Object re-identification during image tracking

A method, comprising:

• tracking, over a period of time using top-view images generated by at least one of a plurality of sensors, a first global position of a first object in the space, based on pixel coordinates of a first contour associated with the first object, wherein each sensor of the plurality of sensors is configured to generate top-view images of at least a portion of a space;
• tracking, over the period of time using top-view images generated by at least one of the plurality of sensors, a second global position of a second object in the space, based on pixel coordinates of a second contour associated with the second object;
• at a first time stamp corresponding to a time within the period of time, detecting a collision event between the first and second object, wherein the collision event corresponds to the first tracked position being within a threshold distance of the second tracked position;
• after detecting the collision event, receiving a first top-view image from a first sensor of the plurality of sensors, the first top-view image comprising a top-view image of the first object;
• determining, based on the first top-view image, a first descriptor for the first object, the first descriptor comprising at least one value associated with an observable characteristic of the first contour;
• determining that criteria are not satisfied for distinguishing the first object from the second object based on the first descriptor, wherein the criteria are not satisfied when a difference, during a time interval associated with the collision event, between a first value of the first descriptor and a second value of a second descriptor associated with the second object is less than a minimum value;
• in response to determining that the criteria are not satisfied, determining a third descriptor for the first contour, wherein the third descriptor comprises a value generated by an artificial neural network configured to identify objects in top-view images;
• and
• determining, based on the third descriptor, that a first identifier from a set of predefined identifiers corresponds to the first object.

A tracking subsystem communicatively coupled to a plurality of sensors, each sensor of the plurality of sensors configured generate top-view images of at least a portion of a space, the tracking subsystem configured to:

• track, over a period of time using top-view images generated by at least one of the plurality of sensors, a first global position of a first object in the space, based on pixel coordinates of a first contour associated with the first object;
• track, over the period of time using top-view images generated by at least one of the plurality of sensors, a second global position of a second object in the space, based on pixel coordinates of a second contour associated with the second object;
• at a first time stamp corresponding to a time within the period of time, detect a collision event between the first and second object, wherein the collision event corresponds to the first tracked position being within a threshold distance of the second tracked position;
• after detecting the collision event, receive a first top-view image from a first sensor of the plurality of sensors, the first top-view image comprising a top-view image of the first object;
• determine, based on the first top-view image, a first descriptor for the first object, the first descriptor comprising at least one value associated with an observable characteristic of the first contour;
• determine that criteria are not satisfied for distinguishing the first object from the second object based on the first descriptor, wherein the criteria are not satisfied when a difference, during a time interval associated with the collision event, between a first value of the first descriptor and a second value of a second descriptor associated with the second object is less than a minimum value;
• in response to determining that the criteria are not satisfied, determine a third descriptor for the first contour, wherein the third descriptor comprises a value generated by an artificial neural network configured to identify objects in top-view images;
• and
• determine, based on the third descriptor, that a first identifier from a set of predefined identifiers corresponds to the first object.