A neural network-based image classification method and system

Abstract

The present application discloses a neural network-based image classification method and system, which mainly relates to the field of image classification technology, and is used to solve the problems that existing multi-stage neural network mechanism methods easily lead to waste of computing resources or omission of key features, and existing dimensionality reduction processing methods may lose key discriminant information. It includes: cutting the image to be recognized corresponding to the erroneous first recognition result, and then inputting it into the image recognition model to obtain a second recognition result; when the second recognition result is consistent with the first recognition result, according to the relationship between the weight values ​​of the distinguishing target information, the next distinguishing point is determined from the distinguishing target information in sequence, and the next distinguishing point that can first determine that the first recognition result is correct is determined as the distinguishing point to be adjusted; and adjusting the distinguishing target weight value of the distinguishing point to be adjusted to be greater than the distinguishing target weight value of the first distinguishing point.

Description

Technical Field

[0001] The present application relates to the technical field of image classification, and in particular to a neural network-based image classification method and system. Background Art

[0002] Existing neural network-based image classification techniques generally use end-to-end deep learning models for feature extraction and classification decisions, but they have significant limitations when dealing with easily confused categories with similar visual features. Traditional methods typically improve classification accuracy by increasing model depth or introducing attention mechanisms, but such solutions lead to a surge in computational complexity and make it difficult to optimize for specific error-prone categories.

[0003] Some improvement schemes attempt to improve reliability through multi-stage neural network mechanisms. For example, a neural network module is superimposed after preliminary classification. However, there is a defect of rigid verification strategy, which is mainly reflected in the fixed neural network logic and the fixed neural network judgment benchmark. It consumes a lot of computing power during operation and has a long running time, which easily leads to waste of computing resources or omission of key features.

[0004] Furthermore, while existing dimensionality reduction methods (such as principal component analysis or convolutional downsampling) can reduce the amount of input data, they are not deeply integrated with the need to enhance the features of error-prone categories, and may lose key discriminative information. For example, in garbage sorting scenarios, the combination of local reflective pixels of plastic and glass bottles is crucial for classification, while traditional dimensionality reduction operations can weaken such details. Summary of the Invention

[0005] In response to the above-mentioned deficiencies in the prior art, the present application provides a neural network-based image classification method and system to solve the problems that the existing multi-stage neural network mechanism method easily leads to waste of computing resources or omission of key features, and the existing dimensionality reduction processing method may lose key discriminant information.

[0006] In a first aspect, the present application provides an image classification method based on a neural network, the method comprising:

[0007] The image to be identified is subjected to dimensionality reduction and shearing processing, and then input into a preset lightweight recognition model to obtain a first recognition result. When the first recognition result belongs to a preset error-prone type, a preset distinction database is called to obtain distinction target information and distinction target weight values. The distinction target information includes at least: a target pixel combination, a target contour, texture features, color distribution features, and an edge gradient distribution. The distinction target weight values ​​include at least: a target pixel weight value, a target contour weight value, a texture feature weight value, a color distribution feature weight value, and an edge gradient distribution weight value. Based on the relationship between the distinction target weight values, a first distinction point is determined from the distinction target information. Based on whether the first distinction point exists in the image to be identified, whether the first recognition result is correct is determined. The image to be identified corresponding to the erroneous first recognition result is sheared and then input into the image recognition model to obtain a second recognition result. When the second recognition result is consistent with the first recognition result, the next distinction point is determined from the distinction target information in sequence based on the relationship between the distinction target information weight values. The next distinction point that can first determine that the first recognition result is correct is determined as the distinction point to be adjusted. The distinction target weight value of the distinction point to be adjusted is adjusted to be greater than the distinction target weight value of the first distinction point.

[0008] In one implementation of the present application, before performing dimensionality reduction and shearing processing on the image to be recognized and then inputting the image into a preset lightweight recognition model to obtain a first recognition result, the method further includes:

[0009] Build a preset lightweight recognition model with 5 convolution layers. The kernel size of each layer is set to 3×3 and the stride is 2×2. Depthwise separable convolution is used instead of standard convolution operation.

[0010] Add a global average pooling layer instead of a fully connected layer at the end of the network of the preset lightweight recognition model.

[0011] In one implementation of the present application, the image to be recognized is subjected to dimensionality reduction and shearing processing, and then input into a preset lightweight recognition model to obtain a first recognition result, which specifically includes:

[0012] The convolutional downsampling layer compresses the resolution of the image to be identified to the preset resolution while retaining high-frequency features. At the same time, the Grad-CAM algorithm is used to locate the salient area and crop it to obtain the identification area.

[0013] The principal component analysis technique is used to map the RGB three channels of the recognition area into a single-channel grayscale image to obtain the image after dimensionality reduction and shearing processing.

[0014] In one implementation of the present application, when the first recognition result belongs to a preset error-prone type, before calling a preset distinction database to obtain distinction target information, the method further includes:

[0015] Through the preset information adjustment interface, add, delete, modify and query the distinguishing target information.

[0016] In one implementation of the present application, determining the first distinguishing point from the distinguishing target information based on the relationship between the distinguishing target weight values ​​specifically includes:

[0017] The distinguishing target information with the largest distinguishing target weight value is determined as the first distinguishing point.

[0018] In a second aspect, the present application provides an image classification system based on a neural network, the system comprising:

[0019] The acquisition module is used to perform dimensionality reduction and shearing processing on the image to be identified, and then input the preset lightweight recognition model to obtain the first recognition result; the acquisition module is used to call the preset distinction database when the first recognition result belongs to the preset error-prone type to obtain the distinction target information and the distinction target weight value; wherein the distinction target information at least includes: target pixel combination, target outline, texture feature, color distribution feature, edge gradient distribution, and the distinction target weight value at least includes: target pixel weight value, target outline weight value, texture feature weight value, color distribution feature weight value, edge gradient distribution weight value; the result module is used to calculate the difference between the target weight values ​​according to the relationship between the distinction target weight values. A system is provided for determining a first distinguishing point from the distinguishing target information, and determining whether a first recognition result is correct according to whether the first distinguishing point exists in the image to be recognized; the image to be recognized corresponding to the erroneous first recognition result is cut, and then input into the image recognition model to obtain a second recognition result; an adjustment module is provided for determining the next distinguishing point from the distinguishing target information in turn according to the relationship between the weight values ​​of the distinguishing target information when the second recognition result is consistent with the first recognition result, and determining the next distinguishing point that can first determine that the first recognition result is correct as the distinguishing point to be adjusted; and adjusting the distinguishing target weight value of the distinguishing point to be adjusted to be greater than the distinguishing target weight value of the first distinguishing point.

[0020] In one implementation of the present application, the acquisition module includes a model building unit,

[0021] It is used to build a preset lightweight recognition model with 5 convolution layers. The kernel size of each layer is set to 3×3, the stride is 2×2, and the depthwise separable convolution is used instead of the standard convolution operation.

[0022] Add a global average pooling layer instead of a fully connected layer at the end of the network of the preset lightweight recognition model.

[0023] In one implementation of the present application, the acquisition module includes a shearing unit,

[0024] It is used to compress the resolution of the image to be identified to the preset resolution through the convolution downsampling layer and retain the high-frequency features. At the same time, it combines the Grad-CAM algorithm to locate the salient area and crop it to obtain the identification area;

[0025] The principal component analysis technique is used to map the RGB three channels of the recognition area into a single-channel grayscale image to obtain the image after dimensionality reduction and shearing processing.

[0026] In one implementation of the present application, the acquisition module includes an acquisition unit,

[0027] Used to add, delete, modify and query distinguishing target information through the preset information adjustment interface.

[0028] In one implementation of the present application, the adjustment module includes an adjustment unit,

[0029] The distinction target information with the largest distinction target weight value is used to determine the first distinction point.

[0030] Those skilled in the art will appreciate that this application has at least the following beneficial effects:

[0031] This application provides a neural network-based image classification method and system:

[0032] 1. Improvements to the problem of wasted computing resources:

[0033] This application uses a combination of "dimensionality reduction and shearing processing + lightweight recognition model" to reduce the computational complexity by first reducing the data volume and then using a lightweight model for preliminary recognition. Compared with the full-scale calculation mode of traditional multi-stage neural networks, this application only calls complex models for error-prone types, effectively avoiding resource waste. This technical feature achieves a balance between computational efficiency and recognition accuracy through a dynamic resource allocation strategy.

[0034] When the initial recognition result is an error-prone type, a progressive verification process of "differentiated database comparison → weight adjustment → secondary recognition" is used to only call high-energy recognition models when necessary. This layered approach avoids the drawbacks of blindly using complex models in existing technologies.

[0035] 2. Improvements to the problem of missing key features:

[0036] This application incorporates five types of target distinguishing information: target pixel combinations, texture features, and color distribution features. Key discriminant points are dynamically determined through weighted relationships (for example, the first discriminant point that can be verified as correct is used as the point to be adjusted). Compared to traditional fixed-weight dimensionality reduction methods, this application retains high-frequency key features through a weighted feedback mechanism, resolving the problem of discriminant information loss caused by static feature selection in existing technologies.

[0037] By distinguishing between the pre-set weights of error-prone features in the database (such as the weight of edge gradient distribution), this essentially forms a feature enhancement mechanism tailored to specific scenarios. By adjusting the weights with gradients (e.g., the weight of the point to be adjusted is greater than the weight of the previous distinguishing point), adaptive optimization of feature discrimination is achieved.

[0038] 3. Systematic improvement of technical effects:

[0039] This application uses the coordinated cooperation of lightweight models and complex models to assign most routine recognition tasks to low-energy models while ensuring recognition accuracy (secondary recognition verification). The overall efficiency improvement is in line with the judgment basis of the three-step method on the actual solution effect of technical problems.

[0040] The weight adjustment mechanism enables the system to dynamically optimize feature discrimination priorities based on actual recognition results (for example, adjusting the color distribution weight value to be greater than the texture weight value). This closed-loop feedback mechanism overcomes the limitations of the traditional neural network's fixed parameter model. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. 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 any creative work.

[0042] Figure 1 This is a flowchart of a neural network-based image classification method provided in an embodiment of the present application.

[0043] Figure 2 This is a schematic diagram of the internal structure of a neural network-based image classification system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0044] It should be understood by those skilled in the art that the embodiments described below are merely preferred embodiments of the present disclosure and do not imply that the present disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely intended to explain the technical principles of the present disclosure and are not intended to limit the scope of protection of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of the present disclosure.

[0045] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0046] The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0047] The embodiment provides an image classification method based on a neural network, such as Figure 1 As shown, the method provided in the embodiment of the present application mainly includes the following steps:

[0048] Step 110 : Perform dimensionality reduction and shearing processing on the image to be recognized, and then input it into a preset lightweight recognition model to obtain a first recognition result.

[0049] In some embodiments, before performing dimensionality reduction and shearing processing on the image to be recognized and then inputting the image into a preset lightweight recognition model to obtain the first recognition result, the method further includes:

[0050] Build a preset lightweight recognition model with 5 convolution layers. The kernel size of each layer is set to 3×3 and the stride is 2×2. Depthwise separable convolution is used instead of standard convolution operation.

[0051] Add a global average pooling layer instead of a fully connected layer at the end of the network of the preset lightweight recognition model.

[0052] It's understandable that this step, by using a small 3×3 convolution kernel and a 2×2 stride, reduces the number of model parameters and computational complexity, thereby improving recognition speed. Depthwise separable convolution decomposes the standard convolution into two steps: depthwise convolution and pointwise convolution, reducing the number of model parameters and memory requirements. The global average pooling layer avoids overfitting caused by fully connected layers while preserving important features in the image, helping to improve recognition accuracy.

[0053] The image to be recognized is subjected to dimensionality reduction and shearing processing, and then input into a preset lightweight recognition model to obtain a first recognition result, specifically including:

[0054] The convolutional downsampling layer compresses the resolution of the image to be identified to the preset resolution while retaining high-frequency features. At the same time, the Grad-CAM algorithm is used to locate the salient area and crop it to obtain the identification area.

[0055] The principal component analysis technique is used to map the RGB three channels of the recognition area into a single-channel grayscale image to obtain the image after dimensionality reduction and shearing processing.

[0056] It should be noted that due to the use of lightweight recognition models and dimensionality reduction and shearing processing, the entire recognition process is more efficient and can complete the image recognition task in a shorter time. By focusing on key areas and removing redundant information, the model can more accurately identify target objects or scenes in the image.

[0057] For example,

[0058] Consider a license plate recognition application scenario where the license plate needs to be identified in a vehicle image. In this scenario, the steps described above can be followed: Build a lightweight recognition model with five convolutional layers, each with a kernel size of 3×3 and a stride of 2×2, using depthwise separable convolution. Add a global average pooling layer at the end of the model. Perform dimensionality reduction on the vehicle image to remove noise and redundant information. Then, crop the license plate based on its position in the image, retaining only the license plate area. The reduced and cropped license plate image is input into the lightweight recognition model, which performs feature extraction and classification on the license plate image and outputs the license plate number as the first recognition result. This approach improves license plate recognition efficiency while maintaining accuracy, meeting the requirements of practical application scenarios.

[0059] Step 120: When the first recognition result belongs to the preset error-prone type, call the preset distinction database to obtain distinction target information and distinction target weight values.

[0060] It should be noted that the target information for distinguishing includes at least: target pixel combination, target contour, texture features, color distribution features, and edge gradient distribution; the target weight value for distinguishing includes at least: target pixel weight value, target contour weight value, texture feature weight value, color distribution feature weight value, and edge gradient distribution weight value.

[0061] It's understandable that this step, by pre-setting the error-prone type, can quickly identify potential recognition errors, providing the prerequisite for subsequent invocation of the differentiation database. The differentiation database stores specific information and weights for error-prone types, and invoking this database provides strong support for subsequent feature comparison and weight adjustment. By obtaining detailed target information and corresponding weights, image features can be more precisely compared and analyzed, thereby improving recognition accuracy. Specialized processing for error-prone types enhances the system's robustness to complex images and noise.

[0062] When the first recognition result belongs to the preset error-prone type, before calling the preset distinction database to obtain the distinction target information, the method may further include:

[0063] Through the preset information adjustment interface, add, delete, modify and query the distinguishing target information.

[0064] It is understandable that this application allows users to flexibly manage the distinguishing target information through the interface, which improves the maintainability and adaptability of the system. Adjusting the distinguishing target information according to actual needs helps optimize recognition performance and improve recognition efficiency and accuracy.

[0065] Specific examples:

[0066] Consider a license plate recognition application scenario where certain special license plates are prone to misidentification due to their special format and complex characters. In this case, you can follow these steps:

[0067] First, the license plate image is preliminarily recognized to obtain a first recognition result.

[0068] Determine whether the first recognition result belongs to a preset error-prone type.

[0069] If the license plate is prone to error, the system calls a pre-set discrimination database to obtain the discrimination target information and weight values ​​for these special license plates. For example, for police license plates, the system may need to focus on the plate's contour features, color distribution features, and edge gradient distribution.

[0070] Through the preset information adjustment interface, necessary addition, deletion, modification and query operations are performed on the obtained distinguishing target information to ensure the accuracy and completeness of the information.

[0071] The obtained distinguishing target information and weight values ​​are used to identify and analyze the license plate image again, and finally an accurate recognition result is obtained.

[0072] In this way, the accuracy and robustness of license plate recognition can be improved, especially when dealing with special license plates, better recognition results can be achieved.

[0073] Step 130: Determine a first distinguishing point from the distinguishing target information based on the relationship between the distinguishing target weight values, and determine whether the first recognition result is correct based on whether the first distinguishing point exists in the image to be recognized; cut the image to be recognized corresponding to the erroneous first recognition result, and then input it into the image recognition model to obtain a second recognition result.

[0074] It should be noted that in the image recognition process, the first distinction point is determined from the distinction target information based on the relationship between the distinction target weight values, and the correctness of the first recognition result is judged accordingly. When the first recognition result is wrong, the corresponding image to be recognized is cut and re-recognized, which enhances the robustness of the recognition.

[0075] By focusing on the most weighted discriminant target information, key features in the image can be more accurately identified, thereby improving recognition accuracy. Focusing recognition on the most weighted discriminant target information helps reduce unnecessary calculations and optimize the recognition process.

[0076] ‌Example‌:

[0077] For example, in a license plate recognition scenario, the target distinguishing information includes features such as the license plate's outline, color, and characters, with corresponding weights of 0.3, 0.2, and 0.5, respectively. In this case, the character feature has the highest weight and is therefore determined as the first distinguishing factor. During the recognition process, the system prioritizes the features of the license plate's characters to improve recognition accuracy.

[0078] By checking whether the first distinguishing point exists, the correctness of the first recognition result can be quickly verified without complex comparison and analysis. This helps reduce unnecessary repeated recognition and improves overall recognition efficiency.

[0079] ‌Example‌:

[0080] Continuing with the example of license plate recognition, if the first recognition result incorrectly identifies the license plate characters, and the first distinguishing point (character feature) does exist in the image to be recognized but is not correctly recognized, then the system will determine that the first recognition result is wrong.

[0081] When the first recognition result is judged to be wrong, the corresponding image to be recognized is cropped to remove redundant information, and then re-input into the image recognition model for recognition to obtain a second recognition result.

[0082] Removing redundant information through cropping helps the model more accurately identify key features in the image and enhances recognition robustness. Re-recognition helps correct errors in the first recognition result and improve overall recognition accuracy.

[0083] ‌Example‌:

[0084] In a license plate recognition scenario, if the first recognition result incorrectly identifies the license plate characters as other characters, the system will crop the license plate image to remove redundant information such as background, and then re-enter the recognition model for recognition. Assuming the second recognition result correctly identifies the license plate characters, the second recognition result is considered the correct recognition result.

[0085] The first distinguishing point is determined from the distinguishing target information according to the relationship between the distinguishing target weight values, which may be specifically:

[0086] The distinguishing target information with the largest distinguishing target weight value is determined as the first distinguishing point.

[0087] Step 140: When the second recognition result is consistent with the first recognition result, the next distinction point is determined from the distinction target information in sequence according to the relationship between the weight values ​​of the distinction target information, and the next distinction point that can first determine that the first recognition result is correct is determined as the distinction point to be adjusted; the distinction target weight value of the distinction point to be adjusted is adjusted to be greater than the distinction target weight value of the first distinction point.

[0088] It should be noted that by sequentially determining the next distinguishing point, the accuracy of recognition can be gradually verified to ensure that the recognition result is correct, which helps to optimize the recognition process and improve recognition efficiency.

[0089] ‌Example‌:

[0090] Assume that in a license plate recognition scenario, the target distinguishing information includes features such as the license plate's outline, color, characters, and position, with corresponding weights of 0.3, 0.2, 0.4, and 0.1, respectively. In the first round of recognition, the character features are determined as the first distinguishing point, and the recognition result is determined based on this. In the second round of recognition, if the second recognition result is consistent with the first, the system will consider features such as outline, color, and position as the next distinguishing points for verification.

[0091] By determining the distinguishing points to be adjusted, the reliability of recognition can be further improved, ensuring the accuracy of the recognition results. This helps to optimize the weight distribution of the recognition model and improve the recognition performance of the model.

[0092] ‌Example‌:

[0093] Continuing with the license plate recognition example, suppose that in the second round of recognition, the contour feature is verified as the next distinguishing point and is the first feature that can confirm the correctness of the first recognition result. At this time, the contour feature will be determined as the distinguishing point to be adjusted.

[0094] By adjusting the weight values, you can optimize the weight distribution of the recognition model, making the model pay more attention to features that have a significant impact on the recognition results. This helps improve the adaptability of the model, enabling it to better cope with complex and changing recognition scenarios.

[0095] ‌Example‌:

[0096] In license plate recognition scenarios, if contour features are identified as distinguishing points to be adjusted, their corresponding weights will be adjusted to be greater than those of character features. This way, in subsequent recognition processes, the model will pay more attention to the contour features of the license plate, thereby improving recognition accuracy and reliability.

[0097] Based on the foregoing description, this embodiment adopts a combination of "dimensionality reduction and shearing processing + lightweight recognition model" to reduce the amount of data by first reducing the dimension, and then using a lightweight model for preliminary recognition, which can reduce the computational complexity. Compared with the full-scale calculation mode of traditional multi-stage neural networks, this application only calls complex models for error-prone types, effectively avoiding waste of resources. This technical feature achieves a balance between computational efficiency and recognition accuracy through a dynamic resource allocation strategy. When the preliminary recognition result is an error-prone type, the high-energy recognition model is called only when necessary through the progressive verification process of "differentiated database comparison → weight adjustment → secondary recognition". This layered processing avoids the defects of blindly using complex models in the prior art.

[0098] This embodiment includes five types of target information, such as "target pixel combination, texture features, and color distribution features", and dynamically determines key discrimination points through weight value relationships (such as "the first discrimination point that can be verified to be correct is used as the point to be adjusted"). Compared with the traditional fixed-weight dimensionality reduction method, this solution retains high-frequency key features through a weight feedback mechanism, solving the problem of loss of discrimination information caused by static feature selection in the existing technology. The preset error-prone feature weights in the discrimination database (such as the edge gradient distribution weight value) essentially form a feature enhancement mechanism for specific scenarios. Through the gradient adjustment of the weight value (such as "the weight value of the point to be adjusted is greater than the previous discrimination point"), adaptive optimization of feature discrimination capabilities is achieved.

[0099] This embodiment utilizes a lightweight model in conjunction with a complex model to allocate the majority of routine recognition tasks to the low-energy model while ensuring recognition accuracy (secondary recognition verification). This overall efficiency improvement aligns with the three-step approach's criteria for determining the effectiveness of technical problem solving. A weight adjustment mechanism enables the system to dynamically optimize feature discrimination priorities based on actual recognition results (e.g., adjusting "color distribution weight value > texture weight value"). This closed-loop feedback mechanism overcomes the limitations of traditional neural network fixed-parameter models.

[0100] In addition, this application Figure 2 The present invention provides an image classification system based on a neural network. Figure 2 As shown, the system provided in the embodiment of the present application mainly includes:

[0101] The acquisition module 210 is used to perform dimensionality reduction and shearing processing on the image to be recognized, and then input it into a preset lightweight recognition model to obtain a first recognition result.

[0102] The acquisition module 210 includes a model building unit for constructing a preset lightweight recognition model including 5 layers of convolution, and the convolution kernel size of each layer is set to 3×3, the step size is 2×2, and the depth-separable convolution is used instead of the standard convolution operation; a global average pooling layer is added at the end of the network of the preset lightweight recognition model instead of the fully connected layer.

[0103] The acquisition module 210 includes a shearing unit, which is used to compress the resolution of the image to be identified to a preset resolution and retain high-frequency features through a convolutional downsampling layer, and at the same time combine the Grad-CAM algorithm to locate the salient area and crop to obtain the identification area; use the principal component analysis technology to map the RGB three channels of the identification area into a single-channel grayscale image to obtain the image after dimensionality reduction and shearing processing.

[0104] The acquisition module 220 is used to call the preset distinction database to obtain the distinction target information and the distinction target weight value when the first recognition result belongs to the preset error-prone type; wherein the distinction target information includes at least: target pixel combination, target contour, texture feature, color distribution feature, edge gradient distribution, and the distinction target weight value includes at least: target pixel weight value, target contour weight value, texture feature weight value, color distribution feature weight value, edge gradient distribution weight value.

[0105] The acquisition module 220 includes an acquisition unit, which is used to perform addition, deletion, modification and query operations on the distinguishing target information through a preset information adjustment interface.

[0106] The result module 230 is used to determine the first distinguishing point from the distinguishing target information based on the relationship between the distinguishing target weight values, and determine whether the first recognition result is correct based on whether the first distinguishing point exists in the image to be recognized; the image to be recognized corresponding to the erroneous first recognition result is cut and then input into the image recognition model to obtain the second recognition result.

[0107] The adjustment module 240 is used to determine the next distinction point from the distinction target information in sequence according to the relationship between the weight values ​​of the distinction target information when the second recognition result is consistent with the first recognition result, and determine the next distinction point that can first determine that the first recognition result is correct as the distinction point to be adjusted; adjust the distinction target weight value of the distinction point to be adjusted to be greater than the distinction target weight value of the first distinction point.

[0108] The adjustment module 240 includes an adjustment unit configured to determine the distinguishing target information with the largest distinguishing target weight value as the first distinguishing point.

[0109] Thus far, the technical solutions of the present disclosure have been described in conjunction with the foregoing multiple embodiments. However, it is easy for those skilled in the art to understand that the scope of protection of the present disclosure is not limited to these specific embodiments. Without departing from the technical principles of the present disclosure, those skilled in the art may split and combine the technical solutions in the above-mentioned various embodiments, and may also make equivalent changes or replacements to the relevant technical features. Any changes, equivalent replacements, improvements, etc. made within the technical concepts and / or technical principles of the present disclosure will fall within the scope of protection of the present disclosure.

Claims

1. A neural network-based image classification method, characterized in that: The method comprises: Performing dimensionality reduction and shearing processing on the image to be recognized, and then inputting it into a preset lightweight recognition model to obtain a first recognition result; When the first recognition result belongs to a preset error-prone type, calling a preset distinction database to obtain distinction target information and distinction target weight values; wherein the distinction target information includes at least: target pixel combination, target outline, texture feature, color distribution feature, edge gradient distribution; the distinction target weight values ​​include at least: target pixel weight value, target outline weight value, texture feature weight value, color distribution feature weight value, edge gradient distribution weight value; Determine a first distinguishing point from the distinguishing target information based on the relationship between the distinguishing target weight values, and determine whether the first recognition result is correct based on whether the first distinguishing point exists in the image to be recognized; crop the image to be recognized corresponding to the incorrect first recognition result and then input it into the image recognition model to obtain a second recognition result; When the second recognition result is consistent with the first recognition result, the next distinction point is determined from the distinction target information in turn according to the relationship between the weight values ​​of the distinction target information, and the next distinction point that can first determine that the first recognition result is correct is determined as the distinction point to be adjusted; the distinction target weight value of the distinction point to be adjusted is adjusted to be greater than the distinction target weight value of the first distinction point.

2. The neural network-based image classification method according to claim 1, characterized in that: Before performing dimensionality reduction and shearing processing on the image to be recognized and then inputting the image into a preset lightweight recognition model to obtain a first recognition result, the method further includes: Build a preset lightweight recognition model with 5 convolution layers. The kernel size of each layer is set to 3×3 and the stride is 2×2. Depthwise separable convolution is used instead of standard convolution operation. Add a global average pooling layer instead of a fully connected layer at the end of the network of the preset lightweight recognition model.

3. The neural network-based image classification method according to claim 1, characterized in that: The image to be recognized is subjected to dimensionality reduction and shearing processing, and then input into a preset lightweight recognition model to obtain a first recognition result, specifically including: The convolutional downsampling layer compresses the resolution of the image to be identified to the preset resolution while retaining high-frequency features. At the same time, the Grad-CAM algorithm is used to locate the salient area and crop it to obtain the identification area. The principal component analysis technique is used to map the RGB three channels of the recognition area into a single-channel grayscale image to obtain the image after dimensionality reduction and shearing processing.

4. The neural network-based image classification method according to claim 1, characterized in that: When the first recognition result belongs to the preset error-prone type, before calling the preset distinction database to obtain distinction target information, the method further includes: Through the preset information adjustment interface, add, delete, modify and query the distinguishing target information.

5. The image classification method based on neural network according to claim 1, characterized in that: Determining a first distinguishing point from the distinguishing target information based on a relationship between distinguishing target weight values ​​specifically includes: The distinguishing target information with the largest distinguishing target weight value is determined as the first distinguishing point.

6. A neural network-based image classification system, characterized in that: The system comprises: An acquisition module is used to perform dimensionality reduction and shearing processing on the image to be recognized, and then input it into a preset lightweight recognition model to obtain a first recognition result; an acquisition module, configured to, when the first recognition result belongs to a preset error-prone type, call a preset distinction database to obtain distinction target information and a distinction target weight value; wherein the distinction target information includes at least: a target pixel combination, a target outline, a texture feature, a color distribution feature, and an edge gradient distribution; and the distinction target weight value includes at least: a target pixel weight value, a target outline weight value, a texture feature weight value, a color distribution feature weight value, and an edge gradient distribution weight value; A result module is configured to determine a first distinguishing point from the distinguishing target information based on the relationship between the distinguishing target weight values, and determine whether the first recognition result is correct based on whether the first distinguishing point exists in the image to be recognized; the image to be recognized corresponding to the incorrect first recognition result is cropped and then input into the image recognition model to obtain a second recognition result; An adjustment module is used to determine the next distinction point from the distinction target information in sequence according to the relationship between the weight values ​​of the distinction target information when the second recognition result is consistent with the first recognition result, and determine the next distinction point that can first determine that the first recognition result is correct as the distinction point to be adjusted; and adjust the distinction target weight value of the distinction point to be adjusted to be greater than the distinction target weight value of the first distinction point.

7. The neural network-based image classification system according to claim 6, characterized in that: The acquisition module includes model building units, It is used to build a preset lightweight recognition model with 5 convolution layers. The kernel size of each layer is set to 3×3, the stride is 2×2, and the depthwise separable convolution is used instead of the standard convolution operation. Add a global average pooling layer instead of a fully connected layer at the end of the network of the preset lightweight recognition model.

8. The neural network-based image classification system according to claim 6, characterized in that: The acquisition module includes a shear unit, It is used to compress the resolution of the image to be identified to the preset resolution through the convolution downsampling layer and retain the high-frequency features. At the same time, it combines the Grad-CAM algorithm to locate the salient area and crop it to obtain the identification area; The principal component analysis technique is used to map the RGB three channels of the recognition area into a single-channel grayscale image to obtain the image after dimensionality reduction and shearing processing.

9. The neural network-based image classification system according to claim 6, characterized in that: The acquisition module includes an acquisition unit, Used to add, delete, modify and query distinguishing target information through the preset information adjustment interface.

10. The neural network-based image classification system according to claim 6, characterized in that: The adjustment module includes an adjustment unit, The distinction target information with the largest distinction target weight value is used to determine the first distinction point.

Images