Method of image processing

Patent metadata

Field	Value
Publication number	US-10062013-B2
Application number	US-201615389858-A
Country	US
Kind code	B2
Filing date	Dec 23, 2016
Priority date	Dec 23, 2015
Publication date	Aug 28, 2018
Grant date	Aug 28, 2018

Abstract

According to an aspect of the present disclosure, there is provided a method of image processing. The method comprises receiving image data comprising a set of feature vectors of a first dimensionality, the feature vectors corresponding to a class of objects. A variable projection is applied to each feature vector in the set of feature vectors to generate a set of projected vectors of a second dimensionality. The method then comprises processing the set of projected vectors to generate a model for the class of objects. A projection is applied to the model to generate an object classification model, of the first dimensionality, for the class of objects.

First claim

The invention claimed is:

1. A method of processing image data, the method comprising:
receiving image data comprising a set of feature vectors of a first dimensionality, the feature vectors corresponding to a class of objects;
generating a set of projected vectors of a second dimensionality, lower than the first dimensionality, by applying a variable projection to each feature vector in the set of feature vectors;
processing the set of projected vectors to generate a model for the class of objects; and
applying a projection to the model to generate an object classification model, of the first dimensionality, for the class of objects.

2. The method according to claim 1 , wherein the variable projection comprises a random, or pseudorandom, projection.

3. The method according to claim 1 , wherein generating the set of projected vectors comprises generating a plurality of sets of projected vectors of the second dimensionality by applying each of a plurality of variable projections to each feature vector in the set of feature vectors.

4. The method according to claim 3 , wherein processing the set of projected vectors comprises processing each of the plurality of sets of projected vectors to generate a plurality of models for the class of objects.

5. The method according to claim 4 , wherein applying the projection to the model comprises applying a projection to each of the plurality of models to generate a plurality of object classification models, of the first dimensionality, for the class of objects.

6. The method according to claim 4 , comprising testing each model of the plurality of models and indicating an accuracy value for each model based on the testing.

7. The method according to claim 6 , comprising selecting a subset of the plurality of models based on the accuracy values of the models.

8. The method according to claim 7 , comprising applying a projection to each model in the subset of the plurality of models to generate a plurality of object classification models, of the first dimensionality, for the class of objects.

9. The method according to claim 1 , wherein the processing the set of projected vectors to generate a model for the class of objects uses a linear classification model.

10. The method according to claim 9 , wherein the linear classification model comprises at least one of:
a support vector machine;
a two neuron classifier; or
a Fisher discriminant.

11. The method according to claim 1 , the image data being captured by an image sensor.

12. The method according to claim 11 , the image data representing at least part of one or more images, wherein each of the one or more images comprises an object of the class of objects, wherein the image data is processed by a feature extractor to produce the image data comprising the set of feature vectors corresponding to the class of objects.

13. The method according to claim 1 , wherein receiving the image data comprises:
capturing image data representing at least part of an image using an image sensor;
on receiving an indication from an object detector that the image comprises an object of the class of objects, tracking the object and capturing, using the image sensor, image data corresponding to video frames comprising the object; and
processing the image data using a feature extractor to produce the image data comprising the set of feature vectors corresponding to the class of objects.

14. A non-transitory, computer-readable storage medium comprising a set of computer-readable instructions stored thereon which, when executed by at least one processor, cause the at least one processor to:
receive image data comprising a set of feature vectors of a first dimensionality, the feature vectors corresponding to a class of objects;
generate a set of projected vectors of a second dimensionality, lower than the first dimensionality, by applying a variable projection to each feature vector in the set of feature vectors;
process the set of projected vectors to generate a model for the class of objects; and
apply a projection to the model to generate an object classification model, of the first dimensionality, for the class of objects.

15. A computer vision apparatus comprising a classifier, the classifier comprising at least one of a plurality of object classification models generated by:
receiving image data comprising a set of feature vectors of a first dimensionality, the feature vectors corresponding to a class of objects;
generating a set of projected vectors of a second dimensionality, lower than the first dimensionality, by applying a variable projection to each feature vector in the set of feature vectors;
processing the set of projected vectors to generate a model for the class of objects; and
applying a projection to the model to generate an object classification model, of the first dimensionality, for the class of objects.

16. The computer vision apparatus according to claim 15 , comprising a feature extractor configured to:
receive image data representing at least part of an image; and
produce image data comprising a plurality of feature vectors.

17. The computer vision apparatus according to claim 16 , comprising an image sensor, wherein the image data representing at least part of an image is captured by the image sensor.

18. The computer vision apparatus according to claim 16 , wherein the classifier is configured to:
process the image data comprising the plurality of feature vectors; and
determine, using the at least one of the plurality of object classification models, whether the image data comprises an object in the class of objects corresponding to the object classification models.

19. The computer vision apparatus according to claim 18 , wherein the classifier is configured to indicate whether the image data comprises an object in the class of objects corresponding to the object classification models.

Assignees

• Apical Ltd
• Apical Ltd

Inventors

• Romanenko Ilya
• TYUKIN Ivan
• Gorban Alexander
• SOFEIKOV Konstantin

Classifications

• G06T7/70

  Determining position or orientation of objects or cameras (camera calibration G06T7/80) · CPC title

• G06F18/211 Primary

  Selection of the most significant subset of features · CPC title

• G06F18/24 Primary

  Classification techniques · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06F18/2431

  Multiple classes · CPC title