Agent re-identification

A computing system, comprising:

• a processor;
• and
• a memory coupled to the processor and storing program instructions that when executed by the processor cause the processor to at least:
• receive first image data that includes a plurality of pixels corresponding to an anterior representation of an agent;
• partition the first image data into:
• a first partition of the first image data that includes a first portion of the first image data corresponding to a first portion of the agent;
• and
• a second partition of the first image data that includes a second portion of the first image data corresponding to a second portion of the agent;
• generate, based at least in part on the first partition and the second partition, a first agent color signature representative of color values corresponding to the plurality of pixels;
• compare the first agent color signature with a second agent color signature corresponding to a second image data to determine a similarity between the first agent color signature and the second agent color signature;
• determine a weighted similarity such that a portion of the first image data corresponding to a head of the agent represented in the first image data and a portion of the second image data corresponding to a head of the agent represented in the second image data has a higher weight;
• and
• determine, based at least in part on the similarity or the weighted similarity, that the agent represented in the first image data is likely the agent represented in the second image data.

A computer implemented method, comprising:

• under control of one or more computing systems configured with executable instructions,
• receiving first image data from a first image capture device located within a materials handling facility, wherein the first image data includes a plurality of pixels illustrating an anterior view of at least a portion of an agent located at a position within the materials handling facility;
• partitioning the first image data into:
• a first partition that includes a first portion of the first image data corresponding to a head region of the agent represented in the first image data;
• and
• a second partition that includes a second portion of the first image data;
• generating a first color signature corresponding to color values of pixels included in the first partition and representative of the head region of the agent;
• generating a second color signature corresponding to color values of pixels included in the second partition;
• comparing the first color signature with a first stored color signature corresponding to color values of pixels of stored image data representative of the head region of the agent represented in the stored image data to determine a first similarity between the first color signature and the first stored color signature;
• comparing the second color signature with a second stored color signature corresponding to color values of pixels of stored image data representative of a second region of the agent represented in the stored image data to determine a second similarity between the second color signature and the second stored color signature;
• determining, based at least in part on the first similarity and the second similarity that the agent represented in the first image data is potentially the agent represented in the stored image data;
• and
• including an identification of the agent represented in stored image data in a list of candidate agents that are potentially the agent represented in the first image data such that the list only includes agents that are determined to potentially be the agent represented in the first image data.

A method, comprising:

• receiving first image data that includes a plurality of pixels corresponding to a representation of an agent;
• generating a first agent color signature representative of color values corresponding to the plurality of pixels, wherein the first agent color signature is determined based at least in part on a k-means algorithm;
• and
• comparing the first agent color signature with a second agent color signature corresponding to a second image data that includes a second plurality of pixels corresponding to a representation of a second agent to determine a similarity between the first agent color signature and the second agent color signature, wherein the similarity is determined based at least in part on an earth mover's distance algorithm.