Efficient agricultural product purchasing method based on AI visual analysis
By using AI visual analysis technology in the process of agricultural product acquisition, the identity verification, image acquisition and feature extraction are automated, and the problems of artificial error and subjectivity in traditional acquisition methods are solved, and efficient and accurate agricultural product acquisition and quality evaluation are achieved.
Patent Information
- Application Number
- CN202411768064.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-04
- Publication Date
- 2025-03-28
AI Technical Summary
Traditional agricultural product purchase methods rely on manual judgment, which is prone to artificial errors and subjectivity, which affects the accuracy of agricultural product classification and quality evaluation.
An efficient agricultural product acquisition method based on AI visual analysis is adopted, and identity verification is performed through intelligent hardware devices, agricultural product images are collected using high-resolution cameras, image preprocessing and feature extraction are combined with convolutional neural networks, and classification and quality evaluation are performed using support vector machines and random forests.
The agricultural product acquisition process has been automated, the error and subjectivity of manual intervention have been reduced, and the accuracy and efficiency of agricultural product acquisition have been greatly improved. The classification accuracy is as high as 90% and the identification accuracy is as high as 95%.
Smart Images
Figure CN119722222A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of visual analysis of agricultural products, and in particular to an efficient agricultural product purchasing method based on AI visual analysis. Background Art
[0002] With the development of agricultural production and the growth of agricultural product trade, the efficiency and accuracy of agricultural product procurement have become increasingly important. Traditional agricultural product procurement methods mainly rely on manual identification and measurement, which have certain limitations and defects, such as manual errors, time-consuming, and high costs. In recent years, the development of computer vision and machine learning technologies has brought new opportunities for agricultural product procurement. The efficient agricultural product procurement method based on AI visual analysis can realize automated agricultural product procurement, improve procurement efficiency and accuracy, reduce manual errors and costs, and improve farmers' satisfaction. This method realizes automated agricultural product procurement and purchase document generation through steps such as identity recognition, image acquisition, image preprocessing, visual analysis, feature recognition, weight measurement, and data integration.
[0003] Traditional agricultural product procurement methods rely on manual judgment, which is prone to human error and subjectivity. This not only affects the accuracy of agricultural product classification and quality assessment, but also slows down procurement efficiency. Summary of the invention
[0004] In view of the shortcomings of the prior art, the present invention provides an efficient agricultural product purchasing method based on AI visual analysis, which solves the problem that manual judgment is prone to human errors and subjectivity, affecting the accuracy of agricultural product classification and quality assessment.
[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: an efficient agricultural product purchasing method based on AI visual analysis, comprising:
[0006] a. Through the intelligent hardware device, the farmer’s ID card information is associated, and the farmer’s business license or ID card information is automatically identified through optical character recognition technology. In order to improve the accuracy of identity verification, face recognition technology is used to compare the face image collected on site with the identity information in the database. The face recognition process is realized through a convolutional neural network, which can be calculated by the following formula:
[0007] f(x)=W·x+b;
[0008] Where f(x) is the output of the identity verification, x is the input face image feature, W is the weight of the convolutional neural network, and b is the bias term. The network optimizes the weights through back propagation to minimize the verification error.
[0009] b. After identity verification, a high-resolution camera installed on an electronic scale is used to collect images of agricultural products. The images are used as input data for subsequent analysis. The appearance features of agricultural products collected by the image sensor are converted into digital signals and transmitted to the back-end computing system;
[0010] c. Image preprocessing denoising is achieved through Gaussian filter or median filter. The calculation formula of this process is:
[0011]
[0012] Among them, I(x,y) is the original image, I preprocessed (x, y) is the preprocessed image, k is the size of the filter;
[0013] d. After the image is preprocessed, a convolutional neural network is used to extract features from the image. The convolutional neural network can automatically extract multi-level features from the image, including low-level features and high-level features. Feature extraction and dimensionality reduction are performed through the combination of convolutional layers and pooling layers, and finally the characteristic identification of agricultural products is obtained. The operation of the convolutional layer can be performed using the following formula:
[0014] y=ReLU(Σ i,j I(i,j)·K(i,j)+b);
[0015] Among them, I(i,j) is the input image, K(i,j) is the convolution kernel, b is the bias term, y is the output feature map, and ReLU is the activation function;
[0016] e. After feature extraction, the agricultural products are classified using the traditional machine learning methods of support vector machine and random forest. By training the classification model, the types of agricultural products can be identified based on the extracted image features, and the quality parameters of agricultural products can be predicted. The specific classification process can be identified by the objective function formula of the support vector machine:
[0017]
[0018] Among them, w is the weight vector of the support vector machine, x i is the input feature vector, y i is the category label and b is the bias term.
[0019] Regression analysis is used to predict continuous variables such as weight, size, and maturity of agricultural products. The regression model can be identified using the following formula:
[0020] y=W·x+b;
[0021] Among them, W is the weight of the regression model, x is the agricultural product feature vector, b is the bias term, and y is the predicted value;
[0022] f. Use smart electronic scales to measure the weight of agricultural products. The sensor will collect weight data and transmit it to the background system. The measured weight data will be processed by the calibration module to ensure the accuracy and consistency of the data. The weight calibration process uses a regression model for calibration. The formula is as follows:
[0023] w corrected =α·w raw +β;
[0024] Among them, w corrected is the corrected weight, w raw is the original measurement value, α and β are the coefficients of the regression model;
[0025] g. Automatically generate purchase orders.
[0026] Preferably, the image preprocessing process includes denoising, contrast adjustment, and edge enhancement operations to improve image quality, and the image preprocessing step also includes color correction and lighting compensation to further enhance the recognizability of agricultural product images.
[0027] Preferably, the low-level features are edges and textures, and the high-level features are object shapes and colors.
[0028] Preferably, the quality prediction of the agricultural product is performed by regression analysis, and the size, weight and maturity quality parameters of the agricultural product are predicted based on its appearance characteristics.
[0029] Preferably, the image acquisition step uses multiple sensors to work simultaneously to acquire images of agricultural products at different angles to improve image quality and recognition accuracy.
[0030] Preferably, the convolutional neural network includes at least three convolutional layers, two pooling layers and one fully connected layer to achieve multi-level extraction of image features. The convolutional neural network uses Dropout technology to prevent overfitting and improve the generalization ability of the model.
[0031] Preferably, the support vector machine adopts a Gaussian kernel function to adapt to nonlinear data distribution and improve classification accuracy, and the regression analysis adopts a gradient boosting regression tree model to improve the accuracy and robustness of predicting continuous variables, and the continuous variables are weight, size, and maturity.
[0032] Preferably, the weight correction model is based on historical weight data and temperature and humidity environmental parameters, and uses multiple regression analysis to perform weight correction.
[0033] The present invention provides an efficient agricultural product purchasing method based on AI visual analysis. It has the following beneficial effects:
[0034] This efficient agricultural product acquisition method based on AI visual analysis combines intelligent hardware devices with AI visual analysis technology to automatically perform identity authentication, agricultural product image acquisition and processing, feature extraction, classification and quality assessment, reducing the errors and subjectivity of manual intervention, and greatly improving the accuracy and efficiency of the agricultural product acquisition process. Especially in the agricultural product classification and quality prediction process, the AI model can make high-precision judgments based on image features and historical data, thereby ensuring accurate evaluation of agricultural products.
[0035] An algorithm based on deep learning is used to analyze images of agricultural products and accurately identify the types of agricultural products, with an identification accuracy rate of over 95%. In the visual analysis step, computer vision technology is used to extract the color, shape, size, quality and other characteristic information of agricultural products through image processing and analysis, and classify and evaluate them, with a classification accuracy rate of over 90%. By integrating the visual analysis results, feature recognition results and weight measurement results, detailed information of agricultural products is generated, and purchase documents are automatically generated, reducing manual errors and costs and improving farmers' satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 It is a schematic diagram of the internal structure of the present invention. DETAILED DESCRIPTION
[0037] The following will be combined with the drawings in 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.
[0038] like Figure 1 As shown, the embodiment of the present invention provides an efficient agricultural product acquisition method based on AI visual analysis, including: a. Associating the farmer's ID card information with the intelligent hardware device, automatically identifying the farmer's business license or ID card information through optical character recognition technology, and using face recognition technology to compare the face image collected on site with the identity information in the database in order to improve the accuracy of identity authentication. The face recognition process is implemented by a convolutional neural network, which can be specifically calculated by the following formula:
[0039] f(x)=W·x+b;
[0040] Among them, f(x) is the output of identity verification, x is the input face image feature, W is the weight of the convolutional neural network, and b is the bias term. The network optimizes the weights through back propagation to minimize the verification error.
[0041] b. After identity verification, a high-resolution camera installed on an electronic scale is used to capture images of agricultural products. The images are used as input data for subsequent analysis. The appearance features of the agricultural products captured by the image sensor are converted into digital signals and transmitted to the back-end computing system. The image acquisition process is combined with automatic focus and exposure adjustment technology to minimize image quality instability caused by changes in ambient light, further improving data reliability.
[0042] c. Image preprocessing denoising is achieved through Gaussian filter or median filter. The calculation formula of this process is:
[0043]
[0044] Among them, I(x,y) is the original image, I preprocessed (x, y) is the preprocessed image, k is the size of the filter, and the image preprocessing process includes denoising, contrast adjustment, and edge enhancement operations to improve image quality. The image preprocessing steps also include color correction and illumination compensation to further enhance the recognizability of agricultural product images. This technology can automatically adjust the enhancement intensity according to the characteristic differences of different agricultural products to ensure that the key information of different categories of agricultural product images is optimally extracted and the recognition accuracy is improved.
[0045] d. After the image is preprocessed, a convolutional neural network is used to extract features from the image. The convolutional neural network can automatically extract multi-level features from the image, including low-level features and high-level features. The low-level features are edges and textures, and the high-level features are object shapes and colors. The convolutional layer and the pooling layer are combined to extract features and reduce dimensions, and finally obtain the characteristic identification of agricultural products. The operation of the convolutional layer can be performed by the following formula:
[0046] y=ReLU(∑ i,j I(i,j)·K(i,j)+b);
[0047] Among them, I(i,j) is the input image, K(i,j) is the convolution kernel, b is the bias term, y is the output feature map, ReLU is the activation function, the image acquisition step uses multiple sensors to work simultaneously to collect agricultural product images at different angles to improve image quality and recognition accuracy, the convolutional neural network includes at least three convolutional layers, two pooling layers and one fully connected layer to achieve multi-level extraction of image features, the convolutional neural network uses Dropout technology to prevent overfitting and improve the generalization ability of the model, in order to further improve the training efficiency and recognition accuracy of the convolutional neural network, the present invention adopts a transfer learning strategy, based on a pre-trained large-scale image data set, fine-tunes the convolutional neural network model to accelerate the agricultural product image classification training process and improve recognition accuracy.
[0048] e. After feature extraction, the agricultural products are classified using the traditional machine learning methods of support vector machine and random forest. By training the classification model, the types of agricultural products can be identified based on the extracted image features, and the quality parameters of agricultural products can be predicted. The specific classification process can be identified by the objective function formula of the support vector machine:
[0049]
[0050] Among them, w is the weight vector of the support vector machine, x i is the input feature vector, y i is the category label, b is the bias term. When using support vector machines for classification, an integrated learning method is used to combine the prediction results of multiple SVM models for weighted averaging, thereby improving the stability and accuracy of classification. This is especially suitable for situations where there are many categories of agricultural products and diverse characteristics.
[0051] Regression analysis is used to predict continuous variables such as weight, size, and maturity of agricultural products. The regression model can be identified using the following formula:
[0052] y=W·x+b;
[0053] Among them, W is the weight of the regression model, x is the feature vector of agricultural products, b is the bias term, and y is the predicted value. The support vector machine uses the Gaussian kernel function to adapt to nonlinear data distribution and improve classification accuracy. The regression analysis uses the gradient boosting regression tree model to improve the accuracy and robustness of predicting continuous variables. The continuous variables are weight, size, and maturity.
[0054] f. Use smart electronic scales to measure the weight of agricultural products. The sensor will collect weight data and transmit it to the background system. The measured weight data will be processed by the calibration module to ensure the accuracy and consistency of the data. The weight calibration process uses a regression model for calibration. The formula is as follows:
[0055] w corrected =α·w raw +β;
[0056] Among them, w corrected is the corrected weight, w raw is the original measurement value, α and β are the coefficients of the regression model. The quality prediction of agricultural products is carried out through regression analysis, and the size, weight and maturity quality parameters of agricultural products are predicted according to their appearance characteristics. The weight correction model is based on historical weight data and temperature and humidity environmental parameters, and multiple regression analysis is used for weight correction.
[0057] g. Automatically generate purchase orders.
[0058] 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.
Claims
1. An efficient agricultural product purchasing method based on AI visual analysis, characterized in that: include: a. Through the intelligent hardware device, the farmer’s ID card information is associated, and the farmer’s business license or ID card information is automatically identified through optical character recognition technology. In order to improve the accuracy of identity verification, face recognition technology is used to compare the face image collected on site with the identity information in the database. The face recognition process is realized through a convolutional neural network, which can be calculated by the following formula: f(x)=W·x+b; Where f(x) is the output of the identity verification, x is the input face image feature, W is the weight of the convolutional neural network, and b is the bias term. The network optimizes the weights through back propagation to minimize the verification error. b. After identity verification, a high-resolution camera installed on an electronic scale is used to collect images of agricultural products. The images are used as input data for subsequent analysis. The appearance features of agricultural products collected by the image sensor are converted into digital signals and transmitted to the back-end computing system; c. Image preprocessing denoising is achieved through Gaussian filter or median filter. The calculation formula of this process is: Among them, I(x,y) is the original image, I preprocessed (x, y) is the preprocessed image, k is the size of the filter; d. After the image is preprocessed, a convolutional neural network is used to extract features from the image. The convolutional neural network can automatically extract multi-level features from the image, including low-level features and high-level features. Feature extraction and dimensionality reduction are performed through the combination of convolutional layers and pooling layers, and finally the characteristic identification of agricultural products is obtained. The operation of the convolutional layer can be performed using the following formula: Among them, I(i,j) is the input image, K(i,j) is the convolution kernel, b is the bias term, y is the output feature map, and ReLU is the activation function; e. After feature extraction, the agricultural products are classified using the traditional machine learning methods of support vector machine and random forest. By training the classification model, the types of agricultural products can be identified based on the extracted image features, and the quality parameters of agricultural products can be predicted. The specific classification process can be identified by the objective function formula of the support vector machine: Among them, w is the weight vector of the support vector machine, x i is the input feature vector, y i is the category label, b is the bias term; Regression analysis is used to predict continuous variables such as weight, size, and maturity of agricultural products. The regression model can be identified using the following formula: y=W·x+b; Among them, W is the weight of the regression model, x is the agricultural product feature vector, b is the bias term, and y is the predicted value; f. Use smart electronic scales to measure the weight of agricultural products. The sensor will collect weight data and transmit it to the background system. The measured weight data will be processed by the calibration module to ensure the accuracy and consistency of the data. The weight calibration process uses a regression model for calibration. The formula is as follows: w corrected =α·w raw +b; Among them, w corrected is the corrected weight, w raw is the original measurement value, α and β are the coefficients of the regression model; g. Automatically generate purchase orders.
2. The method for efficiently purchasing agricultural products based on AI visual analysis according to claim 1, characterized in that: The image preprocessing process includes denoising, contrast adjustment, and edge enhancement operations, aiming to improve image quality. The image preprocessing step also includes color correction and illumination compensation to further enhance the recognizability of agricultural product images.
3. The method for efficiently purchasing agricultural products based on AI visual analysis according to claim 1 is characterized in that: The low-level features are edges and textures, and the high-level features are object shapes and colors.
4. The method for efficiently purchasing agricultural products based on AI visual analysis according to claim 1 is characterized in that: The quality prediction of the agricultural products is carried out through regression analysis, and the size, weight and maturity quality parameters of the agricultural products are predicted based on their appearance characteristics.
5. The method for efficiently purchasing agricultural products based on AI visual analysis according to claim 1 is characterized in that: The image acquisition step uses multiple sensors to work simultaneously to acquire images of agricultural products at different angles to improve image quality and recognition accuracy.
6. The method for efficiently purchasing agricultural products based on AI visual analysis according to claim 1 is characterized in that: The convolutional neural network includes at least three convolutional layers, two pooling layers and one fully connected layer to achieve multi-level extraction of image features. The convolutional neural network uses Dropout technology to prevent overfitting and improve the generalization ability of the model.
7. The method for efficiently purchasing agricultural products based on AI visual analysis according to claim 1 is characterized in that: The support vector machine adopts a Gaussian kernel function to adapt to nonlinear data distribution and improve classification accuracy. The regression analysis adopts a gradient boosting regression tree model to improve the accuracy and robustness of predicting continuous variables, and the continuous variables are weight, size, and maturity.
8. The method for efficiently purchasing agricultural products based on AI visual analysis according to claim 1 is characterized in that: The weight correction model is based on historical weight data and temperature and humidity environmental parameters, and uses multiple regression analysis to perform weight correction.