Image processing system for identifying anthocyanin content of colored potatoes

Abstract

The invention discloses an image processing system for identifying the anthocyanin content of colored potatoes, which is characterized in that three types of images of visible light, epidermis texture and spectrum are acquired by setting an image acquisition unit, and weighted analysis is carried out, so that the representativeness and distinction degree of image data are enhanced; the illumination correction unit dynamically adjusts an image compensation strategy according to the environment illumination intensity, and feature extraction deviation caused by uneven illumination is remarkably reduced; the feature selection unit fuses multi-source image data and extracts a target comprehensive feature vector, and complementary information among multi-modal data is fully mined; and the prediction unit generates an anthocyanin content function and a distribution heat map based on the target comprehensive feature vector, so that the recognition precision and the visualization capability are improved, the defects that a traditional method is sensitive to illumination and poor in nonlinear modeling capability are overcome, and efficient, accurate and stable recognition of the anthocyanin content of the potatoes is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an image processing system for identifying anthocyanin content in colored potatoes. Background Art

[0002] Identifying the anthocyanin content of colored potatoes is essential for assessing their nutritional value, selecting varieties, and processing applications. Anthocyanins are natural antioxidants with health benefits such as enhancing immunity, delaying aging, and protecting cardiovascular and vision. The types and contents of anthocyanins vary significantly between varieties and parts of the potato, directly impacting their commercial value and functional properties. Accurate testing can guide the cultivation of high-anthocyanin varieties, optimize processing techniques, and provide a scientific basis for the development of functional foods, while also meeting consumer demand for healthier diets.

[0003] In existing technologies, due to the high diversity of potato skin texture and uneven color distribution, traditional models find it difficult to accurately capture the nonlinear relationship between color parameters and anthocyanin content, and are easily affected by light interference, resulting in feature extraction deviations and reduced recognition accuracy. Summary of the Invention

[0004] The purpose of the present invention is to provide an image processing system for identifying the anthocyanin content of colored potatoes, so as to solve the technical problems that traditional models are difficult to accurately capture the nonlinear relationship between color parameters and anthocyanin content, and are easily affected by light interference, resulting in feature extraction deviation and reduced recognition accuracy.

[0005] The technical solution of the present invention is achieved as follows:

[0006] An image processing system for identifying anthocyanin content in colored potatoes, comprising: an image acquisition unit, an illumination correction unit, a feature selection unit, and a prediction unit;

[0007] The image acquisition unit is used to acquire a visible light image, a skin texture image and a spectral image of a potato and perform weighted analysis on the images to generate a target visible light image, a target skin texture image and a target spectral image;

[0008] The illumination correction unit is used to adaptively correct the compensation strategies corresponding to the potato visible light image, the potato skin texture image and the spectral image according to the illumination intensity of the potato surrounding environment;

[0009] The feature selection unit is used to fuse the target visible light image, the target epidermal texture image and the target spectral image to generate target image data, and select a target comprehensive feature vector of the target image data based on a target neural network;

[0010] The prediction unit performs mathematical calculation based on the target comprehensive feature vector to output a target content function and distribution heat map data of potato anthocyanins.

[0011] A further technical solution is that the image acquisition unit includes:

[0012] An image acquisition module, which uses a visible light camera, a three-dimensional scanner and a near-infrared spectral imager to acquire the visible light image, the skin texture image and the spectral image of the potato;

[0013] An image weighting module, configured to combine the potato visible light image, the potato skin texture image, and the potato spectral image into a feature matrix, and assign weights according to the information contribution rate of each modality in the feature matrix;

[0014] a data processing module for eliminating low-weight and high-weight values of the information contribution rate of each modality of the feature matrix, reintegrating the remaining weight values, and establishing a new splicing relationship with the potato visible light image, skin texture image, and spectral image to construct a weighted covariance matrix;

[0015] The target output module is used to generate the corresponding target visible light image, target skin texture image and target spectral image from the potato visible light image and skin texture image according to the weighted covariance matrix.

[0016] A further technical solution is that the image weighting module performs the following steps:

[0017] Step S11, respectively extracting the color histogram, gray-level co-occurrence matrix features, and band reflectance curve corresponding to the potato visible light image, skin texture image, and spectral image;

[0018] Step S12: inputting the color histogram, gray-level co-occurrence matrix features, and band reflectance curve into a multi-scale feature pyramid network for modal feature interaction to generate a hybrid feature vector containing local details and global semantics;

[0019] Step S13: Calculate the feature importance score of each modality for the target detection task through a decision tree model, and generate a weight coefficient based on the sensitivity of the spectral data to anthocyanin detection;

[0020] Step S14: input the weight coefficient into the gated attention mechanism, and perform nonlinear weighting on the feature channels of different modalities through learnable attention weights.

[0021] A further technical solution is that the data processing module executes the steps including:

[0022] Step S21: Standardize all feature data and filter out data that has little impact on the results or has abnormal fluctuations;

[0023] Step S22: extracting color, texture, and spectral features respectively, and unifying the feature information into the same frequency dimension through multi-scale decomposition;

[0024] Step S23: Combine the multi-source data of the same frequency dimension, calculate the correlation between the multi-source data, assign different weights according to importance, and establish data relationships.

[0025] A further technical solution is that the illumination correction unit performs the following steps:

[0026] Step S31: using a common camera and a spectrometer to measure the light intensity of the potato's surrounding environment in real time;

[0027] Step S32: Dynamically adjust the brightness of the visible light image, extract fine surface texture features and reflection information of different color bands, and preliminarily identify anthocyanin features;

[0028] Step S33: amplifying the anthocyanin feature and reducing noise in the image under conditions of dim light or strong contrast between light and dark by using an intelligent algorithm;

[0029] Step S34: automatically adjusting the weight ratio of each band according to the sensitivity of anthocyanin to a specific color band, and jointly optimizing the weight ratio.

[0030] A further technical solution is that step S33 specifically includes:

[0031] Step 331: Separate the illumination component and the anthocyanin reflection component of the low-light image;

[0032] Step 332: Using a loss function to constrain the structural similarity between the illumination component and the anthocyanin reflectance component, generate a preliminarily processed image, and simultaneously remove random noise in the image;

[0033] Step 333: Based on the LLCCNN architecture, perform multi-channel analysis on the processed image to generate an enhanced image that can highlight the characteristics of anthocyanins;

[0034] Step 334: Strengthen the features related to anthocyanins in the enhanced image and filter the remaining noise again.

[0035] A further technical solution is that the feature selection unit performs the following steps:

[0036] Step S41: adjusting the target visible light image, target skin texture image, and target spectral image to a uniform clarity and color range, eliminating the influence of light, and extracting common features of the target visible light image, target skin texture image, and target spectral image;

[0037] Step S42: analyzing color changes of visible light, extracting surface concave-convex features of skin texture, and calculating band data corresponding to color depth from spectral images;

[0038] Step S43: assigning importance according to the degree of association of target features of each image of the target visible light image, the target epidermal texture image, and the target spectral image, and eliminating duplicate or irrelevant information;

[0039] Step S44: splicing the key information extracted from the target visible light image, the target epidermal texture image, and the target spectral image into a complete data packet, retaining the core data that best represents the target characteristics;

[0040] Step S45: inputting the core data into the neural network model, adjusting the neural network model parameters to find the best feature combination, and verifying whether the feature combination accurately determines different target features;

[0041] Step S46: Output the target feature and generate a comprehensive feature vector.

[0042] A further technical solution is that step S45 specifically includes:

[0043] Step S451: input the core data into a preset model;

[0044] Step S452: automatically adjusting the internal weights and biases of the preset model to try different feature combinations;

[0045] Step S453: Use the validation set to monitor the distinguishing ability of the feature combination in real time and record the optimal parameter configuration;

[0046] Step S454: using a test set to check whether the preset model can accurately identify different target features;

[0047] Step S455: If the recognition accuracy does not meet the standard, return to step S452 to readjust the parameters until the feature combination achieves the expected effect.

[0048] A further technical solution is that the prediction unit performs the following steps:

[0049] Step S51: inputting the target comprehensive feature vector into a preset analysis model;

[0050] Step S52: Automatically calculate and generate an objective function of anthocyanin content using the preset analysis model;

[0051] Step S53: verify the actual test results of the sample and adjust the objective function parameters;

[0052] Step S54: Calculate the anthocyanin concentration value of each pixel according to the objective function, and generate corresponding distribution heat map data using color depth;

[0053] Step S55: Integrate the objective function and the heat map data into an outputtable result file.

[0054] A further technical solution is that step S52 specifically includes:

[0055] Step S521: taking the target comprehensive feature vector as input data;

[0056] Step S522: automatically identifying the mathematical relationship between the input data and the anthocyanin content through a clustering algorithm;

[0057] Step S523: Iteratively optimize the model parameters using known sample data to minimize the error between the predicted result and the actual value;

[0058] Step S524: output the final anthocyanin content calculation formula;

[0059] Step S525: Encapsulate the calculation formula into a directly callable function module for subsequent batch calculation or real-time prediction.

[0060] The beneficial effects of the present invention are:

[0061] By setting up an image acquisition unit to obtain three types of images: visible light, epidermal texture, and spectrum, and performing weighted analysis, the representativeness and discrimination of the image data are enhanced; the illumination correction unit dynamically adjusts the image compensation strategy according to the ambient light intensity, significantly reducing the feature extraction deviation caused by uneven illumination; the feature selection unit fuses multi-source image data and extracts the target comprehensive feature vector, fully mining the complementary information between multimodal data; the prediction unit generates anthocyanin content function and distribution heat map based on the target comprehensive feature vector, improving the recognition accuracy and visualization ability, overcoming the defects of traditional methods such as sensitivity to light and weak nonlinear modeling ability, and realizing efficient, accurate and stable identification of potato anthocyanin content. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0063] Figure 1 This is a schematic diagram of the structure of an image processing system for identifying anthocyanin content in colored potatoes according to the present invention;

[0064] Figure 2 This is a structural schematic diagram of an image acquisition unit in an image processing system for identifying anthocyanin content in colored potatoes according to the present invention;

[0065] Figure 3 A flow chart of the execution steps of an image weighting module in an image processing system for identifying anthocyanin content in colored potatoes according to the present invention;

[0066] Figure 4 A flow chart of the execution steps of a data processing module in an image processing system for identifying anthocyanin content in colored potatoes according to the present invention;

[0067] Figure 5 A flow chart of the steps executed by the illumination correction unit in the image processing system for identifying the anthocyanin content of colored potatoes according to the present invention;

[0068] Figure 6 A flow chart of the execution steps of a feature selection unit in an image processing system for identifying anthocyanin content in colored potatoes according to the present invention;

[0069] Figure 7 A flow chart of the execution steps of a prediction unit in an image processing system for identifying anthocyanin content in colored potatoes according to the present invention. DETAILED DESCRIPTION

[0070] In order to better understand the technical content of the present invention, specific embodiments are provided below, and the present invention is further described in conjunction with the accompanying drawings.

[0071] See also Figures 1 to 7 The present invention provides an image processing system for identifying the anthocyanin content of colored potatoes, comprising: an image acquisition unit, an illumination correction unit, a feature selection unit, and a prediction unit; the image acquisition unit is used to acquire a visible light image, a skin texture image, and a spectral image of a potato and perform weighted analysis on the images respectively to generate a target visible light image, a target skin texture image, and a target spectral image; the illumination correction unit is used to adaptively correct the compensation strategy corresponding to the visible light image, the skin texture image, and the spectral image of the potato according to the illumination intensity of the surrounding environment of the potato; the feature selection unit is used to fuse the target visible light image, the target skin texture image, and the target spectral image to generate target image data, and select a target comprehensive feature vector of the target image data based on a target neural network; the prediction unit is used to perform mathematical calculations based on the target comprehensive feature vector to output a target content function of potato anthocyanins and distribution heat map data.

[0072] The present invention realizes accurate identification of the anthocyanin content of colored potatoes by sequentially connecting an image acquisition unit, an illumination correction unit, a feature selection unit and a prediction unit, which work together. The image acquisition unit synchronously acquires visible light, epidermal texture and spectral images, and enhances anthocyanin-related bands and surface texture through weighted analysis. The illumination correction unit dynamically adjusts the compensation strategy based on the ambient light intensity to eliminate image distortion caused by uneven illumination. The feature selection unit fuses multi-source image data and uses a neural network to extract a comprehensive feature vector containing color, texture and spectrum, and mines nonlinear associations. The prediction unit outputs anthocyanin content function and distribution heat map through mathematical modeling to achieve quantitative analysis and visualization.

[0073] It is worth noting that multimodal data fusion makes up for the limitations of a single image and improves feature representation capabilities; adaptive illumination correction technology effectively reduces environmental interference and ensures data stability; neural network-driven feature extraction and modeling are significantly superior to traditional linear methods, accurately capturing the complex relationship between anthocyanin content and color parameters, and ultimately achieving high-precision, low-error anthocyanin detection.

[0074] In an embodiment of the present invention, an image acquisition unit is set to obtain three types of images, namely visible light, epidermal texture and spectrum, and perform weighted analysis, thereby enhancing the representativeness and discrimination of the image data; the illumination correction unit dynamically adjusts the image compensation strategy according to the ambient light intensity, significantly reducing the feature extraction deviation caused by uneven illumination; the feature selection unit fuses multi-source image data and extracts the target comprehensive feature vector, fully mining the complementary information between multimodal data; the prediction unit generates anthocyanin content function and distribution heat map based on the target comprehensive feature vector, thereby improving the recognition accuracy and visualization capability, overcoming the defects of traditional methods in being sensitive to illumination and having weak nonlinear modeling capabilities, and realizing efficient, accurate and stable recognition of potato anthocyanin content.

[0075] Preferably, the image acquisition unit includes: an image acquisition module, which uses a visible light camera, a three-dimensional scanner and a near-infrared spectral imager to acquire a visible light image, an epidermal texture image and a spectral image of a potato; an image weighting module, which is used to splice the visible light image, the epidermal texture image and the spectral image of the potato into a feature matrix, and assign weights according to the information contribution rate of each mode of the feature matrix; a data processing module, which is used to eliminate the low-weight values and high-weight values of the information contribution rate of each mode of the feature matrix, re-integrate the remaining weight values, and establish a new splicing relationship with the visible light image, the epidermal texture image and the spectral image of the potato to construct a weighted covariance matrix; a target output module, which is used to generate a corresponding target visible light image, a target epidermal texture image and a target spectral image from the visible light image and the epidermal texture image according to the weighted covariance matrix.

[0076] In this embodiment, the information contribution rate can be that the spectral data is more sensitive to anthocyanins than visible light; the key feature can be the spectral response of a specific band; the low weight can be noise or irrelevant features; and the high weight can be the risk of overfitting.

[0077] The image acquisition module uses a visible light camera, a three-dimensional scanner and a near-infrared spectral imager to synchronously collect visible light images of potato characterization color, microstructure epidermal texture images and wavelength characteristic spectral images, achieving complementary coverage of multi-dimensional data; the image weighting module dynamically assigns weights by analyzing the information contribution rate of each mode in the feature matrix to highlight key features; the data processing module retains robust features with medium weights by eliminating low-weight and high-weight values, and re-integrates multimodal data based on the remaining weights to construct a weighted covariance matrix, effectively eliminating redundant information interference; the target output module uses the covariance matrix to reconstruct visible light and texture images to generate target visible light images (enhanced color contrast), target epidermal texture images (refined surface structure) and target spectral images (extracted key bands), ultimately providing high-quality input for subsequent modeling.

[0078] Through the dynamic weighted fusion of multimodal data, the problem of insufficient modeling of the coupling relationship between color and texture by traditional single image processing methods is solved. At the same time, through the construction of a weighted covariance matrix, the robustness of the model to illumination changes and noise is significantly improved, achieving high precision and anti-interference ability in the identification of anthocyanin content.

[0079] Preferably, the image weighting module performs the steps including:

[0080] Step S11, respectively extracting the color histogram, gray-level co-occurrence matrix features, and band reflectance curve corresponding to the potato visible light image, skin texture image, and spectral image;

[0081] Step S12: inputting the color histogram, gray-level co-occurrence matrix features, and band reflectance curve into a multi-scale feature pyramid network for modal feature interaction to generate a hybrid feature vector containing local details and global semantics;

[0082] Step S13: Calculate the feature importance score of each modality for the target detection task through the decision tree model, and generate a weight coefficient based on the sensitivity of the spectral data to anthocyanin detection;

[0083] Step S14: Input the weight coefficient into the gated attention mechanism, and perform nonlinear weighting on the feature channels of different modalities through learnable attention weights.

[0084] In this embodiment, the color histogram can be used to characterize the macroscopic color distribution; the grayscale co-occurrence matrix feature can be used to capture the microscopic texture patterns; the band reflectance curve can be used to quantify the spectral response; the target detection task can be the sensitivity of spectral data to anthocyanins; and nonlinear weighting of feature channels can be used to adjust the weights of different colors, textures, and spectra in the image, so that the model can focus more on key information and weaken interference, thereby improving the accuracy and efficiency of identifying anthocyanin content.

[0085] By extracting color histograms, gray-level co-occurrence matrix features, and band reflectance curves, a multi-dimensional feature foundation is provided for subsequent fusion. A multi-scale feature pyramid network is used to break down information barriers between modalities. Through top-down semantic guidance and lateral connection detail enhancement, the color features of visible light, the spatial correlation of texture, and the wavelength characteristics of the spectrum deeply interact at the local detail and global semantic levels, generating a hybrid feature vector. A decision tree model is used to quantify the contribution of each modality to the object detection task, quantifying that spectral data scores higher due to its high sensitivity to anthocyanins. Weight coefficients are set based on domain knowledge to avoid the subjectivity of traditional manual weighting. A gated attention mechanism is introduced to nonlinearly weight the feature channels of different modalities using learnable attention weights, dynamically suppressing illumination noise and enhancing the anthocyanin band.

[0086] This module breaks through the limitations of traditional methods in which multimodal feature fusion relies on fixed weights (such as linear weighting or simple splicing), and combines multi-scale feature interaction, dynamic weight generation and adaptive attention mechanism to achieve intelligent and adaptive feature fusion.

[0087] Preferably, the data processing module performs the steps including:

[0088] Step S21: Standardize all feature data and filter out data that has little impact on the results or has abnormal fluctuations;

[0089] Step S22: extract color, texture, and spectral features respectively, and unify the feature information into the same frequency dimension through multi-scale decomposition;

[0090] Step S23: Combine multi-source data of the same frequency dimension, calculate the correlation between the multi-source data, assign different weights according to importance, and establish data relationships.

[0091] In this embodiment, the color, texture, and spectral features are Z-score normalized to eliminate dimensional differences between different modalities. At the same time, statistical thresholds are used to filter out pixel noise data caused by low contribution rates or sudden changes in illumination to ensure the reliability of subsequent analysis. NSCT decomposition technology is used to unify the macroscopic color distribution, texture microstructure, and spectral wavelength response characteristics into a low-frequency band to retain the overall trend of color and spectrum, and a high-frequency band to retain texture details, thereby preventing fusion bias caused by inconsistent multimodal feature scales. By calculating the correlation between multi-source data, weights are dynamically assigned, that is, spectral features are given higher weights due to their high sensitivity to anthocyanins, and data relationships are constructed to achieve collaborative modeling between features.

[0092] Preferably, the illumination correction unit performs the steps including:

[0093] Step S31: using a common camera and a spectrometer to measure the light intensity of the potato's surrounding environment in real time;

[0094] Step S32: Dynamically adjust the brightness of the visible light image, extract the fine texture features of the surface and the reflection information of different color bands, and preliminarily lock the anthocyanin characteristics;

[0095] Step S33: using an intelligent algorithm to amplify anthocyanin features and reduce noise in the image under conditions of dim light or strong contrast between light and dark;

[0096] Step S34: automatically adjust the weight ratio of each band according to the sensitivity of anthocyanin to a specific color band, and jointly optimize the weight ratio.

[0097] In this embodiment, the visible light intensity is captured in real time by an ordinary camera and the wavelength distribution of the ambient light is quantified by a spectrometer to achieve the color temperature difference between natural light and artificial light sources. Combined with adaptive histogram equalization adjustment and extraction of fine lines and color band reflection information, the preliminary positioning of anthocyanin features is achieved in visible light images. An attention mechanism algorithm based on U-Net is used to enhance anthocyanin features and suppress noise in scenes with alternating light and dark contrast shadows and strong light, avoiding the problem of feature loss caused by sudden changes in illumination. According to the high sensitivity of anthocyanins to the 520-560nm band, the contribution of each color channel is dynamically adjusted (such as enhancing the blue-green band and suppressing the red band) through an adaptive weight distribution algorithm to achieve accurate weighted optimization of spectral features.

[0098] Combining multi-sensor environmental perception, dynamic feature enhancement, intelligent noise reduction and band adaptive weighting can avoid the problem of light color temperature fluctuation of ±2000K caused by the temperature difference between morning and evening, the distortion of anthocyanin characteristics and strong noise interference.

[0099] Preferably, step S33 specifically includes:

[0100] Step 331: Separate the illumination component and the anthocyanin reflection component of the low-light image;

[0101] Step 332: Using a loss function to constrain the structural similarity between the illumination component and the anthocyanin reflectance component, generate a preliminarily processed image, and simultaneously remove random noise in the image;

[0102] Step 333: Based on the LLCCNN architecture, perform multi-channel analysis on the processed image to generate an enhanced image that can highlight the characteristics of anthocyanins;

[0103] Step 334: Enhance the features related to anthocyanins in the enhanced image and filter out the remaining noise again.

[0104] In this embodiment, the illumination component and the anthocyanin reflectance component of the low-light image are separated based on the Retinex theory, and the artifacts caused by sudden illumination changes are suppressed by introducing a structural similarity loss function. The non-local mean filter is simultaneously used to remove random noise to overcome the noise amplification defect caused by the division operation. The RGB and spectral channel analysis are processed in parallel based on a lightweight convolutional neural network. The reflectivity of the purple area of the anthocyanin is dynamically enhanced through a learnable channel attention mechanism to reduce the amount of calculation. The enhanced image is subjected to secondary denoising by combining frequency domain filtering and spatial domain adaptive denoising to avoid the coupling interference problem between the residual noise and the anthocyanin features, thereby achieving accurate extraction of anthocyanin features under low light conditions.

[0105] Preferably, the feature selection unit performs the steps including:

[0106] Step S41: Adjust the target visible light image, target skin texture image, and target spectral image to a uniform clarity and color range to eliminate the influence of light, and extract common features of the target visible light image, target skin texture image, and target spectral image;

[0107] Step S42: analyzing color changes of visible light, extracting surface concave-convex features of skin texture, and calculating band data corresponding to color depth from spectral images;

[0108] Step S43: assigning importance according to the degree of association of target features of each image of the target visible light image, the target epidermal texture image, and the target spectral image, and eliminating duplicate or irrelevant information;

[0109] Step S44: splicing the key information extracted from the target visible light image, the target skin texture image, and the target spectral image into a complete data packet, retaining the core data that best represents the target characteristics;

[0110] Step S45: input the core data into the neural network model, adjust the neural network model parameters to find the best feature combination, and verify whether the feature combination accurately determines different target features;

[0111] Step S46: Output the target features and generate a comprehensive feature vector.

[0112] In this embodiment, illumination interference is eliminated through standardization processing, and alignment of visible light, epidermal texture, and spectral images in terms of clarity and color range is achieved. Color gradient, surface unevenness, and spectral reflectance are extracted respectively, and a multi-dimensional feature description system is constructed. The strong correlation between the color and spectrum in the purple region between each feature is dynamically calculated based on the mutual information matrix. Duplicate information is eliminated by retaining features with information entropy > 0.8. An adaptive splicing strategy is adopted to retain the coupling relationship between the reflectance of the purple band of anthocyanin and the surface roughness, avoiding the dimensionality explosion caused by traditional simple splicing. Through the convolution kernel weights of ResNet-18, the synergistic enhancement mode of the high reflectance band of anthocyanin features and specific texture features is automatically searched. The generated comprehensive feature vector can be directly input into the classification model to realize end-to-end feature-prediction mapping, thereby avoiding the distortion of anthocyanin features due to illumination fluctuations of ±30% and the poor generalization ability of the model.

[0113] Preferably, step S45 specifically includes:

[0114] Step S451: input the core data into the preset model;

[0115] Step S452: Try different feature combinations by automatically adjusting the internal weights and biases of the preset model;

[0116] Step S453: Use the validation set to monitor the distinguishing ability of the feature combination in real time and record the optimal parameter configuration;

[0117] Step S454: Use the test set to check whether the preset model can accurately identify different target features;

[0118] Step S455: If the recognition accuracy does not meet the standard, return to step S452 to readjust the parameters until the feature combination achieves the expected effect.

[0119] In this example, a pre-set model automatically adjusts weights and biases, exploring different combinations of color bands and texture features in synergistic patterns. A validation set is used to monitor the model's discriminative capabilities in real time, recording the optimal parameter configuration. If the test set's recognition accuracy is less than 90%, a recurrent optimization mechanism is triggered until the model output is stable and accurate, avoiding feature combination failures caused by overfitting or underfitting.

[0120] Preferably, the prediction unit performs the steps including:

[0121] Step S51: input the target comprehensive feature vector into a preset analysis model;

[0122] Step S52: Automatically calculate and generate an objective function for anthocyanin content using a preset analysis model;

[0123] Step S53: verify the actual test results of the sample and adjust the objective function parameters;

[0124] Step S54: Calculate the anthocyanin concentration value of each pixel according to the objective function, and generate corresponding distribution heat map data using color depth;

[0125] Step S55: Integrate the objective function and the heat map data into an outputtable result file.

[0126] In this example, a mathematical model is automatically generated by using a pre-set analysis model driven by a comprehensive feature vector, and the weight coefficients are dynamically adjusted through a validation set. Ultimately, anthocyanin concentrations are calculated with pixel-level accuracy and a heat map is generated. This achieves non-destructive testing, spatial distribution visualization, and robustness in complex environments.

[0127] Preferably, step S52 specifically includes:

[0128] Step S521: taking the target comprehensive feature vector as input data;

[0129] Step S522: automatically identifying the mathematical relationship between the input data and the anthocyanin content through a clustering algorithm;

[0130] Step S523: Iteratively optimize the model parameters using known sample data to minimize the error between the predicted result and the actual value;

[0131] Step S524: output the final anthocyanin content calculation formula;

[0132] Step S525: encapsulate the calculation formula into a directly callable function module for subsequent batch calculation or real-time prediction.

[0133] In this embodiment, a clustering algorithm is used to automatically mine the coupling relationship between the target feature and the anthocyanin content in terms of color depth and reflectance band, and the model parameters are iteratively optimized through known samples to ultimately generate a highly interpretable mathematical formula:

[0134]

[0135] Where C is the anthocyanin content calculated by clustering algorithm; K is the total number of groups after clustering; i is the cluster group index, ranging from 1 to k; n i is the number of samples in the i-th cluster group; a i The minimum value of the input data for the i-th cluster group; b iis the maximum value of the input data of the i-th cluster group; x is a continuous input variable; μ i is the mean of the input data of the i-th cluster group; σ i is the standard deviation of the input data of the i-th cluster group; exp is the exponential function used to construct the weight function in the form of Gaussian distribution; is the sample mean of the known anthocyanin content in the i-th cluster group; x ij The jth input data of the i-th cluster group j ranges from 1 to n i .

[0136] The C value calculated by the formula represents the anthocyanin content, and its value range is [0, +∞]. When C = 0, the correlation strength between the input data of all cluster groups and the sample mean of anthocyanin content is zero, indicating that the sample does not contain anthocyanins. When C > 0, a larger value indicates a higher anthocyanin content. The numerator weights the data distribution within the cluster group through Gaussian integral, and the denominator normalizes the data dispersion within the group through standard deviation. The overall reflection reflects the comprehensive mapping relationship between the input data distribution and anthocyanin content, making it suitable for rapid quantitative analysis.

[0137] Clustering-driven feature-content correlation modeling, dynamic parameter optimization and formula encapsulation are combined to solve the problems of poor adaptability and low computational efficiency of complex visible light + spectrum + texture data.

[0138] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An image processing system for identifying anthocyanin content in colored potatoes, characterized in that: include: Image acquisition unit, illumination correction unit, feature selection unit and prediction unit; The image acquisition unit is used to acquire a visible light image, a skin texture image and a spectral image of a potato and perform weighted analysis on the images to generate a target visible light image, a target skin texture image and a target spectral image; The illumination correction unit is used to adaptively correct the compensation strategies corresponding to the potato visible light image, the potato skin texture image and the spectral image according to the illumination intensity of the potato surrounding environment; The feature selection unit is used to fuse the target visible light image, the target epidermal texture image and the target spectral image to generate target image data, and select a target comprehensive feature vector of the target image data based on a target neural network; The prediction unit performs mathematical calculation based on the target comprehensive feature vector to output a target content function and distribution heat map data of potato anthocyanins.

2. The image processing system for identifying anthocyanin content in colored potatoes according to claim 1, characterized in that: The image acquisition unit includes: An image acquisition module, which uses a visible light camera, a three-dimensional scanner and a near-infrared spectral imager to acquire the visible light image, the skin texture image and the spectral image of the potato; An image weighting module, configured to combine the potato visible light image, the potato skin texture image, and the potato spectral image into a feature matrix, and assign weights according to the information contribution rate of each modality in the feature matrix; a data processing module for eliminating low-weight and high-weight values of the information contribution rate of each modality of the feature matrix, reintegrating the remaining weight values, and establishing a new splicing relationship with the potato visible light image, skin texture image, and spectral image to construct a weighted covariance matrix; The target output module is used to generate the corresponding target visible light image, target skin texture image and target spectral image from the potato visible light image and skin texture image according to the weighted covariance matrix.

3. The image processing system for identifying anthocyanin content in colored potatoes according to claim 2, characterized in that: The image weighting module performs the following steps: Step S11, respectively extracting the color histogram, gray-level co-occurrence matrix features, and band reflectance curve corresponding to the potato visible light image, skin texture image, and spectral image; Step S12: inputting the color histogram, gray-level co-occurrence matrix features, and band reflectance curve into a multi-scale feature pyramid network for modal feature interaction to generate a hybrid feature vector containing local details and global semantics; Step S13: Calculate the feature importance score of each modality for the target detection task through a decision tree model, and generate a weight coefficient based on the sensitivity of the spectral data to anthocyanin detection; Step S14: input the weight coefficient into the gated attention mechanism, and perform nonlinear weighting on the feature channels of different modalities through learnable attention weights.

4. The image processing system for identifying anthocyanin content in colored potatoes according to claim 2, characterized in that: The data processing module performs the following steps: Step S21: Standardize all feature data and filter out data that has little impact on the results or has abnormal fluctuations; Step S22: extracting color, texture, and spectral features respectively, and unifying the feature information into the same frequency dimension through multi-scale decomposition; Step S23: Combine the multi-source data of the same frequency dimension, calculate the correlation between the multi-source data, assign different weights according to importance, and establish data relationships.

5. The image processing system for identifying anthocyanin content in colored potatoes according to claim 1, characterized in that: The illumination correction unit performs the following steps: Step S31: using a common camera and a spectrometer to measure the light intensity of the potato's surrounding environment in real time; Step S32: Dynamically adjust the brightness of the visible light image, extract fine surface texture features and reflection information of different color bands, and preliminarily identify anthocyanin features; Step S33: amplifying the anthocyanin feature and reducing noise in the image under conditions of dim light or strong contrast between light and dark by using an intelligent algorithm; Step S34: automatically adjusting the weight ratio of each band according to the sensitivity of anthocyanin to a specific color band, and jointly optimizing the weight ratio.

6. The image processing system for identifying anthocyanin content in colored potatoes according to claim 5, characterized in that: The step S33 specifically includes: Step 331: Separate the illumination component and the anthocyanin reflection component of the low-light image; Step 332: Using a loss function to constrain the structural similarity between the illumination component and the anthocyanin reflectance component, generate a preliminarily processed image, and simultaneously remove random noise in the image; Step 333: Based on the LLCCNN architecture, perform multi-channel analysis on the processed image to generate an enhanced image that can highlight the characteristics of anthocyanins; Step 334: Strengthen the features related to anthocyanins in the enhanced image and filter the remaining noise again.

7. The image processing system for identifying anthocyanin content in colored potatoes according to claim 1, characterized in that: The feature selection unit performs the following steps: Step S41: adjusting the target visible light image, target skin texture image, and target spectral image to a uniform clarity and color range, eliminating the influence of light, and extracting common features of the target visible light image, target skin texture image, and target spectral image; Step S42: analyzing color changes of visible light, extracting surface concave-convex features of skin texture, and calculating band data corresponding to color depth from spectral images; Step S43: assigning importance according to the degree of association of target features of each image of the target visible light image, the target epidermal texture image, and the target spectral image, and eliminating duplicate or irrelevant information; Step S44: splicing the key information extracted from the target visible light image, the target epidermal texture image, and the target spectral image into a complete data packet, retaining the core data that best represents the target characteristics; Step S45: inputting the core data into the neural network model, adjusting the neural network model parameters to find the best feature combination, and verifying whether the feature combination accurately determines different target features; Step S46: Output the target feature and generate a comprehensive feature vector.

8. The image processing system for identifying anthocyanin content in colored potatoes according to claim 7, characterized in that: The step S45 specifically includes: Step S451: input the core data into a preset model; Step S452: automatically adjusting the internal weights and biases of the preset model to try different feature combinations; Step S453: Use the validation set to monitor the distinguishing ability of the feature combination in real time and record the optimal parameter configuration; Step S454: using a test set to check whether the preset model can accurately identify different target features; Step S455: If the recognition accuracy does not meet the standard, return to step S452 to readjust the parameters until the feature combination achieves the expected effect.

9. The image processing system for identifying anthocyanin content in colored potatoes according to claim 1, characterized in that: The prediction unit performs the following steps: Step S51: inputting the target comprehensive feature vector into a preset analysis model; Step S52: Automatically calculate and generate an objective function of anthocyanin content using the preset analysis model; Step S53: verify the actual test results of the sample and adjust the objective function parameters; Step S54: Calculate the anthocyanin concentration value of each pixel according to the objective function, and generate corresponding distribution heat map data using color depth; Step S55: Integrate the objective function and the heat map data into an outputtable result file.

10. The image processing system for identifying anthocyanin content in colored potatoes according to claim 9, characterized in that: The step S52 specifically includes: Step S521: taking the target comprehensive feature vector as input data; Step S522: automatically identifying the mathematical relationship between the input data and the anthocyanin content through a clustering algorithm; Step S523: Iteratively optimize the model parameters using known sample data to minimize the error between the predicted result and the actual value; Step S524: output the final anthocyanin content calculation formula; Step S525: Encapsulate the calculation formula into a directly callable function module for subsequent batch calculation or real-time prediction.

Images