Adaptive edge-like feature selection during object detection

Patent metadata

Field	Value
Publication number	US-9684831-B2
Application number	US-201514625588-A
Country	US
Kind code	B2
Filing date	Feb 18, 2015
Priority date	Feb 18, 2015
Publication date	Jun 20, 2017
Grant date	Jun 20, 2017

Abstract

A method of recognizing an object of interest in an image includes extracting a first set of features from within the image. Each extracted feature in the first set of features is then categorized as either blob-like or edge-like. A second set of features is then taken from the first set, where a number of the edge-like features to include in the second set of features is based on a relative number of edge-like features to blob-like features included in the first set of extracted features. An object of interest within the image is detected according to the second set of features.

First claim

What is claimed is:

1. A computer-implemented method of recognizing an object of interest in an image, the method comprising:
extracting a first set of features from within the image;
determining whether each feature included in the first set of features is a blob-like feature or an edge-like feature;
determining a second set of features taken from the first set of features, wherein a number of the edge-like features included in the second set of features is based on a ratio of edge-like features to blob-like features included in the first set of features; and
determining whether the object of interest is in the image according to the second set of features.

2. The computer-implemented method of claim 1 , wherein determining whether the object of interest is in the image comprises querying a feature database using the second set of features.

3. The computer-implemented method of claim 1 , wherein extracting the first set of features from within the image includes applying a Scale-invariant feature transform (SIFT) algorithm to the image.

4. The computer-implemented method of claim 1 , wherein determining whether a feature is a blob-like feature or an edge-like feature includes:
computing a first eigenvalue and a second eigenvalue of the feature;
computing a ratio of the first eigenvalue to the second eigenvalue; and
comparing the ratio to a threshold.

5. The computer-implemented method of claim 1 , wherein the number of the edge-like features included in the second set of features is a function of a ratio of the number of edge-like features to the number of blob-like features included in the first set of features.

6. The computer-implemented method of claim 5 , wherein the function is a piecewise function such that the number of the edge-like features included in the second set of features is zero when the ratio is less than a lower threshold and all the edge-like features of the first set are included in the second set when the ratio is greater than an upper threshold.

7. The computer-implemented method of claim 6 , wherein the number of edge-like features included in the second set of features increases linearly when the ratio is between the lower and upper threshold.

8. The computer-implemented method of claim 1 , further comprising:
segmenting the image into a plurality of regions; and
selecting the features from the first set that are to be included in the second set of features such that the second set of features are distributed among the plurality of regions.

9. The computer-implemented method of claim 1 , further comprising:
capturing the image with a camera; and
updating a pose of the camera in response to determining that the object of interest is in the image.

10. A device for recognizing an object of interest in an image, the device comprising:
memory adapted to store program code for recognizing an object of interest in the captured image; and
at least one processing unit connected to the memory, wherein the program code is configured to cause the at least one processing unit to:
extract a first set of features from within the image;
determine whether each feature included in the first set of features is a blob-like feature or an edge-like feature;
determine a second set of features taken from the first set of features, wherein a number of the edge-like features included in the second set of features is based on a ratio of edge-like features to blob-like features included in the first set of features; and
determine whether the object of interest is in the image according to the second set of features.

11. The device of claim 10 , wherein the instructions to determine whether the object of interest is in the image comprises querying a feature database using the second set of features.

12. The device of claim 10 , wherein the instructions to extract the first set of features from within the image includes instructions to apply a Scale-invariant feature transform (SIFT) algorithm to the image.

13. The device of claim 10 , wherein the instructions to determine whether a feature is a blob-like feature or an edge-like feature includes instructions to:
compute a first eigenvalue and a second eigenvalue of the feature;
compute a ratio of the first eigenvalue to the second eigenvalue; and
compare the ratio to a threshold.

14. The device of claim 10 , wherein the number of the edge-like features included in the second set of features is a function of a ratio of the number of edge-like features to the number of blob-like features included in the first set of features.

15. The device of claim 14 , wherein the function is a piecewise function such that the number of the edge-like features included in the second set of features is zero when the ratio is less than a lower threshold and all the edge-like features of the first set are included in the second set when the ratio is greater than an upper threshold.

16. The device of claim 15 , wherein the number of edge-like features included in the second set of features increases with increases in the ratio when the ratio is between the lower and upper threshold.

17. The device of claim 10 , wherein the program code further comprises instructions to:
segment the image into a plurality of regions; and
select the features from the first set that are to be included in the second set of features such that the second set of features are distributed among the plurality of regions.

18. The device of claim 10 , wherein the program code further comprises instructions to:
capture the image with a camera; and
update a pose of the camera in response to determining that the object of interest is in the image.

19. A non-transitory computer-readable medium including program code stored thereon for recognizing an object of interest in an image, the program code comprising instructions to:
extract a first set of features from within the image;
determine whether each feature included in the first set of features is a blob-like feature or an edge-like feature;
determine a second set of features taken from the first set of features, wherein a number of the edge-like features included in the second set of features is based on a ratio of edge-like features to blob-like features included in the first set of features; and
determine whether the object of interest is in the image according to the second set of features.

20. The non-transitory computer-readable medium of claim 19 , wherein the instructions for determining whether the object of interest is in the image comprises querying a feature database using the second set of features, and wherein the instructions for extracting the first set of features from within the image includes applying a Scale-invariant feature transform (SIFT) algorithm to the image.

21. The non-transitory computer-readable medium of claim 19 , wherein the instructions to determine whether a feature is a blob-like feature or an edge-like feature includes instructions to:
compute a first eigenvalue and a second eigenvalue of the feature;
compute a ratio of the first eigenvalue to the second eigenvalue; and
compare the ratio to a threshold.

22. The non-transitory computer-readable medium of claim 19 , wherein the number of the edge-like features included in the second set of features is a function of a ratio of the number of edge-like features to the number of blob-like features included in the first set of features.

23. The non-transitory computer-readable medium of

Assignees

• Qualcomm Inc

Inventors

• Krishnamoorthi Raghuraman
• Vrcelj Bojan

Classifications

• G06V10/771

  Feature selection, e.g. selecting representative features from a multi-dimensional feature space · CPC title

• G06V10/462 Primary

  Salient features, e.g. scale invariant feature transforms [SIFT] · CPC title

• G06F18/211

  Selection of the most significant subset of features · CPC title

• G06T7/13

  Edge detection · CPC title

• G06K9/4604

  Physics · mapped topic