Determining candidate object identities during image tracking

A tracking subsystem communicatively coupled to the plurality of sensors, plurality of sensors comprising at least a first sensor and a second sensor, each sensor configured to generate a top-view image of at least a portion of a space, wherein the tracking subsystem is configured to:

• receive frames of top-view images generated by the plurality of sensors;
• track, based on at least a portion of the received frames, a first object associated with a first identifier;
• track, based on at least a portion of the received frames, a second object associated with a second identifier;
• track, based on at least a portion of the received frames, a third object associated with a third identifier;
• detect that the first object is within a threshold distance of the second object;
• in response to detecting that the first object is within the threshold distance of the second object:
• determine a first probability that the first object switched identifiers with the second object;
• update a first candidate list for the first object, based on the first probability that the first object switched identifiers with the second object, wherein the updated first candidate list comprises a second probability that the first object is associated with the first identifier and a third probability that the first object is associated with the second identifier;
• update a second candidate list for the second object, based on the first probability that the first object switched identifiers with the second object, wherein the updated second candidate list comprises a fourth probability that the second object is associated with the first identifier and a fifth probability that the second object is associated with the second identifier;
• following updating the first and second candidate lists, detect that the first object is within a threshold distance of the third object;
• in response to detecting that the first object is within the threshold distance of the third object:
• determine a sixth probability that the first object switched identifiers with the third object;
• further update the first candidate list for the first object, based on the sixth probability that the first object switched identifiers with the third object, wherein the further updated first candidate list comprises an updated second probability that the first object is associated with the first identifier, an updated third probability that the first object is associated with the second identifier, and a seventh probability that the first object is associated with the third identifier;
• and
• update a third candidate list for the third object, based on the sixth probability that the first object switched identifiers with the third object and the first candidate list prior to detecting that the first object is within the threshold distance of the third object, wherein the updated third candidate list comprises an eighth probability that the third object is associated with the first identifier, a ninth probability that the third object is associated with the second identifier, and a tenth probability that the third object is associated with the third identifier.

A method, comprising:

• receiving frames of top-view images generated by a plurality of sensors, the plurality of sensors comprising at least a first sensor and a second sensor, each sensor configured to generate a top-view image of at least a portion of a space;
• tracking, based on at least a portion of the received frames, a first object associated with a first identifier;
• tracking, based on at least a portion of the received frames, a second object associated with a second identifier;
• tracking, based on at least a portion of the received frames, a third object associated with a third identifier;
• detecting that the first object is within a threshold distance of the second object;
• in response to detecting that the first object is within the threshold distance of the second object:
• determining a first probability that the first object switched identifiers with the second object;
• updating a first candidate list for the first object, based on the first probability that the first object switched identifiers with the second object, wherein the updated first candidate list comprises a second probability that the first object is associated with the first identifier and a third probability that the first object is associated with the second identifier;
• updating a second candidate list for the second object, based on the first probability that the first object switched identifiers with the second object, wherein the updated second candidate list comprises a fourth probability that the second object is associated with the first identifier and a fifth probability that the second object is associated with the second identifier;
• following updating the first and second candidate lists, detecting that the first object is within a threshold distance of the third object;
• in response to detecting that the first object is within the threshold distance of the third object:
• determining a sixth probability that the first object switched identifiers with the third object;
• further updating the first candidate list for the first object, based on the sixth probability that the first object switched identifiers with the third object, wherein the further updated first candidate list comprises an updated second probability that the first object is associated with the first identifier, an updated third probability that the first object is associated with the second identifier, and a seventh probability that the first object is associated with the third identifier;
• and
• updating a third candidate list for the third object, based on the sixth probability that the first object switched identifiers with the third object and the first candidate list prior to detecting that the first object is within the threshold distance of the third object, wherein the updated third candidate list comprises an eighth probability that the third object is associated with the first identifier, a ninth probability that the third object is associated with the second identifier, and a tenth probability that the third object is associated with the third identifier.