International Business Machines Corporation

Toshiba Global Commerce Solutions Holdings Corporation

Toshiba Global Commerce

Solutions Holdings Corporation

Yun Zhai

Rogerio Schmidt Feris

Russell Patrick Bobbitt

Sharathchandra UmapathiRao Pankanti

Arun Hampapur

Ankur Datta

Quanfu Fan

Akira Yanagawa

Lisa Maire Brown

Frederik C Kjeldsen

Zhe Yan

Zhuo Zhao

Xiao Long Chen

Na Lv

Chiao-Fe Shu

Lisa Marie Brown

Sachiko Miyazawa

Longbin Chen

Ira L Allen

Balamanohar Paluri

Ying-Li Tian

Frederik C M Kjeldsen

Florian Pinel

Andrew William Senior

Myron Dale Flickner

Sharathchandra Umapathirao Pankanti

Charles A Otto

Lei Ding

Unsang Park

Chetan Mishra

Peter Alan Coldicott

Christopher Scott Milite

Zuoxuan Lu

Bruce Henry Hyre

Curtis Howard Brobst

Stephen R Russo

Behjat Siddiquie

Raymond A Cooke

Ying-Ii Tian

A method includes activating a first object detector and a second object detector. The first object detector is operable to detect a first type of object in a set of image frames, and the second object detector is operable to detect a second type of object in the set of image frames. The method also includes activating a first set of attribute classifiers and a second set of attribute classifiers. The method also includes generating first statistics for the first object detector and second statistics for the second object detector. The method also includes modifying activation of at least the second set of attribute classifiers based on a comparison of the first statistics and the second statistics.

Attribute classifiers for image classification

Using a distance sensor, a set of distances to a target object is measured, a distance in the set being measured at a time. From the set of distances and corresponding times, a trajectory of the target object is predicted. Based on receiving a signal identifying a friendly device, a correlation of the trajectory with a trajectory of the friendly device is determined. Responsive to the trajectory being predicted to intercept a defined volume around the distance sensor and responsive to the correlation being above a threshold degree of correlation, the target object is classified as a threat. Responsive to the trajectory being predicted to intercept a defined volume around the distance sensor and responsive to the correlation being below a threshold degree of correlation, a threat level corresponding to the target object is identified based on the trajectory of the target object and trajectory of the friendly device.

Trajectory based threat alerting with friendly device augmentation

A method, a computer program product, and a computer system for determining a location using street view images. A mobile device obtains an image captured by the mobile device, obtains a set of street view images, and obtains a first graph of the image captured by the mobile device and a plurality of second graphs of the set of the street view images. The first graph includes nodes representing interest points in the image captured by the mobile device, and the plurality of the second graphs includes nodes representing interest points in the set of the street view images. The mobile device determines a location of the mobile device, based on relevance between the first graph and the plurality of the second graphs.

Location determination using street view images

Techniques for dataset processing are provided. A plurality of feature vectors is generated for a plurality of pixel regions from a plurality of images, and a plurality of clusters is generated based on the plurality of feature vectors. A score is assigned to each respective pixel region in the plurality of pixel regions based at least in part on a cluster the respective pixel region is associated with. A unified dataset is created by, for each respective cluster in the plurality of clusters, computing a representative index for each pixel region in the respective cluster by comparing the pixel regions in the cluster, and modifying the score of the pixel regions in the cluster based on the computed representative indices. A confidence index is generated for the unified dataset based at least in part on a level of fragmentation in the unified dataset.

Automated and unsupervised curation of image datasets

Datasets containing a plurality of images are processed. A plurality of regions within the images are determined, where each of the plurality of regions corresponds to a facial region. A feature vector is generated for each of the plurality of pixel regions, and the plurality of pixel regions are clustered into a plurality of clusters based on the generated feature vectors. An initial score is assigned to each respective pixel region in the plurality of pixel regions, and the plurality of pixel regions are sorted based on the assigned scores. A representative index is computed for each respective pixel region by comparing each respective pixel regions with each other pixel region in the respective cluster. The score of each pixel region is modified based on the computed representative indices, and the pixel regions are sorted based on the modified scores. A confidence index is generated for the unified dataset.

Aspects determining anomalous events, wherein processors determine a trajectory of tracked movement of an object through an image field of a camera partitioned into a matrix grid of different local units. The aspects generate anomaly confidence decision values for image features extracted from video data of the tracked movement of the object as a function of fitting extracted image features to normal patterns of local motion pattern models defined by dominant distributions of extracted image features. The aspects further extract trajectory features from the video data relative to the trajectory of the tracked movement of the object, and generate global anomaly confidence decision values for the object trajectory as a function of fitting the extracted trajectory features to a normal learned motion trajectory model. The aspects determine anomalous events as a function of the generated global anomaly confidence decision value and the anomaly confidence decision values.

Anomalous event retriever

Methods and system are provided for monitoring a queue. A method includes receiving by sensors a non-visual identifier at predefined locations of a queue. Further, the method includes capturing by image capture devices images of an object possessing the non-visual identifier at the predefined locations of the queue. Further, the method includes visually tracking another object in the queue based on transformations of a predefined feature extracted from the images of the object possessing the non-visual identifier at the predefined locations.

Visual monitoring of queues using auxillary devices

Techniques are described for video redaction. In one aspect, techniques include receiving a video for redaction; analyzing the video to generate an appearance model for the video, providing a user interface (UI) allowing a user to modify the appearance model, and responsive to a user selecting an object from the appearance model, extending and completing a trajectory of the selected object with enhanced marking based on the user input.

Object-centric video redaction

Machine logic that pre-processes and post-processes images for visual object detection by performing the following steps: receiving a set of image(s); filtering the set of image(s) using a set of multimodal integral filter(s), thereby removing at least a portion of the set of image(s) and resulting in a filtered set of image(s); performing object detection on the filtered set of image(s) to generate a set of object-detected image(s); assembling a first plurality of object-detected image(s) from the set of object-detected image(s); and upon assembling the first plurality of object-detected image(s), performing non-maximum suppression on the assembled first plurality of object-detected image(s).

Filtering methods for visual object detection

Embodiments are directed to an object detection system having at least one processor circuit configured to receive a series of image regions and apply to each image region in the series a detector, which is configured to determine a presence of a predetermined object in the image region. The object detection system performs a method of selecting and applying the detector from among a plurality of foreground detectors and a plurality of background detectors in a repeated pattern that includes sequentially selecting a selected one of the plurality of foreground detectors; sequentially applying the selected one of the plurality of foreground detectors to one of the series of image regions until all of the plurality of foreground detectors have been applied; selecting a selected one of the plurality of background detectors; and applying the selected one of the plurality of background detectors to one of the series of image regions.

Self-optimized object detection using online detector selection

A method to automatically calibrate a static camera in a vehicle is provided. The method may include receiving a captured image file. The method may also include detecting a plurality of vehicles in the captured image file. The method may further include determining a 2D projected shape, size, location, direction of travel, and a plurality of features for each vehicle at various locations in the captured image file. The method may additionally include inferring a plurality of geometry scenes associated with the captured image file, whereby the plurality of geometry scenes is inferred based on the determined 2D projected shape, size, location, direction of travel, and the plurality of features of each vehicle within the detected plurality of vehicles as projected onto the captured image file. The method may include calibrating the static camera based on the inferred plurality of geometry scenes associated with the captured image file.

Self-calibration of a static camera from vehicle information

A system for detecting a change in object positioning by processing images by detecting a first entity tagged with a first visually unique identifier, and detecting a second entity tagged with a second visually unique identifier distinguishable from the first visually unique identifier. The system receives a first image, from a first image capturing device, containing the first visually unique identifier and the second visually unique identifier. The system analyzes the first image to determine a distance between the first visually unique identifier and the second visually unique identifier. The system receives a second image containing the first visually unique identifier. The system analyzes the second image to determine a location of the second visually unique identifier relative to the first visually unique identifier to form a distance assessment. Based on the distance assessment, the system determines a change in proximity between the first entity and the second entity.

Low-cost method to reliably determine relative object position

Growing community of inventors

Pound Ridge, NY, United States of America

Yun Zhai