Method and system for dynamically monitoring drying quality of fruits and vegetables by fusing chromatic aberration and shrinkage rate

Abstract

The invention discloses a method for dynamically monitoring the drying quality of fruits and vegetables by fusing chromatic aberration-shrinkage rate analysis. The method comprises the following steps: placing a fruit and vegetable sample on a rotatable objective table, drying according to a preset program, and carrying out multi-angle shooting by using a digital camera according to a timestamp to obtain an RGB color image; the method comprises the following steps: preprocessing by an image analysis system, converting an image into a Lab chromaticity space, calculating a chromatic aberration value delta E and a comprehensive shrinkage rate S, integrating data to generate a change curve, and constructing an appearance form model. In actual drying, real-time data and the model are compared, and early warning is carried out when the color difference value exceeds 3.5 or the shrinkage rate deviation exceeds 15% and image acquisition is abnormal. The drying equipment is provided with a high-precision electric rotary objective table, a specific annular light source array, a high-resolution camera and a standard calibration color card, and illumination and color measurement accuracy is ensured through calibration. Through multi-step operation, accurate and dynamic monitoring of fruit and vegetable drying quality is achieved, and drying process optimization is assisted.

Description

Technical Field

[0001] The present invention relates to the technical field of fruit and vegetable drying, and more specifically, to a method and system for dynamically monitoring the drying quality of fruits and vegetables by integrating color difference and shrinkage. Background Art

[0002] In the fruit and vegetable drying industry, the impact of the drying process on the appearance of fruits and vegetables is a key factor in determining product quality. With the growing market demand for high-quality dried fruits and vegetables, it is crucial to accurately understand the changes in appearance during the drying process.

[0003] Traditional drying process monitoring methods have serious shortcomings. For color measurement, the material is typically removed and analyzed using a color analyzer. This method not only interrupts the drying process and is cumbersome to operate, but also fails to track dynamic color changes during the drying process in real time. For morphological monitoring, traditional methods that rely on volume changes to measure material shrinkage, such as the vernier caliper method and the displacement method, are not only complex to operate but also difficult to guarantee measurement accuracy for small or irregularly shaped fruits and vegetables. While infrared scanning methods are simple to operate, the instrument is expensive, the scanning speed significantly affects the measurement results, and repeatability is poor.

[0004] More critically, current technology is unable to achieve dynamic monitoring of the fruit and vegetable drying process. During the actual drying process, due to the lack of effective dynamic monitoring methods, it is difficult to grasp the continuous changes in the appearance of fruits and vegetables, such as color and shrinkage, in real time, and it is impossible to promptly detect problems caused by improper drying conditions, such as excessive color changes, excessive shrinkage, or uneven drying. Moreover, existing technology is unable to provide early warning for abnormal signals that may appear during the drying process, such as abnormal drying quality due to drying equipment failure, sudden changes in environmental factors, and situations where the stacking and adhesion of fruits and vegetables affect the accuracy of image acquisition and analysis. This results in a lack of effective monitoring and adjustment of the drying process, resulting in a large number of fruits and vegetables having their quality degraded due to excessive or insufficient drying, seriously affecting product quality and economic benefits.

[0005] Prior art, such as the invention patent with publication number CN115372373A, proposes a method for rapidly measuring the appearance of fruits and vegetables during the drying process based on image analysis technology. This method uses a digital camera to capture images and perform color analysis, enabling rapid, non-destructive measurement and providing technical support for optimizing drying conditions. However, this method still has limitations. It fails to establish a comprehensive dynamic monitoring system, is unable to monitor the entire fruit and vegetable drying process in real time and comprehensively, and lacks an early warning mechanism for abnormal drying signals, making it difficult to meet the actual production needs for precise control of the drying process and ensuring product quality. Summary of the Invention

[0006] An object of the present invention is to solve at least the above problems and to provide at least the advantages which will be described hereinafter.

[0007] Another object of the present invention is to provide a dynamic monitoring method for fruit and vegetable drying quality that integrates color difference and shrinkage rate analysis. The method can quickly and non-destructively measure the appearance of fruits and vegetables during the drying process, realize dynamic monitoring and abnormality warning, and provide strong support for optimizing drying conditions and ensuring product quality.

[0008] In order to achieve these purposes and other advantages according to the present invention, a method for dynamic monitoring of fruit and vegetable drying quality integrating color difference and shrinkage rate analysis is provided, comprising: S1. Placing fruit and vegetable samples of the same type as the fruits and vegetables to be dried on a loading platform of a drying device and drying them according to a preset drying program. The loading platform is rotatable, and the drying device is provided with a digital camera facing the loading platform, and the digital camera is communicatively connected to a computer. S2. Setting a timestamp for the drying process, and using the digital camera to take real-time, multi-angle photos of the fruit and vegetable samples at each drying stage according to the timestamp to obtain a color image containing three channels of RGB; S3, using a computer image analysis system to pre-process the three-channel color image and extract the effective area of ​​the fruit and vegetable samples; based on a preset color space conversion algorithm, convert the RGB image into Lab color space and extract the lightness L of the fruit and vegetable sample area * Value, red and green axis a * Value, and yellow-blue axis b * value, and calculate its color difference value ΔE; based on multimodal image analysis and three-dimensional dynamic modeling, the comprehensive shrinkage rate S during the drying process of fruits and vegetables is determined; S4, the L of each drying stage * value, a * value, b * The color difference value ΔE and the comprehensive shrinkage rate S are integrated into time series data to generate the color change curve and the comprehensive shrinkage rate S change curve, and the appearance morphology model of the fruits and vegetables to be dried is obtained; S5. The fruits and vegetables to be dried are placed on a loading platform of the drying equipment and dried according to a box drying program. The digital camera is used to capture multi-angle images of the fruits and vegetables to be dried in real time during the drying process. The images are dynamically compared with the appearance morphology model of the fruits and vegetables to be dried obtained in step S4 to provide a real-time warning of abnormal drying signals.

[0009] Preferably, in step S5, the appearance morphology model of the fruits and vegetables to be dried is used for dynamic comparison, and the specific operation of real-time warning of drying abnormality signal is as follows: L calculated in real time during the drying process of the fruits and vegetables to be dried is used as the * value, a * value, b *The color difference value ΔE and the comprehensive shrinkage rate S are matched with the time series data in the appearance morphology model. When it is detected that the color difference value ΔE exceeds the preset threshold of 3.5 or the deviation between the measured value and the predicted value of the comprehensive shrinkage rate S exceeds 15%, a drying abnormality alarm is triggered. At the same time, the stacking / adhesion status of fruits and vegetables is judged based on image feature recognition technology. If the coverage rate of the effective detection area is less than 90%, an image acquisition quality alarm is issued simultaneously, and the time node of the abnormality and the quantitative deviation parameters are marked on the human-computer interface.

[0010] Preferably, the specific process of calculating the color difference value ΔE in step S3 is as follows: Based on the CIE76 color difference standard, the three-channel RGB image is converted into the Lab color space, and the average chromaticity value L0 of the fruit and vegetable sample area at the initial drying stage, that is, at t=0, is used as the color space. * , a0 * , and b0 * As a benchmark, calculate the color difference value ΔE at time t according to formula (1) t : (1); Among them, L t * 、a t * , and b t * is the mean chromaticity of the fruit and vegetable sample area in the current drying stage.

[0011] Preferably, in step S3, the specific process of determining the comprehensive shrinkage rate S during the fruit and vegetable drying process based on multimodal image analysis and three-dimensional dynamic modeling is as follows: S31. Obtain the 2D projection area of ​​the fruit and vegetable sample in the X, Y, and Z directions by orthogonal projection method, and calculate the projection area ratio A at each moment: ; S32, using structured light scanning and binocular vision fusion algorithm, by matching multi-angle image feature points to generate a 3D point cloud model, calculate the 3D shrinkage rate V: ; S33, weighted fusion of the three-dimensional shrinkage rate V and the projected area ratio A to obtain a comprehensive shrinkage rate S: S = mA + nV; Among them, A0 is the initial projection area, A t is the projected area at time t; V0 is the initial volume, V t is the volume at time t; m is the weight coefficient of the projected area ratio A, and n is the weight coefficient of the three-dimensional shrinkage rate V. Based on the least squares fitting method, m is taken as 0.4 and n is taken as 0.6.

[0012] Preferably, the specific process of step S4 is: S41, Lab value L for each drying stage* 、a * and b * To perform standardization: ; S42. Normalize the comprehensive shrinkage rate S: ; 43. Generate color change curve and calculate comprehensive chromaticity change index C(t): Take drying time t as the horizontal axis and plot L * (t), a * (t), and b * (t) three change curves; Comprehensive chromaticity ; S44, generate shrinkage rate change curve and calculate shrinkage rate V s (t): With drying time t as the horizontal axis, draw the S(t) variation curve; Shrinkage rate ; S45. Establishing a color-time relationship model: ; S406: Establishing a morphology-time relationship model: ; Among them, X norm is the degree of deviation of the original data point X from the mean, μ X is the mean, σ X is the standard deviation, X is the original data point; S norm is the normalized value of the comprehensive shrinkage rate S, S norm ∈ [0,1]; S min is the minimum value of shrinkage rate; S max is the maximum value of shrinkage rate; L * (t) represents the brightness value of fruits and vegetables at drying time t; a * (t) represents the chromaticity value of fruits and vegetables on the red-green axis at drying time t; b * (t) represents the chromaticity value of fruits and vegetables on the yellow-blue axis at drying time t; g is the lightness change rate coefficient, g∈[0.1,5.0]; f is the red-green axis change amplitude coefficient, f∈[0.01,0.5]; j is the yellow-blue axis change amplitude coefficient j∈[0.1,3.0]; γ is the red-green axis change curve morphology index γ∈[0.5,2.0]; δ is the yellow-blue axis attenuation rate constant, δ∈[0.01,0.5]h -1 , h -1 is the reciprocal of each hour; k is the drying rate constant.

[0013] Preferably, step S4 further comprises the following steps: Take fruit and vegetable samples of the same type as the fruits and vegetables to be dried multiple times, repeat steps S1 to S4, and obtain multiple color change curves and multiple comprehensive shrinkage rate S change curves. Further standardize the multiple color change curves and multiple comprehensive shrinkage rate S change curves, specifically: S401, preprocessing the original curves obtained from multiple experiments, including outlier removal and Savitzky-Golay filtering smoothing; S402, selecting the curve with the highest signal-to-noise ratio as the reference curve, aligning the time axes of the curves using a dynamic time warping algorithm, and achieving uniform sampling point density using cubic spline interpolation; S403. The aligned curves are fused based on the Gaussian regression model to generate a standard curve and a 95% confidence interval. The kernel function is a radial basis function, and the length scale parameter is optimized by maximum likelihood estimation: The optimization objective function of Gaussian process regression is: ; The confidence interval calculation formula is: ; Among them, K is the kernel function matrix, y is the data collected at a time point, is the average value of the data at that time point. The optimization process uses the L-BFGS-B algorithm to iterate 10 to 50 times; N is the number of repeated experiments; M is the total number of observation data points, M = N × Ms, Ms is the number of observation data points in a single experiment; σ(t) is the standard deviation of each time point, .

[0014] Preferably, the method for dynamic monitoring of fruit and vegetable drying quality integrating color difference-shrinkage rate analysis further includes verifying the standard curve generated in S403: Randomly retain 1 to 2 experimental data as the validation set, calculate the dynamic time warping distance between the validation set curve and the standard curve, and if the distance exceeds the threshold, re-adjust the preprocessed data and repeat steps S401 to S404 until the dynamic time warping distance of all validation sets does not exceed the threshold.

[0015] Preferably, the specific operation of converting the RGB image into the Lab color space in step S3 is: Based on the color space conversion algorithm specified by the CIE standard, the RGB image is linearized to eliminate the influence of gamma correction, and then mapped to the CIE XYZ color space using the conversion matrix corresponding to the D65 standard light source. Finally, the color space containing lightness L is generated through nonlinear transformation and reference white point normalization. * Value, red and green axis a* Value, yellow-blue axis b * Lab color space image of the values.

[0016] Preferably, the drying equipment in step S1 is provided with: Electric rotating stage with angle scale, rotation positioning accuracy ≤ 0.1°; A ring-shaped light source array consisting of 12 groups of LED light sources, of which 6 groups are equipped with combined lenses to form structured light. The LED light sources with combined lenses are spaced apart from the LED light sources without combined lenses. The color temperature of each LED light source is 5500K ± 50K, the illumination is 2000lux ± 2%, and the illumination uniformity is ≥ 95%; Full-frame digital camera, equipped with a 50mm fixed-focus lens, object distance 30cm±0.5cm, resolution ≥50 megapixels; Standard 24-color calibration color chart, compliant with ISO 12641.

[0017] Preferably, step S1 further includes calibration of the drying equipment, specifically: S11. Place the standard 24-color calibration color card at the center of the stage and adjust the LED light source so that the illumination deviation between the center area and the edge area is ≤2%; S12. Set the camera white balance to the D65 standard light source and ensure that the measured values ​​of the white block area of ​​the standard 24-color calibration color chart meet the following requirements: .

[0018] The present invention further claims protection for a monitoring system for the method for dynamically monitoring the drying quality of fruits and vegetables by integrating color difference and shrinkage rate, comprising: The image processing module is configured to: receive the multi-angle RGB image sequence of the fruit and vegetable drying process transmitted by the external image acquisition device, pre-process the RGB image and extract the effective area of ​​the fruit and vegetable, and perform RGB-Lab color space conversion to output the color parameter L * value, a * value, b * value and color difference value ΔE; generate a 3D point cloud model based on the multi-angle image sequence, calculate the projection area ratio A and 3D shrinkage rate V through orthogonal projection and volume reconstruction algorithm, and fuse the output of the comprehensive shrinkage rate S; The dynamic modeling module is configured to: transform the chromaticity parameter L * value, a * value, b * The values, color difference values ​​ΔE, and combined shrinkage rate S are integrated into a time series data set according to timestamps; the standard color change curve and the comprehensive shrinkage rate S change curve are generated based on Gaussian process regression, and an appearance morphology prediction model with confidence intervals is established; a composite drying kinetics model is embedded in the color-time relationship function and the morphology-time relationship function; The monitoring and early warning module is configured to: receive real-time colorimetric and shrinkage data from the current drying process, and perform dynamic time-warping matching with the appearance morphology prediction model; trigger a drying anomaly signal when the color difference value ΔE exceeds a threshold of 3.5 or the deviation between the measured and predicted values ​​of the comprehensive shrinkage rate S is greater than 15%, and generate a graded alarm based on the deviation parameters; perform edge contour analysis on the input image sequence, and when the fruit and vegetable area coverage is less than 90%, feedback is given indicating image quality anomaly and the data resampling protocol is activated; The burnt visualization module is configured to: provide an abnormal signal timeline marking interface, allowing users to review historical data and adjust drying program parameters; and integrate a calibration interface to receive standard color card image data to update colorimetric conversion parameters, ensuring that the ΔE calculation error is ≤1.5%.

[0019] The present invention has at least the following beneficial effects: First, the present invention installs a digital camera within the drying equipment to take photos of fruit and vegetable samples from multiple angles according to timestamps. Combined with an image analysis system and an established appearance morphology model, data can be collected and compared in real time. A drying anomaly alarm is triggered when the color difference value ΔE exceeds a preset threshold of 3.5 or when the measured value of the comprehensive shrinkage rate S deviates from the predicted value by more than 15%. Using image feature recognition technology, if the effective detection area coverage rate falls below 90%, an image acquisition quality alarm is simultaneously issued, and the abnormal time and quantitative deviation parameters are marked. This enables dynamic monitoring of the drying process and timely warning of abnormal situations, facilitating operators to adjust drying conditions in a timely manner to ensure product quality. Secondly, the present invention converts the RGB image into Lab color space based on the preset color space conversion algorithm, which can accurately extract the brightness L of the fruit and vegetable sample area. * Value, red and green axis a * Value, yellow-blue axis b * The color difference ΔE is calculated to accurately reflect the color change. The orthogonal projection method and structured light scanning are combined with binocular vision fusion algorithm, and the comprehensive shrinkage rate S is calculated in combination with weighted fusion. Compared with traditional methods, it can more comprehensively and accurately measure the shrinkage of fruits and vegetables during the drying process, providing reliable data support for studying the impact of the drying process on the appearance of fruits and vegetables. Third, the present invention further integrates the Lab values, color difference values ​​ΔE, and comprehensive shrinkage rate S at each drying stage into time series data, generates color change curves and comprehensive shrinkage rate S change curves, and establishes color-time relationship models and morphology-time relationship models. These models can intuitively display the changes in the appearance of fruits and vegetables over time during the drying process, provide guidance for drying similar fruits and vegetables, provide a scientific basis for optimizing the drying process, and help improve drying efficiency and product quality. Fourth, the present invention standardizes multiple color change curves and comprehensive shrinkage rate S curves through repeated experiments, including outlier removal, filtering and smoothing, time axis alignment, sampling point density standardization, and curve fusion, to generate a standard curve and 95% confidence interval. The standard curve is also verified to ensure the reliability and universality of the experimental data, making the research results more convincing and providing a reference standard and method for similar fruit and vegetable drying research under different conditions.

[0020] Other advantages, objectives and features of the present invention will be reflected in part from the following description and will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The figure is a flow chart of the method for dynamic monitoring of fruit and vegetable drying quality integrating color difference and shrinkage rate analysis according to the present invention. DETAILED DESCRIPTION

[0022] The present invention will be described in further detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0023] It should be understood that terms such as “having”, “including” and “comprising” used herein do not exclude the existence or addition of one or more other elements or combinations thereof.

[0024] like Figure 1 The present invention provides a method for dynamically monitoring the drying quality of fruits and vegetables by integrating color difference and shrinkage rate analysis, comprising: S1. Placing fruit and vegetable samples of the same type as the fruits and vegetables to be dried on a loading platform of a drying device and drying them according to a preset drying program. The loading platform is rotatable, and the drying device is provided with a digital camera facing the loading platform, and the digital camera is communicatively connected to a computer. S2. Setting a timestamp for the drying process, and using the digital camera to take real-time, multi-angle photos of the fruit and vegetable samples at each drying stage according to the timestamp to obtain a color image containing three channels of RGB; S3, using a computer image analysis system to pre-process the three-channel color image and extract the effective area of ​​the fruit and vegetable samples; based on a preset color space conversion algorithm, convert the RGB image into Lab color space and extract the lightness L of the fruit and vegetable sample area * Value, red and green axis a * Value, and yellow-blue axis b * value, and calculate its color difference value ΔE; based on multimodal image analysis and three-dimensional dynamic modeling, the comprehensive shrinkage rate S during the drying process of fruits and vegetables is determined; S4, the L of each drying stage * value, a* value, b * The color difference value ΔE and the comprehensive shrinkage rate S are integrated into time series data to generate the color change curve and the comprehensive shrinkage rate S change curve, and the appearance morphology model of the fruits and vegetables to be dried is obtained; S5. The fruits and vegetables to be dried are placed on a loading platform of the drying equipment and dried according to a box drying program. The digital camera is used to capture multi-angle images of the fruits and vegetables to be dried in real time during the drying process. The images are dynamically compared with the appearance morphology model of the fruits and vegetables to be dried obtained in step S4 to provide a real-time warning of abnormal drying signals.

[0025] In the above technical solution, a drying device with a rotatable stage is selected. A light source and a full-frame digital camera are installed within the drying device. The full-frame digital camera is equipped with a 50mm fixed-focus lens, with an adjustable object distance of 30cm±0.5cm to ensure a resolution of ≥50 megapixels. Calibration using a 24-color calibration color chart is required before drying begins. Fruits and vegetables similar to the ones to be dried (e.g., apples) are placed on the drying device's stage. The device is started according to a preset drying program, and drying process timestamps are set (e.g., every 5 minutes). At each timestamp, the rotatable stage is used to capture images from multiple angles (e.g., 0°, 45°, and 90°). Alternatively, the stage is rotated starting at 0° and in increments of 20-30° to complete a 360° circular image sequence. Before each shot, contrast focus is performed at the center of the corresponding angle of the sample to ensure image clarity meets requirements. A three-channel RGB color image is acquired and transmitted to a computer in real time. The collected three-channel color images were preprocessed (such as filtering, denoising, etc.) using a computer image analysis system to remove noise and interference information in the image, enhance image quality, and extract the effective area of ​​fruits and vegetables. Based on the color space conversion algorithm specified by the CIE standard, the RGB image was linearized to eliminate the influence of gamma correction, and then mapped to the CIE XYZ color space using the conversion matrix corresponding to the D65 standard light source. Finally, a color space containing lightness L was generated through nonlinear transformation and reference white point normalization. * Value, red and green axis a * Value, and yellow-blue axis b * The Lab color space image of the fruit and vegetable area is used as the benchmark for the average color value of the fruit and vegetable area at the initial drying stage (t=0). The color difference value ΔE at time t is calculated. At the same time, the comprehensive shrinkage rate S of the fruit and vegetable during the drying process is determined based on multimodal image analysis and three-dimensional dynamic modeling.

[0026] In the above technical solution, the L * value, a * value, b * The values, color difference values ​​ΔE and comprehensive shrinkage rate S are integrated into time series data, and the L* value, a * value, b * After normalizing the comprehensive shrinkage rate, the drying time t is used as the horizontal axis to plot L * (t), a * (t) and b * (t) three color change curves and calculate the comprehensive color change index C(t). Plot the S(t) shrinkage rate change curve and calculate the shrinkage rate, and establish the color-time relationship model and the morphology-time relationship model.

[0027] In the above technical solution, the fruits and vegetables to be dried are placed on the loading platform of the drying equipment and dried according to the same drying procedure. During the drying process, multi-angle images are collected in real time by a digital camera and transmitted to the computer. The online appearance morphology model (color-time relationship model, morphology-time relationship model, etc.) is used to calculate the L * value, a * value, b * The color difference value, ΔE, and comprehensive shrinkage rate S are compared with the time series data in the model. If the color difference value ΔE exceeds the preset threshold of 3.5 or the comprehensive shrinkage rate S deviates from the measured value by more than 15%, a drying abnormality alarm is triggered. At the same time, image feature recognition technology is used to determine whether the apple slices are stacked or stuck together. If the effective detection area coverage rate is less than 90%, an image acquisition quality alarm is issued, and the time point of the abnormality and the quantitative deviation parameters are marked on the human-computer interface.

[0028] The above technical solution has at least the following beneficial effects: 1. The use of multi-angle shooting combined with multimodal image analysis and three-dimensional dynamic modeling can fully obtain the appearance information of fruits and vegetables during the drying process, comprehensively considering the two key indicators of color and shrinkage, and more accurately reflect the changes in the drying quality of fruits and vegetables compared with the traditional single indicator monitoring method; 2. By setting timestamps for real-time shooting and dynamic comparison, the whole drying process can be dynamically monitored, and abnormal changes in the drying process can be discovered in time, providing a basis for timely adjustment of drying conditions, avoiding quality degradation due to excessive or insufficient drying, and improving product quality stability; by setting the threshold value of color difference value and shrinkage rate, and The judgment criteria for image acquisition quality and timely alarm when abnormalities occur help operators take quick measures to reduce losses, ensure the smooth progress of the drying process and the quality of the product; 4. By establishing an appearance morphology model and analyzing the changes in color and wrinkle over time, data support is provided for optimizing drying process parameters, which helps to improve drying efficiency, reduce energy consumption, and enhance the economic benefits of the enterprise; 5. The above technical solution is suitable for drying quality monitoring of various fruits and vegetables. It only needs to adjust the drying procedures and related parameters according to the characteristics of different fruits and vegetables. It has strong versatility and adaptability and can be widely used in the fruit and vegetable drying processing industry.

[0029] In one of the technical solutions, in step S5, the appearance morphology model of the fruits and vegetables to be dried is used for dynamic comparison, and the specific operation of real-time warning of drying abnormality signals is as follows: L calculated in real time during the drying process of the fruits and vegetables to be dried is used to compare the appearance morphology model of the fruits and vegetables to be dried. * value, a * value, b * The color difference value ΔE and the comprehensive shrinkage rate S are matched with the time series data in the appearance morphology model. When it is detected that the color difference value ΔE exceeds the preset threshold of 3.5 or the deviation between the measured value and the predicted value of the comprehensive shrinkage rate S exceeds 15%, a drying abnormality alarm is triggered. At the same time, the stacking / adhesion status of fruits and vegetables is judged based on image feature recognition technology. If the coverage rate of the effective detection area is less than 90%, an image acquisition quality alarm is issued simultaneously, and the time node of the abnormality and the quantitative deviation parameters are marked on the human-computer interface.

[0030] In the above technical solution, during the drying process of the fruits and vegetables to be dried, a digital camera is used to capture images from multiple angles in real time, using the same time intervals and image acquisition methods as when establishing the appearance morphology model. After acquisition, the images are transferred to a computer and the same image processing and analysis methods as those used in the model establishment phase are applied, namely, image preprocessing is first performed, including denoising, enhancement, segmentation, and other operations; then, the RGB image is converted to Lab color space based on a preset color space conversion algorithm, and the lightness L of the fruit and vegetable sample area is extracted. * Value, red and green axis a * Value, and yellow-blue axis b * The values ​​were obtained and the color difference ΔE was calculated; the comprehensive shrinkage rate S was determined by multimodal image analysis and three-dimensional dynamic modeling.

[0031] Further, the L calculated in real time * value, a * value, b * The color difference ΔE, as well as the comprehensive shrinkage rate S, are compared one-to-one with the time series data from the appearance morphology model at the corresponding time points. For example, if the appearance morphology model records data for the 30th minute of drying, the real-time data at that time is compared when real-time monitoring indicates that drying has reached the 30th minute. A preset threshold for the color difference ΔE is set to 3.5. When the real-time calculated color difference ΔE exceeds this threshold, an abnormal drying alarm is triggered. Alarms can be implemented in various ways, such as displaying a prompt window on the computer interface with a message such as "Abnormal color changes of fruits and vegetables during drying. Please check drying conditions." Alternatively, an audible alarm can be configured to play a specific warning tone to attract the operator's attention. For the comprehensive shrinkage rate S, the deviation between the measured value and the predicted value from the appearance morphology model is calculated. If the deviation exceeds 15%, a drying abnormality alarm is also triggered, indicating that the drying speed may be too fast or too slow, affecting the quality of the fruit and vegetable.

[0032] To identify the stacking / sticking of fruits and vegetables (or fruit and vegetable slices, such as apple slices) and to alert on image acquisition quality, image feature recognition technology is used to determine the stacking / sticking status of fruits and vegetables. First, image segmentation is used to separate the fruits and vegetables from the background. Then, the boundaries and contact areas between the fruits and vegetables are analyzed. For example, if the boundaries between adjacent fruits and vegetables are blurred or there are large areas of overlap, it can be determined that stacking or sticking has occurred. Simultaneously, the effective detection area coverage is calculated—the ratio of the detectable area of ​​the fruits and vegetables to the entire image acquisition area. If the effective detection area coverage falls below 90%, an image acquisition quality alert is issued. Similar to the drying anomaly alert, the human-machine interface displays a prominent message, such as "Image acquisition quality abnormality. Fruits and vegetables may be stacked / sticking, affecting monitoring accuracy." The time of the anomaly is also noted, such as "Anomaly occurred at the 45th minute of drying." Deviation parameters are also quantified, such as the specific value by which the current color difference value ΔE exceeds the threshold and the specific percentage deviation between the measured and predicted values ​​of the comprehensive shrinkage rate S, to help operators understand the severity of the anomaly.

[0033] In the above technical solution, by real-time monitoring of the color difference ΔE and the comprehensive shrinkage rate S, deviations from the normal range during the drying process can be promptly detected. Excessive color difference indicates abnormal color changes in fruits and vegetables, possibly due to inappropriate drying conditions such as temperature and humidity, leading to increased enzymatic or non-enzymatic browning, affecting the product's appearance and taste. Excessive deviation in the comprehensive shrinkage rate indicates abnormal morphological changes in fruits and vegetables, potentially resulting in uneven drying or overdrying, affecting the product's texture and rehydration. Promptly triggering an alarm allows operators to adjust drying conditions, helping to ensure stable and consistent product quality. Furthermore, by assessing the stacking and sticking of fruits and vegetables and monitoring image acquisition quality, the accuracy of image analysis can be prevented from being affected by abnormal fruit and vegetable placement. Stacking or sticking of fruits and vegetables can prevent accurate color and morphological information from being captured in certain areas, causing deviations in the calculated color difference ΔE and comprehensive shrinkage rate S, which can affect the judgment of the drying process. When abnormal image acquisition quality is detected, an alarm is issued to prompt operators to adjust the fruit and vegetable placement, ensuring the reliability of the monitoring data. Abnormality alerts and quantitative parameter annotations enable operators to quickly identify problems and take appropriate action. Compared to periodic manual inspections of the drying process, this real-time, automatic early warning mechanism significantly shortens the time to problem detection and reduces product losses and production delays caused by drying anomalies. For example, if an alarm indicates an abnormal overall shrinkage rate, the operator can immediately check the drying equipment's ventilation, heating, and other systems and adjust parameters promptly to avoid over-drying or under-drying, thereby improving production efficiency and reducing costs.

[0034] In one of the technical solutions, the specific process of calculating the color difference value ΔE in step S3 is as follows: Based on the CIE76 color difference standard, the three-channel RGB image is converted into the Lab color space, and the average chromaticity value L0 of the fruit and vegetable sample area at the initial drying stage, that is, at t=0, is used as the color space. * , a0 * , and b0 * As a benchmark, calculate the color difference value ΔE at time t according to formula (1) t : (1); Among them, L t * 、a t * , and b t * is the mean chromaticity of the fruit and vegetable sample area in the current drying stage.

[0035] In the above technical solution, in the monitoring of fruit and vegetable drying quality, by calculating the color difference value based on the CIE76 color difference standard, it is possible to effectively evaluate the color changes of fruits and vegetables during the drying process, ensure product quality, provide strong support for optimizing the drying process, and guide the precise control of the drying process, thereby improving the quality stability and consistency of dried fruit and vegetable products. By calculating ΔE in real time during the drying process and comparing it with the preset threshold, real-time monitoring of the drying process and abnormality warnings can be achieved. When ΔE exceeds the threshold, the system can promptly issue an alarm, prompting the operator to take measures to avoid product quality degradation due to excessive color changes. In large-scale production, this function can effectively improve production efficiency and reduce scrap rates.

[0036] In one technical solution, the specific process of determining the comprehensive shrinkage rate S during the fruit and vegetable drying process based on multimodal image analysis and three-dimensional dynamic modeling in step S3 is as follows: S31. Obtain the 2D projection area of ​​the fruit and vegetable sample in the X, Y, and Z directions by orthogonal projection method, and calculate the projection area ratio A at each moment: ; S32, using structured light scanning and binocular vision fusion algorithm, by matching multi-angle image feature points to generate a 3D point cloud model, calculate the 3D shrinkage rate V: ; S33, weighted fusion of the three-dimensional shrinkage rate V and the projected area ratio A to obtain a comprehensive shrinkage rate S: S = mA + nV; Among them, A0 is the initial projection area, A t is the projected area at time t; V0 is the initial volume, V t is the volume at time t; m is the weight coefficient of the projected area ratio A, and n is the weight coefficient of the three-dimensional shrinkage rate V. Based on the least squares fitting method, as a preferred option, m is 0.4 and n is 0.6.

[0037] In the above technical solution, before the drying experiment begins, a high-precision imaging device is used to photograph the fruit and vegetable samples from three mutually orthogonal directions, X, Y, and Z, to obtain a two-dimensional image in the initial state, and the projection area at this time is recorded as A0. At each time point t during the drying process, the fruit and vegetable samples are photographed again from these three directions under the same shooting conditions to obtain the projection area A at the corresponding time. tTo ensure image quality, it is necessary to ensure uniform and stable lighting, a fixed camera position and accurate calibration during shooting to avoid image deformation or blurring due to external factors. The collected images are pre-processed, including denoising (such as using Gaussian filtering to remove random noise in the image), contrast enhancement (such as using histogram equalization technology to improve image clarity), and image segmentation (using threshold segmentation, edge detection and other algorithms to extract the contours of fruit and vegetable samples). Using image processing software or writing code (such as using Python's OpenCV library), by counting the number of pixels within the sample contour and combining the camera's calibration parameters (such as the ratio of pixels to actual size), the number of pixels is converted into the actual projected area value. Finally, according to The projection area ratio A at each moment is calculated to reflect the size changes of fruits and vegetables on the two-dimensional plane over time.

[0038] In the above technical solution, during the drying experiment, multiple cameras were used to simultaneously capture fruit and vegetable samples from different angles, establishing a multi-angle visual acquisition system. Simultaneously, a structured light projector was used to project structured light with a specific pattern (such as stripes or Gray code) onto the fruit and vegetable samples. Structured light provides more texture information for the images, facilitating subsequent feature point matching and 3D reconstruction. Feature extraction algorithms (such as SIFT, SURF, or ORB) were used to process the multi-angle images and extract feature points. Then, a feature point matching algorithm (such as a descriptor-based matching method, which calculates feature point descriptors and performs a similarity measurement) accurately matched identical feature points in different images and established correspondences between them. Based on binocular vision principles and information acquired from structured light scanning, the 3D coordinates of each feature point were calculated. A 3D point cloud model of the fruit and vegetable sample was constructed using the 3D coordinates of these numerous feature points. This process requires precise camera calibration to obtain intrinsic parameters (such as focal length and principal point coordinates) and extrinsic parameters (such as rotation matrices and translation vectors) to ensure the accuracy of the 3D coordinate calculations. The generated 3D point cloud model is processed by using a voxel-based method to divide the 3D space into small voxel units. The volume is estimated by counting the number of voxels containing the fruit and vegetable samples; or the volume is calculated after building a surface model using the Delaunay triangulation algorithm. The volume at the initial moment V0 and the volume at time t V are calculated. t , and according to the formula Calculate the three-dimensional shrinkage rate V to quantify the volume change of fruits and vegetables in three-dimensional space. Using the least squares fitting method, determine the weight coefficients m = 0.4 and n = 0.6. Substitute the previously calculated projected area ratio A and the three-dimensional shrinkage rate V into the formula S = mA + nV to calculate the comprehensive shrinkage rate S.

[0039] In the above-mentioned technical solution, a single projection area ratio only reflects the dimensional changes of fruits and vegetables in the two-dimensional plane, ignoring changes in depth. Relying solely on the three-dimensional shrinkage ratio may not accurately reflect local deformation due to measurement errors or model inaccuracies. Combining and weightedly fusing the two methods, the comprehensive shrinkage ratio S can comprehensively capture the shrinkage information of fruits and vegetables during the drying process in both two and three dimensions, more accurately reflecting the actual shrinkage degree of fruits and vegetables, and providing more precise data support for studying the impact of the drying process on fruit and vegetable morphology. The structured light scanning and binocular vision fusion algorithm combines the high-precision measurement of structured light with the multi-angle observation advantages of binocular vision. Through multi-angle image acquisition and feature point matching, it can more accurately obtain three-dimensional information of fruits and vegetables, reduce measurement errors, and improve the reliability of three-dimensional shrinkage ratio calculation. Furthermore, the orthogonal projection method is relatively simple and stable for two-dimensional projection area measurement. The two complement each other, making the calculation of the comprehensive shrinkage ratio S more reliable and stable, and enhancing the anti-interference ability of the entire measurement method. Different fruits and vegetables have different shapes, and traditional single measurement methods cannot accurately measure their shrinkage during the drying process. The technical solution provided by this invention, through multimodal image analysis and 3D dynamic modeling, can accommodate fruits and vegetables of various complex shapes. Whether regular or irregular in shape, the method can capture image information from multiple angles, construct a 3D model, and calculate the projected area and volume change, thereby accurately determining the comprehensive shrinkage rate. This method offers strong versatility and adaptability.

[0040] In one of the technical solutions, the specific process of step S4 is: S41, Lab value L for each drying stage * 、a * and b * To perform standardization: ; S42. Normalize the comprehensive shrinkage rate S: ; S43. Generate a color change curve and calculate the comprehensive chromaticity change index C(t): Take drying time t as the horizontal axis and plot L * (t), a * (t), and b * (t) three change curves; Comprehensive chromaticity ; S44, generate shrinkage rate change curve and calculate shrinkage rate V s (t): With drying time t as the horizontal axis, draw the S(t) variation curve; Shrinkage rate ; S45. Establishing a color-time relationship model: ; S406: Establishing a morphology-time relationship model: ; Among them, X norm is the degree of deviation of the original data point X from the mean, μ X is the mean, σ X is the standard deviation, X is the original data point; S norm is the normalized value of the comprehensive shrinkage rate S, S nor ∈ [0,1]; S min is the minimum value of shrinkage rate; S max is the maximum value of shrinkage rate; L * (t) represents the brightness value of fruits and vegetables at drying time t; a * (t) represents the chromaticity value of fruits and vegetables on the red-green axis at drying time t; b * (t) represents the chromaticity value of fruits and vegetables on the yellow-blue axis at drying time t; g is the lightness change rate coefficient, g∈[0.1,5.0]; f is the red-green axis change amplitude coefficient, f∈[0.01,0.5]; j is the yellow-blue axis change amplitude coefficient j∈[0.1,3.0]; γ is the red-green axis change curve morphology index γ∈[0.5,2.0]; δ is the yellow-blue axis attenuation rate constant, δ∈[0.01,0.5]h -1 , h -1 is the reciprocal of each hour; k is the drying rate constant.

[0041] In the above technical solution, the Lab values ​​(L * value, a * value, b * value) data, calculate all L * Mean of the values and standard deviation , , for each L * The value is standardized, and the same is true for a * value, b * The value is also standardized. Statistically analyze the data of the comprehensive shrinkage rate S and find the maximum value S max and minimum value S min , according to the formula Normalize each S value so that the value range of S is between [0,1], which is convenient for subsequent comparison and analysis. * (t), a * (t), and b * (t) The data corresponding to the time point are plotted separately.* (t), a * (t), and b * (t) three changing curves. At the same time, according to the formula The comprehensive chromaticity change index C(t) is calculated to quantify the comprehensive degree of color change during the drying process. With the drying time t as the horizontal axis, the normalized comprehensive shrinkage rate S(t) data corresponding to the time point is plotted as the shrinkage rate change curve. At the same time, according to the formula Calculate the shrinkage rate. Set the brightness change rate coefficient g (g∈[0.1,5.0]), the red-green axis change amplitude coefficient f (f∈[0.01,0.5]), the red-green axis change curve morphology index γ (γ∈[0.5,2.0]), the yellow-blue axis change amplitude coefficient j (j∈[0.1,3.0]), the yellow-blue axis attenuation rate constant δ (δ∈[0.01,0.5]h -1 ), combined with the Lab value of the initial drying stage to establish a color-time relationship model, by fitting the actual data, constantly adjust the parameters to make the model more consistent with the actual color change, or adjust the parameters through multiple groups of experiments. Set the drying rate constant k, combined with the maximum value S from the comprehensive week max , a morphology-time relationship model is established, where the k value can be determined by the least squares method, so that the model can accurately describe the morphological changes during the drying process of fruits and vegetables. Among them, the values ​​of the color change rate coefficient g, the red-green axis change amplitude coefficient f, the red-green axis change curve morphology index γ, the yellow-blue axis change amplitude coefficient j, and the yellow-blue axis decay rate constant δ can be determined based on the experience of long-term drying experiments. Alternatively, the least squares method and nonlinear regression algorithm can be used to substitute the data into the color-time relationship model. By adjusting the parameters to minimize the error between the model prediction value and the experimental data, reasonable values ​​of each parameter can be obtained. Alternatively, the curve_fit function of the Scipy library can be used in Python to input the experimental data and the model function, and perform nonlinear regression fitting to obtain reasonable values ​​of each parameter.

[0042] In the above technical solution, the Lab values ​​and comprehensive shrinkage rates are standardized and normalized, eliminating dimensional differences between different data and making the data comparable. The standardized Lab values ​​can more intuitively reflect the relative position and change trend of each color index in the overall data. The normalized comprehensive shrinkage rate unifies the shrinkage rates of different experimental conditions or different fruits and vegetables into the range of [0,1], facilitating the comparison of the degree of shrinkage during different drying processes. At the same time, by plotting color change curves and shrinkage rate change curves, the present invention can intuitively display the changes in color and shrinkage rate over time during the drying process, helping researchers quickly understand the changes in the appearance and morphology of fruits and vegetables during the drying process. Calculating the comprehensive chromaticity change index and shrinkage rate further quantifies the degree and speed of color and shrinkage changes, providing a quantitative basis for accurate analysis of the drying process. Finally, by establishing a color-time relationship model and a morphology-time relationship model, the present invention can mathematically describe the changes in color and morphology over time during the drying process. The model not only helps to understand the intrinsic mechanism of the drying process, but can also be used to predict the color and morphological changes of fruits and vegetables under different drying times, providing theoretical support for optimizing the drying process and controlling the drying process, thereby improving the controllability of the drying process and the stability of product quality.

[0043] In one of the technical solutions, step S4 further includes the following steps: Take fruit and vegetable samples of the same type as the fruits and vegetables to be dried multiple times, repeat steps S1 to S4, and obtain multiple color change curves and multiple comprehensive shrinkage rate S change curves. Further standardize the multiple color change curves and multiple comprehensive shrinkage rate S change curves, specifically: S401, preprocessing the original curves obtained from multiple experiments, including outlier removal and Savitzky-Golay filtering smoothing; S402, selecting the curve with the highest signal-to-noise ratio as the reference curve, aligning the time axes of the curves using a dynamic time warping algorithm, and achieving uniform sampling point density using cubic spline interpolation; S403. The aligned curves are fused based on the Gaussian regression model to generate a standard curve and a 95% confidence interval. The kernel function is a radial basis function, and the length scale parameter is optimized by maximum likelihood estimation: The optimization objective function of Gaussian process regression is: ; The confidence interval calculation formula is: ; Among them, K is the kernel function matrix, y is the data collected at a time point, is the average value of the data at that time point. The optimization process uses the L-BFGS-B algorithm to iterate 10 to 50 times; N is the number of repeated experiments; M is the total number of observation data points, M=N×Ms, Ms is the number of observation data points in a single experiment; σ(t) is the standard deviation of each time point, .

[0044] In the above technical solution, multiple fruit and vegetable samples of the same type as the fruits and vegetables to be dried are prepared. The size and maturity of the samples are as consistent as possible with the fruits and vegetables to be dried. The steps S1 to S4 are strictly followed for each sample. During the drying process, a digital camera is used to take real-time multi-angle photos of the fruit and vegetable samples at each drying stage according to the set timestamp to obtain a color image containing three channels of RGB. The obtained image is processed to extract the Lab value (L * value, a * value, b * The color difference (ΔE), color difference (ΔE), and comprehensive shrinkage (S) values ​​were integrated into time series data to generate multiple color change curves and comprehensive shrinkage (S) change curves. The raw curve data obtained from multiple experiments were individually examined. If, at a specific point in time, a specific comprehensive shrinkage (S) value significantly deviated from the other data, this could be due to unforeseen factors during the experiment (such as equipment failure or improper sample placement). These outliers were identified and removed using statistical methods (such as the 3σ principle, which considers data points that deviate more than three standard deviations from the mean) to ensure data accuracy and reliability. The Savitzky-Golay filter was used to smooth the curves after removing outliers. This filter, based on the least squares principle, smoothes the data by fitting a polynomial. The signal-to-noise ratio (SNR) of each curve was then calculated. The SNR can be calculated as the ratio of signal intensity to noise intensity (the specific calculation method depends on the actual data characteristics). The curve with the highest SNR was selected as the reference curve, as it is least affected by noise and best reflects the true trend. Taking the reference curve as the benchmark, the DTW algorithm is used to align the time axis of other curves. The DTW algorithm makes the two curves as similar as possible in shape by finding the optimal matching path in the time series, thereby achieving time axis alignment. The cubic spline interpolation method is used to unify the sampling point density of the aligned curves. Cubic spline interpolation ensures the continuity of the function value, first-order derivative and second-order derivative at each node by constructing a piecewise cubic polynomial, thereby achieving smooth interpolation of the data. Then, the aligned curves with unified sampling point density are fused based on the Gaussian regression model. First, the kernel function matrix K is constructed based on the given radial basis function (RBF) as the kernel function, and the optimization objective function of the Gaussian process regression is used. , using the L-BFGS-B algorithm for iterative optimization, setting the number of iterations to 10~50 times. After obtaining the optimal length scale parameter and other model parameters through optimization, the standard curve is obtained according to the model prediction. At the same time, according to the confidence interval calculation formula Calculate 95% confidence intervals.

[0045] In this technical solution, outlier removal eliminates erroneous data potentially caused by experimental error, equipment failure, and other factors, ensuring the authenticity and reliability of the curve data. Savitzky-Golay filtering and smoothing reduce noise in the data, allowing the curve to more accurately reflect the true trends during the fruit and vegetable drying process, providing a reliable data foundation for subsequent analysis. The curve with the highest signal-to-noise ratio is selected as the reference curve, and a dynamic time warping algorithm is used to align the time axes of the curves. This eliminates time series inconsistencies caused by experimental manipulation and sample differences, making the curves obtained from different experiments comparable in the temporal dimension. Cubic spline interpolation standardizes the sampling point density, further ensuring curve consistency during data processing and analysis, facilitating comparison and fusion between curves. Curves are fused using a Gaussian regression model to generate a standard curve. This integrates data from multiple experiments to better represent the general trends in the drying process for this type of fruit and vegetable. Furthermore, calculation of a 95% confidence interval assesses the reliability of the standard curve and the degree of data dispersion, providing more comprehensive information for subsequent drying process monitoring and analysis. Researchers can judge the stability and reliability of experimental data based on the confidence interval. When new experimental data exceeds the confidence interval, it indicates that there may be abnormalities in the drying process or changes in experimental conditions, which helps to detect problems in a timely manner and take corresponding measures.

[0046] In one of the technical solutions, the dynamic monitoring method for fruit and vegetable drying quality integrating color difference and shrinkage analysis also includes verification of the standard curve generated by S403: Randomly retain 1 to 2 experimental data as the validation set, calculate the dynamic time warping distance between the validation set curve and the standard curve, and if the distance exceeds the threshold, re-adjust the preprocessed data and repeat steps S401 to S404 until the dynamic time warping distance of all validation sets does not exceed the threshold.

[0047] In the above technical solution, data from one or two experiments are randomly selected from multiple experimental data sets as a validation set. The randomness of the validation set avoids bias caused by artificial data selection. Next, a dynamic time warping algorithm is used to calculate the distance between the validation set curve and the standard curve. This algorithm can find the optimal matching path when time series data exhibits timeline stretching or distortion, thereby accurately measuring the similarity between the two curves. In specific implementation, the coordinate data of the validation set curve and the standard curve are input, and the algorithm calculates a distance value representing the difference between the two. A suitable threshold is also set as a judgment criterion. If the calculated dynamic time warping distance exceeds the threshold, it indicates that the validation set curve differs significantly from the standard curve, and the preprocessed data needs to be adjusted. Adjustment methods include rechecking the rationality of outlier removal, optimizing Savitzky-Golay filter parameters, realigning the timeline, or adjusting the interpolation method. Steps S401-S404 are then repeated, i.e., re-preprocessing, curve alignment, uniform sampling point density, curve fusion, and distance calculation are performed until the dynamic time warping distance of all validation sets does not exceed the threshold.

[0048] In the above technical solution, the standard curve is verified and adjusted through the validation set, which can ensure that the generated standard curve has high accuracy and representativeness. Only when the distance between the validation set curve and the standard curve is within a reasonable range can it be said that the standard curve can better reflect the color and shrinkage rate changes of similar fruits and vegetables during the drying process, thereby improving the accuracy and reliability of the entire monitoring method. Designing a validation set can also effectively identify and process anomalies in experimental data. When the validation set curve differs too much from the standard curve, by adjusting the preprocessing data and repeating the relevant steps, the model can be made more adaptable to the fruit and vegetable drying data of different batches and individuals, thereby enhancing the robustness of the model and reducing erroneous judgments caused by data fluctuations or anomalies. Only after multiple verifications and adjustments, so that the validation set and the standard curve achieve a good degree of matching, can it be ensured that the monitoring model established based on these data can be accurately applied to the actual dynamic monitoring of fruit and vegetable drying quality, providing reliable guidance for actual production.

[0049] In one of the technical solutions, the specific operation of converting the RGB image to the Lab color space in step S3 is: Based on the color space conversion algorithm specified by the CIE standard, the RGB image is linearized to eliminate the influence of gamma correction, and then mapped to the CIE XYZ color space using the conversion matrix corresponding to the D65 standard light source. Finally, the color space containing lightness L is generated through nonlinear transformation and reference white point normalization. * Value, red and green axis a * Value, yellow-blue axis b *The Lab color space representation is more consistent with human perception of color, enabling more accurate description and comparison of color differences. In monitoring the quality of fruit and vegetable drying, it can more intuitively reflect changes in fruit and vegetable color. The RGB color space is associated with specific devices (such as cameras and monitors), and RGB values ​​may vary between devices. CIE XYZ and Lab color spaces, on the other hand, are device-independent. Converting RGB images to Lab color space eliminates color errors caused by device differences, ensuring consistency and comparability in the colors of images collected and processed on different devices, thereby improving the accuracy and reliability of monitoring results.

[0050] In one technical solution, the drying equipment in step S1 is provided with: Electric rotating stage with angle scale, rotation positioning accuracy ≤ 0.1°; A ring-shaped light source array consisting of 12 groups of LED light sources, of which 6 groups are equipped with combined lenses to form structured light. The LED light sources with combined lenses are spaced apart from the LED light sources without combined lenses. The color temperature of each LED light source is 5500K ± 50K, the illumination is 2000lux ± 2%, and the illumination uniformity is ≥ 95%; Full-frame digital camera, equipped with a 50mm fixed-focus lens, object distance 30cm±0.5cm, resolution ≥50 megapixels; Standard 24-color calibration color chart, compliant with ISO 12641.

[0051] In the above technical solution, 12 groups of LED light sources are divided into two groups. At a certain time stamp, 6 groups of LED light sources without combined lenses are started. The stage rotates one circle, and the digital camera takes real-time multi-angle photos of the fruit and vegetable samples at each drying stage to obtain a color image containing three channels of RGB. The RGB image is converted into Lab color space, and the brightness L of the fruit and vegetable sample area is extracted. * Value, red and green axis a * Value, and yellow-blue axis b *The device then generates structured light using six LED light sources with integrated lenses. The stage rotates one full revolution, and a digital camera captures the fruit and vegetable samples at each drying stage in real time from multiple angles. Based on multimodal image analysis and 3D dynamic modeling, the overall shrinkage rate (S) of the fruit and vegetable samples during the drying process is measured. The rotatable function of the motorized rotating stage enables the drying device to accommodate fruit and vegetable samples of varying shapes and sizes, providing more comprehensive information through multi-angle capture. Furthermore, the use of structured light (provided by the LED light sources with integrated lenses) enables the measurement and analysis of the 3D morphology of fruits and vegetables, expanding the device's functionality and application range. Furthermore, the adjustable ring light array (brightness, light source combination, etc.) and the configurable camera capture parameters enable the device to adapt to diverse experimental conditions and requirements, such as different drying stages and fruit and vegetable varieties, enhancing its adaptability and flexibility.

[0052] In one technical solution, step S1 also includes calibration of the drying equipment, specifically: S11. Place the standard 24-color calibration color card at the center of the stage and adjust the LED light source so that the illumination deviation between the center area and the edge area is ≤2%; S12. Set the camera white balance to the D65 standard light source and ensure that the measured values ​​of the white block area of ​​the standard 24-color calibration color chart meet the following requirements: Through the calibration step, the image acquisition of the entire drying equipment is in a stable and repeatable working state, which not only ensures the uniformity of lighting, but also ensures the accuracy of color measurement, effectively reducing the error caused by the equipment itself.

[0053] The following is a specific embodiment of the present invention 1. Equipment calibration: A drying device equipped with an electric rotating stage with an angle scale is selected, and the stage rotation positioning accuracy is 0.1°. A ring light source array consisting of 12 groups of LED light sources is installed inside the drying device, of which 6 groups of LED light sources are installed with combined lenses to form structured light, and the combined lenses are installed at intervals with the uninstalled LED light sources. Ensure that the color temperature of any LED light source is stable at 5500K, the illumination is 2000lux and the error is controlled at ±2%, and the illumination uniformity reaches 95%. Install a full-frame digital camera with a 50mm fixed-focus lens, adjust the object distance to 30cm, and set the camera resolution to 50 million pixels. At the same time, place a standard 24-color calibration color card that complies with the ISO 12641 standard inside the drying device.

[0054] Place the standard 24-color calibration color card in the center of the stage and carefully adjust the LED light source to control the illumination deviation between the center area and the edge area within 2%. Set the camera white balance to D65 standard light source and measure the white block area of ​​the standard 24-color calibration color card to ensure the color difference value .

[0055] 2. Sample drying and data collection: Select multiple fruit and vegetable samples of the same type as the fruits and vegetables to be dried, place them on the loading platform, and start the drying equipment according to the preset drying program.

[0056] Set timestamps for the drying process, such as every 15 minutes. At each timestamp, use a digital camera to capture the fruit and vegetable samples from multiple angles, acquiring three-channel RGB color images. These images are then transferred to a computer connected to the camera. Six sets of LED light sources with integrated lenses are activated to generate structured light. The stage rotates one full revolution, and the digital camera captures the fruit and vegetable samples at each drying stage in real time from multiple angles. The images are then transferred to the computer.

[0057] 3. Image analysis and parameter calculation: The three-channel color images were preprocessed using a computer image analysis system, and edge detection, threshold segmentation and other algorithms were used to accurately extract the effective areas of the fruit and vegetable samples.

[0058] Based on the color space conversion algorithm specified by the CIE standard, the RGB image is linearized to eliminate the influence of gamma correction. Then, the conversion matrix corresponding to the D65 standard light source is used to map the image to the CIE XYZ color space. After nonlinear transformation and reference white point normalization, the Lab color space image is generated to extract the lightness L of the fruit and vegetable sample area. * Value, red and green axis a * Value, and yellow-blue axis b * The average chromaticity value L0 of the fruit and vegetable sample area at the initial drying stage (t=0) * , a0 * , and b0 * As a benchmark, according to Calculate the color difference value ΔE at the moment t .

[0059] The 2D projected areas of fruit and vegetable samples along the X, Y, and Z axes were obtained using orthogonal projection, and the projected area ratio A was calculated at each moment. A structured light scanning and binocular vision fusion algorithm was used to match multi-angle image feature points to generate a 3D point cloud model, and the 3D shrinkage ratio V was calculated. The 3D shrinkage ratio V and the projected area ratio A were weightedly fused (with m set to 0.4 and n set to 0.6) to obtain the comprehensive shrinkage ratio S (S = mA + nV).

[0060] 4. Data integration and model building After normalizing the Lab values ​​of each drying stage and the comprehensive shrinkage rate S, the L * (t), a * (t), and b * (t) three change curves and calculate the comprehensive chromaticity change index C(t). With the drying time t as the horizontal axis, the S(t) change curve is drawn and the shrinkage rate Vs(t) is calculated. At the same time, the color-time relationship model and the morphology-time relationship model are established.

[0061] Repeat the above steps multiple times to standardize the resulting color change curves and the comprehensive shrinkage rate S curve. The original curves are first preprocessed by removing outliers and smoothing with a Savitzky-Golay filter. The curve with the highest signal-to-noise ratio is selected as the reference curve. The time axes of the curves are aligned using a dynamic time warping algorithm, and cubic spline interpolation is used to achieve uniform sampling point density. The aligned curves are then fused using a Gaussian regression model to generate standard curves and 95% confidence intervals.

[0062] 5. Dryness monitoring and early warning The fruits and vegetables to be dried are placed on the loading platform and dried according to the drying program of the chamber. The camera captures multi-angle images of the fruits and vegetables in real time during the drying process.

[0063] Real-time calculation of the L of fruits and vegetables to be dried * value, a * value, and b * The image quality, color difference ΔE, and overall shrinkage ratio S are matched against the time series data from the previously constructed appearance morphology model. A drying anomaly alarm is triggered if the color difference ΔE exceeds the preset threshold of 3.5, or if the measured and predicted values ​​of overall shrinkage ratio S deviate by more than 15%. Simultaneously, image feature recognition technology is used to determine the stacking or adhesion of fruits and vegetables. If the effective detection area coverage falls below 90%, an image acquisition quality alarm is issued, and the time point of the anomaly occurrence and the quantitative deviation parameters are annotated on the human-computer interface.

[0064] The present invention further claims protection for a monitoring system for the method for dynamically monitoring the drying quality of fruits and vegetables by integrating color difference and shrinkage rate, comprising: The image processing module is configured to: receive the multi-angle RGB image sequence of the fruit and vegetable drying process transmitted by the external image acquisition device, pre-process the RGB image and extract the effective area of ​​the fruit and vegetable, and perform RGB-Lab color space conversion to output the color parameter L * value, a * value, b *value and color difference value ΔE; generate a 3D point cloud model based on the multi-angle image sequence, calculate the projection area ratio A and 3D shrinkage rate V through orthogonal projection and volume reconstruction algorithm, and fuse the output of the comprehensive shrinkage rate S; The dynamic modeling module is configured to: transform the chromaticity parameter L * value, a * value, b * The values, color difference values ​​ΔE, and combined shrinkage rate S are integrated into a time series data set according to timestamps; the standard color change curve and the comprehensive shrinkage rate S change curve are generated based on Gaussian process regression, and an appearance morphology prediction model with confidence intervals is established; a composite drying kinetics model is embedded in the color-time relationship function and the morphology-time relationship function; The monitoring and early warning module is configured to: receive real-time colorimetric and shrinkage data from the current drying process, and perform dynamic time-warping matching with the appearance morphology prediction model; trigger a drying anomaly signal when the color difference value ΔE exceeds a threshold of 3.5 or the deviation between the measured and predicted values ​​of the comprehensive shrinkage rate S is greater than 15%, and generate a graded alarm based on the deviation parameters; perform edge contour analysis on the input image sequence, and when the fruit and vegetable area coverage is less than 90%, feedback is given indicating image quality anomaly and the data resampling protocol is activated; The burnt visualization module is configured to: provide an abnormal signal timeline marking interface, allowing users to review historical data and adjust drying program parameters; and integrate a calibration interface to receive standard color card image data to update colorimetric conversion parameters, ensuring that the ΔE calculation error is ≤1.5%.

[0065] According to the above technical solution, the RGB image sequence is converted into Lab colorimetric parameters and a three-dimensional point cloud is reconstructed through the image processing module. The dual indicator analysis of color difference ΔE and comprehensive shrinkage rate S overcomes the limitations of traditional single indicator detection and reduces the error in drying quality assessment. The dynamic modeling module uses Gaussian process regression to generate a standard curve with a confidence interval. Combined with a composite drying kinetic model, it can adapt to the characteristics of different fruit and vegetable varieties. The prediction model confidence reaches 95%, significantly improving the sensitivity of anomaly detection. The monitoring and early warning module achieves millisecond-level response through dynamic time warping matching. When ΔE>3.5 or S deviation>15%, multi-level alarms are triggered to detect drying anomalies in advance. At the same time, the stacking detection function ensures effective data coverage and avoids misjudgment. The embedded calibration interface automatically updates the colorimetric conversion parameters through the standard color card, so that the ΔE calculation error is continuously stable at ≤1.5%, reducing the drift rate of continuous operation of the system.

[0066] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A dynamic monitoring method for fruit and vegetable drying quality integrating color difference and shrinkage rate, characterized in that: include: S1. Placing fruit and vegetable samples of the same type as the fruits and vegetables to be dried on a loading platform of a drying device and drying them according to a preset drying program. The loading platform is rotatable, and the drying device is provided with a digital camera facing the loading platform, and the digital camera is communicatively connected to a computer. S2. Setting a timestamp for the drying process, and using the digital camera to take real-time, multi-angle photos of the fruit and vegetable samples at each drying stage according to the timestamp to obtain a color image containing three channels of RGB; S3, using a computer image analysis system to pre-process the three-channel color image and extract the effective area of ​​the fruit and vegetable samples; based on a preset color space conversion algorithm, convert the RGB image into Lab color space and extract the lightness L of the fruit and vegetable sample area * Value, red and green axis a * Value, and yellow-blue axis b * value, and calculate its color difference value ΔE; based on multimodal image analysis and three-dimensional dynamic modeling, the comprehensive shrinkage rate S during the drying process of fruits and vegetables is determined; S4, the L of each drying stage * value, a * value, b * The color difference value ΔE and the comprehensive shrinkage rate S are integrated into time series data to generate the color change curve and the comprehensive shrinkage rate S change curve, and the appearance morphology model of the fruits and vegetables to be dried is obtained; S5. The fruits and vegetables to be dried are placed on a loading platform of the drying equipment and dried according to a box drying program. The digital camera is used to capture multi-angle images of the fruits and vegetables to be dried in real time during the drying process. The images are dynamically compared with the appearance morphology model of the fruits and vegetables to be dried obtained in step S4 to provide a real-time warning of abnormal drying signals.

2. The method for dynamic monitoring of fruit and vegetable drying quality by integrating color difference and shrinkage rate according to claim 1, characterized in that: In step S5, the appearance model of the fruits and vegetables to be dried is used for dynamic comparison, and the specific operation of real-time warning of drying abnormality signals is as follows: L calculated in real time during the drying process of the fruits and vegetables to be dried is used to compare the appearance model of the fruits and vegetables to be dried. * value, a * value, b * The color difference value ΔE and the comprehensive shrinkage rate S are matched with the time series data in the appearance morphology model. When it is detected that the color difference value ΔE exceeds the preset threshold of 3.5 or the deviation between the measured value and the predicted value of the comprehensive shrinkage rate S exceeds 15%, a drying abnormality alarm is triggered. At the same time, the stacking / adhesion status of fruits and vegetables is judged based on image feature recognition technology. If the coverage rate of the effective detection area is less than 90%, an image acquisition quality alarm is issued simultaneously, and the time node of the abnormality and the quantitative deviation parameters are marked on the human-computer interface.

3. The method for dynamic monitoring of fruit and vegetable drying quality by integrating color difference and shrinkage rate according to claim 2, characterized in that: The specific process of calculating the color difference value ΔE in step S3 is as follows: Based on the CIE76 color difference standard, the three-channel RGB image is converted into the Lab color space, and the average chromaticity value L0 of the fruit and vegetable sample area at the initial drying stage, that is, at t=0, is used as the color space. * , a0 * , and b0 * As a benchmark, calculate the color difference value ΔE at time t according to formula (1) t : (1)； Among them, L t * 、a t * , and b t * is the mean chromaticity of the fruit and vegetable sample area in the current drying stage.

4. The method for dynamic monitoring of fruit and vegetable drying quality by integrating color difference and shrinkage rate according to claim 3, characterized in that: In step S3, the specific process of measuring the comprehensive shrinkage rate S during the fruit and vegetable drying process based on multimodal image analysis and three-dimensional dynamic modeling is as follows: S31. Obtain the 2D projection area of ​​the fruit and vegetable sample in the X, Y, and Z directions by orthogonal projection method, and calculate the projection area ratio A at each moment: ; S32, using structured light scanning and binocular vision fusion algorithm, by matching multi-angle image feature points to generate a 3D point cloud model, calculate the 3D shrinkage rate V: ; S33, weighted fusion of the three-dimensional shrinkage rate V and the projected area ratio A to obtain a comprehensive shrinkage rate S: S = mA + nV; Among them, A0 is the initial projection area, A t is the projected area at time t; V0 is the initial volume, V t is the volume at time t; m is the weight coefficient of the projected area ratio A, and n is the weight coefficient of the three-dimensional shrinkage rate V. Based on the least squares fitting method, m is taken as 0.4 and n is taken as 0.

6.

5. The method for dynamic monitoring of fruit and vegetable drying quality by integrating color difference and shrinkage rate according to claim 4, characterized in that: The specific process of step S4 is: S41, Lab value L for each drying stage * 、a * and b * To perform standardization: ； S42. Normalize the comprehensive shrinkage rate S: ; S43. Generate a color change curve and calculate the comprehensive chromaticity change index C(t): Take drying time t as the horizontal axis and plot L * (t), a * (t), and b * (t) three change curves; Comprehensive chromaticity ; S44, generate shrinkage rate change curve and calculate shrinkage rate V s (t): With drying time t as the horizontal axis, draw the S(t) variation curve; Shrinkage rate ; S45. Establishing a color-time relationship model: ； S406: Establishing a morphology-time relationship model: ； Among them, X norm is the degree of deviation of the original data point X from the mean, μ X is the mean, σ X is the standard deviation, X is the original data point; S norm is the normalized value of the comprehensive shrinkage rate S, S norm ∈[0,1]; S min is the minimum value of shrinkage rate; S max is the maximum value of shrinkage rate; L * (t) represents the brightness value of fruits and vegetables at drying time t; a * (t) represents the chromaticity value of fruits and vegetables on the red-green axis at drying time t; b * (t) represents the chromaticity value of fruits and vegetables on the yellow-blue axis at drying time t; g is the lightness change rate coefficient, g∈[0.1,5.0]; f is the red-green axis change amplitude coefficient, f∈[0.01,0.5]; j is the yellow-blue axis change amplitude coefficient j∈[0.1,3.0]; γ is the red-green axis change curve morphology index γ∈[0.5,2.0]; δ is the yellow-blue axis attenuation rate constant, δ∈[0.01,0.5]h -1 , h -1 is the reciprocal of each hour; k is the drying rate constant.

6. The method for dynamic monitoring of fruit and vegetable drying quality by integrating color difference and shrinkage rate according to claim 5, characterized in that: Step S4 further includes the following steps: Take fruit and vegetable samples of the same type as the fruits and vegetables to be dried multiple times, repeat steps S1 to S4, and obtain multiple color change curves and multiple comprehensive shrinkage rate S change curves. Further standardize the multiple color change curves and multiple comprehensive shrinkage rate S change curves, specifically: S401, preprocessing the original curves obtained from multiple experiments, including outlier removal and Savitzky-Golay filtering smoothing; S402, selecting the curve with the highest signal-to-noise ratio as the reference curve, aligning the time axes of the curves using a dynamic time warping algorithm, and achieving uniform sampling point density using cubic spline interpolation; S403. The aligned curves are fused based on the Gaussian regression model to generate a standard curve and a 95% confidence interval. The kernel function is a radial basis function, and the length scale parameter is optimized by maximum likelihood estimation: The optimization objective function of Gaussian process regression is: ； The confidence interval calculation formula is: ； Among them, K is the kernel function matrix, y is the data collected at a time point, is the average value of the data at that time point. The optimization process uses the L-BFGS-B algorithm to iterate 10 to 50 times; N is the number of repeated experiments; M is the total number of observation data points, M=N×Ms, Ms is the number of observation data points in a single experiment; σ(t) is the standard deviation of each time point, .

7. The method for dynamic monitoring of fruit and vegetable drying quality by integrating color difference and shrinkage rate according to claim 6, characterized in that: It also includes verification of the standard curve generated by S403: Randomly retain 1 to 2 experimental data as the validation set, calculate the dynamic time warping distance between the validation set curve and the standard curve, and if the distance exceeds the threshold, re-adjust the preprocessed data and repeat steps S401 to S404 until the dynamic time warping distance of all validation sets does not exceed the threshold.

8. The method for dynamic monitoring of fruit and vegetable drying quality by integrating color difference and shrinkage rate according to claim 7, characterized in that: The drying equipment in step S1 is provided with: Electric rotating stage with angle scale, rotation positioning accuracy ≤ 0.1°; A ring-shaped light source array consisting of 12 groups of LED light sources, of which 6 groups are equipped with combined lenses to form structured light. The LED light sources with combined lenses are spaced apart from the LED light sources without combined lenses. The color temperature of each LED light source is 5500K ± 50K, the illumination is 2000lux ± 2%, and the illumination uniformity is ≥ 95%; Full-frame digital camera, equipped with a 50mm fixed-focus lens, object distance 30cm±0.5cm, resolution ≥50 megapixels; Standard 24-color calibration color chart, compliant with ISO 12641.

9. The method for dynamic monitoring of fruit and vegetable drying quality integrating color difference and shrinkage rate according to claim 8, characterized in that: Step S1 also includes calibration of the drying equipment, specifically: S11. Place the standard 24-color calibration color card at the center of the stage and adjust the LED light source so that the illumination deviation between the center area and the edge area is ≤2%; S12. Set the camera white balance to the D65 standard light source and ensure that the measured values ​​of the white block area of ​​the standard 24-color calibration color chart meet the following requirements: 。 10. A monitoring system based on the method for dynamically monitoring the drying quality of fruits and vegetables by integrating color difference and shrinkage rate according to any one of claims 1 to 9, characterized in that: include: The image processing module is configured to: receive the multi-angle RGB image sequence of the fruit and vegetable drying process transmitted by the external image acquisition device, pre-process the RGB image and extract the effective area of ​​the fruit and vegetable, and perform RGB-Lab color space conversion to output the color parameter L * value, a * value, b * value and color difference value ΔE; generate a 3D point cloud model based on the multi-angle image sequence, calculate the projection area ratio A and 3D shrinkage rate V through orthogonal projection and volume reconstruction algorithm, and fuse the output of the comprehensive shrinkage rate S; The dynamic modeling module is configured to: transform the chromaticity parameter L * value, a * value, b * The values, color difference values ​​ΔE, and combined shrinkage rate S are integrated into a time series data set according to timestamps; the standard color change curve and the comprehensive shrinkage rate S change curve are generated based on Gaussian process regression, and an appearance morphology prediction model with confidence intervals is established; a composite drying kinetics model is embedded in the color-time relationship function and the morphology-time relationship function; The monitoring and early warning module is configured to: receive real-time colorimetric and shrinkage data from the current drying process, and perform dynamic time-warping matching with the appearance morphology prediction model; trigger a drying anomaly signal when the color difference value ΔE exceeds a threshold of 3.5 or the deviation between the measured and predicted values ​​of the comprehensive shrinkage rate S is greater than 15%, and generate a graded alarm based on the deviation parameters; perform edge contour analysis on the input image sequence, and when the fruit and vegetable area coverage is less than 90%, feedback is given indicating image quality anomaly and the data resampling protocol is activated; The burnt visualization module is configured to: provide an abnormal signal timeline marking interface, allowing users to review historical data and adjust drying program parameters; and integrate a calibration interface to receive standard color card image data to update colorimetric conversion parameters, ensuring that the ΔE calculation error is ≤1.5%.

Images