Intelligent image recognition and commodity classification system

Abstract

The invention belongs to the technical field of information recommendation, and particularly relates to an intelligent image recognition and commodity classification system. Comprising an image acquisition module, an image processing and feature extraction module, a deep learning model training module, a commodity classification and identification module, a user interface module, a database module, a system integration and deployment module, an update and maintenance module and a security module. The invention provides an intelligent image recognition and commodity classification system, and a multi-layer perceptron is adopted as a feature representation unit in the intelligent image recognition and commodity classification system, so that the system can capture a more complex nonlinear relationship in an image. Therefore, abstract features in the image can be better expressed, and the recognition capability of the system on complex modes is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of information recommendation, and specifically refers to an intelligent image recognition and commodity classification system. Background Art

[0002] With the booming development of e-commerce, people are increasingly relying on online platforms during the shopping process, making product image recognition and classification an important issue in improving shopping experience and efficiency. As a technical means to meet this demand, the intelligent image recognition and product classification system aims to use computer vision and machine learning technology to automatically identify and classify product images, providing more convenient and accurate product information.

[0003] The system first involves an image acquisition module, which is responsible for collecting product images from different channels or sources. These images may come from e-commerce platforms, social media or other online channels. The image processing and feature extraction module pre-processes the collected product images, including resizing, noise removal and other operations, and extracts image features to prepare for subsequent classification work.

[0004] However, the existing intelligent image recognition and product classification systems still have some shortcomings. First, the processing and feature extraction capabilities of some systems for product images still need to be improved, especially when dealing with complex scenes and diverse product images. Second, some systems may be limited by the quality and scale of the data set when training models, resulting in insufficient model generalization capabilities. Summary of the invention

[0005] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent image recognition and commodity classification system. In order to solve the problem that the processing and feature extraction capabilities of the existing system for commodity images still need to be improved, especially when processing complex scenes and diversified commodity images, the present invention proposes an intelligent image recognition and commodity classification system, which uses a multi-layer perceptron as a feature representation unit in the intelligent image recognition and commodity classification system, so that it can capture more complex nonlinear relationships in the image. This helps to better express the abstract features in the image and improve the system's ability to recognize complex patterns.

[0006] The technical solution adopted by the present invention is as follows: The present invention provides an intelligent image recognition and commodity classification system, including:

[0007] Image acquisition module: responsible for collecting product images from different channels or sources;

[0008] Image processing and feature extraction module: pre-process the collected product images;

[0009] Deep learning model training module: select appropriate deep learning algorithms and establish product classification models;

[0010] Product classification and identification module: implements image recognition algorithm and uses trained models to classify product images;

[0011] User interface module: provides a user-friendly interface that enables users to upload product images or take photos using a camera;

[0012] Database module: stores product data, model parameters and other system information;

[0013] System integration and deployment module: integrate various modules into a complete system;

[0014] Update and maintenance module: provide regular model updates and system maintenance to adapt to new products and improve classification accuracy; Security module: consider conducting security assessments on the system to ensure that images and data uploaded by users are properly handled to prevent malicious attacks.

[0015] Furthermore, the image acquisition module includes:

[0016] The data source interface is responsible for connecting with different data sources or channels and providing interfaces to obtain product image data from these data sources; the image acquisition unit is used to obtain product images from data sources, obtain images from online stores through APIs, and obtain from web pages through crawler technology;

[0017] Data cleaning and preprocessing unit, which cleans and preprocesses product images obtained from different data sources to ensure image quality and consistency;

[0018] The image storage unit stores the cleaned and pre-processed images in the system for subsequent processing and training. Furthermore, the image processing and feature extraction module includes:

[0019] An image preprocessing unit further preprocesses the commodity images obtained from the image acquisition module to ensure image quality and consistency;

[0020] The feature extraction unit uses image processing technology to extract the features of the product image and the visual features related to the product category;

[0021] The feature representation unit represents the features extracted from the image in a format that can be processed by the deep learning model;

[0022] The feature storage unit stores the extracted features in the system for subsequent model training and image classification.

[0023] Furthermore, the deep learning model training module includes:

[0024] Model selection unit, which selects appropriate deep learning algorithms to meet the needs of product classification;

[0025] The data set division and preparation unit divides the data obtained from the image acquisition module and the image processing and feature extraction module into training sets, validation sets, and test sets, prepares the data, ensures that its format meets the input requirements of the selected model, and performs label encoding and other processing;

[0026] Training configuration unit, configures the hyperparameters of model training and defines the loss function and optimization algorithm during training;

[0027] The model training unit uses the training set to train the selected model, monitors the performance during the training process, and uses the validation set to perform early stopping and other strategies on the model;

[0028] Model evaluation and optimization unit, which uses the test set to evaluate the trained model, checks its performance on unseen data, and adjusts and optimizes the model based on the evaluation results to improve its accuracy and generalization ability;

[0029] The model saving and exporting unit saves the trained model in a format that can be used for inference after training is completed for subsequent use.

[0030] Furthermore, the image preprocessing unit adopts a Gaussian blur algorithm, and the algorithm formula is as follows:

[0031]

[0032] in:

[0033] G(x, y) is the value of the Gaussian kernel;

[0034] σ is the standard deviation of the Gaussian kernel, which controls the degree of blur;

[0035] (x, y) is the coordinate of the pixel;

[0036] Where e is the base of the natural logarithm. In image processing, Gaussian blur is achieved by applying this kernel at each pixel position in the image. The blur effect depends on the size of σ. A larger σ will result in a larger blur radius.

[0037] Furthermore, the deep learning model training module includes:

[0038] Model selection unit, which selects appropriate deep learning algorithms to meet the needs of product classification;

[0039] The data set division and preparation unit divides the data obtained from the image acquisition module and the image processing and feature extraction module into training sets, validation sets, and test sets, prepares the data, ensures that its format meets the input requirements of the selected model, and may perform label encoding and other processing;

[0040] The training configuration unit configures the hyperparameters of model training, such as learning rate and batch size, and defines the loss function and optimization algorithm during training.

[0041] The model training unit uses the training set to train the selected model, monitors the performance during the training process, and uses the validation set to perform the early stopping strategy of the model.

[0042] Model evaluation and optimization unit, which uses the test set to evaluate the trained model, checks its performance on unseen data, and adjusts and optimizes the model based on the evaluation results to improve its accuracy and generalization ability;

[0043] The model saving and exporting unit saves the trained model in a format that can be used for inference after training is completed for subsequent use.

[0044] Furthermore, the commodity classification and identification module includes:

[0045] The image input unit receives product images uploaded by users or real-time images captured by cameras, and passes the images to subsequent processing units for classification and recognition;

[0046] The classification algorithm unit uses the trained machine learning or deep learning model to classify the image. This unit uses the parameters and weights of the previously trained model to map the image to the corresponding product category.

[0047] A result output unit, which presents the classification results to the user;

[0048] User feedback unit collects user feedback on classification results.

[0049] Furthermore, the classification algorithm unit adopts a support vector machine algorithm, and the algorithm formula is as follows:

[0050] Assume that there are training samples (x 1 ,y 1 ), (x 2 ,y 2 ),...,(x n ,y n ), where x i is the input feature vector, y i is the corresponding class label, y i ∈{-1, 1};

[0051] Decision function of linear SVM:

[0052] The decision function of SVM is defined as:

[0053] f(x)=sign(w·x+b);

[0054] Where w is the weight vector, x is the input feature vector, b is the bias (intercept), · represents the dot product of the vectors, and sign is the sign function, with the output being -1 or 1;

[0055] The optimization goal of linear SVM is:

[0056] The goal of SVM is to find a maximum margin hyperplane such that for all i, the following conditions are satisfied:

[0057] y i (w·x i +b)≥1;

[0058] Among them, y i is the category label of sample i;

[0059] Optimization problem of linear SVM:

[0060] The optimization problem of SVM can be expressed as:

[0061]

[0062] Satisfy the constraints: y i (w·x i +b)≥1 for all i;

[0063] in, is a regularization term that aims to minimize the norm of the weight vector to prevent overfitting.

[0064] Furthermore, the feature representation unit adopts a multi-layer perceptron algorithm, and the algorithm formula is as follows:

[0065] Let the input be X=(x 1 , x 2 , ..., x n ), the weight of the hidden layer is W (1) , the bias of the hidden layer is b (2) , the weight of the output layer is W (2) , the bias of the output layer is b (2) , the activation function of the hidden layer is f, the activation function of the output layer is g; the output of the hidden layer is:

[0066] Z (1) =f(W (1) ·X+b (1) );

[0067] Where W (1) X represents the dot product of the weight and the input, b (1) is the bias of the hidden layer, f is the activation function of the hidden layer; the output of the output layer:

[0068] Z (2) =g(W (2) ·Z (1) +b (2) );

[0069] Where W (2) ·Z (1) represents the dot product of the weight and the hidden layer output, b (2) is the bias of the output layer, and g is the activation function of the output layer;

[0070] Loss function:

[0071] If it is a supervised learning task, a loss function can be used to measure the difference between the model output and the actual label;

[0072] Backward Propagation:

[0073] Using optimization algorithms such as gradient descent, the gradient of the loss function with respect to the network parameters is calculated through backpropagation, and then the weights and biases are updated to minimize the loss function.

[0074] This solution also discloses an operation method of the intelligent image recognition and commodity classification system, which mainly includes the following steps:

[0075] Step A1: Image acquisition, collect product images from different channels or sources;

[0076] Step A2: Data cleaning and preprocessing: Data cleaning and preprocessing of the collected product images to ensure image quality and consistency;

[0077] Step A3: Feature extraction, using image processing technology to extract the features of product images for subsequent classification work;

[0078] Step A4: Model training: select an appropriate deep learning algorithm, build a product classification model, and use the training data set to train the model so that it can accurately identify different product categories;

[0079] Step A5: Model evaluation, evaluate the trained model to ensure it has good performance on the validation set or test set; Step A6: User interface design, provide a user-friendly interface that allows users to upload product images or take photos with a camera; Step A7: Image classification and recognition, implement image recognition algorithms, and use the trained model to classify product images;

[0080] Step A8: Result display, displaying the recognition results and related information to optimize the user experience;

[0081] Step A9: Data storage, storing product data, model parameters and other system information, used for persistent storage of product classification results and product information uploaded by users;

[0082] Step A10: System deployment and maintenance, integrating various modules into a complete system.

[0083] The beneficial effects achieved by the present invention using the above structure are as follows: The present invention provides an intelligent image recognition and commodity classification system, achieving the following beneficial effects:

[0084] (1) Using multi-layer perceptron as the feature representation unit in the intelligent image recognition and product classification system enables it to capture more complex nonlinear relationships in the image. This helps to better express the abstract features in the image and improve the system's ability to recognize complex patterns.

[0085] (2) Using multi-layer perceptron as the feature representation unit in the intelligent image recognition and product classification system, the hidden layer of MLP can automatically learn to adapt to features of different scales and directions, thereby capturing more comprehensive information in the image. This is very helpful for the size differences and direction changes that may exist in product images.

[0086] (3) Multilayer perceptrons are used as feature representation units in intelligent image recognition and product classification systems. By stacking multiple hidden layers, MLP can gradually perform multi-level feature abstraction. This means that the system can gradually learn from simple features such as edges and textures at the bottom level to higher-level abstract features, which helps to understand product images more comprehensively.

[0087] (4) The use of multi-layer perceptrons as feature representation units in intelligent image recognition and product classification systems can be adjusted according to the needs of specific tasks. By adjusting the number of neurons and layers in the hidden layer, the system can adapt to different data and tasks, improving the adaptability of the system.

[0088] (5) Multilayer perceptron is used as the feature representation unit in intelligent image recognition and product classification systems. Product classification usually involves multiple categories, and MLP can easily adapt to multi-category problems in the output layer.

[0089] (6) Support vector machine is used as the classification algorithm unit in the intelligent image recognition and product classification system. In image recognition, the image data after feature extraction is often high-dimensional. SVM can effectively handle this situation, especially for the complex features output by the feature representation unit in the system.

[0090] (7) Using support vector machines as the classification algorithm unit in intelligent image recognition and product classification systems is very helpful in solving the complex nonlinear relationships and patterns existing in product images, and improves the system's ability to fit image data.

[0091] (8) The support vector machine is used as the classification algorithm unit in the intelligent image recognition and product classification system. It has strong generalization ability and can maintain good classification performance when facing unseen images. This is of great significance for the application of the system in real scenarios.

[0092] (9) The use of support vector machine as the classification algorithm unit in the intelligent image recognition and product classification system has good adaptability to small sample problems, which is beneficial to the sample imbalance and small sample problems that may be faced in the product classification system.

[0093] (10) Using support vector machine as the classification algorithm unit in the intelligent image recognition and product classification system can resist noise and interference in the image to a certain extent and improve the system's requirements for image quality and consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 This is a flow chart of the intelligent image recognition and commodity classification system proposed in the present invention.

[0095] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0096] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0097] Embodiment 1:

[0098] See also Figure 1 As shown, this embodiment is an intelligent image recognition and commodity classification system, including an image acquisition module, an image processing and feature extraction module, a deep learning model training module, a commodity classification and recognition module, a user interface module, a database module, a system integration and deployment module, an update and maintenance module and a security module;

[0099] Among them, the image acquisition module is responsible for collecting product images from different channels or sources;

[0100] Among them, the image processing and feature extraction module: pre-processes the collected product images;

[0101] Among them, the deep learning model training module: selects appropriate deep learning algorithms and establishes a product classification model;

[0102] Among them, the commodity classification and identification module: implements the image recognition algorithm and uses the trained model to classify the commodity images; Among them, the user interface module: provides a user-friendly interface, allowing users to upload commodity images or take pictures with a camera; Among them, the database module: stores commodity data, model parameters and other system information;

[0103] Among them, the system integration and deployment module: integrates each module into a complete system;

[0104] Among them, the update and maintenance module: provides regular model updates and system maintenance to adapt to new products and improve classification accuracy;

[0105] Among them, the security module: consider conducting a security assessment of the system to ensure that the images and data uploaded by users are properly handled to prevent malicious attacks.

[0106] Embodiment 2:

[0107] See also Figure 1 As shown, this embodiment is a method for using the intelligent image recognition and commodity classification system, comprising the following steps:

[0108] Step A1: Image acquisition, collect product images from different channels or sources;

[0109] Step A2: Data cleaning and preprocessing: Data cleaning and preprocessing of the collected product images to ensure image quality and consistency;

[0110] Step A3: Feature extraction, using image processing technology to extract the features of product images for subsequent classification work;

[0111] Step A4: Model training: select an appropriate deep learning algorithm, build a product classification model, and use the training data set to train the model so that it can accurately identify different product categories;

[0112] Step A5: Model evaluation, evaluate the trained model to ensure it has good performance on the validation set or test set; Step A6: User interface design, provide a user-friendly interface that allows users to upload product images or take photos with a camera; Step A7: Image classification and recognition, implement image recognition algorithms, and use the trained model to classify product images;

[0113] Step A8: Result display, displaying the recognition results and related information to optimize the user experience;

[0114] Step A9: Data storage, storing product data, model parameters and other system information, used for persistent storage of product classification results and product information uploaded by users;

[0115] Step A10: System deployment and maintenance, integrating various modules into a complete system.

[0116] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0117] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. Intelligent image recognition and commodity classification system, characterized by: include: Image acquisition module: responsible for collecting product images from different channels or sources; Image processing and feature extraction module: pre-process the collected product images; Deep learning model training module: select appropriate deep learning algorithms and establish product classification models; Product classification and identification module: implements image recognition algorithm and uses trained models to classify product images; User interface module: provides a user-friendly interface that enables users to upload product images or take photos using a camera; Database module: stores product data, model parameters and other system information; System integration and deployment module: integrate various modules into a complete system; Update and maintenance module: provide regular model updates and system maintenance to adapt to new products and improve classification accuracy; Security module: consider conducting security assessments on the system to ensure that images and data uploaded by users are properly handled to prevent malicious attacks.

2. The intelligent image recognition and commodity classification system according to claim 1, characterized in that: The image acquisition module includes: a data source interface, which is responsible for connecting with different data sources or channels and providing an interface to obtain product image data from these data sources; an image acquisition unit, which is used to obtain product images from data sources, obtain pictures from online stores through APIs, and obtain from web pages through crawler technology; Data cleaning and preprocessing unit, which cleans and preprocesses product images obtained from different data sources to ensure image quality and consistency; The image storage unit stores the cleaned and preprocessed images into the system for subsequent processing and training.

3. The intelligent image recognition and commodity classification system according to claim 2, characterized in that: Image processing and feature extraction modules include: An image preprocessing unit further preprocesses the commodity images obtained from the image acquisition module to ensure image quality and consistency; The feature extraction unit uses image processing technology to extract the features of the product image and the visual features related to the product category; The feature representation unit represents the features extracted from the image in a format that can be processed by the deep learning model; The feature storage unit stores the extracted features in the system for subsequent model training and image classification.

4. The intelligent image recognition and commodity classification system according to claim 3, characterized in that: The deep learning model training module includes: Model selection unit, which selects appropriate deep learning algorithms to meet the needs of product classification; The data set division and preparation unit divides the data obtained from the image acquisition module and the image processing and feature extraction module into training sets, validation sets, and test sets, prepares the data, ensures that its format meets the input requirements of the selected model, and performs label encoding and other processing; Training configuration unit, configures the hyperparameters of model training and defines the loss function and optimization algorithm during training; The model training unit uses the training set to train the selected model, monitors the performance during the training process, and uses the validation set to perform early stopping and other strategies on the model; Model evaluation and optimization unit, which uses the test set to evaluate the trained model, checks its performance on unseen data, and adjusts and optimizes the model based on the evaluation results to improve its accuracy and generalization ability; The model saving and exporting unit saves the trained model in a format that can be used for inference after training is completed for subsequent use.

5. The intelligent image recognition and commodity classification system according to claim 4, characterized in that: The image preprocessing unit adopts Gaussian blur algorithm, and the algorithm formula is as follows: in: G(x, y) is the value of the Gaussian kernel; σ is the standard deviation of the Gaussian kernel, which controls the degree of blur; (x, y) is the coordinate of the pixel; Where e is the base of the natural logarithm. In image processing, Gaussian blur is achieved by applying this kernel at each pixel position in the image. The blur effect depends on the size of σ. A larger σ will result in a larger blur radius.

6. The intelligent image recognition and commodity classification system according to claim 5, characterized in that: The deep learning model training module includes: Model selection unit, which selects appropriate deep learning algorithms to meet the needs of product classification; The data set division and preparation unit divides the data obtained from the image acquisition module and the image processing and feature extraction module into training sets, validation sets, and test sets, prepares the data, ensures that its format meets the input requirements of the selected model, and may perform label encoding and other processing; The training configuration unit configures the hyperparameters of model training, such as learning rate and batch size, and defines the loss function and optimization algorithm during training. The model training unit uses the training set to train the selected model, monitors the performance during the training process, and uses the validation set to perform the model's early stopping strategy; Model evaluation and optimization unit, which uses the test set to evaluate the trained model, checks its performance on unseen data, and adjusts and optimizes the model based on the evaluation results to improve its accuracy and generalization ability; The model saving and exporting unit saves the trained model in a format that can be used for inference after training is completed for subsequent use.

7. The intelligent image recognition and commodity classification system according to claim 6, characterized in that: The commodity classification and identification module includes: The image input unit receives product images uploaded by users or real-time images captured by cameras, and passes the images to subsequent processing units for classification and recognition; The classification algorithm unit uses the trained machine learning or deep learning model to classify the image. This unit uses the parameters and weights of the previously trained model to map the image to the corresponding product category. A result output unit, which presents the classification results to the user; User feedback unit collects user feedback on classification results.

8. The intelligent image recognition and commodity classification system according to claim 7, characterized in that: The classification algorithm unit adopts a support vector machine algorithm, and the algorithm formula is as follows: Assume that there are training samples (x1, y1), (x2, y2), ..., (x n ,y n ), where x i is the input feature vector, y i is the corresponding class label, y i ∈{-1, 1}; Decision function of linear SVM: The decision function of SVM is defined as: f(x)=sign(w.x+b); Where w is the weight vector, x is the input feature vector, b is the bias (intercept), · represents the dot product of the vectors, and sign is the sign function, with the output being -1 or 1; The optimization goal of linear SVM is: The goal of SVM is to find a maximum margin hyperplane such that for all i, the following conditions are satisfied: y i (w·x i +b)≥1; Among them, y i is the category label of sample i; Optimization problem of linear SVM: The optimization problem of SVM can be expressed as: Satisfy the constraints: y i (w·x i +b)≥1 for all i; in, is a regularization term that aims to minimize the norm of the weight vector to prevent overfitting.

9. The intelligent image recognition and commodity classification system according to claim 8, characterized in that: The feature representation unit adopts a multi-layer perceptron algorithm, and the algorithm formula is as follows: Let the input be X = (x1, x2, ..., x n ), the weight of the hidden layer is W (1) , the bias of the hidden layer is b (1) , the weight of the output layer is W (2) , the bias of the output layer is b (2) , the activation function of the hidden layer is f, and the activation function of the output layer is g; Output of the hidden layer: Z (1) =f(W (1) ·X+b (1) ); Where W (1) X represents the dot product of the weight and the input, b (1) is the bias of the hidden layer, and f is the activation function of the hidden layer; Output of the output layer: Z (2) =g(W (2) ·Z (1) +b (2) ); Where W (2) ·Z (1) represents the dot product of the weight and the hidden layer output, b (2) is the bias of the output layer, and g is the activation function of the output layer; Loss function: If it is a supervised learning task, a loss function can be used to measure the difference between the model output and the actual label; Backward Propagation: Using optimization algorithms such as gradient descent, the gradient of the loss function with respect to the network parameters is calculated through backpropagation, and then the weights and biases are updated to minimize the loss function.

10. The method for operating the intelligent image recognition and commodity classification system is characterized by: The operation using the intelligent image recognition and commodity classification system of claim 9 mainly includes the following steps: Step A1: Image acquisition, collect product images from different channels or sources; Step A2: Data cleaning and preprocessing: Data cleaning and preprocessing of the collected product images to ensure image quality and consistency; Step A3: Feature extraction, using image processing technology to extract the features of product images for subsequent classification work; Step A4: Model training: select an appropriate deep learning algorithm, build a product classification model, and use the training data set to train the model so that it can accurately identify different product categories; Step A5: Model evaluation: evaluate the trained model to ensure that it has good performance on the validation set or test set; Step A6: User interface design, providing a user-friendly interface that allows users to upload product pictures or take photos using a camera; Step A7: Image classification and recognition, implement image recognition algorithm, and use the trained model to classify product images; Step A8: Result display, displaying the recognition results and related information to optimize the user experience; Step A9: Data storage, storing product data, model parameters and other system information, used for persistent storage of product classification results and product information uploaded by users; Step A10: System deployment and maintenance, integrating various modules into a complete system.

Images