Multi-element fusion type mango maturity detection method, device and equipment and storage medium

Abstract

The invention discloses a multivariate fusion type mango maturity detection method, device and equipment and a storage medium, and relates to the technical field of fruit maturity evaluation.The multivariate fusion type mango maturity detection method comprises the steps that a first gas concentration data matrix is obtained, and a second gas concentration data matrix is obtained based on an optimal data analysis strategy; dividing the second gas concentration data matrix into a plurality of third arrays according to columns; calculating a correlation coefficient and a variation coefficient of each third array, and deleting an abnormal third array; fusing the remaining third array to obtain gas sensitive data; screening an optimal combination of the model and the optimizer, and obtaining image data based on the mango images in different periods and the combination; inputting the gas-sensitive data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity; obtaining second maturity based on physical characteristics of mangoes; and verifying the first maturity based on the second maturity. According to the method, the mango maturity detection efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of fruit maturity evaluation, and in particular to a multi-integrated mango maturity detection method, device, equipment and storage medium. Background Art

[0002] At present, people at home and abroad are paying more and more attention to the quality of fruits. High-quality fruits are often more popular with people, and methods for fruit quality testing are also gradually developing.

[0003] Traditional mango maturity detection mainly relies on manual sensory evaluation, such as observing the appearance, color, smelling the fragrance, and listening to the sound by tapping. This method is highly subjective and uncertain, and can easily lead to errors in the test results. In the mango picking, packaging, storage and other operations, maturity detection needs to be completed quickly and accurately. However, the existing detection devices may still be insufficient in flexibility and convenience, and it is difficult to meet the needs of actual production.

[0004] Therefore, how to improve the efficiency of mango maturity detection is a problem that needs to be solved urgently.

[0005] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0006] The main purpose of this application is to provide a multi-integrated mango maturity detection method, device, equipment and storage medium, aiming to solve the technical problem of how to improve the efficiency of mango maturity detection.

[0007] To achieve the above-mentioned purpose, the present application proposes a multi-integrated mango maturity detection method, which is applied to a multi-integrated mango maturity detection device, characterized in that the device comprises: a host computer, an arc conveyor belt, a chain plate with grooves, a collection cover, a high-speed camera, a flexible push rod, a cylindrical air chamber with a plurality of gas sensors and a needle probe, and the method comprises:

[0008] Acquire a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor;

[0009] Processing the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix;

[0010] Dividing the second gas concentration data matrix into a plurality of third arrays by columns, wherein each gas sensor corresponds to one third array;

[0011] Calculate the correlation coefficient and the coefficient of variation of each of the third arrays, and delete the third arrays whose correlation coefficients do not belong to the preset correlation coefficient interval and whose coefficients of variation do not belong to the preset coefficient of variation interval;

[0012] fusing the remaining third arrays to obtain gas-sensing data;

[0013] Obtaining mango images at different times, screening an optimal combination of a model and an optimizer, and obtaining image data based on the mango images and the combination;

[0014] Inputting the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity;

[0015] Acquiring physical properties of mangoes, and obtaining a second degree of maturity based on the physical properties;

[0016] Based on the second maturity, the first maturity is verified.

[0017] In one embodiment, obtaining the first gas concentration matrix includes:

[0018] collecting voltage changes by using a plurality of gas sensors, wherein the voltage changes are caused by concentration changes of components contained in the mango gas;

[0019] Convert each of the voltage changes into a digital signal to obtain a plurality of first arrays;

[0020] Taking an average value of a fixed number of consecutive elements in each of the first arrays, and performing standardization processing on the average value to obtain a plurality of second arrays;

[0021] The plurality of second arrays are used as a first gas concentration data matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the second arrays.

[0022] In one embodiment, the step of processing the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix includes:

[0023] Get multiple data analysis strategies;

[0024] Using the data analysis strategy to process the first gas concentration data matrix respectively, and determine a fixed number of principal components of the gas;

[0025] Calculating the variance contribution rate of each of the principal components to obtain the sum of the variance contribution rates;

[0026] determining an optimal data analysis strategy based on the sum;

[0027] Processing the first gas concentration data matrix according to the optimal data analysis strategy to divide maturity standard values, wherein the maturity includes unripe, partially ripe, and ripe;

[0028] The first gas concentration data matrix processed by the optimal data analysis strategy is used as the second gas concentration data matrix.

[0029] In one embodiment, the calculating of the correlation coefficient and the coefficient of variation of each of the third arrays, and deleting the third arrays whose correlation coefficient does not belong to a preset correlation coefficient interval and whose coefficient of variation does not belong to a preset coefficient of variation interval, comprises:

[0030] Calculating a plurality of first average values ​​of the elements of each of the third arrays and a second average value of the maturity standard value;

[0031] Obtain a plurality of correlation coefficients according to each of the third arrays, each of the first average values, and the second average values;

[0032] If the correlation coefficient does not belong to the preset correlation coefficient interval, removing the third array corresponding to the correlation coefficient;

[0033] Calculating a plurality of sample standard deviations of each of the third arrays;

[0034] Obtaining a plurality of coefficients of variation according to the corresponding first mean values ​​and the sample standard deviations;

[0035] If the coefficient of variation does not belong to the preset coefficient of variation interval, the third array corresponding to the coefficient of variation is removed.

[0036] In one embodiment, the acquiring of mango images at different periods, selecting an optimal combination of a model and an optimizer, and obtaining image data based on the mango images and the combination, comprises:

[0037] Get images of mangoes at different times;

[0038] Preprocessing the mango image and annotating the mango image to obtain a data set;

[0039] Dividing the data set into a training set and a validation set;

[0040] Acquire multiple models and multiple optimizers to obtain different combinations of the models and the optimizers;

[0041] Training each of the combinations according to the training set and the validation set to obtain a plurality of training results, wherein the training results correspond one-to-one to the combination of the model and the optimizer;

[0042] Obtaining the accuracy and loss value of the corresponding combination according to each of the training results;

[0043] Determining the optimal combination of the optimizer and the model according to the accuracy and loss values;

[0044] The model optimized by the optimizer of the optimal combination is used as the final model;

[0045] The data set is used as input of the final model to obtain image data.

[0046] In one embodiment, the step of inputting the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity includes:

[0047] Extracting features of the gas-sensitive data;

[0048] Combining the extracted features to obtain a first feature vector;

[0049] Counting the number of data corresponding to each color in the image data to obtain multiple color features;

[0050] extracting geometric features of the image data;

[0051] Obtaining a second feature vector according to the color feature and the geometric feature;

[0052] Concatenate the first eigenvector and the second eigenvector to obtain a third eigenvector;

[0053] The third eigenvector is used as an input of a preset mango maturity comprehensive detection model to obtain a first maturity.

[0054] In one embodiment, the step of obtaining the physical properties of the mango and obtaining the second maturity based on the physical properties includes:

[0055] Obtaining a mango region sample, wherein the region includes the head, middle, and tail of the mango;

[0056] Obtaining physical properties based on the mango region sample, wherein the physical properties include strength, hardness and toughness of the mango peel;

[0057] Performing correlation analysis based on the physical properties to obtain the physical property correlation coefficient;

[0058] Setting the physical property weight;

[0059] Based on the weight and the correlation coefficient, a weighted score is calculated, and the second maturity is obtained according to the weighted score. The calculation formula is as follows:

[0060] Weighted score = hardness correlation coefficient × W1 + strength correlation coefficient × W2 + toughness correlation coefficient × W3

[0061] Among them, the physical property correlation coefficient includes a hardness correlation coefficient, a strength correlation coefficient and a toughness correlation coefficient, W1 represents the hardness weight, W2 represents the strength weight, W3 represents the toughness weight, and satisfies W1+W2+W3=1.

[0062] In addition, to achieve the above-mentioned purpose, the present application also proposes a multi-integrated mango maturity detection device, characterized in that the device comprises:

[0063] An acquisition module, acquiring a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor;

[0064] A preprocessing module, configured to process the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix; and divide the second gas concentration data matrix into a plurality of third arrays according to columns, wherein each gas sensor corresponds to one third array;

[0065] A screening module, used for calculating the correlation coefficient and the coefficient of variation of each of the third arrays, and deleting the third arrays whose correlation coefficient does not belong to a preset correlation coefficient interval and whose coefficient of variation does not belong to a preset coefficient of variation interval;

[0066] A fusion module, used for fusing the remaining third arrays to obtain gas sensing data;

[0067] An image recognition module, used to obtain mango images at different times, select an optimal combination of a model and an optimizer, and obtain image data based on the mango images and the combination;

[0068] The output module inputs the gas-sensitive data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity; obtains the physical properties of the mango, and obtains a second maturity based on the physical properties; verifies the first maturity based on the second maturity; and verifies the first maturity based on the second maturity.

[0069] In addition, to achieve the above purpose, the present application also proposes a multi-integrated mango maturity detection device, characterized in that the device includes: a host computer, an arc conveyor belt, a chain plate with grooves, a collection cover, a high-speed camera, a flexible push rod, a cylindrical air chamber with a plurality of gas sensors and a needle probe, a temperature and humidity controller is arranged in the collection cover, and a dynamic seal is adopted at the junction of the collection cover and the groove:

[0070] The host computer is used to obtain a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor;

[0071] The host computer is further used to process the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix;

[0072] The host computer is further used to divide the second gas concentration data matrix into a plurality of third arrays according to the columns, wherein each gas sensor corresponds to one third array;

[0073] The host computer is further used to calculate the correlation coefficient and the coefficient of variation of each of the third arrays, and delete the third arrays whose correlation coefficient does not belong to the preset correlation coefficient interval and whose coefficient of variation does not belong to the preset coefficient of variation interval;

[0074] The host computer is also used to fuse the remaining third arrays to obtain gas-sensitive data;

[0075] The host computer is also used to obtain mango images at different times, select the best combination of the model and the optimizer, and obtain image data based on the mango images and the combination;

[0076] The host computer is further used to input the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity;

[0077] The host computer is further used to obtain physical properties of mangoes, and obtain a second maturity based on the physical properties;

[0078] The host computer is further used to verify the first maturity based on the second maturity.

[0079] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the multi-integrated mango maturity detection method as described above are implemented.

[0080] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of the multi-integrated mango maturity detection method as described above are implemented.

[0081] One or more technical solutions proposed in this application have at least the following technical effects:

[0082] The present application provides a multi-integrated mango maturity detection device, characterized in that the device comprises: a host computer, an arc conveyor belt, a chain plate with grooves, a collection cover, a high-speed camera, a flexible push rod, a cylindrical air chamber with a plurality of gas sensors and a needle probe, wherein a temperature and humidity controller is arranged in the collection cover, and a dynamic seal is adopted at the joint between the collection cover and the groove, and the concentration change data of several specific gases exhaled by the detection sample are obtained by the gas sensor, and the optimal data analysis strategy is screened out to process the concentration change data, and then based on the correlation coefficient and the coefficient of variation of the processed data, the unreasonable data is deleted, the remaining data is integrated to obtain the gas data, and mango pictures of different periods are obtained, and the mango pictures are processed by the most suitable model to obtain the picture data, and the obtained gas data and image data are processed and integrated, and the integrated data are input into the comprehensive detection of mango maturity for comprehensive judgment. In addition, the present invention also evaluates the maturity of mango by obtaining the physical properties of mango peel. Due to the combination of gas sensors and image recognition technology, the two methods can complement each other to obtain more comprehensive data analysis with higher accuracy. Both gas sensors and image recognition are automated collection processes, which can reduce manual and manual operation errors. Combined with the method of judging the maturity of mangoes through physical properties, a comprehensive judgment of the maturity of mangoes can be achieved, while improving the efficiency of mango maturity detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0084] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0085] Figure 1 This is a schematic diagram of the structure of the multi-integrated mango maturity detection device of this application;

[0086] Figure 2 The structure of the collecting cover and the groove of the multi-integrated mango maturity detection device of this application;

[0087] Figure 3 The air chamber of the multi-integrated mango maturity detection device of this application;

[0088] Figure 4 A schematic diagram of the process provided for Example 1 of the multi-integrated mango maturity detection method of the present application;

[0089] Figure 5 This is a schematic diagram of the module structure of the multi-integrated mango maturity detection device according to an embodiment of the present application.

[0090] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0091] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0092] At present, the arc conveyor belt of the multi-integrated mango maturity detection equipment includes multiple chain plates, and a groove is arranged on the surface of each chain plate (the groove can be set to a circle with a surface area of ​​25 cm and a depth of 5 cm). The bottom material of the groove is selected to be a relatively soft material (such as silicone rubber). First, the host computer controls the high-speed camera to complete the image acquisition work, collects mango images at different periods, so as to obtain image data later, and pushes the mango into the groove on the surface through a flexible push rod (the material can be selected to be silicone rubber). The collection cover will drop to form a seal with the groove. A small hole is opened above the collection cover, and an air pump is connected to the air pump through a pipeline to evacuate the enclosed space formed by the groove and the collection cover. The gas enters the air chamber equipped with a gas sensitive sensor along the pipeline. After the gas collection is completed, sampling and detection are carried out. The host computer controls the needle probe to puncture the sample to obtain data.

[0093] Reference Figure 1 , Figure 1 This is the structural diagram of the multi-integrated mango maturity detection equipment.

[0094] It should be noted that if Figure 2 As shown, a dynamic sealing structure can be used where the groove and the collection cover are combined. A sealing groove can be set at the contact edge of the groove and the collection cover, and an O-ring seal can be installed at the opening edge of the groove. When the collection cover falls into place, the seal will enter the sealing groove of the upper collection cover under the action of dynamic pressure, thereby forming a seal. At the same time, a temperature and humidity controller is set in the collection cover to adjust the sealing environment. Exemplarily, the sealing material can be nitrile rubber, and the sealing grade can be selected from ISO 6149, which is not limited in this embodiment. Figure 3 As shown, the shape of the air chamber is cylindrical, and a ring distribution is adopted according to the central axis of the cylindrical air chamber. The arrangement of the sensors is completed by setting the angle between the sensors and the distance between the sensors along the central axis (the angle between the sensors can be between 10 degrees and 30 degrees, and the distance between the sensors along the central axis can be between 3cm and 5cm). This embodiment does not impose any restrictions on this.

[0095] It should also be noted that the above-mentioned host computer is a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device that can realize the above functions, and this embodiment does not limit this.

[0096] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0097] The main solution of the embodiment of the present application is: obtain a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in the mango gas collected by each gas sensor; process the first gas concentration matrix based on the optimal data analysis strategy to obtain a second gas concentration data matrix; divide the second gas concentration data matrix into multiple third arrays according to the columns, wherein each gas sensor corresponds to a third array; calculate the correlation coefficient and the coefficient of variation of each of the third arrays, delete the third arrays whose correlation coefficient does not belong to the preset correlation coefficient interval and whose coefficient of variation does not belong to the preset coefficient of variation interval; merge the remaining third arrays to obtain gas-sensitive data; obtain mango images of different periods, screen the optimal combination of model and optimizer, and obtain image data based on the mango image and the combination; input the gas-sensitive data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity; obtain the physical properties of mango, obtain a second maturity based on the physical properties, and verify the first maturity based on the second maturity.

[0098] In this embodiment, for the convenience of description, the following description is made by taking the identification of the multi-integrated mango maturity detection device as the execution subject.

[0099] Since traditional mango maturity detection mainly relies on manual sensory evaluation, such as observing the appearance, color, smelling the fragrance, and tapping to listen to the sound, this method has great subjectivity and uncertainty, which can easily lead to errors in the test results. In the operation links such as mango picking, packaging, and storage, maturity detection needs to be completed quickly and accurately. However, the existing detection devices may still be insufficient in flexibility and convenience, and it is difficult to meet the needs of actual production. Therefore, how to improve the efficiency of mango maturity detection is a problem that needs to be solved urgently.

[0100] The present application provides a solution, which uses a gas sensor to obtain the concentration change data of several specific gases exhaled by the test sample, selects the optimal data analysis strategy to process the concentration change data, and then deletes the unreasonable data based on the correlation coefficient and coefficient of variation of the processed data, fuses the remaining data, obtains gas-sensitive data, obtains mango pictures of different periods, and processes the mango pictures through the most suitable model to obtain picture data, fuses the obtained gas-sensitive data and image data after processing, and inputs the fused data into the comprehensive detection of mango maturity for comprehensive judgment. In addition, the present invention also evaluates the maturity of mango by obtaining the physical properties of mango peel. Due to the combination of gas-sensitive sensors and image recognition technology, the two methods can complement each other, obtain more comprehensive data analysis, and have higher accuracy. Both gas-sensitive sensors and image recognition are automated collection processes, which can reduce manual and manual operation errors. Combined with the method of judging the maturity of mangoes by physical properties, the maturity of mangoes can be comprehensively judged, and the efficiency of mango maturity detection is improved.

[0101] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, or an electronic device capable of realizing the above functions, a multi-integrated mango maturity detection device, etc. The following takes the multi-integrated mango maturity detection device as an example to illustrate this embodiment and the following embodiments.

[0102] Based on this, the present application embodiment provides a multi-integrated mango maturity detection method, referring to Figure 4 , Figure 4 This is a flow chart of the first embodiment of the multi-integrated mango maturity detection method of the present application.

[0103] In this embodiment, the multi-integrated mango maturity detection method is applied to a multi-integrated mango maturity detection device, characterized in that the device at least includes: a host computer, an arc conveyor belt, a chain plate with grooves, a collection cover, a high-speed camera, a flexible push rod, a cylindrical air chamber with a plurality of gas sensors and a needle probe, a temperature and humidity controller is provided in the collection cover, and a dynamic seal is used at the junction of the collection cover and the groove. The multi-integrated mango maturity detection method includes steps S10 to S80:

[0104] Step S10, obtaining a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor;

[0105] It should be noted that a gas sensor is a sensor that detects the type and concentration of specific gases. When the concentration of a specific gas in the air changes, it will cause the voltage of the corresponding odor sensor to change, and convert the electrical signal obtained by the sensor into a digital signal. Each column of the first gas concentration data represents the digital signal data corresponding to the concentration change of different components.

[0106] It can be understood that since the data collected by the gas sensor is processed into a matrix, the two-dimensional data can be easily represented and processed, and converting multiple arrays into matrices can make the storage and operation of the data more standardized.

[0107] In a feasible implementation manner, the step S10 may include steps A11 to A14:

[0108] Step A11, collecting voltage changes through a plurality of gas sensors, wherein the voltage changes are caused by concentration changes of components contained in mango gas;

[0109] It should be noted that the voltage change of the gas sensor refers to the change in electrical properties, especially resistance or voltage, caused by the interaction between the target gas molecules and the solid material on the sensor surface during the operation of the gas sensor. These changes can be detected and analyzed by an external circuit to achieve gas detection.

[0110] It is understandable that, since the gas emitted during the ripening process of mangoes includes multiple components (such as alcohols, alkenes, lipids, sulfides, etc.), the concentration of each component will change, so step S10 is performed, and the change in the concentration of each component will cause the voltage of the corresponding gas sensor to change.

[0111] Step A12, converting each of the voltage changes into a digital signal to obtain a plurality of first arrays;

[0112] It should be noted that there is a one-to-one correspondence between the gas sensors and the first arrays.

[0113] It can be understood that converting electrical signals into digital signals can intuitively reflect changes in concentration.

[0114] Step A13, taking an average value of a fixed number of consecutive elements in each of the first arrays, and performing standardization processing on the average value to obtain a plurality of second arrays;

[0115] It should be noted that the continuous fixed number of elements may be 4 consecutive elements in the first array.

[0116] Exemplarily, the fixed number of elements is 4, and the data of the first array is X1, X2, X3, X4, X5, ... Xn.

[0117] The first data point average:

[0118]

[0119] The second data point average:

[0120]

[0121] And so on, the calculation is completed, and the new data set is [Y1, Y2, Y3...Yn].

[0122] Then, standardization is performed. For the newly obtained data set [Y1, Y2, Y3...Yn], the mean μ and standard deviation σ corresponding to the data set are first calculated, and then standardized:

[0123]

[0124] Z i is the standardized data, and the new standardized data set is [Z1, Z2, ..., Z n ], the data collected by each gas sensor needs to be preprocessed, as shown in Table 1:

[0125] Table 1

[0126]

[0127] It is understandable that after data preprocessing, random fluctuations in the data can be reduced, making it easier to compare and summarize the data.

[0128] Step A14: using the plurality of second arrays as a first gas concentration data matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the second arrays.

[0129] The above is only a feasible implementation of step S10 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S10.

[0130] Step S20, processing the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix;

[0131] It should be noted that the data analysis strategy may be a data processing method, including logarithmic method, difference method, partial fraction method, and relative difference method, and the second gas concentration matrix may be data processed by one of the four data analysis strategies.

[0132] It can be understood that by processing the first gas concentration matrix through the optimal data analysis strategy, more reliable data can be obtained, which is beneficial to the subsequent data processing.

[0133] In a feasible implementation, step S30 may include steps B11 to B16:

[0134] Step B11, obtaining multiple data analysis strategies;

[0135] Step B12, using the data analysis strategy to process the first gas concentration data matrix respectively to determine a fixed number of principal components of the gas;

[0136] It should be noted that the fixed number can be any real number, for example, 3. The principal component is the component with a higher concentration in the gas emitted by mango.

[0137] Exemplarily, the relative difference method is used as a reference, as follows: the preprocessed data is selected as V(t), and the relative difference method is used for calculation:

[0138]

[0139] Among them, △V(t) represents the relative change at the current moment, V(t) is the first gas concentration matrix at the current moment, and V(t-1) is the first gas concentration matrix at the previous moment. The preprocessed data is calculated point by point to obtain the relative difference value at each moment, and a relative difference value curve is drawn. The threshold ranges corresponding to the three maturity levels are set through preliminary experiments, such as: unripe represents below threshold A; partially mature represents between A and B; mature represents above threshold B.

[0140] Taking the results after relative difference method processing as an example, we first need to integrate the data processed by the relative difference method, stack all the data sets processed by the relative difference method to form a relative difference result matrix, in which each column represents the relative difference data of a sensor, and then calculate the covariance matrix, and perform eigendecomposition on the covariance matrix to obtain its corresponding eigenvalues ​​and eigenvectors, and select the first three principal components according to the size of the eigenvalues.

[0141] The matrix form of the relative difference method result is:

[0142] Sensor 1 Sensor 2 …… Sensor D1 E1 …… Z1 D2 E2 …… Z2 D3 E3 …… Z3 …… …… …… …… Dn En …… Zn

[0143] First, the average reading of each sensor is calculated, and then the mean is subtracted from the reading of each sensor to get the bias matrix. The bias matrix is ​​then multiplied by the transpose of the bias matrix and divided by the number of samples minus one to get the covariance matrix.

[0144] The deviation matrix D is in the form:

[0145] Sensor 1 Sensor 2 …… Sensor <![CDATA[D1-μ1]]> <![CDATA[E1-μ2]]> …… <![CDATA[Z1-μ n ]]> <![CDATA[D2-μ1]]> <![CDATA[E2-μ2]]> …… <![CDATA[Z2-μ n ]]> <![CDATA[D3-μ1]]> <![CDATA[E3-μ2]]> …… <![CDATA[Z3-μ n ]]> …… …… …… …… <![CDATA[Dn-μ1]]> <![CDATA[En-μ2]]> …… <![CDATA[Zn-μ n ]]>

[0146] Among them, μ i is the average reading for each sensor.

[0147] Calculate the covariance matrix Σ:

[0148]

[0149] Among them, D T is the transpose of the deviation matrix D, and n is the number of data.

[0150] The matrix is ​​then solved to find its characteristics and eigenvectors, and the eigenvalues ​​are used to determine which principal components are more representative of the data, and the top three principal components are selected.

[0151] Step B13, calculating the variance contribution rate of each of the principal components to obtain the sum of the variance contribution rates;

[0152] It should be noted that the variance contribution rate is the ratio of its eigenvalue to the sum of all eigenvalues, which is used to calculate the variance contribution rate of each principal component.

[0153] Step B14, determining an optimal data analysis strategy based on the sum;

[0154] It should be noted that the data analysis strategy with the highest total is selected as the optimal analysis strategy.

[0155] Step B15, processing the first gas concentration data matrix according to the optimal data analysis strategy to divide the maturity standard value, wherein the maturity includes unripe, partially ripe, and ripe;

[0156] It should be noted that threshold ranges corresponding to three levels of maturity are set, such as: unripe, below threshold A; partially ripe, between A and B; ripe: above threshold B. By observing the changing trend of the relative difference value, the standard values ​​of mango maturity can be divided. For example, the standard value of unripe is A.

[0157] Step B16: using the first gas concentration data matrix processed by the optimal data analysis strategy as the second gas concentration data matrix.

[0158] Exemplarily, the relative variance method result matrix is ​​used as the second gas concentration data matrix.

[0159] In this embodiment, the result matrix of the data processing strategy is obtained, and the result matrix of the preprocessing of the four strategies is compared and verified by using the principal component analysis method, and the variance contribution rate of each of the four methods is calculated to evaluate which strategy is more effective, thereby improving the accuracy of mango ripeness detection.

[0160] The above is only a feasible implementation of step S20 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S20.

[0161] Step S30, dividing the second gas concentration data matrix into a plurality of third arrays according to the columns, wherein each gas sensor corresponds to one third array;

[0162] Step S40, calculating the correlation coefficient and the coefficient of variation of each of the third arrays, and deleting the third arrays whose correlation coefficients do not belong to a preset correlation coefficient interval and whose coefficients of variation do not belong to a preset coefficient of variation interval;

[0163] It should be noted that the correlation coefficient can be the relationship between the component concentration corresponding to the sensor and the maturity of the mango. The closer the correlation value is to 1 or -1, the stronger the correlation is. The coefficient of variation reflects the degree of discreteness of the third array data corresponding to the sensor. The larger the coefficient of variation, the higher the degree of discreteness.

[0164] Correlation (R) calculation:

[0165]

[0166] Among them, X represents the gas concentration data output by each sensor, that is, the third array, and Y is the value corresponding to the maturity standard given by the relative difference method. are the average values ​​of the above two variables.

[0167] Coefficient of variation (RSD) calculation:

[0168]

[0169] in, is the above average value, SD is the sample standard deviation of the third array, and the output of different sensors under the same conditions is evaluated by evaluating the size of the RSD value, and the sensor with a lower RSD value is selected.

[0170] It is understandable that a reasonable preset correlation coefficient interval and preset variation coefficient interval can be set to remove abnormal data, improve data quality, reduce noise and errors, and facilitate subsequent data processing.

[0171] Step S50, fusing the remaining third arrays to obtain gas sensing data;

[0172] It should be noted that fusing the third array to obtain gas-sensitive data is an iterative update to gradually improve the estimation accuracy.

[0173] Exemplarily, the remaining third arrays are fused based on the Kalman filter method to obtain gas-sensitive data, and step S50 may include steps C01 to C10;

[0174] Step C01, based on each of the initial errors, setting an initial covariance matrix to a diagonal matrix, wherein the diagonal elements are the initial errors;

[0175] It should be noted that the initial error is a fixed ratio corresponding to the reading range of the sensor. For example, if the reading range of the sensor is 1-10, the fixed ratio can be set to 10%, resulting in an initial error of 1.

[0176] Step C02, setting a unit matrix according to the number of each of the third arrays, and using the unit matrix as a state transfer matrix;

[0177] It should be noted that the state transfer matrix can be used to predict the state at the next moment, that is, the data information collected by the gas sensor at the next moment. By defining the initial state unit matrix, it means that the state remains unchanged.

[0178] Step C03, using each of the third arrays as a column of a matrix to obtain an observation value matrix, and setting the element value of the corresponding position of the observation matrix according to the element value of the observation value matrix;

[0179] It should be noted that the observation matrix can be the actual data collected by the gas sensor, that is, the processed third array. If an element of the observation matrix is ​​an invalid value, the element value at the corresponding position in the observation matrix is ​​set to 0, otherwise, it is set to 1.

[0180] Step C04, calculating the standard deviation of each of the remaining third arrays, and determining the observation noise covariance matrix based on the standard deviation;

[0181] It should be noted that the standard deviation can be the positive square root of the average (i.e., variance) of the sum of squares of the differences between each value in the third array and its mean, which reflects the degree of dispersion between individuals in the group. The observation noise covariance matrix is ​​obtained by performing noise characteristic analysis on the third array data.

[0182] Step C05, obtaining a first predicted state matrix according to the state transfer matrix;

[0183] It should be noted that the first prediction state matrix is ​​calculated according to the formula, where F k represents the state transfer matrix, X k|k-1 represents the first prediction state matrix, x k-1 Represents the historical state estimation matrix, that is, the matrix corresponding to the third array data after fusion in the previous round of data processing.

[0184] x k|k-1 =F k x k-1

[0185] Step C06, obtaining a prediction covariance matrix according to the state transfer matrix, the historical covariance matrix and the process noise covariance matrix;

[0186] It should be noted that the prediction covariance matrix is ​​calculated according to the formula, where P k-1 represents the historical covariance matrix, that is, the covariance matrix in the previous round of data processing, Q k represents the process noise covariance matrix, P k|k-1 Represents the forecast covariance matrix.

[0187]

[0188] Step C07, obtaining a Kalman gain according to the prediction covariance matrix, the observation matrix and the observation noise covariance matrix;

[0189] It should be noted that the Kalman gain is calculated according to the formula, where H k represents the observation matrix, K k stands for Kalman gain.

[0190]

[0191] Step C08, updating the prediction covariance matrix according to the identity matrix, the Kalman gain, and the prediction covariance matrix;

[0192] It should be noted that the prediction covariance matrix is ​​updated according to the formula, where P k represents the updated prediction covariance matrix, and I is the identity matrix.

[0193] P k =(IK k H k ) k|k-1

[0194] Step C09, obtaining a second prediction state matrix according to each of the observation value matrices, the Kalman gain, the observation matrix and the first prediction state matrix;

[0195] It should be noted that the predicted state matrix is ​​calculated according to the formula, where x k represents the second prediction state matrix, z k Represents the observation matrix.

[0196] x k =x k|k-1 +K k (z k -H k x k|k-1 )

[0197] Step C10: using the second predicted state matrix as gas-sensing data.

[0198] In this implementation, the Kalman filter method is used to fuse the data of the remaining sensors, thereby improving the accuracy of the mango ripeness detection data.

[0199] The above is only a feasible implementation of step S50 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S50.

[0200] Step S60, obtaining mango images of different periods, selecting an optimal combination of a model and an optimizer, and obtaining image data based on the mango images and the combination;

[0201] It should be noted that the model can be a Google network (GoogLeNet), a mobile network (MobileNet), a visual geometry group network (Visual Geometry Group, VGG194) and other models; the optimizer is used to optimize the model structure, which can be Stochastic Gradient Descent (SGD), Adaptive Gradient Algorithm (Adagrad) and Adaptive Moment Estimatiom (Adam).

[0202] It can be understood that by comparing the data processing effects of different model and optimizer combinations and selecting the best combination, the accuracy of the data is further improved.

[0203] In a feasible implementation, step S60 may include steps D01 to D09:

[0204] Step D01, obtaining images of mangoes at different periods;

[0205] It should be noted that mangoes at different stages can be of different degrees of maturity.

[0206] Step D02, preprocessing the mango image and annotating the mango image to obtain a data set;

[0207] It should be noted that the dataset is the processed mango image information, including the color channel of the image, the image size, etc.

[0208] Exemplarily, a mango image is loaded, and the channel order is reversed using the Open Source Computer Vision Library (OpenCV) to convert the image into the HSV format, the image size is scaled and adjusted to make the image reach an input size suitable for the model (such as 256×256), and then the image is processed by calling a Gaussian filter to make the image smooth, completing the image preprocessing, and according to the Technical Specification for Mango Quality Evaluation (National Standard NYT4416-2023), the maturity of mangoes is divided into three conditions: unripe, partially ripe, and ripe through the HSV color space, where the H channel is the hue measured in angle, the S channel is the saturation, and the V channel represents the brightness of the color, and the labeling is completed.

[0209] Step D03, dividing the data set into a training set and a validation set;

[0210] It should be noted that the training set can be a data set used to train the model, that is, the model learns how to predict or make decisions through these data, and the validation set is a data set used to evaluate the performance of the model, but does not participate in the training process of the model.

[0211] Step D04, acquiring multiple models and multiple optimizers, and obtaining different combinations of the models and the optimizers;

[0212] Step D05, training each of the combinations according to the training set and the validation set to obtain a plurality of training results, wherein the training results correspond one to one with the combination of the model and the optimizer;

[0213] It should be noted that the training results include the accuracy and loss value of the model.

[0214] Step D06, obtaining the accuracy and loss value of the corresponding combination according to each of the training results;

[0215] It should be noted that accuracy is an indicator to measure the correctness of model prediction, and loss value is an indicator to measure the difference between the model prediction value and the true value.

[0216] Step D07, determining the optimal combination of the optimizer and the model according to the accuracy and loss values;

[0217] Step D08, taking the model optimized by the optimizer of the optimal combination as the final model;

[0218] It should be noted that the final model can be used to process mango images.

[0219] Step D09, using the data set as input of the final model to obtain image data.

[0220] It should be noted that the image data includes image pixel value, size, resolution, color space and other data.

[0221] The above is only a feasible implementation of step S60 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S60.

[0222] Step S70, inputting the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity;

[0223] It should be noted that the preset mango maturity comprehensive detection model can be a random forest model.

[0224] It is understandable that the comprehensive use of gas-sensitive data and image data to judge the maturity of mangoes, and the combination of gas-sensitive sensors and image recognition technology, can allow the two methods to complement each other, obtain more comprehensive data analysis, and have higher accuracy.

[0225] In a feasible implementation, step S70 may include performing feature extraction on the gas-sensitive data; combining the extracted features to obtain a first feature vector; counting the number of data corresponding to each color in the image data to obtain multiple color features; extracting geometric features of the image data; obtaining a second feature vector based on the color features and the geometric features; concatenating the first feature vector and the second feature vector to obtain a third feature vector; and using the third feature vector as input to a random forest model to obtain a first maturity.

[0226] The above is only a feasible implementation of step S70 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S70.

[0227] It should be noted that the features can be basic statistical features extracted by calculations such as mean, median, standard deviation, etc., or they can be waveform features in advance by determining the peaks and troughs of the data and calculating the peaks and slopes. The color features can be the number of pixels of each color in the image.

[0228] For example, time domain feature extraction is selected to extract features from the fused gas-sensing data, and basic statistical features are extracted by calculating the mean, median, standard deviation, etc. The peak and trough of the data are determined, and the peak-to-peak value and slope are calculated to complete the extraction of waveform features. Appropriate features are selected for training, and the selected features are combined into a feature vector X sensor(k1, k2, ..., kL), the image will be subjected to color features (h1, h2, ..., hN) through the convolutional neural network (by counting the number of pixels of each color in the image) and geometric features (p1, p2, ..., pM) using edge detection, and the image features will be assembled into a feature vector X image (h1,h2,...,hN,p1,p2,...,pM). sensor and X image Splice into a new feature vector X = [X sensor ,X image ](k1,k2,...,kL,h1,h2,...,hN,p1,p2,...pM) sends the new feature vector to the random forest model for maturity judgment.

[0229] Step S80, obtaining physical properties of the mango, and obtaining a second maturity based on the physical properties;

[0230] It should be noted that the physical property may be the strength, hardness and toughness of the mango peel, and the second maturity is to test the physical properties of the selected mangoes through a self-made mechanical platform to obtain a judgment on the maturity of the mangoes.

[0231] It is understandable that when using the self-made mechanical platform for physical property testing, mangoes are selected for testing by sampling, which adds a reference indicator to the mango maturity detection and improves the reliability of maturity detection.

[0232] In a feasible implementation, step S80 may include obtaining a mango region sample, wherein the region includes the head, middle and tail of the mango; obtaining physical properties based on the mango region sample, wherein the physical properties include the strength, hardness and toughness of the mango peel; performing correlation analysis based on the physical properties to obtain a correlation coefficient of the physical properties; setting a weight for the physical properties; calculating a weighted score based on the weight and the correlation coefficient, and obtaining a second maturity according to the weighted score, wherein the calculation formula is as follows:

[0233] Weighted score = hardness correlation coefficient × W1 + strength correlation coefficient × W2 + toughness correlation coefficient × W3

[0234] Among them, the physical property correlation coefficient includes a hardness correlation coefficient, a strength correlation coefficient and a toughness correlation coefficient, W1 represents the hardness weight, W2 represents the strength weight, W3 represents the toughness weight, and satisfies W1+W2+W3=1.

[0235] The above is only a feasible implementation of step S80 provided in this embodiment, and this embodiment does not specifically limit the specific implementation of step S80.

[0236] Exemplarily, the Spearman rank correlation coefficient is used to perform correlation analysis on physical characteristics. First, the observed values ​​of each variable are sorted in order of numerical value. After the sorting is completed, each observed value is given a rank according to the sorting, starting from 1. For observations with the same value, their average rank can be taken.

[0237]

[0238] where d i is the difference between the ranks of the two variables, and n is the number of samples.

[0239] The calculated Spearman correlation coefficient indicates whether there is a correlation between the physical properties of different parts of the sample and the strength of the correlation. The correlation coefficient value is between -1 and 1. A value close to 1 indicates a strong positive correlation, a value close to -1 indicates a strong negative correlation, and a value close to zero indicates no correlation. For example, it is found that there is a strong positive correlation between the puncture force of several parts of the sample, indicating that the hardness of mangoes in these parts is relatively consistent. On the contrary, it indicates that there are certain differences in the hardness of different parts. By pre-measuring the peel strength, hardness and toughness of mangoes with different maturity, the intervals of strength, toughness and hardness corresponding to their different maturity stages are established, and the intervals of strength, toughness and hardness correspond to the three levels of immature, partially mature and mature, and set corresponding maturity levels for them, with immature as 1, relatively mature as 2, and mature as 3. The hardness, strength and toughness are given weights W1, W2 and W3 (W1+W2+W3=1) respectively, and their weighted scores are calculated:

[0240] Weighted score = hardness correlation coefficient × W1 + strength correlation coefficient × W2 + toughness correlation coefficient × W3

[0241] The maturity level is determined by assigning weighted scores to corresponding maturity levels.

[0242] Step S90: verifying the first maturity based on the second maturity.

[0243] It is understandable that when a mango reaches the second maturity, the upper machine will compare the first maturity of the mango obtained based on the gas-sensitive data and image data, and use the second maturity to verify the first maturity, thereby improving the reliability of mango maturity detection.

[0244] This embodiment provides a multi-integrated mango maturity detection method, which uses a gas sensor to obtain the concentration change data of several specific gases exhaled by the detection sample, selects the optimal data analysis strategy to process the concentration change data, and then deletes the unreasonable data based on the correlation coefficient and coefficient of variation of the processed data, fuses the remaining data, obtains gas-sensitive data, obtains mango pictures of different periods, and processes the mango pictures through the most suitable model to obtain picture data, fuses the obtained gas-sensitive data and image data after processing, and inputs the fused data into the comprehensive detection of mango maturity for comprehensive judgment. In addition, the present invention also evaluates the maturity of mango by obtaining the physical properties of mango peel. Due to the combination of gas sensors and image recognition technology, the two methods can complement each other, obtain more comprehensive data analysis, and have higher accuracy. Both gas sensors and image recognition are automated collection processes, which can reduce manual and manual operation errors. Combined with the method of judging the maturity of mango by physical properties, the maturity of mango can be comprehensively judged, and the efficiency of mango maturity detection is improved.

[0245] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the multi-integrated mango maturity detection method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0246] This application also provides a multi-integrated mango maturity detection device, please refer to Figure 5 The multi-integrated mango maturity detection device comprises:

[0247] An acquisition module 10 is used to acquire a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor;

[0248] A preprocessing module 20 is used to process the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix; and divide the second gas concentration data matrix into a plurality of third arrays according to columns, wherein each gas sensor corresponds to one third array;

[0249] A screening module 30, configured to calculate the correlation coefficient and the coefficient of variation of each of the third arrays, and delete the third arrays whose correlation coefficients do not belong to a preset correlation coefficient interval and whose coefficients of variation do not belong to a preset coefficient of variation interval;

[0250] A fusion module 40, used for fusing the remaining third arrays to obtain gas sensing data;

[0251] An image recognition module 50 is used to obtain mango images of different periods, select an optimal combination of a model and an optimizer, and obtain image data based on the mango images and the combination;

[0252] The output module 60 is used to input the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity; obtain the physical properties of the mango, and obtain a second maturity based on the physical properties; and verify the first maturity based on the second maturity.

[0253] The multi-integrated mango maturity detection device provided by the present application adopts the multi-integrated mango maturity detection method in the above embodiment, which can solve the technical problem of multi-integrated mango maturity detection. Compared with the prior art, the beneficial effects of the multi-integrated mango maturity detection device provided by the present application are the same as the beneficial effects of the multi-integrated mango maturity detection method provided by the above embodiment, and other technical features in the multi-integrated mango maturity detection device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0254] In one embodiment, the acquisition module 10 is further used to collect voltage changes through multiple gas sensors, wherein the voltage changes are caused by changes in the concentration of components contained in mango gas; convert each of the voltage changes into a digital signal to obtain multiple first arrays; take the average value of a continuous fixed number of elements in each of the first arrays, and standardize the average value to obtain multiple second arrays; use the multiple second arrays as a first gas concentration data matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the second arrays.

[0255] In one embodiment, the preprocessing module 20 is also used to obtain multiple data analysis strategies; use the data analysis strategies to process the first gas concentration data matrix respectively to determine a fixed number of principal components of the gas; calculate the variance contribution rate of each of the principal components to obtain the sum of the variance contribution rates; determine the optimal data analysis strategy based on the sum; process the first gas concentration data matrix according to the optimal data analysis strategy to divide the maturity standard values, wherein the maturity includes unripe, partially ripe, and ripe; and use the first gas concentration data matrix processed by the optimal data analysis strategy as the second gas concentration data matrix.

[0256] In one embodiment, the screening module 30 is also used to calculate multiple first average values ​​of the elements of each of the third arrays and the second average value of the maturity standard value; obtain multiple correlation coefficients based on each of the third arrays, each of the first average values ​​and the second average value; if the correlation coefficient does not belong to a preset correlation coefficient interval, remove the third array corresponding to the correlation coefficient; calculate multiple sample standard deviations of each of the third arrays; obtain multiple coefficients of variation based on the corresponding first average values ​​and each of the sample standard deviations; if the coefficient of variation does not belong to the preset coefficient of variation interval, remove the third array corresponding to the coefficient of variation.

[0257] In one embodiment, the image recognition module 50 is also used to obtain mango images of different periods; preprocess the mango images and annotate the mango images to obtain a data set; divide the data set into a training set and a validation set; obtain multiple models and multiple optimizers to obtain different combinations of the model and the optimizer; train each of the combinations according to the training set and the validation set to obtain multiple training results, wherein the training results correspond to the combination of the model and the optimizer one-to-one; obtain the accuracy and loss value of the corresponding combination according to each of the training results; determine the optimal combination of the optimizer and the model according to each of the accuracy and loss values; use the model optimized by the optimizer of the optimal combination as the final model; use the data set as the input of the final model to obtain image data.

[0258] In one embodiment, the output module 60 is also used to extract features of the gas-sensitive data; combine the extracted features to obtain a first feature vector; count the number of data corresponding to each color in the image data to obtain multiple color features; extract geometric features of the image data; obtain a second feature vector based on the color features and the geometric features; concatenate the first feature vector and the second feature vector to obtain a third feature vector; and use the third feature vector as an input to a random forest model to obtain a first maturity.

[0259] In one embodiment, the output module 60 is further used to obtain a mango region sample, wherein the region includes the head, middle and tail of the mango; obtain physical properties based on the mango region sample, wherein the physical properties include the strength, hardness and toughness of the mango peel; perform correlation analysis based on the physical properties to obtain the physical property correlation coefficient; set the physical property weight; calculate a weighted score based on the weight and the correlation coefficient, and obtain the second maturity according to the weighted score, and the calculation formula is as follows:

[0260] Weighted score = hardness correlation coefficient × W1 + strength correlation coefficient × W2 + toughness correlation coefficient × W3

[0261] Among them, the physical property correlation coefficient includes a hardness correlation coefficient, a strength correlation coefficient and a toughness correlation coefficient, W1 represents the hardness weight, W2 represents the strength weight, W3 represents the toughness weight, and satisfies W1+W2+W3=1.

[0262] The multi-integrated mango maturity detection device provided by the present application adopts the multi-integrated mango maturity detection method in the above embodiment, which can solve the technical problem of multi-integrated mango maturity detection. Compared with the prior art, the beneficial effects of the multi-integrated mango maturity detection device provided by the present application are the same as the beneficial effects of the multi-integrated mango maturity detection method provided by the above embodiment, and the other technical features of the multi-integrated mango maturity detection device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0263] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0264] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0265] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the multi-integrated mango maturity detection method in the above-mentioned embodiment.

[0266] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0267] The computer-readable storage medium may be included in the multi-integrated mango maturity detection device; or may exist independently without being assembled into the multi-integrated mango maturity detection device.

[0268] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the multi-integrated mango maturity detection device, the multi-integrated mango maturity detection device:

[0269] Acquire a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor;

[0270] Processing the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix;

[0271] Dividing the second gas concentration data matrix into a plurality of third arrays by columns, wherein each gas sensor corresponds to one third array;

[0272] Calculate the correlation coefficient and the coefficient of variation of each of the third arrays, and delete the third arrays whose correlation coefficients do not belong to the preset correlation coefficient interval and whose coefficients of variation do not belong to the preset coefficient of variation interval;

[0273] fusing the remaining third arrays to obtain gas-sensing data;

[0274] Obtaining mango images at different times, screening an optimal combination of a model and an optimizer, and obtaining image data based on the mango images and the combination;

[0275] Inputting the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity;

[0276] Acquiring physical properties of mangoes, and obtaining a second degree of maturity based on the physical properties;

[0277] Based on the second maturity, the first maturity is verified.

[0278] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0279] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0280] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0281] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned multi-integrated mango maturity detection method, and can solve the technical problem of multi-integrated mango maturity detection. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the multi-integrated mango maturity detection method provided in the above-mentioned embodiment, and will not be repeated here.

[0282] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned multi-integrated mango maturity detection method when executed by a processor.

[0283] The computer program product provided by the present application can solve the technical problem of multi-integrated mango maturity detection. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the multi-integrated mango maturity detection method provided by the above embodiment, which will not be repeated here.

[0284] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A multi-integrated mango maturity detection method, characterized in that: The method is applied to a multi-integrated mango maturity detection device, which at least comprises: an arc conveyor belt, a chain plate with grooves, a collection cover, a high-speed camera, a flexible push rod, a cylindrical air chamber with a plurality of gas sensors, and a needle probe, and the method comprises: Acquire a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor; Processing the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix; Dividing the second gas concentration data matrix into a plurality of third arrays by columns, wherein each gas sensor corresponds to one third array; Calculate the correlation coefficient and the coefficient of variation of each of the third arrays, and delete the third arrays whose correlation coefficients do not belong to the preset correlation coefficient interval and whose coefficients of variation do not belong to the preset coefficient of variation interval; fusing the remaining third arrays to obtain gas-sensing data; Obtaining mango images at different times, screening an optimal combination of a model and an optimizer, and obtaining image data based on the mango images and the combination; Inputting the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity; Acquiring physical properties of mangoes, and obtaining a second degree of maturity based on the physical properties; Based on the second maturity, the first maturity is verified.

2. The method according to claim 1, characterized in that The obtaining of the first gas concentration matrix comprises: collecting voltage changes by using a plurality of gas sensors, wherein the voltage changes are caused by concentration changes of components contained in the mango gas; Convert each of the voltage changes into a digital signal to obtain a plurality of first arrays; Taking an average value of a fixed number of consecutive elements in each of the first arrays, and performing standardization processing on the average value to obtain a plurality of second arrays; The plurality of second arrays are used as a first gas concentration data matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the second arrays.

3. The method according to claim 1, characterized in that The step of processing the first gas concentration matrix based on the optimal data analysis strategy to obtain a second gas concentration data matrix includes: Get multiple data analysis strategies; Using the data analysis strategy to process the first gas concentration data matrix respectively, and determine a fixed number of principal components of the gas; Calculating the variance contribution rate of each of the principal components to obtain the sum of the variance contribution rates; determining an optimal data analysis strategy based on the sum; Processing the first gas concentration data matrix according to the optimal data analysis strategy to divide the maturity standard values, wherein the maturity includes unripe, partially ripe, and ripe; The first gas concentration data matrix processed by the optimal data analysis strategy is used as the second gas concentration data matrix.

4. The method according to claim 1, characterized in that The step of calculating the correlation coefficient and the coefficient of variation of each of the third arrays, and deleting the third arrays whose correlation coefficient does not belong to a preset correlation coefficient interval and whose coefficient of variation does not belong to a preset coefficient of variation interval, comprises: Calculating a plurality of first average values ​​of the elements of each of the third arrays and a second average value of the maturity standard value; Obtain a plurality of correlation coefficients according to each of the third arrays, each of the first average values, and the second average values; If the correlation coefficient does not belong to the preset correlation coefficient interval, removing the third array corresponding to the correlation coefficient; Calculating a plurality of sample standard deviations of each of the third arrays; Obtaining a plurality of coefficients of variation according to the corresponding first mean values ​​and the sample standard deviations; If the coefficient of variation does not belong to the preset coefficient of variation interval, the third array corresponding to the coefficient of variation is removed.

5. The method according to claim 1, characterized in that The step of acquiring mango images of different periods, selecting an optimal combination of a model and an optimizer, and obtaining image data based on the mango images and the combination includes: Get images of mangoes at different times; Preprocessing the mango image and annotating the mango image to obtain a data set; Dividing the data set into a training set and a validation set; Acquire multiple models and multiple optimizers to obtain different combinations of the models and the optimizers; Training each of the combinations according to the training set and the validation set to obtain a plurality of training results, wherein the training results correspond one to one to the combination of the model and the optimizer; Obtaining the accuracy and loss value of the corresponding combination according to each of the training results; Determining the optimal combination of the optimizer and the model according to the accuracy and loss values; The model optimized by the optimizer of the optimal combination is used as the final model; The data set is used as input of the final model to obtain image data.

6. The method according to claim 1, characterized in that The step of inputting the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity includes: Extracting features of the gas-sensitive data; Combining the extracted features to obtain a first feature vector; Counting the number of data corresponding to each color in the image data to obtain multiple color features; extracting geometric features of the image data; Obtaining a second feature vector according to the color feature and the geometric feature; Concatenate the first eigenvector and the second eigenvector to obtain a third eigenvector; The third eigenvector is used as an input of a preset mango maturity comprehensive detection model to obtain a first maturity.

7. The method according to claim 1, characterized in that The step of obtaining the physical properties of the mango and obtaining the second maturity based on the physical properties includes: Obtaining a mango region sample, wherein the region includes the head, middle, and tail of the mango; Obtaining physical properties based on the mango region sample, wherein the physical properties include strength, hardness and toughness of the mango peel; Performing correlation analysis based on the physical properties to obtain the physical property correlation coefficient; Setting the physical property weight; Based on the weight and the correlation coefficient, a weighted score is calculated, and the second maturity is obtained according to the weighted score. The calculation formula is as follows: Weighted score = hardness correlation coefficient × W1 + strength correlation coefficient × W2 + toughness correlation coefficient × W3 Among them, the physical property correlation coefficient includes a hardness correlation coefficient, a strength correlation coefficient and a toughness correlation coefficient, W1 represents the hardness weight, W2 represents the strength weight, W3 represents the toughness weight, and satisfies W1+W2+W3=1.

8. A multi-integrated mango maturity detection device, characterized in that: The device comprises: An acquisition module, acquiring a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor; A preprocessing module, configured to process the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix; and divide the second gas concentration data matrix into a plurality of third arrays according to columns, wherein each gas sensor corresponds to one third array; A screening module, used for calculating the correlation coefficient and the coefficient of variation of each of the third arrays, and deleting the third arrays whose correlation coefficient does not belong to a preset correlation coefficient interval and whose coefficient of variation does not belong to a preset coefficient of variation interval; A fusion module, used for fusing the remaining third arrays to obtain gas sensing data; An image recognition module, used to obtain mango images at different times, select an optimal combination of a model and an optimizer, and obtain image data based on the mango images and the combination; The output module inputs the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity; obtains the physical properties of the mango, and obtains a second maturity based on the physical properties; and verifies the first maturity based on the second maturity.

9. A multi-integrated mango maturity detection device, characterized in that: The equipment includes: a host computer, an arc conveyor belt, a chain plate with grooves, a collection cover, a high-speed camera, a flexible push rod, a cylindrical air chamber with a plurality of gas sensors and a needle probe. A temperature and humidity controller is arranged in the collection cover, and a dynamic seal is used at the joint between the collection cover and the groove: The host computer is used to obtain a first gas concentration matrix, wherein the columns of the first gas concentration data matrix correspond one-to-one to the concentration data of a component contained in mango gas collected by each gas sensor; The host computer is further used to process the first gas concentration matrix based on an optimal data analysis strategy to obtain a second gas concentration data matrix; The host computer is further used to divide the second gas concentration data matrix into a plurality of third arrays according to the columns, wherein each gas sensor corresponds to one third array; The host computer is further used to calculate the correlation coefficient and the coefficient of variation of each of the third arrays, and delete the third arrays whose correlation coefficient does not belong to the preset correlation coefficient interval and whose coefficient of variation does not belong to the preset coefficient of variation interval; The host computer is also used to fuse the remaining third arrays to obtain gas-sensitive data; The host computer is also used to obtain mango images at different times, select the best combination of the model and the optimizer, and obtain image data based on the mango images and the combination; The host computer is further used to input the gas-sensing data and the image data into a preset mango maturity comprehensive detection model to obtain a first maturity; The host computer is further used to obtain physical properties of mangoes, and obtain a second maturity based on the physical properties; The host computer is further used to verify the first maturity based on the second maturity.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the multi-integrated mango maturity detection method according to any one of claims 1 to 7 are implemented.

Images