Method, system, and non-transitory computer-readable recording medium, for monitoring object
AssigneeBodit Inc.
InventorsKwang Jae Choo, Min Yong Shin, Heung Jong Yoo, Yoon Chul Choi, Seongjin Kim, Nayeon Kim
Life of this patent
Estimating the term…
Abstract
A method for monitoring an object is provided. The method includes the steps of: specifying, from sensor data measured by a sensor for at least one object, pattern data on an unidentified behavior of the at least one object that is not recognized by a machine learning-based behavior recognition model; and determining at least a part of the pattern data as target data on the basis of at least one of a measurement frequency and a measurement magnitude of the pattern data, and providing information on the target data to a user.
Claims3
A method performed in a system for monitoring an object, the system comprising one or more processors and the method comprising the steps of:
- by the one or more processors, specifying, from sensor data measured by a sensor for at least one object, pattern data on an unidentified behavior among behaviors of the at least one object, wherein the unidentified behavior is a behavior that is not recognized by a machine learning-based behavior recognition model;
- by the one or more processors, determining at least a part of the pattern data as target data on the basis of at least one of a measurement frequency and a measurement magnitude of the pattern data, wherein the measurement frequency is a frequency at which the pattern data appears in the sensor data measured for a predetermined period of time, and providing information on the target data to a user;
- and
- by the one or more processors, training the behavior recognition model for the unidentified behavior using the target data and the information on the target data, on the basis of feedback acquired from the user regarding the information on the target data,
- wherein in the specifying step, data units that are clustered into a cluster other than a cluster of data units for a behavior of the at least one object recognized by the behavior recognition model are specified as the pattern data on the unidentified behavior, using a clustering algorithm,
- wherein the information on the target data includes context information on the unidentified behavior, and
- wherein the context information includes environmental information on the at least one object acquired in relation to a period in which the unidentified behavior occurs, information on behaviors of the at least one object recognized in relation to the period in which the unidentified behavior occurs, information on health status of the at least one object, and information on management status of the at least one object.
A system for monitoring an object, the system comprising one or more processors configured to:
- specify, from sensor data measured by a sensor for at least one object, pattern data on an unidentified behavior among behaviors of the at least one object, wherein the unidentified behavior is a behavior that is not recognized by a machine learning-based behavior recognition model;
- determine at least a part of the pattern data as target data on the basis of at least one of a measurement frequency and a measurement magnitude of the pattern data, wherein the measurement frequency is a frequency at which the pattern data appears in the sensor data measured for a predetermined period of time, and provide information on the target data to a user;
- and
- train the behavior recognition model for the unidentified behavior using the target data and information on the target data, on the basis of feedback acquired from the user regarding the information on the target data,
- wherein the one or more processors are configured to specify data units that are clustered into a cluster other than a cluster of data units for a behavior of the at least one object recognized by the behavior recognition model as the pattern data on the unidentified behavior, using a clustering algorithm,
- wherein the information on the target data includes context information on the unidentified behavior, and
- wherein the context information includes environmental information on the at least one object acquired in relation to a period in which the unidentified behavior occurs, information on behaviors of the at least one object recognized in relation to the period in which the unidentified behavior occurs, information on health status of the at least one object, and information on management status of the at least one object.
Dependencies are inferred from the claim text.