Honey pomelo withered water disease detection method and device, electronic equipment and storage medium

Through the extraction of spectral data and feature wavelength screening, combined with partial least squares discriminant analysis, the automated detection of honey pomelo dry water disease is realized, solving the problems of high manual screening cost and high misjudgment rate, and improving the detection efficiency and accuracy.

CN118914130APending Publication Date: 2024-11-08INTELLIGENT EQUIPMENT RESEARCH CENTER BEIJING ACADEMY OF AGRICULTURE AND FORESTRY SCIENCES
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410973992.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

In the prior art, the qualitative judgment of honey pomelo dry water mainly relies on manual screening, with high cost and high misjudgment rate, and lack of effective automated detection methods.

Method used

By obtaining the spectral data of honey pomelo, extracting the characteristic wavelengths, and inputting them into the qualitative discriminant model of honey pomelo for discrimination, the model is constructed using partial least squares discriminant analysis method, combining convolutional smoothing method and stability competition adaptive heavy Weighted sampling method performs preprocessing and feature wavelength screening to achieve automated detection.

Benefits of technology

You can accurately identify dry water diseases without destroying honey pomelo, reduce manpower and material costs, and improve detection efficiency and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118914130A_ABST
    Figure CN118914130A_ABST
Patent Text Reader

Abstract

The invention provides a honey pomelo withered disease detection method and device, electronic equipment and a storage medium, and relates to the technical field of honey pomelo detection. The honey pomelo withered disease detection method comprises the following steps: acquiring spectral data of honey pomelo; extracting a characteristic wavelength from the spectral data; and inputting the characteristic wavelength into a honey pomelo dry water qualitative discrimination model to obtain a discrimination result output by the honey pomelo dry water qualitative discrimination model. According to the honey pomelo dry-water disease detection method and device, the electronic equipment and the storage medium, the characteristic wavelength is extracted from the spectrum data of the honey pomelo, whether the honey pomelo has the dry-water disease or not can be judged by inputting the characteristic wavelength into the honey pomelo dry-water disease qualitative judgment model, the honey pomelo does not need to be damaged in the whole detection process, and compared with manual screening, manpower and material resources can be effectively reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of honey pomelo detection, and in particular to a honey pomelo dry-weed disease detection method, device, electronic equipment and storage medium. Background Art

[0002] Pomelo is a traditional fruit in China. It is a citrus fruit of the Rutaceae family. Most of the fruits are spherical and pear-shaped, with high nutritional value. Waterlogging is a common physiological disease of citrus fruits after they mature and during the post-harvest storage period. It is manifested by irregular swelling, fibrosis, corkification, water loss, hardening, and drying of the pulp juice cells, lightening of color, decreased pulp juice yield, and dry and tasteless taste, which seriously affects the commercial quality of the fruit. The granulation of juice cells in citrus fruits is usually related to the formation of secondary walls and thickening of cell walls in juice vesicles. Harvest time also has an important impact on the internal quality of the fruit. Late harvest time may increase the degree of juice cell granulation in honey pomelo. Since the causes of waterlogging are relatively complex, the current research on honey pomelo waterlogging is relatively limited. The traditional qualitative judgment of honey pomelo waterlogging is mainly manual screening, but it requires rich experience and has a large misjudgment rate. Summary of the invention

[0003] The invention provides a honey pomelo dry water disease detection method, device, electronic equipment and storage medium, which are used to solve the defect that the conventional honey pomelo dry water disease qualitative identification in the prior art relies on manual screening and has high cost.

[0004] The present invention provides a method for detecting honey pomelo dry water disease, comprising:

[0005] Get the spectrum data of honey pomelo;

[0006] extracting characteristic wavelengths from the spectral data;

[0007] The characteristic wavelength is input into a pomelo low water qualitative discrimination model to obtain a discrimination result output by the pomelo low water qualitative discrimination model.

[0008] According to a method for detecting dry pomelo disease provided by the present invention, obtaining spectral data of pomelo specifically comprises:

[0009] Acquire multiple sets of original spectral data when the honey pomelo is rotated to different positions;

[0010] The average value of the multiple groups of original spectral data is used as the spectral data of the honey pomelo.

[0011] According to a method for detecting dry pomelo disease provided by the present invention, extracting characteristic wavelengths from the spectral data specifically includes:

[0012] The spectral data is preprocessed, and the characteristic wavelength is extracted from the preprocessed spectral data.

[0013] According to a method for detecting pomelo dry-swell disease provided by the present invention, the spectral data is preprocessed by using a convolution smoothing method.

[0014] According to a method for detecting dry pomelo disease provided by the present invention, the step of extracting characteristic wavelengths from the spectral data specifically comprises:

[0015] A stability competitive adaptive reweighted sampling method is used to extract characteristic wavelengths from the spectral data.

[0016] According to a method for detecting dry pomelo disease provided by the present invention, the qualitative discriminant model for dry pomelo disease is constructed based on dry pomelo samples, healthy pomelo samples, characteristic wavelength samples of the dry pomelo samples and characteristic wavelength samples of the healthy pomelo samples, using partial least squares discriminant analysis.

[0017] According to a method for detecting dry pomelo disease provided by the present invention, before constructing the dry pomelo qualitative discriminant model by using partial least squares discriminant analysis, abnormal samples in the dry pomelo samples and the healthy pomelo samples are firstly eliminated.

[0018] The present invention also provides a honey pomelo dry water disease detection device, comprising:

[0019] An acquisition unit, used for acquiring spectrum data of honey pomelo;

[0020] An extraction unit, used for extracting characteristic wavelengths from the spectral data;

[0021] The discrimination unit is used to input the characteristic wavelength into the honey pomelo low water qualitative discrimination model to obtain the discrimination result output by the honey pomelo low water qualitative discrimination model.

[0022] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for detecting waterlogging disease of pomelo as described in any one of the above are implemented.

[0023] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method for detecting waterlogging disease of pomelo as described in any one of the above are implemented.

[0024] The honey pomelo drought disease detection method, device, electronic device and storage medium provided by the present invention extract characteristic wavelengths from the spectral data of honey pomelo, and input the characteristic wavelengths into the honey pomelo drought qualitative discrimination model to determine whether the honey pomelo has drought disease. The entire detection process does not require destroying the honey pomelo, and can effectively reduce manpower and material resources compared to manual screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 This is one of the flow diagrams of the method for detecting honey pomelo dry water disease provided by the present invention;

[0027] Figure 2 This is the second flow chart of the method for detecting honey pomelo dry water disease provided by the present invention;

[0028] Figure 3 This is the third flow chart of the method for detecting honey pomelo dry water disease provided by the present invention;

[0029] Figure 4 is the original spectral data graph of each sample in the sample set of the present invention;

[0030] Figure 5 It is the original spectrum data diagram after the abnormal samples are removed from the sample set of the present invention;

[0031] Figure 6 This is a spectrum data diagram of a pomelo sample pre-treated by SGS in the present invention;

[0032] Figure 7 This is a spectrum data diagram of a pomelo sample pre-processed by SGS+SNV in the present invention;

[0033] Figure 8 This is a spectrum data diagram of a honey pomelo sample pretreated by SGS+MSC in the present invention;

[0034] Fig. 9 It is a root mean square error diagram obtained by cross-validation after SPA characteristic wavelength screening of full-band sample spectrum data in the present invention;

[0035] Fig.10 It is a characteristic wavelength distribution position diagram obtained after the SPA characteristic wavelength screening of the present invention;

[0036] Fig.11 It is a characteristic variable screening process diagram when the present invention uses the SCARS method to screen the sample spectrum data pre-processed by SGS;

[0037] Fig.12 It is the curve of RMSECV changing with the number of iterations in the process of screening characteristic wavelengths using the BOSS algorithm in the present invention;

[0038] Fig.13 This is a characteristic wavelength diagram screened out by the BOSS algorithm in the present invention.

[0039] Fig.14 It is a structural diagram of a honey pomelo dry water disease detection device provided by the present invention;

[0040] Fig.15 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Combine the following Figure 1-Figure 13 The invention describes a method for detecting waterlogging disease of pomelo.

[0043] The present invention provides a method for detecting honey pomelo dry water disease, such as Figure 1 As shown, it includes the following steps:

[0044] Step 100, obtaining spectrum data of honey pomelo;

[0045] Step 200, extracting characteristic wavelengths from spectral data;

[0046] Step 300, inputting the characteristic wavelength into a pomelo low water qualitative discrimination model to obtain a discrimination result output by the pomelo low water qualitative discrimination model.

[0047] The spectral data of honey pomelo can be directly detected by a spectral measurement system, or can be obtained by analyzing and processing the spectral information obtained by the detection. Specifically, the spectral measurement system includes a light source, a spectrometer, a transmission system, a photoelectric sensor and an industrial control computer. Among them, the light source is a 150W, 15V halogen lamp. The spectral range of the spectrometer is 615-1044nm, and the spectral resolution is 5nm. The transmission system is a conveyor belt or other conveying device. A blackout curtain is set on the transmission channel of the transmission system to form a closed detection space. The light source and the spectrometer are located in the detection space and are respectively set on both sides of the sample. The spectrometer collects the full transmission spectrum signal of honey pomelo to eliminate the interference of fruit mirror reflection. The photoelectric sensor is set in the detection space to detect whether the honey pomelo reaches the detection position. The industrial control computer is connected to the light source, the transmission system, and the photoelectric sensor respectively to control the switch of the light source and the operation of the transmission system according to the detection information of the photoelectric sensor. When the honey pomelo is transported to the detection position by the transmission system, the photoelectric sensor generates a trigger signal, and the industrial control computer controls the transmission system to stop accordingly, and after the spectrometer detection is completed, the transmission system is started again. In order to improve the shading effect of the detection space, a double-layer shading curtain can be used to effectively avoid interference from external ambient light. Among them, the spectrometer can be turned on with the help of hardware structure control. The industrial control computer calculates the time when the honey pomelo blocks the photoelectric sensor and determines the working time of the spectrometer, thereby avoiding the spectrometer from collecting useless information and improving the effectiveness of the spectrum.

[0048] Pomelo is large in size and thick in skin, with a thick sponge layer, so it is difficult to obtain the transmission spectrum that carries the information of the central part of the fruit pulp. In order to obtain the spectral information of pomelo flesh more comprehensively, the general spectrum acquisition method uses a longer integration time and a stronger light source. The above-mentioned spectral measurement system has an integration time of 10ms and a 150W halogen lamp as the light source. The energy of the transmitted light is low. Compared with the long integration time acquisition mode, it can continuously collect the transmission spectra of different positions of the pomelo in a very short time interval.

[0049] During the detection process, in order to ensure the detection effect, the spectrometer is started first, and the first preheating time is used to make the equipment reach a stable working state. Among them, the first time length is 10min, 15min, 20min or any time length between 10-20min. During the detection, the honey pomelo is placed on the turntable and then placed on the conveyor system, and the conveyor system moves at a constant speed. Among them, the turntable is black. When the honey pomelo is transported to the detection position, the photoelectric sensor receives the trigger signal, and the industrial control computer controls the spectrometer to continuously obtain the transmission spectrum data of the honey pomelo according to the trigger signal and the moving speed detected by the encoder.

[0050] The placement of the pomelo affects the final identification result. To ensure a more accurate identification result, the pomelo is placed on the turntable with its pointed end facing upwards.

[0051] Taking the near-infrared spectral region as an example, the overlapping absorption peaks correspond to the CH, OH and NH chemical bonds of water and carbohydrates in the fruit. In the short-wave near-infrared spectral region (900-1100nm), the overlapping absorption peaks appear around 928nm-950nm, and the overlapping absorption peaks are the absorption bands of water. The juice cysts of the dry honey pomelo have lost more water, so the absorption peaks of the dry honey pomelo are lower than those of the healthy honey pomelo. The health status of honey pomelo can be determined by analyzing part of the spectral region in the spectral data, so before entering the honey pomelo dry water qualitative discrimination model, the characteristic wavelengths are first extracted from the spectral data.

[0052] The pomelo dry water qualitative discrimination model can distinguish whether the pomelo is dry water pomelo or healthy pomelo. After inputting the characteristic wavelength into the pomelo dry water qualitative discrimination model, if it is determined to be a healthy pomelo, the output discrimination result is a healthy pomelo; if it is determined to be a dry water pomelo, the output discrimination result is a dry water pomelo. In this way, it is possible to determine whether the pomelo is dry water without destroying the overall structure of the pomelo, which is convenient to operate.

[0053] The honey pomelo drought disease detection method provided in the embodiment of the present invention extracts characteristic wavelengths from the spectral data of the honey pomelo, and inputs the characteristic wavelengths into the honey pomelo drought qualitative discrimination model to determine whether the honey pomelo has drought disease. The entire detection process does not require destroying the honey pomelo, and can effectively reduce manpower and material resources compared to manual screening.

[0054] In one embodiment, if Figure 2 As shown, step 100, obtaining the spectrum data of honey pomelo specifically includes:

[0055] Step 110, obtaining multiple groups of original spectral data when the honey pomelo is rotated to different positions;

[0056] Step 120: taking the average value of the multiple groups of original spectral data as the spectral data of honey pomelo.

[0057] In order to avoid insufficient information in a single set of original spectral data, multiple sets of original spectral data are collected. During the collection process of honey pomelo, the spectral information is easily interfered by the sensor and other external factors, resulting in saturation of the information at the beginning and end of the spectral data. Therefore, before processing and analyzing the spectral data, the first four and last four spectra of each set of original spectral data can be removed, and then the average spectrum is calculated as the spectral data of honey pomelo.

[0058] In a specific embodiment, if Figure 3 As shown, step 200, extracting characteristic wavelengths from spectral data specifically includes:

[0059] Step 210, preprocessing the spectral data;

[0060] Step 220: extracting characteristic wavelengths from the preprocessed spectral data.

[0061] In addition to the target information, the spectral information collected by the spectrometer also contains a lot of non-target information. The use of appropriate preprocessing methods can not only filter the noise in the spectrum and retain important information, but also improve the discrimination accuracy of the honey pomelo dry water qualitative discrimination model to a certain extent. Optionally, one or more of Savitzky-Golay smoothing (SGS), multiplicative scatter correction (MSC), and standard normal variate (SNV) are used for preprocessing. After comparison, the spectral data preprocessed by SGS is input into the honey pomelo dry water qualitative discrimination model to obtain a more accurate result.

[0062] The characteristic wavelengths can be extracted from the spectral data by using any of the following methods: stability competitive adaptive reweighted sampling (SCARS), successive projections algorithm (SPA), or bootstrapping soft shrinkage (BOSS).

[0063] Specifically, step 200, extracting characteristic wavelengths from spectral data specifically includes:

[0064] The characteristic wavelengths are extracted from spectral data using a stability competitive adaptive reweighted sampling method.

[0065] The spectral data collected by the spectrometer often contains more wavelength variables, and there are many collinearities and wavelength variables that are irrelevant to the object to be measured. In order to eliminate these irrelevant variables and improve the detection speed, a characteristic wavelength screening method is needed.

[0066] In one embodiment, SPA is used to screen characteristic wavelengths. By projecting and mapping the spectral data, a few wavelengths in the spectral data are selected, which can effectively eliminate the collinearity effect between many wavelength variables and summarize as much spectral information as possible.

[0067] In another embodiment, SCARS is used to screen characteristic wavelengths. SCARS is a method for screening characteristic variables of near-infrared spectral data based on competitive adaptive reweighted sampling (CARS), and continues the variable screening process of the CARS method. The SCARS algorithm uses the stability of the variable as a measurement indicator. The larger the stability value, the greater the possibility of the variable being selected. The entire calculation includes N cycles, in which the (ii≤N)th cycle first calculates the stability of each wavelength variable, and then uses mandatory variable screening and adaptive reweighted sampling (ARS) to screen out variables with greater stability to form a variable subset.

[0068] In another embodiment, the BOSS variable selection method is used to screen the preprocessed spectral data to improve the prediction accuracy of the spectral data. BOSS is derived from the ideas of weighted bootstrap sampling (WBS) and model population analysis (MPA), and the weights of the variables are determined according to the absolute values ​​of the regression coefficients. WBS is applied to generate sub-models according to the weights, and MPA is used to analyze the sub-models to update the weights of the variables. The optimization process follows the soft shrinkage rule, that is, less important variables are not directly eliminated, but smaller weights are assigned. The algorithm iteratively runs and terminates until the number of variables reaches 1. The optimal variable set with the smallest cross-validation root mean square error is selected.

[0069] Among them, the method of using SCARS to screen characteristic wavelengths is better than the other two methods.

[0070] Specifically, the qualitative discrimination model of honey pomelo drought is constructed based on the dry honey pomelo samples, healthy honey pomelo samples, characteristic wavelength samples of dry honey pomelo samples and characteristic wavelength samples of healthy honey pomelo samples by partial least squares discriminant analysis. In addition, the qualitative discrimination model of honey pomelo drought can also be constructed by least squares support vector machine and random forest algorithm.

[0071] Before using the partial least squares discriminant analysis method to construct the qualitative discriminant model of honey pomelo drought, the abnormal samples in the dry honey pomelo samples and healthy honey pomelo samples were first eliminated.

[0072] For example, a sample set was constructed with 157 honey pomeloes, including 112 healthy honey pomeloes and 45 dry honey pomeloes. Abnormal sample data will appear during the spectral data collection process. The abnormal sample data may contain abnormal chemical components or the component content is relatively extreme, which is significantly different from other samples. If abnormal samples are used to establish a qualitative discrimination model for honey pomelo dry water, the performance of the qualitative discrimination model for honey pomelo dry water will be reduced. Therefore, before establishing a reliable model, it is necessary to remove outliers in the sample set.

[0073] Specifically, the Mahalanobis distance algorithm is used to remove abnormal samples in the sample set, and a total of 5 abnormal samples are proposed. Figure 4 As shown in the figure, the original spectral data of each sample in the sample set. Figure 5 As shown in Figure 1, this is the original spectral data graph after removing abnormal samples from the sample set.

[0074] After removing the abnormal samples, the remaining 152 samples were divided into the training set (110 samples) and the test set (42 samples) according to the ratio of about 3:1 between the number of samples in the training set and the number of samples in the prediction set. The sample set division results and labels are shown in Table 1.

[0075] Table 1 Distinction and sample number of different categories of training sets and prediction sets

[0076]

[0077] The spectral data of honey pomelo samples after SGS preprocessing are as follows: Figure 6 As shown in Figure 2, the spectral data of honey pomelo samples preprocessed by SGS+SNV is shown in Figure 2. Figure 7 As shown in Figure 2, the spectral data of the pomelo sample after SGS+MSC pretreatment is as follows: Figure 8 shown.

[0078] The SPA characteristic wavelengths were screened for the full-band sample spectral data, and the root mean square error obtained by cross-validation was as follows: Fig. 9 As shown, the selected characteristic wavelength distribution position is as follows Fig.10 shown.

[0079] The SCARS method was used to screen the characteristic variables of the sample spectral data preprocessed by SGS. The simulation times, sampling times and sampling rate were set to 1000, 50 and 0.8 respectively. The whole characteristic variable screening process was as follows: Fig.11 shown. Fig.11 The top curve in the figure shows the trend of the number of retained variables during the variable screening process. Fig.11 From the top curve, we can see that as the number of cycles increases, the number of selected wavelength variables gradually decreases, and the downward trend changes from fast to slow, reflecting the two processes of rough selection and fine selection of wavelength variables. Fig.11 The curve in the middle shows the changing trend of RMSECV value during variable screening. Fig.11It can be seen from the curve in the middle that the RMSECV value decreases slowly during the 1st to 20th cycles, indicating that the process gradually eliminates unimportant variables, and after 25 cycles, the RMSECV value gradually increases. The RMSECV value is the lowest at the 20th cycle, and the variable subset obtained after this cycle is the optimal variable subset. The number of variables corresponding to the 20th cycle is 121.

[0080] The BOSS algorithm is used to screen the characteristic wavelengths of the preprocessed spectral data. Fig.12 The figure shows the process of BOSS algorithm screening characteristic wavelengths. The number of wavelength variables and RMSECV value of the sub-model change with the increase of iteration number. Fig.12 In the figure, as the number of iterations increases, the RMSECV value of the sub-model decreases first and then increases. At the 10th iteration, RMSECV reaches the minimum value, indicating that the prediction effect of the model built by the characteristic band at this time is the best. Therefore, the characteristic variables at the 10th iteration are selected as the best modeling variables, with a total of 84 characteristic bands, accounting for about 4.2% of the total bands. Fig.13 The importance of different characteristic wavelengths is shown, and the more important bands are concentrated around 780-890nm and 1000-1100nm.

[0081] The study used PLS-DA, LS-SVM, and RF models to discriminate honey pomelo samples. Due to the different effects of different preprocessing methods on spectral information and spectral quality, the original spectrum and the spectral data processed by SGS, SGS+SNV, and SGS+MSC were used as input variables to establish full-band spectral PLS-DA, LS-SVM, and RF discrimination models. Table 2 summarizes the detection results of the classification models established after different preprocessing methods. From the results, it can be seen that different preprocessing methods have a certain impact on the accuracy of the discrimination model. By comparing the prediction results of the models built by different preprocessing methods, it can be found that the detection accuracy of the three models after SGS preprocessing has improved, among which the classification accuracy of the prediction set of PLS-DA increased from 87.18% to 89.74%, the classification accuracy of the prediction set of LS-SVM was 94.87%, and the classification accuracy of the prediction set of RF increased from 92.31% to 94.87%. Therefore, SGS was selected to preprocess all spectral data and used in subsequent studies.

[0082] Table 2 Model classification results based on different preprocessing methods

[0083]

[0084]

[0085] Among them, partial least squares discriminant analysis (PLS-DA) is a multivariate statistical data analysis method widely used in spectral data processing. Its purpose is to find some linear combinations of components with the highest degree of correlation in the independent variables and the dependent variables, so as to better explain the variation information of the dependent variables. PLS-DA is similar to the principle of PLS. It decomposes the spectral matrix and the category label matrix (equivalent to the concentration variable matrix in the PLS model) at the same time, which helps the role of category information in the spectral decomposition process to extract the spectral information most relevant to the sample category, that is, the algorithm can extract the differences between spectra of different categories. If the PLS prediction value of an unknown sample is between ±0.5 of the sample category label value, it means that the sample is correctly discriminated.

[0086] Before modeling, the SCARS, SPA, and BOSS algorithms were used to screen the wavelengths of the full-band spectrum, and then the PLS-DA model was established and the modeling effects were compared. Table 3 shows the classification accuracy comparison of the PLS-DA model under different characteristic wavelength screening methods. After comparison, the modeling effect after using the SCARS algorithm to screen the characteristic wavelengths was the best. The number of characteristic wavelengths was 121, the number of principal components was 7, the classification accuracy of the training set reached 100%, and the classification accuracy of the prediction set reached 97.44%.

[0087] Table 3 Comparison of PLS-DA models based on different characteristic wavelength screening methods

[0088]

[0089]

[0090] Least Squares Support Vector Machine (LS-SVM) is a method used to solve problems such as pattern classification and function estimation. The Least Squares Support Vector Machine method uses the least squares linear system as the loss function, replacing the quadratic programming method used by the traditional support vector machine. In addition, since LS-SVM uses the least squares method, its operation speed is significantly faster than other versions of the support vector machine.

[0091] LS-SVM generally uses two kernel functions, radial basis kernel function (RBF-Kernel) and linear kernel function (Lin-Kernel) to establish the model. Specifically, the radial basis kernel function (RBF-Kernel) is used to establish the model, and γ and σ2 are important parameters for LS-SVM model evaluation.

[0092] Table 4 Comparison of classification accuracy of LS-SVM model under different characteristic wavelength screening methods. The results show that the BOSS wavelength selection method has a better effect on the LS-SVM model. Although the model has been simplified, the classification accuracy of the prediction set has not been significantly improved.

[0093] Table 4 Comparison of LS-SVM models based on different characteristic wavelength screening methods

[0094]

[0095] Random Forest Algorithm (RF) is a machine learning algorithm based on the combination of decision tree algorithm and bagging algorithm. Ensemble learning is its core content. Its essence is to combine several decision trees to form a classifier, and then make the classification performance of the strong classifier model more stable by combining multiple random decision trees. Finally, the classification results of all decision trees are combined to determine the output value of the model.

[0096] Table 5 compares the classification accuracy of the RF classification model using different characteristic wavelength screening methods. The results show that for the RF model, the modeling effect after using the BOSS wavelength screening method is better than the other two methods. The classification accuracy of the training set is 100% and the classification accuracy of the prediction set is 94.87%.

[0097] Table 5 Comparison of RF models based on different characteristic wavelength screening methods

[0098]

[0099] A comprehensive comparison of the three modeling methods shows that the overall modeling effect of PLS-DA is better than that of LS-SVM and RF. The classification accuracy of the PLS-DA model prediction set after SCARS characteristic wavelength screening reaches 97.44%. Therefore, the SCARS-PLS-DA model is more suitable for distinguishing the internal drought of pomelo.

[0100] The present invention also provides a device for detecting waterlogging disease of honey pomelo, such as Fig.14 As shown, it includes:

[0101] An acquisition unit 510 is used to acquire spectrum data of honey pomelo;

[0102] An extraction unit 520, for extracting characteristic wavelengths from spectral data;

[0103] The discrimination unit 530 is used to input the characteristic wavelength into the honey pomelo low water qualitative discrimination model to obtain the discrimination result output by the honey pomelo low water qualitative discrimination model.

[0104] Fig.15 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Fig.15As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic commands in the memory 630 to execute the following method: obtaining the spectrum data of honey pomelo; extracting the characteristic wavelength from the spectrum data; inputting the characteristic wavelength into the honey pomelo dry water qualitative discrimination model to obtain the discrimination result output by the honey pomelo dry water qualitative discrimination model.

[0105] In addition, the logic commands in the above-mentioned memory 630 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several commands to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0106] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the honey pomelo dry-soil disease detection method provided by the above-mentioned methods, and the method includes: obtaining spectral data of honey pomelo; extracting characteristic wavelengths from the spectral data; inputting the characteristic wavelengths into a honey pomelo dry-soil qualitative discrimination model to obtain a discrimination result output by the honey pomelo dry-soil qualitative discrimination model.

[0107] On the other hand, an embodiment of the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it is implemented to execute the methods provided in the above embodiments, for example, including: obtaining spectral data of honey pomelo; extracting characteristic wavelengths from the spectral data; inputting the characteristic wavelengths into a qualitative discrimination model for honey pomelo dry water, and obtaining a discrimination result output by the qualitative discrimination model for honey pomelo dry water.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting dry water disease of pomelo, characterized in that: include: Get the spectrum data of honey pomelo; extracting characteristic wavelengths from the spectral data; The characteristic wavelength is input into a pomelo low water qualitative discrimination model to obtain a discrimination result output by the pomelo low water qualitative discrimination model.

2. The method for detecting dry water disease of honey pomelo according to claim 1, characterized in that: The obtaining of the spectral data of honey pomelo specifically includes: Acquire multiple sets of original spectral data when the honey pomelo is rotated to different positions; The average value of the multiple groups of original spectral data is used as the spectral data of the honey pomelo.

3. The method for detecting dry water disease of honey pomelo according to claim 1, characterized in that: The extracting characteristic wavelength from the spectral data specifically includes: The spectral data is preprocessed, and the characteristic wavelength is extracted from the preprocessed spectral data.

4. The method for detecting dry water disease of honey pomelo according to claim 3, characterized in that: The spectral data were preprocessed using convolution smoothing method.

5. The method for detecting dry water disease of pomelo according to claim 1 or 3, characterized in that: The step of extracting characteristic wavelengths from the spectral data specifically includes: A stability competitive adaptive reweighted sampling method is used to extract characteristic wavelengths from the spectral data.

6. The method for detecting dry water disease of pomelo according to claim 1, characterized in that: The pomelo dry water qualitative discrimination model is constructed based on dry pomelo samples, healthy pomelo samples, characteristic wavelength samples of the dry pomelo samples and characteristic wavelength samples of the healthy pomelo samples by using partial least squares discriminant analysis method.

7. The method for detecting dry water disease of honey pomelo according to claim 6, characterized in that: Before using the partial least squares discriminant analysis method to construct the honey pomelo dry water qualitative discriminant model, the abnormal samples in the dry honey pomelo samples and the healthy honey pomelo samples are first eliminated.

8. A device for detecting dry water disease of honey pomelo, characterized in that: include: An acquisition unit, used for acquiring spectrum data of honey pomelo; An extraction unit, used for extracting characteristic wavelengths from the spectral data; The discrimination unit is used to input the characteristic wavelength into the honey pomelo low water qualitative discrimination model to obtain the discrimination result output by the honey pomelo low water qualitative discrimination model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for detecting pomelo dry-stomach disease as described in any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting pomelo dry-stomach disease as described in any one of claims 1 to 7 are implemented.