Intelligent detection method and system for muskmelon pulp hardness based on low-field nuclear magnetic resonance

Abstract

The invention relates to the technical field of fruit detection, in particular to a muskmelon pulp hardness intelligent detection method and system based on low-field nuclear magnetic resonance. Searching a healthy muskmelon pulp hardness detection value by taking the muskmelon variety, the planting scheme and the planting duration as constraints; when the target muskmelon pulp low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, setting a healthy muskmelon pulp hardness detection value as a target muskmelon pulp hardness value; and when the target muskmelon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, processing the target muskmelon pulp low-field nuclear magnetic resonance signal through a muskmelon pulp hardness predictor to obtain a target muskmelon pulp hardness value. The hardness detection can be completed without damaging the muskmelon, the efficiency of large-scale detection is greatly improved, and the accuracy and the reliability of a detection result are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fruit detection, and in particular to an intelligent detection method and system for melon pulp hardness based on low-field nuclear magnetic resonance. Background Art

[0002] Currently, the primary method for testing fruit hardness is using a fruit hardness tester. This method involves inserting the tester's probe into the fruit flesh and measuring the resistance encountered during insertion. This is a contact-based, destructive testing method. However, this method suffers from technical issues such as physical damage to the fruit, relatively cumbersome operation, and difficulty accurately inferring the hardness distribution and overall quality of an entire batch of fruit using only a small sample. Summary of the Invention

[0003] The present invention addresses the technical problems in existing fruit hardness detection technology, such as physical damage to the fruit, relatively cumbersome operation, and difficulty in accurately inferring the hardness distribution and overall quality of a whole batch of fruit using a small number of samples. The present invention provides an intelligent melon pulp hardness detection method and system based on low-field nuclear magnetic resonance to solve these problems.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides an intelligent detection method for melon flesh hardness based on low-field nuclear magnetic resonance, comprising: retrieving a healthy melon flesh hardness detection value based on melon variety, planting plan and planting time as constraints, wherein the healthy melon flesh hardness detection value has a label identifying a standard low-field nuclear magnetic resonance signal; when the target melon flesh low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, setting the healthy melon flesh hardness detection value as the target melon flesh hardness value, wherein the consistent characterization type attribute signals are all the same, and the quantitative attribute signal deviations are all less than or equal to the corresponding quantitative attribute signal deviation threshold; when the target melon flesh low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, processing the target melon flesh low-field nuclear magnetic resonance signal through a melon flesh hardness predictor to obtain the target melon flesh hardness value, wherein the melon flesh hardness predictor is generated by machine learning training using multiple sets of data, and any set of the multiple sets of data includes a melon flesh low-field nuclear magnetic resonance recording signal and a label identifying the melon flesh hardness value.

[0005] Optionally, based on the melon variety, planting plan and planting time as constraints, the healthy melon flesh hardness test value is retrieved, wherein the healthy melon flesh hardness test value has a label that identifies the standard low-field nuclear magnetic resonance signal, including: obtaining an initial low-field nuclear magnetic resonance signal attribute set; based on the melon flesh hardness, performing a correlation analysis on the initial low-field nuclear magnetic resonance signal attribute set to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set; based on the initial low-field nuclear magnetic resonance signal attribute correlation degree set, sorting initial low-field nuclear magnetic resonance signal attribute set from the initial low-field nuclear magnetic resonance signal attribute set. A set of associated low-field nuclear magnetic resonance signal attributes whose correlation degree of field nuclear magnetic resonance signal attributes is greater than or equal to a correlation degree threshold is prepared; a set of healthy melon flesh hardness record values ​​is retrieved with melon variety, planting plan and planting time as constraints, a box plot analysis is performed, and a box interval of the healthy melon flesh hardness record value is obtained, which is set as the healthy melon flesh hardness test value; based on the associated low-field nuclear magnetic resonance signal attribute set, the low-field nuclear magnetic resonance signal of the healthy melon flesh that meets the healthy melon flesh hardness test value is extracted, and a label for identifying a standard low-field nuclear magnetic resonance signal is constructed.

[0006] Wherein, based on the hardness of melon flesh, the initial low-field nuclear magnetic resonance signal attribute set is subjected to correlation analysis to obtain the initial low-field nuclear magnetic resonance signal attribute correlation degree set, including: loading the one-to-one corresponding melon flesh hardness record value set, the first attribute initial low-field nuclear magnetic resonance signal detection value set until the Qth attribute initial low-field nuclear magnetic resonance signal detection value set, wherein Q represents the number of attributes of the initial low-field nuclear magnetic resonance signal attribute set; de-dimensionalizing the melon flesh hardness record value set to obtain a melon flesh hardness characteristic value sequence; traversing the first attribute initial low-field nuclear magnetic resonance signal detection value set. The first attribute detection value set up to the Qth attribute initial low-field nuclear magnetic resonance signal detection value set are respectively de-dimensionalized to obtain the first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence up to the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence; the melon pulp hardness characteristic value sequence is used as the reference sequence, and the first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence up to the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence is used as the comparison sequence, a grey correlation matrix is ​​constructed, and grey correlation analysis is performed to obtain the initial low-field nuclear magnetic resonance signal attribute correlation set.

[0007] Among them, based on the associated low-field nuclear magnetic resonance signal attribute set, the low-field nuclear magnetic resonance signal of the healthy melon pulp that meets the healthy melon pulp hardness detection value is extracted, and a label for identifying the standard low-field nuclear magnetic resonance signal is constructed, including: obtaining several first-associated attribute low-field nuclear magnetic resonance signal detection values ​​of several healthy melon pulps whose hardness detection values ​​belong to the healthy melon pulp hardness detection value, up to several P-th associated attribute low-field nuclear magnetic resonance signal detection values, where P represents the total number of attributes of the associated low-field nuclear magnetic resonance signal attribute set; performing a centralized value evaluation on the several first-associated attribute low-field nuclear magnetic resonance signal detection values ​​to obtain a first-associated attribute standard low-field nuclear magnetic resonance signal characteristic value; performing a centralized value evaluation on the several P-th associated attribute low-field nuclear magnetic resonance signal detection values ​​to obtain a P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value; and constructing a label for identifying the standard low-field nuclear magnetic resonance signal based on the first-associated attribute standard low-field nuclear magnetic resonance signal characteristic value up to the P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value.

[0008] The method comprises the following steps: processing the target melon pulp low-field nuclear magnetic resonance signal through a melon pulp hardness predictor to obtain the target melon pulp hardness value; and generating the melon pulp hardness predictor through machine learning training using multiple sets of data. Any set of the multiple sets of data includes a melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value. The method comprises the following steps: constructing a signal input branch set based on an associated low-field nuclear magnetic resonance signal attribute set; configuring a backbone network based on a fully connected neural network; connecting the output nodes of the signal input branch set in parallel to the backbone network to obtain a melon pulp hardness predictor architecture; placing the label identifying the melon pulp hardness value at the output layer of the melon pulp hardness predictor architecture, placing the melon pulp low-field nuclear magnetic resonance recording signal at the input layer of the melon pulp hardness predictor architecture, retrieving the multiple sets of data, training the melon pulp hardness predictor architecture, and obtaining the melon pulp hardness predictor.

[0009] Among them, based on the associated low-field nuclear magnetic resonance signal attribute set, a signal input branch set is constructed, including: extracting a first associated low-field nuclear magnetic resonance signal attribute from the associated low-field nuclear magnetic resonance signal attribute set; when the storage format of the first associated low-field nuclear magnetic resonance signal attribute is an image, configuring a signal input convolution branch; when the storage format of the first associated low-field nuclear magnetic resonance signal attribute is a numerical value, configuring a signal input fully connected branch.

[0010] Optionally, when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, and then the method further includes: associating and storing the target melon pulp low-field nuclear magnetic resonance signal and the target melon pulp hardness value, using the target melon pulp low-field nuclear magnetic resonance signal as an index condition, and constructing a melon pulp hardness calibration database with the target melon pulp hardness value; wherein, after the melon pulp hardness calibration database is constructed: when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, hardness value calibration is performed by the melon pulp hardness calibration database to obtain a melon pulp hardness value calibration result; when the melon pulp hardness value calibration result is empty, processing the target melon pulp low-field nuclear magnetic resonance signal by a melon pulp hardness predictor to obtain the target melon pulp hardness value.

[0011] In a second aspect, the present invention provides an intelligent detection system for melon pulp hardness based on low-field nuclear magnetic resonance, comprising: A healthy melon flesh hardness calibration module is used to retrieve healthy melon flesh hardness test values ​​based on melon variety, planting plan, and planting duration, wherein the healthy melon flesh hardness test values ​​have a label identifying a standard low-field nuclear magnetic resonance signal; a target melon pulp hardness matching module, configured to set the healthy melon pulp hardness detection value as the target melon pulp hardness value when the target melon pulp low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, wherein the consistent characterization type attribute signals are the same and the quantitative attribute signal deviations are less than or equal to the corresponding quantitative attribute signal deviation threshold; The target melon pulp hardness prediction module is used to process the target melon pulp low-field nuclear magnetic resonance signal through a melon pulp hardness predictor to obtain the target melon pulp hardness value when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, wherein the melon pulp hardness predictor is generated by machine learning training through multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value.

[0012] By implementing the present invention, it is possible to retrieve healthy melon flesh hardness test values ​​based on melon variety, planting plan, and planting duration as constraints. The healthy melon flesh hardness test values ​​have labels that identify standard low-field nuclear magnetic resonance signals. The setting of the constraint conditions can ensure that the retrieved healthy melon flesh hardness test values ​​are targeted and accurate, avoiding a reduction in the reference value of the test values ​​due to differences in variety, planting conditions, etc., and the application of correlation analysis and box plot analysis can screen out nuclear magnetic resonance signal attributes that are highly correlated with flesh hardness, reduce interference from irrelevant signals, and improve subsequent detection efficiency. By implementing the present invention, it can be achieved that when the target melon pulp low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, the healthy melon pulp hardness detection value is set as the target melon pulp hardness value, wherein the consistent characterization type attribute signals are all the same, and the quantitative attribute signal deviations are all less than or equal to the corresponding quantitative attribute signal deviation threshold. The clear consistency judgment standard ensures the rigor of the matching result, avoids misjudgment due to partial signal similarity, and improves detection accuracy. The healthy melon pulp hardness detection value is directly used without the need for complex calculations or additional detection processes, which can significantly shorten the detection time and achieve rapid detection. It is particularly suitable for batch detection scenarios to improve detection efficiency. By implementing the present invention, when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, wherein the melon pulp hardness predictor is generated by machine learning training using multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value. The construction of the melon pulp hardness predictor fully considers the properties of nuclear magnetic resonance signals in different formats, and respectively configures convolution branches and fully connected branches, which can effectively process different types of signal data and improve prediction applicability. The machine learning training based on a large amount of data enables the melon pulp hardness predictor to have strong data analysis and processing capabilities. Even if there is a difference between the target signal and the standard signal, the pulp hardness value can be accurately predicted, thereby expanding the detection range and ensuring that melons with non-standard signals can also be effectively detected.

[0013] In summary, by implementing the present invention, hardness testing can be completed without destroying the melons, and the efficiency of large-scale testing is greatly accelerated, thereby improving the accuracy and reliability of the test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic flow chart of an intelligent detection method for melon pulp hardness based on low-field nuclear magnetic resonance provided by the present invention; Figure 2This is a structural schematic diagram of a melon pulp hardness intelligent detection system based on low-field nuclear magnetic resonance provided by the present invention.

[0015] In the accompanying drawings, the components represented by the reference numerals are as follows: Healthy melon pulp hardness calibration module 11, target melon pulp hardness matching module 12, target melon pulp hardness prediction module 13. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0017] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Therefore, features specified as "first" or "second" may explicitly or implicitly include one or more features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0018] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0019] Example 1, as Figure 1 As shown, the embodiment of the present invention provides a method and system for intelligently detecting melon pulp hardness based on low-field nuclear magnetic resonance, comprising: S100: Retrieving a healthy melon flesh hardness test value based on the melon variety, planting plan, and planting duration as constraints, wherein the healthy melon flesh hardness test value has a label identifying a standard low-field nuclear magnetic resonance signal; S200: When the target melon pulp low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, setting the healthy melon pulp hardness detection value as the target melon pulp hardness value, wherein the consistent characterization type attribute signals are the same, and the quantitative attribute signal deviations are less than or equal to the corresponding quantitative attribute signal deviation threshold; S300: When the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, wherein the melon pulp hardness predictor is generated by machine learning training using multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value.

[0020] In step S100 of the embodiment of the present application, the healthy melon flesh hardness test value is retrieved based on the melon variety, planting plan and planting time as constraints, wherein the healthy melon flesh hardness test value has a label identifying a standard low-field nuclear magnetic resonance signal, including: obtaining an initial low-field NMR signal property set; Based on the hardness of the melon pulp, a correlation analysis is performed on the initial low-field nuclear magnetic resonance signal attribute set to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set; Based on the initial low-field nuclear magnetic resonance signal attribute correlation degree set, sorting, from the initial low-field nuclear magnetic resonance signal attribute set, associated low-field nuclear magnetic resonance signal attribute sets whose initial low-field nuclear magnetic resonance signal attribute correlation degrees are greater than or equal to a correlation degree threshold; Using the melon variety, planting plan, and planting duration as constraints, retrieve a set of healthy melon flesh hardness record values, perform box plot analysis, and obtain the box intervals of the healthy melon flesh hardness record values, which are set as the healthy melon flesh hardness test values; Based on the associated low-field nuclear magnetic resonance signal attribute set, the low-field nuclear magnetic resonance signal of the healthy melon pulp that meets the healthy melon pulp hardness test value is extracted, and a label for identifying the standard low-field nuclear magnetic resonance signal is constructed.

[0021] In the embodiment of the present application, the purpose of step S100 is to screen out low-field nuclear magnetic resonance signal attributes that are strongly correlated with the hardness of melon flesh, and to determine the flesh hardness range of healthy melons under specific growing conditions, and to match the corresponding standard low-field nuclear magnetic resonance signal for the hardness range to form a reusable healthy melon reference label.

[0022] First, the initial low-field NMR signal attribute set needs to be obtained. Specifically, using low-field NMR testing equipment, signals are collected from healthy melon pulp of different varieties, planting plans, and planting times. All detectable low-field NMR signal attributes, such as signal intensity, relaxation time, and signal waveform, are recorded and summarized to form the initial low-field NMR signal attribute set.

[0023] It is further necessary to perform a correlation analysis on the initial low-field nuclear magnetic resonance signal attribute set based on the hardness of the melon pulp to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set.

[0024] In step S100 of the embodiment of the present application, a correlation analysis is performed on the initial low-field nuclear magnetic resonance signal attribute set based on the hardness of the melon pulp to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set, including: Load the corresponding melon flesh hardness record value set, the first attribute initial low-field nuclear magnetic resonance signal detection value set, and the Qth attribute initial low-field nuclear magnetic resonance signal detection value set, where Q represents the number of attributes in the initial low-field nuclear magnetic resonance signal attribute set; performing dimensionless processing on the melon pulp hardness record value set to obtain a melon pulp hardness characteristic value sequence; Traversing the first attribute initial low-field nuclear magnetic resonance signal detection value set until the Qth attribute initial low-field nuclear magnetic resonance signal detection value set, respectively performing dimensionless processing to obtain a first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence until the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence; Taking the melon pulp hardness characteristic value sequence as the reference sequence and the first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence to the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence as the comparison sequence, a grey correlation matrix is ​​constructed, and grey correlation analysis is performed to obtain the initial low-field nuclear magnetic resonance signal attribute correlation set.

[0025] In the embodiment of the present application, the purpose of the above steps is to quantify the degree of correlation between each attribute in the initial low-field nuclear magnetic resonance signal attribute set and the hardness of the melon pulp, eliminate dimensional interference by loading matching data and standardizing processing, and then use gray correlation analysis to construct the correlation relationship between pulp hardness and signal attributes, and finally generate a correlation set to provide a quantitative basis for the subsequent screening of strongly correlated signal attributes.

[0026] First, you need to load a one-to-one correspondence between the melon flesh hardness record value set and the first attribute initial low-field NMR signal detection value set, through the Qth attribute initial low-field NMR signal detection value set. This means extracting two types of data from the same batch of healthy melon samples from the historical test database, and the two types of data must strictly correspond. The first type of data is the melon flesh hardness record value set, which contains the actual flesh hardness test results for each healthy melon sample; the second type of data is the initial low-field NMR signal detection value set. This data set is categorized by attribute and contains the initial low-field NMR signal detection values ​​for the first attribute through the Qth attribute, where Q is the total number of attributes in the initial signal attribute set. These attributes can include low-field NMR signal attributes such as signal intensity and relaxation time.

[0027] Furthermore, it is necessary to perform dimensionless processing on the melon flesh hardness record value set to obtain a melon flesh hardness characteristic value sequence.

[0028] In the actual process of measuring melon flesh hardness, different testing equipment may use units such as "kg / cm²" or "N," resulting in unit discrepancies. Converting hardness records to a uniformly scaled characteristic value avoids imbalances in data weights during subsequent correlation analysis due to unit differences and ensures the objectivity of the analysis results. Specifically, standardization methods such as Z-score and Min-Max standardization can be used to de-dimensionalize all hardness records and aggregate them into a sequence of melon flesh hardness characteristic values. Taking Min-Max standardization as an example, the calculation formula is x' = x - min(X) / [max(X) - min(X)]. Here, x is a single hardness record, min(X) is the minimum value of the hardness record set, max(X) is the maximum value of the hardness record set, and x' is the de-dimensionalized hardness characteristic value.

[0029] Next, it is necessary to traverse the first attribute initial low-field nuclear magnetic resonance signal detection value set until the Qth attribute initial low-field nuclear magnetic resonance signal detection value set and perform dimensionless processing on each of them to obtain the first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence until the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence; Similarly, the Min-Max normalization method can be used to process the initial low-field nuclear magnetic resonance signal detection value sets of the first attribute to the Qth attribute one by one. The specific process includes: First, for the initial low-field NMR signal detection value set of each attribute, the minimum and maximum values ​​of the set are calculated; Each initial low-field NMR signal detection value in the set is converted into an eigenvalue according to the above-mentioned Min-Max normalization dimensionless formula; The characteristic values ​​of each attribute are summarized respectively to form an initial low-field nuclear magnetic resonance signal characteristic value sequence from the first attribute to the Qth attribute.

[0030] Furthermore, it is necessary to use the melon pulp hardness characteristic value sequence as the benchmark sequence, and the first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence to the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence as the comparison sequence, construct a grey correlation matrix, perform grey correlation analysis, and obtain the initial low-field nuclear magnetic resonance signal attribute correlation set.

[0031] First, the melon flesh hardness feature value sequence needs to be set as the benchmark sequence, denoted as X0, which represents the target variable of the analysis; the initial signal feature value sequence from the first attribute to the Qth attribute needs to be set as the comparison sequence, denoted as (X1, X2, ..., X Q ), representing the influencing variables of the analysis.

[0032] Next, we need to calculate the correlation coefficient for each alignment sequence X i (i=1,2,...,Q), calculate the correlation coefficient ξ with the benchmark sequence at each sample point i (k), where k is the sample number, and the calculation formula is:

[0033] Among them, min i min k |X0(k)-X i (k)| is the minimum difference between the two levels, ζmax i max k |X0(k)-X i (k)|| is the maximum difference between the two levels, and ζ is the resolution coefficient, which is usually taken as 0.5 and is used to adjust the sensitivity of the correlation coefficient.

[0034] Then, each aligned sequence X i Correlation coefficient ξ with the benchmark sequence X0 i (1),ξ i (2),...ξ i (n) are arranged in rows to form a Q×n grey relational matrix, where n is the number of samples.

[0035] Furthermore, the correlation coefficient of each comparison sequence Xi is averaged to obtain the correlation degree ri between the sequence and the reference sequence, which is: i =(1 / n)∑ n k=1 ξ i (k).

[0036] Finally, the correlation degrees r1, r2, ..., r of all aligned sequences are calculated. Q Summarize and form the initial low-field NMR signal attribute correlation set.

[0037] Furthermore, it is necessary to sort out associated low-field NMR signal attribute sets having initial low-field NMR signal attribute correlation degrees greater than or equal to a correlation threshold from the initial low-field NMR signal attribute set based on the initial low-field NMR signal attribute correlation degree set.

[0038] A correlation threshold is preset. The specific setting can be based on the detection accuracy requirements and the actual data distribution: if higher detection accuracy is required, the correlation threshold can be set to 0.7-0.8; if both detection efficiency and accuracy are required, the correlation threshold can be set to 0.5-0.6, etc. Then, the initial low-field NMR signal attribute correlations are concentrated, and all signal attributes with correlation values ​​greater than or equal to the threshold are extracted to form a set of correlated low-field NMR signal attributes.

[0039] Furthermore, it is necessary to retrieve a set of healthy melon flesh hardness record values ​​based on the melon variety, planting plan, and planting time, perform box plot analysis, and obtain the box interval of the healthy melon flesh hardness record values, which is set as the healthy melon flesh hardness detection value.

[0040] First, it is necessary to use melon variety, planting plan, and planting time as joint constraints, retrieve the flesh hardness record values ​​of all healthy melons under these conditions from the historical detection database, and form a set of healthy melon flesh hardness record values.

[0041] Then, a box plot analysis is performed to determine a reasonable interval. That is, a box plot analysis is performed on the retrieved set of recorded values ​​of healthy melon flesh hardness. The quartiles in the data are identified through the box plot, namely the lower quartile Q1, the median Q2, and the upper quartile Q3. The Q1-Q3 interval, that is, the box interval, is used as the reasonable range of healthy melon flesh hardness. This interval can exclude outliers, such as excessively high / excessively low values ​​caused by detection errors. Ultimately, the box interval is set as the healthy melon flesh hardness detection value under specific constraints.

[0042] In step S100 of the embodiment of the present application, based on the associated low-field nuclear magnetic resonance signal attribute set, the low-field nuclear magnetic resonance signal of the healthy melon pulp that meets the healthy melon pulp hardness test value is extracted, and a label identifying the standard low-field nuclear magnetic resonance signal is constructed, including: Obtaining a plurality of first associated attribute low-field nuclear magnetic resonance signal detection values ​​up to a plurality of P-th associated attribute low-field nuclear magnetic resonance signal detection values ​​of a plurality of healthy melon pulps whose hardness detection values ​​belong to the healthy melon pulp hardness detection value, where P represents the total number of attributes in the associated low-field nuclear magnetic resonance signal attribute set; Performing a centralized value evaluation on the plurality of first correlation attribute low-field nuclear magnetic resonance signal detection values ​​to obtain a first correlation attribute standard low-field nuclear magnetic resonance signal characteristic value; Until the plurality of P-th associated attribute low-field nuclear magnetic resonance signal detection values ​​are collectively evaluated to obtain the P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value; Based on the first associated attribute standard low-field nuclear magnetic resonance signal characteristic value up to the P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value, a label identifying the standard low-field nuclear magnetic resonance signal is constructed.

[0043] In step S100 of this embodiment, the goal is to establish a precise binding signature between strongly correlated signal attributes and the range of healthy melon flesh firmness. This is achieved by extracting the strongly correlated NMR signals corresponding to melon samples that meet the healthy melon flesh firmness criteria, calculating the typical eigenvalues ​​of each signal attribute, and ultimately forming a standardized signal signature. This signature serves as the core benchmark for subsequent target melon firmness testing, ensuring that only melons that meet the dual criteria of healthy melon flesh firmness and strongly correlated signal are used for rapid matching testing, fundamentally ensuring the accuracy and pertinence of test results.

[0044] First, it is necessary to obtain several first-related attribute low-field nuclear magnetic resonance signal detection values ​​of several healthy melon pulps whose hardness detection values ​​belong to the healthy melon pulp hardness detection values, up to several P-th related attribute low-field nuclear magnetic resonance signal detection values, where P represents the total number of attributes in the related low-field nuclear magnetic resonance signal attribute set.

[0045] That is, a healthy melon sample that meets two conditions at the same time is retrieved from the historical detection database. First, the flesh hardness detection value belongs to the healthy melon flesh hardness detection value interval determined in the previous step, that is, the [Q1, Q3] box interval; second, the detection values ​​of all attributes in the associated low-field nuclear magnetic resonance signal attribute set of the healthy melon sample have been recorded, of which there are P attributes in total, such as the first associated attribute to the Pth associated attribute.

[0046] Then, it is necessary to perform a centralized value evaluation on the several first-association attribute low-field nuclear magnetic resonance signal detection values ​​to obtain the first-association attribute standard low-field nuclear magnetic resonance signal characteristic value. First, it is necessary to eliminate outliers on the several first-association attribute signal detection values, for example, by calculating the mean μ and standard deviation σ of the group of data, and eliminating extreme values ​​that exceed the range of "μ±2σ". Next, it is necessary to select a suitable centralized value evaluation method according to the signal characteristics of the first-association attribute. If the attribute signal data is evenly distributed and has no obvious skewness, the arithmetic mean is used as the centralized value. If the attribute signal data has a skewed distribution, such as some signal intensity data, the median can be used as the centralized value to avoid the influence of extreme values ​​on the centralized value. Finally, the calculated centralized value is used as the first-association attribute standard low-field nuclear magnetic resonance signal characteristic value, which represents the typical signal level of the first-association attribute under the condition of healthy melon flesh hardness.

[0047] Furthermore, consistent with the processing logic of the first associated attribute, the standard low-field nuclear magnetic resonance signal characteristic values ​​are calculated for all P associated attributes respectively to ensure that each strongly associated attribute has a corresponding mapping between the hardness of healthy melon flesh and the standard low-field nuclear magnetic resonance signal characteristic value, providing standard data for all attributes for the subsequent construction of a complete label.

[0048] Finally, the P scattered associated attribute standard low-field NMR signal eigenvalues ​​are integrated into a complete standard low-field NMR signal eigenvalue label that can be directly used for comparison, clarifying the standard low-field NMR signal morphology corresponding to the hardness of healthy melon flesh, and providing a clear and quantifiable basis for the subsequent consistency judgment between the target low-field NMR signal and the standard low-field NMR signal in S200.

[0049] In step S300 of the embodiment of the present application, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, wherein the melon pulp hardness predictor is generated by machine learning training using multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value, including: Based on the associated low-field NMR signal attribute set, a signal input branch set is constructed; Based on the fully connected neural network, configure the backbone network; Connecting the output nodes of the signal input branch set in parallel to the backbone network to obtain a melon flesh firmness predictor architecture; A label identifying the melon pulp hardness value is placed in the output layer of the melon pulp hardness predictor architecture, and a low-field nuclear magnetic resonance recording signal of the melon pulp is placed in the input layer of the melon pulp hardness predictor architecture. The multiple sets of data are retrieved, and the melon pulp hardness predictor architecture is trained to obtain the melon pulp hardness predictor.

[0050] In the embodiment of the present application, the purpose of step S300 is to construct a machine learning melon flesh hardness predictor that is adapted to strongly correlated low-field nuclear magnetic resonance signals: by customizing the input branches to adapt to different types of correlated low-field nuclear magnetic resonance signal attributes, combining a fully connected neural network to build a prediction architecture, and then using low-field nuclear magnetic resonance signal-melon flesh hardness annotation data to train the melon flesh hardness predictor, and finally obtain a prediction tool that can accurately process non-standard signals and output the target melon flesh hardness value, thereby solving the special scene detection problem that cannot be covered by the standard signal matching in S200, and ensuring the comprehensiveness and accuracy of the detection.

[0051] To implement the above steps, it is first necessary to construct a signal input branch set based on the associated low-field NMR signal attribute set.

[0052] In step S300 of the embodiment of the present application, a signal input branch set is constructed based on the associated low-field nuclear magnetic resonance signal attribute set, including: extracting a first correlated low-field nuclear magnetic resonance signal attribute from the correlated low-field nuclear magnetic resonance signal attribute set; When the storage format of the first associated low-field nuclear magnetic resonance signal attribute is an image, configuring a signal input convolution branch; When the storage format of the first associated low-field nuclear magnetic resonance signal attribute is a numerical value, a signal input fully connected branch is configured.

[0053] Specifically, first, a first associated low-field nuclear magnetic resonance signal attribute is extracted from the associated low-field nuclear magnetic resonance signal attribute set, and then a storage format of the extracted first associated low-field nuclear magnetic resonance signal attribute is identified to determine whether it is in an image format or a numerical format. The image format may be a nuclear magnetic resonance image or a signal heat map, and the numerical format may be a specific value of a relaxation time or a signal intensity.

[0054] If the first associated low-field NMR signal attribute is in image format, a signal input convolution branch is configured, consisting of a convolutional layer and a pooling layer. The signal input convolution branch includes three one-dimensional convolutional layers with a 7×1 kernel size and increasing channels at each layer. If the attribute is in numeric format, a signal input fully connected branch is configured, consisting of a densely connected layer. This logic is used to traverse all attributes in the associated low-field NMR signal attribute set, completing the configuration of the input branch corresponding to each attribute, and summarizing them to form a signal input branch set.

[0055] Furthermore, a backbone network based on a fully connected neural network is required. Specifically, the backbone network structure should be designed based on the number of associated low-field NMR signal attributes and the data complexity, typically using 3-5 fully connected layers.

[0056] For example, the first layer is used to receive the fusion features of multi-branch outputs. The number of nodes is set to 256 to adapt to the requirements of multi-feature fusion. The activation function uses ReLU to solve the gradient disappearance problem. The number of nodes in the middle layer is gradually reduced, for example, from 128 to 64. The activation function still uses ReLU, and the feature map is optimized by gradually reducing the dimension. The number of nodes in the penultimate layer is set to 32, and a Dropout layer, such as Dropout(0.2), is added to prevent the model from overfitting.

[0057] The last layer of the backbone network needs to reserve an input interface that matches the output dimension of the signal input branch set to ensure that the features extracted by multiple branches can be smoothly transmitted to the backbone network for fusion calculation.

[0058] Furthermore, it is necessary to connect the output nodes of the signal input branch set in parallel to the backbone network to obtain the melon flesh hardness predictor architecture, so as to form a complete end-to-end prediction architecture, ensure that different types of correlated low-field nuclear magnetic resonance signal features can be used synergistically, and improve the melon flesh hardness predictor's ability to process complex signals.

[0059] Next, the output features of each input branch need to be dimensionalized. Specifically, the output of the signal input convolutional branch is converted into a one-dimensional feature vector using a Flatten layer. Since the output of the signal input fully connected branch is already a one-dimensional vector, its original dimensionality can be retained. A Concatenate layer is then used to concatenate all the adjusted branch output feature vectors in parallel to form a fused feature vector. This fused feature vector is then directly input into the first fully connected layer of the backbone network, completing the connection between the signal input branch set and the backbone network, thus obtaining the melon flesh firmness predictor architecture.

[0060] Furthermore, it is necessary to train the melon flesh hardness predictor architecture with labeled data to obtain the melon flesh hardness predictor, that is, place the label identifying the melon flesh hardness value at the output layer of the melon flesh hardness predictor architecture, place the melon flesh low-field nuclear magnetic resonance recording signal at the input layer of the melon flesh hardness predictor architecture, retrieve the multiple sets of data, train the melon flesh hardness predictor architecture, and obtain the melon flesh hardness predictor.

[0061] First, we need to retrieve multiple sets of low-field NMR recordings of melon flesh and their corresponding label data identifying the melon flesh hardness values ​​as training data for the melon flesh hardness predictor. The input of each set of data must satisfy the requirement that the low-field NMR signal record contains all the attributes in the associated low-field NMR signal attribute set; the label must be the actual flesh hardness value corresponding to the low-field NMR signal, such as a specific value in "kg / cm²", which corresponds one-to-one with the input low-field NMR signal. The training data is divided into a training set and a validation set in a 7:3 ratio.

[0062] In the parameter setting of the melon pulp hardness predictor, it is also necessary to add an output layer to the last layer of the melon pulp hardness predictor architecture, that is, at the end of the backbone network. The number of nodes is set to 1, corresponding to a single melon pulp hardness value output, and the activation function is Linear; the loss function of the melon pulp hardness predictor is the mean square error (MSE); the optimizer uses the Adam optimizer, and the learning rate is set to 0.001.

[0063] In the training of the melon flesh hardness predictor, the batch size is set to 32, the training rounds are 50-100, and the early stopping mechanism is set. When the validation set loss does not decrease for 50 consecutive times, the training is stopped to judge the model convergence and save the model parameters of the melon flesh hardness predictor at this time.

[0064] The trained and saved model parameters are loaded into the melon flesh firmness predictor architecture, creating a melon flesh firmness predictor that can be directly used to process the target signal. During subsequent testing, the predictor outputs the corresponding melon flesh firmness value by inputting the target melon's low-field NMR signal.

[0065] In step S300 of the embodiment of the present application, when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, the target melon pulp low-field nuclear magnetic resonance signal is processed by the melon pulp hardness predictor to obtain the target melon pulp hardness value, and then the method further includes: The target melon pulp low-field nuclear magnetic resonance signal and the target melon pulp hardness value are associated and stored, and a melon pulp hardness calibration database is constructed using the target melon pulp low-field nuclear magnetic resonance signal as an index condition and the target melon pulp hardness value; Wherein, after the melon pulp hardness calibration database is constructed: When the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, performing hardness value calibration through the melon pulp hardness calibration database to obtain a melon pulp hardness value calibration result; When the calibration result of the melon pulp hardness value is empty, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value.

[0066] In the embodiment of the present application, the purpose of the above steps is to construct a low-field nuclear magnetic resonance signal-melon flesh hardness calibration database and form a detection logic of database priority matching + melon flesh hardness predictor backup: on the one hand, by storing the non-standard low-field nuclear magnetic resonance signal-melon flesh hardness value data of historical detection, a rapid matching basis is provided for subsequent similar non-standard signals, thereby improving detection efficiency; on the other hand, it avoids resource consumption caused by repeated use of the melon flesh hardness predictor, and at the same time optimizes the detection coverage by continuously accumulating data, further ensuring the efficiency and stability of detection in non-standard signal scenarios.

[0067] First, it is necessary to associate and store the target melon pulp low-field nuclear magnetic resonance signal and the target melon pulp hardness value, and construct a melon pulp hardness calibration database using the target melon pulp low-field nuclear magnetic resonance signal as an index condition and the target melon pulp hardness value.

[0068] Specifically, the target melon pulp low-field NMR signal processed by the melon pulp firmness predictor and the corresponding target melon pulp firmness value data need to be extracted, and an associated identifier is added to each data set. A calibration database is then constructed using a structured database such as MySQL or SQLite. A storage structure consisting of index fields and data fields is designed, using the key features of the target melon pulp low-field NMR signal as the core index. Constraint fields such as variety, planting plan, and planting duration are also associated to ensure accurate subsequent retrieval. The target melon pulp firmness value corresponding to the index signal is then stored. This completes the construction of the melon pulp firmness calibration database.

[0069] Furthermore, after the melon pulp hardness calibration database is constructed, when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, hardness calibration is performed using the melon pulp hardness calibration database to obtain a melon pulp hardness calibration result. Specifically, using the target melon pulp low-field nuclear magnetic resonance signal as an index, the corresponding melon pulp hardness value is directly extracted from the melon pulp hardness calibration database and output as the melon pulp hardness calibration result.

[0070] Furthermore, when the melon flesh hardness calibration result is empty, the target melon flesh low-field nuclear magnetic resonance signal is processed by a melon flesh hardness predictor to obtain the target melon flesh hardness value. Specifically, if no record meeting the index criteria is found after searching the melon flesh hardness calibration database, the melon flesh hardness calibration result is determined to be empty. The target melon flesh low-field nuclear magnetic resonance signal is then input into the melon flesh hardness predictor to predict the target melon flesh hardness value.

[0071] Example 2, as Figure 2 As shown, based on the same inventive concept as the method for intelligently detecting melon pulp hardness based on low-field nuclear magnetic resonance provided in Example 1, an embodiment of the present invention further provides an intelligent system for detecting melon pulp hardness based on low-field nuclear magnetic resonance, comprising: A healthy melon flesh hardness calibration module 11 is used to retrieve a healthy melon flesh hardness test value based on the melon variety, planting plan, and planting duration, wherein the healthy melon flesh hardness test value has a label identifying a standard low-field nuclear magnetic resonance signal; a target melon pulp hardness matching module 12, configured to set the healthy melon pulp hardness detection value as the target melon pulp hardness value when the target melon pulp low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, wherein the consistent characterization type attribute signals are the same and the quantitative attribute signal deviations are less than or equal to the corresponding quantitative attribute signal deviation threshold; The target melon pulp hardness prediction module 13 is used to process the target melon pulp low-field nuclear magnetic resonance signal through a melon pulp hardness predictor to obtain the target melon pulp hardness value when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, wherein the melon pulp hardness predictor is generated by machine learning training through multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value.

[0072] Furthermore, the healthy melon pulp hardness calibration module 11 includes the following execution steps: obtaining an initial low-field NMR signal property set; Based on the hardness of the melon pulp, a correlation analysis is performed on the initial low-field nuclear magnetic resonance signal attribute set to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set; Based on the initial low-field nuclear magnetic resonance signal attribute correlation degree set, sorting, from the initial low-field nuclear magnetic resonance signal attribute set, associated low-field nuclear magnetic resonance signal attribute sets whose initial low-field nuclear magnetic resonance signal attribute correlation degrees are greater than or equal to a correlation degree threshold; Using the melon variety, planting plan, and planting duration as constraints, retrieve a set of healthy melon flesh hardness record values, perform box plot analysis, and obtain the box intervals of the healthy melon flesh hardness record values, which are set as the healthy melon flesh hardness test values; Based on the associated low-field nuclear magnetic resonance signal attribute set, the low-field nuclear magnetic resonance signal of the healthy melon pulp that meets the healthy melon pulp hardness test value is extracted, and a label for identifying the standard low-field nuclear magnetic resonance signal is constructed.

[0073] The method comprises performing a correlation analysis on the initial low-field nuclear magnetic resonance signal attribute set based on the hardness of the melon pulp to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set, including: Load the corresponding melon flesh hardness record value set, the first attribute initial low-field nuclear magnetic resonance signal detection value set, and the Qth attribute initial low-field nuclear magnetic resonance signal detection value set, where Q represents the number of attributes in the initial low-field nuclear magnetic resonance signal attribute set; performing dimensionless processing on the melon pulp hardness record value set to obtain a melon pulp hardness characteristic value sequence; Traversing the first attribute initial low-field nuclear magnetic resonance signal detection value set until the Qth attribute initial low-field nuclear magnetic resonance signal detection value set, respectively performing dimensionless processing to obtain a first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence until the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence; Taking the melon pulp hardness characteristic value sequence as the reference sequence and the first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence to the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence as the comparison sequence, a grey correlation matrix is ​​constructed, and grey correlation analysis is performed to obtain the initial low-field nuclear magnetic resonance signal attribute correlation set.

[0074] The method comprises extracting the low-field nuclear magnetic resonance signal of healthy melon pulp that meets the healthy melon pulp hardness test value based on the associated low-field nuclear magnetic resonance signal attribute set, and constructing a label for identifying the standard low-field nuclear magnetic resonance signal, including: Obtaining a plurality of first associated attribute low-field nuclear magnetic resonance signal detection values ​​up to a plurality of P-th associated attribute low-field nuclear magnetic resonance signal detection values ​​of a plurality of healthy melon pulps whose hardness detection values ​​belong to the healthy melon pulp hardness detection value, where P represents the total number of attributes in the associated low-field nuclear magnetic resonance signal attribute set; Performing a centralized value evaluation on the plurality of first correlation attribute low-field nuclear magnetic resonance signal detection values ​​to obtain a first correlation attribute standard low-field nuclear magnetic resonance signal characteristic value; Until the plurality of P-th associated attribute low-field nuclear magnetic resonance signal detection values ​​are collectively evaluated to obtain the P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value; Based on the first associated attribute standard low-field nuclear magnetic resonance signal characteristic value up to the P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value, a label identifying the standard low-field nuclear magnetic resonance signal is constructed.

[0075] Furthermore, the target melon pulp hardness prediction module 13 includes the following execution steps: Based on the associated low-field NMR signal attribute set, a signal input branch set is constructed; Based on the fully connected neural network, configure the backbone network; Connecting the output nodes of the signal input branch set in parallel to the backbone network to obtain a melon flesh firmness predictor architecture; A label identifying the melon pulp hardness value is placed in the output layer of the melon pulp hardness predictor architecture, and a low-field nuclear magnetic resonance recording signal of the melon pulp is placed in the input layer of the melon pulp hardness predictor architecture. The multiple sets of data are retrieved, and the melon pulp hardness predictor architecture is trained to obtain the melon pulp hardness predictor.

[0076] Among them, based on the associated low-field nuclear magnetic resonance signal attribute set, a signal input branch set is constructed, including: extracting a first correlated low-field nuclear magnetic resonance signal attribute from the correlated low-field nuclear magnetic resonance signal attribute set; When the storage format of the first associated low-field nuclear magnetic resonance signal attribute is an image, configuring a signal input convolution branch; When the storage format of the first associated low-field nuclear magnetic resonance signal attribute is a numerical value, a signal input fully connected branch is configured.

[0077] Wherein, when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, and then the method further includes: The target melon pulp low-field nuclear magnetic resonance signal and the target melon pulp hardness value are associated and stored, and a melon pulp hardness calibration database is constructed using the target melon pulp low-field nuclear magnetic resonance signal as an index condition and the target melon pulp hardness value; Wherein, after the melon pulp hardness calibration database is constructed: When the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, performing hardness value calibration through the melon pulp hardness calibration database to obtain a melon pulp hardness value calibration result; When the calibration result of the melon pulp hardness value is empty, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value.

[0078] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0079] Those skilled in the art will appreciate that embodiments of the present invention may provide methods, systems, or computer program products. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0080] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0081] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0083] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0084] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An intelligent detection method for melon pulp hardness based on low-field nuclear magnetic resonance, characterized in that: include: Retrieving healthy melon flesh hardness test values ​​based on melon variety, planting plan, and planting duration, wherein the healthy melon flesh hardness test values ​​have a label identifying a standard low-field nuclear magnetic resonance signal; When the target melon pulp low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, the healthy melon pulp hardness detection value is set as the target melon pulp hardness value, wherein the consistent characterization type attribute signals are the same, and the quantitative attribute signal deviations are less than or equal to the corresponding quantitative attribute signal deviation threshold; When the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, wherein the melon pulp hardness predictor is generated by machine learning training through multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value.

2. The method according to claim 1, wherein The healthy melon flesh hardness test value is retrieved based on the melon variety, planting plan, and planting duration as constraints, wherein the healthy melon flesh hardness test value has a label identifying a standard low-field nuclear magnetic resonance signal, including: obtaining an initial low-field NMR signal property set; Based on the hardness of the melon pulp, a correlation analysis is performed on the initial low-field nuclear magnetic resonance signal attribute set to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set; Based on the initial low-field nuclear magnetic resonance signal attribute correlation degree set, sorting, from the initial low-field nuclear magnetic resonance signal attribute set, associated low-field nuclear magnetic resonance signal attribute sets whose initial low-field nuclear magnetic resonance signal attribute correlation degrees are greater than or equal to a correlation degree threshold; Using the melon variety, planting plan, and planting duration as constraints, retrieve a set of healthy melon flesh hardness record values, perform box plot analysis, and obtain the box intervals of the healthy melon flesh hardness record values, which are set as the healthy melon flesh hardness test values; Based on the associated low-field nuclear magnetic resonance signal attribute set, the low-field nuclear magnetic resonance signal of the healthy melon pulp that meets the healthy melon pulp hardness test value is extracted, and a label for identifying the standard low-field nuclear magnetic resonance signal is constructed.

3. The method according to claim 2, wherein Based on the hardness of the melon pulp, a correlation analysis is performed on the initial low-field nuclear magnetic resonance signal attribute set to obtain an initial low-field nuclear magnetic resonance signal attribute correlation degree set, including: Load the corresponding melon flesh hardness record value set, the first attribute initial low-field nuclear magnetic resonance signal detection value set, and the Qth attribute initial low-field nuclear magnetic resonance signal detection value set, where Q represents the number of attributes in the initial low-field nuclear magnetic resonance signal attribute set; performing dimensionless processing on the melon pulp hardness record value set to obtain a melon pulp hardness characteristic value sequence; Traversing the first attribute initial low-field nuclear magnetic resonance signal detection value set until the Qth attribute initial low-field nuclear magnetic resonance signal detection value set, respectively performing dimensionless processing to obtain a first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence until the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence; Taking the melon pulp hardness characteristic value sequence as the reference sequence and the first attribute initial low-field nuclear magnetic resonance signal characteristic value sequence to the Qth attribute initial low-field nuclear magnetic resonance signal characteristic value sequence as the comparison sequence, a grey correlation matrix is ​​constructed, and grey correlation analysis is performed to obtain the initial low-field nuclear magnetic resonance signal attribute correlation set.

4. The method according to claim 2, wherein Based on the associated low-field nuclear magnetic resonance signal attribute set, extracting the low-field nuclear magnetic resonance signal of healthy melon pulp that meets the healthy melon pulp hardness test value, and constructing a label for identifying the standard low-field nuclear magnetic resonance signal, including: Obtaining a plurality of first associated attribute low-field nuclear magnetic resonance signal detection values ​​up to a plurality of P-th associated attribute low-field nuclear magnetic resonance signal detection values ​​of a plurality of healthy melon pulps whose hardness detection values ​​belong to the healthy melon pulp hardness detection value, where P represents the total number of attributes in the associated low-field nuclear magnetic resonance signal attribute set; Performing a centralized value evaluation on the plurality of first correlation attribute low-field nuclear magnetic resonance signal detection values ​​to obtain a first correlation attribute standard low-field nuclear magnetic resonance signal characteristic value; Until the plurality of P-th associated attribute low-field nuclear magnetic resonance signal detection values ​​are collectively evaluated to obtain the P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value; Based on the first associated attribute standard low-field nuclear magnetic resonance signal characteristic value up to the P-th associated attribute standard low-field nuclear magnetic resonance signal characteristic value, a label identifying the standard low-field nuclear magnetic resonance signal is constructed.

5. The method according to claim 2, wherein The target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, wherein the melon pulp hardness predictor is generated by machine learning training using multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value, including: Based on the associated low-field NMR signal attribute set, a signal input branch set is constructed; Based on the fully connected neural network, configure the backbone network; Connecting the output nodes of the signal input branch set in parallel to the backbone network to obtain a melon flesh firmness predictor architecture; A label identifying the melon pulp hardness value is placed in the output layer of the melon pulp hardness predictor architecture, and a low-field nuclear magnetic resonance recording signal of the melon pulp is placed in the input layer of the melon pulp hardness predictor architecture. The multiple sets of data are retrieved, and the melon pulp hardness predictor architecture is trained to obtain the melon pulp hardness predictor.

6. The method according to claim 5, wherein Based on the associated low-field NMR signal attribute set, a signal input branch set is constructed, including: extracting a first correlated low-field nuclear magnetic resonance signal attribute from the correlated low-field nuclear magnetic resonance signal attribute set; When the storage format of the first associated low-field nuclear magnetic resonance signal attribute is an image, configuring a signal input convolution branch; When the storage format of the first associated low-field nuclear magnetic resonance signal attribute is a numerical value, a signal input fully connected branch is configured.

7. The method according to claim 1, wherein When the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value, and then the method further includes: The target melon pulp low-field nuclear magnetic resonance signal and the target melon pulp hardness value are associated and stored, and a melon pulp hardness calibration database is constructed using the target melon pulp low-field nuclear magnetic resonance signal as an index condition and the target melon pulp hardness value; Wherein, after the melon pulp hardness calibration database is constructed: When the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, performing hardness value calibration through the melon pulp hardness calibration database to obtain a melon pulp hardness value calibration result; When the calibration result of the melon pulp hardness value is empty, the target melon pulp low-field nuclear magnetic resonance signal is processed by a melon pulp hardness predictor to obtain the target melon pulp hardness value.

8. An intelligent detection system for melon pulp hardness based on low-field nuclear magnetic resonance, characterized in that: The system is used to implement the intelligent detection method for melon pulp hardness based on low-field nuclear magnetic resonance as described in any one of claims 1 to 7, comprising: A healthy melon flesh hardness calibration module is used to retrieve healthy melon flesh hardness test values ​​based on melon variety, planting plan, and planting duration, wherein the healthy melon flesh hardness test values ​​have a label identifying a standard low-field nuclear magnetic resonance signal; a target melon pulp hardness matching module, configured to set the healthy melon pulp hardness detection value as the target melon pulp hardness value when the target melon pulp low-field nuclear magnetic resonance signal is consistent with the standard low-field nuclear magnetic resonance signal, wherein the consistent characterization type attribute signals are the same and the quantitative attribute signal deviations are less than or equal to the corresponding quantitative attribute signal deviation threshold; The target melon pulp hardness prediction module is used to process the target melon pulp low-field nuclear magnetic resonance signal through a melon pulp hardness predictor to obtain the target melon pulp hardness value when the target melon pulp low-field nuclear magnetic resonance signal is inconsistent with the standard low-field nuclear magnetic resonance signal, wherein the melon pulp hardness predictor is generated by machine learning training through multiple sets of data, and any set of the multiple sets of data includes the melon pulp low-field nuclear magnetic resonance recording signal and a label identifying the melon pulp hardness value.

Images