ORIENTATION-BASED HASHING FOR FAST ITEM ORIENTATION SENSING

• US 20160362211A1
• Filed: 06/11/2015
• Published: 12/15/2016
• Est. Priority Date: 06/11/2015
• Status: Active Grant

• Current Assignee: Empire Technology Development LLC (Allied Inventors Management, LLC)
• Original Assignee: Empire Technology Development LLC (Allied Inventors Management, LLC)
• Inventors: Kruglick, Ezekiel
• Granted Patent: US 9,969,514 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: B65B 35/58 Turning articles by positively-acting means, e.g. to present labelled portions in uppermost position
• CPC Class Codes: B65B 57/02 responsive to absence, presence, abnormal feed, or misplacement of binding or wrapping material, containers, or packages
• CPC Class Codes: G05B 15/02 electric
• CPC Class Codes: G05B 19/4183 characterised by data acquisition, e.g. workpiece identification
• CPC Class Codes: G05B 2219/35084 Geometric feature extraction, concave and convex regions, object recognition
• CPC Class Codes: G05B 2219/40554 Object recognition to track object on conveyor
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

Technologies are generally described for orientation-based hashing to determine item orientation. In some examples, a conveyance system may be configured to adjust the orientations of various types of items before sorting. The conveyance system may include a computer vision system that determines whether the orientation of an item should be adjusted based on an item-feature hash table that relates item image features to item orientations or actuator signals. In other examples, the computer vision system may determine item image features associated with a particular item and perform an item-feature hash table lookup using the item image features to determine an item orientation or actuator action. The item-feature hash table may be separately trained, allowing the computer vision system to rapidly determine the appropriate item orientation or actuator action.

Claims

1. A system configured to adjust item orientation, the system comprising:

a memory configured to store an item-feature hash table;

an imaging device configured to determine at least one feature associated with an item;

an actuator configured to adjust an orientation of the item based on an actuation signal; and

a processor coupled to the memory, the imaging device, and the actuator, the processor configured to;

determine a hash value based on the feature determined by the imaging device;

determine the actuation signal based on the hash value used as a key input into the item-feature hash table; and

provide the actuation signal to the actuator.

2. The system of claim 1, wherein the actuator is configured to adjust the orientation of the item through at least one of a tilt and a rotation of the item.

3. The system of claim 2, further comprising a conveyor module configured to transport the item, wherein the actuator is configured to adjust the orientation of the item with respect to the conveyor module.

4. The system of claim 1, wherein the at least one feature is associated with a particular orientation of the item.

5. The system of claim 1, wherein the imaging device is configured to determine the at least one feature by use of at least one of:

a speeded-up robust features (SURF) process;

a scale-invariant feature transform (SIFT) process; and

a histogram of oriented gradients (HOG) process.

6. The system of claim 1, wherein the item-feature hash table is configured to group similar actuation signals into nearby hash buckets.

7. The system of claim 1, wherein the item-feature hash table is implemented by use of locality-sensitive hashing and/or parameter sketching.

8. The system of claim 1, wherein the item-feature hash table is one of a single-entry hash table, a multiple-entry hash table, and a cuckoo hash table.

9. A system configured to determine item orientation, the system comprising:

a memory configured to store an item-feature hash table, the item-feature hash table configured to relate feature-hash values to item orientation parameters; and

a processor configured to;

determine a first feature-hash value based on at least one feature associated with an item; and

determine a first item orientation parameter associated with the first feature-hash value from the item-feature hash table.

10. The system of claim 9, wherein the at least one feature is uniquely associated with a particular orientation of the item.

11. The system of claim 9, wherein the at least one feature is at least one of a SURF feature, a SIFT feature, and a HOG feature.

12. The system of claim 9, wherein similar item orientation parameters are grouped into nearby hash buckets within the item-feature hash table.

13. The system of claim 9, wherein the item-feature hash table is implemented by use of locality-sensitive hashing and/or parameter sketching.

14. The system of claim 9, wherein the item-feature hash table is one of a single-entry hash table, a multiple-entry hash table, and a cuckoo hash table.

15. The system of claim 9, wherein the first item orientation parameter indicates at least one of an orientation of the item and an actuation signal associated with the item.

16. A method to determine item orientation, the method comprising:

determining at least one feature associated with an orientation of an item;

determining a hash value based on the feature; and

using the hash value as a key for an item-feature hash table to determine an orientation signal, wherein the item-feature hash table is configured to relate feature-hash values to item orientation parameters.

17. The method of claim 16, wherein the at least one feature is at least one of a SURF feature, a SIFT feature, and a HOG feature.

18. The method of claim 16, further comprising:

generating the item-feature hash table based on locality-sensitive hashing and/or parameter sketching.

19. The method of claim 18, wherein generating the item-feature hash table further comprises generating the item-feature hash table using sparse matrices.

20. The method of claim 18, wherein generating the item-feature hash table comprises:

collecting, for at least one item, a plurality of orientations and associated features; and

determining at least one hash transformation configured to collapse multiple, similar orientations in the plurality of orientations together into nearby hash buckets.

21. The method of claim 20, wherein determining the at least one hash transformation comprises:

determining at least one hash metric based on a type of the item;

performing a decomposition based on the at least one hash metric to form at least one eigenvector; and

using the at least one eigenvector to collapse the multiple, similar orientations together.

22. The method of claim 16, wherein the item-feature hash table is one of a single-entry hash table, a multiple-entry hash table, and a cuckoo hash table.

23. The method of claim 16, wherein the orientation signal indicates at least one of the orientation of the item and an actuation signal associated with the item.