Action detection during image tracking
A system includes a sensor, a weight sensor, and a tracking subsystem. The tracking subsystem receives an image feed of top-view images generated by the sensor and weight measurements from the weight sensor. The tracking subsystem detects an event associated with an item being removed from a rack in which the weight sensor is installed. The tracking subsystem determines that a first person or a second person may be associated with the event. In response to determining that the first or second person may be associated with the event, buffer frames are stored of top-view images generated by the sensor during a time period associated with the event. The tracking subsystem then determines, using at least one of the stored buffer frames and a first action-detection algorithm, whether an action associated with the event was performed by the first person or the second person.
1. A system comprising:
a plurality of sensors, each sensor of the plurality of sensors configured to generate images of at least a portion of a space; and
a tracking subsystem communicatively coupled to the plurality of sensors, the tracking subsystem comprising a processor configured to:
receive the images generated by the plurality of sensors;
track a first object and one or more other objects in the space using at least a portion of the images generated by the plurality of sensors;
determine that re-identification of the tracked first object is needed based at least upon a probability that an identifier of the tracked first object is associated with the first object is less than a threshold probability value;
in response to determining that re-identification of the tracked first object is needed, determine candidate identifiers for the tracked first object, wherein the candidate identifiers comprise a subset of identifiers of all tracked objects, the subset comprising possible identifiers of the tracked first object based at least in part upon a history of movements of the tracked first object;
receive, from a first sensor of the plurality of sensors, a first image of the tracked first object;
determine a first descriptor for the tracked first object, the first descriptor comprising at least one value associated with the tracked first object;
compare the first descriptor to a set of predetermined descriptors associated with the candidate identifiers determined for the tracked first object;
based on results of the comparison, determine an updated identifier for the tracked first object, wherein the updated identifier is the predetermined descriptor from the set of predetermined descriptors with a value that is within a threshold range of a first descriptor value; and
assign the updated identifier to the tracked first object.
2. The system of claim 1 , wherein:
the first descriptor comprises a first data vector associated with characteristics of the tracked first object in a frame;
each of the predetermined descriptors from the set of predetermined descriptors comprises a corresponding predetermined data vector; and
the processor of the tracking subsystem is further configured to:
compare the first descriptor to each of the predetermined descriptors from the set of predetermined descriptors associated with the candidate identifiers by calculating a first cosine similarity value between the first data vector and each of the predetermined data vectors; and
determine the updated identifier as a particular candidate identifier that corresponds to the first cosine similarity value nearest one.
3. The system of claim 2 , wherein the processor of the tracking subsystem is further configured to determine that each of the first cosine similarity values is less than a threshold similarly value and, in response:
determine a second descriptor value for each of the one or more other objects, wherein each second descriptor value comprises a second data vector;
determine a second cosine similarity value between each of the second data vectors and each of the predetermined descriptor values; and
determine second updated identifiers for the tracked first object and each of the other objects, based on the first and second cosine similarity values.
4. The system of claim 1 , wherein the first descriptor is determined based on a portion of the first image, the portion corresponding to a predefined field-of-view comprising a central sub-region of a full field-of-view addressed by the first sensor.
5. The system of claim 1 , wherein the tracked first object is a first person.
6. The system of claim 1 , wherein the processor of the tracking subsystem is further configured to, prior to determining that re-identification of the tracked first object is needed, periodically determine updated predetermined descriptors associated with the candidate identifiers.
7. The system of claim 6 , wherein the processor of the tracking subsystem is further configured to:
in response to determining the updated predetermined descriptors, determine that a first updated predetermined descriptor is different by at least a threshold amount than a corresponding previously predetermined descriptor; and
store in a memory both the updated descriptor and the corresponding previous predetermined descriptor.
8. A method comprising:
receiving images generated by a plurality of sensors, each sensor of the plurality of sensors configured to generate images of at least a portion of a space;
tracking a first object and one or more other objects in the space using at least a portion of the images generated by the plurality of sensors;
determining that re-identification of the tracked first object is needed based at least upon a probability that an identifier of the tracked first object is associated with the first object is less than a threshold probability value;
in response to determining that re-identification of the tracked first object is needed, determining candidate identifiers for the tracked first object, wherein the candidate identifiers comprise a subset of identifiers of all tracked objects, the subset comprising possible identifiers of the tracked first object based at least in part upon a history of movements of the tracked first object;
receiving, from a first sensor of the plurality of sensors, a first image of the tracked first object;
determining a first descriptor for the tracked first object, the first descriptor comprising at least one value associated with the tracked first object;
comparing the first descriptor to a set of predetermined descriptors associated with the candidate identifiers determined for the tracked first object;
based on results of the comparison, determining an updated identifier for the tracked first object, wherein the updated identifier is the predetermined descriptor from the set of predetermined descriptors with a value that is within a threshold range of a first descriptor value; and
assigning the updated identifier to the tracked first object.
9. The method of claim 8 , wherein:
the first descriptor comprises a first data vector associated with characteristics of the tracked first object in a frame;
each of the predetermined descriptors from the set of predetermined descriptors comprises a corresponding predetermined data vector; and
the method further comprises:
comparing the first descriptor to each of the predetermined descriptors from the set of predetermined descriptors associated with the candidate identifiers by calculating a first cosine similarity value between the first data vector and each of the predetermined data vectors; and
determining the updated identifier as a particular candidate identifier that corresponds to the first cosine similarity value nearest one.
10. The system of claim 9 , further comprising determining that each of the first cosine similarity values is less than a threshold similarity value and, in response:
determining a second descriptor value for each of the one or more other objects, wherein each second descriptor value comprises a second data vector;
determining a second cosine similarity value between each of the second data vectors and each of the predetermined descriptor values; and
determining second updated identifiers for the tracked first object and each of the other objects, based on the first and second cosine similarity values.
11. The method of claim 8 , determining the first descriptor based on a portion of the first image, the portion corresponding to a predefined field-of-view comprising a central sub-region of a full field-of-view addressed by the first sensor.
12. The method of claim 8 , wherein the tracked first object is a first person.
13. The method of claim 8 , further comprising, prior to determining that re-identification of the tracked first object is needed, periodically determining updated predetermined descriptors associated with the candidate identifiers.
14. The method of claim 13 , further comprising:
in response to determining the updated predetermined descriptors, determining that a first updated predetermined descriptor is different by at least a threshold amount than a corresponding previously predetermined descriptor; and
storing in a memory both the updated descriptor and the corresponding previous predetermined descriptor.
15. A tracking subsystem communicatively coupled to a plurality of sensors, each sensor of the plurality of sensors configured to generate images of at least a portion of a space, the tracking subsystem comprising a processor configured to:
receive images generated by the plurality of sensors;
track a first object and one or more other objects in the space using at least a portion of the images generated by the plurality of sensors;
determine that re-identification of the tracked first object is needed based at least upon a probability that an identifier of the tracked first object is associated with the first object is less than a threshold probability value;
in response to determining that re-identification of the tracked first object is needed, determine candidate identifiers for the tracked first object, wherein the candidate identifiers comprise a subset of identifiers of all tracked objects, the subset comprising possible identifiers of the tracked first object based at least in part upon a history of movements of the tracked first object;
receive, from a first sensor of the plurality of sensors, a first image of the tracked first object;
determine a first descriptor for the tracked first object, the first descriptor comprising at least one value associated with the tracked first object;
compare the first descriptor to a set of predetermined descriptors associated with the candidate identifiers determined for the tracked first object;
based on results of the comparison, determine an updated identifier for the tracked first object, wherein the updated identifier is the predetermined descriptor from the set of predetermined descriptors with a value that is within a threshold range of a first descriptor value; and
assign the updated identifier to the tracked first object.
16. The tracking subsystem of claim 15 , wherein:
the first descriptor comprises a first data vector associated with characteristics of the tracked first object in a frame;
each of the predetermined descriptors from the set of predetermined descriptors comprises a corresponding predetermined data vector; and
the processor of the tracking subsystem is further configured to:
compare the first descriptor to each of the predetermined descriptors from the set of predetermined descriptors associated with the candidate identifiers by calculating a first cosine similarity value between the first data vector and each of the predetermined data vectors; and
determine the updated identifier as a particular candidate identifier that corresponds to the first cosine similarity value nearest one.
17. The tracking subsystem of claim 16 , wherein the processor is further configured to determine that each of the first cosine similarity values is less than a threshold similarity value and, in response:
determine a second descriptor value for each of the one or more other objects, wherein each second descriptor value comprises a second data vector;
determine a second cosine similarity value between each of the second data vectors and each of the predetermined descriptor values; and
determine second updated identifiers for the tracked first object and each of the other objects, based on the first and second cosine similarity values.
18. The tracking subsystem of claim 15 , wherein the first descriptor is determined based on a portion of the first image, the portion corresponding to a predefined field-of-view comprising a central sub-region of a full field-of-view addressed by the first sensor.
19. The tracking subsystem of claim 15 , wherein the tracked first object is a first person.
20. The tracking subsystem of claim 15 , wherein the processor is further configured to, prior to determining that re-identification of the tracked first object is needed, periodically determine updated predetermined descriptors associated with the candidate identifiers.