A food quality control method and system based on intelligent image analysis

Abstract

This application provides a food quality control method and system based on intelligent image analysis. The method involves collecting real-time multi-angle, multi-spectral image streams from different locations on a food production line using a multi-sensor system, forming a comprehensive image dataset. Based on this comprehensive image dataset, a convolutional autoencoder is used for image denoising and contrast enhancement, combined with super-resolution reconstruction technology to improve image clarity and detail, generating a high-quality image dataset. Based on this high-quality image dataset, a deep residual network is used for anomaly detection processing to identify abnormal regions, and dynamic analysis is performed using time-series data analysis technology to generate anomaly detection results. Based on these anomaly detection results, the data is sent to the food production line control system to adjust production parameters and generate a food quality control strategy. The technical solution provided by this application improves product quality, reduces operating costs, and enhances market competitiveness.

Description

Technical Field

[0001] This application relates to the field of food quality control technology, and in particular to a food quality control method and system based on intelligent image analysis. Background Technology

[0002] With the rapid development of the food industry and the increasing demands of consumers for food safety and quality, there is a need for real-time monitoring of the appearance, color, and shape of food to detect problems such as foreign objects, spoilage, and damage. Traditional manual inspection methods are not only inefficient but also susceptible to human error, leading to inconsistent results. To achieve efficient and accurate quality control, a method is needed that can automatically acquire multi-angle, multispectral image streams and process them in real time using intelligent image analysis technology. This method should have functions such as image denoising, contrast enhancement, and super-resolution reconstruction to improve image quality. Simultaneously, it needs to incorporate deep learning models for anomaly detection and provide real-time feedback on detection results, supporting adjustments to production line parameters.

[0003] Currently, some food production companies use quality control systems based on traditional image processing techniques and simple machine learning algorithms. These systems use a single camera or sensor to collect images at a fixed location, and then perform preliminary analysis using image processing algorithms. They also utilize traditional image processing techniques (such as filtering and edge detection) for image denoising and enhancement, and apply some basic machine learning models (such as support vector machines and random forests) for anomaly detection.

[0004] Image data acquired by a single sensor cannot comprehensively reflect all situations on the production line, especially in complex environments with multiple angles and spectra, where the integrity and accuracy of the data are difficult to guarantee. Traditional image processing techniques have limited effectiveness in denoising and enhancement, and low-quality images directly affect the accuracy of anomaly detection, thereby reducing the effectiveness of product quality control and resulting in low product quality stability. Existing basic machine learning models are weak in feature extraction and pattern recognition, especially when dealing with complex and variable image data, which easily leads to false detections and missed detections. Existing solutions often only focus on the analysis of static images, ignoring dynamic factors that change during the production process, and cannot provide comprehensive anomaly detection results. Due to the speed limitations of data processing and analysis, existing solutions are difficult to achieve true real-time feedback, resulting in lag in production parameter adjustments, which affects production efficiency and product quality. Summary of the Invention

[0005] This application provides a food quality control method and system based on intelligent image analysis to solve the problem of poor food quality control in the prior art.

[0006] In a first aspect, embodiments of this application provide a food quality control method based on intelligent image analysis, comprising:

[0007] A multi-sensor system is used to collect real-time multi-angle, multi-spectral image streams from different locations on the food production line, forming a comprehensive image dataset.

[0008] Based on the comprehensive image dataset, a convolutional autoencoder is used to perform image denoising and contrast enhancement processing, combined with super-resolution reconstruction technology to improve image clarity and detail, and generate a high-quality image dataset.

[0009] Based on the high-quality image dataset, a deep residual network is used for anomaly detection to identify abnormal regions in the high-quality image dataset. Dynamic analysis is then performed using time-series data analysis techniques to generate anomaly detection results.

[0010] Based on the anomaly detection results, the data is sent to the food production line control system in real time via an encrypted interface to support timely adjustments to production parameters and the generation of food quality control strategies.

[0011] Optionally, based on the comprehensive image dataset, image denoising and contrast enhancement processing is performed using a convolutional autoencoder, combined with super-resolution reconstruction technology, to improve image clarity and detail, generating a high-quality image dataset, including:

[0012] Based on the comprehensive image dataset, random noise removal processing is performed using a convolutional autoencoder to improve image purity and generate a denoised image dataset.

[0013] Based on the denoised image dataset, the non-linear mapping capability of the convolutional autoencoder is utilized to analyze and adjust the image histogram, perform contrast enhancement processing, and generate a contrast-enhanced image dataset.

[0014] Based on the contrast-enhanced image dataset, combined with super-resolution reconstruction technology, high-resolution image detail prediction is performed to improve image resolution and clarity, and generate high-resolution images.

[0015] Based on the high-resolution image, local contrast is adjusted, edge sharpening is performed, and microscopic details are enhanced to improve image detail and generate a high-quality image dataset.

[0016] Optionally, the step of analyzing and adjusting the image histogram based on the denoised image dataset, utilizing the nonlinear mapping capability of a convolutional autoencoder, and performing contrast enhancement processing to generate a contrast-enhanced image dataset includes:

[0017] Based on the denoised image dataset, multi-scale key feature extraction processing is performed by pre-training a convolutional autoencoder to generate image dataset features;

[0018] Based on the features of the image dataset, the non-linear mapping capability of the convolutional autoencoder is used to transform the features of the image dataset to adapt to the image histogram adjustment requirements, thereby generating a feature-transformed image dataset.

[0019] Based on the aforementioned feature-transformed image dataset, the specific distribution of image histograms is analyzed, the contrast enhancement optimization space is identified, and image histogram analysis results are generated.

[0020] Based on the image histogram analysis results, the brightness of low-light areas is enhanced and the contrast of high-contrast areas is adjusted to improve the overall visual effect and generate a contrast-enhanced image dataset.

[0021] Optionally, based on the features of the image dataset, the nonlinear mapping capability of a convolutional autoencoder is used to transform the features of the image dataset to adapt to the image histogram adjustment requirements, generating a feature-transformed image dataset, including:

[0022] Based on the features of the image dataset, convolution operations are performed on the features of the image dataset using convolution kernels of different sizes to extract key features at different scales.

[0023] By fusing the key features at different scales, noise suppression is performed using a smoothing filter to remove high-frequency noise, thereby generating intermediate results of feature changes;

[0024] The intermediate results of the feature transformation are calculated using the following formula:

[0025]

[0026] Where T(x) is the intermediate result of the feature transformation; i is the index of the feature map, from 1 to N; N is the number of feature maps; W 1i x is the convolution kernel corresponding to the i-th feature map; i b1 is the i-th feature map; b1 is the bias term; tanh is the hyperbolic tangent activation function; α is the scaling factor; γ is the parameter controlling the exponential decay rate; j is the feature index participating in the exponential decay calculation, from 1 to M; M is the number of features participating in the exponential decay calculation; c is the center point value; x j For the j-th feature map;

[0027] Based on the intermediate results of the feature transformation, further nonlinear transformation is performed to enhance the feature representation capability. The most representative subset is obtained through the feature selection method to reduce redundant information and generate an enhanced feature representation.

[0028] The enhanced feature representation is calculated using the following formula:

[0029]

[0030] Where EFR(T(x)) is the enhanced feature representation; k is the index of the processed feature map, from 1 to K; K is the number of processed feature maps; W 2k T is the convolution kernel corresponding to the k-th feature map; k (x) represents the k-th processed feature map; b2 is the bias term; σ is the activation function; β is the scaling factor; ω is the parameter controlling the oscillation frequency of the sine wave; l is the feature index participating in the sine wave calculation, from 1 to L; L is the number of features participating in the sine wave calculation; T l (x) represents the l-th processed feature map; d represents the center point value of the sine wave; η represents the weighting coefficient of the logarithmic term; m represents the feature index for logarithmic calculation, from 1 to P; P represents the number of features involved in the logarithmic calculation; |T m (x)| represents the absolute value of the m-th feature map; δ is the scaling factor for the sigmoid term; θ is the parameter controlling the slope of the sigmoid function; n is the feature index participating in the sigmoid calculation, from 1 to Q; Q is the number of features participating in the sigmoid calculation; T n (x) represents the nth processed feature map; φ is the scaling factor for the square root term; o is the feature index participating in the square root calculation, from 1 to R; R is the number of features participating in the square root calculation; e is the center point value for the square root calculation; T o (x) is the o-th processed feature map;

[0031] Based on the enhanced feature representation, the image space is restored through inverse transformation, and histogram adjustment is performed to ensure image quality and visual effect, generating a feature-transformed image dataset.

[0032] Optionally, the step of predicting high-resolution image details based on the contrast-enhanced image dataset, combined with super-resolution reconstruction technology, to improve image resolution and sharpness and generate a high-resolution image includes:

[0033] Based on the contrast-enhanced image dataset, an image upsampling method is used to enlarge the image and generate a preliminary image framework.

[0034] Based on the preliminary image framework, and combined with the detailed information of the contrast-enhanced image dataset, interpolation techniques are used to fill in the missing pixels in the image to generate an optimized image framework.

[0035] Based on the optimized image framework, the super-resolution reconstruction technique using a deep learning mechanism is applied to further optimize the image framework, predict and supplement potential fine structures and texture features, and generate a reconstructed image framework.

[0036] Based on the reconstructed image framework, the overall sharpness and resolution are adjusted to ensure that there is no distortion during the image magnification process, thereby generating a high-resolution image.

[0037] Optionally, the step of using a deep residual network for anomaly detection processing based on the high-quality image dataset to identify anomalous regions in the high-quality image dataset, and combining this with time-series data analysis techniques for dynamic analysis to generate anomaly detection results includes:

[0038] Based on the high-quality image dataset, the vanishing gradient problem in deep learning is solved by introducing residual blocks, which enhances the model's learning ability and generates a preprocessed image dataset.

[0039] Based on the preprocessed image dataset, a deep residual network is used for anomaly detection. The anomaly probability value is calculated through forward propagation to generate preliminary anomaly detection results.

[0040] Based on the preliminary anomaly detection results, the time series patterns of the abnormal events are identified, dynamic analysis is performed, and dynamic analysis results are generated.

[0041] Based on the dynamic analysis results, a comprehensive evaluation is performed by combining the multi-dimensional key features of the abnormal event to generate anomaly detection results.

[0042] Optionally, the step of using a deep residual network for anomaly detection based on the preprocessed image dataset, calculating anomaly probability values ​​through forward propagation, and generating preliminary anomaly detection results includes:

[0043] Based on the preprocessed image dataset, the directional gradient histogram method is used to extract key features and generate a feature representation dataset.

[0044] Based on the aforementioned feature representation dataset, a deep residual network is used to compute the feature representations through forward propagation, generating a deep feature representation dataset.

[0045] Based on the aforementioned deep feature representation dataset, and combined with a deep residual network classification layer, anomaly probability values ​​are obtained, and anomaly probability value datasets are generated.

[0046] Based on the abnormal probability value dataset, a threshold judgment is performed within a preset reasonable threshold range, and the abnormal probability values ​​are marked and recorded to generate preliminary abnormal detection results.

[0047] Optionally, based on the feature representation dataset, a deep residual network is used to compute the feature representations through forward propagation to generate a deep feature representation dataset, including:

[0048] Based on the aforementioned feature representation dataset, the input feature map is preprocessed, and filtering techniques are used to reduce input noise and highlight key information in the input feature map in order to generate an intermediate output vector.

[0049] The intermediate output vector is calculated using the following formula:

[0050]

[0051] Among them, A i The intermediate output vector; N is the number of input feature maps; W ij X is the convolution kernel between the i-th output vector and the j-th input feature map; j b is the j-th feature map in the input feature vector; i V is the bias term of the i-th output vector; σ is the ReLU activation function; α and β are parameters controlling the strength of the nonlinear term; M is the number of input feature maps participating in the nonlinear transformation; V ik X represents the weights used for the nonlinear transformation between the i-th output vector and the k-th input feature map; j is the index of the input feature map, from 1 to N; k is the index of the input feature map, from 1 to M; k This represents the k-th input feature map that participates in the nonlinear transformation;

[0052] Based on the intermediate output vector, multiple levels of nonlinear transformations are used to capture complex patterns. Additional convolutional and pooling layers are introduced to increase expressive power and generalization performance in order to generate the final output vector.

[0053] The final output vector is calculated using the following formula:

[0054]

[0055] Among them, D i The final output vector; N is the number of input feature maps; W' il A is the convolution kernel between the i-th output vector and the l-th intermediate vector; l A is the l-th intermediate vector; m Let A be the m-th intermediate vector; n A is the nth intermediate vector; o b' is the o-th intermediate vector; i σ is the bias term of the i-th output vector; σ is the ReLU activation function; γ, δ, and θ are parameters controlling the strength of the nonlinear term; O is the number of intermediate vectors participating in the nonlinear transformation; U im Z represents the weights used for nonlinear transformation between the i-th output vector and the m-th intermediate vector; ∈ is a parameter controlling the strength of the fractional term; P and Q are the number of intermediate vectors involved in the operation in the numerator and denominator, respectively; in and Y io , respectively, are the weights between the i-th output vector in the numerator and the n-th or o-th intermediate vector; l is the index of the input feature map, from 1 to N; m is the index of the intermediate vector, from 1 to O; n is the index of the intermediate vector, from 1 to P; o is the index of the intermediate vector, from 1 to Q;

[0056] Based on the final output vector, standardization is performed, and dimensionality reduction is carried out through principal component analysis to map the high-dimensional feature vector into a low-dimensional space for visualization analysis, thereby generating a deep feature representation dataset.

[0057] Optionally, the step of sending the anomaly detection results to the food production line control system in real time via an encrypted interface to support timely adjustment of production parameters and generation of food quality control strategies includes:

[0058] Based on the anomaly detection results, the anomaly detection report is transmitted to the food production line control system in real time using an encrypted interface, generating an encrypted transmission result.

[0059] Based on the encrypted transmission results, combined with historical production line data and current production status, a comprehensive analysis is conducted to generate production line adjustment suggestions;

[0060] Based on the production line adjustment suggestions, the food production line control system automatically adjusts the production parameters and generates adjusted production parameter settings.

[0061] Based on the adjusted production parameter settings, the actual operating performance indicators are recorded. Combined with the anomaly detection results, the system is continuously adjusted and optimized through a feedback mechanism to generate a food quality control strategy.

[0062] Secondly, embodiments of this application provide a food quality control system based on intelligent image analysis, comprising:

[0063] The collection module is used to collect real-time image streams from different locations on the food production line from multiple angles and in multiple spectra using a multi-sensor system, forming a comprehensive image dataset;

[0064] The processing module is used to perform image denoising and contrast enhancement processing based on the comprehensive image dataset using a convolutional autoencoder, and combined with super-resolution reconstruction technology to improve image clarity and detail, thereby generating a high-quality image dataset.

[0065] The analysis module is used to perform anomaly detection processing based on the high-quality image dataset using a deep residual network, identify abnormal regions in the high-quality image dataset, and perform dynamic analysis by combining time-series data analysis techniques to generate anomaly detection results.

[0066] The sending module is used to send the anomaly detection results to the food production line control system in real time through an encrypted interface, so as to support timely adjustment of production parameters and generation of food quality control strategies.

[0067] In this embodiment, a multi-sensor system collects real-time multi-angle, multi-spectral image streams from different locations on the food production line to form a comprehensive image dataset. Based on this comprehensive image dataset, a convolutional autoencoder is used for image denoising and contrast enhancement, combined with super-resolution reconstruction technology to improve image clarity and detail, generating a high-quality image dataset. Based on this high-quality image dataset, a deep residual network is used for anomaly detection to identify abnormal regions in the high-quality image dataset, and dynamic analysis is performed using time-series data analysis technology to generate anomaly detection results. Based on these anomaly detection results, they are sent to the food production line control system in real time via an encrypted interface to support timely adjustments to production parameters and the generation of food quality control strategies. By employing a convolutional autoencoder to denoise and enhance the contrast of collected multi-angle, multispectral image streams, combined with super-resolution reconstruction technology, the system significantly improves image clarity and detail, providing more accurate foundational data for subsequent analysis. Anomaly detection is performed using a deep residual network model, which possesses powerful feature extraction capabilities and can effectively identify potential problem areas in the food production process. Combined with time-series data analysis technology, the system can not only detect anomalies at a static level but also capture dynamic trends, further improving the comprehensiveness and accuracy of anomaly detection. Anomaly detection results are rapidly fed back to the production line control system via an encrypted interface, supporting immediate adjustments to relevant production parameters and enabling rapid response to quality issues. The use of an encrypted interface for transmitting anomaly detection information ensures the security of sensitive data, preventing unauthorized access or tampering and safeguarding the company's information security. The entire solution embodies a highly integrated intelligent control concept, achieving an integrated solution from data acquisition and processing to decision support through effective monitoring and management of each stage of the production process. This promotes the digital transformation of the food industry and improves overall management levels.

[0068] Furthermore, random noise removal using a convolutional autoencoder effectively reduces noise in the image, improving its purity and aiding subsequent image processing steps, thus ensuring a solid foundation for image quality. Utilizing the nonlinear mapping capability of the convolutional autoencoder, the image histogram is adjusted for contrast enhancement, making details more apparent and improving visual appeal and recognizability. Combined with super-resolution reconstruction technology, high-resolution image detail prediction is performed on the denoised and contrast-enhanced image, significantly improving resolution and clarity, and facilitating more accurate capture and analysis of minute defects or anomalies in food production processes. Local contrast adjustment and edge sharpening of the high-resolution image further enhances the image's microscopic layers, making fine structures clearer and improving the accuracy and reliability of anomaly detection. The resulting high-quality image dataset not only possesses high definition and rich detail but also excellent contrast and purity, providing a high-quality data foundation for subsequent deep learning models, thereby improving the overall system performance.

[0069] Furthermore, by introducing residual blocks, the vanishing gradient problem, a common issue in deep learning, is addressed, enhancing the model's learning ability. This allows deep residual networks to better learn complex features, improving the accuracy of anomaly detection. Using deep residual networks for forward propagation to calculate anomaly probability values ​​generates preliminary anomaly detection results, enabling rapid identification of potential anomaly regions and providing a foundation for subsequent dynamic analysis. Based on these preliminary anomaly detection results, identifying time-series patterns of anomalous events and performing dynamic analysis captures the development trends and changing patterns of these events, improving the comprehensiveness and accuracy of anomaly detection. Combining multi-dimensional key features of anomalous events for comprehensive evaluation generates the final anomaly detection results, providing a comprehensive assessment of anomalous events from multiple perspectives and improving the reliability and robustness of anomaly detection. The generated anomaly detection results are sent to the food production line control system in real time via an encrypted interface, supporting timely adjustments to production parameters. This facilitates rapid response to quality issues during production, improves product quality stability, reduces losses, and optimizes the production process.

[0070] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0071] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0072] Figure 1A flowchart illustrating a food quality control method based on intelligent image analysis, provided as an embodiment of this application;

[0073] Figure 2 A schematic diagram of a food quality control system based on intelligent image analysis provided in an embodiment of this application;

[0074] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0075] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0076] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0077] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0078] Figure 1 A flowchart of a food quality control method based on intelligent image analysis is provided in this application embodiment, as shown below. Figure 1 As shown, the method includes:

[0079] 101. Collect real-time image streams from multiple angles and spectra at different locations on the food production line using a multi-sensor system to form a comprehensive image dataset;

[0080] The multi-sensor system includes multiple cameras and other sensors (such as infrared sensors, ultraviolet sensors, etc.) to collect real-time image data of the food production line from different positions and angles. This data includes various types of images such as visible light images, infrared images, and ultraviolet images, which are used to comprehensively monitor various situations in the production process.

[0081] A comprehensive image dataset refers to a dataset that integrates multi-angle, multi-spectral image data to form a complete dataset for subsequent image processing and analysis.

[0082] In this embodiment, firstly, a multi-sensor system installs multiple cameras and other sensors at different locations on the food production line to ensure that image data can be acquired from multiple angles and within multiple spectral ranges; secondly, these sensors acquire images in real time and transmit the data to a central processing system; thirdly, the central processing system performs preliminary processing on the acquired image data, including timestamp marking and location information recording, to ensure data integrity and traceability; finally, the processed image data is integrated into a comprehensive image dataset, providing a high-quality data foundation for subsequent steps such as image denoising, contrast enhancement, and anomaly detection.

[0083] Suppose there is a food production line that requires quality control during the packaging process;

[0084] First, the aforementioned cameras and other sensors are installed at key locations on the production line and calibrated to ensure they can accurately capture the required image data. Second, each sensor begins acquiring image data in real time; for example, the visible light camera acquires 30 frames per second, the infrared camera acquires 10 frames per second, and the ultraviolet camera acquires 5 frames per second. Third, the acquired image data is transmitted to the central processing system via wired or wireless network. The central processing system performs preliminary processing on the image data, including timestamp marking, location information recording, and basic format conversion. Finally, the central processing system integrates all the acquired image data into a comprehensive image dataset. This dataset contains all images acquired by each sensor at different time points, and each image is accompanied by a timestamp and location information to facilitate subsequent analysis and processing.

[0085] Through the above steps, a comprehensive image dataset containing real-time images from multiple angles and spectra is formed, providing a solid foundation for subsequent image processing and quality control.

[0086] 102. Based on the comprehensive image dataset, a convolutional autoencoder is used to perform image denoising and contrast enhancement processing. Combined with super-resolution reconstruction technology, the image clarity and detail are improved to generate a high-quality image dataset.

[0087] A convolutional autoencoder is a deep learning model used for unsupervised learning that learns representations of images through encoding and decoding processes.

[0088] Image denoising refers to removing noise from an image to improve its purity.

[0089] Contrast enhancement refers to adjusting the brightness distribution of an image to make the details in the image more apparent.

[0090] Super-resolution reconstruction is an image processing technique that uses algorithms to convert low-resolution images into high-resolution images, improving image clarity and detail.

[0091] High-quality image datasets refer to image datasets that, after the above processing, possess high resolution, rich detail, and good contrast.

[0092] In this step, firstly, a convolutional autoencoder is used to remove random noise from the comprehensive image dataset, generating a denoised image dataset. Secondly, the nonlinear mapping capability of the convolutional autoencoder is used to analyze and adjust the image histogram, performing contrast enhancement processing to generate a contrast-enhanced image dataset. Thirdly, combined with super-resolution reconstruction technology, high-resolution image detail prediction is performed on the contrast-enhanced image to improve the image resolution and sharpness, generating a high-resolution image. Finally, local contrast adjustment and edge sharpening processing are performed on the high-resolution image to further enhance the microscopic level of the image, generating the final high-quality image dataset.

[0093] Optionally, step 102, which involves using a convolutional autoencoder to perform image denoising and contrast enhancement processing based on the comprehensive image dataset, combined with super-resolution reconstruction technology to improve image clarity and detail, and generate a high-quality image dataset, includes: performing random noise removal processing using a convolutional autoencoder based on the comprehensive image dataset to improve image purity and generate a denoised image dataset; using the nonlinear mapping capability of the convolutional autoencoder to analyze and adjust the image histogram based on the denoised image dataset to perform contrast enhancement processing and generate a contrast-enhanced image dataset; using super-resolution reconstruction technology based on the contrast-enhanced image dataset to perform high-resolution image detail prediction, improve image resolution and clarity, and generate a high-resolution image; and adjusting local contrast and performing edge sharpening processing based on the high-resolution image to enhance microscopic levels and improve image detail, thereby generating a high-quality image dataset.

[0094] The step of generating a contrast-enhanced image dataset by analyzing and adjusting the image histogram using the non-linear mapping capability of a convolutional autoencoder based on the denoised image dataset includes: extracting key features at multiple scales using a pre-trained convolutional autoencoder to generate image dataset features; transforming the image dataset features using the non-linear mapping capability of the convolutional autoencoder to adapt to the image histogram adjustment requirements to generate a feature-transformed image dataset; analyzing the specific distribution of the image histogram based on the feature-transformed image dataset, identifying the contrast enhancement optimization space, and generating image histogram analysis results; and enhancing the brightness of low-light areas and adjusting the contrast of high-contrast areas based on the image histogram analysis results to improve the overall visual effect and generate a contrast-enhanced image dataset.

[0095] A convolutional autoencoder is a deep learning model used for unsupervised learning that learns representations of images through encoding and decoding processes.

[0096] Image denoising refers to removing noise from an image to improve its clarity. Contrast enhancement refers to adjusting the brightness distribution of an image to make details more apparent.

[0097] Super-resolution reconstruction is an image processing technique that uses algorithms to convert low-resolution images into high-resolution images, improving image clarity and detail.

[0098] High-quality image datasets refer to image datasets that, after the above processing, possess high resolution, rich detail, and good contrast.

[0099] In this embodiment, firstly, a convolutional autoencoder is used to remove random noise from the comprehensive image dataset to improve image purity and generate a denoised image dataset. Then, the nonlinear mapping capability of the convolutional autoencoder is used to extract key features at multiple scales, generating image dataset features. Secondly, based on these features, the nonlinear mapping capability of the convolutional autoencoder is used to transform the image dataset features to adapt to image histogram adjustment requirements, generating a feature-transformed image dataset. The specific distribution of the image histogram is analyzed to identify the contrast enhancement optimization space, generating image histogram analysis results. Thirdly, based on the image histogram analysis results, the brightness of low-light areas is enhanced, and the contrast of high-contrast areas is adjusted to improve the overall visual effect, generating a contrast-enhanced image dataset. Combined with super-resolution reconstruction technology, high-resolution image detail prediction is performed to improve image resolution and clarity, generating a high-resolution image. Finally, based on the high-resolution image, local contrast is adjusted, and edge sharpening is performed to enhance microscopic levels, improving image detail representation and generating a high-quality image dataset.

[0100] Suppose an organic food production line needs to control the quality of the packaging process. The production line is equipped with multiple cameras and other sensors, forming a comprehensive image dataset.

[0101] First, a pre-trained convolutional autoencoder is used to process the collected image data, removing random noise and generating a relatively clean image dataset. The autoencoder extracts key features and uses its non-linear mapping capabilities to transform these features to meet subsequent histogram adjustment requirements. Second, based on the transformed features, the system analyzes the specific distribution of the image histogram, identifying areas for contrast optimization. Based on the analysis, the system enhances the brightness of low-light areas while adjusting the contrast of high-contrast areas, thereby improving the overall visual effect and generating a contrast-enhanced image dataset. Third, based on the contrast-enhanced image dataset, the system employs super-resolution reconstruction technology to predict and generate higher-resolution images, further improving image clarity and detail. Finally, the system performs local contrast adjustment and edge sharpening on the high-resolution images, enhancing the fine structures within the images and ensuring that every detail is clearly visible, ultimately generating a high-quality image dataset.

[0102] Through the above steps, the generated high-quality image dataset not only has high definition and rich details, but also good contrast and purity, providing a high-quality data foundation for subsequent deep learning models, thereby improving the overall system performance.

[0103] This application takes into account that, in image processing and feature extraction, convolutional autoencoders transform the features of image datasets through nonlinear mapping capabilities to adapt to the image histogram adjustment requirements, ultimately generating feature-transformed image datasets. Through complex nonlinear transformations, the feature representation capability is improved while reducing redundant information, ensuring image quality and visual effects.

[0104] Optionally, based on the features of the image dataset, the nonlinear mapping capability of a convolutional autoencoder is used to transform the features of the image dataset to adapt to the image histogram adjustment requirements, generating a feature-transformed image dataset, including:

[0105] Based on the features of the image dataset, convolution operations are performed on the features of the image dataset using convolution kernels of different sizes to extract key features at different scales.

[0106] By fusing the key features at different scales, noise suppression is performed using a smoothing filter to remove high-frequency noise, thereby generating intermediate results of feature changes;

[0107] The intermediate results of the feature transformation are calculated using the following formula:

[0108]

[0109] Where T(x) is the intermediate result of the feature transformation; i is the index of the feature map, from 1 to N; N is the number of feature maps; W 1i x is the convolution kernel corresponding to the i-th feature map; i b1 is the i-th feature map; b1 is the bias term; tanh is the hyperbolic tangent activation function; α is the scaling factor; γ is the parameter controlling the exponential decay rate; j is the feature index participating in the exponential decay calculation, from 1 to M; M is the number of features participating in the exponential decay calculation; c is the center point value; x j For the j-th feature map;

[0110] Based on the intermediate results of the feature transformation, further nonlinear transformation is performed to enhance the feature representation capability. The most representative subset is obtained through the feature selection method to reduce redundant information and generate an enhanced feature representation.

[0111] The enhanced feature representation is calculated using the following formula:

[0112]

[0113] Where EFR(T(x)) is the enhanced feature representation; k is the index of the processed feature map, from 1 to K; K is the number of processed feature maps; W 2k T is the convolution kernel corresponding to the k-th feature map; k (x) represents the k-th processed feature map; b2 is the bias term; σ is the activation function; β is the scaling factor; ω is the parameter controlling the oscillation frequency of the sine wave; l is the feature index participating in the sine wave calculation, from 1 to L; L is the number of features participating in the sine wave calculation; T l (x) represents the l-th processed feature map; d represents the center point value of the sine wave; η represents the weighting coefficient of the logarithmic term; m represents the feature index for logarithmic calculation, from 1 to P; P represents the number of features involved in the logarithmic calculation; |T m (x)| represents the absolute value of the m-th feature map; δ is the scaling factor for the sigmoid term; θ is the parameter controlling the slope of the sigmoid function; n is the feature index participating in the sigmoid calculation, from 1 to Q; Q is the number of features participating in the sigmoid calculation; T n (x) represents the nth processed feature map; φ is the scaling factor for the square root term; o is the feature index participating in the square root calculation, from 1 to R; R is the number of features participating in the square root calculation; e is the center point value for the square root calculation; To ( x) is the o-th processed feature map;

[0114] Based on the enhanced feature representation, the image space is restored through inverse transformation, and histogram adjustment is performed to ensure image quality and visual effect, generating a feature-transformed image dataset.

[0115] This method aims to extract more representative and discriminative features from the original image data through a series of mathematical transformations. It can capture key information in the image, reduce the impact of noise and redundant data, and thus improve the accuracy of subsequent processing tasks (such as classification and recognition).

[0116] In the intermediate results of feature transformation, the convolution operation W 1i *x i +b1: Extracts key features at different scales using convolutional kernels of different sizes; exponential decay term. Control the weight distribution of features so that features near the center point have greater weight;

[0117] Where N is the number of feature maps, which can be preset through the model architecture; W 1i The convolution kernel corresponding to the i-th feature map is obtained through optimization using the backpropagation algorithm during training; x i b1 is the i-th feature map, obtained directly from the input image or the output of the previous network layer; b1 is the bias term, also optimized through the backpropagation algorithm during training; α is the scaling factor, which can be adjusted experimentally to achieve optimal performance; γ is a parameter controlling the exponential decay rate, which also needs to be determined experimentally; c is the center point value, representing the desired feature center position, usually set to 0.5 or adjusted according to the actual problem; x k The j-th feature map is obtained directly from the input image or the output of the previous layer of the network.

[0118] In enhanced feature representation, nonlinear transformation The feature representation capability is further enhanced through a new convolutional kernel; sinusoidal oscillation term Introducing periodic variations enhances the diversity of features; for several terms Increase the dynamic range of the feature through the logarithmic function; S-shaped term A smooth nonlinear transformation is introduced through the sigmoid function; square root term The variance of the features is controlled by the square root function;

[0119] Where K is the number of processed feature maps, determined by the model architecture; W 2k The convolution kernel corresponding to the k-th feature map is learned during training; T k(x) is the k-th processed feature map, which comes from the output of the previous step; b2 is the bias term, which is optimized during training; β, ω, d, η, P, m, δ, θ, Q, n, φ, R, o, e are various parameters, including scaling factor, frequency, center point, etc., which need to be tuned according to the specific task and experimental results.

[0120] Suppose a green food production line needs to perform quality control on food packaging and has collected a large number of images as an image dataset;

[0121] Assuming parameter N = 5, W 1i Assuming randomized 3×3 and 5×5 convolution kernels, x i The i-th feature map is obtained directly from the input image or the output of the previous network layer, with b1 = 0.2, α = 0.5, γ = 0.1, and c = 0.5; assuming x i = [0.1, 0.2, 0.3, 0.4, 0.5]; Assume K = 4; W 2k Let T be the convolution kernel corresponding to the k-th feature map, assumed to be a randomly initialized 3×3 convolution kernel; k (x) is the k-th processed feature map, derived from the output of the previous step; b2 = 0.3, β = 0.4, ω = 2, d = 0.7, η = 0.3, P = 2, δ = 0.2, θ = 1, Q = 3, φ = 0.1, R = 2, e = 0.6, T k (x) = [0.7, 0.8, -0.6, 0.9];

[0122]

[0123] Assuming a threshold of 0.8 is set, since most feature values ​​in the results are higher than the preset threshold of 0.8, it indicates that there are obvious quality problems with the currently inspected packaging, such as cracks, contamination, or labeling errors. High feature values ​​indicate that the features contained in the input image closely match the expected defect patterns, thus it can be inferred that there are significant anomalies in the packaging materials. Through the above steps, the production line can more accurately detect minor defects in packaging materials, such as cracks, contamination, or labeling errors, ensuring the consistency and stability of product quality.

[0124] Optionally, the step of predicting high-resolution image details based on the contrast-enhanced image dataset and combining it with super-resolution reconstruction technology to improve image resolution and sharpness and generate a high-resolution image includes: performing image upsampling processing based on the contrast-enhanced image dataset to generate a preliminary image framework; using interpolation technology to fill in missing pixels in the image based on the preliminary image framework and combining it with detail information from the contrast-enhanced image dataset to generate an optimized image framework; using a deep learning mechanism of super-resolution reconstruction technology to further optimize the optimized image framework, predicting and supplementing potential fine structures and texture features to generate a reconstructed image framework; and adjusting the overall sharpness and resolution based on the reconstructed image framework to ensure no distortion during image upsampling processing and generate a high-resolution image.

[0125] A contrast-enhanced image dataset refers to an image dataset that has undergone contrast enhancement processing to improve the visual effect of images.

[0126] Super-resolution reconstruction is an image processing technique that uses algorithms to convert low-resolution images into high-resolution images, improving image clarity and detail.

[0127] Image upsampling refers to techniques that enlarge low-resolution images to higher resolutions through interpolation or other methods.

[0128] Interpolation is a method of estimating unknown data points between known data points, and it is often used for pixel filling in the process of image magnification.

[0129] Deep learning mechanisms refer to techniques that utilize deep neural networks for feature learning and prediction.

[0130] In this embodiment, firstly, based on the contrast-enhanced image dataset, image upsampling is performed to generate a preliminary image framework; secondly, based on the preliminary image framework, combined with the detailed information of the contrast-enhanced image dataset, interpolation techniques are used to fill in the missing pixels in the image to generate an optimized image framework; thirdly, based on the optimized image framework, a deep learning mechanism of super-resolution reconstruction technology is used to further optimize the optimized image framework, predict and supplement potential fine structures and texture features, and generate a reconstructed image framework; finally, based on the reconstructed image framework, the overall sharpness and resolution are adjusted to ensure that there is no distortion during image upsampling, generating the final high-resolution image.

[0131] Suppose a food processing company needs to control the quality of meat products on its production line to ensure product safety and consistency;

[0132] First, the contrast-enhanced image dataset is magnified using image upsampling methods (such as bilinear interpolation) to generate a preliminary image framework. Second, based on the preliminary image framework and combined with the detailed information in the contrast-enhanced image dataset, more advanced interpolation techniques (such as bicubic interpolation) are used to fill in the missing pixels in the image, generating an optimized image framework. Third, based on the optimized image framework, a pre-trained super-resolution reconstruction model is used to further optimize the optimized image framework through a deep learning mechanism, predicting and supplementing potential fine structures and texture features to generate a reconstructed image framework. Finally, based on the reconstructed image framework, the overall sharpness and resolution are adjusted to ensure that the image is not distorted during magnification, generating the final high-resolution image.

[0133] Through the above steps, the generated high-resolution image not only has higher clarity and richer details, but also maintains the quality of the original image, providing a high-quality data foundation for subsequent anomaly detection and quality control.

[0134] 103. Based on the high-quality image dataset, use a deep residual network for anomaly detection processing to identify abnormal regions in the high-quality image dataset, and combine time-series data analysis technology for dynamic analysis to generate anomaly detection results.

[0135] Deep residual networks are a type of deep learning model that addresses the vanishing gradient problem by introducing residual blocks, thereby enhancing the model's learning ability.

[0136] Anomaly detection refers to identifying abnormal areas or defects in an image.

[0137] Time series data analysis technology refers to the analysis of time series data to capture the patterns and trends of data changes over time.

[0138] Anomaly detection results refer to the abnormal areas and related information identified after the above processing.

[0139] In this step, firstly, the vanishing gradient problem in deep learning is solved by introducing residual blocks to enhance the model's learning ability and generate a preprocessed image dataset; secondly, a deep residual network is used to perform forward propagation calculations on the preprocessed image dataset to generate preliminary anomaly detection results; thirdly, based on the preliminary anomaly detection results, the time series patterns of anomalous events are identified, dynamic analysis is performed, and dynamic analysis results are generated; finally, a comprehensive evaluation is performed by combining the multi-dimensional key features of anomalous events to generate the final anomaly detection results.

[0140] Optionally, step 103, which involves using a deep residual network to perform anomaly detection processing based on the high-quality image dataset, identifying anomalous regions in the high-quality image dataset, and performing dynamic analysis using time-series data analysis techniques to generate anomaly detection results, includes: generating a preprocessed image dataset by introducing residual blocks to solve the vanishing gradient problem in deep learning and enhancing the model's learning ability; performing anomaly detection processing using a deep residual network based on the preprocessed image dataset, calculating anomaly probability values ​​through forward propagation, and generating preliminary anomaly detection results; identifying time-series patterns of anomalous events based on the preliminary anomaly detection results, performing dynamic analysis, and generating dynamic analysis results; and performing a comprehensive evaluation based on the dynamic analysis results, combining multi-dimensional key features of the anomalous events, to generate the final anomaly detection result.

[0141] The step of performing anomaly detection processing using a deep residual network based on the preprocessed image dataset, and calculating anomaly probability values ​​through forward propagation to generate preliminary anomaly detection results, includes: extracting key features using the histogram of oriented gradients (HARQ) method based on the preprocessed image dataset to generate a feature representation dataset; calculating feature representations through forward propagation using a deep residual network based on the feature representation dataset to generate a deep feature representation dataset; obtaining anomaly probability values ​​by combining the deep residual network classification layer with the deep feature representation dataset to generate an anomaly probability value dataset; and performing threshold judgment based on a preset reasonable threshold range using the anomaly probability value dataset, marking and recording the anomaly probability values ​​to generate preliminary anomaly detection results.

[0142] A high-quality image dataset refers to a collection of images that have undergone a series of processing steps (such as noise reduction and contrast enhancement) and have clear details and high resolution.

[0143] Deep residual networks are a type of neural network architecture that uses residual blocks to alleviate the gradient vanishing problem that occurs during the training of deep networks, thereby improving the model's learning ability.

[0144] Anomaly detection is the process of identifying data points or regions that are significantly different from normal patterns, based on machine learning or deep learning methods.

[0145] Time series data analysis is a technique for analyzing data that changes over time to discover trends, periods, or other patterns.

[0146] Histogram of Oriented Gradients (HGP) is a feature descriptor commonly used for object detection. It describes the shape of an object by calculating the gradient direction statistics within a local region of an image.

[0147] In this embodiment, firstly, a high-quality image dataset is preprocessed to solve the gradient vanishing problem using residual blocks, generating a preprocessed image that is easier to process in subsequent steps. Secondly, key features are extracted from the preprocessed image, and the Histogram of Oriented Gradients (HGP) method is used to further convert them into deep feature representations through a deep residual network. Thirdly, based on these deep features, the probability value of each sample becoming an anomaly is calculated, and a preliminary anomaly detection result is determined according to a set threshold. Finally, by combining time series analysis techniques, the time series patterns of abnormal events are dynamically analyzed, the anomaly situation is comprehensively evaluated, and a final anomaly detection report is generated.

[0148] A frozen food company wanted to improve product quality by monitoring the integrity of product packaging on its production line, and the company collected a large number of product packaging images as a high-quality image dataset.

[0149] First, the dataset is preprocessed using the residual block structure in a deep residual network, enhancing the model's ability to learn complex patterns and generating a preprocessed image dataset. Second, key visual features are extracted from the preprocessed images using the histogram of oriented gradients (HARQ) method, constructing a feature representation dataset. Third, this feature representation is input into the deep residual network, and the depth feature representation of each image and its corresponding anomaly probability value are obtained through the forward propagation process. When the probability exceeds a preset threshold, it is marked as a potential anomaly. Finally, by combining time-series data analysis techniques, the temporal distribution characteristics of the marked anomalies are analyzed. Considering factors such as the frequency, duration, and interval of anomaly events, a comprehensive evaluation is conducted to generate a detailed anomaly detection report.

[0150] Through the above steps, the company is able to promptly identify and repair potential product packaging defects on the production line, ensuring product quality standards.

[0151] This application takes into account that in the feature representation computation based on deep residual networks, the feature representation is computed through forward propagation to generate a deep feature representation dataset. This process includes preprocessing the input feature map to reduce noise and highlight key information, capturing complex patterns through complex nonlinear transformations, and generating the final deep feature representation dataset through standardization and dimensionality reduction techniques.

[0152] Optionally, based on the feature representation dataset, a deep residual network is used to compute the feature representations through forward propagation to generate a deep feature representation dataset, including:

[0153] Based on the aforementioned feature representation dataset, the input feature map is preprocessed, and filtering techniques are used to reduce input noise and highlight key information in the input feature map in order to generate an intermediate output vector.

[0154] The intermediate output vector is calculated using the following formula:

[0155]

[0156] Among them, A i The intermediate output vector; N is the number of input feature maps; W ij X is the convolution kernel between the i-th output vector and the j-th input feature map; j b is the j-th feature map in the input feature vector; i V is the bias term of the i-th output vector; σ is the ReLU activation function; α and β are parameters controlling the strength of the nonlinear term; M is the number of input feature maps participating in the nonlinear transformation; V ik X represents the weights used for the nonlinear transformation between the i-th output vector and the k-th input feature map; j is the index of the input feature map, from 1 to N; k is the index of the input feature map, from 1 to M; j This represents the k-th input feature map that participates in the nonlinear transformation;

[0157] Based on the intermediate output vector, multiple levels of nonlinear transformations are used to capture complex patterns. Additional convolutional and pooling layers are introduced to increase expressive power and generalization performance in order to generate the final output vector.

[0158] The final output vector is calculated using the following formula:

[0159]

[0160] Among them, D i The final output vector; N is the number of input feature maps; W' il A is the convolution kernel between the i-th output vector and the l-th intermediate vector; l A is the l-th intermediate vector; m Let A be the m-th intermediate vector; n A is the nth intermediate vector; o b' is the o-th intermediate vector; i σ is the bias term of the i-th output vector; σ is the ReLU activation function; γ, δ, and θ are parameters controlling the strength of the nonlinear term; O is the number of intermediate vectors participating in the nonlinear transformation; U im Z represents the weights used for nonlinear transformation between the i-th output vector and the m-th intermediate vector; ∈ is a parameter controlling the strength of the fractional term; P and Q are the number of intermediate vectors involved in the operation in the numerator and denominator, respectively; in and Y io , respectively, are the weights between the i-th output vector in the numerator and the n-th or o-th intermediate vector; l is the index of the input feature map, from 1 to N; m is the index of the intermediate vector, from 1 to O; n is the index of the intermediate vector, from 1 to P; o is the index of the intermediate vector, from 1 to Q;

[0161] Based on the final output vector, standardization is performed, and dimensionality reduction is carried out through principal component analysis to map the high-dimensional feature vector into a low-dimensional space for visualization analysis, thereby generating a deep feature representation dataset.

[0162] This method aims to generate intermediate output vectors through complex nonlinear transformations (such as logarithmic and sine terms), introduce additional convolutional and pooling layers for multi-level nonlinear transformations to enhance the model's expressive power and generalization performance, and map high-dimensional features to a low-dimensional space through standardization and principal component analysis for easier visualization and further analysis.

[0163] In the intermediate output vector, the convolution operation W ij ×X j : Extract key information from the input feature map using convolutional kernels; bias term b i Introducing an offset for each output vector increases the model's flexibility; nonlinear terms Enhance feature representation by nonlinear transformation and control feature weight distribution;

[0164] Where N is the number of input feature maps, determined by the model architecture; W ij X is the convolution kernel between the i-th output vector and the j-th input feature map, optimized through the backpropagation algorithm during training; j The j-th feature map in the input feature vector is obtained directly from the input image or the output of the previous network layer; b i The bias term for the i-th output vector is also optimized through backpropagation during training; α and β are parameters controlling the strength of the nonlinear term, usually determined through experimental tuning; M is the number of input feature maps participating in the nonlinear transformation, set according to the actual task requirements; V ik The weights used for nonlinear transformation between the i-th output vector and the k-th input feature map are optimized through the backpropagation algorithm during the training process.

[0165] In the final output vector, the additional convolutional layer and pooling layer W' il ×A l +b' i : By employing multi-level nonlinear transformations, more complex patterns are captured, increasing the model's expressive power and generalization performance; the nonlinear term γ·sin(δ·∑U im ·A m +θ): Introduces periodic variations to enhance feature diversity; fractional term The dynamic range of features can be controlled by adjusting the importance of the features through fractional terms.

[0166] Where N is the number of input feature maps, determined by the model architecture; W' il The convolution kernel between the i-th output vector and the l-th intermediate vector is obtained through backpropagation during training; A l b' is the l-th intermediate vector, derived from the output of the previous step. i The bias term for the i-th output vector is also optimized through the backpropagation algorithm during training; γ, δ, and θ are parameters controlling the strength of the nonlinear term, usually determined through experimental tuning; O is the number of intermediate vectors participating in the nonlinear transformation, set according to the actual task requirements; U im Z represents the weights used for nonlinear transformation between the i-th output vector and the m-th intermediate vector, optimized through backpropagation during training; ∈ is a parameter controlling the strength of the fractional term, usually determined through experimental tuning; P and Q are the number of intermediate vectors involved in the calculation in the numerator and denominator, set according to the actual task requirements; in ,Y io The weights between the i-th output vector in the numerator and denominator and the nth or o-th intermediate vector are obtained through the backpropagation algorithm during the training process.

[0167] Suppose a fresh food processing company needs to monitor the quality of its products on its production line in real time, especially the quality control of the packaging process;

[0168] Assume parameter N = 3; W ij A 3×3 convolution kernel is randomly initialized; X j =[0.1,0.2,0.3]; b i =0.2; α=0.5; β=0.2; M=2; V ik Weights initialized randomly;

[0169]

[0170] Assume parameter N = 3; W' il A is a randomly initialized 3×3 convolution kernel; l = [1.02, 1.02, 1.02]; b' i =0.3; γ = 0.4; δ = 0.3; θ = 0.1; O = 2; U im The weights are randomly initialized; ∈ = 0.3; P = 2; Q = 2; Z in ,Y io Weights initialized randomly;

[0171] Because of D iThe result of 4.4229 is significantly higher than the preset threshold of 4.0, which indicates that the features in the input image are highly matched with known defect patterns, meaning that there are obvious quality problems in the packaging materials, such as damage, contamination, or inaccurate label information. Through the above steps, the production line can achieve high-precision detection of minor defects in packaging materials, specifically covering typical defect types such as cracks, contamination, and label errors, ensuring the uniformity and long-term stability of product quality.

[0172] 104. Based on the anomaly detection results, the data is sent to the food production line control system in real time via an encrypted interface to support timely adjustment of production parameters and generation of food quality control strategies.

[0173] An encrypted interface refers to a secure data transmission protocol that ensures data is not tampered with or leaked during transmission.

[0174] A food production line control system is a system used to monitor and control a food production line, capable of adjusting production parameters based on received data.

[0175] Food quality control strategies refer to a series of measures developed based on abnormal detection results to ensure stable product quality and optimize production processes.

[0176] In this step, firstly, the anomaly detection results are securely transmitted through an encrypted interface to ensure data security during transmission; secondly, the food production line control system receives and parses these anomaly detection results; thirdly, the control system automatically generates corresponding production parameter adjustment instructions based on the anomaly detection results; finally, the control system executes these instructions, adjusts production parameters in a timely manner, and generates food quality control strategies to ensure product quality stability and optimize the production process.

[0177] Optionally, step 104, which involves sending the anomaly detection results to the food production line control system in real time via an encrypted interface to support timely adjustment of production parameters and generation of a food quality control strategy, includes: transmitting the anomaly detection report to the food production line control system in real time via an encrypted interface based on the anomaly detection results, generating an encrypted transmission result; performing a comprehensive analysis based on the encrypted transmission result, combining historical data of the production line with the current production status, and generating production line adjustment suggestions; automatically adjusting production parameters through the food production line control system based on the production line adjustment suggestions, generating adjusted production parameter settings; and recording actual operating performance indicators based on the adjusted production parameter settings, continuously adjusting and optimizing them through a feedback mechanism in conjunction with the anomaly detection results, and generating a food quality control strategy.

[0178] Encrypted interfaces are a secure data transmission method that ensures data is not tampered with or leaked during transmission.

[0179] A food production line control system is a system used to monitor and control a food production line, and can adjust production parameters based on the received data.

[0180] Production line historical data includes information such as past production records, equipment status, and quality inspection results.

[0181] Current production status refers to the current operating status of the production line, including equipment operating status, production speed, product quality, etc.

[0182] Comprehensive analysis combines multiple data sources for multi-dimensional analysis to generate more accurate recommendations.

[0183] The feedback mechanism is a continuous optimization process that monitors the actual results and constantly adjusts the strategy to achieve optimal performance.

[0184] In this embodiment, firstly, an anomaly detection report is transmitted to the food production line control system in real time using an encrypted interface to ensure data security; secondly, based on the encrypted transmission results, combined with historical data and current production status of the production line, a comprehensive analysis is performed to generate specific production line adjustment suggestions; thirdly, the food production line control system automatically adjusts production parameters to respond to these adjustment suggestions; finally, actual operating performance indicators are recorded, and combined with the anomaly detection results, continuous adjustments and optimizations are made through a feedback mechanism to generate the final food quality control strategy.

[0185] Suppose a food processing plant needs to perform quality control on the biscuit packaging process on its production line. The plant has already generated anomaly detection results using image analysis technology.

[0186] First, an encrypted interface is used to transmit anomaly detection reports to the food production line control system in real time, ensuring data security during transmission and generating encrypted transmission results. Second, based on the encrypted transmission results, combined with historical data from the production line (such as past packaging defect records and equipment maintenance records) and current production status (such as equipment operating status and production speed), a comprehensive analysis is performed to generate specific production line adjustment suggestions, such as adjusting the packaging machine speed or increasing the frequency of quality inspections. Third, the food production line control system automatically adjusts production parameters, such as slowing down the packaging machine speed or increasing sensor sensitivity, to respond to these adjustment suggestions. Finally, actual operating performance indicators, such as packaging defect rate and production efficiency, are recorded, and combined with anomaly detection results, production parameters are continuously adjusted and optimized through a feedback mechanism to generate the final food quality control strategy, ensuring the stability of the production line and the consistency of product quality.

[0187] Through the above steps, the factory is able to promptly identify and resolve problems in the packaging process, thereby improving product quality and production efficiency.

[0188] Figure 2 This application provides a schematic diagram of the structure of a food quality control system based on intelligent image analysis, as shown in the embodiment. Figure 2 As shown, the device includes:

[0189] The collection module 21 is used to collect real-time image streams from different locations on the food production line from multiple angles and in multiple spectra through a multi-sensor system, forming a comprehensive image dataset;

[0190] Processing module 22 is used to perform image denoising and contrast enhancement processing based on the comprehensive image dataset using a convolutional autoencoder, and combined with super-resolution reconstruction technology to improve image clarity and detail, thereby generating a high-quality image dataset.

[0191] Analysis module 23 is used to perform anomaly detection processing based on the high-quality image dataset using a deep residual network, identify abnormal regions in the high-quality image dataset, and perform dynamic analysis by combining time series data analysis technology to generate anomaly detection results.

[0192] The sending module 24 is used to send the anomaly detection results to the food production line control system in real time through an encrypted interface, so as to support timely adjustment of production parameters and generation of food quality control strategies.

[0193] Figure 2 The aforementioned food quality control system based on intelligent image analysis can execute... Figure 1 The implementation principle and technical effects of the food quality control method based on intelligent image analysis described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit of the food quality control system based on intelligent image analysis in the above embodiments are described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0194] In one possible design, Figure 2 The food quality control system based on intelligent image analysis shown in the embodiment can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0195] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0196] The processing component 32 is used to: collect real-time image streams from different locations on the food production line using a multi-sensor system, forming a comprehensive image dataset; based on the comprehensive image dataset, perform image denoising and contrast enhancement processing using a convolutional autoencoder, and combine it with super-resolution reconstruction technology to improve image clarity and detail, generating a high-quality image dataset; based on the high-quality image dataset, perform anomaly detection processing using a deep residual network to identify abnormal regions in the high-quality image dataset, and combine it with time-series data analysis technology for dynamic analysis to generate anomaly detection results; based on the anomaly detection results, send them to the food production line control system in real time through an encrypted interface to support timely adjustment of production parameters and generate food quality control strategies.

[0197] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0198] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0199] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0200] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0201] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0202] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0203] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The illustrated embodiment is a food quality control method based on intelligent image analysis.

[0204] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0205] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0206] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0207] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for food quality control based on intelligent image analysis, characterized in that, The application relates to a food production line quality control method and system. Collecting multi-angle and multi-spectrum real-time image streams at different positions of a food production line through a multi-sensor system to form a comprehensive image dataset; Based on the comprehensive image dataset, using a convolutional autoencoder to perform image denoising and contrast enhancement processing, wherein the contrast enhancement processing includes using the nonlinear mapping capability of the convolutional autoencoder; The use of the nonlinear mapping capability of the convolutional autoencoder includes: Using convolution kernels of different sizes to perform convolution operations to extract key features of different scales; Fusing the different scale key features, suppressing noise through a smoothing filter to remove high-frequency noise, and generating a feature transformation intermediate result; Based on the feature transformation intermediate result, performing further nonlinear transformation to enhance feature expression capability, obtaining the strongest representative subset through a feature selection method to reduce redundant information, and generating an enhanced feature representation; ； in, This is an intermediate result of feature transformation; The index of the feature map, from 1 to ; The number of feature maps; The kernel is the convolution kernel corresponding to the i-th feature map; For the first Each feature map; Here, tanh is the bias term; tanh is the hyperbolic tangent activation function. This is the scaling factor; Parameters used to control the rate of exponential decay; For the feature index participating in the exponential decay calculation, from 1 to ; The number of features involved in the exponential decay calculation; The value at the center point; For the first Each feature map; Based on the enhanced feature representation, restoring to the image space through inverse transformation, adjusting the histogram to ensure image quality and visual effect, and generating a feature transformation image dataset; Based on the feature transformation image dataset, generating a contrast enhancement image dataset, and combining a super-resolution reconstruction technology to improve image clarity and detail performance, and generating a high-quality image dataset; ； wherein, is an enhanced feature representation; is a processed feature map index from 1 to ; is a number of processed feature maps; is a convolution kernel corresponding to the th feature map; is the th processed feature map; is a bias term; is an activation function; is a scaling factor; is a parameter to control the frequency of sine wave oscillation; is a feature index participating in the sine wave calculation from 1 to ; is a number of features participating in the sine wave calculation; is the th processed feature map; is a center point value of the sine wave; is a weight coefficient of the logarithmic term; is a feature index participating in the logarithmic calculation from 1 to ; is a number of features participating in the logarithmic calculation; is an absolute value of the th feature map; is a scaling factor of the S-shaped term; is a parameter to control the slope of the S-shaped function; is a feature index participating in the S-shaped calculation from 1 to ; is a number of features participating in the S-shaped calculation; is the th processed feature map; is a scaling factor of the square root term; is a feature index participating in the square root calculation from 1 to ; is a number of features participating in the square root calculation; is a center point value of the square root calculation; is the th processed feature map; Based on the high-quality image dataset, using a deep residual network to perform anomaly detection processing, identifying abnormal areas in the high-quality image dataset, combining time series data analysis technology for dynamic analysis, and generating an anomaly detection result; Based on the anomaly detection result, sending the result to a food production line control system in real time through an encryption interface to support timely adjustment of production parameters and generate a food quality control strategy. The use of the convolutional autoencoder to perform image denoising and contrast enhancement processing based on the comprehensive image dataset, combining a super-resolution reconstruction technology to improve image clarity and detail performance, and generating a high-quality image dataset includes: Based on the comprehensive image dataset, performing random noise removal processing of the convolutional autoencoder to improve image purity and generate a denoised image dataset; 2. The method of claim 1, wherein, Based on the denoised image dataset, using the nonlinear mapping capability of the convolutional autoencoder to analyze and adjust the image histogram, performing contrast enhancement processing, and generating a contrast enhancement image dataset; Based on the contrast enhancement image dataset, combining a super-resolution reconstruction technology to perform high-resolution image detail prediction, improving image resolution and clarity, and generating a high-resolution image; Based on the high-resolution image, adjusting local contrast, performing edge sharpening processing, and enhancing microscopic levels to improve image detail performance, and generating a high-quality image dataset. The use of the nonlinear mapping capability of the convolutional autoencoder based on the denoised image dataset to analyze and adjust the image histogram, perform contrast enhancement processing, and generate a contrast enhancement image dataset includes: Based on the denoised image dataset, performing multi-scale key feature extraction processing through a pre-trained convolutional autoencoder to generate image dataset features; 3. The method of claim 2, wherein, ​ ​ Based on the image dataset features, the convolutional autoencoder nonlinear mapping capability is used to transform the image dataset features to adapt to the image histogram adjustment requirements, and a feature transformed image dataset is generated; Based on the feature transformed image dataset, the specific distribution of the image histogram is analyzed, the contrast enhancement optimization space is identified, and the image histogram analysis result is generated; Based on the image histogram analysis result, the brightness of the low light area is enhanced, and the contrast of the high contrast area is adjusted to improve the overall visual effect, and a contrast enhanced image dataset is generated.

4. The method of claim 2, wherein, Based on the contrast enhanced image dataset, combined with the super resolution reconstruction technology, the high resolution image details are predicted, the image resolution and clarity are improved, and a high resolution image is generated, including: Based on the contrast enhanced image dataset, an image magnification process is performed through an image upsampling method to generate a preliminary image framework; Based on the preliminary image framework, combined with the detail information of the contrast enhanced image dataset, an interpolation technique is used to fill in the missing pixel points of the image to generate an optimized image framework; Based on the optimized image framework, the super resolution reconstruction technology deep learning mechanism is used to optimize the processing of the optimized image framework, to predict and supplement potential subtle structures and texture features, and to generate a reconstructed image framework; Based on the reconstructed image framework, the overall clarity and resolution are adjusted to ensure that there is no distortion in the image magnification process, and a high resolution image is generated.

5. The method of claim 1, wherein, Based on the high-quality image dataset, the deep residual network is used for anomaly detection processing to identify the abnormal area of the high-quality image dataset, and the time series data analysis technology is used for dynamic analysis to generate an anomaly detection result, including: Based on the high-quality image dataset, the deep learning gradient vanishing problem is solved by introducing a residual block to enhance the model learning ability and generate a preprocessed image dataset; Based on the preprocessed image dataset, the deep residual network is used for anomaly detection processing, and the abnormal probability value is calculated through forward propagation to generate a preliminary anomaly detection result; Based on the preliminary anomaly detection result, the time series pattern of the abnormal event is identified and dynamically analyzed to generate a dynamic analysis result; Based on the dynamic analysis result, the abnormal event multidimensional key features are combined for comprehensive evaluation to generate an anomaly detection result.

6. The method of claim 5, wherein, Based on the preprocessed image dataset, the deep residual network is used for anomaly detection processing, and the abnormal probability value is calculated through forward propagation to generate a preliminary anomaly detection result, including: Based on the preprocessed image dataset, the histogram of oriented gradients method is used for key feature extraction processing to generate a feature representation dataset; Based on the feature representation dataset, the deep residual network is used to calculate the feature representation through forward propagation traversal to generate a deep feature representation dataset; Based on the deep feature representation dataset, combined with the deep residual network classification layer, the abnormal probability value is obtained to generate an abnormal probability value dataset; Based on the abnormal probability value dataset, a reasonable threshold interval is preset for threshold judgment, and the abnormal probability value exceeding the threshold is marked and recorded to generate a preliminary anomaly detection result.

7. The method of claim 6, wherein, Based on the feature representation dataset, a deep residual network is used to calculate the feature representation through forward propagation to generate a deep feature representation dataset, including: Based on the feature representation dataset, the input feature map is preprocessed to reduce input noise and highlight key information of the input feature map through filtering technology to generate an intermediate output vector; The intermediate output vector is calculated by the following formula: ； wherein, is an intermediate output vector; is the number of input feature maps; is a convolution kernel between the th output vector and the th input feature map; is the th feature map in the input feature vector; is a bias term for the th output vector; is a ReLU activation function; and is a parameter to control the strength of the non-linear term; is the number of input feature maps participating in the non-linear transformation; is a weight for the non-linear transformation between the th output vector and the th input feature map; is the index of the input feature map, from 1 to is the index of the input feature map, from 1 to ; denotes the th input feature map participating in the non-linear transformation; Based on the intermediate output vector, multiple levels of nonlinear transformation are used to capture complex patterns, additional convolution layers and pooling layers are introduced to increase expression ability and generalization performance to generate a final output vector; The final output vector is calculated by the following formula: ； wherein, is the final output vector; is the number of input feature maps; is the convolution kernel between the th output vector and the th intermediate vector; is the th intermediate vector; is the th intermediate vector; is the th intermediate vector; is the th intermediate vector; is the th output vector; is the ReLU activation function; and are parameters to control the strength of the non-linear term; is the number of intermediate vectors involved in the non-linear transformation; is the weight between the th output vector and the th intermediate vector for the non-linear transformation; is the parameter to control the strength of the fractional term; and are the number of intermediate vectors involved in the numerator and denominator, respectively; and are the weight between the th output vector and the th or th intermediate vector in the numerator and denominator, respectively; is the index of input feature maps, from 1 to is the index of intermediate vectors, from 1 to ; is the index of intermediate vectors, from 1 to is the index of intermediate vectors, from 1 to ; Based on the final output vector, standardization processing is performed, dimensionality reduction is performed through principal component analysis, high-dimensional feature vectors are mapped to low-dimensional space, and visual analysis is performed to generate a deep feature representation dataset.

8. The method of claim 1, wherein, Based on the anomaly detection result, the encrypted interface is used to send the result to the food production line control system in real time to support timely adjustment of production parameters and generate food quality control strategies, including: Based on the anomaly detection result, an anomaly detection report is transmitted to the food production line control system in real time using an encrypted interface to generate an encrypted transmission result; Based on the encrypted transmission result, comprehensive analysis is performed by combining historical data and current production status to generate production line adjustment suggestions; Based on the production line adjustment suggestions, the food production line control system automatically adjusts the production parameters to generate adjusted production parameter settings; Based on the adjusted production parameter settings, actual performance indicators are recorded, and the feedback mechanism is continuously adjusted and optimized based on the anomaly detection result to generate food quality control strategies.

9. A food quality control system based on intelligent image analysis, for performing a food quality control method based on intelligent image analysis according to any one of claims 1 to 8, characterized in that, Including: The collection module is used to collect multi-angle multi-spectral real-time image streams at different positions of the food production line through a multi-sensor system to form a comprehensive image dataset; The processing module is used to generate a high-quality image dataset based on the comprehensive image dataset by using a convolutional autoencoder for image denoising and contrast enhancement processing, and combining super-resolution reconstruction technology to improve image clarity and detail performance; The analysis module is used to perform anomaly detection processing on the high-quality image dataset using a deep residual network to identify abnormal areas in the high-quality image dataset and generate an anomaly detection result through dynamic analysis combined with time series data analysis technology; The sending module is used to send the anomaly detection result to the food production line control system in real time through an encrypted interface to support timely adjustment of production parameters and generate food quality control strategies.

Images