Image identification system and image identification method

Patent metadata

Field	Value
Publication number	US-11270166-B2
Application number	US-201916454488-A
Country	US
Kind code	B2
Filing date	Jun 27, 2019
Priority date	Dec 27, 2016
Publication date	Mar 8, 2022
Grant date	Mar 8, 2022

Abstract

The embodiment of the present disclosure provides an image identification system and an image identification method, and relates to the field of identification technology for improving the classification accuracy of the existing image classification model. The system includes an identification module, a retrieval module and a training module, wherein the identification module is configured to identify an image sample by using an image classification model to obtain an image category confidence coefficient of the image sample; the retrieval module is configured to retrieve similar artificial identification examples of the image sample when it is determined that the image category confidence coefficient is less than a first predetermined threshold; and the training module is configured to train the image classification model according to training samples in a training sample library; wherein the training samples include an artificial identification example and an image retrieval example with a high confidence coefficient.

First claim

The invention claimed is:

1. An image identification method, comprising:
identifying an image sample by using an image classification model to obtain an image category confidence coefficient of the image sample;
retrieving similar artificial identification examples of the image sample when it is determined that the image category confidence coefficient is less than a first predetermined threshold, and using an identification result of a target artificial identification example having a highest confidence coefficient among the similar artificial identification examples as the identification result of the image sample;
obtaining image features of the image sample after identifying the image sample by using the image classification model; and
retrieving similar artificial identification examples of the image sample for analysis to obtain the identification result of the image sample, wherein the retrieving for analysis to obtain the identification result of the image sample comprises:
retrieving the similar artificial identification examples of the image sample from an image sample library according to the image features of the image sample obtained by an identification module, wherein a similarity between the image features contained in the identification results of the similar artificial identification examples and the image features of the image sample is greater than a second predetermined threshold, and wherein the image sample library is used for storing one or more artificial identification examples.

2. The method according to claim 1 , wherein after using the identification result of the target artificial identification example having the highest confidence coefficient among the similar artificial identification examples as the identification result of the image sample, the method further comprises:
storing the image sample in a training sample library of training samples as an image retrieval example with a high confidence coefficient when the image category confidence coefficient of the image sample is greater than or equal to the first predetermined threshold;
wherein the training samples comprise an artificial identification example and the image retrieval example with the high confidence coefficient.

3. The method according to claim 2 , further comprising:
training the image classification model by using training samples in the training sample library, wherein the training samples comprise the artificial identification example and the image retrieval example with the high confidence coefficient.

4. The method according to claim 3 , further comprising:
judging whether a sample size of the training samples stored in the training sample library satisfies a second predetermined threshold; and
if yes, obtaining the identification results of the training samples from the training sample library, and training the image classification model in the identification module based thereon.

5. The method according to claim 1 , wherein after identifying the image sample by using the image classification model to obtain the identification result, the method further comprises:
outputting the identified identification result of the image sample when it is determined that the image category confidence coefficient of the image sample is greater than or equal to the first predetermined threshold.

6. An image identification system, wherein the system comprises:
a memory; and
a processor,
wherein the memory is configured to store a computer execution code,
wherein the processor, upon execution of the computer execution code, is configured to:
identify an image sample by using an image classification model to obtain an image category confidence coefficient of the image sample;
retrieve similar artificial identification examples of the image sample it is determined that the image category confidence coefficient is less than a first predetermined threshold, and using an identification result of a target artificial identification example having a highest confidence coefficient among the similar artificial identification examples as the identification result of the image sample;
obtaining image features of the image sample after identifying the image sample by using the image classification model; and
retrieving similar artificial identification examples of the image sample for analysis to obtain the identification result of the image sample, wherein the retrieving for analysis to obtain the identification result of the image sample comprises:
retrieving the similar artificial identification examples of the image sample from an image sample library according to the image features of the image sample obtained by an identification module, wherein a similarity between the image features contained in the identification results of the similar artificial identification examples and the image features of the image sample is greater than a second predetermined threshold, and wherein the image sample library is used for storing one or more artificial identification examples.

7. A non-transitory computer storage medium configured to store computer software instructions used by an image identification system comprising a memory and a processor, wherein the computer software instructions, upon execution by the processor, results in the processor being configured to:
identify an image sample by using an image classification model to obtain an image category confidence coefficient of the image sample; and
retrieve similar artificial identification examples of the image sample it is determined that the image category confidence coefficient is less than a first predetermined threshold, and using an identification result of a target artificial identification example having a highest confidence coefficient among the similar artificial identification examples as the identification result of the image sample;
obtaining image features of the image sample after identifying the image sample by using the image classification model; and
retrieving similar artificial identification examples of the image sample for analysis to obtain the identification result of the image sample, wherein the retrieving for analysis to obtain the identification result of the image sample comprises:
retrieving the similar artificial identification examples of the image sample from an image sample library according to the image features of the image sample obtained by an identification module, wherein a similarity between the image features contained in the identification results of the similar artificial identification examples and the image features of the image sample is greater than a second predetermined threshold, and wherein the image sample library is used for storing one or more artificial identification examples.

Assignees

• Cloudminds Shanghai Robotics Co Ltd
• Cloudminds Shenzhen Robotics Systems Co Ltd

Inventors

• Chai Lunshao
• Lian Shiguo

Classifications

• G06V20/35

  Categorising the entire scene, e.g. birthday party or wedding scene · CPC title

• G06V10/774

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

• G06V10/772 Primary

  Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries · CPC title

• G06F18/24

  Classification techniques · CPC title

• G06F18/2415 Primary

  based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate · CPC title