Image object recognition based on a feature vector with context information

Patent metadata

Field	Value
Publication number	US-9165220-B2
Application number	US-201213717706-A
Country	US
Kind code	B2
Filing date	Dec 18, 2012
Priority date	Dec 18, 2012
Publication date	Oct 20, 2015
Grant date	Oct 20, 2015

Abstract

Examples disclosed herein relate to image object recognition based on a feature vector with context information. A processor may create an expanded feature vector related to a first area of an image including context information related to the first area. The processor may determine the presence of an object in the image based on the feature vector and output information about the determined object.

First claim

The invention claimed is:

1. A computing system, comprising:
a processor to:
overlay a grid on an image, the grid having a plurality of positions;
determine a feature vector representing each position of the grid, the feature vector having a plurality of vector positions, each vector position comprising a vector item representing a characteristic of the area of the image covered by the feature vector,
the feature vector comprising context information at an end thereof, the context information including derivative information of the feature vector, the derivative information including a velocity and an acceleration for each vector item,
the velocity and the acceleration for each vector item describing a change of the vector item in an area in the image surrounding the position represented by the feature vector, and including derivatives in multiple directions within the area;

determine the presence of an object in the image based on the feature vector; and
output information about the determined object.

2. The computing system of claim 1 , wherein the processor is further to apply a dimensionality reduction technique to the expanded feature vector.

3. A method, comprising:
overlaying, by a processor, a grid on an image, the grid having a plurality of positions;
determining, by the processor, a feature vector representing each position of the grid, the feature vector having a plurality of vector positions, each vector position comprising a vector item representing a characteristic of the area of the image covered by the feature vector,
the feature vector comprising context information at an end thereof, the context information including derivative information of the feature vector, the derivative information including a velocity and an acceleration for each vector item,
the velocity and the acceleration for each vector item describing a change of the vector item in an area in the image surrounding the position represented by the feature vector, and including derivatives in multiple directions within the area;

determining, by the processor, the presence of an object in the image based on the feature vector; and
outputting, by the processor, information about the determined object.

4. The method of claim 3 , further comprising applying a dimensionality reduction method to the feature vector.

5. The method of claim 3 , further comprising:
determining a window surrounding the area;
concatenating feature vectors for areas within the window to the feature vector.

6. The method of claim 5 , wherein the window comprises a circular or rectangular window around the area.

7. A machine-readable non-transitory storage medium comprising instructions executable by a processor to:
overlay a grid on an image, the grid having a plurality of positions;
determine a feature vector representing each position of the grid, the feature vector having a plurality of vector positions, each vector position comprising a vector item representing a characteristic of the area of the image covered by the feature vector,
the feature vector comprising context information at an end thereof, the context information including derivative information of the feature vector, the derivative information including a velocity and an acceleration for each vector item,
the velocity and the acceleration for each vector item describing a change of the vector item in an area in the image surrounding the position represented by the feature vector, and including derivatives in multiple directions within the area;

determine the presence of an object in the image based on the feature vector; and
output information about the determined object.

8. The machine-readable non-transitory storage medium of claim 7 , further comprising instructions to apply a dimensionality reduction method to the stacked feature vector.

Assignees

• Hewlett Packard Development Co

Inventors

• Tang Henry “Hoa”

Classifications

• G06V10/50 Primary

  by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis · CPC title

• G06V10/768

  using context analysis, e.g. recognition aided by known co-occurring patterns · CPC title

• G06K9/4642

  Physics · mapped topic

• G06K9/72 Primary

  Physics · mapped topic