Intelligent image processing and analyzing system for AI machine vision enhancement

Abstract

The invention provides an intelligent image processing and analysis system for AI machine vision enhancement. The intelligent image processing and analysis system comprises an image acquisition module, a preprocessing module, a merging module, a feature extraction module, a raw and cooked degree identification module and a result output module. The acquisition module acquires food image information; the preprocessing module performs de-noising and contrast enhancement; the merging module merges the images; the feature extraction module extracts features; the uncooked-cooked degree identification module identifies the uncooked degree of the food; and the result output module outputs an identification result. According to the invention, the accuracy and efficiency of raw and cooked food identification in the food processing process can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent image processing and analysis, and more specifically, to an AI-enhanced intelligent image processing and analysis system. Background Technology

[0002] In the field of food processing and quality control, accurately and quickly identifying the raw or cooked state of food is crucial for ensuring food safety and improving production efficiency. Traditional methods for identifying the raw or cooked state of food mainly rely on manual visual inspection. This method is not only inefficient but also easily influenced by subjective judgment, making it difficult to guarantee the accuracy and consistency of the identification results. With the development of technology, machine vision technology has been introduced into the food processing field to identify the raw or cooked state of food through image acquisition and analysis. However, existing technologies still have certain limitations in image acquisition, preprocessing, feature extraction, and raw / cooked state identification. For example, image acquisition is limited in its level, preprocessing algorithms have limited denoising and contrast enhancement effects, feature extraction accuracy is insufficient, and raw / cooked state identification algorithms have limited ability to penetrate different levels of detail. These problems limit the application effectiveness and scope of machine vision technology in food raw / cooked state identification.

[0003] In implementing the embodiments of the present invention, the inventors discovered that the prior art has at least the following problems or defects: when processing food images, the existing system often fails to effectively merge image information at different levels, resulting in insufficient feature extraction; in the process of raw and cooked food identification, the algorithm often lacks in-depth analysis and weight allocation of features in different dimensions, resulting in insufficient accuracy and robustness of the identification results; in addition, when faced with multiple sample categories with high similarity, the existing system lacks an effective decision-making mechanism to determine the final raw and cooked food category, affecting the reliability of the identification results. Summary of the Invention

[0004] This invention provides an AI-enhanced intelligent image processing and analysis system, comprising: Image acquisition module, image preprocessing module, multi-level image merging module, feature extraction module, raw / cooked recognition module, and result output module; The image acquisition module is used to acquire image information at different levels during food processing and transmit it to the image preprocessing module. The image preprocessing module performs noise reduction and contrast enhancement on the received image before transmitting it to the multi-level image merging module. The multi-level image merging module merges images from different levels and transmits the merged image to the feature extraction module. The feature extraction module extracts image feature information and transmits it to the raw / cooked food recognition module; the raw / cooked food recognition module performs layer-by-layer recognition of food raw / cooked status based on a preset algorithm and transmits the recognition result to the result output module. The result output module outputs the results of food raw / cooked identification.

[0005] Furthermore, the image acquisition module includes an optical imaging unit and an image transmission unit. The optical imaging unit is used to image different layers of the food, and the image transmission unit is used to transmit the image data after imaging to the image preprocessing module.

[0006] Furthermore, the image preprocessing module includes a denoising unit and a contrast enhancement unit. The denoising unit uses a median filtering algorithm to remove image noise, and the contrast enhancement unit uses a grayscale stretching algorithm to enhance image contrast.

[0007] Furthermore, the multi-level image merging module includes an image registration unit and an image fusion unit. The image registration unit registers images at different levels based on a feature point matching algorithm, and the image fusion unit uses a weighted average method to fuse the registered images at different levels to obtain a merged image. The formula for the weighted average method is: Let the images at different levels be... The corresponding weight is The merged image ,in The number of image layers. Indicates the first The weights of the layer images, and .

[0008] Furthermore, the feature extraction module includes a texture feature extraction unit, a color feature extraction unit, and a shape feature extraction unit. The texture feature extraction unit extracts the texture features of the food image, the color feature extraction unit extracts the color features of the food image, and the shape feature extraction unit extracts the shape features of the food image. These features are then integrated and transmitted to the raw / cooked food recognition module.

[0009] Furthermore, the preset algorithm of the raw / cooked food identification module includes the following steps: Step 1: Establish a raw and cooked food sample library, collect image data of foods at different stages of rawness or cookedness, and label them as raw or cooked categories, denoted as the raw and cooked category set. ,in Indicates the first Raw and cooked varieties, The total number of categories; simultaneously, multi-dimensional features of the image are extracted to form a feature vector set. ,in Indicates the first 1 eigenvector The dimension of the feature vector; Step 2: For the food image to be identified, extract its feature vector. ; Step 3: Calculate the feature vector of the image to be recognized. With each feature vector in the sample library similarity The calculation formula is:

[0010] in, This represents the vector dot product operation. Represents the magnitude of a vector; Step 4: Determine the raw / cooked category of the food to be identified based on similarity. Select the category of the sample with the highest similarity as the raw / cooked category of the food to be identified.

[0011] in, This indicates the raw / cooked category of the identification result.

[0012] Furthermore, in step one, the extracted multi-dimensional features include the image's gray-level co-occurrence matrix features, color histogram features, and edge contour features.

[0013] Furthermore, in step three, when calculating similarity, different weights can be assigned to features of different dimensions. The revised similarity calculation formula is as follows:

[0014] in, and The first and second feature vectors of the image to be identified and the second feature vector of the sample, respectively, represent the first and second feature vectors of the image to be identified. Features in 100 dimensions.

[0015] Furthermore, in step four, if multiple sample categories with high and close similarity exist, a voting mechanism is used to determine the final category (raw or matured), i.e., the categories ranked by similarity to the feature vector of the image to be identified are determined. The number of raw / cooked categories in each sample category is used as the final result, with the category having the highest number of raw / cooked categories. This is a preset threshold.

[0016] Furthermore, the result output module includes a display unit and a data storage unit. The display unit is used to intuitively display the food raw / cooked identification results, and the data storage unit is used to store the identification results and related image data.

[0017] The embodiments of the present invention have at least the following beneficial effects: This AI machine vision-enhanced intelligent image processing and analysis system, by integrating multiple modules such as image acquisition, preprocessing, multi-level image merging, feature extraction, raw / cooked food recognition, and result output, can improve the accuracy and efficiency of food raw / cooked food recognition. The system employs an optical imaging unit and an image transmission unit, enabling it to acquire image information at different levels during food processing. Preprocessing the images through a denoising unit and a contrast enhancement unit improves image quality. The multi-level image merging module utilizes an image registration unit and an image fusion unit, based on a feature point matching algorithm and a weighted average method, to register and fuse images at different levels, enhancing the image's sense of depth and detail, and providing richer and more accurate image data for subsequent feature extraction and raw / cooked food recognition.

[0018] Furthermore, the system's feature extraction module can comprehensively extract texture, color, and shape features from food images and integrate these features before transmitting them to the raw / cooked food recognition module. Based on a preset algorithm, the raw / cooked food recognition module establishes a food raw / cooked food sample library and feature vector set to perform layer-by-layer recognition of food raw / cookedness and outputs the recognition results. This process reduces interference from human factors, improves the objectivity and consistency of recognition, and can adapt to the recognition needs of different foods and different degrees of rawness / cooking, enhancing the system's applicability and flexibility. The results output module includes a display unit and a data storage unit, which can not only intuitively display the food raw / cooked food recognition results but also store the recognition results and related image data, facilitating subsequent data analysis and quality control. Attached Figure Description

[0019] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of the invention are illustrated in the drawings by way of example and not limitation, wherein: Figure 1 This is a schematic diagram of the structure of an AI machine vision-enhanced intelligent image processing and analysis system provided in an embodiment of the present invention. Detailed Implementation

[0020] The principles and spirit of the invention will now be described with reference to several exemplary embodiments. It should be understood that these embodiments are provided merely to enable those skilled in the art to better understand and implement the invention, and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided to make the invention more thorough and complete, and to fully convey the scope of the invention to those skilled in the art.

[0021] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0022] It should be noted that the number of any elements in the accompanying drawings is for illustrative purposes only and not as a limitation, and any naming is for distinction only and has no limiting meaning.

[0023] The following is for reference. Figure 1 , Figure 1 This is a schematic diagram of the structure of an AI machine vision-enhanced intelligent image processing and analysis system provided in an embodiment of the present invention. Figure 1 As shown, an AI machine vision-enhanced intelligent image processing and analysis system 100 includes: Image acquisition module 101, image preprocessing module 102, multi-level image merging module 103, feature extraction module 104, raw / cooked recognition module 105, and result output module 106; The image acquisition module is used to acquire image information at different levels during food processing and transmit it to the image preprocessing module. The image preprocessing module performs noise reduction and contrast enhancement on the received image before transmitting it to the multi-level image merging module. The multi-level image merging module merges images from different levels and transmits the merged image to the feature extraction module. The feature extraction module extracts image feature information and transmits it to the raw / cooked food recognition module; the raw / cooked food recognition module performs layer-by-layer recognition of food raw / cooked status based on a preset algorithm and transmits the recognition result to the result output module. The result output module outputs the results of food raw / cooked identification.

[0024] It should be noted that the system includes an image acquisition module, an image preprocessing module, a multi-layer image merging module, a feature extraction module, a raw / cooked food recognition module, and a result output module. The image acquisition module is the component used to acquire image information at different levels during food processing; it includes an optical imaging unit and an image transmission unit. The optical imaging unit is responsible for imaging different layers of the food, while the image transmission unit is responsible for transmitting the image data to the image preprocessing module.

[0025] Specifically, the optical imaging unit in the image acquisition module can employ a high-resolution camera to ensure that the captured images are rich in detail and sufficiently clear. The image transmission unit can be wireless or wired, with the most suitable transmission method selected based on the actual food processing environment and equipment layout. The image preprocessing module performs denoising and contrast enhancement on the received images. The denoising unit uses a median filtering algorithm to remove image noise, while the contrast enhancement unit uses a grayscale stretching algorithm to enhance image contrast.

[0026] Preferably, the image registration unit in the multi-level image merging module can register images at different levels based on a feature point matching algorithm to ensure image alignment accuracy. The image fusion unit uses a weighted average method to fuse the registered images at different levels to obtain the merged image. In the weighted average method, the weights corresponding to different levels of images can be dynamically adjusted according to the importance and quality of the images to optimize the fusion effect. For example, images containing more detailed information can be assigned higher weights to ensure that this information is better preserved in the final merged image.

[0027] In some embodiments, the image acquisition module includes an optical imaging unit and an image transmission unit. The optical imaging unit is used to image different layers of food, and the image transmission unit is used to transmit the image data after imaging to an image preprocessing module.

[0028] It should be noted that the image acquisition module includes an optical imaging unit and an image transmission unit. The optical imaging unit is used to image different layers of the food, while the image transmission unit is used to transmit the image data to the image preprocessing module. The optical imaging unit refers to a device that captures images using optical principles; it can be a camera or scanner specifically designed to capture images of food. The image transmission unit is the device that transmits the image data captured by the optical imaging unit to the next processing stage; this can be via cable, wireless network, or other data transmission technologies.

[0029] Specifically, the optical imaging unit can employ a high-resolution camera, with parameters including, but not limited to, exposure time, aperture size, and focal length, to adapt to different food processing environments and lighting conditions. The image transmission unit can be a high-speed Ethernet cable, a Wi-Fi module, or a Bluetooth module; the specific transmission method chosen depends on the electromagnetic compatibility and transmission distance at the site. When setting parameters, the size of the image data and the processing speed need to be considered to ensure that the image data can be transmitted to the image preprocessing module in a timely and accurate manner.

[0030] Preferably, the optical imaging unit can be equipped with autofocus and auto exposure functions to adapt to food imaging under different distances and lighting conditions. Furthermore, the image transmission unit can optionally use 5G wireless communication technology to achieve faster data transmission speeds and longer transmission distances.

[0031] Furthermore, in certain situations where there is strong electromagnetic interference in the food processing environment, fiber optic cables can be considered as the image transmission unit to improve signal stability and anti-interference capabilities. These alternatives can be selected and adjusted according to specific application scenarios and requirements.

[0032] In some embodiments, the image preprocessing module includes a denoising unit and a contrast enhancement unit. The denoising unit uses a median filtering algorithm to remove image noise, and the contrast enhancement unit uses a grayscale stretching algorithm to enhance image contrast.

[0033] It should be noted that the image preprocessing module includes a denoising unit and a contrast enhancement unit. The denoising unit uses a median filtering algorithm to remove image noise, while the contrast enhancement unit uses a grayscale stretching algorithm to enhance image contrast. Here, the denoising unit refers to the component used to reduce random noise in the image, while the contrast enhancement unit refers to the component used to improve the contrast between different pixels in the image. The median filtering algorithm is a commonly used denoising technique that reduces noise by replacing each pixel value with the median of its neighborhood. The grayscale stretching algorithm is a contrast enhancement technique that enhances image contrast by adjusting the range of grayscale values.

[0034] Specifically, the median filtering algorithm in the denoising unit can adjust the denoising effect by setting the filter size, which is typically an odd number, such as a 3x3 or 5x5 window. The grayscale stretching algorithm in the contrast enhancement unit can be set by determining the range of input and output grayscale values. For example, the input grayscale value range from 0 to 255 can be mapped to a new range, such as 0 to 255, to enhance the image contrast. These parameters can be adjusted according to the specific characteristics of the image and the preprocessing objectives.

[0035] Preferably, the denoising unit can be further optimized by using an adaptive median filtering algorithm to adapt to the noise characteristics of different images. This algorithm dynamically adjusts the filter window size based on the local statistical characteristics of the image to achieve better denoising results. The contrast enhancement unit can also employ a more advanced contrast-limited adaptive histogram equalization (CLAHE) algorithm, which limits the contrast enhancement during histogram equalization to avoid noise amplification caused by excessive enhancement.

[0036] Furthermore, machine learning methods can be used to automatically adjust the parameters for denoising and contrast enhancement to adapt to different food images and environmental conditions. These alternatives can be selected and adjusted based on the specific application requirements and desired results.

[0037] In some embodiments, the multi-level image merging module includes an image registration unit and an image fusion unit. The image registration unit registers images of different levels based on a feature point matching algorithm, and the image fusion unit fuses the registered images of different levels using a weighted average method to obtain a merged image. The formula for the weighted average method is: Let the images at different levels be... The corresponding weight is The merged image ,in The number of image layers. Indicates the first The weights of the layer images, and .

[0038] It should be noted that the multi-level image merging module includes an image registration unit and an image fusion unit. The image registration unit registers images at different levels based on a feature point matching algorithm, while the image fusion unit uses a weighted average method to fuse the registered images at different levels to obtain the merged image. Here, the image registration unit refers to the component used to adjust different images to align them, while the image fusion unit refers to the component that merges multiple images into one. The feature point matching algorithm is a computer vision technique used to find and match identical feature points in different images. The weighted average method is a mathematical method used to combine multiple data points, each contributing differently to the final result according to its weight.

[0039] Specifically, the feature point matching algorithm in the image registration unit can be implemented using algorithms such as SIFT (Scale Invariant Feature Transform), SURF (Speed-Up Robust Feature Transform), or ORB (Oriented Fast and Rotated BRIEF). These algorithms can detect and describe key points in the image, and then register the image by matching these key points. The weights in the weighted average method can be set according to the quality, resolution, or other importance indicators of each image layer. For example, if an image layer provides more detail, it can be assigned a higher weight. The weight settings must satisfy the condition that the sum of all weights is 1.

[0040] Preferably, the image registration unit can employ more advanced deep learning methods, such as feature point matching based on convolutional neural networks, to improve the accuracy and robustness of registration. In addition to the weighted average method, the image fusion unit can also consider using other fusion techniques, such as Laplacian pyramid fusion or nonlinear fusion methods, which can provide better visual effects in different application scenarios.

[0041] Furthermore, the determination of weights can also be automated through learning algorithms. For example, machine learning models can be used to learn the optimal weights for each image layer based on historical data to achieve better fusion results. These alternatives can be selected and adjusted according to specific application requirements and image characteristics.

[0042] In some embodiments, the feature extraction module includes a texture feature extraction unit, a color feature extraction unit, and a shape feature extraction unit. The texture feature extraction unit extracts the texture features of the food image, the color feature extraction unit extracts the color features of the food image, and the shape feature extraction unit extracts the shape features of the food image. These features are then integrated and transmitted to the raw / cooked food recognition module.

[0043] It should be noted that the feature extraction module includes a texture feature extraction unit, a color feature extraction unit, and a shape feature extraction unit. These units are responsible for extracting the texture, color, and shape features of the food image, respectively, and then integrating these features before transmitting them to the raw / cooked food recognition module. Texture features refer to recurring patterns or structures in the image, color features involve the distribution and composition of colors in the image, and shape features describe the contours and geometric properties of objects in the image. These features are crucial for recognizing the raw / cooked state of food because they reflect the physical and chemical changes in the food.

[0044] Specifically, the texture feature extraction unit can use the Gray-Level Co-occurrence Matrix (GLCM) to analyze the image's texture information, the color feature extraction unit can use a color histogram to statistically analyze color distribution, and the shape feature extraction unit can use edge detection algorithms such as Canny edge detection to extract the contour information of the object. The parameter settings for these feature extraction methods, such as the angle and distance parameters of the GLCM, the number of bins in the color histogram, and the edge detection threshold, need to be adjusted according to the specific food type and image quality to achieve the best extraction results.

[0045] Preferably, the feature extraction module can be further refined. For example, the texture feature extraction unit can combine Local Binary Pattern (LBP) and Gabor filter to extract richer texture information; the color feature extraction unit can use color moments or color correlation analysis to provide more comprehensive color features; and the shape feature extraction unit can combine Hough transform to identify specific geometric shapes.

[0046] Furthermore, machine learning algorithms, such as convolutional neural networks (CNNs), can be incorporated into the feature extraction process to automatically learn and extract the features most useful for raw / cooked food identification. These alternatives can be selected and tailored to different food characteristics and identification needs to improve the accuracy and efficiency of identification.

[0047] In some embodiments, the preset algorithm of the raw / cooked food identification module includes the following steps: Step 1: Establish a raw and cooked food sample library, collect image data of foods at different stages of rawness or cookedness, and label them as raw or cooked categories, denoted as the raw and cooked category set. ,in Indicates the first Raw and cooked varieties, The total number of categories; simultaneously, multi-dimensional features of the image are extracted to form a feature vector set. ,in Indicates the first 1 eigenvector The dimension of the feature vector; Step 2: For the food image to be identified, extract its feature vector. ; Step 3: Calculate the feature vector of the image to be recognized. With each feature vector in the sample library similarity The calculation formula is:

[0048] in, This represents the vector dot product operation. Represents the magnitude of a vector; Step 4: Determine the raw / cooked category of the food to be identified based on similarity. Select the category of the sample with the highest similarity as the raw / cooked category of the food to be identified.

[0049] in, This indicates the raw / cooked category of the identification result.

[0050] It should be noted that the preset algorithm of the raw / cooked food recognition module includes establishing a raw / cooked food sample library, extracting multi-dimensional features from images, calculating the similarity between the feature vector of the food image to be recognized and each feature vector in the sample library, and determining the raw / cooked category of the food to be recognized based on the similarity. The raw / cooked food sample library refers to a dataset containing images of foods at different levels of rawness / cooking and their corresponding labels. Multi-dimensional features refer to a series of descriptive parameters extracted from the image, which can comprehensively represent the characteristics of the image. Similarity calculation refers to evaluating the closeness of two feature vectors in the feature space.

[0051] Specifically, establishing a food raw / cooked sample database requires collecting images of various foods in different raw / cooked states and labeling each image with its raw / cooked category. Extracting multi-dimensional features from the images can include texture, color, and shape features, with each feature vector composed of these features. Similarity can be calculated using vector dot product and the magnitude of the vectors. Parameter settings include selecting an appropriate distance metric (such as Euclidean distance or cosine similarity) and determining the dimensions of the feature vectors. These parameter settings directly affect the accuracy of similarity calculations and the results of raw / cooked identification.

[0052] Preferably, the operational steps of the raw / cooked identification module can be further refined. For example, when building the sample library, data augmentation techniques can be used to expand the number of samples and improve the model's generalization ability. In the feature extraction stage, a deep learning model can be introduced to automatically generate and select the most representative features. In similarity calculation, kernel tricks, such as Gaussian kernels, can be introduced to handle non-linearly separable feature spaces.

[0053] Furthermore, ensemble learning methods, such as random forests or gradient boosting machines, can be considered to integrate the decisions of multiple classifiers, thereby improving the accuracy and robustness of recognition. These alternatives can be selected and adjusted according to the specific application scenario and data characteristics to achieve the best recognition results.

[0054] In some embodiments, in step one, the extracted multidimensional features include the gray-level co-occurrence matrix features, color histogram features, and edge contour features of the image.

[0055] It should be noted that the extracted multi-dimensional features include the image's gray-level co-occurrence matrix (GLCM) features, color histogram features, and edge contour features. GLCM features describe image texture by analyzing the gray-level relationships between pixel pairs; color histogram features provide a statistical description of color distribution in the image; and edge contour features describe the shape and position of object edges in the image. These features together constitute a comprehensive description used to identify the raw / cooked state of food.

[0056] Specifically, the gray-level co-occurrence matrix (GLCM) feature can be obtained by calculating the difference in gray values ​​of pixel pairs in an image at specific directions and distances, with parameters including direction, distance, and number of gray levels. The color histogram feature can be obtained by counting the number of pixels of each color in the image, with parameters including color space (e.g., RGB, HSV) and the number of bins in the histogram. Edge contour features can be obtained using edge detection algorithms such as the Canny algorithm, with parameters including the edge detection threshold and Gaussian filtering parameters. These parameters need to be adjusted according to the specific characteristics of the food and the image quality to achieve the best extraction results.

[0057] Preferably, the operational steps can be further refined. For example, when extracting gray-level co-occurrence matrix features, different angles and spacings can be considered to capture the multi-directional characteristics of image texture. When extracting color histogram features, adaptive thresholding can be used to improve the discriminative power of color features. When extracting edge contour features, morphological operations can be combined to remove noise and fill holes to obtain more accurate edge information.

[0058] Furthermore, machine learning methods can be used to automatically select and optimize these features to improve recognition accuracy. These alternatives can be selected and adjusted according to different food characteristics and recognition needs to improve recognition accuracy and efficiency.

[0059] In some embodiments, in step three, when calculating similarity, different weights can be assigned to features of different dimensions. The revised similarity calculation formula is as follows:

[0060] in, and The first and second feature vectors of the image to be identified and the second feature vector of the sample, respectively, represent the first and second feature vectors of the image to be identified. Features in 100 dimensions.

[0061] It should be noted that when calculating similarity, different weights can be assigned to features of different dimensions. The similarity calculation formula can be further expressed as:

[0062] in, Indicates the first Weights of each feature dimension and The first and second feature vectors of the image to be identified and the second feature vector of the sample, respectively, represent the first and second feature vectors of the image to be identified. This approach utilizes multiple dimensions of features. It allows the system to adjust the similarity calculation based on the importance of the features, thereby improving recognition accuracy.

[0063] Specifically, the weights in similarity calculation The weighting can be set based on the feature's contribution to the classification. For example, if a feature shows a higher ability to distinguish between raw and cooked foods in historical data, then that feature can be given a higher weight. The weighting can be determined through expert experience, statistical analysis, or machine learning methods. Furthermore, the numerator in similarity calculations is the dot product of the feature vectors, and the denominator is the product of the magnitudes of the two vectors. These vector operations need to be performed precisely to ensure the accuracy of the similarity calculation.

[0064] Preferably, the operational steps can be further refined. For example, in determining the weights... When calculating similarity, feature selection methods based on information gain or model performance can be used to automatically determine weights. In similarity calculations, regularization terms can be introduced to avoid overfitting and improve the model's generalization ability.

[0065] Furthermore, different similarity metrics, such as cosine similarity or Jaccard similarity, can be considered to adapt to different types of features and data distributions. These alternatives can be selected and adjusted according to specific application scenarios and data characteristics to achieve the best recognition results.

[0066] In some embodiments, in step four, if there are multiple sample categories with high and close similarity, a voting mechanism is used to determine the final category, i.e., the categories ranked by similarity to the feature vector of the image to be identified are determined. The number of raw / cooked categories in each sample category is used as the final result, with the category having the highest number of raw / cooked categories. This is a preset threshold.

[0067] It should be noted that in step four, if multiple sample categories with high similarity exist, a voting mechanism is used to determine the final raw / cooked category. The voting mechanism is a decision-making method used to select the option that appears most frequently from multiple candidates as the final result. Here, it is used to address the problem of determining the final raw / cooked category when multiple sample categories have high similarity to the food image to be identified.

[0068] Specifically, the voting mechanism can rank the top performers based on the similarity of their feature vectors to the image to be identified. This is achieved by counting the number of raw and cooked categories in each sample category. This is a preset threshold, representing the number of sample categories considered in the similarity ranking. For example, if the image to be identified has a high similarity to three categories in the sample library, then the frequency of each raw / cooked category within those three categories can be counted, and the raw / cooked category with the highest frequency will be determined as the final raw / cooked category of the food to be identified. Parameter The settings can be adjusted according to actual application needs and the size of the sample library to balance the accuracy and complexity of decision-making.

[0069] Preferably, the operation steps can be further refined. For example, in determining the threshold... In this case, adjustments can be made dynamically based on the diversity of the sample library and the distribution of food categories. If the similarity between categories in the sample library is high, a larger adjustment may be needed. Values ​​are used to reduce misjudgments.

[0070] Furthermore, a confidence threshold can be introduced, whereby a sample category is only counted in the vote if the similarity exceeds a certain threshold. This can reduce misclassifications caused by noise or irrelevant features. Combining the voting mechanism with other decision-making algorithms, such as Bayesian classifiers or support vector machines, can also be considered to improve the accuracy of the final decision. These alternatives can be selected and adjusted according to the specific application scenario and data characteristics to achieve the best recognition results.

[0071] In some embodiments, the result output module includes a display unit and a data storage unit. The display unit is used to visually display the food raw / cooked identification result, and the data storage unit is used to store the identification result and related image data.

[0072] It should be noted that the results output module includes a display unit and a data storage unit, used to intuitively display the food raw / cooked identification results and store the identification results and related image data. The display unit refers to the interface used to show the identification results to the operator, while the data storage unit refers to the storage system used to save the identification results and related image data. These units ensure the traceability of the identification results and facilitate subsequent analysis.

[0073] Specifically, the display unit can be a touchscreen display or a computer monitor, used to display the real-time results of food raw / cooked identification. The data storage unit can be a local database, cloud storage service, or hard drive, used to permanently store the identification results and image data. Parameter settings include determining the data storage format, database access permissions, and data backup frequency. For example, data can be stored in a structured format in an SQL database for easy querying and analysis. Simultaneously, appropriate data encryption and access control mechanisms need to be set up to protect data security and privacy.

[0074] Preferably, the operation steps can be further refined. For example, the display unit can integrate a graphical user interface (GUI) to provide intuitive charts and statistical information to help the operator quickly understand the recognition results. The data storage unit can employ a distributed storage system to improve data reliability and access speed.

[0075] Furthermore, data compression techniques can be introduced to reduce storage space requirements while ensuring data integrity and readability. Machine learning algorithms can also be used to analyze stored data to optimize the performance of recognition algorithms. These alternatives can be selected and adjusted according to specific application scenarios and data characteristics to achieve the best user experience and system performance.

[0076] The above embodiments of the present invention have the following beneficial effects: The AI ​​machine vision-enhanced intelligent image processing and analysis system of the present invention can provide an efficient and accurate solution for identifying the raw and cooked nature of food. Through an integrated image acquisition module, the system can acquire image information at different levels during food processing. The image preprocessing module improves image quality through denoising and contrast enhancement, laying a solid foundation for subsequent analysis. The multi-level image merging module merges images at different levels through image registration and fusion techniques, enhancing image detail and depth, and providing more comprehensive data support for feature extraction. The feature extraction module comprehensively extracts texture, color, and shape features from food images, providing rich feature information for raw / cooked food identification. The raw / cooked food identification module uses a preset algorithm based on a sample library and feature vector set to achieve penetrating layer recognition of raw / cooked food, improving the accuracy and reliability of identification.

[0077] Furthermore, during the raw / cooked food identification process, the system can assign different weights based on the importance of features across different dimensions, optimizing the similarity calculation formula and resulting in more accurate identification results. When faced with multiple highly similar sample categories, the system can employ a voting mechanism to determine the final raw / cooked category, enhancing the stability of the identification results. The results output module not only displays the identification results intuitively but also stores relevant image data, facilitating subsequent data analysis and quality control, thereby improving the efficiency and effectiveness of the entire food processing and quality monitoring process.

[0078] Furthermore, the storage medium in the embodiments of this application stores program instructions capable of implementing all the above methods. These program instructions can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.

[0079] The above description is merely an explanation of some preferred embodiments of the present invention and the technical principles employed. Those skilled in the art should understand that the scope of the invention as described in the embodiments of the present invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. An AI-enhanced intelligent image processing and analysis system, characterized in that, It includes an image acquisition module, an image preprocessing module, a multi-level image merging module, a feature extraction module, a raw / cooked recognition module, and a result output module; The image acquisition module is used to acquire image information at different levels during food processing and transmit it to the image preprocessing module. The image preprocessing module performs noise reduction and contrast enhancement on the received image before transmitting it to the multi-level image merging module. The multi-level image merging module merges images from different levels and transmits the merged image to the feature extraction module. The feature extraction module extracts image feature information and transmits it to the raw / cooked food recognition module; the raw / cooked food recognition module performs layer-by-layer recognition of food raw / cooked status based on a preset algorithm and transmits the recognition result to the result output module. The result output module outputs the results of food raw / cooked identification.

2. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 1, characterized in that, The image acquisition module includes an optical imaging unit and an image transmission unit. The optical imaging unit is used to image different layers of food, and the image transmission unit is used to transmit the image data after imaging to the image preprocessing module.

3. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 2, characterized in that, The image preprocessing module includes a denoising unit and a contrast enhancement unit. The denoising unit uses a median filtering algorithm to remove image noise, and the contrast enhancement unit uses a grayscale stretching algorithm to enhance image contrast.

4. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 3, characterized in that, The multi-level image merging module includes an image registration unit and an image fusion unit. The image registration unit registers images at different levels based on a feature point matching algorithm. The image fusion unit uses a weighted average method to fuse the registered images at different levels to obtain a merged image. The weighted average method is used to fuse registered images of different levels, including: assuming the images of different levels are... The corresponding weight is The formula for calculating the fused image is: ； in, The number of image layers. Indicates the first The weights of the layer images, and .

5. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 4, characterized in that, The feature extraction module includes a texture feature extraction unit, a color feature extraction unit, and a shape feature extraction unit. The texture feature extraction unit extracts the texture features of the food image, the color feature extraction unit extracts the color features of the food image, and the shape feature extraction unit extracts the shape features of the food image. These features are then integrated and transmitted to the raw / cooked food recognition module.

6. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 5, characterized in that, The preset algorithm of the raw / cooked food identification module includes the following steps: Step 1: Establish a raw and cooked food sample library, collect image data of foods at different stages of rawness or cookedness, and label them as raw or cooked categories, denoted as the raw and cooked category set. ,in, Indicates the first Raw and cooked varieties, The total number of categories; simultaneously, multi-dimensional features of the image are extracted to form a feature vector set. ,in Indicates the first 1 eigenvector The dimension of the feature vector; Step 2: For the food image to be identified, extract its feature vector. ; Step 3: Calculate the feature vector of the image to be recognized. With each feature vector in the sample library similarity The calculation formula is: ； in, This represents the vector dot product operation. Represents the magnitude of a vector; Step 4: Determine the raw / cooked category of the food to be identified based on similarity. Select the category of the sample with the highest similarity as the raw / cooked category of the food to be identified. ； in, This indicates the raw / cooked category of the identification result.

7. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 6, characterized in that, In step one, the extracted multi-dimensional features include the image's gray-level co-occurrence matrix features, color histogram features, and edge contour features.

8. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 7, characterized in that, In step three, when calculating similarity, different weights are assigned to features of different dimensions. The similarity calculation formula further includes: ； in, and The first and second feature vectors of the image to be identified and the second feature vector of the sample, respectively, represent the first and second feature vectors of the image to be identified. Features in 100 dimensions.

9. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 8, characterized in that, In step four, if multiple sample categories with high and close similarity exist, a voting mechanism is used to determine the final category (raw or matured), i.e., the categories ranked by similarity to the feature vector of the image to be identified are selected. The number of raw / cooked categories in each sample category is used as the final result, with the category having the highest number of raw / cooked categories. This is a preset threshold.

10. The AI ​​machine vision-enhanced intelligent image processing and analysis system according to claim 1, characterized in that, The result output module includes a display unit and a data storage unit. The display unit is used to intuitively display the food raw / cooked identification results, and the data storage unit is used to store the identification results and related image data.

Images