High-precision nondestructive testing method and device for sugar degree of pineapple sorting

Abstract

The invention discloses a sugar degree high-precision nondestructive testing method and device for pineapple sorting, and the method comprises the steps: for a certain pineapple variety, obtaining a corresponding pineapple subinterval spectral data sample set and a pineapple subinterval local sugar degree value sample set according to different pineapple fruit measurement subintervals; giving an interval fuzzy preference relationship of different prediction models to pineapple fruit measurement subintervals, and aggregating the interval fuzzy preference relationship into an accurate value by using an importance weight continuous generalized ordered weighted average operator; determining a pineapple fruit recommendation subinterval and a recommendation prediction model for pineapple sorting; according to the sugar degree high-precision nondestructive testing method and device for pineapple sorting, the near infrared spectrum data in the pineapple fruit recommendation subintervals in the sorting process are processed by constructing the global sugar degree prediction model, and therefore sugar degree high-precision nondestructive testing is provided for detected pineapples.

Description

Technical Field

[0001] This invention relates to the field of fruit internal quality testing technology, and in particular to a high-precision non-destructive testing method and apparatus for sugar content detection in pineapple sorting. Background Technology

[0002] Pineapple, a terrestrial herbaceous plant belonging to the genus *Ananas* of the family Bromeliaceae, is one of the world's major tropical fruits, widely cultivated in tropical regions. It boasts a sweet and delicious flavor and is rich in nutrients. Besides being sold fresh, its processing prospects for high-quality juice, jam, and frozen fruit are promising. Sugar content testing of pineapple fruit is a key method for assessing fruit ripeness, quality, and taste, and is crucial for ensuring the quality and market competitiveness of pineapple products. Therefore, pineapple sugar content testing helps ensure the quality, ripeness, and market competitiveness of the fruit, and is of great significance for promoting the sustainable development of the pineapple industry.

[0003] In recent years, online near-infrared spectroscopy has gained increasing attention in the field of non-destructive testing and sorting of fruit quality. Online fruit testing systems are characterized by high efficiency and automation, enabling rapid real-time monitoring using near-infrared spectrometers, and are currently developing rapidly towards automation and intelligence. While near-infrared spectroscopy has achieved good results in detecting internal quality parameters such as sugar content, acidity, and firmness in fruits, fruits, as complex natural products, exhibit variations in chemical composition and physical properties across different parts of the fruit. These sample correlation factors pose challenges to the robustness of internal quality testing models. During pineapple growth, factors such as light conditions and soil moisture lead to different sugar content distributions in different parts of the fruit. Therefore, neglecting the testing location during internal quality non-destructive testing can significantly impact sugar content detection, resulting in inaccurate quality grading. Finding the recommended testing location for pineapple sugar content is crucial for improving the accuracy of non-destructive testing. Thus, how to quickly, accurately, and stably detect the sugar content of pineapple is a key issue that needs to be addressed in the current development of the pineapple industry.

[0004] Considering the influence of irrelevant information in the Raman spectra acquired by near-infrared spectrometers, leading to baseline drift and spectral overlap, and interfering with the stability of spectral modeling, spectral preprocessing algorithms can be used to eliminate errors in near-infrared diffuse reflectance spectra caused by light scattering. Due to the large amount of spectral feature data, the significant collinearity in wavelength variables can affect the model's prediction accuracy and stability. Wavelength variable screening methods can be used to extract effective information from the original full-band spectrum and optimize the original data. Because near-infrared spectroscopy suffers from problems such as spectral bandwidth, severe overlap, weak absorption signals, and complex information analysis, spectral data modeling methods can be used to predict the near-infrared spectra of unknown samples to obtain predicted values ​​for various components. Therefore, different spectral preprocessing algorithms, wavelength variable screening methods, and spectral data modeling methods can often be combined to form a prediction model, thereby achieving non-destructive testing of sugar content. However, determining the optimal detection location and the best prediction model for non-destructive testing of pineapple sugar content is a key research focus in achieving non-destructive, rapid, short-cycle, and high-precision online detection during pineapple sorting.

[0005] Fuzzy multi-attribute decision-making methods can utilize uncertain or fuzzy information in the decision-making process, improving the scientific rigor and practicality of multi-attribute decision-making. An effective and appropriate aggregation operator ensures that attribute values ​​are integrated without information loss, thus reflecting the decision-making effect. In solving decision problems, the decision-maker's judgment information is given in the form of interval numbers, and an importance weight function can be defined to express the decision-maker's preference for data within each interval. The Importance Weighted Continuous Generalized Ordered Weighted Averaging (IW-CGOWA) operator, as an averaging aggregation operator, can integrate interval data by aggregating information and can improve prediction accuracy in uncertain multi-attribute decision-making applications. Summary of the Invention

[0006] The purpose of this invention is to provide a high-precision non-destructive testing method and device for pineapple sorting. By using fuzzy multi-attribute decision-making to find a suitable detection range and prediction model, the sugar content of pineapples can be detected, achieving the effect of rapid, accurate and stable high-precision detection of pineapple fruit sugar content, thereby providing an internal quality testing method for pineapple fruit sorting.

[0007] The first embodiment of the present invention provides a high-precision non-destructive testing method for sugar content detection in pineapple sorting, comprising:

[0008] For a certain pineapple variety, after dividing the pineapple fruit measurement area into different pineapple fruit measurement sub-intervals, we obtain the spectral data sample set and local sugar content sample set of the pineapple sub-intervals corresponding to the different pineapple fruit measurement sub-intervals, and obtain the global sugar content sample set of the pineapple.

[0009] By comparing the degree of preference of different prediction models in the pineapple sugar content prediction model set for different pineapple fruit measurement sub-intervals with respect to accuracy, the interval fuzzy preference relationship of different prediction models for pineapple fruit measurement sub-intervals is given;

[0010] The importance weight continuous generalized ordered weighted average operator is used to aggregate the preference degree in the interval fuzzy preference relationship of different prediction models for pineapple fruit measurement sub-intervals from the interval number into an accurate value;

[0011] The weight coefficients of different prediction models for pineapple fruit measurement sub-intervals in the pineapple sugar content prediction model set are given, the evaluation index of different pineapple fruit measurement sub-intervals is obtained, and the recommended sub-intervals and recommended prediction models for pineapple sorting are determined. In this way, a global sugar content prediction model is established to provide high-precision non-destructive detection of sugar content for pineapple sorting.

[0012] Optionally, in the first embodiment of the present invention, for a certain pineapple variety, after dividing the pineapple fruit measurement area into different pineapple fruit measurement sub-intervals, obtaining the spectral data sample set of the pineapple sub-intervals and the local sugar content value sample set of the pineapple sub-intervals corresponding to the different pineapple fruit measurement sub-intervals, and obtaining the global sugar content value sample set of the pineapple, specifically includes:

[0013] For a specific pineapple variety, when performing non-destructive testing of the pineapple's internal quality using a near-infrared spectrometer, the measurement area for the pineapple fruit is defined as a rectangular region excluding the core area, with the height of the pineapple fruit as the height and the minimum diameter of the pineapple fruit as the width. The total number of sub-intervals for the pineapple fruit measurement is then set. After that, the pineapple can be identified. Regarding the measurement sub-interval of pineapple fruit The scope is:

[0014] A k , l = { ( i k , j k ) | i k ∈ [ − M k 2 , − N k 2 ] ∪ [ N k 2 , M k 2 ], j k ∈ [ ( l − 1 ) ⋅ H k L , l ⋅ H k L ] }

[0015] in, Number the sub-intervals for measuring pineapple fruits. This refers to the numbering of the pineapple samples used for sugar content testing. Pineapple samples facing the near-infrared spectrometer The x and y coordinates are set, and the following settings are provided. At this time, it is at the bottom center of the pineapple fruit. Pineapple The maximum diameter of the core, the minimum diameter of the fruit, and the height of the fruit;

[0016] By extracting different pineapple samples from the pineapple fruit measurement sub-region Near-infrared spectral data within the specified range can be used to obtain measurement sub-intervals for pineapple fruits. The corresponding spectral data sample set for pineapple seeds is as follows:

[0017]

[0018] in, This represents the total number of samples of this pineapple variety used for sugar content testing. To extract pineapple samples When facing a near-infrared spectrometer, the detection range is: Near-infrared spectral data;

[0019] By measuring different pineapple samples in the pineapple fruit measurement sub-region The sugar content value can be used to obtain the measurement sub-interval of the pineapple fruit. The corresponding local sugar content sample set for pineapple seeds is as follows:

[0020]

[0021] in, The cutting height for measuring the sugar content of pineapple. For pineapple samples The height of the fruit cut from the middle is The sugar content value was obtained by measuring the pulp.

[0022] By removing the core from different pineapple samples, juicing them whole, and then measuring the sugar content, a global sugar content sample set of pineapples can be obtained:

[0023]

[0024] in, For pineapple samples The sugar content value was measured after the whole fruit was juiced after removing the core;

[0025] Optionally, in the first embodiment of the present invention, the comparison of the degree of preference of different prediction models in the pineapple sugar content prediction model set for different pineapple fruit measurement sub-intervals with respect to accuracy, and the provision of the interval fuzzy preference relationship of different prediction models for pineapple fruit measurement sub-intervals, specifically includes:

[0026] Set up a set of pineapple sugar content prediction models Among them, the number is Prediction model It can be composed of a combination of spectral preprocessing algorithms, wavelength variable selection methods, and spectral data modeling methods. Given the number of available prediction models, when using a prediction model... A certain pineapple fruit measurement sub-interval Measurement sub-intervals of other pineapple fruits The degree of preference can be expressed as:

[0027] p ˜ l , l ′ t = [ minutes 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) ), max 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )]

[0028] in, Number the sub-intervals for measuring other pineapple fruits. ,when hour, p ˜ l , l ′ t = [ 0 . 5 , 0 . 5 ] , To pass the above and Can the use of a prediction model be given? Time measurement sub-intervals of pineapple fruits The accuracy in establishing a local sugar content prediction model, as described in this invention, can be represented by the probability that all samples are correctly classified in the model evaluation method;

[0029] This leads to the use of predictive models. The interval fuzzy preference relation for measuring different pineapple fruits in the sub-intervals is as follows:

[0030] A ˜ t = ( p ˜ l , l ′ t ) L × L = ([ minutes 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) ), max 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )]) L × L

[0031] Optionally, in the first embodiment of the present invention, the step of using the importance weight continuous generalized ordered weighted average (IW-CGOWA) operator to aggregate the preference degree in the interval fuzzy preference relationship of different prediction models for the measurement sub-intervals of pineapple fruit from the number of intervals into an accurate value specifically includes:

[0032] For the IW-CGOWA operator F Q ([ a , b ], f 0 ) = [ ∫ 0 1 dQ ( y ) y ( H − 1 ( y )) l day ] 1 l ,in, The main function symbol can be set as follows: For set [ a , b ] Non-negative functions on, The upper limit of the set, The lower bound of the set, As the independent variable, For the power exponent, x ∈ [ a , b ] ,like Then the variable , and For function, As an intermediate variable, It is a monotonic function of the basic unit interval; set , Importance weight function Then the IW-CGOWA operator F Q ([ a , b ], f 0 ) It can be represented as [( a 2 / 3 ) + ( 2 b 2 / 3 )] 1 / 2 ;

[0033] When the IW-CGOWA operator F Q ([ a , b ], f 0 ) for [( a 2 / 3 ) + ( 2 b 2 / 3 )] 1 / 2 At that time, through the IW-CGOWA operator F Q ([ a , b ], f 0 ) The preference levels in the interval fuzzy preference relations of different pineapple fruit measurement sub-intervals can be aggregated from the number of intervals into precise values, and then a prediction model can be used. A certain pineapple fruit measurement sub-interval Measurement sub-intervals of other pineapple fruits The precise value of the degree of preference can be expressed as:

[0034] q l , l ′ t = {[ minutes 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )] 2 / 3 + 2 [ max 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )] 2 / 3 } 1 / 2

[0035] Therefore, after the interval numbers are converted into exact values ​​by the IW-CGOWA operator, the prediction model is used. The interval fuzzy preference relation for measuring sub-intervals of pineapple fruit can be expressed as:

[0036] A t = ( q l , l ′ t ) L × L = ({[ minutes 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )] 2 / 3 + 2 [ max 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )] 2 / 3 } 1 / 2 ) L × L

[0037] Optionally, in the first embodiment of the present invention, the step of providing the weight coefficients of different prediction models in the pineapple sugar content prediction model set for the pineapple fruit measurement sub-interval, obtaining the evaluation index of different pineapple fruit measurement sub-intervals, determining the recommended pineapple fruit sub-intervals and recommended prediction models for pineapple sorting, thereby establishing a global sugar content prediction model to provide high-precision non-destructive detection of sugar content for pineapple sorting, specifically includes:

[0038] For the prediction models in the pineapple sugar content prediction model set The above can be given The weighting coefficients for the pineapple fruit measurement sub-intervals are:

[0039] w t = ∑ 1 ≤ l ≤ L [ 1 − m ( l , t )] − β / ∑ 1 ≤ t ′ ≤ t ∑ 1 ≤ l ≤ L [ 1 − m ( l , t ' )] − β

[0040] in, As an intermediate variable, The weighting coefficient is the power exponent;

[0041] Through the above and The measurement sub-intervals of pineapple fruits can be obtained. The evaluation indicators are:

[0042]

[0043] Therefore, Time corresponding This can be set as the recommended sub-interval number for pineapple fruit used for sugar content detection. ;

[0044] Considering the robustness of the prediction model for detecting the entire pineapple region, Time corresponding This can be set as the number of the recommended prediction model for sugar content detection. :

[0045] Therefore, recommended sub-intervals for pineapple fruits can be extracted during pineapple sorting. The near-infrared spectral data within the range, and then the... Corresponding recommendation prediction model Through the above and To construct a global sugar content prediction model, thereby providing high-precision non-destructive detection of sugar content for pineapple sorting;

[0046] A second embodiment of the present invention provides a high-precision non-destructive testing device for sugar content detection in pineapple sorting, comprising:

[0047] The sample data fusion module is used to input the spectral data sample set of pineapple sub-intervals, the local sugar content value sample set of pineapple sub-intervals, and the global sugar content value sample set of pineapple. By setting the measurement area of ​​pineapple fruit and the total number of pineapple fruit measurement sub-intervals, it provides recommended sub-intervals and recommended prediction models for pineapple fruit sugar content detection.

[0048] The spectral data extraction module is used to extract near-infrared spectral data of the pineapple fruit within a recommended sub-interval based on the ratio of the maximum diameter of the core, the minimum diameter of the fruit, and the height of the fruit to the near-infrared spectral image after the near-infrared spectrometer has acquired the near-infrared spectral image of the pineapple being tested.

[0049] The sugar content non-destructive testing module is used to construct a global sugar content prediction model by using the spectral preprocessing algorithm, wavelength variable screening method, and spectral data modeling method in the recommended prediction model, as well as the spectral data sample set of the pineapple sub-interval corresponding to the recommended sub-interval of the pineapple fruit and the global sugar content value sample set of the pineapple. Then, the global sugar content prediction model processes the near-infrared spectral data of the pineapple fruit in the recommended sub-interval during the sorting process to give the sugar content value of the inspected pineapple. Attached Figure Description

[0050] Figure 1 This is a schematic flowchart of a high-precision non-destructive testing method for sugar content in pineapple sorting provided by an embodiment of the present invention;

[0051] Figure 2 This is a schematic diagram illustrating how a pineapple fruit measurement area is divided into pineapple fruit measurement sub-intervals according to an embodiment of the present invention.

[0052] Figure 3 This is a schematic diagram of a high-precision non-destructive testing device for sugar content in pineapple sorting provided in an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the present invention will be further described below with reference to the accompanying drawings in the embodiments of the present invention.

[0055] Please see Figure 1 and Figure 2 In the first embodiment of the present invention, as follows is provided Figure 1 The method shown is a high-precision non-destructive testing method for sugar content in pineapple sorting, comprising:

[0056] S1. For a certain pineapple variety, after dividing the pineapple fruit measurement area into different pineapple fruit measurement sub-intervals, obtain the spectral data sample set of the pineapple sub-intervals and the local sugar content sample set of the pineapple sub-intervals corresponding to the different pineapple fruit measurement sub-intervals, and obtain the global sugar content sample set of the pineapple.

[0057] In this embodiment of the invention, when facing a near-infrared spectrometer during non-destructive testing of the internal quality of a fragrant pineapple, a rectangular area excluding the core region, with the fruit height as the height and the minimum diameter of the fruit as the width, is set as the pineapple fruit measurement area. The total number of pineapple fruit measurement sub-intervals is set. After setting it to 5, the measurement area for the pineapple fruit can be divided into: Figure 2 The pineapple fruit measurement sub-intervals shown can be used to determine the pineapple's size. Regarding the measurement sub-interval of pineapple fruit The scope is:

[0058] A k , l = { ( i k , j k ) | i k ∈ [ − M k 2 , − N k 2 ] ∪ [ N k 2 , M k 2 ], j k ∈ [ ( l − 1 ) ⋅ H k L , l ⋅ H k L ] }

[0059] in, Number the sub-intervals for measuring pineapple fruits. This refers to the numbering of the pineapple samples used for sugar content testing. Pineapple samples facing the near-infrared spectrometer The x and y coordinates are set, and the following settings are provided. At this time, it is at the bottom center of the pineapple fruit. Pineapple The maximum diameter of the core, the minimum diameter of the fruit, and the height of the fruit;

[0060] The ENVI remote sensing image processing software can be used to extract the measurement sub-regions of pineapple fruits from different pineapple samples. Near-infrared spectral data within the specified range can be used to obtain measurement sub-intervals for pineapple fruits. The corresponding spectral data sample set for pineapple seeds is as follows:

[0061]

[0062] The total number of samples of this pineapple variety used for sugar content testing It can be set to 300. To extract pineapple samples When facing a near-infrared spectrometer, the detection range is: Near-infrared spectral data;

[0063] By measuring different pineapple samples in the pineapple fruit measurement sub-region The sugar content value can be used to obtain the measurement sub-interval of the pineapple fruit. The corresponding local sugar content sample set for pineapple seeds is as follows:

[0064]

[0065] in, The cutting height for measuring the sugar content of pineapple. For pineapple samples The height of the fruit cut from the middle is The sugar content value was obtained by measuring the pulp.

[0066] By removing the core from different pineapple samples, juicing them whole, and then measuring the sugar content, a global sugar content sample set of pineapples can be obtained:

[0067]

[0068] in, For pineapple samples The sugar content value was measured after the whole fruit was juiced after removing the core;

[0069] S2. Compare the degree of preference of different prediction models in the pineapple sugar content prediction model set for different pineapple fruit measurement sub-intervals with respect to accuracy, and give the interval fuzzy preference relationship of different pineapple fruit measurement sub-intervals.

[0070] In this embodiment of the invention, a set of pineapple sugar content prediction models is set. Let the number of available prediction models be... The value is 3, where the prediction model is 3. The prediction model is composed of a spectral preprocessing algorithm (BC, Baseline Correction), a wavelength variable selection method (CARS, Competitive Adaptive Reweighted Sampling), and a spectral data modeling method (PLSR, Partial Least Squares Regression). The prediction model is composed of a combination of the spectral preprocessing algorithm MSC (Multiplicative Scatter Correction), the wavelength variable elimination method UVE (Uninformative Variable Elimination), and the spectral data modeling method LS-SVM (Least Squares Support Vector Machines). It can be composed of the spectral preprocessing algorithm MSC, the wavelength variable screening method CARS, and the spectral data modeling method LSSVM;

[0071] When using prediction models A certain pineapple fruit measurement sub-interval Measurement sub-intervals of other pineapple fruits The degree of preference can be expressed as:

[0072] p ˜ l , l ′ t = [ minutes 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) ), max 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )]

[0073] in, Number the sub-intervals for measuring other pineapple fruits. ,when hour, p ˜ l , l ′ = [ 0 . 5 , 0 . 5 ] , To pass the above and Can the use of a prediction model be given? Time measurement sub-intervals of pineapple fruits Accuracy in establishing a local sugar content prediction model;

[0074] By analyzing and comparing the accuracy of different pineapple fruit measurement sub-intervals using different prediction models, the corresponding interval fuzzy preference relationship can be given as follows:

[0075] A ˜ 1 = ( [ 0 . 5 , 0 . 5 ] [ 0 . 31 , 0 . 64 ] [ 0 . 42 , 0 . 53 ] [ 0 . 12 , 0 . 45 ] [ 0 . 47 , 0 . 58 ] [ 0 . 27 , 0 . 34 ] [ 0 . 5 , 0 . 5 ] [ 0 . 34 , 0 . 75 ] [ 0 . 27 , 0 . 54 ] [ 0 . 69 , 0 . 85 ] [ 0 . 49 , 0 . 75 ] [ 0 . 15 , 0 . 76 ] [ 0 . 5 , 0 . 5 ] [ 0 . 48 , 0 . 54 ] [ 0 . 27 , 0 . 59 ] [ 0 . 22 , 0 . 79 ] [ 0 . 57 , 0 . 65 ] [ 0 . 35 , 0 . 49 ] [ 0 . 5 , 0 . 5 ] [ 0 . 67 , 0 . 75 ] [ 0 . 39 , 0 . 65 ] [ 0 . 18 , 0 . 27 ] [ 0 . 46 , 0 . 69 ] [ 0 . 35 , 0 . 82 ] [ 0 . 5 , 0 . 5 ] )

[0076] A ˜ 2 = ( [ 0 . 5 , 0 . 5 ] [ 0 . 31 , 0 . 71 ] [ 0 . 43 , 0 . 87 ] [ 0 . 16 , 0 . 38 ] [ 0 . 48 , 0 . 69 ] [ 0 . 55 , 0 . 76 ] [ 0 . 5 , 0 . 5 ] [ 0 . 18 , 0 . 69 ] [ 0 . 44 , 0 . 56 ] [ 0 . 43 , 0 . 64 ] [ 0 . 27 , 0 . 60 ] [ 0 . 47 , 0 . 75 ] [ 0 . 5 , 0 . 5 ] [ 0 . 51 , 0 . 65 ] [ 0 . 16 , 0 . 26 ] [ 0 . 49 , 0 . 84 ] [ 0 . 22 , 0 . 38 ] [ 0 . 50 , 0 . 69 ] [ 0 . 5 , 0 . 5 ] [ 0 . 54 , 0 . 72 ] [ 0 . 24 , 0 . 35 ] [ 0 . 19 , 0 . 66 ] [ 0 . 11 , 0 . 25 ] [ 0 . 33 , 0 . 47 ] [ 0 . 5 , 0 . 5 ] )

[0077] A ˜ 3 = ( [ 0 . 5 , 0 . 5 ] [ 0 . 25 , 0 . 58 ] [ 0 . 37 , 0 . 75 ] [ 0 . 56 , 0 . 78 ] [ 0 . 46 , 0 . 71 ] [ 0 . 16 , 0 . 56 ] [ 0 . 5 , 0 . 5 ] [ 0 . 26 , 0 . 30 ] [ 0 . 41 , 0 . 65 ] [ 0 . 68 , 0 . 74 ] [ 0 . 58 , 0 . 82 ] [ 0 . 27 , 0 . 44 ] [ 0 . 5 , 0 . 5 ] [ 0 . 49 , 0 . 77 ] [ 0 . 73 , 0 . 81 ] [ 0 . 18 , 0 . 40 ] [ 0 . 14 , 0 . 55 ] [ 0 . 26 , 0 . 38 ] [ 0 . 5 , 0 . 5 ] [ 0 . 14 , 0 . 43 ] [ 0 . 41 , 0 . 73 ] [ 0 . 33 , 0 . 77 ] [ 0 . 39 , 0 . 61 ] [ 0 . 32 , 0 . 57 ] [ 0 . 5 , 0 . 5 ] )

[0078] S3. The importance weighted continuous generalized ordered weighted average (IW-CGOWA) operator is used to aggregate the preference degree in the interval fuzzy preference relationship of different prediction models for the measurement sub-interval of pineapple fruit from the interval number into an accurate value;

[0079] In an embodiment of the present invention, it is set that For set [ a , b ] Non-negative functions on, The upper limit of the set, The lower bound of the set, As the independent variable, For the power exponent, x ∈ [ a , b ] ,like ,but ,in For function, As an intermediate variable, It is a monotonic function of the basic unit interval; set , Importance weight function Then the IW-CGOWA operator F Q ([ a , b ], f 0 ) It can be represented as [( a 2 / 3 ) + ( 2 b 2 / 3 )] 1 / 2 ;

[0080] Therefore, the IW-CGOWA operator can be used. F Q ([ a , b ], f 0 ) The preference levels in the interval fuzzy preference relations of different pineapple fruit measurement sub-intervals can be aggregated from the number of intervals into precise values, and then a prediction model can be used. A certain pineapple fruit measurement sub-interval Measurement sub-intervals of other pineapple fruits The precise value of the degree of preference can be expressed as:

[0081] q l , l ′ t = {[ minutes 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )] 2 / 3 + 2 [ max 1 ≤ l , l ′ ≤ L ( m ( l , t ) m ( l , t ) + m ( l ′ , t ) )] 2 / 3 } 1 / 2

[0082] Therefore, the interval fuzzy preference relationship of different prediction models for the pineapple fruit measurement sub-intervals after the interval numbers are converted into precise values ​​by the IW-CGOWA operator can be expressed as:

[0083]

[0084]

[0085]

[0086] S4. Give the weight coefficients of different prediction models for the measurement sub-intervals of pineapple fruit in the set of pineapple sugar content prediction models, obtain the evaluation index of different measurement sub-intervals of pineapple fruit, determine the recommended sub-intervals and recommended prediction models of pineapple fruit for pineapple sorting, and thus establish a global sugar content prediction model to provide high-precision non-destructive detection of sugar content for pineapple sorting.

[0087] In this embodiment of the invention, for the prediction models in the pineapple sugar content prediction model set... When the weight coefficient is a power exponent When the value is 2, let the weighting coefficients for different prediction models for the pineapple fruit measurement sub-intervals be:

[0088]

[0089] Through the above and The evaluation metrics for obtaining different measurement sub-intervals of pineapple fruits are as follows:

[0090]

[0091] Therefore, Time corresponding This can be set as the recommended sub-interval number for pineapple fruit used for sugar content detection. ;

[0092] Considering the robustness of the prediction model for detecting the entire pineapple region, Time corresponding This can be set as the number of the recommended prediction model for sugar content detection. :

[0093] Therefore, recommended sub-intervals for pineapple fruits can be extracted during pineapple sorting. Near-infrared spectral data within the range, i.e. Figure 2 The pineapple fruit measurement sub-interval in the middle of the diagram is shown, and then the method described is used. Corresponding recommendation prediction model This means that a prediction model composed of the MSC-CARS-LSSVM combination can be used, through the aforementioned and A global sugar content prediction model was constructed. The model performance was evaluated after the number of calibration sets and prediction sets were set at a ratio of 3:1. The correlation coefficient of the modeling set, the correlation coefficient of the prediction set, the root mean square error of the modeling set, and the root mean square error of the prediction set were 0.941, 0.926, 0.407, and 0.826, respectively. This indicates that the global sugar content prediction model has high accuracy and is relatively stable, and can provide high-precision non-destructive testing of sugar content for pineapple sorting.

[0094] Please see Figure 3 The second embodiment of the present invention provides a high-precision non-destructive testing device for sugar content detection in pineapple sorting, comprising:

[0095] The sample data fusion module 10 is used to input the spectral data sample set of pineapple sub-intervals, the local sugar content sample set of pineapple sub-intervals, and the global sugar content sample set of pineapple. It also provides recommended sub-intervals and a recommended prediction model for pineapple fruit sugar content detection by setting the measurement area of ​​pineapple fruit and the total number of measurement sub-intervals of pineapple fruit.

[0096] In this embodiment of the invention, by inputting the spectral data sample set of the pineapple sub-interval corresponding to the different pineapple fruit measurement sub-intervals in step S1 of the first embodiment and the local sugar content sample set of the pineapple sub-interval, the recommended sub-interval for sugar content detection of pineapple fruit can be given as the third pineapple fruit measurement sub-interval, and the recommended prediction model is the prediction model composed of the combination of MSC-CARS-LSSVM.

[0097] The spectral data extraction module 20 is used to extract near-infrared spectral data of the pineapple fruit within a recommended sub-interval based on the ratio of the maximum diameter of the core, the minimum diameter of the fruit, and the height of the fruit to the near-infrared spectral image after the near-infrared spectrometer has acquired the near-infrared spectral image of the pineapple under test.

[0098] In this embodiment of the invention, the range of the recommended sub-interval of the pineapple fruit can be calculated based on the maximum diameter of the core, the minimum diameter of the fruit, and the height of the fruit. Then, after performing black and white correction on the near-infrared spectral image of the pineapple, the relevant areas of the recommended sub-interval of the pineapple fruit on the near-infrared spectral image are marked proportionally, and the spectral data of the marked areas are extracted.

[0099] The sugar content non-destructive testing module 30 is used to construct a global sugar content prediction model by using the spectral preprocessing algorithm, wavelength variable screening method, and spectral data modeling method in the recommended prediction model, as well as the spectral data sample set of the pineapple sub-interval corresponding to the recommended sub-interval of the pineapple fruit and the global sugar content value sample set of the pineapple. Then, the global sugar content prediction model processes the near-infrared spectral data of the pineapple fruit in the recommended sub-interval during the sorting process to give the sugar content value of the inspected pineapple.

[0100] In this embodiment of the invention, for the spectral data sample set of pineapple sub-intervals corresponding to the recommended sub-intervals of pineapple fruits, the data quality can be optimized, interference reduced, and features highlighted by the MSC spectral preprocessing algorithm in the recommendation prediction model. Then, the most representative wavelength is selected as the wavelength variable using the CARS wavelength variable screening method. Subsequently, the LSSVM spectral data modeling method is combined with the global sugar content sample set of pineapples and the least squares method is used to fit the data to generate a regressive function with the wavelength variable as the independent variable to predict the target value of the input data, thereby constructing a global sugar content prediction model. For the pineapples being inspected during the sorting process, after the spectral data extraction module 20 outputs the near-infrared spectral data within the recommended sub-intervals of pineapple fruits to the sugar content non-destructive detection module 30, the sugar content value of the inspected pineapple can be output according to the regressive function through the global sugar content prediction model.

[0101] All parts not explicitly stated in the above embodiments can be implemented using existing technologies. The above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Any changes, modifications, substitutions, and variations made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A high-precision non-destructive testing method for sugar content in pineapple sorting, characterized in that, include: For a certain pineapple variety, after dividing the pineapple fruit measurement area into different pineapple fruit measurement sub-intervals, we obtain the spectral data sample set and local sugar content sample set of the pineapple sub-intervals corresponding to the different pineapple fruit measurement sub-intervals, and obtain the global sugar content sample set of the pineapple. By comparing the degree of preference of different prediction models in the pineapple sugar content prediction model set for different pineapple fruit measurement sub-intervals with respect to accuracy, the interval fuzzy preference relationship of different prediction models for pineapple fruit measurement sub-intervals is given; The importance weight continuous generalized ordered weighted average operator is used to aggregate the preference degree in the interval fuzzy preference relationship of different prediction models for pineapple fruit measurement sub-intervals from the interval number into an accurate value; The weight coefficients of different prediction models for pineapple fruit measurement sub-intervals in the pineapple sugar content prediction model set are given, the evaluation index of different pineapple fruit measurement sub-intervals is obtained, and the recommended sub-intervals and recommended prediction models for pineapple sorting are determined. In this way, a global sugar content prediction model is established to provide high-precision non-destructive detection of sugar content for pineapple sorting.

2. The method according to claim 1, characterized in that, For a specific pineapple variety, after dividing the pineapple fruit measurement area into different pineapple fruit measurement sub-intervals, the spectral data sample sets and local sugar content value sample sets corresponding to the different pineapple fruit measurement sub-intervals are obtained, and the global sugar content value sample set of the pineapple is obtained. Specifically: For a specific pineapple variety, when performing non-destructive testing of the pineapple's internal quality using a near-infrared spectrometer, the measurement area for the pineapple fruit is defined as a rectangular region excluding the core area, with the height of the pineapple fruit as the height and the minimum diameter of the pineapple fruit as the width. The total number of sub-intervals for the pineapple fruit measurement is then set. After that, the pineapple can be identified. Regarding the measurement sub-interval of pineapple fruit The scope is: A k , l = { ( i k , j k ) | i k ∈ [ − M k 2 , − N k 2 ] ∪ [ N k 2 , M k 2 ], j k ∈ [ ( l − 1 ) ⋅ H k L , l ⋅ H k L ] } in, Number the sub-intervals for measuring pineapple fruits. This refers to the numbering of the pineapple samples used for sugar content testing. Pineapple samples facing the near-infrared spectrometer x and y coordinates Pineapple The maximum diameter of the core, the minimum diameter of the fruit, and the height of the fruit; By extracting different pineapple samples from the pineapple fruit measurement sub-region Near-infrared spectral data within the specified range can be used to obtain measurement sub-intervals for pineapple fruits. The corresponding spectral data sample set for pineapple seeds is as follows: in, This represents the total number of samples of this pineapple variety used for sugar content testing. To extract pineapple samples When facing a near-infrared spectrometer, the detection range is: Near-infrared spectral data; By measuring different pineapple samples in the pineapple fruit measurement sub-region The sugar content value can be used to obtain the measurement sub-interval of the pineapple fruit. The corresponding local sugar content sample set for pineapple seeds is as follows: in, The cutting height for measuring the sugar content of pineapple. For pineapple samples The height of the fruit cut from the middle is The sugar content value was obtained by measuring the pulp. By removing the core from different pineapple samples, juicing them whole, and then measuring the sugar content, a global sugar content sample set of pineapples can be obtained: in, For pineapple samples The sugar content was measured after the whole fruit was juiced after removing the core.

3. The method according to claim 1, characterized in that, The comparison of the degree of preference of different prediction models in the pineapple sugar content prediction model set for different pineapple fruit measurement sub-intervals with respect to accuracy yields the interval fuzzy preference relationship of different prediction models for pineapple fruit measurement sub-intervals, specifically as follows: Set up a set of pineapple sugar content prediction models Among them, the number is Prediction model It can be composed of a combination of spectral preprocessing algorithms, wavelength variable selection methods, and spectral data modeling methods. Given the number of available prediction models, when using a prediction model... A certain pineapple fruit measurement sub-interval Measurement sub-intervals of other pineapple fruits The degree of preference can be expressed as: p ˜ l , l ′ t = [ min 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) ), max 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) )] in, Number the sub-intervals for measuring other pineapple fruits. ,when hour, p ˜ l , l ′ t = [ 0 . 5 , 0 . 5 ] , To pass the above and Can the use of a prediction model be given? Time measurement sub-intervals of pineapple fruits Accuracy in establishing a local sugar content prediction model; This leads to the use of predictive models. The interval fuzzy preference relation for measuring different pineapple fruits in the sub-intervals is as follows: A ˜ t = ( p ˜ l , l ′ t ) L × L = ([ min 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) ), max 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) )]) L × L 4. The method according to claim 1, characterized in that, The method of using the Importance Weighted Continuous Generalized Ordered Weighted Average (IW-CGOWA) operator to aggregate the preference degree in the interval fuzzy preference relationship of different prediction models for pineapple fruit measurement sub-intervals from the number of intervals into an exact value is as follows: When the IW-CGOWA operator F Q ([ a , b ], f 0 ) for [( a 2 / 3 ) + ( 2 b 2 / 3 )] 1 / 2 At that time, among them, Main function symbol, The upper limit of the set, The lower bound of the set, the importance weight function , As the independent variable, If the basic unit interval is a monotonic function, then the IW-CGOWA operator can be used to convert the preference degree in the interval fuzzy preference relationship of different pineapple fruit measurement sub-intervals from the interval number to a precise value, and then a prediction model can be used. A certain pineapple fruit measurement sub-interval Measurement sub-intervals of other pineapple fruits The precise value of the degree of preference can be expressed as: q l , l ′ t = {[ min 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) )] 2 / 3 + 2 [ max 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) )] 2 / 3 } 1 / 2 Therefore, after the interval numbers are converted into exact values ​​by the IW-CGOWA operator, the prediction model is used. The interval fuzzy preference relation for measuring sub-intervals of pineapple fruit can be expressed as: A t = ( q l , l ′ t ) L × L = ({[ min 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) )] 2 / 3 + 2 [ max 1 ≤ l , l ′ ≤ L ( μ ( l , t ) μ ( l , t ) + μ ( l ′ , t ) )] 2 / 3 } 1 / 2 ) L × L 5. The method according to claim 1, characterized in that, The method provides weight coefficients for different prediction models in the pineapple sugar content prediction model set for the pineapple fruit measurement sub-interval, obtains evaluation indicators for different pineapple fruit measurement sub-intervals, determines recommended sub-intervals and recommended prediction models for pineapple sorting, and thus establishes a global sugar content prediction model to provide high-precision non-destructive detection of sugar content for pineapple sorting. Specifically: For the prediction models in the pineapple sugar content prediction model set The above can be given The weighting coefficients for the pineapple fruit measurement sub-intervals are: w t = ∑ 1 ≤ l ≤ L [ 1 − μ ( l , t )] − β / ∑ 1 ≤ t ′ ≤ τ ∑ 1 ≤ l ≤ L [ 1 − μ ( l , t ' )] − β in, As an intermediate variable, The weighting coefficient is the power exponent; Through the above and The measurement sub-intervals of pineapple fruits can be obtained. The evaluation indicators are: Therefore, Time corresponding This can be set as the recommended sub-interval number for pineapple fruit used for sugar content detection. , Time corresponding This can be set as the number of the recommended prediction model for sugar content detection. : Therefore, recommended sub-intervals for pineapple fruits can be extracted during pineapple sorting. The near-infrared spectral data within the range, and then the... Corresponding recommendation prediction model Through the above and To construct a global sugar content prediction model, thereby providing high-precision non-destructive detection of sugar content for pineapple sorting.

6. A high-precision non-destructive testing device for sugar content in pineapple sorting, characterized in that, include: The sample data fusion module is used to input the spectral data sample set of pineapple sub-intervals, the local sugar content value sample set of pineapple sub-intervals, and the global sugar content value sample set of pineapple. By setting the measurement area of ​​pineapple fruit and the total number of pineapple fruit measurement sub-intervals, it provides recommended sub-intervals and recommended prediction models for pineapple fruit sugar content detection. The spectral data extraction module is used to extract near-infrared spectral data of the pineapple fruit within a recommended sub-interval based on the ratio of the maximum diameter of the core, the minimum diameter of the fruit, and the height of the fruit to the near-infrared spectral image after the near-infrared spectrometer has acquired the near-infrared spectral image of the pineapple being tested. The sugar content non-destructive testing module is used to construct a global sugar content prediction model by using the spectral preprocessing algorithm, wavelength variable screening method, and spectral data modeling method in the recommended prediction model, as well as the spectral data sample set of the pineapple sub-interval corresponding to the recommended sub-interval of the pineapple fruit and the global sugar content value sample set of the pineapple. Then, the global sugar content prediction model processes the near-infrared spectral data of the pineapple fruit in the recommended sub-interval during the sorting process to give the sugar content value of the inspected pineapple.

Images