Method and apparatus for image processing to avoid counting shelf edge promotional labels when counting product labels

• US 9,424,482 B2
• Filed: 10/31/2013
• Issued: 08/23/2016
• Est. Priority Date: 06/12/2013
• Status: Active Grant

• Current Assignee: Symbol Technologies, LLC (Zebra Technologies Corporation)
• Original Assignee: Symbol Technologies, LLC (Zebra Technologies Corporation)
• Inventors: Patel, Ankur R., Super, Boaz J.
• Primary Examiner(s): Desire, Gregory M
• Application Number:
  US14/068,495
• Publication Number:
  US 20150117788A1
• Time in Patent Office: 1,027 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/583 using metadata automatically derived from the content
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 18/23 Clustering techniques
• CPC Class Codes: G06F 18/2321 using statistics or function optimisation, e.g. modelling of probability density functions
• CPC Class Codes: G06F 18/23211 with adaptive number of clusters
• CPC Class Codes: G06Q 10/087 Inventory or stock management, e.g. order filling, procurement or balancing against orders
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/763 Non-hierarchical techniques, e.g. based on statistics of modelling distributions
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69

Abstract

A method and apparatus for image processing to avoid counting shelf edge promotional labels when counting product labels. Shelf edges are identified by detecting shelf edge content in a captured image by comparing the image to reference images of shelf edge content. Detected occurrences of shelf edge content are demarcated using a geometric pattern having corners at coordinates corresponding to positions around the identified shelf edge content. Corresponding corners are grouped into clusters, and the clusters are analyzed to define the upper and lower bounds of a shelf region in the image.

Claims

1. A method for identifying edge regions in an image, comprising:

receiving an image containing at least one edge region, the edge region containing patterned content;

identifying, using a processor, individual occurrences of the patterned content by comparing sections of the image to a reference image of the patterned content;

demarcating, using the processor, each identified individual occurrence of the patterned content in the image with a commonly oriented geometric pattern having at least one corner associated with a first side of the corresponding pattern content and at least one corner associated with a second side of the corresponding patterned content; and

splitting, using the processor, the corners into first and second sets, with the first set indicative of a first boundary of an edge and the second set indicative of a second boundary of the edge; and

generating, using the processor, an indication of the first and second boundaries to enable exclusion of the edge from a recognition process.

2. The method of claim 1, wherein demarcating each individual occurrence of the patterned content with the commonly oriented geometric pattern comprises demarcating each individual occurrence of the patterned content with an estimated quadrilateral.

3. The method of claim 1, wherein comparing sections of the image to the reference image of the patterned content comprises:

determining interest point matches between the image and the reference image; and

applying a geometric model fitting and verification algorithm to estimate a quadrilateral based on the interest point matches.

4. The method of claim 1, further comprising fitting a first line to define the first boundary and a second line to define the second boundary, and wherein a space between the first line and the second line is identified as the edge.

5. The method of claim 1, further comprising clustering corners of different occurrences of the patterned content by selecting a kernel bandwidth in a horizontal direction of the image that is larger than a kernel bandwidth in a vertical direction of the image.

6. The method of claim 5, wherein clustering the corners further comprises associating corners to a cluster based on minimizing a distance measure between vertical co-ordinates of the corners and a center of the cluster.

7. The method of claim 1, wherein generating the indication of the first and second boundaries comprises generating a processed image having the edge obscured.

8. The method of claim 1, wherein generating the indication of the first and second boundaries comprises generating metadata that identifies the edge in the image, wherein the recognition process is to use the metadata to avoid analyzing the edge.

9. A product shelf image processing system, comprising:

a camera that captures an image containing at least one product shelf having an edge upon which shelf edge content is present; and

a processor to;

receive the image from the camera;

identify individual occurrences of the shelf edge content by comparing sections of the image to at least one reference image of the shelf edge content;

demarcate each identified individual occurrence of the shelf edge content in the image with a commonly oriented geometric pattern having at least a first corner and a second corner opposing the first corner; and

split the corners into first and second sets, with the first set indicative a first boundary of the shelf edge and the second set indicative of a second boundary of the shelf edge; and

generate an indication of the first and second boundaries to enable exclusion of the edge from a recognition process.

10. The product shelf image processing system of claim 9, wherein the processor is to not count shelf edge content as product labels while it recognizes and to count product labels in the image.

11. The product shelf image processing system of claim 9, where the processor is to determine interest point matches between the image and the reference image, and apply a geometric model fitting and verification algorithm to estimate a quadrilateral based on the interest point matches.

12. The product shelf image processing system of claim 9, wherein generating the indication of the first and second boundaries comprises generating a processed image having the edge obscured.

13. The product shelf image processing system of claim 9, wherein generating the indication of the first and second boundaries comprises generating metadata that identifies the edge in the image, wherein the recognition processing is to use the metadata to avoid analyzing the edge.