Image feature extraction and recognition method, electronic equipment and storage medium
By using multiple convolution and normalized image feature extraction methods in railway truck fault detection, the problem of few training samples is solved, and the accuracy and efficiency of fault detection are improved.
Patent Information
- Application Number
- CN202411284170.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-13
- Publication Date
- 2025-02-18
AI Technical Summary
In the detection of railway truck faults, the existing technology faces the problems of few training samples, high development costs and unsatisfactory results, especially in the application of deep learning.
A method of image feature extraction is adopted to obtain multiple features by multiple convolution and normalization, and superimposing the results of multiple convolution and normalization. The method includes acquiring an image, performing a first feature extraction, performing feature extraction and aggregation through a plurality of CBS modules and ELAN modules, and finally inputting the features into a classifier for identification.
Through this method, the accuracy and efficiency of image recognition can be improved while fewer training samples are required, and the development cost of deep learning models can be reduced.
Smart Images
Figure CN119478438A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image recognition technology, and in particular to an image feature extraction and recognition method, an electronic device and a storage medium. Background Art
[0002] With the rapid development of machine vision technology, technologies such as computer vision, target detection, and deep learning have been widely used in the field of image detection.
[0003] Railway freight cars currently mainly use traditional fault detection methods, which rely on manual inspection and empirical judgment. This method is time-consuming and prone to missed and misjudgment, and work efficiency cannot be improved.
[0004] With the rapid development of the truck transportation industry, improving operational efficiency and safety has become a necessary requirement, and it has become a trend to improve the efficiency and accuracy of truck fault detection through new technologies.
[0005] However, compared with other fields, railway train fault monitoring faces problems such as few training samples, high development costs, and unsatisfactory results in the application of deep learning.
[0006] Based on this, it is necessary to develop and design an image feature extraction method. Summary of the invention
[0007] The embodiments of the present invention provide an image feature extraction and recognition method, an electronic device and a storage medium, which are used to solve the problem of insufficient training samples in the prior art.
[0008] In a first aspect, an embodiment of the present invention provides an image feature extraction method, comprising:
[0009] acquiring a first image;
[0010] Inputting the first image into a plurality of first CBS modules (channel compression modules, ChannelBlock Squeeze) connected in sequence to extract the first feature, wherein the first CBS module includes: a first convolution layer, a first normalization layer and an activation function layer, the first convolution layer convolves the input data to obtain a convolution result, the first normalization layer normalizes the convolution result to obtain a normalized result, and the activation function layer processes the normalized result to obtain the first feature;
[0011] Inputting the first feature into a first ELAN module (Effective Long-Range Aggregation Network), and using the output of the first ELAN module as a second feature, wherein the first ELAN module includes a plurality of second CBS modules, the second CBS modules have the same structure as the first CBS module, and the plurality of second CBS modules are mixed and connected to output the second feature;
[0012] The second features are respectively input into four intermediate layers connected in series, and the output of each intermediate layer is used as an image feature, so as to obtain multiple image features, wherein each intermediate layer includes an MP1 module and a second ELAN module whose input end is connected to the output end of the MP1 module (Max Pool), and the second ELAN module has the same structure as the first ELAN module.
[0013] In a possible implementation, the plurality of second CBS modules form a first branch, a second branch and a merging branch;
[0014] The first branch includes five second CBS modules connected in series, the second branch is one second CBS module, and the confluence is one second CBS module;
[0015] The output of the first second CBS module, the output of the third second CBS module, the output of the fifth second CBS module on the first branch and the output of the second CBS module on the second branch are combined as the input of the second CBS module on the merging path;
[0016] The second CBS module of the merging path outputs a second feature.
[0017] In one possible implementation, the second CBS module of the second branch is a convolution layer with a 1×1 convolution kernel, the first second CBS module on the first branch is a convolution layer with a 1×1 convolution kernel, and the other second CBS modules on the first branch are convolution layers with a 3×3 convolution kernel.
[0018] In a possible implementation, the MP1 module includes a third branch and a fourth branch;
[0019] The third branch includes a maximum pooling layer and a third CBS module whose input end is connected to the output end of the maximum pooling layer. The fourth branch includes a fourth CBS module and a fifth CBS module whose input end is connected to the output end of the fourth CBS module. The input end of the maximum pooling layer is connected to the input end of the fourth CBS module, and the output of the third CBS module is merged with the output of the fifth CBS module as the output of the MP1 module.
[0020] In one possible implementation, the first convolution layer of the third CBS module is a convolution layer with a 1×1 convolution kernel, the first convolution layer of the fourth CBS module is a convolution layer with a 1×1 convolution kernel, and the first convolution layer of the third CBS module is a convolution layer with a 3×3 convolution kernel and a stride of 2.
[0021] In a possible implementation, the first normalization layer of each second CBS module is normalized according to a first formula, wherein the first formula is:
[0022]
[0023] In the formula, is the normalized data of the ith i is the data before normalization of the ith data, σ i is the i-th standard deviation, m is the total number of data that need to be normalized, C is a constant, S i is the data set that needs to be normalized.
[0024] In a possible implementation, the data set to be normalized is determined according to a second formula, wherein the second formula is:
[0025]
[0026] In the formula, G is the number of groups, C is the number of channels in the data set, This is a lower truncation operation.
[0027] In a second aspect, an embodiment of the present invention provides an image recognition method, comprising:
[0028] Receiving a plurality of image features as described in the first aspect or any possible implementation manner of the first aspect;
[0029] The multiple image features are sent to a classifier to obtain a recognition result.
[0030] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the steps of the method described in the first aspect or any possible implementation method of the first aspect.
[0031] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect or any possible implementation of the first aspect are implemented.
[0032] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:
[0033] The image feature extraction method implementation of the present invention first acquires a first image; then inputs the first image into a plurality of first CBS modules connected in sequence to extract a first feature, wherein the first CBS module includes: a first convolution layer, a first normalization layer, and an activation function layer, wherein the first convolution layer performs multiplication on the input data to obtain a convolution result, the first normalization layer performs normalization processing on the convolution result to obtain a normalized result, and the activation function layer processes the normalized result to obtain the first feature; then the first feature is input into a first ELAN module, and the output of the first ELAN module is used as a second feature, wherein the first ELAN module includes a plurality of second CBS modules, wherein the second CBS module has the same structure as the first CBS module, and the plurality of second CBS modules are mixed and connected to output the second feature; finally, the second feature is respectively input into four intermediate layers connected in series in sequence, and the output of each intermediate layer is used as an image feature, thereby obtaining a plurality of image features, wherein each intermediate layer includes an MP1 module and a second ELAN module whose input end is connected to the output end of the MP1 module, and the second ELAN module has the same structure as the first ELAN module. The implementation manner of the present invention obtains multiple features through multiple convolutions and normalizations, and superimposes the results of the multiple convolutions and normalizations. The levels and number of features retained are relatively large, which is conducive to improving the classification results of the classifier. Compared with the deep learning training model of the prior art, fewer training samples are required. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0035] Figure 1 is a flow chart of an image feature extraction method provided by an embodiment of the present invention;
[0036] Figure 2 is a structural diagram of a feature extraction system provided by an embodiment of the present invention;
[0037] Figure 3 is a CBS module structure diagram provided by an embodiment of the present invention;
[0038] Figure 4 is a CBS module structure diagram provided by an embodiment of the present invention;
[0039] Figure 5is a functional block diagram of an image feature extraction device provided by an embodiment of the present invention;
[0040] Figure 6 It is a functional block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0041] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, and methods are omitted so as not to obscure the description of the present invention with unnecessary details.
[0042] In order to make the purpose, technical solutions and advantages of the present invention more clear, a specific implementation method will be described below in conjunction with the accompanying drawings.
[0043] The following is a detailed description of an embodiment of the present invention. This example is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.
[0044] Figure 1 A flowchart of an image feature extraction method provided in an embodiment of the present invention.
[0045] like Figure 1 As shown, it shows a flowchart of the implementation of the image feature extraction method provided by an embodiment of the present invention, which is described in detail as follows:
[0046] In step 101, a first image is acquired.
[0047] In step 102, the first image is input into a plurality of first CBS modules connected in sequence to perform first feature extraction, wherein the first CBS module includes: a first convolution layer, a first normalization layer, and an activation function layer, the first convolution layer convolves the input data to obtain a convolution result, the first normalization layer normalizes the convolution result to obtain a normalized result, and the activation function layer processes the normalized result to obtain the first feature.
[0048] Exemplarily, the structure diagram of the feature extraction system in the embodiment of the present invention is as follows: Figure 2 As shown, it includes multiple CBS modules, ELAN modules and MP1 modules. After the image is input from the upper CBS module, the first ELAN module outputs a feature, the second ELAN module outputs a feature, and the third ELAN module outputs a feature.
[0049] like Figure 3As shown, the CBS module includes a convolution layer, a normalization layer and an activation function layer. The data is first convolved, then normalized, and finally activated to generate the first feature. For the ELAN module and the MP1 module, the embodiments of the present invention will be discussed in detail below.
[0050] In step 103, the first feature is input into the first ELAN module, and the output of the first ELAN module is used as the second feature, wherein the first ELAN module includes multiple second CBS modules, wherein the second CBS modules have the same structure as the first CBS module, and the multiple second CBS modules are mixed and connected to output the second feature.
[0051] In some embodiments, the plurality of second CBS modules form a first branch, a second branch, and a merged path;
[0052] The first branch includes five second CBS modules connected in series, the second branch is one second CBS module, and the confluence is one second CBS module;
[0053] The output of the first second CBS module, the output of the third second CBS module, the output of the fifth second CBS module on the first branch and the output of the second CBS module on the second branch are combined as the input of the second CBS module on the merging path;
[0054] The second CBS module of the merging path outputs a second feature.
[0055] In some embodiments, the second CBS module of the second branch is a convolution layer with a 1×1 convolution kernel, the first second CBS module on the first branch is a convolution layer with a 1×1 convolution kernel, and the other second CBS modules on the first branch are convolution layers with a 3×3 convolution kernel.
[0056] For example, Figure 4 As shown, the ELAN module is a feature extraction layer composed of multiple CBS modules, wherein the multiple CBS modules are divided into three paths, wherein the first branch is composed of five CBS modules connected in series, and the second branch is a CBS module. The outputs of the first CBS module, the third CBS module and the fifth CBS module in the first branch are superimposed with the output of the CBS module in the second branch and input to the confluence path. In fact, the confluence path is a CBS module, and the output of this CBS module is used as the second feature.
[0057] In fact, the settings of these CBS modules are different. The second CBS module of the second branch is a convolution layer with a 1×1 convolution kernel, the first second CBS module on the first branch is a convolution layer with a 1×1 convolution kernel, and the other second CBS modules on the first branch are convolution layers with a 3×3 convolution kernel. The advantage of this is that it reduces the amount of convolution calculation and retains more features.
[0058] In step 104, the second feature is respectively input into four intermediate layers connected in series, and the output of each intermediate layer is used as an image feature, thereby obtaining multiple image features, wherein each intermediate layer includes an MP1 module and a second ELAN module whose input end is connected to the output end of the MP1 module, and the second ELAN module has the same structure as the first ELAN module.
[0059] In some embodiments, the MP1 module includes a third branch and a fourth branch;
[0060] The third branch includes a maximum pooling layer and a third CBS module whose input end is connected to the output end of the maximum pooling layer. The fourth branch includes a fourth CBS module and a fifth CBS module whose input end is connected to the output end of the fourth CBS module. The input end of the maximum pooling layer is connected to the input end of the fourth CBS module, and the output of the third CBS module is merged with the output of the fifth CBS module as the output of the MP1 module.
[0061] In some embodiments, the first convolution layer of the third CBS module is a convolution layer with a 1×1 convolution kernel, the first convolution layer of the fourth CBS module is a convolution layer with a 1×1 convolution kernel, and the first convolution layer of the third CBS module is a convolution layer with a 3×3 convolution kernel and a stride of 2.
[0062] In some embodiments, the first normalization layer of each second CBS module is normalized according to a first formula, wherein the first formula is:
[0063]
[0064] In the formula, is the normalized data of the ith i is the data before normalization of the ith data, σ i is the i-th standard deviation, m is the total number of data that need to be normalized, C is a constant, S i is the data set that needs to be normalized.
[0065] In some implementations, the data set to be normalized is determined according to a second formula, wherein the second formula is:
[0066]
[0067] In the formula, G is the number of groups, C is the number of channels in the data set, This is a lower truncation operation.
[0068] Exemplarily, on the MP1 module, it includes two branches, one of which is input to a maximum pooling layer, and the output of the maximum pooling layer is input to a CBS module, which is a convolutional layer with a 1×1 convolution kernel. The other branch is two CBS modules connected in series, where the front-end CBS module is a convolutional layer with a 1×1 convolution kernel, and the back-end CBS module is a convolutional layer with a 3×3 convolution kernel and a stride of 2.
[0069] The outputs of the two branches are combined as the output of the MP1 module.
[0070] As mentioned above, the CBS module includes a convolution layer, a normalization layer and an activation function layer, wherein the features extracted by the convolution kernel in the convolution layer are normalized and then sent to the activation function layer to perform feature conversion and fusion on the data. As an important part of it, the normalization layer has a variety of different normalization methods, and the embodiment of the present invention adopts the first formula for normalization:
[0071]
[0072] In the formula, is the normalized data of the ith i is the data before normalization of the ith data, σ i is the i-th standard deviation, m is the total number of data that need to be normalized, C is a constant, S i is the data set that needs to be normalized.
[0073] As mentioned above, when the features extracted from the image are applied to a model based on deep learning with a small sample size, if the traditional normalization method is used, the features will become very few. Therefore, the embodiment of the present invention uses group normalization to normalize the data obtained by the convolution layer. The group normalization process is expressed by the formula:
[0074]
[0075] In the formula, G is the number of groups, C is the number of channels in the data set, This is a lower truncation operation.
[0076] The image feature extraction method implementation of the present invention first acquires a first image; then inputs the first image into a plurality of first CBS modules connected in sequence to extract a first feature, wherein the first CBS module includes: a first convolution layer, a first normalization layer, and an activation function layer, wherein the first convolution layer performs multiplication on the input data to obtain a convolution result, the first normalization layer performs normalization processing on the convolution result to obtain a normalized result, and the activation function layer processes the normalized result to obtain the first feature; then the first feature is input into a first ELAN module, and the output of the first ELAN module is used as a second feature, wherein the first ELAN module includes a plurality of second CBS modules, wherein the second CBS module has the same structure as the first CBS module, and the plurality of second CBS modules are mixed and connected to output the second feature; finally, the second feature is respectively input into four intermediate layers connected in series in sequence, and the output of each intermediate layer is used as an image feature, thereby obtaining a plurality of image features, wherein each intermediate layer includes an MP1 module and a second ELAN module whose input end is connected to the output end of the MP1 module, and the second ELAN module has the same structure as the first ELAN module. The implementation manner of the present invention obtains multiple features through multiple convolutions and normalizations, and superimposes the results of the multiple convolutions and normalizations. The levels and number of features retained are relatively large, which is conducive to improving the classification results of the classifier. Compared with the deep learning training model of the prior art, fewer training samples are required.
[0077] In addition, an embodiment of the present invention also provides an image recognition method, which inputs the features obtained in the image feature extraction method of the first aspect into a classifier, for example, a classifier built based on a deep learning model, to obtain a recognition result.
[0078] It should be understood that the size of the serial numbers of the steps in the above implementation does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation method of the present invention.
[0079] The following is an implementation of the device of the present invention. For details not described in detail, reference may be made to the corresponding method implementation described above.
[0080] Figure 5 is a functional block diagram of an image feature extraction device provided by an embodiment of the present invention, referring to Figure 5 The image feature extraction device includes: an image acquisition module 501, a first feature extraction module 502, a second feature extraction module 503 and an image feature extraction module 504, wherein:
[0081] An image acquisition module 501 is used to acquire a first image;
[0082] A first feature extraction module 502 is used to input the first image into a plurality of first CBS modules connected in sequence to perform first feature extraction, wherein the first CBS module includes: a first convolution layer, a first normalization layer, and an activation function layer, the first convolution layer convolves the input data to obtain a convolution result, the first normalization layer normalizes the convolution result to obtain a normalized result, and the activation function layer processes the normalized result to obtain the first feature;
[0083] A second feature extraction module 503 is used to input the first feature into a first ELAN module, and use the output of the first ELAN module as a second feature, wherein the first ELAN module includes a plurality of second CBS modules, wherein the second CBS module has the same structure as the first CBS module, and the plurality of second CBS modules are mixed and connected to output the second feature;
[0084] as well as,
[0085] The image feature extraction module 504 is used to input the second feature into four intermediate layers connected in series respectively, and use the output of each intermediate layer as the image feature, so as to obtain multiple image features, wherein each intermediate layer includes an MP1 module and a second ELAN module whose input end is connected to the output end of the MP1 module, and the second ELAN module has the same structure as the first ELAN module.
[0086] Figure 6 is a functional block diagram of an electronic device provided by an embodiment of the present invention. Figure 6 As shown, the electronic device 6 of this embodiment includes: a processor 600 and a memory 601, wherein the memory 601 stores a computer program 602 that can be run on the processor 600. When the processor 600 executes the computer program 602, the steps in the above-mentioned image feature extraction methods and embodiments are implemented, for example Figure 1 Steps 101 to 104 are shown.
[0087] Exemplarily, the computer program 602 may be divided into one or more modules / units, and the one or more modules / units are stored in the memory 601 and executed by the processor 600 to implement the present invention.
[0088] The electronic device 6 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device 6 may include, but is not limited to, a processor 600 and a memory 601. Those skilled in the art will appreciate that Figure 6It is only an example of the electronic device 6 and does not constitute a limitation of the electronic device 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 6 may also include input and output devices, network access devices, buses, etc.
[0089] The processor 600 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.
[0090] The memory 601 may be an internal storage unit of the electronic device 6, such as a hard disk or memory of the electronic device 6. The memory 601 may also be an external storage device of the electronic device 6, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 6. Further, the memory 601 may also include both an internal storage unit and an external storage device of the electronic device 6. The memory 601 is used to store the computer program 602 and other programs and data required by the electronic device 6. The memory 601 may also be used to temporarily store data that has been output or is to be output.
[0091] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the implementation method can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method implementation method, which will not be repeated here.
[0092] In the above-mentioned embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0093] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0094] In the embodiments provided by the present invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0095] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0096] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0097] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned implementation method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various methods and device implementation methods. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.
[0098] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A method for extracting image features, characterized in that: include: acquiring a first image; Inputting the first image into a plurality of first CBS modules connected in sequence to extract a first feature, wherein the first CBS module includes: a first convolution layer, a first normalization layer, and an activation function layer, wherein the first convolution layer performs convolution on the input data to obtain a convolution result, the first normalization layer performs normalization processing on the convolution result to obtain a normalized result, and the activation function layer processes the normalized result to obtain the first feature; Inputting the first feature into a first ELAN module, and using the output of the first ELAN module as a second feature, wherein the first ELAN module includes a plurality of second CBS modules, the second CBS modules have the same structure as the first CBS module, and the plurality of second CBS modules are mixed and connected to output the second feature; The second feature is respectively input into four intermediate layers connected in series, and the output of each intermediate layer is used as an image feature, so as to obtain multiple image features, wherein each intermediate layer includes an MP1 module and a second ELAN module whose input end is connected to the output end of the MP1 module, and the second ELAN module has the same structure as the first ELAN module.
2. The image feature extraction method according to claim 1, characterized in that: The plurality of second CBS modules form a first branch, a second branch and a merging path; The first branch includes five second CBS modules connected in series, the second branch is one second CBS module, and the confluence is one second CBS module; The output of the first second CBS module, the output of the third second CBS module, the output of the fifth second CBS module on the first branch and the output of the second CBS module on the second branch are combined as the input of the second CBS module on the merging path; The second CBS module of the merging path outputs a second feature.
3. The image feature extraction method according to claim 2, characterized in that: The second CBS module of the second branch is a convolution layer with a 1×1 convolution kernel, the first second CBS module on the first branch is a convolution layer with a 1×1 convolution kernel, and the other second CBS modules on the first branch are convolution layers with a 3×3 convolution kernel.
4. The image feature extraction method according to claim 1, characterized in that: The MP1 module includes a third branch and a fourth branch; The third branch includes a maximum pooling layer and a third CBS module whose input end is connected to the output end of the maximum pooling layer. The fourth branch includes a fourth CBS module and a fifth CBS module whose input end is connected to the output end of the fourth CBS module. The input end of the maximum pooling layer is connected to the input end of the fourth CBS module, and the output of the third CBS module is merged with the output of the fifth CBS module as the output of the MP1 module.
5. The image feature extraction method according to claim 4, characterized in that: The first convolution layer of the third CBS module is a convolution layer with a 1×1 convolution kernel, the first convolution layer of the fourth CBS module is a convolution layer with a 1×1 convolution kernel, and the first convolution layer of the third CBS module is a convolution layer with a 3×3 convolution kernel and a stride of 2.
6. The image feature extraction method according to claims 1-5, characterized in that: The first normalization layer of each second CBS module is normalized according to a first formula, wherein the first formula is: In the formula, is the normalized data of the ith i is the data before normalization of the ith data, σ i is the i-th standard deviation, m is the total number of data that need to be normalized, C is a constant, S i is the data set that needs to be normalized.
7. The image feature extraction method according to claim 6, characterized in that: The data set to be normalized is determined according to a second formula, wherein the second formula is: In the formula, G is the number of groups, C is the number of channels in the data set, This is a lower truncation operation.
8. An image recognition method, characterized in that: include: Receiving a plurality of image features as described in any one of claims 1 to 7; The multiple image features are sent to a classifier to obtain a recognition result.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method as claimed in any one of claims 1 to 7 are implemented.