Visual sensor-based actinidia arguta leaf screening system

Abstract

The invention relates to the field of actinidia arguta leaf screening, in particular to an actinidia arguta leaf screening system based on a visual sensor, comprising: an optical detection module for acquiring first optical image data and first infrared image data of a target actinidia arguta leaf; the preprocessing module is used for preprocessing the first optical image data and the first infrared image data to generate second optical image data and second infrared image data; through fusion of optical appearance detection and near infrared spectrum internal component analysis, accurate quantitative scoring of the comprehensive value of the leaves is realized, full-dimensional quality grading can be executed, it is ensured that high-end raw materials have both excellent appearance and high-activity components, special high-value leaves generated due to light stress can be intelligently identified, and the method is suitable for large-scale popularization and application. Resource misjudgment is avoided, the efficiency bottleneck is broken through, quality fluctuation is completely eradicated, and meanwhile the raw materials are finely distributed according to the real commercial value of the raw materials.

Description

Technical Field

[0001] This invention relates to the field of screening of hardy kiwi leaves, specifically a screening system for hardy kiwi leaves based on a vision sensor. Background Technology

[0002] Currently, the screening of kiwifruit leaves before processing mainly relies on mechanical screening based on size or color and manual visual sorting. These traditional methods can only perform rough classification based on the external physical characteristics of the leaves, and cannot detect the content of their internal active ingredients (such as flavonoids and polyphenols) in a non-destructive and rapid manner, resulting in inconsistent quality and efficacy of the final product. Meanwhile, manual sorting is inefficient, costly, and inconsistent in standards, while existing automated vision equipment remains at the level of appearance judgment and lacks the ability to identify special high-value raw materials that may be affected by biological stress. Therefore, the industry urgently needs an intelligent screening technology and system that can simultaneously eliminate appearance defects and grade internal quality, in order to overcome the bottleneck of single raw material value assessment and low utilization rate of high-quality resources, and meet the needs of refined processing of high-value-added products. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a vision sensor-based kiwifruit leaf screening system, which solves the technical problems mentioned in the background section.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A vision sensor-based kiwifruit leaf screening system includes an optical detection module, a preprocessing module, an image processing module, and a screening module.

[0006] The optical detection module is used to acquire the first optical image data and the first infrared image data of the target kiwifruit leaf;

[0007] The preprocessing module is used to preprocess the first optical image data and the first infrared image data to generate the second optical image data and the second infrared image data.

[0008] The image processing module is used to calculate the value score of the target kiwifruit leaf based on the second optical image data and the second infrared image data. ;

[0009] The screening module is used to score values. The leaves of the target hardy kiwifruit were screened.

[0010] Furthermore, in the optical detection module, an industrial area array color camera and a near-infrared spectral imaging camera are used to acquire first optical image data and first infrared image data, respectively.

[0011] Furthermore, in the preprocessing module, the specific steps for preprocessing the first optical image data are as follows:

[0012] S111. The dark current correction method is used to perform image correction on the first optical image data to generate the first optical image data to be used.

[0013] S112. The first optical image data to be used is enhanced by contrast stretching method to generate the second optical image data to be used.

[0014] S113. The second optical image data to be used is processed by an edge detection algorithm to generate a binary mask. Morphological operations are performed on the binary mask to generate the second optical image data.

[0015] Furthermore, in the preprocessing module, the specific steps for preprocessing the first infrared image data are as follows:

[0016] S121. The first infrared image data is spectrally corrected using the dark spectrum correction method to generate the first infrared image data to be used.

[0017] S122. The first infrared image data to be used is subjected to spectral processing using the multivariate scattering correction method to generate the second infrared image data to be used.

[0018] S123. Use the Mahalanobis distance method to detect and remove abnormal spectral data to generate second infrared image data.

[0019] Furthermore, the image processing module specifically includes the following steps:

[0020] S21. Calculate the appearance quality score of the target kiwifruit leaf based on the second optical image data. ;

[0021] S22. Calculate the intrinsic quality score of the target kiwifruit leaves based on the second infrared image data. ;

[0022] S23. Calculate the health status correction score of the target kiwifruit leaves. The calculation formula is as follows: ;

[0023] In the formula, and These represent the first and second negative scores, respectively; and Both represent indicator functions;

[0024] S24, Preset appearance weight parameters and intrinsic weight parameters ;

[0025] S25. Classified according to appearance quality and intrinsic qualities Calculate the coupling correction term ;

[0026] S26. Calculate the value score of the target kiwifruit leaves. .

[0027] Furthermore, step S21 specifically includes the following steps:

[0028] S211. Calculate the color of the target kiwifruit leaf based on the second optical image data. The calculation formula is as follows: ;

[0029] In the formula, This represents the average hue of the second optical image data in the HSV color space; The average value representing the saturation of the second optical image data; and They represent the different colors. The first and second weighting coefficients; and They represent about and Table lookup functions;

[0030] S212. Calculate the integrity score of the target kiwifruit leaf. The calculation formula is as follows: ;

[0031] In the formula, =1;

[0032] S213, Calculate the size and specifications of the target kiwifruit leaves. The calculation formula is as follows: ;

[0033] In the formula, express The scoring function;

[0034] S214. Calculate the deduction points for defects in the leaves of the target kiwifruit. The calculation formula is as follows: ;

[0035] In the formula, , and These represent the pixel area ratios of insect spots, spots, and mold spots, respectively. , and These represent the points deducted for defects. The first, second, and third penalty coefficients;

[0036] S215. According to color Completeness score Size and specifications Deductions for defects Calculate appearance quality score .

[0037] Furthermore, step S22 specifically includes the following steps:

[0038] S221. Calculate the predicted total flavonoid content score based on the second infrared image data. The calculation formula is as follows: ;

[0039] In the formula, This indicates the predicted total flavonoid content; express The scoring function;

[0040] S222, Calculate the predicted polyphenol content score of the target kiwifruit leaves. ;

[0041] S223, Scoring based on predicted total flavonoid content Polyphenol Predicted Content Score Calculate intrinsic quality score The calculation formula is as follows: ;

[0042] In the formula, This indicates the intrinsic quality score of the target kiwifruit leaves.

[0043] Furthermore, step S25 specifically includes the following steps:

[0044] S251, Regarding appearance quality and intrinsic qualities Normalization is performed to generate normalized appearance quality scores. and normalized intrinsic quality And calculate the first correction term. The calculation formula is as follows: ;

[0045] In the formula, Indicates the deviation penalty coefficient; Indicates normalized intrinsic quality score The expected function;

[0046] S252. Deduct points based on defects. Calculate the second correction term The calculation formula is as follows: ;

[0047] In the formula, Indicates the metabolic gain reward coefficient; Points will be deducted for defects. Gain start function; Indicates an abnormal gain in intrinsic quality; Represents the gain intensity function;

[0048] S253, Preset top-tier appearance quality threshold and intrinsic top quality threshold And calculate the third correction term based on it. The calculation formula is as follows: ;

[0049] In the formula, Indicates the synergistic multiplication factor; and All are indicator functions;

[0050] S254, according to the first amendment Second Amendment and the third amendment Calculate the coupling correction term .

[0051] Further, in step S26, value scoring The calculation formula is: ;

[0052] In the formula, Used to indicate the value score of the target kiwifruit leaves.

[0053] Compared with the prior art, the present invention provides a vision sensor-based leaf screening system for hardy kiwifruit, which has the following advantages:

[0054] This invention achieves precise quantitative scoring of the comprehensive value of leaves by integrating optical appearance inspection with near-infrared spectral internal component analysis. It can not only perform full-dimensional quality grading to ensure that high-end raw materials have both excellent appearance and high active ingredients, but also intelligently identify special high-value leaves produced by mild stress, avoiding misjudgment of resources. While breaking through efficiency bottlenecks and eliminating quality fluctuations, it can also finely allocate raw materials according to their true commercial value. Attached Figure Description

[0055] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0056] Figure 1This is a schematic diagram of a vision sensor-based kiwifruit leaf screening system according to the present invention. Detailed Implementation

[0057] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding and implementation of how the present application uses technical means to solve technical problems and achieve technical effects.

[0058] Those skilled in the art will understand that all or part of the steps in the methods of the following embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0059] The hardy kiwifruit, also known as the heart-leaf kiwifruit or purple kiwifruit, is a large deciduous vine belonging to the genus Actinidia in the family Actinidiaceae. After harvesting, the leaves of the hardy kiwifruit can be used for tea and ingredient extraction. However, before processing into tea or extracts, the content and composition of effective active ingredients (such as flavonoids, polyphenols, and polysaccharides) vary significantly depending on the leaf's growth stage (young leaves, mature leaves, old leaves) and the location of the leaf. Young leaves typically contain higher levels of active substances, but may also have higher levels of irritating components such as tannins; older leaves are more fibrous, and the effective ingredient content may decrease. Mixed processing can lead to batch-to-batch inconsistencies in the quality of the final product (such as tea or extracts). In addition, leaves of different sizes and maturity have different cell wall structures, water content, and drying rates. Leaves of uniform size ensure even heating and consistent reactions during processes such as blanching, rolling, drying, crushing, and extraction, thus guaranteeing stable product color, flavor, and extraction efficiency. Therefore, it is necessary to sieve the leaves of the hardy kiwifruit. However, the common method of manual sieving is not only labor-intensive but also prone to misclassification. Therefore, please refer to [the relevant documentation / reference needed]. Figure 1 As shown, this invention proposes a sieving system for leaves of hardy kiwifruit based on a vision sensor, comprising: an optical detection module, a preprocessing module, an image processing module, and a sieving module;

[0060] The optical detection module is used to acquire the first optical image data and the first infrared image data of the target kiwifruit leaves. Specifically, the intelligent screening of kiwifruit leaves using vision + near-infrared spectroscopy is currently the most advanced and effective industrial solution. In the optical detection module, an industrial area array color camera and a near-infrared spectral imaging camera are used to acquire the first optical image data and the first infrared image data, respectively. Specifically, the industrial area array color camera is an industrial camera based on a CMOS sensor, which is cost-effective, fast, and has low power consumption. The near-infrared spectral imaging camera is a near-infrared spectral imaging system. It should be noted that both the industrial camera based on a CMOS sensor and the near-infrared spectral imaging system are mature existing technologies and will not be elaborated here.

[0061] The preprocessing module is used to preprocess the first optical image data and the first infrared image data to generate second optical image data and second infrared image data; specifically, the preprocessing steps for the first optical image data in the preprocessing module are as follows:

[0062] S111. The first optical image data is corrected using the dark current correction method to generate the first optical image data to be used. Specifically, the dark current correction method involves taking a "dark field" image under completely dark conditions and subtracting the dark field from the first optical image data to eliminate the inherent thermal noise of the industrial camera based on the CMOS sensor. It should be noted that the dark current correction method is a common existing technology and will not be described in detail here.

[0063] S112. The first optical image data to be used is enhanced by the contrast stretching method to generate the second optical image data to be used. Specifically, the contrast stretching method is used to adjust the grayscale range of the image to make the contrast between the target kiwifruit leaf and the background and the different color areas on the leaf more distinct. It should be noted that the contrast stretching method is a common existing technology and will not be described in detail here.

[0064] S113. The second optical image data to be used is processed by an edge detection algorithm to generate a binary mask, and morphological operations are performed on the binary mask to generate the second optical image data. Specifically, the edge detection algorithm and morphological operations are common existing technologies, and will not be elaborated here.

[0065] The specific steps for preprocessing the first infrared image data in the preprocessing module are as follows:

[0066] S121. The first infrared image data is spectrally corrected using the dark spectrum correction method to generate the first infrared image data to be used. Specifically, the dark spectrum correction method is similar to the dark current correction method. The light source is turned off, and the spectrum of ambient light and detector noise is collected as a dark reference and subtracted from the first infrared image data. It should be noted that the dark spectrum correction method is an existing technology and will not be described in detail here.

[0067] S122. The first infrared image data to be used is processed by the multivariate scattering correction method to generate the second infrared image data to be used. Specifically, the multivariate scattering correction method calculates the average spectrum of all spectra, regresses the spectrum of each sample with the average spectrum, and corrects the scattering effect. It should be noted that the multivariate scattering correction method is a common existing technology and will not be described in detail here.

[0068] S123. The Mahalanobis distance method is used to detect and remove abnormal spectral data to generate second infrared image data; specifically, the Mahalanobis distance method is a common existing technology and will not be elaborated here.

[0069] The image processing module is used to calculate the value score of the target kiwifruit leaf based on the second optical image data and the second infrared image data. Specifically, common methods for evaluating the value of kiwifruit leaves only consider the value of optical and infrared images, without taking into account the interaction between external and internal factors (i.e., optical and infrared images). For example: 1. In the early stages of the color change from "light green" to "dark green," the flavonoid content may increase, but in the "yellow-green" aging stage, the flavonoid content will decrease. Therefore, there is a non-linear functional relationship between appearance score and internal score; 2. A leaf with a slight insect bite may accumulate a higher concentration of polyphenols around the wound as a defense mechanism. In this case, a simple "appearance score + internal score" may underestimate its value. Therefore, it is not possible to simply calculate the external and internal scores and add them together. To address this, the image processing module includes the following steps:

[0070] S21. Calculate the appearance quality score of the target kiwifruit leaf based on the second optical image data. Specifically, step S21 includes the following steps:

[0071] S211. Calculate the color of the target kiwifruit leaf based on the second optical image data. The calculation formula is as follows: ;

[0072] In the formula, This represents the average hue of the second optical image data in the HSV color space; The average value representing the saturation of the second optical image data; and They represent the different colors. The first and second weighting coefficients; and They represent about and The lookup table function; specifically, and The values ​​are 0.7 and 0.3 respectively; The possible values ​​are shown in Table 1; The values ​​are shown in Table 2;

[0073] Table 1

[0074] Table 2

[0075] S212. Calculate the integrity score of the target kiwifruit leaf. The calculation formula is as follows: ;

[0076] In the formula, =1;

[0077] S213, Calculate the size and specifications of the target kiwifruit leaves. The calculation formula is as follows: ;

[0078] In the formula, express The scoring function; specifically, The values ​​are shown in Table 3;

[0079] Table 3

[0080] S214. Calculate the deduction points for defects in the leaves of the target kiwifruit. The calculation formula is as follows: ;

[0081] In the formula, , and These represent the pixel area ratios of insect spots, spots, and mold spots, respectively. , and These represent the points deducted for defects. The first, second, and third penalty coefficients; specifically, , and The values ​​are 0.1, 0.2, and 0.3 respectively;

[0082] S215. According to color Completeness score Size and specifications Deductions for defects Calculate appearance quality score The calculation formula is as follows: ;

[0083] In the formula, The appearance quality score represents the appearance of the leaves of the target hardy kiwifruit.

[0084] S22. Calculate the intrinsic quality score of the target kiwifruit leaves based on the second infrared image data. Specifically, step S22 includes the following steps:

[0085] S221. Calculate the predicted total flavonoid content score based on the second infrared image data. The calculation formula is as follows: ;

[0086] In the formula, This indicates the predicted total flavonoid content; express The scoring function; specifically, because the chemical bonds inside the leaves of the hardy kiwifruit (such as OH, CH, NH) absorb light of specific wavelengths, the detector in the instrument (commonly indium gallium arsenide) receives the light reflected back from the leaves. The detector converts the light signal into an electrical signal, which is then processed into a spectrum. The horizontal axis of this spectrum is wavelength, and the vertical axis is absorbance. From 900nm to 1700nm, a point is taken in every 1nm interval, resulting in 801 wavelength points. The absorbance values ​​corresponding to each wavelength point are arranged in wavelength order, resulting in a sequence list of 801 numbers, i.e., the original spectral vector. ; For the original spectral vector Preprocessing is performed to generate standard spectral vectors. , ; The method for obtaining the standard spectral vector is as follows: collect n leaves of the hardy kiwifruit tree, and obtain the standard spectral vector for each leaf. The total flavonoid content of each kiwifruit leaf was tested. We obtain n pairs of data: Partial least squares regression was used to fit n pairs of data to obtain the constant term. and model coefficients The standard spectral vector of the target kiwifruit leaves was input into partial least squares regression to obtain the predicted total flavonoid content. ; The values ​​are shown in Table 4;

[0087] Table 4

[0088] S222, Calculate the predicted polyphenol content score of the target kiwifruit leaves. Specifically, the predicted polyphenol content score. The calculation method is the same as the total flavonoid content prediction score. Exactly similar;

[0089] S223, Scoring based on predicted total flavonoid content Polyphenol Predicted Content Score Calculate intrinsic quality score The calculation formula is as follows: ;

[0090] In the formula, This indicates the intrinsic quality score of the target kiwifruit leaves.

[0091] S23. Calculate the health status correction score of the target kiwifruit leaves. The calculation formula is as follows: ;

[0092] In the formula, and These represent the first and second negative scores, respectively; and Both represent indicator functions; specifically, and The values ​​are -1000 and -800 respectively; whether there is severe internal mold and abnormally high moisture content in the leaves of the hardy kiwifruit can be determined by near-infrared spectroscopy. If severe internal mold is present in the leaves of the hardy kiwifruit, then... If the moisture content of the leaves of the hardy kiwifruit is abnormally high, then... It is 1 if it is true, otherwise it is 0.

[0093] S24, Preset appearance weight parameters and intrinsic weight parameters Specifically, appearance weight parameters and intrinsic weight parameters Certified by relevant technical experts;

[0094] S25. Classified according to appearance quality and intrinsic qualities Calculate the coupling correction term Specifically, step S25 includes the following steps:

[0095] S251, Regarding appearance quality and intrinsic qualities Normalization is performed to generate normalized appearance quality scores. and normalized intrinsic quality And calculate the first correction term. The calculation formula is as follows: ;

[0096] In the formula, Indicates the deviation penalty coefficient; Indicates normalized intrinsic quality score The expected function; specifically, Certified by relevant technical experts; The values ​​are shown in Table 5;

[0097] Table 5

[0098] S252. Deduct points based on defects. Calculate the second correction term The calculation formula is as follows: ;

[0099] In the formula, Indicates the metabolic gain reward coefficient; Points will be deducted for defects. Gain start function; Indicates an abnormal gain in intrinsic quality; This represents the gain intensity function; specifically, , This represents the defect sensitivity attenuation coefficient. The value was determined by relevant technical experts; , , Indicates the gain saturation coefficient. The value was determined by relevant technical experts;

[0100] S253, Preset top-tier appearance quality threshold and intrinsic top quality threshold And calculate the third correction term based on it. The calculation formula is as follows: ;

[0101] In the formula, Indicates the synergistic multiplication factor; and All are indicator functions; specifically, , and All were calibrated by relevant technical experts; if If it is established, then The value is 1 if it is not 0 otherwise; If it is established, then The value is 1 if it is set to 1, and 0 otherwise.

[0102] S254, according to the first amendment Second Amendment and the third amendment Calculate the coupling correction term The calculation formula is as follows: ;

[0103] S26. Calculate the value score of the target kiwifruit leaves. Specifically, in step S26, value scoring... The calculation formula is: ;

[0104] In the formula, Used to indicate the value score of the target kiwifruit leaves.

[0105] Traditional methods for screening the leaves of hardy kiwifruit rely on size-measuring sieves, manual color sorting, or simple camera identification of color / size. These methods can only remove obviously inferior products and cannot determine the content of the internal effective ingredients.

[0106] This invention achieves precise quantitative scoring of the comprehensive value of leaves by integrating optical appearance inspection with near-infrared spectral internal component analysis. It can not only perform full-dimensional quality grading to ensure that high-end raw materials have both excellent appearance and high activity, but also intelligently identify special high-value leaves produced by mild stress, avoiding misjudgment of resources. While breaking through efficiency bottlenecks and eliminating quality fluctuations, it can finely allocate raw materials according to their true commercial value, thereby completely realizing the industrial upgrade from "extensive rejection" to "value-based classification", providing key technical support for improving product competitiveness and raw material utilization.

[0107] The screening module is used to score values. The target kiwifruit leaves were screened; specifically, the value score was divided into several scoring levels according to the requirements, and the value score was then used for further analysis. The target kiwifruit leaves are classified into corresponding grades and then selected for the corresponding production line to complete the screening.

[0108] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A vision sensor-based leaf sieving system for kiwifruit, characterized in that, It includes an optical detection module, a preprocessing module, an image processing module, and a sieving module; The optical detection module is used to acquire the first optical image data and the first infrared image data of the target kiwifruit leaf; The preprocessing module is used to preprocess the first optical image data and the first infrared image data to generate the second optical image data and the second infrared image data. The image processing module is used to calculate the value score of the target kiwifruit leaf based on the second optical image data and the second infrared image data. ; The screening module is used to score values. The leaves of the target hardy kiwifruit were screened.

2. The vision sensor-based leaf sieving system for hardy kiwifruit according to claim 1, characterized in that, In the optical detection module, an industrial area array color camera and a near-infrared spectral imaging camera are used to acquire first optical image data and first infrared image data, respectively.

3. The vision sensor-based leaf sieving system for hardy kiwifruit according to claim 1, characterized in that, The specific steps for preprocessing the first optical image data in the preprocessing module are as follows: S111. The dark current correction method is used to perform image correction on the first optical image data to generate the first optical image data to be used. S112. The first optical image data to be used is enhanced by contrast stretching method to generate the second optical image data to be used. S113. The second optical image data to be used is processed by an edge detection algorithm to generate a binary mask. Morphological operations are performed on the binary mask to generate the second optical image data.

4. The vision sensor-based leaf sieving system for hardy kiwifruit according to claim 1, characterized in that, The specific steps for preprocessing the first infrared image data in the preprocessing module are as follows: S121. The first infrared image data is spectrally corrected using the dark spectrum correction method to generate the first infrared image data to be used. S122. The first infrared image data to be used is subjected to spectral processing using the multivariate scattering correction method to generate the second infrared image data to be used. S123. Use the Mahalanobis distance method to detect and remove abnormal spectral data to generate second infrared image data.

5. The vision sensor-based leaf sieving system for hardy kiwifruit according to claim 1, characterized in that, The image processing module specifically includes the following steps: S21. Calculate the appearance quality score of the target kiwifruit leaf based on the second optical image data. ; S22. Calculate the intrinsic quality score of the target kiwifruit leaves based on the second infrared image data. ; S23. Calculate the health status correction score of the target kiwifruit leaves. The calculation formula is as follows: ; In the formula, and These represent the first and second negative scores, respectively; and Both represent indicator functions; S24, Preset appearance weight parameters and intrinsic weight parameters ; S25. Classified according to appearance quality and intrinsic qualities Calculate the coupling correction term ; S26. Calculate the value score of the target kiwifruit leaves. .

6. The vision sensor-based leaf sieving system for kiwifruit as described in claim 5, characterized in that, Step S21 specifically includes the following steps: S211. Calculate the color of the target kiwifruit leaf based on the second optical image data. The calculation formula is as follows: ; In the formula, This represents the average hue of the second optical image data in the HSV color space; The average value representing the saturation of the second optical image data; and They represent the different colors. The first and second weighting coefficients; and They represent about and Table lookup functions; S212. Calculate the integrity score of the target kiwifruit leaf. The calculation formula is as follows: ; In the formula, =1; S213, Calculate the size and specifications of the target kiwifruit leaves. The calculation formula is as follows: ; In the formula, express The scoring function; S214. Calculate the deduction points for defects in the leaves of the target kiwifruit. The calculation formula is as follows: ; In the formula, , and These represent the pixel area ratios of insect spots, spots, and mold spots, respectively. , and These represent the points deducted for defects. The first, second, and third penalty coefficients; S215. According to color Completeness score Size and specifications Deductions for defects Calculate appearance quality score .

7. The vision sensor-based leaf sieving system for hardy kiwifruit according to claim 5, characterized in that, Step S22 specifically includes the following steps: S221. Calculate the predicted total flavonoid content score based on the second infrared image data. The calculation formula is as follows: ; In the formula, This indicates the predicted total flavonoid content; express The scoring function; S222, Calculate the predicted polyphenol content score of the target kiwifruit leaves. ; S223, Scoring based on predicted total flavonoid content Polyphenol Predicted Content Score Calculate intrinsic quality score The calculation formula is as follows: ; In the formula, This indicates the intrinsic quality score of the target kiwifruit leaves.

8. The vision sensor-based leaf sieving system for kiwifruit as described in claim 5, characterized in that, Step S25 specifically includes the following steps: S251, Regarding appearance quality and intrinsic qualities Normalization is performed to generate normalized appearance quality scores. and normalized intrinsic quality And calculate the first correction term. The calculation formula is as follows: ; In the formula, Indicates the deviation penalty coefficient; Indicates normalized intrinsic quality score The expected function; S252. Deduct points based on defects. Calculate the second correction term The calculation formula is as follows: ; In the formula, Indicates the metabolic gain reward coefficient; Points will be deducted for defects. Gain start function; Indicates an abnormal gain in intrinsic quality; Represents the gain intensity function; S253, Preset top-tier appearance quality threshold and intrinsic top quality threshold And calculate the third correction term based on it. The calculation formula is as follows: ; In the formula, Indicates the synergistic multiplication factor; and All are indicator functions; S254, according to the first amendment Second amendment and the third amendment Calculate the coupling correction term .

9. The vision sensor-based leaf sieving system for kiwifruit as described in claim 1, characterized in that, In step S26, value scoring The calculation formula is: ; In the formula, Used to indicate the value score of the target kiwifruit leaves.

Images