CN108022239B - Bubbling wine browning process detection device and method based on machine vision - Google Patents

Abstract

The invention proposes a device and method for detecting the browning process of sparkling wine based on machine vision, which is used to solve the problems of complex detection process, high cost and poor applicability in the prior art; The sample on the diffuse light source panel at the bottom, the digital image of sparkling wine is obtained, and the digital image is processed by grayscale, binarization, corrosion, etc. through machine vision and image processing technology, and the three channels of the image RGB are analyzed. , the blue channel decay percentage was used as a quality marker for sparkling wine, which was highly correlated with the absorbance at 420 nm and the content of 5-hydroxymethyl-2-furfural. The invention only needs to acquire the transmission images of sparkling wine samples at different browning stages, and can analyze multiple brands of sparkling wine samples at the same time, avoid complex sample processing, and can perform multi-sample analysis, with high speed, low cost, and no need for Expensive professional instruments and other chemical reagents.

Description

Device and method for detecting browning process of sparkling wine based on machine vision

technical field

The invention relates to the technical field of sparkling wine browning and image processing, in particular to a device and method for detecting the browning process of sparkling wine based on machine vision.

Background technique

At this stage, sparkling wines are being produced almost everywhere in the world, and consumers’ favor and demand for sparkling wines is also growing. The principle of sparkling wine is to add sugar and yeast to the brewed wine for a second alcoholic fermentation in a closed container, and the carbon dioxide produced during the fermentation process is confined in the bottle and becomes the source of bubbles in the wine (Serra-Cayuela A, Jourdes M, Riu-Aumatell M, et al. Kinetics of browning, phenolics, and 5-hydroxymethylfurfural in commercial sparkling wines. [J]. Journal of Agricultural & Food Chemistry, 2014, 62(5):1159-1166). The added yeast converts the added sugars into alcohol and carbon dioxide, causing the bubbles to churning, one of the most important characteristics of sparkling wine.

Sparkling wine consumption is growing rapidly, according to the International Wine Organization, sparkling wine production has grown by 40% in recent years, and sparkling wine accounts for 8% of global wine production. China and the United States have become the main force driving this growth.

Spanish Cava and French Champagne are the most famous sparkling wine brands. Cava is a high-quality sparkling wine produced by traditional methods and protected by the Designation of Origin system. The grapes used in Cava are carefully selected and great care must be taken during the juicing process in order to obtain the best grape juice at the desired ripeness. The grape juice is then poured into stainless steel barrels for low-temperature fermentation, and then tasted to determine its quality and selectively blended. After the final blending, the wine is bottled and left in the wine cabinet for at least 9 months, but usually longer. During this period, the wine will undergo secondary fermentation in the bottle, forming the wine's complex organoleptic properties such as aroma, color and foaming properties, becoming sparkling wine, and producing yeast deposits. When the aging period is over, the yeasts can be carefully removed, filled with the same batch of wine, cork and seal. In addition, there are two different categories based on the duration of ageing in the bottle: Reserva of at least 15 months and Gran Reserva of at least 30 months (Serra-Cayuela A, Aguilera-Curiel M A, Riu-Aumatell M, et al. Browning during biological aging and commercial storage of Cava sparkling wine and the use of 5-HMF as a quality marker [J]. Food Research International, 2013, 53(1):226-231). In addition to improving grape quality, aging time also affects production costs, with the final price of aged wines higher than new wines.

In terms of physicochemical and organoleptic properties, wine is a dynamic product and as such undergoes complex chemical changes during biological aging and storage. These chemical changes may cause changes in sensory characteristics, especially aroma, flavor and color (Wang Yufeng. Research on the factors affecting wine aroma [D]. Jinan: Shandong Institute of Light Industry, 2010).

Browning is an oxidative process involving sugars, lipids, amino acids and phenols (Li H, Guo A, Wang H. Mechanisms of oxidative browning of wine [J]. Food chemistry, 2008, 108(1): 1-13.), Browning reduces the organoleptic quality of wine (changes in color, flavor and aroma, and increases in astringency), so it must be controlled during processing and storage. Once the bottle is sealed, the browning can no longer be regulated. Therefore, wineries must find markers and parameters that can indicate the browning of sparkling wines, and use technical means to detect the browning of sparkling wines. In practice, it is necessary to develop fast and reliable methods for monitoring the degree of browning and to define parameters that indicate the quality of sparkling wines.

Existing studies have proposed several methods for quantifying and characterizing the browning process in sparkling wines, based on different quality markers and measurement techniques. Such as UV-visible spectroscopy (Ultraviolet-visible spectroscopy, referred to as UV-VIS) and high performance liquid chromatography (High Performance Liquid Chromatography, referred to as HPLC) (Serra-Cayuela A, Aguilera-Curiel M A, Riu-Aumatell M, et al. .Browning duringbiological aging and commercial storage of Cava sparkling wine and the use of 5-HMF as a quality marker[J].Food Research International,2013,53(1):226-231.).

Absorbance at 420 nm wavelength (A ₄₂₀ ) was used as a parameter for browning monitoring of white wine (Kallithraka S, Salacha MI, Tzourou I. Changes in phenolic composition and antioxidant activity of white wine during bottle storage: Accelerated browning test versus bottle storage [J _[ J]. Journal of the Science of Food & Agriculture, 2000, 80(8):1162-1168.). However, in a previous study, the A ₄₂₀ parameter, which is a quality marker for cava sparkling wine, was shown to have low sensitivity and low specificity, after which the 5-hydroxymethyl-2-furfural (5-HMF) content was used instead as More efficient markers (Serra-Cayuela A, Aguilera-Curiel MA, Riu-Aumatell M, et al. Browning during biologicalaging and commercial storage of Cava sparkling wine and the use of 5-HMF as aquality marker [J]. Food Research International , 2013, 53(1): 226-231.), but the chromatographic analysis of this compound in the laboratory requires a lot of time and reagents, which is expensive.

Among the existing research methods, fluorescence excitation-emission spectroscopy combined with parallel factor analysis was successfully used to monitor the _browning process of four sparkling wines. A quick alternative to detection. To investigate a better alternative, the researchers proposed a new method based on colorimetric techniques for monitoring the browning process in cava sparkling wines in an accelerated browning test. The degree of browning was measured using the Lab color model defined by the International Commission on Illumination (CIE) or the Lab color model created by Richard. S. Hunter (Raquel P FG, Joao Barroca M. Evaluation of the Browning Kinetics for Bananas and Pears Submitted to Convective Drying [J]. Current Biochemical Engineering, 2014, 1: 165-172.), but these methods require the use of a dedicated colorimeter or spectrophotometer.

In recent years, with the improvement of the sensor performance of smart phones, the application research based on smart phones is also increasing. Compared with professional laboratory instruments and equipment, smartphones have the advantages of low cost, easy portability, on-site collection and analysis, and easy sharing of results.

SUMMARY OF THE INVENTION

Aiming at the technical problems that the browning detection of the existing sparkling wine requires the use of a special colorimeter or a spectrophotometer, the detection process is complicated, and the applicability is not strong, the present invention proposes a machine vision-based detection device for the browning process of sparkling wine And the method, using machine vision and image processing technology to detect the browning process of sparkling wine, only need to obtain transmission images of sparkling wine samples at different browning stages, and can simultaneously analyze multiple brands of sparkling wine samples , the device is cheap, without expensive professional equipment and other chemical reagents.

In order to achieve the above object, the technical scheme of the present invention is achieved in this way: a machine vision-based sparkling wine browning process detection device and method, the steps are as follows:

Step 1: Build an experimental device and prepare samples;

Step 2: use the camera of the smartphone to obtain the sample image, and perform grayscale and binarization processing on the obtained RGB image IG _i to obtain the binarized image BIN _i , i=1,2,...,6;

Step 3: Invert all the pixels in the image BIN _i to obtain the image IR _i ; use structural elements to corrode the IR _i to obtain the image IER _i , and obtain the shrinkage area of the hole containing the sample;

Step 4: Mark the connected area of the image IER _i , and identify each type of sample according to the connected area mark;

Step 5: Calculate the correction factor based on the mean value of the R channel, G channel and B channel of the shrinkage sample hole area of the standard original sample of the image IG ₁ , and perform color correction on each channel of the image IG _i to obtain the corrected image IGC _i , i=2,...6;

Step 6: Calculate the mean value of the R, G, B channels in the area of the shrinkage hole of each brand sample of the corrected image IGC _i , and select the B channel as the parameter of browning;

Step 7: Calculate the B channel color invariant and the blue decay percentage. The blue decay percentage has a linear relationship with time, and the blue decay percentage is consistent with the browning process of sparkling wine as a quality marker for browning detection.

The samples are Cava sparkling wines of four best-selling brands of Brut, Brut Reserva, Brut Gran Reserva and Semiseco, four wines with different sugar content and production years; 10mL were filled into four 20mL amber vials and degassed under _N2 flow; place the 4 amber vials containing different brands of sparkling wine in a dark environment and keep the ambient temperature at 8°C; place the different brands of the 4 bottles The sparkling wine was used as the standard sample for the subsequent 10-day experiment, and the 4 amber vials containing the standard samples were named Bottle1, Bottle2, Bottle3, Bottle4; and then 10mL of each of the four brands of sparkling wine were taken respectively. Filled into 4 20mL amber vials and degassed under a stream of _N2 ; place these 4 amber vials with different brands of sparkling wine in complete darkness, then place in an oven, set The accelerated browning test was carried out at a constant temperature of 65±1°C; the 4 amber vials containing the accelerated browning samples were named Bottle5, Bottle6, Bottle7, Bottle8; after 0, 2, 4, 6, 8 and 10 The sample information was collected at the sampling time point of each day, and the collection time interval was 48 hours; at each sampling time point, (1) firstly extract one standard sample of Brut, BrutReserva, Brut Gran Reserva and Semiseco, in order from top to bottom Placed in the middle well of the 96-well plate, arranged in a vertical column, and then named the first column; (2) Then extract 4 samples of accelerated browning of Brut, Brut Reserva, Brut GranReserva and Semiseco, and then Put 1 column for each brand, and put them in the 2nd to 5th columns of the 96-well plate in turn.

The experimental device includes a collection box and a light source, the light source is arranged below the collection box, the lower part of the collection box is provided with an orifice plate for holding samples, the top of the collection box is provided with a collection hole, and the collection hole is located directly above the orifice plate.

The collection box is made of black foam core board, and the inside of the collection box is covered with matte black velvet paper; the collection box is a cuboid structure with a height of 80cm, and the difference between the diagonal and the center vertical light path is less than 5%; the light source It is a diffuse reflection light source with controllable intensity; the orifice plate is a 96-well plate, which is made of imported optically transparent pure polystyrene and sterilized by gamma rays.

The method of acquiring the sample image is as follows: use a 0.3ml micropipette to arrange the cava sparkling wine sample in the middle of the well plate, place the well plate on the panel of the light source, cover the collection box on the well plate, and the collection hole is located in the well plate. Above the center of the plate, place the smartphone on the collection hole at the top of the collection box, and use the smartphone's built-in camera to acquire images at the highest resolution and save as JPEG; the standard raw samples are arranged in the well plate and acquired All samples into the same image.

The method of graying and binarizing the obtained RGB image is as follows:

The average value of the three components of R, G, and B of each pixel in the RGB image IG _i obtained by the smartphone is taken as the gray value of the image, that is, Gray _i (x,y)=(R _IG,i (x,y )+G _IG,i (x,y)+B _IG,i (x,y))/3 to obtain a grayscale image Gray _i ;

For the grayscale image Gray _i, use the Otsu algorithm to calculate the optimal threshold T. The optimal threshold T is the grayscale value gt that maximizes σ: if the grayscale of the grayscale image is L, the grayscale range is [0, L -1], using the Ostu method to calculate the optimal threshold for grayscale images: σ=Max[w ₀ (gt)×(u ₀ (gt)-u)^2+w ₁ (gt)×(u ₁ (gt )-u)^2)], where w ₀ is the foreground ratio, u ₀ is the average foreground gray level, w ₁ is the background proportion, u ₁ is the background gray average, and u is the average of the entire gray image.

T is the best threshold for dividing the image. According to the best threshold T, the grayscale image is divided into two parts, and the binarized image BIN _i (x, y) is obtained:

Among them, Gray _i (x, y) is the value of the pixel at (x, y) in the grayscale image;

Invert the binarized image BIN _i (the pixel value of 1 becomes 0, and the pixel value of 0 becomes 1) to obtain the binarized image IR _i ; use the imerode function in Matlab to perform the binarized image IR _i . Erosion operation: IER _i =imerode(IR _i , D ₁ ), where IER _i represents the image obtained after the binarized image IR _i is eroded by the circular structuring element D ₁ with a diameter of 8, i=1,2, 3, . . . , 6; D ₁ =strel('disk', 4), where disk represents a circular shape.

The method for labeling the connected regions of the image IER _i is: using the connected region labeling function bwlabel in Matlab: [LJ _i , LNUM _i ]=bwlabel(IER _i , 8), where LNUM _i represents the number of connected regions in the image IER _i , LJ _i represents a matrix of the same size as the image IER _i , the matrix LJ _i contains the class labels of each connected region in the labeled image IER _i , the values of these labels are 1, 2, ..., LNUM _i , 8 means that the Neighborhood to find connected regions;

The method for identifying the position of each type of sample is: the connected region marks each contracted sample region as a connected region, and for each connected region, the regionprops function in Matlab is adopted: STATS _i =regionprops(LJ _i , 'Centroid'), the 'Centroid' attribute is the centroid of the connected region, STATS _i contains the centroid coordinates (RX _i,nu ,LY _i,nu ) of each connected region (in the image IER _i ), where the subscript nu The value of 1 to 20.

The acquisition method of the corrected image is: based on the mean values R _1,ave , G _1,ave and B _1,ave of the R channel, G channel and B channel of the area of the shrinkage sample hole of the standard original sample of the image IG ₁ as the benchmark, The mean values of R channel, G channel and B channel in the area of the shrinkage sample hole where the standard original sample is located in the images collected at different time sampling points are R _{i, ave} , G _i , _ave and B _{i, ave} is a reference , calculate the three correction factors C _i,R , C _i,G and C _i,B :

Wherein, C _i,R , C _i,G , C _i,B are the correction factors of the R channel, G channel and B channel of the image IG _i , i=2,3,...,6;

Multiply the R channel R _IG,i of all pixels of the image IG _i by C _i,R , the G channel G _IG,i by C _i,G , the B channel B _IG,i by C _i,B , i= 2, 3, ..., 6, three new RGB channels are obtained, and the composed image is the corrected image IGC _i .

Set the pixel points of the connected area where the column coordinates of the centroid coordinates in the image IER ₁ are LY _1,1 to LY _1,4 are set to 1, and the other pixel points are set to 0, to obtain a matrix JZ ₁ , and set all the pixels in the original image IG ₁ to 1. The value of the R channel of the pixel is taken as the matrix R _IG,1 , the value of the G channel of all the pixels in the original image IG ₁ is taken as the matrix G _IG,1 , and the B channel of all the pixels in the original image IG ₁ is taken as the matrix G IG,1 . The value is taken as the matrix B _IG,1 ; the matrix JZ ₁ is dot-multiplied by the matrix R _IG,1 to obtain the matrix DCR ₁ , the matrix JZ ₁ is dot-multiplied by the matrix G _IG,1 to obtain the matrix DCG ₁ , and the matrix JZ ₁ is dot-multiplied by the matrix B _{IG, 1} to get the matrix DCB ₁ ;

R _{1, ave} = sum(sum(DCR ₁ ))/sum(MJ ₁ ),

G _{1, ave} = sum(sum(DCG ₁ ))/sum(MJ ₁ ),

B _{1, ave} = sum(sum(DCB ₁ ))/sum(MJ ₁ ),

where sum(sum(DCR ₁ )) is the sum of all element values in matrix DCR ₁ , sum(sum(DCG ₁ )) is the sum of all element values in matrix DCG ₁ , and sum(sum(DCB ₁ )) is the matrix The sum of all element values in DCB ₁ , sum(MJ ₁ ) is the total number of pixels in the connected regions whose column coordinates are LY _1,1 to LY _1,4 in the image IER ₁ , and MJ ₁ =regionprops(JZ ₁ ,'Area'), the 'Area' property is the number of pixels in each area of the image.

Convert the B channel of the sample image IGC _i to the corresponding color invariant: b _i =B _IGC,i,ave /(R _IGC,i,ave +G _IGC,i,ave +B _IGC,i,ave ), calculate the blue Percent Color Decay % _Bt :

where b _t0 is the B channel invariant at t=0, b _t is the color invariant of the B channel at time t; the percentage of blue decay % B _t represents the browning process as: Y=Y ₀ +kt,

where Y is the absorbance _at 420 nm or 5-HMF content or %Bt, Y0 is the initial value of absorbance at 420 nm or 5-HMF content or % _Bt , _t is time, and k is the rate constant.

The beneficial effects of the present invention are as follows: the sparkling wine sample is placed in the hole in the middle of the 96-well plate, the orifice plate is placed on the diffused light source panel at the bottom of the collection box, and a smart phone is used as the image collection device. Digital images of sparkling wines were obtained in the acquisition wells, digital image processing and analysis were performed to detect the browning of sparkling wines in the RGB color space, and the degree of browning of sparkling wines was determined; with two widely used studies of absorbance at 420 nm Compared with the method to obtain 5-HMF content, the blue channel decay percentage is highly correlated with the commonly used browning index, which is a new marker of browning in sparkling wine. The invention only needs to acquire transmission images of sparkling wine samples of different browning stages, and can simultaneously analyze the browning samples of sparkling wine of multiple brands, avoids complex sample processing, and can simultaneously perform multi-sample analysis, The speed is fast, the price of the device is low, and there is no need for expensive professional instruments and other chemical reagents.

Description of drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

FIG. 1 is a flow chart of the present invention.

FIG. 2 is a schematic diagram of the mechanism of the image capturing device of the present invention.

Fig. 3 is the arrangement sequence of the samples of the present invention.

Figure 4 is a grayscale image of a sample of the present invention.

FIG. 5 is a binarized image of a sample of the present invention.

Figure 6 is an image of the present invention after etching.

Figure 7 is a schematic diagram of the mean value changes of R, G, and B channels of each brand sample area caused by accelerated browning.

Figure 8 is an image of the percentage of decay and heating time in channel B for each brand of sample.

Figure 9 is a graph showing the relationship between the absorbance at 420 nm and the attenuation percentage of the blue channel for each brand sample.

Figure 10 is a graph showing the relationship between the content of 5-HMF and the decay percentage of the blue channel in each brand sample.

Figure 11 is a qualitative comparison of several algorithms.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

As shown in Figure 1, a device and method for detecting the browning process of sparkling wine based on machine vision, the steps are as follows:

Step 1: Build the experimental device and prepare the samples.

The study used four best-selling brands of cava sparkling wine, including Brut, Brut Reserva, Brut Gran Reserva and Semiseco, with different sugar levels and production years. At initial sampling, the quality parameters of each type of cava wine are shown in Table 1. Sugar, alcohol content, pH, free and total sulfur dioxide were measured for each sample using the standard methods of the International Organization of Vine and Wine. Among them, the alcohol content corresponds to the value indicated on the cava bottle.

Table 1 Quality parameters of 4 brands of Cava sparkling wine

Four 20-mL amber vials were filled with 10 mL of each of the four brands of sparkling wine and degassed under a stream of _N2 . The 4 amber vials containing different brands of sparkling wine were then placed in a dark environment at a temperature of 8°C where browning was negligible for 10 days. The different brands of sparkling wine in these 4 bottles were used as standard samples for subsequent experiments, and the 4 amber vials containing the standard samples were named Bottle1, Bottle2, Bottle3, Bottle4.

Immediately thereafter, 10 mL of each of the four brands of sparkling wine were filled into four 20 mL amber vials and degassed under a stream of _N2 . The 4 amber vials containing different brands of sparkling wine were placed in complete darkness and then placed in an oven with a constant temperature of 65±1°C for accelerated browning tests. The 4 amber vials containing accelerated browning samples were named Bottle5, Bottle6, Bottle7, Bottle8.

Sample information was collected on days 0, 2, 4, 6, 8, and 10 (6 sampling time points) with a time interval of 48 hours. At each sampling time point, (1) firstly extract one standard sample of Brut, Brut Reserva, Brut Gran Reserva and Semiseco, and place them in the middle wells of the 96-well plate from top to bottom in a vertical arrangement of 1 Column, and then named column 1; (2) Then extract 4 samples of accelerated browning of Brut, Brut Reserva, Brut Gran Reserva and Semiseco, and then put 1 column for each brand, and put them into the first column of the 96-well plate in turn. 2 to 5 columns, as shown in Figure 3. Therefore, a total of 120 samples were analyzed (6 sampling time points x 4 Cava brands x 5 (4 accelerated browning samples and 1 standard)).

The experimental device is shown in Figure 2, including a collection box 2 and a light source 4. The light source 4 is arranged below the collection box 2, the lower part of the collection box 2 is provided with a 96-well plate 3 for holding samples, and the top of the collection box 2 is provided with a collection hole 1. The collection hole 1 is located directly above the orifice plate 3. The collection box 1 is made of black foam core board and the inside is covered with matte black velvet paper to eliminate internal light reflections. The collection box 1 is a cuboid with a height of 80 cm, so that the difference between the diagonal and the center vertical light path is less than 5%, in order to reduce the difference caused by the light passing through different paths. The light source 4 is a diffuse reflection light source with controllable intensity. Orifice plate 3 is a 96-well plate, made of imported optically transparent pure polystyrene and sterilized by gamma rays.

Step 2: Use the camera of the smartphone to obtain the sample image, and perform grayscale and binarization processing on the obtained RGB image.

Use a micropipette (0.3ml) to arrange the cava sparkling wine samples in the holes in the middle of the well plate 3, add 0.15ml of sample to each well, place the well plate 3 on the panel of the light source 4, put the collection box 2 The cover is on the orifice plate 3 , and the collection hole 1 is located just above the orifice plate 3 . Place the smartphone on the collection hole 1 at the top of the collection box, use the smartphone's built-in camera to acquire the image at the highest resolution (3888×5152=20030976 pixels), and save it in JPEG format. Acquire all samples in the well plate into the same image for each acquisition to avoid effects due to fluctuations in the illumination source or shaking of the smartphone. For efficiency and to enable comparison between different sample images, standard raw samples are placed in column 1 of the plate each time an image is acquired, taken from Bottle1, Bottle2, Bottle3, Bottle4 for each adopted time point .

Place the samples in the center of the 96-well plate and arrange them in order from left to right. The first column is the standard original sample (1 for each sparkling wine, respectively from Bottle1, Bottle2, Bottle3 and Bottle4, a total of 4 samples), the first column Column 2 has 4 samples of Brut (from Bottle5), column 3 has 4 samples of Brut Reserva (from Bottle6), column 4 has 4 samples of Brut GranReserva (from Bottle7), and column 5 has 4 samples of Semiseco (from Bottle7) Bottle8), as shown in Figure 3.

Images of samples containing accelerated browning sparkling wine samples and standard sparkling wines were acquired first, and images containing browning samples and standard samples were acquired at subsequent sampling time points. By acquiring 6 images at 6 sampling time points, in the 1st time, the standard sample in the well plate and the image IG1 _of the sample with accelerated browning after 0 days are acquired; in the 2nd time, the standard sample in the well plate is acquired and images IG ₂ of the samples with accelerated browning over 2 days; on the 3rd, images of the standard sample in the well plate and images IG ₃ of the samples with accelerated browning over 4 days; Image IG4 of standard sample and samples with accelerated browning over ₆ days; at _5th time, image IG5 of standard sample in well plate and sample with accelerated browning over 8 days; at 6th time, acquire well plate Images of standard samples in IG6 and samples with accelerated browning over ₁₀ days. Before acquiring images at each sampling time point, samples should be taken from Bottle1 to Bottle8 and placed in the corresponding position of the well plate, as shown in Figure 3. When collecting the image of the sample, try to ensure that each column containing the sample is in a vertical state in the preview window of the camera, so that the obtained image is close to the ideal position of the sample in Figure 3, so as to facilitate subsequent image processing.

The uniformity of illumination and measurement was studied using a whiteboard, which was placed on a diffuse light source, and the image IB was obtained at the highest resolution (3888×5152=20030976 pixels) using the smartphone’s built-in camera, and the image IB was equally divided into 322 ×243=78246 blocks (each block is represented as BK _{num1, num2} , the value of num1 is 1˜322, and the value of num2 is 1˜243), and the size of each block is 16×16=256 pixels. Statistics BK _{num1, num2} RGB three-channel mean BK _{num1, num2, ave} .

Among them, R _IB (x, y), G _IB (x, y) and B _IB (x, y) represent the value of R channel, the value of G channel and the value of R channel at pixel point (x, y) in image IB, respectively. The value of the B channel. The value range of x is 0 to 3887, and the value range of y is 0 to 5151.

Statistics of the maximum value MAX _bk and the minimum value MIN _bk in BK _{num1, num2 and ave} , after calculation, the maximum difference (MAX _bk -MIN _bk )/MAX _bk is less than 1%, indicating that the influence of external lighting is small, which satisfies the present invention Requirements for uniformity of lighting. The present invention uses the image data of the center position of the orifice plate to minimize errors.

Convert the RGB image IG _i (the value of i is 1 to 6) obtained by the smartphone into a grayscale image Gray _i , and use the average value method, that is, extract the average value of the three RGB channels as the gray value:

Gray _i (x,y)＝( _RIG,i (x,y)+ _GIG,i (x,y)+ _BIG,i (x,y))/3

Among them, R _IG,i (x,y), G _IG,i (x,y), B _IG,i (x,y) are R, G at the pixel point (x, y) in the image IG _i , respectively , the value of the B channel; Gray _i (x, y) represents the gray value of the gray image at the pixel point (x, y), as shown in Figure 4.

The threshold value needs to be determined to binarize the gray image Gray _i . The maximum inter-class variance method (Ostu) is simple, stable and effective in calculation, and is often used in practical applications to determine the threshold value. Ostu algorithm is considered to be the best algorithm for threshold selection in image segmentation. The algorithm is not affected by image brightness and contrast, so it has been widely used in digital image processing. The Ostu algorithm divides the image into two parts: background and foreground according to the grayscale characteristics of the image. The greater the inter-class variance between the background and the foreground, the greater the difference between the two parts of the image. When part of the foreground is mistakenly classified as the background or part of the background is mistakenly classified as the foreground, the difference between the two parts will become smaller. The split with the largest variance means the smallest probability of misclassification.

Assuming that the gray level of the grayscale image is L, the grayscale range is [0, L-1], and the optimal threshold for calculating the grayscale image using the Ostu algorithm is:

σ=Max[w ₀ (gt)×(u ₀ (gt)-u)^2+w ₁ (gt)×(u ₁ (gt)-u)^2)]

Among them, w ₀ is the foreground ratio, u ₀ is the average foreground gray value, w ₁ is the background ratio, u ₁ is the background gray average value, and u is the average value of the entire gray image, so that the gray value gt with the largest σ is the segmentation The optimal threshold T for the image.

Use the above criteria to find the optimal threshold T for the grayscale image Gray _i (x, y), divide the image into 2 parts, and obtain the binarized image BIN _i (x, y).

Among them, Gray _i (x, y) is the value of the pixel at (x, y) in the grayscale image Gray _i , and the value of i is 1 to 6 (a total of 6 images are collected, and one image is obtained each time), The binarized image BIN _i (x, y) is shown in Figure 5.

Step 3: Invert all the pixels in the image BIN _i to obtain a binarized image IR _i ; use structural elements to corrode the image IR _i to obtain the image IER _i , obtain the shrinkage area of the hole containing the sample, and prevent the sample from being installed. The edges of the wells interfere with the image processing of the sample.

D ₁ represents a circular structuring element with a diameter of 4, and its creation function is:

D ₁ =strel('disk',4)

Its meaning is to create a circular structuring element with a radius of 4 (that is, a diameter of 8), and disk represents a circular shape. The function of the strel function is to construct a structuring element. The so-called structuring element can be regarded as a small image, which is usually used for image morphological operations (such as dilation, erosion, opening and closing operations, etc.).

Invert all the pixels in the image BIN _i to get the image IR _i . That is, if the value of the original pixel in the image BIN _i is 0, it becomes 1; if the value of the original pixel is 1, it becomes 0.

Use the imerode function in Matlab to perform the corrosion operation in the morphological processing on the binarized image IR _i . The specific operation method is as follows:

IER _i =imerode(IR _i ,D ₁ )

Among them, IER _i represents the image obtained after the binarized image IR _i is eroded by the circular structuring element D ₁ with a diameter of 8, i=1, 2, ···, 6. As shown in Figure 6. The circular structuring element D ₁ is used to morphologically corrode the binarized image IR _i to remove the boundaries of the empty holes in the image that do not hold the sample, and also remove the boundary of the holes that hold the sample, so as to obtain the central area of the sample. The image of the sample, the constricted area of the well of the sample, prevents the edge of the well containing the sample from interfering with the image processing of the sample. In the area of the hole containing the sample in the image, the sample area in the image is eroded by using the morphological corrosion method (to achieve the purpose of shrinking the sample area), only the pixels inside the hole are used, and the boundary of the hole is removed. influences.

Step 4: Mark the image IER _i with connected regions, and identify the location of each type of sample according to the connected region markers.

Image IER _i for connected region labeling for processing images of wells containing different samples. If E is connected to F and F is connected to G, then E is connected to G. Visually, points that are connected to each other form one area, and points that are not connected to each other form a different area. Such a set of all connected points is called a connected region. The most important method of binarized image analysis is connected region marking, which makes each individual connected region form a marked block by marking the white pixels (targets) in the binarized image. Geometric parameters such as area, contour, circumscribed rectangle, centroid, and invariant moments.

The present invention adopts the connected region labeling function bwlabel in Matlab: [LJ _i , LNUM _i ]=bwlabel(IER _i , 8), wherein LNUM _i represents the number of connected regions in the image IER _i , since there are 20 samples in each image , so LNUM _i =20. LJ _i represents a matrix of the same size as IER _i . The LJ _i matrix contains the class labels of each connected region in the labeled image IER _i . The values of these labels are 1, 2, ..., LNUM _i , and 8 in the formula means yes Find connected regions by 8 neighborhoods.

The algorithm of the function bwlabel is to traverse the image once, and note the equivalent pairs of consecutive cliques and labels in each row (or column), and then relabel the original image by the equivalent pairs. This algorithm is one of the most efficient at present. The sparse matrix and Dulmage-Mendelsohn decomposition algorithm are used in the algorithm to eliminate equivalent pairs.

The image IER _i is marked as a connected region, and each contracted sample region is regarded as a connected region. For each connected region, the regionprops function in Matlab is used: STATS _i =regionprops(LJ _i ,'Centroid'), 'Centroid' property is the centroid of the region (ie, the centroid of the connected region).

The STATS _i calculated according to the regionprops function contains the centroid coordinates (RX _i,nu ,LY _i,nu ) of each connected region (that is, each sample region) in the image IER _i , where the value of nu is 1 to 20 . The centroids of these connected regions are arranged as LY _i,1 to LY _i,20 according to the column coordinates from small to large, and the four sample regions of LY _i,1 to LY _i,4 with the smallest column coordinates are the sample regions in the sample image. Standard sample area (there are four samples in this column, from top to bottom are Brut standard sample, BrutReserva standard sample, Brut Gran Reserva standard sample, Semiseco standard sample); the column coordinates are LY _i,5 ~ LY _i,8 The four sample areas are the Brut sample areas in the sample image; the column coordinates are LY _i,9 to LY _i,12 , the four sample areas are the Brut Reserva sample areas in the sample image; the column coordinates are LY _i,13 The four sample areas of ~LY _i,16 are the Brut Gran Reserva sample areas in the sample image; the column coordinates are the four sample areas of LY _i,17 ~LY _i,20 , which are the Semiseco sample areas in the sample image .

Step 5: Calculate the correction factor based on the mean value of the R channel, G channel and B channel of the shrinkage sample hole area of the standard original sample of the image IG ₁ , and perform color correction on each channel of the image IG _i to obtain the corrected image IGC _i , i=2,...6.

The column coordinates of the centroid coordinates in the image IER ₁ are set to 1 for the pixels of the connected area of LY _1,1 to LY _1,4 , and the other pixels are set to 0 to obtain a matrix JZ ₁ , and then set the pixels in the original image IG ₁ to 1. The values of the R channels of all pixels are taken as the matrix R _IG,1 , the values of the G channels of all the pixels in the original image IG ₁ are taken as the matrix G _IG,1 , and the B channels of all the pixels in the original image IG ₁ are taken as the matrix G IG,1 . The values as matrix B _IG,1 . Multiply JZ ₁ by R _IG,1 to get DCR ₁ , multiply JZ ₁ by G _IG,1 to get DCG ₁ , and multiply JZ ₁ by B _IG,1 to get DCB ₁ .

Using the regionprops function in matlab: MJ ₁ =regionprops(JZ ₁ ,'Area'), calculate the mean value of the R channel of the area of the shrinkage hole of the standard sample of the image IG ₁ . The 'Area' property contained in this function is the total number of pixels in each area of the image.

R _{1, ave} = sum(sum(DCR ₁ ))/sum(MJ ₁ ),

G _{1, ave} = sum(sum(DCG ₁ ))/sum(MJ ₁ ),

B _{1, ave} = sum(sum(DCB ₁ ))/sum(MJ ₁ ),

where sum(sum(DCR ₁ )) is the sum of all element values in matrix DCR ₁ , sum(sum(DCG ₁ )) is the sum of all element values in matrix DCG ₁ , and sum(sum(DCB ₁ )) is the matrix The sum of all element values in DCB ₁ , sum(MJ ₁ ) is the total number of pixels in the connected region whose column coordinates are LY _1,1 to LY _1,4 in the image IER ₁ . Similarly, G _1,ave and B _1,ave are obtained by calculation.

In the same way, the average values Ri _,ave , Gi _,ave and B _i,ave of the R channel, G channel and B channel in the area of the shrinkage sample hole of the standard sample of the image IG _i are calculated, where i=2,3, …, 6.

Then do color correction. Based on the average values R _1,ave , G _1,ave and B _1,ave of the R channel, G channel and B channel of the shrinkage sample hole area of the standard original sample of the image IG ₁ The mean values of R channel, G channel and B channel in the area of the well where the standard original sample is located in the image IG _i (i=2,3,...,6) are R _i , _ave , Gi _{, ave} and B _{i , ave} is the reference (where i=2,3,...,6), three correction factors C _i,R , C _i,G and C _i,B (where i=2,3,...,6) are calculated.

Among them, C _i,R , C _i,G , C _i,B are the correction factors of the R channel, the G channel and the B channel of the image IG _i .

Multiply the R channel R _IG,i of all pixels of the image IG _i by C _i,R , G channel G _IG , i by C _i,G , B channel B _IG,i by C _i,B , to get a new The three channels of RGB, the composed image is named as the corrected image IGC _i , where i=2, 3, . . . , 6.

Step 6: Calculate the mean value of R, G, and B channels of the shrinkage hole area of each brand sample of the corrected image IGC _i , and select the B channel as the browning parameter.

Using the previous method, the mean values R _IGC,i,ave , G _IGC of the three channels of R, G, and B in the shrinkage area of each brand of accelerated browning samples of the corrected image IGC _i (i=1,2,...,6) were obtained _,i,ave , B _IGC,i,ave , and draw a picture, as shown in Figure 7. Figure 7 shows the evolution of R, G and B channel values for the sample area with heating time. It can be seen that the acceleration changes are mainly reflected in the B channel, while the R and G values remain basically constant. This manifested as a darker yellow, ie a change to a lighter brown, as the heating time increased. The changes in the B channel of Brut and GRes samples showed a clear linear dependence, while in Res and SS, the browning started after 48 h, after which there was a linear dependence again between the blue channel value and time. The original sparkling wine samples showed slight differences in their color, as the Brut and Brut Reserva (Res) samples were more similar to each other. While Semiseco (SS) and BrutGran Reserva (GRes) show larger differences, the largest difference between samples is in the blue channel.

Step 7: Calculate the B channel color invariant and blue decay percentage; the blue decay percentage has a linear relationship with time, which is consistent with the browning process of sparkling wine.

In addition to using the changes in R _IGC,i,ave , G _IGC,i,ave , B _IGC,i,ave values for analysis, browning was studied using R, G, and B color invariants. Convert the B channel of the sample image IGC _i to the corresponding color invariant (b _i =B _IGC,i,ave /(R _IGC,i,ave +G _IGC,i,ave +B _IGC,i,ave )), with Quantify and compare browning. The B channel color invariant is similar to absorbance, and the percent blue decay (% _Bt ) can be calculated. b _i at this time corresponds to b _t .

where b _t0 is the B channel invariant at t=0, and b _t is the color invariant of the B channel at time t. Figure 8 shows the change in % _Bt over time for the samples studied, and it can be seen that browning is reflected in a linear increase in % _Bt over time.

R²为0.99。For Brut,

^R2 was 0.99.

R²为0.93。For Brut Reserva,

^R2 was 0.93.

R²为0.93。For Semiseco,

^R2 was 0.93.

R²为0.96。For Brut Gran Reserva,

R2 was ^0.96 .

where x _day is the time unit (day).

UV-3600，德国杜伊斯堡)中使用10mm路径长度的石英比色杯和双蒸水作为参考，测量4种品牌的Cava起泡葡萄酒在420nm处的吸光度。在420nm(A₄₂₀)的吸光度值乘以1000倍，表示为毫吸光度单位(milli-absorbance units，简称mAU)。根据国际葡萄与葡萄酒组织的方法在每个样品中测定5-HMF，使用具有四级L-7100泵、L-7455二极管阵列检测器及LaChrom色谱柱的日立液相色谱仪(Hitachi)进行4种品牌的Cava起泡葡萄酒的液相色谱分析。本发明中的图像采集设备是华为荣耀7智能手机，具有5.2英寸触摸显示屏，分辨率为1920像素×1080像素，2000万像素摄像头(5152×3888像素图像)，支持PDAF相位检测快速对焦，F2.0光圈。on a Shimadzu UV spectrophotometer (

UV-3600, Duisburg, Germany) using a 10 mm path length quartz cuvette and double distilled water as a reference to measure the absorbance at 420 nm of 4 brands of Cava sparkling wine. The absorbance value at 420 nm (A ₄₂₀ ) was multiplied by a factor of 1000 and expressed as milli-absorbance units (mAU for short). 5-HMF was determined in each sample according to the method of the International Organization of Vine and Wine, using a Hitachi liquid chromatograph (Hitachi) with a quadruple L-7100 pump, L-7455 diode array detector and LaChrom column for 4 Liquid chromatography analysis of branded Cava sparkling wine. The image acquisition device in the present invention is a Huawei Honor 7 smart phone, with a 5.2-inch touch display screen, a resolution of 1920 pixels × 1080 pixels, a 20-megapixel camera (5152 × 3888 pixel images), supports PDAF phase detection and fast focusing, F2 .0 aperture.

In order to evaluate the detection effect of the present invention, the % _Bt was compared with other conventional browning monitoring parameters such as _A420 and 5-HMF content, Figure 9 shows the difference between the absorbance measured at 420nm ( _A420 ) and the % _Bt correlation between.

Linear regression analysis was performed to determine the relationship between these browning indicators:

R²为0.98。For Brut,

^R2 was 0.98.

R²为0.98。For Brut Reserva,

^R2 was 0.98.

R²为0.99。For Semiseco,

^R2 was 0.99.

R²为0.96。For Brut Gran Reserva,

R2 was ^0.96 .

where x is the absorbance of the sample at 420 nm (A ₄₂₀ ) and y is %B _t .

The results showed that the browning index expressed by %B _t and A420 showed a similar trend. It should be noted that there is a high correlation between the two different methods. Considering that 420 nm corresponds to a color in the border between blue and violet, the absorption band will give the sample a yellowish to orange/brown tinge, and accordingly these changes will be reflected in the B channel of the image. It is also remarkable that there is agreement between the different samples with very similar slopes, especially the Brut, Res and SS samples.

On the other hand, Figure 10 shows the correlation between % _Bt and 5-HMF content. Linear regression analysis was performed to determine the relationship between these browning indicators:

R²为0.93。For Brut,

^R2 was 0.93.

R²为0.97。For Brut Reserva,

^R2 was 0.97.

R²为0.99。For Semiseco,

^R2 was 0.99.

R²为0.91。For Brut Gran Reserva,

^R2 was 0.91.

where x is the 5-HMF content of the sample and y is the % _Bt .

It can be seen that there is a high correlation among them, especially the Brut, Res and SS samples. It should be noted that 5-HMF is not the only compound indicative of wine browning, so the difference in slope values can be attributed to the presence of other compounds that also affect wine color.

The results obtained with respect to % _Bt , 5-HMF content and A ₄₂₀ characterize the browning process, which can be described by the following equation:

Y=Y ₀ +kt (8)

where Y is the absorbance at 420 nm (mAU) or 5-HMF content or %B _t , t is the time (in days), k is the rate constant (expressed as mAu/day of A ₄₂₀ , mg of 5-HMF /L/day, percent decay per day for the blue channel of _Bt ).

Zero-order elimination kinetics refers to the elimination of drugs in the body at a constant rate, that is, the amount of drug eliminated per unit time remains unchanged regardless of the plasma drug concentration. The main reason for the zero-order kinetic process is the saturation process of drug metabolizing enzymes, drug transporters, and drug binding to plasma proteins. The zero-order kinetic process has the characteristics of active transport. Zero-order dynamics are built up from the system of natural forces, such as electromagnetism, biomechanics, and gravity. Zero-order dynamics has rich connotations, is natural, has far-reaching inspiration and is widely used. The invention adopts the relevant theory of zero-order kinetics to conduct research on the browning of sparkling wine. The formula of zero-order kinetics is the same as the above formula (at this time, k is the zero-order rate constant, Y ₀ is the initial blood drug concentration, and Y is the blood drug concentration at t). The calculated rate constants and monitoring methods for the zero-order kinetics of each sample are shown in Table 2. As can be seen, zero order kinetics were also observed using % _Bt as a control variable (Table 2 and Figure 8). Furthermore, although the values of the browning rate constants evaluated varied according to the different methods of use, the relationship between the results did follow the same trend.

Table 2 Zero-order kinetic parameters (A ₄₂₀ , 5-HMF, %B _t )

Sparkling wines ferment over time (from Brut Gran Reserva to Brut Reserva), and the browning rate constant increases with increasing sugar content (from Brut to Semiseco). showed a constant rate increase in sugar content, and as expected, 5-HMF formation was highly correlated with initial sugar content because the rate of 5-HMF formation was sugar dependent (Camara J S, Alves M A, Marques J C. Changes in volatile composition of Madeirawines during their oxidative ageing[J].Analytica Chimica Acta,2006,563(1):188-197.). Therefore, the sparkling wines that are least affected by the browning process appear to be the highest quality wines (Brut GranReserva, Brut Reserva). Figure 11 summarizes the results obtained graphically and it can be seen that the present invention enables the characterization of the browning process and is in complete agreement with known methods.

The present invention proposes a colorimetric method for monitoring the browning process of sparkling wine, which is fast and low-cost. The camera of the smartphone is used as the acquisition device, and it is used to analyze the transmission image of the cava wine, which avoids complicated sample processing, and can perform single-step multi-sample analysis. The present invention proposes a new control parameter, the percent decay of the blue channel over time, which enables the study and characterization of browning kinetics. That is, the R and G channels remained almost unchanged during the browning process, and the browning process was time-dependent, and mainly affected the B channel as time increased. The present invention proposes to use the blue channel decay percentage %Bt as a quality marker for sparkling wine, and this value is highly correlated with the absorbance at 420nm and the content of 5-hydroxymethyl-2-furfural, the absorbance at 420nm and 5-hydroxymethyl The content of base-2-furfural is a common quality marker for wine browning, and the obtained results are completely consistent with the existing research methods. Thus, the new parameter is highly correlated with the presence of furfural compounds, reflected in the correlation with A420 and, to a lesser extent, 5-HMF concentration, as this is only one of the coloring compounds involved in the browning process. The results of the present invention show that %Bt is a good browning descriptor, and it is only necessary to obtain transmission images of sparkling wine samples at different browning stages to simultaneously analyze multiple brands of sparkling wine samples. Cheap, without expensive professional equipment and other chemical reagents.

The method proposed in the present invention can be used as an alternative to conventional methods and can be used as a key quality control indicator for sparkling wines. The method proposed in the present invention has proved its usefulness, so further research based on this method will be very meaningful.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the present invention. within the scope of protection.

Claims (10)

1. A method for detecting browning process of sparkling wine based on machine vision is characterized by comprising the following steps:

the method comprises the following steps: building an experimental device and preparing a sample;

step two: acquiring a sample image by using a camera of a smart phone, and acquiring an RGB image IG_iCarrying out graying and binarization processing to obtain a binary image BIN_i，i＝1,2,…,6；

Step three: BIN for binary image by using structural elements_iEtching to obtain image IE_iObtaining a constricted region of the well containing the sample;

step four: image IE_iAll the pixels in the image are inverted to obtain an image IER_i(ii) a Image IER_iCarrying out connected region marking, and identifying each type of sample according to the connected region marking;

step five: by image IG₁The mean value of the R channel, G channel and B channel of the area of the shrinkage hole of the standard original sample as a reference, and calculating a correction factor for the image IG_iColor correction is carried out on each channel to obtain a corrected image IGC_i，i＝2,…6；

Step six: calculating a corrected image IGC_iSelecting a channel B as a browning parameter according to the average value of R, G, B channels in the area of the shrinkage hole of each brand sample;

step seven: calculating the color invariant of the channel B and the blue decay percentage, wherein the blue decay percentage has a linear relation with time, and the blue decay percentage is consistent with the browning process of the sparkling wine and is used as a quality marker for browning detection;

IGC corrected image_iB-channel to B-channel color invariant: b_i＝B_IGC,i,ave/(R_IGC,i,ave+G_IGC,i,ave+B_IGC,i,ave) Calculating the blue decay percentage% B_t：

Wherein, b_t0B is the B channel color invariant when t is 0, B_tIs the B channel color invariant at time t; r_IGC,i,ave、G_IGC,i,ave、B_IGC,i,aveRespectively corrected image IGC_iAccelerated browning of the R, G, B three-channel mean of the constricted region of the sample.

2. A machine vision based detection method of browning process in sparkling wines according to claim 1, characterized in that said samples are Brut, Brut Reserva, Brut Gran Reserva and semi seco four popular brands of kava sparkling wines with different sugar content and production year; 10mL of each of the solutions were taken and charged into 420 mL amber vials in N₂Degassing under airflow; placing the 4 amber vials containing different brands of sparkling wine in a dark environment and maintaining the ambient temperature at 8 ℃; the different brands of sparkling wines in the 4 bottles were used as standard samples for the subsequent 10-day experiment, and the 4 amber vials containing the standard samples were named Bottle1, Bottle2, Bottle3, Bottle 4; then, 10mL of each of the four brands of sparkling wine was filled into 4 amber vials of 20mL and placed in N₂Degassing under airflow; placing the 4 amber vials filled with sparkling wines of different brands in a completely dark environment, then placing the vials in an oven, setting the constant temperature of 65 +/-1 ℃, and performing an accelerated browning test; the 4 amber vials containing the accelerated browning samples were designated Bottle5, Bottle6, Bottle7, Bottle 8; collecting sample information at sampling time points of 0, 2, 4,6, 8 and 10 days, wherein the collection time interval is 48 hours; at each sampling time, (1) 1 part each of the standard samples of Brut, Brut Reserva, Brutgran Reserva and Semiseco was first extracted, and the wells placed in the middle of a 96-well plate in this order from top to bottomIn the column, 1 is vertically arranged and then named as the 1 st column; (2) then 4 parts of samples of Brut, Brut Reserva, Brut Gran Reserva and Semiseco which accelerate browning are extracted, and then 1 column of each brand is put into the 2 nd to 5 th columns of a 96-well plate in sequence.

3. The machine vision-based sparkling wine browning process detection method according to claim 1, wherein the experimental device comprises a collection box (2) and a light source (4), the light source (4) is arranged below the collection box (2), a pore plate (3) for containing a sample is arranged at the lower part of the collection box (2), a collection hole (1) is arranged at the top of the collection box (2), and the collection hole (1) is positioned right above the pore plate (3).

4. The machine vision-based detection method for browning processes of sparkling wine according to claim 3, characterized in that said collection box (2) is made of black foam core board, and the inside of collection box (2) is covered with matt black velvet paper; the collection box (2) is of a cuboid structure with the height of 80cm, and the difference between a diagonal line and a central vertical light path is less than 5%; the light source (4) is a diffuse reflection light source with controllable intensity; the pore plate (3) is a 96 pore plate, and the pore plate (3) is made of imported optical transparent pure polystyrene and is sterilized by gamma rays.

5. A machine vision based detection method for browning processes of sparkling wines according to claim 2 or 3, characterized in that the method of obtaining the image of the sample is: arranging kava bubble grape wine samples in the middle of a pore plate (3) by using a 0.3ml micropipette, placing the pore plate (3) on a panel of a light source (4), covering an acquisition box (2) on the pore plate (3), positioning the acquisition hole (1) right above the pore plate (3), placing a smart phone on the acquisition hole (1) at the top of the acquisition box, and obtaining images at the highest resolution by using a built-in camera of the smart phone and storing the images as JPEG; the standard raw samples are arrayed in a well plate and all samples are taken in the same image.

6. The machine vision-based sparkling wine browning process detection method according to claim 1, wherein the obtained RGB image is subjected to graying and binarization by:

IG (image group) of RGB (Red Green blue) image obtained by smart phone_iThe average value of R, G, B three components of each pixel is used as the Gray value of the image, i.e. Gray_i(x,y)＝(R_IG,i(x,y)+G_IG,i(x,y)+B_IG,i(x, y))/3 to obtain a Gray image Gray_i；

For Gray scale image Gray_iAn optimal threshold T is calculated using Otsu's algorithm, which is a gray value gt that maximizes σ: assuming that the gray level of the gray image is L, the gray range is [0, L-1]]The optimal threshold for calculating the gray level image by using the Ostu method is as follows: max [ w [ ]₀(gt)×(u₀(gt)-u)^2+w₁(gt)×(u₁(gt)-u)^2)]Wherein w is₀As a foreground proportion, u₀Is the mean value of the gray levels of the foreground, w₁As background proportion, u₁Is the background gray level mean value, and u is the mean value of the whole gray level image;

t is the optimal threshold value of the segmentation image, the gray level image is segmented into 2 parts according to the optimal threshold value T, and the binary image BIN is obtained_i(x,y)：

Wherein, Gray_i(x, y) is the value of the pixel at (x, y) in the grayscale image;

BIN binary image_iObtaining a binary image IR by negation_iIR is subjected to the imode function in Matlab_iCarrying out corrosion operation: IER_i＝imerode(IR_i,D₁) Wherein, IER_iRepresenting a binarized image IR_iCircular structural element D with diameter of 8₁The image obtained after etching, i ═ 1,2,3, ·, 6; d₁Stral ('disk',4), disk denotes a circular shape.

7. Machine vision based detection method for browning process of sparkling wine according to claim 6, characterised in that it consists in detecting the browning process of sparkling wineImage IER_iThe method for marking the connected region comprises the following steps: adopting a connected region mark function bwleabel in Matlab: [ LJ_i,LNUM_i]＝bwlabel(IER_i8), wherein, LNUM_iRepresentation image IER_iNumber of interconnected regions, LJ_iRepresentation and image IER_iMatrices of equal size, the matrix LJ_iContains a label image IER_iClass label for each connected region, the labels having values of 1,2, …, LNUM_iAnd 8, searching a connected region according to 8 neighborhoods;

the method for identifying the position of each type of sample comprises the following steps: the connected regions labeling treats each contracted sample region as a connected region, using, for each connected region, the regionprops function in Matlab: STATS_i＝regionprops(LJ_i'Central'), the 'Central' attribute is the Centroid of the connected region, STATS_iContains an image IER_iOf each connected Region (RX)_i,nu,LY_i,nu) And the subscript nu has a value of 1-20.

8. A machine vision based detection method of browning process of sparkling wine according to claim 1, characterized in that the corrected image is obtained by: by image IG₁Of the standard raw sample, the mean value R of the R channel, G channel and B channel of the area of the shrinkage cavity_1，ave、G_1，aveAnd B_1，aveFor reference, the mean value of the R channel, the G channel and the B channel of the area of the shrinkage hole where the standard original sample in the images acquired by the subsequent different time sampling points is R_i，ave、G_i，aveAnd B_i，aveFor reference, three correction factors C are calculated_i,R、C_i,GAnd C_i,B：

Wherein, C_i,R、C_i,G、C_i,BAs an image IG_iOf R channel, G channel and B channelCorrection factors, i ═ 2,3, …, 6;

image IG_iR channels R of all pixel points_IG，iMultiplying by C_i,RG channel G_IG，iMultiplying by C_i,GThe B channel B_IG，iMultiplying by C_i,BI 2,3, …,6, obtaining new RGB three channels, and forming a corrected image IGC_i。

9. Machine vision based detection method for browning processes of sparkling wines according to claim 7 or 8, characterized in that the image IER is₁The column coordinate of the centroid coordinate in (1) is LY_1,1～LY_1,4The pixel point of the connected region is set to 1, and other pixel points are set to 0 to obtain a matrix JZ₁Original image IG₁The values of the R channels of all the pixel points in (1) are used as a matrix R_IG,1Original image IG₁The values of the G channels of all the pixel points in (1) are used as a matrix G_IG,1Original image IG₁The values of the B channels of all the pixel points in (1) are used as a matrix B_IG,1(ii) a Will matrix JZ₁Dot-by-dot matrix R_IG,1Obtaining a matrix DCR₁Will matrix JZ₁Dot-product matrix G_IG,1Obtaining a matrix DCG₁Will matrix JZ₁Dot-by-dot matrix B_IG,1Obtaining a matrix DCB₁；

R_1，ave＝sum(sum(DCR₁))/sum(MJ₁)，

G_1，ave＝sum(sum(DCG₁))/sum(MJ₁)，

B_1，ave＝sum(sum(DCB₁))/sum(MJ₁)，

Of these, sum (DCR)₁) Is a matrix DCR₁Sum of all the element values in (1), sum (DCG)₁) Is a matrix DCG₁Sum of all the element values in (1), sum (DCB)₁) Is a matrix DCB₁Sum of all element values in, sum (MJ)₁) Is an image IER₁The column coordinate of the centroid coordinate in (1) is LY_1,1～LY_1,4Total number of pixels of connected region, MJ₁＝regionprops(JZ₁,'Area') The 'Area' attribute is the number of pixels in each Area of the image.

10. The machine vision-based sparkling wine browning process detection method of claim 1, wherein said blue decay percent% B_tThe browning process is shown as follows: y ═ Y₀+kt，

Wherein Y is the absorbance at 420nm or the 5-HMF content or% B_t，Y₀Is the absorbance at 420nm or the 5-HMF content or% B_tT is time and k is the velocity constant.

Images