Granted Patent B2

US 12,190,613 · App. 16/672,179 · Granted Jan 7, 2025

Inventors: Yan Zhang (Buffalo Grove, IL); Tong Shen (Buffalo Grove, IL); Santiago Romero (Mount Airy, MD)

Assignee: Zebra Technologies Corporation

G06V20/64G06V10/96G06V30/10

Abstract

Claims (53)

1. A system, comprising:

a container recognition component configured to receive an image and perform container type recognition and localization analysis of the image based on a shape of at least one container present in the image, to produce one of:

(i) a first output indicating a container type is identified;

(ii) a second output corresponding to a first set of container types, and

(iii) a third output corresponding to a second set of container types;

a character recognition component configured to perform further analysis of the image, responsive to the container recognition component not being able to determine a container type from the first set of container types, to identify alphanumeric characters and determine a container type from the first set of container types; and

a 3D point cloud component configured to perform further analysis of the image, responsive to the container recognition component not being able to determine a container type from the second set of container types, to determine a container type from the second set of container types,

wherein, responsive to the container recognition component not being able to produce the one of three outputs satisfying a confidence level, the container recognition component is further configured to apply a temporal coherence constraint on the image and a plurality of previous sequential frames to produce the one of three outputs.

2. The system of claim 1 , wherein when the output is that a container is identified, further comprising:

the container recognition component is further configured to identify the container by analyzing the received image with a machine learning algorithm.

3. The system of claim 1 , wherein the character recognition component is configured to perform further analysis of the image by:

analyzing a subset of the image to identify the alphanumeric characters.

4. The system of claim 3 , wherein the identified alphanumeric characters represent a container identifier code.

5. The system of claim 4 , wherein the container identifier code represents an identifiable container type.

6. The system of claim 1 , wherein the 3D point cloud component is configured to perform further analysis of the image by:

analyzing a subset of the image to identify a container ceiling attribute.

7. The system of claim 6 , wherein the 3D point cloud component is configured to identify, based on the container ceiling attribute, a container type.

8. A method, comprising:

receiving, at a container recognition component, an image;

performing, by the container recognition component, container type recognition and localization analysis of the image based on a shape of at least one container present in the image;

generating, based on the container type recognition and localization analysis of the image by the container recognition component, one of:

(i) a first output indicating a container type is identified;

(ii) a second output corresponding to a first set of container types, and

(iii) a third output corresponding to a second set of container types;

performing, by a character recognition component, further analysis of the image, responsive to the container recognition component not being able to determine a container type from the first set of container types, to identify alphanumeric characters and determine a container type from the first set of container types;

performing, by a 3D point cloud component, further analysis of the image, responsive to the container recognition component not being able to determine a container type from the second set of container types, to determine a container type from the second set of container types; and

responsive to the container recognition component not being able to produce the one of three outputs satisfying a confidence level, applying, by the container recognition component, a temporal coherence constraint on the image and a plurality of previous sequential frames to produce the one of three outputs.

9. The method of claim 8 , further comprising:

filtering, at a coherent constraint component, a sequence of images including the received image based on a temporal coherence constraint to remove recognition errors.

10. The method of claim 8 , wherein performing, by the container recognition component, container type recognition and localization analysis of the image based on the shape of the at least one container present in the image further comprises:

analyzing, by the container recognition component, the image using a machine learning algorithm.

11. The method of claim 8 , wherein performing, by the character recognition component, further analysis of the image, further comprises:

analyzing, at the character recognition component, a subset of the image to identify the alphanumeric characters.

12. The method of claim 11 , wherein the identified alphanumeric characters represent a container identifier code.

13. The method of claim 8 , wherein performing, by the 3D point cloud component, further analysis of the image, further comprises:

analyzing, at the 3D point cloud component, a subset of the image to identify a container ceiling attribute.

14. The method of claim 13 , wherein the 3D point cloud component is configured to identify, based on the container ceiling attribute, a container type.

15. A tangible machine-readable medium comprising instructions for image analysis that, when executed, cause a machine to at least:

receive an image;

perform, by a container recognition component, container type recognition and localization analysis of the image based on a shape of at least one container present in the image; and

generate, based on the container type recognition and localization analysis, one of:

(i) a first output indicating a container type is identified by the container recognition component;

(ii) a second output corresponding to a first set of container types, and

(iii) a third output corresponding to a second set of container types;

perform, by a character recognition component, further analysis of the image, responsive to the container recognition component not being able to determine a container type from the first set of container types, to identify alphanumeric characters and determine a container type from the first set of container types;

perform, by a 3D point cloud component, further analysis of the image, responsive to the container recognition component not being able to determine a container type from the second set of container types, to determine a container type from the second set of container types; and

responsive to not being able to generate the one of three outputs satisfying a confidence level, applying a temporal coherence constraint on the image and a plurality of previous sequential frames to produce the one of three outputs.

16. The tangible machine-readable medium of claim 15 , further comprising instructions that, when executed, cause the machine to at least:

perform, by the container recognition component, the container type recognition and localization analysis of the image using a machine learning algorithm.

17. The tangible machine-readable medium of claim 15 , further comprising instructions that, when executed, cause the machine to at least:

analyze a subset of the image to identify the alphanumeric characters utilizing the character recognition component.

18. The tangible machine-readable medium of claim 15 , further comprising instructions that, when executed, cause the machine to at least:

analyze a subset of the image to identify a container ceiling attribute utilizing the 3D point cloud component.

Assignments (3)

Continuity (1)

References Cited (9)