disclosure provides peanut maturity grading systems and methods for quickly, efficiently, and objectively determining a peanut maturity grade for a crop of peanuts and determining an optimal harvest time for the crop. Embodiments of systems and methods of the present disclosure can be performed in the field or field-side and do not require assistance of a trained peanut grading specialist.

Abstract: In an apparatus for assisting a driver of a vehicle in driving the vehicle, a road sign recognizer is configured to recognize a road sign from an image captured by a vehicle-mounted imager. A turn detector is configured to detect a turning action of the vehicle. A notifier is configured to notify the driver of the vehicle of road sign information representing the road sign recognized by the road sign recognizer. The notifier is further configured to, if the turning action of the vehicle is detected by the turn detector, withhold notifying the driver of the vehicle of the road sign information representing the road sign recognized by the road sign recognizer within a predetermined period of time during the turning action.

Abstract: A method for aligning a wafer image with a reference image, comprising: searching for a targeted reference position on the wafer image for aligning the wafer image with the reference image; and in response to a determination that the targeted reference position does not exist: defining a current lock position and an area that encloses the current lock position on the wafer image; computing an alignment score of the current lock position; comparing the alignment score of the current lock position with stored alignment scores of positions previously selected in relation to aligning the wafer image with the reference image; and aligning the wafer image with the reference image based on the comparison.

Abstract: A system for detecting the face of a vehicle driver comprises an instrument cluster for the driver including a carrier and an dashboard directed towards the driver and covered by a glass, said dashboard being embedded in the carrier; an image capture device comprising a camera configured to capture sequential images of the driver's gaze; and a controller configured to analyze the driver's gaze; at least one intrinsic element of the instrument cluster defines a reflecting surface configured to directly reflect the sequential images of the driver's gaze such as a glossy plastic, a chrome-plated element, a display screen or the dashboard glass; the camera comprises a field of view oriented towards the reflecting surface so as to indirectly capture the images of the driver's gaze.

Abstract: Example aspects of the present disclosure are directed to systems and methods for learning data augmentation strategies for improved object detection model performance. In particular, example aspects of the present disclosure are directed to iterative reinforcement learning approaches in which, at each of a plurality of iterations, a controller model selects a series of one or more augmentation operations to be applied to training images to generate augmented images. For example, the controller model can select the augmentation operations from a defined search space of available operations which can, for example, include operations that augment the training image without modification of the locations of a target object and corresponding bounding shape within the image and/or operations that do modify the locations of the target object and bounding shape within the training image.

Abstract: Methods, systems, and techniques for enhancing use of two-dimensional (2D) video analytics by using depth data. Two-dimensional image data representing an image comprising a first object is obtained, as well as depth data of a portion of the image that includes the first object. The depth data indicates a depth of the first object. An initial 2D classification of the portion of the image is generated using the 2D image data without using the depth data. The initial 2D classification is stored as an approved 2D classification when the initial 2D classification is determined consistent with the depth data. Additionally or alternatively, a confidence level of the initial 2D classification may be adjusted depending on whether the initial 2D classification is determined to be consistent with the depth data, or the depth data may be used with the 2D image data for classification.

Abstract: A method for template matching can include iteratively selecting a template set of points to project over a centerline of a candidate symbol; conducting a template matching analysis; assigning a score to each template set; and selecting a template set with a highest assigned score. For example, the score can depend on proximity of the template points to a center and/or boundaries of a principal tracing path of the symbol. Additionally, one or more template sets having a top rank can be selected for a secondary analysis of proximity of the template points to a boundary of a printing of the symbol. The method can further include using the template with the highest score to interpret the candidate symbol.

Abstract: An aspect of the invention includes reading a scale in image data representing an image of physical characteristics and resizing at least a portion of the image data to align with target image data representing a target image based at least in part on the scale to form resized image data representing one or more resized images. Noise reduction is applied to the resized image data to produce test image data representing one or more test images. A best fit analysis is performed on the test image data with respect to the target image data. Test image data having the best fit are stored with training image data representing classification training images indicative of one or more recognized features. An anomaly in unclassified image data representing an unclassified image is identified based at least in part on an anomaly detector as trained using the classification training images.

Abstract: In an approach, data of a plurality of points is sampled in a target area, wherein the data of each point of the plurality of points comprises position information and a height value. A first area of a target area is determined, wherein the height value of each point of the plurality of points in the first area complies with a first range. A second area of the target area is determined, wherein the height value of each point of the plurality of points in the second area complies with a second range. A third area of the target area is determined, wherein the height value of each point of the plurality of points in the third area complies with a third range. A first pattern is generated, wherein the first pattern is a combination of the first area and the third area.

Abstract: In one embodiment, a set of feature vectors can be derived from any biometric data, and then using a deep neural network (“DNN”) on those one-way homomorphic encryptions (i.e., each biometrics' feature vector) an authentication system can determine matches or execute searches on encrypted data. Each biometrics' feature vector can then be stored and/or used in conjunction with respective classifications, for use in subsequent comparisons without fear of compromising the original biometric data. In various embodiments, the original biometric data is discarded responsive to generating the encrypted values. In another embodiment, the homomorphic encryption enables computations and comparisons on cypher text without decryption of the encrypted feature vectors. Security of such privacy enable biometrics can be increased by implementing an assurance factor (e.g., liveness) to establish a submitted biometric has not been spoofed or faked.

Abstract: Provided is a method for blending of agricultural product utilizing hyperspectral imaging. At least one region along a sample of agricultural product is scanned using at least one light source of different wavelengths. Hyperspectral images are generated from the at least one region. A spectral fingerprint for the sample of agricultural product is formed from the hyperspectral images. A plurality of samples of agricultural product is blended based on the spectral fingerprints of the samples according to parameters determined by executing a blending algorithm.

Abstract: A system and a method are disclosed that allow for generation of a model or reconstruction of a model of a subject based upon acquired image data. The image data can be acquired in a substantially mobile system that can be moved relative to a subject to allow for image acquisition from a plurality of orientations relative to the subject. The plurality of orientations can include a first and final orientation and a predetermined path along which an image data collector or detector can move to acquire an appropriate image data set to allow for the model of construction.

Abstract: An image presentation method includes obtaining a first image captured by a first image sensor and a second image captured by a second image sensor; size-processing the first image according to at least one of a target resolution, an aspect ratio of the first image, or an aspect ratio of the second image to generate a size-processed first image having the target resolution; and generating a presenting image at least by combining the size-processed first image and the second image. The presenting image has a preset resolution that is not less than a sum of the target resolution and a resolution of the second image.

Abstract: Systems and methods which provide iterative multi-directional image searching are described. Embodiments utilize a multi-directional searching pattern for defining one or more searching areas within a source image. A searching area of embodiments provides a searching area within a source image for searching for a template image in the multiple directions of the multi-directional searching pattern. A location of the searching area within the source image may be updated iteratively, such as based upon motion vectors derived from the searching, until a template image match position is identified within the source image. Embodiments transform the template image and the searching area within a source image to 1D representations corresponding to each direction of the multi-directional image search. Embodiments accommodate rotation and scale variance of the subject (e.g., object of interest) of the template image within the source image.

Abstract: In one embodiment, a set of feature vectors can be derived from any biometric data, and then using a deep neural network (“DNN”) on those one-way homomorphic encryptions (i.e., each biometrics' feature vector) an authentication system can determine matches or execute searches on encrypted data. Each biometrics' feature vector can then be stored and/or used in conjunction with respective classifications, for use in subsequent comparisons without fear of compromising the original biometric data. In various embodiments, the original biometric data is discarded responsive to generating the encrypted values. In another embodiment, the homomorphic encryption enables computations and comparisons on cypher text without decryption of the encrypted feature vectors. Security of such privacy enable biometrics can be increased by implementing an assurance factor (e.g., liveness) to establish a submitted biometric has not been spoofed or faked.

Abstract: Feedback data useful in prosthodontic procedures associated with the intra oral cavity is provided. First, a 3D numerical model of the target zone in the intra oral cavity is provided, and this is manipulated so as to extract particular data that may be useful in a particular procedure, for example data relating to the finish line or to the shape and size of a preparation. The relationship between this data and the procedure is then determined, for example the clearance between the preparation and the intended crown. Feedback data, indicative of this relationship, is then generated, for example whether the preparation geometry is adequate for the particular type of prosthesis.

Abstract: Detecting the start of a credit roll within video program may allow for the automatic extension of video recordings among other functions. The start of the credit roll may be detected by determining the number of text blocks within a sequence of frames and identifying a point in the sequence of frames where a difference between the number of text blocks in frames occurring before the point and the number of text blocks in frames occurring after the point is greatest and exceeds a specified threshold. Text blocks may be identified within each frame by partitioning the frame into one or more segments and recording the segments having a pixel of a sufficiently high contrast. Contiguous segments may be merged or combined into single blocks, which may then be filtered to remove noise and false positives. Additional content may be inserted into the credit roll frames.