Quantitative evaluation method for raisin plumpness based on image recognition
Through the method based on image recognition, the wrinkle area of the raisin surface and the score is calculated, the problems of unstable and subjective evaluation of raisin fullness in the prior art are solved, and fast and accurate quantitative evaluation is achieved.
Patent Information
- Application Number
- CN202411671585.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-20
- Publication Date
- 2025-03-11
AI Technical Summary
The prior art is difficult to quickly and accurately evaluate the fullness of raisins, mainly because it relies on artificial sensory evaluation, and the results are unstable and greatly affected by subjective factors.
Using an image recognition-based method, the proportion of the wrinkle area on the surface of raisins to the total area and the score is calculated using a specific formula to quantify the fullness of raisins.
The rapid and accurate quantitative evaluation of the fullness of raisins is achieved, which avoids the subjectivity of manual evaluation and improves the stability and objectivity of evaluation.
Smart Images
Figure BDA0005146747800000051 
Figure BDA0005146747800000061 
Figure BDA0005146747800000071
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of food detection, and in particular relates to a quantitative evaluation method for the fullness of raisins based on image recognition. Background Art
[0002] Raisins are an important grape processed product, and fullness is an important evaluation index of raisin quality. Raisins with high fullness have full grains, thick flesh, and good appearance quality and taste; raisins with poor fullness not only have poor appearance quality, thin flesh and thick skin, but also poor tasting experience. The fullness of raisins is affected by many factors such as grape variety characteristics, cultivation measures, and drying technology. At present, the evaluation of the fullness of raisins in production is mainly through sensory evaluation, which can only be simply graded using descriptions such as full, relatively full, and not full; or evaluated by professionals through comparative scoring, and the evaluation results are unstable and greatly affected by the subjective factors of the participants. This restricts the selection of grape varieties for drying, raisin production, sales, and the evaluation of raisin processing effects. Therefore, it is very necessary to provide a method that can quickly and accurately quantify the fullness of raisins. Summary of the invention
[0003] In view of this, the object of the present invention is to provide a raisin fullness quantitative evaluation method based on image recognition, which can evaluate the fullness of raisins simply, quickly and accurately.
[0004] The present invention provides a method for quantitatively evaluating the fullness of raisins, comprising the following steps:
[0005] The ratio of the wrinkle area on the surface of the raisin to the total surface area of the raisin is measured and the quantified score of the fullness of the raisin is calculated by using the formula:
[0006] y=0.0233x 3 -1.4634x 2 +32.121x-158.71,
[0007] Among them, y is the quantitative score of the raisin fullness, and x is the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin;
[0008] y>90 points means full, y between 70 and 90 points means relatively full, y between 40 and 70 points means less full, and y<40 points means extremely less full.
[0009] Preferably, the measurement method is image recognition measurement.
[0010] Preferably, the image recognition measurement comprises the following steps:
[0011] Image acquisition: take pictures to obtain the images of the raisin samples to be tested;
[0012] Image optimization: remove the image background, change the image format to RGB Stack, and then sharpen the image to obtain an optimized image;
[0013] Data collection: Determine and extract data on the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin in the optimized image.
[0014] Preferably, during the image acquisition, the raisin samples to be tested do not overlap or adhere to each other.
[0015] Preferably, when the image is collected, the number of raisin samples to be tested is 30 to 50.
[0016] Preferably, the raisin samples to be tested are randomly selected.
[0017] Preferably, the removing of the picture background is to use photo shop software or Apowersoft cutout software to remove the picture background.
[0018] Preferably, the sharpening process is performed 2 to 4 times using the Sharpen function under the process menu of Image-J software.
[0019] Preferably, the data are measured and extracted using Image-J software, Image-Pro software or MIPAR software.
[0020] The invention also provides application of the evaluation method in quality detection of raisin products.
[0021] Beneficial effects of the present invention:
[0022] The present invention provides a method for quantitatively evaluating the fullness of raisins based on image recognition. The method collects data on the ratio of the wrinkle area on the surface of raisins to the total surface area of raisins through image recognition, and then calculates the fullness score of the raisins quickly and accurately through the formula of the present invention. On the one hand, the cumbersome process of existing manual evaluation is omitted and the evaluation method is simplified. On the other hand, the influence of subjective factors in manual evaluation is avoided, so that the obtained raisin fullness result is more accurate and objective. The present invention provides a new idea for the quality evaluation of the fullness of raisins. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 This is an example picture of the raisin sample to be tested obtained through image acquisition.
[0024] Figure 2 are example images of the image processing process; a is the example image after background removal, b is the example image after changing to RGBStack format, and c is the example optimized image after sharpening.
[0025] Figure 3 This is the fitting curve diagram of the raisin fullness calculation formula. DETAILED DESCRIPTION
[0026] The present invention provides a method for quantitatively evaluating the fullness of raisins, comprising the following steps:
[0027] The ratio of the wrinkle area on the surface of the raisin to the total surface area of the raisin is measured, and the measuring method is preferably image recognition measurement, and the image recognition measurement preferably includes the following steps:
[0028] Image acquisition: Take a picture to obtain a picture of the raisin sample to be tested; the number of grains of the raisin sample to be tested is preferably within the range of 30 to 50 grains, and the raisin sample to be tested is preferably randomly selected. The present invention does not specifically limit the random selection method, and the conventional random selection method in the art can be used. In one embodiment, the 5-point method can be selected to randomly extract 300 to 500g (such as 300g, 400g or 500g) of samples, that is, the sample to be tested is flattened to form a uniformly distributed rectangular single-layer plane, the midpoint of the diagonal of the sample rectangular plane is determined as the central sampling point, and then four points on the diagonal with equal distances from the central sampling point are selected as sampling points for sampling. After sampling, the quartering method is selected for continuous selection, that is: the evenly mixed sample is flattened into a single-layer circle, divided into four equal parts, two opposite parts are taken and mixed together, and then the mixed sample is flattened into a circle, and the above steps are repeated until the raisins reach 30 to 50 grains. After randomly selecting the sample, the sample is preferably placed on a solid color base plate so that the raisins do not overlap or stick to each other to avoid affecting subsequent image optimization and data extraction. The sample to be tested is photographed to obtain the sample image. The sample image of the sample to be tested is shown in FIG. Figure 1 shown.
[0029] Image optimization: Remove the background of the sample image to be tested to avoid errors caused by the background. The background removal is preferably performed using Photoshop or Apowersoft. In one embodiment, the background removal operation in Photoshop or Apowersoft can be selected to remove the background. The sample image after background removal is shown in the figure below. Figure 2 As shown in a in the figure; the picture format is modified to RGB Stack, and the modification of the picture format is preferably performed by using Image-J software or Image-Pro software. In one embodiment, the conventional modification of the picture format in Image-J software or Image-Pro software can be selected to modify the picture format; the example figure after the modification of the format is as shown in Figure 2As shown in b in the figure; the image is then sharpened to further enhance the contrast between the raised and sunken parts of the wrinkles so that the wrinkles can be clearly displayed. The sharpening process is preferably performed 2 to 4 times using the Sharpen function under the process menu of the Image-J software. In one embodiment, the sharpening process can be selected 2, 3 or 4 times to obtain an optimized image. An example of the optimized image is shown in FIG. Figure 2 As shown in c.
[0030] Data collection: Determine and extract data on the ratio of the wrinkle area on the surface of the raisins to the total surface area of the raisins in the optimized image. The data is preferably determined and extracted using Image-J software, Image-Pro software or MIPAR software. In one embodiment, the measure function under the Analyze module of Image-J software can be selected to check Area fraction for determination to obtain the desired ratio of the wrinkle area to the total surface area. The conventional data determination and extraction operation steps of Image-Pro software or MIPAR software can also be selected to determine and extract the ratio of the wrinkle area to the total surface area. The determination is preferably performed 3 to 10 times. In one embodiment, 3, 4, 5, 6, 7, 8, 9 or 10 times can be selected to determine. The average value of the determination is taken to obtain the ratio of the wrinkle area on the surface of the raisins to the total surface area of the raisins, and then the average value is substituted into the formula to calculate the quantitative score of the fullness of the raisins. The formula is:
[0031] y=0.0233x 3 -1.4634x 2 +32.121x-158.71,
[0032] Wherein, y is the quantitative score of raisin fullness, and x is the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin. The quantitative score of raisin fullness preferably reflects the fullness of raisins according to the following standards: y>90 points for fullness, y 70-90 points for relatively fullness, y 40-70 points for relatively less fullness, and y<40 points for extremely less fullness. The formula preferably uses any 20 groups of raisin samples purchased from the market, and measures the "ratio x of the wrinkle area on the raisin surface to the total surface area of the raisin" obtained by the above-mentioned image recognition measurement method. Then, 5 experts conduct sensory evaluation on the fullness of the above 20 groups of raisin samples to obtain the score y. A scatter plot is made between the "sensory evaluation score y" and the "ratio x of the wrinkle area on the raisin surface to the total surface area of the raisin", and a multivariate linear regression fitting is performed to obtain the formula (R 2 >0.98), the fitting results are as follows Figure 3 shown.
[0033] The present invention also provides an application of the above evaluation method in the quality detection of raisin products. The evaluation method of the present invention can be used to obtain the fullness of raisins simply, quickly and accurately, so as to better detect the product quality of raisins.
[0034] The technical solutions provided by the present invention are described in detail below in conjunction with the embodiments, but they should not be construed as limiting the protection scope of the present invention.
[0035] In the following embodiments, unless otherwise specified, all of them are conventional methods.
[0036] Unless otherwise specified, the materials and reagents used in the following examples can be obtained from commercial sources.
[0037] The raisin samples purchased from the market in the following embodiments and the raisin samples used in the fitting formula were purchased from different markets.
[0038] Example 1
[0039] Seedless white raisins were purchased from the market and evaluated using a quantitative evaluation method for raisin fullness based on image recognition:
[0040] (1) Image acquisition: Randomly select 3 groups (group 1, group 2, group 3) of raisins from the raisin samples by using the 5-point method (the 5-point method is to flatten the sample to form a uniformly distributed rectangular single-layer plane, determine the midpoint of the diagonal of the sample rectangular plane as the central sampling point, and then select four points on the diagonal that are equidistant from the central sampling point as sampling points for sampling) , each group weighing 500g. After each group of samples is mixed, the quartering method (the quartering method is to flatten the mixed sample into a single-layer circle, divide it into four equal parts, take two opposite parts and mix them together, then flatten the mixed sample into a circle, and repeat the above steps) is used to select until the number of particles in each group of samples is between 30 and 50. The sample obtained by screening is placed on a solid color base plate so that the raisin particles do not overlap or adhere to each other, and a picture of the sample is obtained by taking a picture.
[0041] (2) Image optimization: First, use Apowersoft to remove the image background. The specific operation is to import the image into the software and use the image recognition function to remove the image background. Then import the image into Image-J software, change the image format to RGB Stack format in the Type menu of the Image module, and then use the Sharpen function in the process menu of Image-J software to sharpen it twice to obtain the optimized image.
[0042] (3) Quantitative data acquisition: The images were optimized and Image-J software was used to measure and extract the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin. The specific operation was to select the measure function under the Analyze module of Image-J software and check Area fraction for measurement to obtain the required ratio of the wrinkle area to the total surface area. Each group was measured 5 times and the average value was taken to obtain the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin.
[0043] (4) Substitute the proportion data obtained in (3) into the formula to obtain the quantitative score result of the fullness of the sample tested.
[0044] The formula is y=0.0233x 3 -1.4634x 2 +32.121x-158.71, where y is the quantitative score of the fullness of the raisins, and x is the ratio of the wrinkle area on the surface of the raisins to the total surface area of the raisins. y>90 is full, y 70-90 is relatively full, y 40-70 is relatively unfull, and y<40 is extremely unfull.
[0045] Comparative Example 1
[0046] The seedless white raisins purchased from the market as described in Example 1 were subjected to sensory scoring, and 5 experts were invited to conduct sensory evaluation on the fullness, with a score of 0-100 points. The higher the score, the fuller the raisins. The average of the scores of the 5 experts represented the scoring result.
[0047] The evaluation results of Example 1 and Comparative Example 1 are shown in Table 1:
[0048] Table 1 Evaluation results of fullness of seedless white raisins
[0049]
[0050] Example 2
[0051] In addition, seedless white raisins were purchased from the market and evaluated using a quantitative evaluation method for the fullness of raisins based on image recognition:
[0052] (1) Image acquisition: Three groups (group 1, group 2, and group 3) of raisins were randomly selected from the raisin samples using the 5-point method (the same operation method as in Example 1), each group weighing 500 g. After each group of samples was mixed, they were continuously selected using the quartering method (the same operation method as in Example 1) until the number of particles in each group of samples was between 30 and 50. The sample obtained by screening was placed on a solid color base plate so that the raisin particles did not overlap or adhere to each other, and a picture of the sample was taken.
[0053] (2) Image optimization: First, use Apowersoft to remove the image background. The specific operation is to import the image into the software and use the image recognition function to remove the image background. Then import the image into Image-J software, change the image format to RGB Stack format in the Type menu of the Image module, and then use the Sharpen function in the process menu of Image-J software to sharpen it three times to obtain the optimized image.
[0054] (3) Quantitative data acquisition: Optimize the image and use Image-J software to measure and extract the data of the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin. The specific operation is as follows: select the measure function under the Analyze module of Image-J software and check Area fraction to measure and obtain the required ratio of the wrinkle area to the total surface area. Each group is measured 3 times and the average value is taken to obtain the data of the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin.
[0055] (4) Substitute the proportion data obtained in (3) into the formula to obtain the quantitative score result of the fullness of the sample tested.
[0056] The formula is y=0.0233x 3 -1.4634x 2 +32.121x-158.71, where y is the quantitative score of the fullness of the raisins, and x is the ratio of the wrinkle area on the surface of the raisins to the total surface area of the raisins. y>90 is full, y 70-90 is relatively full, y 40-70 is relatively unfull, and y<40 is extremely unfull.
[0057] Comparative Example 2
[0058] The seedless white raisins purchased from the market as described in Example 2 were subjected to sensory scoring, and 5 experts were invited to conduct sensory evaluation on the fullness, with a score of 0-100 points. The higher the score, the fuller the raisins. The average of the scores of the 5 experts represented the scoring result.
[0059] The evaluation results of Example 2 and Comparative Example 2 are shown in Table 2:
[0060] Table 2 Evaluation results of fullness of seedless white raisins
[0061]
[0062] Example 3
[0063] The Institute of Horticultural Crops of Xinjiang Academy of Agricultural Sciences used raisins made from seedless white grapes with low maturity to evaluate the plumpness of raisins using a quantitative evaluation method based on image recognition:
[0064] (1) Image acquisition: Three groups (group 1, group 2, and group 3) of raisins were randomly selected from the raisin samples using the 5-point method (the same operation method as in Example 1), each group weighing 300 g. After each group of samples was mixed, they were continuously selected using the quartering method (the same operation method as in Example 1) until the number of particles in each group of samples was between 30 and 50. The sample obtained by screening was placed on a solid color base plate so that the raisin particles did not overlap or adhere to each other, and a picture was taken to obtain the test image.
[0065] (2) Image optimization: First, use Photoshop to remove the image background. Then use Image-Pro to change the image format to RGB Stack. Then use the Sharpen function under the process menu of Image-Pro to sharpen the image four times to obtain an optimized image.
[0066] (3) Quantitative data collection: The images were optimized and MIPAR software was used to measure and extract the data on the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin. Each group was measured 10 times and the average value was taken to obtain the data on the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin.
[0067] (4) Substitute the proportion data obtained in (3) into the formula to obtain the quantitative score result of the fullness of the sample tested.
[0068] The formula is y=0.0233x 3 -1.4634x 2 +32.121x-158.71, where y is the quantitative score of the fullness of the raisins, and x is the ratio of the wrinkle area on the surface of the raisins to the total surface area of the raisins. y>90 is full, y 70-90 is relatively full, y 40-70 is relatively unfull, and y<40 is extremely unfull.
[0069] Comparative Example 3
[0070] The raisins dried by the Institute of Horticultural Crops of Xinjiang Academy of Agricultural Sciences described in Example 3 were subjected to sensory scoring, and 5 experts were invited to conduct sensory evaluation on the fullness, with a score of 0 to 100 points. The higher the score, the fuller the raisins. The average of the scores of the 5 experts represented the scoring result.
[0071] The evaluation results of Example 3 and Comparative Example 3 are shown in Table 3:
[0072] Table 3 Evaluation results of plumpness of air-dried seedless raisins
[0073]
[0074] The evaluation results in Tables 1-3 show that the evaluation method provided by the present invention is close to the results obtained by the manual sensory scoring evaluation method, indicating that the evaluation method of the present invention can accurately reflect the fullness of raisins.
[0075] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A quantitative evaluation method for the fullness of raisins, characterized in that: The following steps are involved: The ratio of the wrinkle area on the surface of the raisin to the total surface area of the raisin is measured and the quantified score of the fullness of the raisin is calculated by using the formula: y=0.0233x 3 -1.4634x 2 +32.121x-158.71, Among them, y is the quantitative score of the raisin fullness, and x is the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin; y>90 points means full, y between 70 and 90 points means relatively full, y between 40 and 70 points means less full, and y<40 points means extremely less full.
2. The evaluation method according to claim 1, characterized in that: The measuring method is image recognition measurement.
3. The evaluation method according to claim 2, characterized in that: The image recognition measurement comprises the following steps: Image acquisition: take pictures to obtain the images of the raisin samples to be tested; Image optimization: remove the image background, change the image format to RGB Stack, and then sharpen the image to obtain an optimized image; Data collection: Determine and extract data on the ratio of the wrinkle area on the raisin surface to the total surface area of the raisin in the optimized image.
4. The evaluation method according to claim 3, characterized in that: During the image acquisition, the raisin samples to be tested do not overlap or adhere to each other.
5. The evaluation method according to claim 4, characterized in that: During the image acquisition, the number of raisin samples to be tested is 30 to 50.
6. The evaluation method according to any one of claims 3 to 5, characterized in that: The raisin samples to be tested are randomly selected.
7. The evaluation method according to claim 3, characterized in that: The removing of the picture background is to use photo shop software or Apowersoft cutout software to remove the picture background.
8. The evaluation method according to claim 3, characterized in that: The sharpening process is to use the Sharpen function under the process menu of Image-J software for 2 to 4 times.
9. The evaluation method according to claim 3, characterized in that: The data were measured and extracted using Image-J software, Image-Pro software or MIPAR software.
10. Application of the evaluation method according to any one of claims 1 to 9 in the quality inspection of raisin products.