Sensormatic Electronics Corporation

Johnson Controls Tyco IP Holdings LLP

Michael C Stewart

Gopi Subramanian

Joseph Celi, Jr

Abdul Razak

Michelangelo Palella

Jeffrey R Rendlen

Bryan W Martin

Bernice Stewart

Karthik Jayaraman

Natasha Mishell Astudillo

Harsh Gaurangbhai Motka

Harish Yadav

William Dubois

A system may be configured to generate real-time analytics using machine learning and computer vision. In some aspects, the system may receive a storage structure video frame and an interaction video frame, determine a storage structure inference based on the storage structure video frame, and determine an interaction inference based on the interaction video frame. Further, the system may determine that the storage structure inference and the interaction inference correspond to a common time period and common location, and generate analytics information based on the storage structure inference and the interaction inference.

Method and system to provide real time interior analytics using machine learning and computer vision

Example implementations include a method, apparatus and computer-readable medium of computer vision configured for person classification, comprising receiving, during a first period of time, a plurality of image frames of an environment, identifying images of persons from each frame of the plurality of image frames, and determining a respective vector representation of each of the images. The implementations include generating a probability distribution indicative of a likelihood of a particular vector representation appearing in the plurality of image frames and identifying an associate vector representation by sampling the probability distribution using a probability model. The implementations include determining an input vector representation of an input image identified in an image frame depicting a person and received during a second period of time. The implementations include comparing the input vector representation with the associate vector representation and classifying, based on the comparison, the person in the input image as an associate.

Vision system for classifying persons based on visual appearance and dwell locations

Example implementations include a method, apparatus and computer-readable medium for person detection by a computer device, comprising receiving a plurality of video frames from a camera located in an environment, wherein the video frames include a depiction of a person in the environment. The implementations further include identifying attributes of the person from the video frames and comparing the attributes with attribute entries in a database of suspicious persons, wherein the database of suspicious persons includes attributes of a plurality of persons associated with an alert in the environment. The implementations include retrieving a profile indicative of an alert history of the suspicious person in response to determining a match between the attributes of the person and an entry in the database corresponding to a suspicious person, and transmitting, to a second computer device, the retrieved profile and a notification that indicates that the suspicious person is in the environment.

Person-of-interest (POI) detection

Aspects of the present disclosure relate generally to a vision system that detects persons exiting an environment, comprising detecting, by a processor using a plurality of image frames from at least one sensor, persons that exited the environment during a first period of time. The vision system further determines, by the processor, an exit count for the first period of time and retrieves, from a database, historical egress data comprising a detected historic exit count and a corrected historic exit count of the environment for a second period of time corresponding to the first period of time. The vision system calculates an error rate for the second period of time based on a ratio of the detected historic exit count and the corrected historic exit count, determines and stores a corrected exit count for the first period of time by adjusting the exit count using the error rate.

Systems and methods for object detection in an environment

A system may be configured to perform enhanced detection of occluded objects in a multiple object detection system. In some aspects, the system may determine a first size of a queue captured in a video frame based on an object detection system and a video capture device, and determine a second size of the queue based on the first size of the queue and a correction model. Further, the system may detect a person at a front of the queue based on an object tracking system, determine a first period of time spent by the person at the front of the queue based on an object tracking system, and estimate a second period of time spent in the queue based at least in part on the first period of time and the second size of the queue.

Queue monitoring in occlusion conditions through computer vision

Example aspects include methods, apparatuses, and computer-readable medium for receiving, by an antenna of a tag tracking/detection system, a tag signal of a tag in the tag tracking/detection system; determining, based on a strength or pattern of the tag signal, whether the tag signal comprises a reflected signal that is reflected by one or more reflective surfaces, wherein the one or more reflective surfaces are configured to focus tag signals of tags in a target location of the tag tracking/detection system toward the antenna of the tag tracking/detection system; and generating an output signal responsive to the tag signal comprising the reflected signal that is reflected by the one or more reflective surfaces, wherein the output signal is indicative of the tag being detected in the target location.

Tag tracking / detection system having a reflective surface

Aspects of the present disclosure provide techniques that utilize artificial intelligence (AI) to preemptively identify and prioritize risks for retailers from shrink loss. Specifically, features of the present disclosure provide actionable prescriptions to prevent shrink based on the output of a hybrid machine learning (ML) system that accurately identifies shrink risk factors and implements a cost-effective shrinkage control plan to reduce the retail theft. The shrinkage control plan may then be displayed to the user on a display device (or user interface) in order to allow for the user (e.g., store manager) to implement the recommendations developed by the recommendation engine.

Techniques for preventing shrink based on hybrid machine learning system

Example aspects include techniques for building a ML model in a use case with prohibitive training data and employing the ML model within the use case. These techniques may include determining training information including a plurality of stray training reads and a plurality of valid training reads, determining modified training information based at least in part on modifying the plurality of valid training reads, and generating a model for distinguishing a valid read from a stray read based on the modified training information and an evolutionary algorithm. In addition, the techniques may include detecting, by a monitoring device, a plurality of tag reads in response to a plurality of interactions between a tag and the monitoring device, and determining, by the monitoring device, a plurality of valid tag reads based on the model and plurality of tag reads.

Facilitating ML algorithm contribution to use cases where data is prohibitive

A system may be configured to implement a probability network for loss prevention sensors. In some aspects, the system may determine a location transition of a tag associated with an article, determine a first predictive value based at least in part on a first predictive structure and the location transition, the first predictive structure corresponding to a valid detection alarm, and determine a second predictive value based at least in part on a second predictive structure and the location transition, the second predictive structure corresponding to an invalid detection alarm. Further, the system may generate a notification based at least in on part on comparing the first predictive value to the second predictive value, the notification indicating unauthorized movement of the tag outside of a geographic area.

Method and system for probabilistic network based loss prevention sensors

Systems and methods for alarm panel analysis include receiving a plurality of alarm events from the respective alarm panels monitoring corresponding buildings. An alarm analysis system may classify the alarm panels by identifying an alarm type for the alarm events. The alarm analysis system may determine a number of occurrences of each alarm type for the alarm panels, and generate a data point for a dataset for the alarm panels. The alarm analysis system may identify statistical dividers in the data points for the data set, which may be used for assigning a ranking for the alarm panels based on their location in relation to the statistical dividers. The alarm analysis system may construct and a monitoring dashboard which includes the rankings of the alarm panels, which may be rendered on a display to an end user.

Automated alarm panel classification using pareto optimization

A system may be configured to reliably identify, track, and collect data on a person of interest. In some aspects, the system may detect a facial representation within an image, generate a bounding box corresponding to the facial representation, determine an enhanced facial representation based at least in part on the bounding box and an image enhancement pipeline, and extract a plurality of facial features from the enhanced facial representation. Further, the system may determine quality information based on the plurality of facial features, determine that the quality information is greater than a threshold, and store the plurality of facial features in a data structure. In some aspects, the facial features may be used to identify red shoppers and green shoppers in a controlled environment.

Method and system for identifying, tracking, and collecting data on a person of interest

A system may be configured to accurately track shelf activity in real-time with support for dynamic shelf size, configuration, and item containment. In some aspects, the system may parse regions of a video frame to determine a region of interest representation corresponding to a physical location (e.g., a shelf compartment), determine an enhanced region of interest representation based at least in part on the region of interest representation and an image enhancement pipeline, determine edge information of one or more objects based on the enhanced region of interest representation, compare a reference representation of the physical location to the edge information, and determine the amount of available space for the physical location based on the comparing.

Real time tracking of shelf activity supporting dynamic shelf size, configuration and item containment

Patient transfer device

Disclosed herein are apparatuses and methods for classifying person type in a visual medium. The implementations include identifying a person in an image, generating a visual boundary around the identified person, and dividing the image into a plurality of image portions such that multiple images portions partially depict the identified person inside the visual boundary. The implementations include generating the divided image on a user interface that prompts a user to select each portion that at least partially depicts a person of a predetermined type. The implementations include receiving a selection of at least one portion. The implementations include generating, for a machine learning model and in response to determining that the selection comprises the identified person, a training input that comprises a training image and a tag indicating that the identified person is of the predetermined type. Additionally, the implementations include storing the training input in memory.

Classification of person type in a visual medium

Disclosed herein are apparatuses and methods for iteratively mapping a layout of an environment. The implementations include receiving a visual stream from a camera installed in the environment, wherein the visual stream depicts a view of the environment, and wherein positional parameters of the camera and dimensions of the environment are set to arbitrary values. The implementations include monitoring a plurality of persons in the visual stream. For each person in the plurality of persons, the implementations further includes identifying a respective path that the person moves along in the view, updating the dimensions of the environment captured in the view, based on an estimated height of the person and movement speed along the respective path, and updating the positional parameters of the camera based on the updated dimensions of the environment. The implementations further includes mapping a layout of the environment captured in the view of the camera.

Growing community of inventors

Delray Beach, FL, United States of America

Michael C Stewart