Abstract: Automated analysis of video data for determination of human behavior includes segmenting a video stream into a plurality of discrete individual frame image primitives which are combined into a visual event that may encompass an activity of concern as a function of a hypothesis. The visual event is optimized by setting a binary variable to true or false as a function of one or more constraints. The visual event is processed in view of associated non-video transaction data and the binary variable by associating the visual event with a logged transaction if associable, issuing an alert if the binary variable is true and the visual event is not associable with the logged transaction, and dropping the visual event if the binary variable is false and the visual event is not associable.

Abstract: Alerts to object behaviors are prioritized for adjudication as a function of relative values of abandonment, foregroundness and staticness attributes. The attributes are determined from feature data extracted from video frame image data. The abandonment attribute indicates a level of likelihood of abandonment of an object. The foregroundness attribute quantifies a level of separation of foreground image data of the object from a background model of the image scene. The staticness attribute quantifies a level of stability of dimensions of a bounding box of the object over time. Alerts are also prioritized according to an importance or relevance value that is learned and generated from the relative abandonment, foregroundness and staticness attribute strengths.

Abstract: A method for summarizing image content from video images received from a moving camera includes detecting foreground objects in the images, determining moving objects of interest from the foreground objects, tracking the moving objects, rating movements of the tracked objects, and generating a list of highly rated segments within the video images based on the ratings.

Abstract: Automated analysis of video data for determination of human behavior includes segmenting a video stream into a plurality of discrete individual frame image primitives which are combined into a visual event that may encompass an activity of concern as a function of a hypothesis. The visual event is optimized by setting a binary variable to true or false as a function of one or more constraints. The visual event is processed in view of associated non-video transaction data and the binary variable by associating the visual event with a logged transaction if associable, issuing an alert if the binary variable is true and the visual event is not associable with the logged transaction, and dropping the visual event if the binary variable is false and the visual event is not associable.

Abstract: Human behavior is determined by sequential event detection by constructing a temporal-event graph with vertices representing primitive images of images of a video stream, and also of idle states associated with the respective primitive images. A human activity event is determined as a function of a shortest distance path of the temporal-event graph vertices.

Abstract: A foreground object blob having a bounding box detected in frame image data is classified by a finite state machine as a background, moving foreground, or temporally static object, namely as the temporally static object when the detected bounding box is distinguished from a background model of a scene image of the video data input and remains static in the scene image for a threshold period. The bounding box is tracked through matching masks in subsequent frame data of the video data input, and the object sub-classified within a visible sub-state, an occluded sub-state, or another sub-state that is not visible and not occluded as a function of a static value ratio. The ratio is a number of pixels determined to be static by tracking in a foreground region of the background model corresponding to the tracked object bounding box over a total number of pixels of the foreground region.

Abstract: Techniques for detecting one or more events are provided. The techniques include using multiple overlapping regions of interest on a video sequence to cover a location for one or more events, wherein each event is associated with at least one of the multiple overlapping regions of interest, applying multiple-instance learning to the video sequence to select one or more of the multiple overlapping regions of interest to construct one or more location-aware event models, and applying the models to the video sequence to detect the one or more events and to determine the one or more regions of interest that are associated with the one or more events.

Abstract: An approach that automatically distinguishes between in-store customers and in-store employees is provided. In one embodiment, there is a learning tool configured to construct a model for an in-store employee; a matching tool configured to match attributes between a particular person and the constructed models for an in-store employee; and a classifying tool configured to classify persons into categories of employees and customers based on amount of matching attributes between a particular person and the model for an in-store employee.

Abstract: Events are classified through string pattern recognition. Text labels are assigned to image primitives in a time-ordered set of training images and to related time-ordered transactions in an associated training transaction log in a combined time-ordered training string of text labels as a function of image types. Transactions are labeled in a training transaction log with a transaction label, a training primitive image of a start of a transaction with a start image text label, a training primitive of an entry of a transaction into the log with an entry image text label, and a training primitive of a conclusion of a transaction with an ending image text label. Positive subset string patterns are discovered representing true events from the combined time-ordered training string of text labels, and negative subset string patterns defined by removing single transaction primitive labels from the positive subset string patterns.

Abstract: Visual content in images captured from a scene by a camera in each of a plurality of different pose settings are analyzed to determine predicted occurrences of a transaction associated with the visual content in each pose, which are compared with actual transaction occurrence data to generate performance values for each pose as a function difference between the predicted and actual transactions. Optimized poses are chosen having the best performance value, wherein a camera controller may place the camera in the optimum pose for use in monitoring the scene and generating the primitives of interest associated with the transactions.

Abstract: A computer identifies a proto-object in a digital image using a background subtraction method, the proto-object being associated with a lighting artifact in the surveillance region. The background subtraction method preserves boundary details and interior texture details of proto-objects associated with lighting artifacts. A plurality of characteristics of the proto-object digital data are determined, the characteristics, individually or in combination, distinguish a proto-object related to a lighting artifact from its background. A learning machine, trained with the plurality of characteristics of proto-objects classified as either foreground or not foreground, determines a likelihood that the plurality of characteristics is associated with a foreground object.

Abstract: Video analytics data is audited through review of selective subsets of visual images from a visual image stream as a function of a temporal relationship of the images to a triggering alert event. The subset comprehends an image contemporaneous with the triggering alert event and one or more other images occurring before or after the contemporaneous image. The generated subset may be presented for review to determine whether the triggering alert event is a true or false alert, or whether additional data from the visual image stream is required to make such a determination. If determined from the presented visual essence that the additional data is required make the true or false determination, then additional data is presented from the visual image stream for review.

Abstract: Techniques for detecting one or more events are provided. The techniques include using multiple overlapping regions of interest on a video sequence to cover a location for one or more events, wherein each event is associated with at least one of the multiple overlapping regions of interest, applying multiple-instance learning to the video sequence to select one or more of the multiple overlapping regions of interest to construct one or more location-aware event models, and applying the models to the video sequence to detect the one or more events and to determine the one or more regions of interest that are associated with the one or more events.

Abstract: Transaction units of video data and transaction data captured from different checkout lanes are prioritized as a function of lane priority values of respective ones of the different checkout lanes from which the transaction units are acquired. Each of the checkout lanes has a different lane priority value. The individual transaction units are processed in the prioritized processing order to automatically detect irregular activities indicated by the transaction unit video and the transaction data of the processed individual transaction units.

Abstract: Human behavior is determined by sequential event detection by constructing a temporal-event graph with vertices representing primitive images of images of a video stream, and also of idle states associated with the respective primitive images. A human activity event is determined as a function of a shortest distance path of the temporal-event graph vertices.

Abstract: Techniques for creating a training technique for an individual are provided. The techniques include obtaining video of one or more events and information from a transaction log that corresponds to the one or more events, wherein the one or more events relate to one or more actions of an individual, classifying the one or more events into one or more event categories, comparing the one or more classified events with an enterprise best practices model to determine a degree of compliance, examining the one or more classified events to correct one or more misclassifications, if any, and revise the one or more event categories with the one or more corrected misclassifications, if any, and using the degree of compliance to create a training technique for the individual.

Abstract: Human behavior alerts are determined from a video stream through application of video analytics that parse a video stream into a plurality of segments, wherein each of the segments are either temporally related to at least one of a plurality of temporally distinct transactions in an event data log; or they are each associated with a pseudo transaction marker if not temporally related to at least one of the temporally distinct transactions and an image analysis indicates a temporal correlation with at least one of the distinct transactions is expected. Visual image features are extracted from the segments and one-SVM classification is performed on the extracted features to categorize segments into inliers or outliers relative to a threshold boundary. Event of concern alerts are issued with respect to the inlier segments associated with the associated pseudo transaction marker.

Abstract: Video analytics data is audited through review of selective visual essence subsets of visual images from a visual image stream as a function of a temporal relationship of the essence subset images to a triggering alert event. The visual essence subset comprehends an image contemporaneous with the triggering alert event and one or more other images occurring before or after the contemporaneous image. The generated visual essence is presented for review to determine whether the triggering alert event is a true or false alert, or whether additional data from the visual image stream is required to make such a determination. If determined from the presented visual essence that the additional data is required make the true or false determination, then additional data is presented from the visual image stream for review.

Abstract: Determination of human behavior from an alignment of data streams includes acquiring visual image primitives from a video input comprising visual information relevant to a human activity. The primitives are temporally aligned to an optimally hypothesized sequence of primitives transformed from a sequence of transactions as a function of a distance metric between the observed primitive sequence and the transformed primitive sequence. More particularly, transforming includes comparing the distance metric costs and choosing and performing the lowest cost of temporally matching the observed primitives to one or more transactions, deleting a primitive, or associating a primitive with a pseudo transaction marker. Accordingly, alerts are issued based on analysis of the transformation of primitives.

Abstract: Automated analysis of video data for determination of human behavior includes segmenting a video stream into a plurality of discrete individual frame image primitives which are combined into a visual event that may encompass an activity of concern as a function of a hypothesis. The visual event is optimized by setting a binary variable to true or false as a function of one or more constraints. The visual event is processed in view of associated non-video transaction data and the binary variable by associating the visual event with a logged transaction if associable, issuing an alert if the binary variable is true and the visual event is not associable with the logged transaction, and dropping the visual event if the binary variable is false and the visual event is not associable.