Methods and apparatus to improve detection and false alarm rate over image segmentation

Patent metadata

Field	Value
Publication number	US-9704260-B2
Application number	US-201514810989-A
Country	US
Kind code	B2
Filing date	Jul 28, 2015
Priority date	Jul 28, 2015
Publication date	Jul 11, 2017
Grant date	Jul 11, 2017

Abstract

Methods, apparatus, systems and articles of manufacture are disclosed herein. An example method to improve object detection and false alarm rate over image segmentation includes overlaying a first object of a first image onto a second image. A first score based on a first chamfer distance between first edges of the first object and second edges in the second image is determined. A second score corresponding to a second chamfer distance between the second edges and a mathematical representation of a plurality of shapes is determined, the second score representing a similarity between the second edges and the plurality of shapes observed simultaneously. A normalized score is determined by normalizing the first score based on the second score. A presence of the second object in the second image matching the first object is detected based on whether the normalized score satisfies a threshold score.

First claim

What is claimed is:

1. A method to detect an object in an image comprising:
retrieving a first image of a first object and a second image of a search area;
determining a first score based on a first chamfer distance between first edges of the first object in the first image and second edges in the second image of the search area, the first chamfer distance being a first measure of pixel distances between portions of the first edges and the second edges;
determining a second score corresponding to a second chamfer distance between the second edges of the search area and a mathematical representation of a plurality of shapes, the second score representing a similarity between the second edges of the search area and the plurality of shapes observed simultaneously, the second chamfer distance being a second measure of pixel distances between portions of the second edges and the mathematical representation of the plurality of shapes;
determining a normalized score by normalizing the first score based on the second score; and
detecting a presence of a second object in the second image of the search area matching the first object when the normalized score satisfies a threshold score.

2. The method of claim 1 , wherein the second score represents a lesser similarity between the second edges and the plurality of shapes observed simultaneously than a second similarity between the second edges and any one of the plurality of shapes.

3. The method of claim 1 , wherein the plurality of shapes corresponding to the mathematical representation are representative of background clutter.

4. The method of claim 1 , wherein the second edges of the second image are edges of a retail store shelf.

5. The method of claim 1 , wherein determining the second score further includes cross-correlating the second edges with a matrix of ones having a same size as the first object.

6. An apparatus to detect an object comprising:
an image retriever to retrieve a first image of a first object and a second image of a search area;
an image scorer to determine a first score based on a first chamfer distance between first edges of the first object in the first image and second edges in the second image of the search area, the first chamfer distance being a first measure of pixel distances between portions of the first edges and the second edges;
the image scorer further to determine a second score corresponding to a second chamfer distance between the second edges of the search area and a mathematical representation of a plurality of shapes, the second score representing a similarity between the second edges of the search area and the plurality of shapes observed simultaneously, the second chamfer distance being a second measure of pixel distances between portions of the second edges and the mathematical representation of the plurality of shapes;
a normalizer to determine a normalized score by normalizing the first score based on the second score; and
an object detector to determine whether a second object matching the first object is present in the second image of the search area based on whether the normalized score satisfies a threshold.

7. The apparatus of claim 6 , wherein the second score represents a lesser similarity between the second edges and the plurality of shapes observed simultaneously than a second similarity between the second edges and any one of the plurality of shapes.

8. The apparatus of claim 6 , wherein the plurality of shapes corresponding to the mathematical representation are representative of background clutter.

9. The apparatus of claim 6 , wherein the second edges of the second image are edges of a retail store shelf.

10. The apparatus of claim 6 , wherein determining the second score further includes cross-correlating the second edges with a matrix of ones having a same size as the first object.

11. A tangible computer readable storage medium comprising instructions, that when executed, cause a computing device to at least:
retrieve a first image of a first object and a second image of a search area;
determine a first score based on a first chamfer distance between first edges of the first object in the first image and second edges in the second image of the search area, the first chamfer distance being a first measure of pixel distances between portions of the first edges and the second edges;
determine a second score corresponding to a second chamfer distance between the second edges of the search area and a mathematical representation of a plurality of shapes, the second score representing a similarity between the second edges of the search area and the plurality of shapes observed simultaneously, the second chamfer distance being a second measure of pixel distances between portions of the second edges and the mathematical representation of the plurality of shapes;
determine a normalized score by normalizing the first score based on the second score; and
determine whether a second object matching the first object is present in the second image of the search area based on whether the normalized score satisfies a threshold.

12. The tangible computer readable storage medium of claim 11 , wherein the second score represents a lesser similarity between the second edges and the plurality of shapes observed simultaneously than a second similarity between the second edges and any one of the plurality of shapes.

13. The tangible computer readable storage medium of claim 11 , wherein the plurality of shapes corresponding to the mathematical representation are representative of background clutter.

14. The tangible computer readable storage medium of claim 11 , wherein the second edges of the second image are edges of a retail store shelf.

15. The tangible computer readable storage medium of claim 11 , wherein determining the second score further includes cross-correlating the second edges with a matrix of ones having a same size as the first object.

Assignees

• Nielsen Co Us Llc

Inventors

• Sullivan Jonathan
• Schiller Brian
• Terrazas Alejandro
• Xie Wei
• Bivins Michael Allen

Classifications

• G06T7/74 Primary

  involving reference images or patches · CPC title

• G06T7/12

  Edge-based segmentation · CPC title

• G06T7/143

  involving probabilistic approaches, e.g. Markov random field [MRF] modelling · CPC title

• G06T2207/30242

  Counting objects in image · CPC title

• G06T7/231 Primary

  using full search · CPC title