Intelligent image recognition and classification system based on deep learning

Abstract

The invention discloses an intelligent image recognition and classification system based on deep learning. The system comprises an image acquisition module, a first feature extraction module, a first classification module, a second feature extraction module, a second classification module and a storage module. When the method is implemented, the first feature extraction module and the second feature extraction module are arranged, the feature information of the images can be collected step by step, and then the purpose of step-by-step classification is achieved, that is, in the first step, the images are classified for the first time, and then the number of the images entering the second feature extraction module is reduced; therefore, the operation pressure of the second feature extraction module is relieved, rapid recognition and classification of the images are facilitated, the first feedback module can continuously complete and supplement the first feature extraction module, and then the first classification module can recognize and classify more images, so that the image classification efficiency is improved. And the number of images identified and classified by the second classification module is reduced.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to an intelligent image recognition and classification system based on deep learning. Background Art

[0002] With the continuous development of electronic information technology, image recognition and classification can quickly display images that meet the requirements when users search for images.

[0003] When processing images, existing image processing systems identify the content of images based on the system's algorithms, and the image recognition process relies on the system's fast computing power to achieve fast recognition. However, when the system's computing power is weak, the recognition speed slows down. How to quickly recognize images without relying on the system's computing power has become a problem that urgently needs to be solved in this field. Summary of the invention

[0004] To this end, the present application provides an intelligent image recognition and classification system based on deep learning to solve the problem in the prior art that chips with stronger computing power must be used to quickly recognize images.

[0005] In order to achieve the above objectives, this application provides the following technical solutions:

[0006] In a first aspect, an intelligent image recognition and classification system based on deep learning includes an image acquisition module, a first feature extraction module, a first classification module, a second feature extraction module, a second classification module, and a storage module;

[0007] The output end of the image acquisition module is connected to the input end of the first feature extraction module, the output end of the first feature extraction module is connected to the input end of the first classification module, the output end of the first classification module is connected to the input end of the second feature extraction module, the output end of the second feature extraction module is connected to the input end of the second classification module, and the output end of the second classification module is connected to the input end of the storage module;

[0008] The image acquisition module is used to acquire images, and the first feature extraction module is used to extract first feature information from the image. After the first feature information is extracted from the acquired image, the first classification module can perform a first classification on the image according to the extracted first feature information. When the image acquired by the image acquisition module has not been extracted by the first feature extraction module with the first feature information, this part of the image is marked as a remaining image, and the remaining image enters the second feature extraction module, and the second feature extraction module extracts the second feature information. The second classification module performs a second classification on the remaining image according to the second feature information, and the storage module can save the images classified by the first classification and the second classification;

[0009] It also includes a first feedback module, the output end of the second classification module is connected to the input end of the first feedback module, and the output end of the first feedback module is connected to the input end of the first classification module;

[0010] After the second classification module classifies the image, the first feedback module can extract feature information on the second classified image and input it into the first classification module.

[0011] The first classification module and the second classification module classify the image into n categories, and both the first classification module and the second classification module include n standard features. The extracted first feature information and the second feature information can be compared with the standard features to determine the classification of the image.

[0012] When the first classification module classifies the image, the first feature information is compared with the standard feature, and when the similarity between the first feature information and the standard feature is greater than 90%, the image corresponding to the first feature information is classified into the category of the standard feature.

[0013] The first feedback module comprises a fuzzy feature extraction unit and a fuzzy feature comparison unit. After the second classification module classifies the remaining image, the fuzzy feature extraction unit can extract feature information from the remaining image and mark the extracted feature information as a fuzzy feature. The fuzzy feature comparison unit can compare the fuzzy feature with the standard feature, and when the similarity between the fuzzy feature and the standard feature is greater than 80%, the fuzzy feature and the standard feature information are saved in parallel.

[0014] It also includes a preprocessing module, the output end of the image acquisition module is connected to the input end of the preprocessing module, and the output end of the preprocessing module is connected to the first feature extraction module. After the image acquisition module takes the image, the preprocessing module can perform noise reduction on the image.

[0015] It also includes a manual review module, the output end of the second classification module is connected to the input end of the manual review module, after the second classification module classifies the image, part of the remaining images are not classified by the second classification module, and this part of the images is marked as uncertain images. The uncertain images can enter the manual review module, and the manual review module can classify the images manually.

[0016] It also includes a second feedback module, the input end of the second feedback module is connected to the output end of the manual review module, and the output end of the second feedback module is connected to the input end of the second classification module. When the manual review module reviews the uncertain image, the manual review module can also manually mark the feature information of the uncertain image, and record the manually marked feature information as manual feature information. The second feedback module extracts the manual feature information and saves it in the second classification module.

[0017] Compared with the prior art, this application has at least the following beneficial effects:

[0018] When the present invention is implemented, by providing a first feature extraction module and a second feature extraction module, the feature information of the image can be collected step by step, thereby achieving the purpose of step-by-step classification, that is, in the first step, the image is classified for the first time, thereby reducing the number of images entering the second feature extraction module, thereby reducing the operating pressure of the second feature extraction module, and facilitating rapid recognition and classification of the image, and the first feedback module can continuously improve and supplement the first feature extraction module, thereby enabling the first classification module to recognize and classify more images, thereby reducing the number of images recognized and classified by the second classification module;

[0019] The first classification module can perform initial recognition and classification of the image based on the determined feature information. When the feature information on the image is not clear enough, the second classification module will perform in-depth recognition and classification, thereby alleviating the computational pressure of the second classification module to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more intuitively illustrate the prior art and the present application, exemplary drawings are given below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing the present application; for example, those skilled in the art are capable of easily making conventional adjustments or further optimizations to the addition / reduction / attribution division, specific shapes, positional relationships, connection methods, dimensional ratios, etc. of certain units (components) based on the technical concepts and exemplary drawings disclosed in the present application.

[0021] Figure 1 A module diagram of an intelligent image recognition and classification system based on deep learning provided in Example 1 of the present application. DETAILED DESCRIPTION

[0022] The present application is further described below in detail through specific embodiments in conjunction with the accompanying drawings.

[0023] An intelligent image recognition and classification system based on deep learning,

[0024] It includes an image acquisition module, which acquires the original image.

[0025] The first feature extraction module extracts features from the image.

[0026] The first classification module classifies the image according to the features extracted by the first feature extraction module.

[0027] The second feature extraction module extracts features from the image.

[0028] The second classification module classifies the image again according to the features extracted by the second feature extraction module.

[0029] The storage module stores the classified images so that users can extract images according to categories;

[0030] The output end of the image acquisition module is connected to the input end of the first feature extraction module, the output end of the first feature extraction module is connected to the input end of the first classification module, the output end of the first classification module is connected to the input end of the second feature extraction module, the output end of the second feature extraction module is connected to the input end of the second classification module, and the output end of the second classification module is connected to the input end of the storage module;

[0031] The image acquisition module is used to acquire images, and the first feature extraction module is used to extract first feature information from the image. After the first feature information is extracted from the acquired image, the first classification module can perform a first classification on the image according to the extracted first feature information. When the image acquired by the image acquisition module has not been extracted by the first feature extraction module with the first feature information, this part of the image is marked as a remaining image, and the remaining image enters the second feature extraction module, and the second feature extraction module extracts the second feature information. The second classification module performs a second classification on the remaining image according to the second feature information, and the storage module can save the images classified by the first classification and the second classification;

[0032] It also includes a first feedback module, the output end of the second classification module is connected to the input end of the first feedback module, and the output end of the first feedback module is connected to the input end of the first classification module;

[0033] After the second classification module classifies the image, the first feedback module can extract feature information on the second classified image and input it into the first classification module.

[0034] When the present invention is implemented, by providing a first feature extraction module and a second feature extraction module, the feature information of the image can be collected step by step, thereby achieving the purpose of step-by-step classification. That is, in the first step, the image is classified for the first time, thereby reducing the number of images entering the second feature extraction module, thereby alleviating the operating pressure of the second feature extraction module, and facilitating rapid identification and classification of the image. The first feedback module can continuously improve and supplement the first feature extraction module, thereby enabling the first classification module to identify and classify more images, thereby reducing the number of images identified and classified by the second classification module.

[0035] The first classification module can perform initial recognition and classification of the image based on the determined feature information. When the feature information on the image is not clear enough, the second classification module will perform in-depth recognition and classification, thereby alleviating the computational pressure of the second classification module to a certain extent.

[0036] The first classification module and the second classification module classify the image into n categories, and both the first classification module and the second classification module include n standard features. The standard features in the first classification module are different from the standard features in the second classification module. The standard features in the second classification module are more detailed than the standard features in the first classification module. The extracted first feature information and the second feature information can be compared with the standard features to determine the classification of the image. After the first feature information and the second feature information are extracted, the first feature information and the second feature information can be compared with the n standard feature information respectively to determine the classification of the image.

[0037] In some embodiments, when the first classification module classifies an image, the first feature information is compared with the standard feature, and when the similarity between the first feature information and the standard feature is greater than 90%, the image corresponding to the first feature information is classified into the category of the standard feature.

[0038] When the first feature extraction module extracts feature information, it extracts feature information containing standard feature information or feature information similar to the standard feature information to obtain first feature information, and then compares the first feature information with the standard feature information. When the similarity between the two is greater than 90%, the first feature information is judged to be standard feature information.

[0039] In other embodiments, when the similarity between the first feature information and the standard feature is greater than A%, the image corresponding to the first feature information is classified into the category of the standard feature, and the value of A is a numerical value defined by the user.

[0040] The first feedback module comprises a fuzzy feature extraction unit and a fuzzy feature comparison unit. After the second classification module classifies the remaining image, the fuzzy feature extraction unit can extract feature information from the remaining image and mark the extracted feature information as a fuzzy feature. The fuzzy feature comparison unit can compare the fuzzy feature with the standard feature, and when the similarity between the fuzzy feature and the standard feature is greater than 80%, the fuzzy feature and the standard feature information are saved in parallel.

[0041] The first feedback module can extract fuzzy features according to the classification results of the second classification module, and then improve the standard features, so that the first classification module shares the computational workload of the second classification module.

[0042] It also includes a preprocessing module, the output end of the image acquisition module is connected to the input end of the preprocessing module, and the output end of the preprocessing module is connected to the first feature extraction module. After the image acquisition module takes the image, the preprocessing module can perform noise reduction on the image.

[0043] It also includes a manual review module, the output end of the second classification module is connected to the input end of the manual review module, after the second classification module classifies the image, part of the remaining images are not classified by the second classification module, and this part of the images is marked as uncertain images. The uncertain images can enter the manual review module, and the manual review module can classify the images manually.

[0044] It also includes a second feedback module, the input end of the second feedback module is connected to the output end of the manual review module, and the output end of the second feedback module is connected to the input end of the second classification module. When the manual review module reviews the uncertain image, the manual review module can also manually mark the feature information of the uncertain image, and record the manually marked feature information as manual feature information. The second feedback module extracts the manual feature information and saves it in the second classification module.

[0045] When this system is implemented, for some images that cannot be recognized by the first classification module and the second classification module, manual classification is used to classify the images. During manual classification, the feature information can also be marked, and then the second feedback module is used to improve the standard feature information in the second classification module, so that this system can recognize more images when implemented.

[0046] The technical features of the above embodiments may be arbitrarily combined (as long as there is no contradiction in the combination of these technical features). To make the description concise, not all possible combinations of the technical features in the above embodiments are described; these embodiments that are not explicitly written should also be considered to be within the scope of this specification.

Claims

1. An intelligent image recognition and classification system based on deep learning, characterized in that: It includes an image acquisition module, a first feature extraction module, a first classification module, a second feature extraction module, a second classification module and a storage module; The output end of the image acquisition module is connected to the input end of the first feature extraction module, the output end of the first feature extraction module is connected to the input end of the first classification module, the output end of the first classification module is connected to the input end of the second feature extraction module, the output end of the second feature extraction module is connected to the input end of the second classification module, and the output end of the second classification module is connected to the input end of the storage module; The image acquisition module is used to acquire images, and the first feature extraction module is used to extract first feature information from the image. After the first feature information is extracted from the acquired image, the first classification module can perform a first classification on the image according to the extracted first feature information. When the image acquired by the image acquisition module has not been extracted by the first feature extraction module with the first feature information, this part of the image is marked as a remaining image, and the remaining image enters the second feature extraction module, and the second feature extraction module extracts the second feature information. The second classification module performs a second classification on the remaining image according to the second feature information, and the storage module can save the images classified by the first classification and the second classification; It also includes a first feedback module, the output end of the second classification module is connected to the input end of the first feedback module, and the output end of the first feedback module is connected to the input end of the first classification module; After the second classification module classifies the image, the first feedback module can extract feature information on the second classified image and input it into the first classification module.

2. The deep learning-based intelligent image recognition and classification system according to claim 1, characterized in that: The first classification module and the second classification module classify the image into n categories, and both the first classification module and the second classification module include n standard features. The extracted first feature information and the second feature information can be compared with the standard features to determine the classification of the image.

3. The deep learning-based intelligent image recognition and classification system according to claim 2, characterized in that: When the first classification module classifies an image, the first feature information is compared with the standard feature, and when the similarity between the first feature information and the standard feature is greater than 90%, the image corresponding to the first feature information is classified into the category of the standard feature.

4. The deep learning-based intelligent image recognition and classification system according to claim 3, characterized in that: The first feedback module comprises a fuzzy feature extraction unit and a fuzzy feature comparison unit. After the second classification module classifies the remaining image, the fuzzy feature extraction unit can extract feature information from the remaining image and mark the extracted feature information as a fuzzy feature. The fuzzy feature comparison unit can compare the fuzzy feature with the standard feature, and when the similarity between the fuzzy feature and the standard feature is greater than 80%, the fuzzy feature and the standard feature information are saved in parallel.

5. The deep learning-based intelligent image recognition and classification system according to claim 1, characterized in that: It also includes a preprocessing module, the output end of the image acquisition module is connected to the input end of the preprocessing module, and the output end of the preprocessing module is connected to the first feature extraction module. After the image acquisition module takes the image, the preprocessing module can perform noise reduction on the image.

6. The deep learning-based intelligent image recognition and classification system according to claim 1, characterized in that: It also includes a manual review module, the output end of the second classification module is connected to the input end of the manual review module, after the second classification module classifies the image, part of the remaining images are not classified by the second classification module, and this part of the images is marked as uncertain images. The uncertain images can enter the manual review module, and the manual review module can classify the images manually.

7. The deep learning-based intelligent image recognition and classification system according to claim 6, characterized in that: It also includes a second feedback module, the input end of the second feedback module is connected to the output end of the manual review module, and the output end of the second feedback module is connected to the input end of the second classification module. When the manual review module reviews the uncertain image, the manual review module can also manually mark the feature information of the uncertain image, and record the manually marked feature information as manual feature information. The second feedback module extracts the manual feature information and saves it in the second classification module.

Images