Method and apparatus for recognizing client feature, and storage medium

• US 9,697,440 B2
• Filed: 09/27/2013
• Issued: 07/04/2017
• Est. Priority Date: 11/27/2012
• Status: Active Grant

• Current Assignee: Tencent Technology Company Limited (Tencent Holdings Limited)
• Original Assignee: Tencent Technology Company Limited (Tencent Holdings Limited)
• Inventors: Xiong, Pengfei, Liu, Hailong, Chen, Bo
• Primary Examiner(s): Rush, Eric
• Application Number:
  US14/647,460
• Publication Number:
  US 20150310308A1
• Time in Patent Office: 1,376 Days
• Field of Search: 382100, 382115-119, 382124, 382156, 382157, 382159, 382160, 382181, 382224
• CPC Class Codes: G06F 18/213 Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
• CPC Class Codes: G06F 18/2132 based on discrimination criteria, e.g. discriminant analysis
• CPC Class Codes: G06F 18/2135 based on approximation criteria, e.g. principal component analysis
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06F 18/2155 characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06F 18/2433 Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
• CPC Class Codes: G06V 10/758 Involving statistics of pixels or of feature values, e.g. histogram matching
• CPC Class Codes: G06V 10/774 Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06V 40/1365 Matching; Classification
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

In the present disclosure, a client feature and a pre-stored template image feature are obtained; the obtained client feature and template image feature are projected according to a preset projection matrix, to generate a projection feature pair, where the projection matrix is formed by training of a first template image feature of a same object and a second template image feature of a different object; and similarity calculation is performed on the projection feature pair according to a preset similarity calculation rule, to generate a similarity result and prompt the similarity result to a client.

Claims

1. A method for recognizing a client feature, comprising:

obtaining a client feature and a pre-stored template image feature;

projecting the obtained client feature and the obtained pre-stored template image feature according to a preset projection matrix, to generate a projection feature pair, the preset projection matrix being formed by training an energy function using first template image features of a same object and second template image features of different objects;

performing similarity calculation on the projection feature pair according to a preset similarity calculation rule, to generate a similarity result; and

prompting the generated similarity result;

wherein the preset similarity calculation rule comprises a similarity probability function, and the similarity probability function is generated according to a preset similarity metric function; and

the energy function comprises the preset projection matrix and the preset projection matrix is obtained by training the energy function until similarity between the first template image features of the same object is the greatest and similarity between the second template image features of the different objects is the smallest;

wherein a formula of the similarity probability function is;

QCS(x_i,x_j)=(1 exp(dist(x_i,x_j)−

b))<sup>−

1</sup>

wherein, (x_i, x_j) is an image feature pair formed by two different template image features, b is a metric parameter, QCS is the similarity probability function, dist is the preset similarity metric function, and exp is an exponential function with a base being a natural logarithm e.

2. The method for recognizing a client feature according to claim 1, wherein the step of generating a similarity result comprises:

performing similarity calculation on the projection feature pair according to the similarity probability function, to generate a similarity probability;

determining whether the similarity probability is greater than or equal to a preset threshold;

if the similarity probability is greater than or equal to the threshold, determining that the client feature and the pre-stored template image feature belong to a same category; and

if the similarity probability is less than the threshold, determining that the client feature and the pre-stored template image feature do not belong to a same category.

3. The method for recognizing a client feature according to claim 1, wherein before forming the preset projection matrix, the method further comprises:

when the client feature and the pre-stored template image feature are obtained, generating the preset similarity metric function, which is used to collect statistics on the pre-stored template image feature;

generating the similarity probability function according to the preset similarity metric function;

generating the energy function according to the preset similarity metric function and the similarity probability function.

4. The method for recognizing a client feature according to claim 3, wherein the step of forming the preset projection matrix comprises:

training the energy function in a manner of gradient descent, to form the preset projection matrix.

5. An apparatus for recognizing a client feature, comprising:

memory;

one or more processors; and

one or more programs stored in the memory and configured for execution by the one or more processors, the one or more programs comprising the following instruction modules;

a feature obtaining module, configured to obtain a client feature, and obtain a pre-stored template image feature;

a projection module, configured to project the obtained client feature and the obtained pre-stored template image feature according to a preset projection matrix, to generate a projection feature pair, the preset projection matrix being formed by training an energy function using first template image features of a same object and second template image features of different objects;

a similarity calculation module, configured to perform similarity calculation on the projection feature pair according to a preset similarity calculation rule, to generate a similarity result; and

a prompting module, configured to prompt the generated similarity result;

wherein the preset similarity calculation rule comprises a similarity probability function, and the similarity probability function is generated according to a preset similarity metric function; and

the energy function comprises the preset projection matrix and the preset projection matrix is obtained by training the energy function until similarity between the first template image features of the same object is the greatest and similarity between the second template image features of the different objects is the smallest;

wherein a formula of the similarity probability function is;

QCS(x_i,x_j)=(1 exp(dist(x_i,x_j)−

b))<sup>−

1</sup>

wherein, (x_i, x_j) is an image feature pair formed by two different template image features, b is a metric parameter, QCS is the similarity probability function, dist is the preset similarity metric function, and exp is an exponential function with a base being a natural logarithm e.

6. The apparatus for recognizing a client feature according to claim 5, wherein the similarity calculation module comprises:

a probability generating module, configured to perform similarity calculation on the projection feature pair according to the similarity probability function, to generate a similarity probability; and

a determining module, configured to determine whether the similarity probability is greater than or equal to a preset threshold;

if the similarity probability is greater than or equal to the preset threshold, determine that the client feature and the pre-stored template image feature belong to a same category; and

if the similarity probability is less than the preset threshold, determine that the client feature and the pre-stored template image feature do not belong to a same category.

7. The apparatus for recognizing a client feature according to claim 5, further comprising:

a metric function generating module, configured to;

generate the preset similarity metric function, and collect statistics on the pre-stored template image feature by using the preset similarity metric function;

a probability function generating module, configured to generate the similarity probability function according to the preset similarity metric function;

an energy function generating module, configured to generate the energy function according to the preset similarity metric function and the similarity probability function; and

a projection matrix generating module, configured to train the energy function using the first template image features of the same object and the second template image features of the different objects.

8. The apparatus for recognizing a client feature according to claim 7, wherein the projection matrix generating module is further configured to train the energy function in a manner of gradient descent, to form the preset projection matrix.

9. A non-transitory storage medium, having a processor executable instruction stored therein, and the processor executable instruction being used to enable a processor to complete the following operations:

obtaining a client feature and a pre-stored template image feature;

projecting the obtained client feature and the obtained pre-stored template image feature according to a preset projection matrix, to generate a projection feature pair, the preset projection matrix being formed by training an energy function using first template image features of a same object and second template image features of different objects;

performing similarity calculation on the projection feature pair according to a preset similarity calculation rule, to generate a similarity result; and

prompting the generated similarity result;

wherein the preset similarity calculation rule comprises a similarity probability function, and the similarity probability function is generated according to a preset similarity metric function; and

the energy function comprises the preset projection matrix and the preset projection matrix is obtained by training the energy function until similarity between the first template image features of the same objet is the greatest and similarity between the second template image features of the different objects is the smallest;

wherein a formula of the similarity probability function is;

QCS(x_i,x_j)=(1 exp(dist(x_i,x_j)−

b))<sup>−

1</sup>

wherein, (x_i, x_j) is an image feature pair formed by two different template image features, b is a metric parameter, QCS is the similarity probability function, dist is the preset similarity metric function, and exp is an exponential function with a base being a natural logarithm e.

10. The non-transitory storage medium according to claim 9, wherein the processor executable instruction stored in the storage medium is further used to enable the processor to complete the following operations:

when similarity calculation is performed on the projection feature pair according to the preset similarity calculation rule to generate the similarity result, first performing similarity calculation on the projection feature pair according to the similarity probability function, to generate a similarity probability, then determining whether the similarity probability is greater than or equal to a preset threshold;

if the similarity probability is greater than or equal to the threshold, determining that the client feature and the pre-stored template image feature belong to a same category; and

if the similarity probability is less than the threshold, determining that the client feature and the pre-stored template image feature do not belong to a same category.

11. The non-transitory storage medium according to claim 9, wherein the processor executable instruction stored in the storage medium is further used to enable the processor to complete the following operations:

when the client feature and the pre-stored template image feature are obtained, generating the preset similarity metric function, which is used to collect statistics on the pre-stored template image feature;

generating the similarity probability function according to the preset similarity metric function;

generating the energy function according to the preset similarity metric function and the similarity probability function; and

training the first template image features of the same object and the second template image features of the different objects by using the energy function, to form the preset projection matrix.

12. The non-transitory storage medium according to claim 11, wherein the processor executable instruction stored in the storage medium is further used to enable the processor to complete the following operations:

when the first template image features of the same object and the second template image features of the different objects are trained by using the energy function, to form the preset projection matrix, training the energy function in a manner of gradient descent, to form the preset projection matrix.