Fruit juice soft sweet texture grading system and method based on visual identification

Abstract

The invention discloses a fruit juice soft sweet texture grading system and method based on visual identification, and belongs to the technical field of material physical property detection.The method comprises the steps that a to-be-detected area of fruit juice soft sweet is positioned through the visual identification technology, and target positioning coordinates are generated; according to the target positioning coordinate, applying a preset energy pulse to the to-be-detected area to excite instantaneous deformation; collecting a dynamic interference fringe image of instantaneous deformation of the to-be-detected area under the action of the energy pulse; phase demodulation is carried out on the dynamic interference fringe image, a microscopic deformation time sequence curve of the to-be-detected area is reconstructed, and dynamic response characteristic parameters are extracted from the microscopic deformation time sequence curve; and determining the texture grade of the fruit juice soft sweets based on the dynamic response characteristic parameters. Through combination of visual technology positioning, energy pulse non-contact excitation and optical interference dynamic analysis, rapid, lossless and automatic online grading of fruit juice soft sweet texture can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material physical property detection, in particular to a fruit juice gummy texture grading system and method based on visual recognition. BACKGROUND

[0002] As a popular candy product, the taste and texture of fruit juice gummy are key factors determining consumer acceptance and product quality. Texture is a comprehensive physical property of food taste evaluation, covering hardness, elasticity, viscosity and other dimensions. In the production process of fruit juice gummy, accurate, stable and efficient control and grading of product texture are important links to ensure product quality uniformity and market competitiveness.

[0003] Currently, fruit juice gummy texture detection mainly relies on two technical means. The first is artificial sensory evaluation, in which trained evaluators subjectively evaluate the hardness and toughness of gummy by chewing. The second is instrument analysis, which generally uses a texture analyzer for texture profile analysis. This method controls the probe to press the sample at a set speed and depth, records the force and displacement curve, and calculates a series of physical indicators such as hardness and elasticity.

[0004] However, the existing technical solutions have certain limitations. The artificial sensory evaluation method is highly subjective, and the evaluation results are easily affected by the physiological and psychological state of the evaluators, resulting in poor repeatability and difficulty in standardization. At the same time, this method is inefficient and costly, and cannot be used for online rapid detection on the production line. Although traditional texture analyzer analysis is objective, it is a contact measurement, and the contact between the probe and the sample can cause irreversible physical damage to the product, making the measured sample unsuitable for sale as finished product. Therefore, this destructive detection can only be used for offline sampling inspection, and cannot achieve online quality monitoring of all products, which poses the risk of missing unqualified products. SUMMARY

[0005] To solve the above problems, the present application provides a fruit juice gummy texture grading system and method based on visual recognition, which combines visual technology positioning, energy pulse non-contact excitation and optical interference dynamic analysis to realize rapid, non-destructive and automatic online grading of fruit juice gummy texture.

[0006] The above objective can be achieved by the following scheme:

[0007] A fruit juice gummy texture grading method based on visual recognition, comprising positioning the to-be-measured area of fruit juice gummy by visual recognition technology, and generating target positioning coordinates;

[0008] According to the target positioning coordinates, a predetermined energy pulse is applied to the to-be-measured area to excite transient deformation;

[0009] collecting a dynamic interference fringe image of the instantaneous deformation of the to-be-tested region under the energy pulse;

[0010] phase demodulating the dynamic interference fringe image, reconstructing a microscopic deformation time sequence curve of the to-be-tested region, and extracting a dynamic response characteristic parameter from the microscopic deformation time sequence curve;

[0011] determining the texture grade of the fruit juice gummy based on the dynamic response characteristic parameter.

[0012] Optionally, the step of positioning the to-be-tested region of the fruit juice gummy by the visual recognition technology and generating the target positioning coordinates comprises:

[0013] capturing a top surface image of the fruit juice gummy, and pre-processing the surface image of the fruit juice gummy to identify a surface flat region;

[0014] determining a center region not containing an edge as the to-be-tested region according to the surface flat region;

[0015] generating the target positioning coordinates based on the geometric center of the to-be-tested region.

[0016] Optionally, the step of applying a preset energy pulse to the to-be-tested region to excite the instantaneous deformation according to the target positioning coordinates comprises:

[0017] using the target positioning coordinates to guide the emission direction of a preset pulse laser to the to-be-tested region;

[0018] configuring the pulse width and energy density parameters of the pulse laser;

[0019] controlling the pulse laser to emit a laser pulse to generate an energy pulse to excite the instantaneous deformation.

[0020] Optionally, the step of collecting the dynamic interference fringe image of the instantaneous deformation of the to-be-tested region under the energy pulse comprises:

[0021] configuring the collection frequency and collection duration to obtain collection parameters;

[0022] capturing a series of interference images within a preset time window before and after the energy pulse is applied according to the collection parameters;

[0023] extracting an image sequence reflecting the instantaneous deformation process from the series of interference images as the dynamic interference fringe image.

[0024] Optionally, the step of extracting the dynamic response characteristic parameter from the microscopic deformation time sequence curve comprises:

[0025] phase demodulating the dynamic interference fringe image to reconstruct a micro-deformation time curve of the to-be-tested region;

[0026] obtaining a maximum deformation depth parameter from the micro-deformation time curve;

[0027] obtaining a deformation recovery half-life parameter from the micro-deformation time curve based on the maximum deformation depth parameter;

[0028] combining the maximum deformation depth parameter and the deformation recovery half-life parameter to generate a dynamic response feature parameter.

[0029] Optionally, the combining the maximum deformation depth parameter and the deformation recovery half-life parameter to generate a dynamic response feature parameter comprises:

[0030] performing viscoelastic analysis using the maximum deformation depth parameter and the deformation recovery half-life parameter to obtain a comprehensive texture score;

[0031] taking the comprehensive texture score as the dynamic response feature parameter.

[0032] Optionally, the performing viscoelastic analysis using the maximum deformation depth parameter and the deformation recovery half-life parameter to obtain a comprehensive texture score comprises:

[0033] calculating an energy absorption rate parameter using the micro-deformation time curve;

[0034] performing viscoelastic analysis using the energy absorption rate parameter, the maximum deformation depth parameter and the deformation recovery half-life parameter to obtain a comprehensive texture score.

[0035] Optionally, the determining the texture grade of the fruit juice gummy based on the dynamic response feature parameter comprises:

[0036] obtaining a texture grading threshold from a preset reference texture database, the reference texture database recording a dynamic response feature parameter distribution range corresponding to a known physical texture grade of a sample;

[0037] comparing the dynamic response feature parameter of the to-be-tested fruit juice gummy with the texture grading threshold in an interval;

[0038] dividing the to-be-tested fruit juice gummy into a corresponding texture grade according to a matching result of the interval comparison.

[0039] Optionally, the method further comprises:

[0040] periodically extracting the classified fruit juice gummy for physical texture re-inspection to obtain re-inspection texture data after determining the texture grade;

[0041] correlate the re-inspection texture data with corresponding dynamic response characteristic parameters to generate a calibration data pair;

[0042] update the reference texture database online by using the calibration data pair, and adjust the texture grading threshold based on the updated reference texture database.

[0043] Based on the same inventive concept, the present application also provides a visual recognition-based fruit juice gummy texture grading system, which comprises:

[0044] a visual recognition module for locating a to-be-tested region of the fruit juice gummy by visual recognition technology and generating target positioning coordinates;

[0045] a pulse excitation module for applying a preset energy pulse to the to-be-tested region to excite transient deformation according to the target positioning coordinates;

[0046] an image acquisition module for acquiring a dynamic interference fringe image of the transient deformation of the to-be-tested region under the action of the energy pulse;

[0047] a data processing module for phase demodulating the dynamic interference fringe image, reconstructing a micro-deformation time series curve of the to-be-tested region, and extracting a dynamic response characteristic parameter from the micro-deformation time series curve;

[0048] a decision module for determining the texture grade of the fruit juice gummy based on the dynamic response characteristic parameter.

[0049] Compared with the prior art, the present application has the following advantages:

[0050] 1. The present application excites transient deformation by energy pulse and measures by optical interference, and the entire detection process does not need to have physical contact with the sample, thereby fundamentally avoiding any physical damage or surface contamination to the fruit juice gummy, ensuring the quality consistency of the products leaving the factory, and improving the food safety level;

[0051] 2. The present application automatically locates the optimal detection region by machine vision, eliminating the random errors caused by manual placement or mechanical positioning; by quantitatively analyzing the micro-deformation time series curve, extracting the dynamic response characteristic parameter, and converting sensory evaluation into repeatable and accurate physical data, the consistency and reliability of the grading results are ensured;

[0052] 3. The present application automatically completes the whole process from sample positioning, excitation, data acquisition to analysis decision without human intervention, greatly improves the detection speed, reduces the dependence on manual quality inspectors and related labor costs, and meets the needs of modern large-scale food production for high-efficiency quality control.

[0053] 4. This invention periodically correlates and compares the system's test results with physical re-inspection data, and updates the reference database and grading thresholds online. This automatically compensates for measurement drift caused by fluctuations in raw material batches or fine adjustments in production processes, ensuring that the grading standards always remain consistent with the actual product characteristics. This enhances the system's adaptability and reliability in complex industrial environments.

[0054] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart illustrating a method for grading the texture of fruit juice gummies based on visual recognition, according to an embodiment of the present invention.

[0057] Figure 2 This is a schematic diagram of a visual recognition-based texture grading system for fruit juice gummies according to an embodiment of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Reference Figure 1 One embodiment of the present invention proposes a visual recognition-based method for grading the texture of fruit juice gummies. By combining visual technology positioning, energy pulse non-contact excitation, and optical interference dynamic analysis, it is possible to achieve rapid, non-destructive, and automated online grading of the texture of fruit juice gummies.

[0060] The method described in this embodiment specifically includes:

[0061] The test area of ​​the fruit juice gummies is located using visual recognition technology, and the target location coordinates are generated.

[0062] applying a preset energy pulse to the to-be-measured region according to the target positioning coordinate to excite transient deformation;

[0063] collecting a dynamic interference fringe image of transient deformation of the to-be-measured region under the action of the energy pulse;

[0064] performing phase demodulation on the dynamic interference fringe image, reconstructing a micro-deformation time curve of the to-be-measured region, and extracting a dynamic response characteristic parameter from the micro-deformation time curve;

[0065] determining the texture grade of the juice gummy based on the dynamic response characteristic parameter.

[0066] Firstly, a flat and central to-be-measured region is automatically determined on the surface of the juice gummy by high-resolution visual recognition technology, providing a spatial reference for subsequent accurate measurement. Subsequently, a controlled energy pulse, such as a pulsed laser, is used to non-contact and non-destructively excite the to-be-measured region, and the energy input will cause the surface of the gummy to produce a small and recoverable dynamic deformation. The occurrence and recovery speed, depth and other dynamic characteristics of this deformation process are directly controlled by the viscoelastic properties of the juice gummy, i.e., its texture. Then, the system captures dynamic interference fringe images of the deformation region in real time through high-speed optical interference measurement technology, and the changes in these fringes contain all the spatiotemporal information of the deformation process. By phase demodulation and calculation on the interference fringe image sequence, the micro-deformation curve of the excitation point with time can be accurately reconstructed. Finally, key parameters that can represent the dynamic response characteristics of the juice gummy, such as the maximum deformation depth and recovery half-time, are extracted from the time curve. These dynamic response characteristic parameters serve as micro-representations of the macro-texture properties of the juice gummy, and are ultimately used for comparison with preset standards to achieve objective determination of the texture grade.

[0067] The present application realizes a solution for juice gummy texture grading by integrating visual positioning, pulse excitation, high-speed interference imaging and dynamic signal analysis technology. This method changes the traditional quality inspection method that relies on artificial senses or contact physical measurement, avoiding the inconsistency caused by subjective judgment and the sample contamination or damage caused by contact measurement. By automatically positioning the to-be-measured region, the consistency of the detection position and the repeatability of the results are ensured. The non-contact excitation and measurement method ensures the non-destructive nature of the detection, making online full detection possible. The accurate quantitative analysis of micro-transient deformation enables the method to sensitively capture subtle texture differences between different juice gummies, thereby achieving grade division. Ultimately, the present application not only improves the quality control level and production efficiency in the production process of juice gummy, but also provides a new technical approach for the digital and standardized representation of texture.

[0068] Optionally, the step of positioning the to-be-tested region of the fruit jelly gummy by the visual recognition technology and generating the target positioning coordinates comprises:

[0069] capturing a top surface image of the fruit jelly gummy and pre-processing the surface image of the fruit jelly gummy to identify a surface flat region;

[0070] determining a center region not containing edges as the to-be-tested region according to the surface flat region;

[0071] generating the target positioning coordinates based on a geometric center of the to-be-tested region.

[0072] Specifically, first, the fruit jelly gummy to be tested is placed on a stage of a detection platform, and a top surface image of the fruit jelly gummy is captured by an image capturing device, such as an industrial high-resolution area array camera. In order to ensure the stability and consistency of image quality, a uniform and shadow-free lighting scheme is adopted in the entire imaging environment, for example, a ring light source or a coaxial light source is used to eliminate environmental light interference and present the details of the gummy surface. The acquired top surface image enters a pre-processing procedure. The pre-processing aims to enhance the effective information and suppress the noise, and the first step is usually to perform a grayscale conversion on the color image, converting the three-channel color information into single-channel grayscale information, to reduce the complexity of subsequent calculations. Subsequently, a Gaussian filter or a median filter algorithm is used to smooth the grayscale image, effectively removing random noise introduced by the image sensor or the environment.

[0073] Then, in order to accurately identify the surface flat region of the fruit jelly gummy, the system will segment and analyze the features of the pre-processed image. By using edge detection algorithms such as Canny operator or Sobel operator, the overall contour of the fruit jelly gummy is extracted, so as to separate it from the background. On this basis, the system uses local variance analysis or texture analysis methods to identify the surface flat region. Specifically, the image within the gummy contour can be divided into several non-overlapping sub-regions, and the variance of the pixel grayscale values in each sub-region is calculated. Generally, the grayscale of the surface flat region changes gently, and its variance value is low; while the region with patterns, characters or uneven surface, its grayscale value changes dramatically, and the variance value is high. By setting a suitable variance threshold, all sub-regions with variance values lower than the threshold can be screened out, and they are merged to form a binary mask, and the region identified by the mask is the surface flat region.

[0074] After the surface flat region is identified, to further eliminate the interference of edge effect on deformation measurement, a central region without edge is determined as the final measurement region. This process can be achieved by performing a morphological erosion operation on the surface flat region mask generated in the previous step. Morphological erosion will reduce the boundary of the target region, and the reduction degree is determined by a preset structure element size. After the erosion operation, the boundary of the original flat region is inwardly retracted to form a smaller and completely surrounded internal connected domain by the original flat region, and the connected domain is the measurement region.

[0075] The last step is to generate the target positioning coordinates for guiding the energy pulse based on the determined measurement region. The coordinates are the geometric center of the measurement region, also known as the centroid. The calculation method is as follows:

[0076] ,

[0077] ,

[0078] wherein, represents the finally generated target positioning coordinates, is the image coordinate of each pixel point in the measurement region, and N is the total number of pixel points constituting the measurement region. The pixel coordinates calculated need to be mapped to the accurate coordinates in the physical world through the camera-mechanical coordinate system conversion matrix established in advance, and the coordinates will be directly used to guide the subsequent pulse excitation module.

[0079] Optionally, the method further comprises:

[0080] using the target positioning coordinates to guide the emission direction of a preset pulse laser to the measurement region;

[0081] configuring the pulse width and energy density parameters of the pulse laser;

[0082] controlling the pulse laser to emit a laser pulse to generate an energy pulse to excite transient deformation.

[0083] Specifically, first, the system uses the target positioning coordinates generated in the previous stage to guide the emission direction of a preset pulse laser. The pulse laser is usually integrated on a two-dimensional galvanometer scanning system. The control computer converts the target positioning coordinates from the image coordinate system to the deflection angle instruction of the galvanometer scanning system to drive the galvanometer mirror to deflect quickly and accurately, so as to accurately aim the focal point of the laser beam at the geometric center of the measurement region determined on the surface of the fruit jelly.

[0084] Secondly, before the laser is emitted, the key parameters of the pulsed laser, i.e. pulse width and energy density, must be configured. The pulse width refers to the duration of a single laser pulse, which is usually selected to be in the order of nanoseconds or picoseconds. The ultra-short pulse is selected to inject energy into the region to be measured in a very short time, so as to realize heat confinement, so as to work in the thermoelastic mechanism, i.e. energy is mainly converted into mechanical stress waves to cause deformation, rather than causing ablation damage such as melting or vaporization of the material, which ensures the non-destructive nature of the measurement. The energy density is a key factor that determines the deformation amplitude, and its calculation formula is:

[0085] ,

[0086] wherein, represents the energy density, represents the total energy of a single laser pulse set by the laser controller, represents the spot area formed on the surface of the gummy after focusing by the optical system. The preset energy density needs to be calibrated through preliminary experiments to ensure that its value is higher than the threshold value that can excite a detectable transient deformation, and at the same time is much lower than the damage threshold of the gummy material. By configuring these two parameters, a standardized energy pulse can be customized for different types or batches of gummies, ensuring the stability and repeatability of the excitation conditions.

[0087] Finally, the system controller sends a synchronization trigger signal to the pulsed laser to control it to emit a laser pulse that meets the preset parameters. The laser pulse serves as an energy pulse, and after reaching the region to be measured, its energy is instantaneously absorbed by the surface layer medium of the gummy, causing a sharp rise in local area temperature and thermal expansion, thereby generating an elastic stress wave propagating into the material, and at the same time forming a transient deformation on the surface. This deformation process usually takes place in the order of microseconds to milliseconds from occurrence to recovery, and its dynamic characteristics directly reflect the viscoelastic property characteristics of the gummy.

[0088] Optionally, the acquisition of the dynamic interference fringe image of the transient deformation of the region to be measured under the action of the energy pulse comprises:

[0089] configuring the acquisition frequency and the acquisition duration to obtain acquisition parameters;

[0090] According to the acquisition parameters, a series of interference images are captured within a preset time window before and after the energy pulse is applied;

[0091] extracting an image sequence reflecting the transient deformation process from the series of interference images as the dynamic interference fringe image.

[0092] Specifically, first, the image acquisition system needs to be configured with parameters to obtain acquisition parameters. The acquisition parameters mainly include acquisition frequency and acquisition duration. The acquisition frequency, i.e. the number of frames captured per second by the image sensor, meets the distortionless recording of the transient deformation process. Since the deformation occurs and recovers extremely quickly after the energy pulse excitation, a high-speed camera must be used, and its acquisition frequency must be set to several kilohertz to several hundred kilohertz to ensure that the complete dynamic profile of the deformation from the beginning, to the peak, to the final recovery can be accurately captured. The acquisition duration defines the total length of the entire recording process. In order to obtain complete deformation data, acquisition must be performed within a preset time window that covers the entire process before, during and after the energy pulse application. Usually, acquisition begins several milliseconds before the pulsed laser emits to record a stable, undeformed surface as a reference, and continues long enough after the pulse is emitted until the surface deformation is completely or substantially recovered.

[0093] Secondly, according to the configured acquisition parameters, a series of interference images are captured within the preset time window before and after the energy pulse is applied. This process is coordinated by a precise timing controller synchronized with the pulsed laser and the high-speed camera. The controller first sends a start acquisition command to the high-speed camera, which then continuously captures images at the preset acquisition frequency. After a short delay set accurately, the controller sends a trigger signal to the pulsed laser to make it emit an energy pulse. During this period, the interference field generated by the optical interference system, such as a Michelson interferometer or a Mach-Zehnder interferometer, is recorded in real time by the high-speed camera. Before the laser pulse arrives, the camera captures static or quasi-static interference fringes. When the laser pulse excites transient deformation, the optical path difference of the surface of the region to be measured changes dramatically, causing the interference fringes to move, bend and change in density. As the surface deformation recovers, the interference fringes gradually return to their initial state. The entire process is recorded as a series of independent digital image frames, i.e. a series of interference images.

[0094] Finally, the system extracts an image sequence reflecting the transient deformation process from the series of interference images as a dynamic interference fringe image. This step is to filter out the effective data segment containing the complete physical process from the original captured video stream. The system automatically identifies the starting frame where deformation begins and the terminating frame where deformation basically ends by analyzing the changes in the fringes between image frames. The entire image sequence from a stable reference frame before the starting frame to a stable frame after the terminating frame is extracted. This accurately extracted, time-ordered image sequence is the dynamic interference fringe image. It is not a single picture, but a data cube containing rich spatiotemporal information, with each frame faithfully recording the two-dimensional spatial distribution of the surface deformation at a specific time.

[0095] Optionally, the extracting the dynamic response characteristic parameter from the micro-deformation time series curve comprises:

[0096] phase demodulating the dynamic interference fringe image to reconstruct a micro-deformation time series curve of the region to be measured;

[0097] acquiring a maximum deformation depth parameter from the micro-deformation time series curve;

[0098] acquiring a deformation recovery half-time parameter from the micro-deformation time series curve based on the maximum deformation depth parameter;

[0099] combining the maximum deformation depth parameter and the deformation recovery half-time parameter to generate a dynamic response characteristic parameter.

[0100] Specifically, the system adopts a digital image processing algorithm, such as Fourier transform method or phase shift method, to analyze the dynamic interference fringe image frame by frame in the phase demodulation of the dynamic interference fringe image. These algorithms can calculate a two-dimensional phase distribution from the fringe distribution of each interference image. By subtracting the phase distribution of each frame from the phase distribution of the reference frame captured before the energy pulse excitation, the phase change amount caused by instantaneous deformation can be obtained . Subsequently, the phase unwrapping algorithm is used to eliminate the 2π ambiguity in phase calculation to obtain a continuous phase change distribution. Finally, the phase change is reconstructed into a physical deformation variable through the following relationship to obtain the three-dimensional topography of the region to be measured at different times.

[0101] ,

[0102] wherein, represents the micro-deformation depth at the image coordinates (x, y) at time point t; is the wavelength of the light source used by the interference measurement system, which is a known constant value; is the phase change amount calculated in the aforementioned phase demodulation and unwrapping steps. By extracting the deformation depth values of all time points at the center point of the energy pulse excitation , the micro-deformation time series curve is constructed . This curve directly shows the complete dynamic process of the excitation point from static, deformation, maximum deformation to gradual recovery.

[0103] Next, the system automatically extracts key characteristic parameters from the micro-deformation time series curve. First, by searching for the extreme value of the entire curve, the maximum deformation depth parameter is directly obtained, which is the peak value on the deformation time series curve and corresponds to the maximum indentation depth that the gummy candy can produce under the action of the standard energy pulse in the physical sense. Then, based on the known maximum deformation depth parameter The system searches along the curve from the peak point forward along the time axis until the deformation depth first recovers to half of the peak value, i.e., 0.5*. Moment Simultaneously record the peak value. point Deformation recovery half-life parameters This is obtained by calculating the time difference between these two points in time, i.e. This parameter reflects the rate of deformation recovery.

[0104] Finally, the obtained maximum deformation depth parameter With the deformation recovery half-life parameter The parameters are then combined to generate the final dynamic response feature parameters. This combination typically involves constructing a two-dimensional feature vector from the two independent scalar parameters. This vector combines the smoothness and resilience of fruit gummy candies.

[0105] Optionally, combining the maximum deformation depth parameter with the deformation recovery half-life parameter to generate dynamic response characteristic parameters includes:

[0106] Viscoelastic analysis was performed using the maximum deformation depth parameter and the deformation recovery half-life parameter to obtain a comprehensive texture score;

[0107] The comprehensive texture score is used as a dynamic response characteristic parameter.

[0108] Specifically, a mathematical model is established to combine the maximum deformation depth parameter, which characterizes softness, and the deformation recovery half-life parameter, which characterizes rebound rate. The ideal texture of fruit gummies typically exhibits moderate hardness and good elasticity; that is, the maximum deformation depth parameter should not be too large, and the deformation recovery half-life parameter should not be too long. After calculation, this comprehensive texture score will serve as the sole dynamic response characteristic parameter for subsequent texture grading.

[0109] Optionally, the step of performing viscoelastic analysis using the maximum deformation depth parameter and the deformation recovery half-life parameter to obtain a comprehensive texture score includes:

[0110] The energy absorption rate parameter was calculated using the aforementioned micro-deformation time-series curve.

[0111] Viscoelastic analysis was performed using the energy absorption rate parameter, the maximum deformation depth parameter, and the deformation recovery half-life parameter to obtain a comprehensive texture score.

[0112] Specifically, the energy absorption rate parameter is calculated using the micro-deformation time curve. The micro-deformation time curve records the change of deformation depth with time. The energy absorption rate parameter aims to quantify the degree of energy dissipation due to internal viscosity during the deformation and recovery cycle, which is directly related to the toughness and damping feeling during chewing. This parameter can be obtained by numerical integration of the micro-deformation time curve. The specific calculation is as follows:

[0113] ,

[0114] where, is the energy absorption rate parameter, is the micro-deformation time curve, the lower limit of integration is the time when deformation starts, and the upper limit is the time when deformation fully recovers. These two time points can be determined by setting a small deformation threshold and automatically finding the time when the deformation value first exceeds and finally falls below the threshold on the time curve. The physical meaning of the energy absorption rate parameter obtained by integration is similar to impulse, and its value reflects the combined effect of deformation depth and duration. A material with higher viscosity or greater energy dissipation will generally exhibit a longer recovery process under similar maximum deformation depth, resulting in a larger energy absorption rate parameter value.

[0115] After calculating the energy absorption rate parameter, the system uses this parameter together with the previously obtained maximum deformation depth parameter and deformation recovery half-time parameter for viscoelastic analysis to generate a more detailed comprehensive texture score. This analysis is achieved through a weighted model containing three characteristic quantities:

[0116] ,

[0117] where, is the final dimensionless comprehensive texture score. and are the known maximum deformation depth and deformation recovery half-time, respectively. is the energy absorption rate parameter calculated above. , and are reference benchmark values set for these three parameters, with the same dimension as the corresponding measured parameters, usually taken from the average measured value of standardized samples with ideal texture, for normalization of each parameter. , and are pre-calibrated dimensionless weight coefficients that reflect the relative importance of softness, rebound rate, and energy dissipation in the final texture evaluation, and satisfy These weight coefficients need to be optimized by multiple linear regression or machine learning modeling with a large number of multi-parameter measurements and professional sensory evaluation data.

[0118] Optionally, the determining the texture grade of the fruit juice gummy based on the dynamic response characteristic parameter comprises:

[0119] obtaining a texture grading threshold value from a preset reference texture database, the reference texture database recording a dynamic response characteristic parameter distribution range corresponding to a sample of a known physical texture grade;

[0120] comparing the dynamic response characteristic parameter of the fruit juice gummy to be tested with the texture grading threshold value in an interval;

[0121] According to the matching result of the interval comparison, the fruit juice gummy to be tested is divided into a corresponding texture grade.

[0122] Specifically, first, the system needs to obtain a texture grading threshold value from a preset reference texture database. This reference texture database is established by a large number of experiments before the system is deployed. The establishment process includes: collecting representative fruit juice gummy samples covering all expected texture grades, testing these samples using traditional authoritative physical texture analyzers (such as TPA texture analyzers), or evaluating them by a professional sensory evaluation team, thereby obtaining the known physical texture grade of each sample, such as being divided into "soft", "medium", "hard" and the like. Subsequently, the same batch of samples is measured using the visual recognition and energy pulse excitation method of the present application, and the dynamic response characteristic parameter corresponding to each sample is calculated, for example, the aforementioned comprehensive texture score. By statistically analyzing the dynamic response characteristic parameters of all samples of the same texture grade in the database, the distribution range of the characteristic parameters corresponding to each grade can be determined, for example, the mean and standard deviation are calculated. Based on these statistical distributions, the system sets a series of texture grading threshold values, which constitute the boundary lines for dividing different grades. For example, for three grades, two threshold values and , where is "soft", is "medium", is "hard". These threshold values are fixed in the reference texture database for online detection.

[0123] Secondly, after the dynamic response characteristic parameter of a fruit juice gummy to be tested is obtained through the above steps, the system compares the parameter with the texture grading threshold value retrieved from the database in an interval. This is a simple numerical comparison process. The system will check in which numerical interval the dynamic response characteristic parameter of the sample to be tested falls, which is divided by the threshold values. For example, the system will judge in turn Is it greater than Is it between and Between, or whether less than .

[0124] Finally, based on the matching results of the interval comparison, the system classifies the tested fruit gummy into the corresponding texture grade. If the overall texture score of the tested sample meets the conditions of a certain interval, the system will automatically mark the sample as the texture grade represented by that interval. For example, if... If the result is negative, the fruit juice gummies are classified as "soft". This final classification information can be used in subsequent automated production line operations, such as rejecting qualified and unqualified products, or classifying and packaging products of different grades.

[0125] Optionally, the method further includes:

[0126] After determining the texture grade, the graded fruit gummies are periodically sampled for physical texture re-inspection to obtain re-inspection texture data;

[0127] The re-inspection texture data is associated with the corresponding dynamic response characteristic parameters to generate calibration data pairs;

[0128] The calibration data is used to update the reference texture database online, and the texture grading threshold is adjusted based on the updated reference texture database.

[0129] Specifically, firstly, after the system automatically grades the texture of the fruit juice gummies and outputs the texture grade, the production quality control process periodically and randomly selects a portion of the graded fruit juice gummies according to a preset ratio. These selected samples are sent to the laboratory for physical texture retesting. The physical texture retesting uses standardized and recognized testing methods, such as texture profile analysis (TPA) using a texture analyzer, to obtain accurate physical indicators such as hardness, elasticity, and viscosity. These accurate indicators, or the authoritative grades derived from them, constitute the retested texture data.

[0130] Next, the system precisely correlates these externally obtained re-inspection texture data with the dynamic response characteristic parameters recorded during the online testing of the sample. Since each tested fruit-juice gummy has a unique identifier, such as a timestamp or serial number, the system can easily retrieve its corresponding dynamic response characteristic parameters, such as the overall texture score, from the historical database. By pairing the re-inspection texture data as the "true value" with the system-measured dynamic response characteristic parameters as the "measured value," a calibration data pair is generated. With continuous periodic sampling, the system accumulates a series of such calibration data pairs.

[0131] Then, the system utilizes these newly generated calibration data to update the preset reference texture database online. Online update means that the process can be automatically executed in the background without interrupting the normal operation of the system. The update algorithm can analyze whether there is a systematic deviation or drift between the revealed system measurement values and the physical true values based on the new calibration data. For example, the algorithm may find that, for the current batch of raw materials, the dynamic response characteristic parameters of all samples are generally high. Based on these new data, the system will update the statistical distribution model of the dynamic response characteristic parameters corresponding to each texture grade in the reference texture database, such as adjusting its mean and variance.

[0132] Finally, based on the updated reference texture database, the system will automatically recalculate and adjust the texture grading threshold values used for grading. Because the characteristic parameter distribution representing each grade in the database has changed, the original grading boundaries may no longer be optimal. The system will use statistical methods, such as recalculating the optimal segmentation line between classes, to generate a new set of texture grading threshold values that better adapt to the current production conditions, and replace the old threshold values.

[0133] Based on the same inventive concept, as shown in Figure 2 The present application also provides a visual recognition-based fruit juice gummy texture grading system, which comprises:

[0134] A visual recognition module for locating the to-be-tested region of the fruit juice gummy through visual recognition technology and generating target positioning coordinates;

[0135] A pulse excitation module for applying a preset energy pulse to the to-be-tested region to excite transient deformation according to the target positioning coordinates;

[0136] An image acquisition module for acquiring the dynamic interference fringe image of the transient deformation of the to-be-tested region under the action of the energy pulse;

[0137] A data processing module for phase demodulating the dynamic interference fringe image, reconstructing the micro-deformation time series curve of the to-be-tested region, and extracting the dynamic response characteristic parameters from the micro-deformation time series curve;

[0138] A decision module for determining the texture grade of the fruit juice gummy based on the dynamic response characteristic parameters.

[0139] Example 1

[0140] To verify the feasibility of the application in practice, the application was applied to the production line of high dietary fiber juice gummy of a certain food production enterprise. The company aims to improve the automation level and consistency of texture control of its high dietary fiber juice gummy products. Currently, the company's texture grading mainly relies on manual sensory evaluation, which has strong subjectivity, low efficiency and cannot achieve full inspection, resulting in fluctuations in taste between product batches.

[0141] In this embodiment, the company integrates the system of the application in the sorting station at the end of its production line. When the high dietary fiber juice gummy enters the detection area through the conveyor belt, the visual recognition module of the system first captures the top image, automatically locates a flat central area that avoids brand imprints and sugar particle scattering as the detection area, and generates target positioning coordinates. Subsequently, the pulse excitation module emits a nanosecond laser pulse with a preset energy to the detection area according to the coordinates, exciting transient deformation. At the same time, the high-speed camera of the image acquisition module captures dynamic interference fringe images of the deformation process at a frequency of tens of kilohertz. The data processing module demodulates the image sequence, reconstructs the micro-deformation time curve, and extracts the maximum deformation depth, deformation recovery half-life and energy absorption rate parameters, and finally calculates a comprehensive texture score. The decision module compares the score with the preset threshold, divides the gummy into "soft", "medium" and "hard" three grades, and controls the sorting mechanical arm to send products of different grades into the corresponding channels.

[0142] To verify the beneficial effects of the application, the company conducted a one-month production line test. The experimental group used the system of the application to automatically grade 10,000 high dietary fiber juice gummies. The control group was manually graded by three experienced sensory panelists. At the same time, a total of 500 samples were randomly selected from both groups and sent to the laboratory for physical testing using a standard texture analyzer (TPA), and the results were used as the "gold standard" to judge the accuracy of the grading.

[0143] In the application of the application, for the ideal product with "medium" texture, the maximum deformation depth measured by the system is generally between 1.1-1.4 microns, the deformation recovery half-life is about 45-55 microseconds, and the calculated comprehensive texture score is higher than 85 points. For "soft" products, their maximum deformation depth increases significantly, such as >1.6 microns, and their deformation recovery half-life also extends accordingly, such as >70 microseconds, resulting in a comprehensive texture score of less than 70 points, which will be removed as defective products. For "hard" products, their maximum deformation depth is smaller, such as <1.0 microns, and their deformation recovery half-life is shorter, such as <40 microseconds, and their comprehensive texture score is between 70 and 85 points, which is classified as standard grade.

[0144] In addition, in the third week of the test, the overall texture of the gummies produced shifted slightly due to the replacement of a new batch of pectin raw material. The system automatically generated a pair of calibration data by periodically sampling and correlating with offline TPA retest data. Using this data pair, the system updated the reference texture database online, and based on the updated data, the comprehensive texture score threshold of the "moderate" level was fine-tuned from 85 points to 83.5 points. This adjustment allowed the system to maintain high accuracy in grading after the change of raw materials, demonstrating its strong adaptive ability and robustness.

[0145] Example 2

[0146] This example aims to illustrate the application of the present application in the food research and development stage for new formula development and texture reference database establishment. A certain food research and development center aims to develop a high dietary fiber juice gummy, which requires accurate evaluation of the impact of different sugar ratios on the final texture of the gummy. The traditional research and development process relies on sensory evaluation by researchers and time-consuming texture analyzer tests, which is inefficient and subjective.

[0147] In this example, the research and development team used the texture grading system of the present application as a research and development tool. First, the team used the system to test a "gold standard" product on the market, and measured its dynamic response characteristic parameters: maximum deformation depth of 1.2 pm, deformation recovery half-life of 50 ps, and comprehensive texture score of 95 points, which served as the target texture for new product development.

[0148] Subsequently, the research and development team prepared five batches of experimental samples A to E, in which the ratio of erythritol and stevioside was systematically adjusted. Each batch of samples was placed in the detection platform of the system of the present application:

[0149] Batch A had too high a proportion of erythritol, and the system detected a maximum deformation depth of 0.8 pm and a deformation recovery half-life of 35 ps. The comprehensive texture score was 76 points. The researchers judged that the texture was too hard, brittle, and lacked elasticity.

[0150] Batch B had a slightly higher proportion of erythritol, a maximum deformation depth of 1.0 pm, a deformation recovery half-life of 42 ps, and a comprehensive texture score of 88 points. The texture was close to the target, but still slightly hard.

[0151] Batch C had an optimized ratio, a maximum deformation depth of 1.22 pm, a deformation recovery half-life of 51 ps, and a comprehensive texture score of 94 points. Its dynamic response characteristic parameters were highly consistent with those of the "gold standard" product. Sensory evaluation also confirmed that this batch of samples had the best hardness and elasticity.

[0152] Batch D had a slightly higher proportion of stevioside, a maximum deformation depth of 1.5 pm, a deformation recovery half-life of 68 ps, and a comprehensive texture score of 82 points. The texture was soft and lacked elasticity.

[0153] E batch steviol glycoside ratio is too high, the maximum deformation depth is 1.9 μm, the deformation recovery half-life is 85 μs, and the comprehensive texture score is 68. The texture is too soft, and there is a sticky sign.

[0154] Through the application of the present application, the research and development team can quantify the subtle texture differences caused by different formulations. The research and development personnel do not need to wait for the feedback of the long sensory evaluation, but can quickly and objectively screen out the C batch as the best formula by comparing the comprehensive texture scores, thereby shortening the research and development cycle. In addition, the five experimental samples and their corresponding dynamic response characteristic parameters are recorded completely, forming the first-hand data for establishing the texture grading standard of the new product line, and laying a solid reference texture database foundation for the online quality control in subsequent large-scale production.

[0155] Embodiment 3

[0156] This embodiment aims to illustrate the application of the present application in the production line for multi-level product sorting to realize the maximization of product value and fine quality management.

[0157] A candy manufacturer produces high-quality bear-shaped high-dietary-fiber juice gummies, and hopes to not only reject unqualified products, but also to subdivide qualified products according to their texture, divide them into “premium level”, “standard level” and “value level” three grades, so as to target different market channels and pricing strategies, thereby improving the overall profit margin.

[0158] In this embodiment, the company downstream of the texture grading system of the present application is configured with a three-channel sorting pneumatic device linked with the decision module. The decision module of the system sets two comprehensive texture score thresholds of 90 points and 75 points according to the pre-established reference texture database.

[0159] When the high-dietary-fiber juice gummies on the production line pass through the detection area in turn, the system completes the whole process from positioning, excitation, collection to analysis in real time, and outputs the comprehensive texture score of each gummy, thereby driving the sorting device.

[0160] When the comprehensive texture score of a certain gummy is detected to be greater than 90 points, it indicates that it has a perfect combination of hardness and elasticity. The decision module issues an instruction, and the gummy is sorted into the “premium level” channel for packaging of its high-end gift box product.

[0161] When the comprehensive texture score of the gummy is detected to be between 75 points and 90 points, it indicates that its texture meets the general requirements of the mass market, but does not reach the top level. These gummies are classified as “standard level” and enter the regular packaging line.

[0162] When the comprehensive texture score of the gummy is less than 75, it indicates that the texture is obviously soft or hard, although it does not belong to the substandard product in safety or morphology, but the taste is not good. These gummies are classified as "value level" for bulk mixed sales or discount channels to recover part of the cost and avoid waste caused by direct disposal as waste products.

[0163] After a month of trial operation, the company found that about 15% of the products were classified as "premium level" through the fine classification of the application, and their selling price was increased by 30% compared with "standard level"; about 10% of the products were classified as "value level" and successfully sold, avoiding the loss of complete scrapping. This scheme not only enables the company to realize differentiated pricing according to the actual quality of the products and improve the gross profit margin, but also generates a proportion distribution diagram of each level product in real time, which becomes a key performance indicator (KPI) for monitoring the stability of the production process. Once the proportion of "value level" abnormally increases, the upstream process parameters can be quickly traced back and adjusted.

[0164] It should be noted that the electrical connection between the above-mentioned units does not necessarily represent the direct connection of the line, and the indirect connection mode can be applied to the embodiments of the present application as long as the purpose of the present application is achieved. The above-described is only an exemplary embodiment of the present application, and cannot limit the scope of the present application.

[0165] That is, any equivalent changes and modifications made according to the teachings of the present application are still within the scope of the present application. Other embodiments of the present application will be readily apparent to those skilled in the art upon considering the specification and practice of the true principles disclosed herein. The present application is intended to cover any variations, uses, or adaptive changes of the present application that follow the general principles of the present application and include common knowledge or conventional technical means in the art not disclosed by the present application.

Claims

1. A method for grading the texture of fruit juice gummies based on visual recognition, characterized in that, The method includes: The test area of ​​the fruit juice gummies is located using visual recognition technology, and the target location coordinates are generated. Based on the target positioning coordinates, a preset energy pulse is applied to the area to be measured to induce instantaneous deformation; Acquire dynamic interference fringe images of the area under test under the action of the energy pulse, showing instantaneous deformation. Phase demodulation is performed on the dynamic interference fringe image to reconstruct the micro-deformation time-series curve of the region under test, and dynamic response feature parameters are extracted from the micro-deformation time-series curve. The texture grade of the fruit juice gummies is determined based on the dynamic response characteristic parameters.

2. The method for grading the texture of fruit juice gummies based on visual recognition according to claim 1, characterized in that, The step of locating the test area of ​​the fruit juice gummy candy using visual recognition technology and generating target positioning coordinates includes: Capture images of the top surface of the gummy candy and preprocess the surface images of the gummy candy to identify flat areas on the surface; Based on the flat surface area, the central region that does not contain edges is determined as the area to be measured. The target positioning coordinates are generated based on the geometric center of the area to be measured.

3. The method for grading the texture of fruit juice gummies based on visual recognition according to claim 2, characterized in that, The step of applying a preset energy pulse to the area to be measured to induce instantaneous deformation based on the target positioning coordinates includes: Using the target positioning coordinates, the emission direction of the preset pulsed laser is guided to the area to be tested; Configure the pulse width and energy density parameters of the pulsed laser; The pulsed laser is controlled to emit laser pulses, generating energy pulses to excite instantaneous deformation.

4. The method for grading the texture of fruit juice gummies based on visual recognition according to claim 3, characterized in that, The dynamic interference fringe image of the measured area under the action of the energy pulse includes: Configure the acquisition frequency and acquisition duration to obtain the acquisition parameters; Based on the acquisition parameters, a series of interference images are captured within a preset time window before and after the energy pulse is applied; The image sequence reflecting the instantaneous deformation process is extracted from the series of interference images and used as a dynamic interference fringe image.

5. A method for grading the texture of fruit juice gummies based on visual recognition according to claim 4, characterized in that, The extraction of dynamic response feature parameters from the micro-deformation time-series curve includes: Phase demodulation is performed on the dynamic interference fringe image to reconstruct the microscopic deformation time-series curve of the region under test; The maximum deformation depth parameter is obtained from the micro-deformation time series curve; Based on the maximum deformation depth parameter, the deformation recovery half-life parameter is obtained from the micro-deformation time series curve; The maximum deformation depth parameter is combined with the deformation recovery half-life parameter to generate dynamic response characteristic parameters.

6. The method for grading the texture of fruit juice gummies based on visual recognition according to claim 5, characterized in that, The step of combining the maximum deformation depth parameter with the deformation recovery half-life parameter to generate dynamic response characteristic parameters includes: Viscoelastic analysis was performed using the maximum deformation depth parameter and the deformation recovery half-life parameter to obtain a comprehensive texture score; The comprehensive texture score is used as a dynamic response characteristic parameter.

7. A method for grading the texture of fruit juice gummies based on visual recognition according to claim 6, characterized in that, The viscoelastic analysis performed using the maximum deformation depth parameter and the deformation recovery half-life parameter yields a comprehensive texture score, including: The energy absorption rate parameter was calculated using the aforementioned micro-deformation time-series curve. Viscoelastic analysis was performed using the energy absorption rate parameter, the maximum deformation depth parameter, and the deformation recovery half-life parameter to obtain a comprehensive texture score.

8. A method for grading the texture of fruit juice gummies based on visual recognition according to claim 7, characterized in that, The determination of the textural grade of the fruit juice gummies based on the dynamic response characteristic parameters includes: The texture grading threshold is obtained from a preset reference texture database, which records the distribution range of dynamic response characteristic parameters corresponding to samples with known physical texture grades. The dynamic response characteristic parameters of the fruit juice gummies to be tested are compared with the texture grading threshold within a range. Based on the matching results of the interval comparison, the fruit juice gummies to be tested are classified into the corresponding texture grades.

9. A method for grading the texture of fruit juice gummies based on visual recognition according to claim 1, characterized in that, The method further includes: After determining the texture grade, the graded fruit gummies are periodically sampled for physical texture re-inspection to obtain re-inspection texture data; The re-inspection texture data is associated with the corresponding dynamic response characteristic parameters to generate calibration data pairs; The calibration data is used to update the reference texture database online, and the texture grading threshold is adjusted based on the updated reference texture database.

10. A visual recognition-based texture grading system for fruit juice gummies, characterized in that, The system includes: The visual recognition module is used to locate the test area of ​​the fruit juice gummies using visual recognition technology and generate target positioning coordinates; The pulse excitation module is used to apply a preset energy pulse to the area to be measured according to the target positioning coordinates to excite instantaneous deformation; The image acquisition module is used to acquire dynamic interference fringe images of the area under test under the action of the energy pulse; The data processing module is used to perform phase demodulation on the dynamic interference fringe image, reconstruct the micro-deformation time-series curve of the region under test, and extract dynamic response feature parameters from the micro-deformation time-series curve. The decision module is used to determine the texture grade of the fruit juice gummies based on the dynamic response characteristic parameters.

Images