Object tracking

Patent metadata

Field	Value
Publication number	US-12530783-B2
Application number	US-202018015981-A
Country	US
Kind code	B2
Filing date	Jul 14, 2020
Priority date	Jul 14, 2020
Publication date	Jan 20, 2026
Grant date	Jan 20, 2026

Abstract

There is provided a method comprising: acquiring ( 110 ) sensor data related to an object; using the first learning module, identifying ( 120 ) the object based on the acquired sensor data using a first learning module and determining ( 130 ) a user associated with the identified object; determining ( 140 ) a timestamped location of the object based on at least one of the acquired sensor data and one or more locations of the one or more sensors; performing ( 150 ) a first analysis to determine whether the current status of the object contains an anomaly based on one or more predefined rules stored in a knowledge base; performing ( 160 ) a second analysis to determine whether the current status of the object contains an anomaly, using a second learning module; and validating ( 170 ) whether the current status of the object contains an anomaly based on results of the first analysis and results of the second analysis.

First claim

The invention claimed is:

1 . A computer-implemented method for object tracking using machine learning, the method comprising:
acquiring sensor data from one or more sensors, wherein the sensor data is related to an object; identifying the object based on the acquired sensor data using a first learning module, wherein the first learning module comprises a supervised or unsupervised learning module; determining a user associated with the identified object using the first learning module; determining a timestamped location of the identified object based on at least one of the acquired sensor data and one or more locations of the one or more sensors; performing a first analysis to determine whether a current status of the identified object contains an anomaly based on one or more predefined rules stored in a knowledge base, wherein the current status is defined by: an identifier of the identified object, the determined user associated with the identified object, and the timestamped location of the identified object; performing a second analysis to determine whether the current status of the identified object contains an anomaly, using a second learning module, wherein the second learning module comprises a semi-supervised or unsupervised learning module; and validating whether the current status of the identified object contains an anomaly based on results of the first analysis and results of the second analysis.

2 . The computer-implemented method according to claim 1 , further comprising detecting a change in location of the object based on the acquired sensor data, wherein the current status of the object is further defined by the detected change in location of the identified object.

3 . The computer-implemented method according to claim 1 , further comprising detecting an entity in the vicinity of the identified object based on the acquired sensor data.

4 . The computer-implemented method according to claim 2 , further comprising detecting a pick-up event of the object based on whether the detected change in location of the identified object involves an instantaneous displacement from a location,
wherein performing the first analysis comprises determining, upon detecting the pick-up event of the object, whether an entity triggering the pick-up event corresponds to the determined user associated with the identified object, based on the one or more predefined rules stored in the knowledge base, and wherein performing the second analysis comprises determining, upon detecting the pick-up event of the object, whether an entity triggering the pick-up event or an entity carrying the object after the pick-up event corresponds to the determined user associated with the identified object, using the second learning module.

5 . The computer-implemented method according to claim 2 , further comprising detecting a drop-off event of the object based on whether the detected change in location of the identified object involves a placement of the object,
wherein performing the first analysis comprises determining, upon detecting the drop-off event of the object, whether an entity triggering the drop-off event corresponds to the determined user associated with the identified object, based on the one or more predefined rules stored in the knowledge base, and wherein performing the second analysis comprises determining, upon detecting the drop-off event the object, whether an entity triggering the drop-off event corresponds to the determined user associated with the identified object, using the second learning module.

6 . The computer-implemented method according to claim 2 , wherein detecting a pick-up event or a drop-off event of the object is based on a third learning module, and wherein the third learning module comprises a supervised learning module that is pre-trained based on at least one of: distances between an object and one or more users or entities associated with the respective object, for one or more objects; audio and/or visual data associated with displacement of one or more objects; and a range of speed and/or acceleration of movement of an object during displacement of the respective object for one or more of objects.

7 . The computer-implemented method according to claim 2 , further comprising determining whether the determined user associated with the identified object is in the vicinity of the object upon detecting a change in location of the object, wherein performing at least one of the first analysis and the second analysis is further based on whether the user associated with the identified object is in the vicinity of the object subsequent to the change in location of the object.

8 . The computer-implemented method according to claim 1 , wherein identifying the object using the first learning module comprises determining that the object does not correspond to a known object with respect to at least one of the first learning module, the second learning module, and the knowledge base, and wherein determining a user associated with the identified object is performed using an inference engine of the first learning module, the method further comprising at least one of the following steps:
updating the knowledge base with an identifier of the identified object and the determined user associated with the identified object; and training the second learning module with an identifier of the identified object and the determined user associated with the identified object.

9 . The computer-implemented method according to claim 1 , further comprising generating an output indicating the anomaly upon validating that the current status contains an anomaly.

10 . The computer-implemented method according to claim 1 , further comprising, if it is validated that the current status of the identified object contains an anomaly, performing at least one of:
updating a database using results of the validation, wherein the database contains at least one of: one or more usual locations with respect to time and/or date for a plurality of objects including the identified object; updating the knowledge base based on results of the validation; and training the second learning module based on of the validation.

11 . The computer-implemented method according to claim 10 , further comprising:
receiving a user query of a location of the object via a user interface; retrieving, from the data base, at least one of the latest updated location of the object and the usual location of the object based on the user query; and outputting the retrieved information via the user interface.

12 . The computer-implemented method according to claim 1 , further comprising, if it is validated that the current status of the identified object does not contain an anomaly, performing at least one of:
updating the knowledge base based on at least one of the results of the first analysis and the results of the second analysis; and training the second learning module based on at least one of the results of the first analysis and the results of the second analysis.

13 . The computer-implemented method according to claim 1 , wherein the second learning module includes clustering of a plurality of object identifiers into a plurality of groups, wherein the plurality of groups are associated with at least one of: a location, a user, and a date and/or time.

14 . The computer-implemented method according to claim 1 , wherein validating whether the current status of the identified object contains an anomaly is further based on a relative weighting of the second learning module, wherein the relative weighting is at least based on a level of output uncerta

Assignees

• Ericsson Telefon Ab L M

Inventors

• Folkesson Mats
• Moradi Farnaz
• Ickin Selim
• Lan Xiaoyu

Classifications

• G06T2207/20081

  Training; Learning · CPC title

• G06T2200/24

  involving graphical user interfaces [GUIs] · CPC title

• G06T7/0002

  Inspection of images, e.g. flaw detection · CPC title

• G06V10/945

  User interactive design; Environments; Toolboxes · CPC title

• G06V10/776

  Validation; Performance evaluation · CPC title