Persistent feature based image rotation and candidate region of interest

Patent metadata

Field	Value
Publication number	US-11373388-B2
Application number	US-201816040344-A
Country	US
Kind code	B2
Filing date	Jul 19, 2018
Priority date	Jul 24, 2017
Publication date	Jun 28, 2022
Grant date	Jun 28, 2022

Abstract

Embodiments of a system and method for sorting and delivering articles in a processing facility based on image data are described. Image processing results such as rotation notation information may be included in or with an image to facilitate downstream processing such as when the routing information cannot be extracted from the image using an unattended system and the image is passed to an attended image processing system. The rotation notation information may be used to dynamically adjust the image before presenting the image via the attended image processing system.

First claim

What is claimed is:

1. A computer-implemented method comprising:
under control of one or more processors,
obtaining an image of a mail piece on a conveyor, the image showing address information affixed thereto;
receiving, from an optical character recognition (OCR) system, an OCR result comprising (i) an electronic text record of characters of the address information shown in the image, and (ii) rotation information for the image including a rotation angle, the rotation information identifying a number of degrees the image should be rotated to show characters of the address information from left to right with the characters in a standardized orientation;
determining that the electronic text record does not include a set of characters of the address information;
in response to determining that the electronic text record does not include the set of characters of the address information, adding the rotation angle to an image file of the image of the mail piece to generate an enhanced image;
transmitting the enhanced image for additional processing in a computing device; and
rotating, at the computing device, the enhanced image by the rotation angle prior to additional processing by the computing device,
wherein the rotation angle is dynamically generated based on a relationship between a maximum pixel density and an average pixel density for a row or a column, and
wherein the relationship is expressed as Equation (1) below:

rotation
⁢

=

30
*

density

w
-
avg

-

density

w
-
max

density

w
-
avg

Equation
⁢

⁢

(
1
)

where density w-max is the maximum density for a row of the image; and
density w-avg is the average density for the rows of the image.

2. The computer-implemented method of claim 1 , further comprising:
generating a grey-scale image from the image; and
providing the grey-scale image to the OCR system.

3. The computer-implemented method of claim 1 , wherein the address information is printed on a label affixed to the mail piece, and wherein the method further comprises:
identifying a type for the label based on an identifier for the label shown in the image;
receiving a layout template for the type, the layout template identifying a region for the address information; and
identifying the rotation information based at least in part on an orientation of the region shown in the image and an orientation of the characters shown within the region in the image.

4. The computer-implemented method of claim 3 , wherein the identifier for the type comprises at least one of a graphic or a scannable code identifying the type for the label.

5. The computer-implemented method of claim 1 , wherein determining that the electronic text record does not include the set of characters comprises evaluating the set of characters using a regular expression, and wherein a result of the evaluating indicates that the electronic text record does not include the set of characters.

6. The computer-implemented method of claim 1 , further comprising:
identifying a processing orientation for the mail piece based at least in part on a type for the mail piece; and
causing activation of the conveyor to rotate the mail piece from an orientation shown in the image to the processing orientation, wherein the activation is based at least in part on a difference between the processing orientation and the rotation information.

7. The computer-implemented method of claim 1 , wherein the image file comprises a rotation notation tagged field including the rotation angle.

8. The computer-implemented method of claim 1 , wherein the additional processing includes causing display of the enhanced image on a display device after the image is rotated by the rotation angle.

9. The computer-implemented method of claim 1 , wherein the rotation angle is adjusted in proportion to the difference between the average pixel density and the maximum pixel density for the row or the column.

10. A computer-implemented method comprising:
under control of one or more processors,
receiving an image showing an item and address information affixed thereto;
generating an electronic text record of the address information based on an optical character recognition (OCR) processing of at least a portion of the image showing the address information;
detecting orientation information for the image based on pixel information included in the image, the orientation information including a rotation angle;
determining that additional processing of the image is needed to identify routing information for the item based at least in part on the electronic text record;
in response to determining that the additional processing of the image is needed, adding the rotation angle to an image file of the image to generate an enhanced image;
transmitting the enhanced image for additional processing in a computing device; and
rotating, at the computing device, the enhanced image by the rotation angle prior to the additional processing by the computing device,
wherein the rotation angle is dynamically generated based on a relationship between a maximum pixel density and an average pixel density for a row or a column and
wherein the relationship is expressed as Equation (1) below:

rotation
⁢

=

30
*

density

w
-
avg

-

density

w
-
max

Assignees

• United States Postal Service

Inventors

• Simpson Ryan J

Classifications

• G06V30/424 Primary

  Postal images, e.g. labels or addresses on parcels or postal envelopes · CPC title

• G06V10/242

  by image rotation, e.g. by 90 degrees · CPC title

• G06V30/40

  Document-oriented image-based pattern recognition · CPC title

• B07C3/20

  Arrangements for facilitating the visual reading of addresses, e.g. display arrangements {coding stations} · CPC title

• G06K7/1404

  Methods for optical code recognition · CPC title