Method and device for estimating leaf weight of tobacco plant based on canopy reflectance spectrum

Abstract

The application provides a tobacco plant leaf weight estimation method and device based on canopy reflectance spectrum, comprising: obtaining a target key spectrum of a to-be-measured tobacco plant; the target key spectrum is determined based on the variety, growth stage and canopy reflectance spectrum of the to-be-measured tobacco plant; inputting the target key spectrum into a tobacco plant leaf weight estimation model to obtain the leaf weight of the to-be-measured tobacco plant output by the tobacco plant leaf weight estimation model; and the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights. The tobacco plant leaf weight estimation method and device based on canopy reflectance spectrum provided by the application realizes tobacco plant leaf weight estimation through a small number of key spectral bands, reduces the complexity of the model, improves the prediction accuracy, and according to the non-contact characteristics of spectral technology, combined with different load platforms, can realize the measurement of the canopy reflectance spectrum of a large range of tobacco plants, so as to quickly obtain the leaf weight of a large range of tobacco plants.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing, and in particular to a tobacco plant leaf weight estimation method and device based on canopy reflectance spectrum. BACKGROUND

[0002] In the field of agriculture, leaves are the direct product of some crops, and leaf weight is the main part of above-ground biomass of vegetation, which is closely related to crop yield.

[0003] Tobacco plant canopy reflectance spectrum contains various information of tobacco plants, such as canopy structure, leaf pigment, and leaf water content, and can be used for tobacco plant leaf weight estimation. At present, the reflectance spectrum of the full spectrum is input into a multivariate regression, machine learning, or other method to establish a tobacco plant leaf weight estimation model, and then the weight of the tobacco plant leaf is predicted.

[0004] However, the above method has the defects of complex model and low precision. SUMMARY

[0005] The tobacco plant leaf weight estimation method and device based on canopy reflectance spectrum provided by the present application solve the defects of complex model and low precision in the prior art, reduce the complexity of the model, and improve the prediction accuracy. According to the non-contact characteristics of spectral technology, combined with different load platforms, the measurement of large-scale tobacco plant canopy reflectance spectrum can be realized, so that the large-scale tobacco plant leaf weight can be quickly obtained.

[0006] The present application provides a tobacco plant leaf weight estimation method based on canopy reflectance spectrum, comprising:

[0007] obtaining a target key spectrum of a tobacco plant to be measured; the target key spectrum is determined based on the variety, growth stage, and canopy reflectance spectrum of the tobacco plant to be measured;

[0008] inputting the target key spectrum into a tobacco plant leaf weight estimation model to obtain the leaf weight of the tobacco plant to be measured output by the tobacco plant leaf weight estimation model; the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights.

[0009] According to the tobacco plant leaf weight estimation method based on canopy reflectance spectrum provided by the present application, the target key spectrum of the tobacco plant to be measured is obtained, comprising:

[0010] determining the spectral range of a plurality of important spectral intervals for leaf weight estimation based on the canopy spectral response of the tobacco plant and the canopy reflectance spectrum of the tobacco plant to be measured;

[0011] combining the plurality of important spectral intervals based on the variety and growth stage of the tobacco plant to be measured to obtain a key spectral interval;

[0012] Band selection is performed in the canopy reflectance spectrum by using the spectral range of the important spectral interval and the key spectral interval to extract the target key spectrum.

[0013] According to the tobacco canopy spectral response and the canopy reflectance spectrum of the to-be-measured tobacco plant, a method for estimating the leaf weight of a tobacco plant based on a canopy reflectance spectrum is provided, and the spectral range of a plurality of important spectral intervals for estimating the leaf weight is determined, including:

[0014] According to the canopy spectral response of the tobacco plant, a plurality of important spectral intervals are determined, and the spectral range of the plurality of important spectral intervals in the canopy reflectance spectrum is determined.

[0015] According to the method for estimating the leaf weight of a tobacco plant based on a canopy reflectance spectrum provided by the application, before the target key spectrum is input into the tobacco leaf weight estimation model, the method further includes:

[0016] According to the variety and growth stage of each sample tobacco plant, the sample key spectrum and the sample leaf weight of a plurality of sample tobacco plants are obtained.

[0017] The initial model is trained by using the sample key spectrum and the sample leaf weight of the plurality of sample tobacco plants to obtain the tobacco leaf weight estimation model.

[0018] According to the method for estimating the leaf weight of a tobacco plant based on a canopy reflectance spectrum provided by the application, the initial model is trained by using the sample key spectrum and the sample leaf weight of the plurality of sample tobacco plants to obtain the tobacco leaf weight estimation model, including:

[0019] The combination of the sample key spectrum and the sample leaf weight of each sample tobacco plant is taken as a sample to obtain a plurality of samples.

[0020] Based on the plurality of samples, a training set and a validation set are constructed.

[0021] The training sample in the training set is input into the initial model to obtain a trained initial model.

[0022] The validation sample in the validation set is input into the trained initial model to obtain the predicted weight corresponding to the validation sample output by the trained initial model.

[0023] A loss value is calculated according to the predicted weight and the sample leaf weight of the training sample by using a preset loss function.

[0024] In a case that the loss value is less than a preset threshold value or a training number of the initial model reaches a preset number, it is determined that the tobacco leaf weight estimation model is obtained.

[0025] According to the tobacco leaf weight estimation method based on the canopy reflectance spectrum provided by the application, before the wave band selection in the canopy reflectance spectrum, the method further comprises the following steps of:

[0026] Obtaining initial canopy spectral data of the tobacco to be measured;

[0027] Performing noise reduction processing on the initial canopy spectral data to generate noise reduction spectral data;

[0028] Performing baseline correction on the noise reduction spectral data to generate the canopy reflectance spectrum of the tobacco to be measured.

[0029] The application further provides a server, wherein a processor is arranged in the server; and the server further comprises a memory and a program or instructions stored in the memory and executable on the processor, and the program or instructions are executed by the processor to perform the tobacco leaf weight estimation method based on the canopy reflectance spectrum according to any one of the above.

[0030] The application further provides a tobacco leaf weight estimation device based on a canopy reflectance spectrum, comprising:

[0031] An acquisition module is configured to acquire target key spectrum of the tobacco to be measured; the target key spectrum is determined based on a variety, a growth stage and a canopy reflectance spectrum of the tobacco to be measured;

[0032] An input module is configured to input the target key spectrum into a tobacco leaf weight estimation model to acquire leaf weight of the tobacco to be measured output by the tobacco leaf weight estimation model; the tobacco leaf weight estimation model is obtained by training a plurality of sample key spectrums with sample leaf weight.

[0033] The application further provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the program to implement the tobacco leaf weight estimation method based on the canopy reflectance spectrum according to any one of the above.

[0034] The application further provides a non-transitory computer readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the tobacco leaf weight estimation method based on the canopy reflectance spectrum according to any one of the above.

[0035] The application also provides a computer program product comprising a computer program which, when executed by a processor, implements any of the above-mentioned methods for estimating tobacco leaf weight based on canopy reflectance spectrum.

[0036] The application provides a method and device for estimating tobacco leaf weight based on canopy reflectance spectrum, which realizes tobacco leaf weight estimation through a small number of key spectral bands, reduces the complexity of the model, improves the prediction accuracy, and realizes measurement of canopy reflectance spectrum of a large range of tobacco plants according to the non-contact characteristics of spectral technology combined with different load platforms, so as to quickly obtain the leaf weight of a large range of tobacco plants. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0038] Figure 1 is one of the flowcharts of the method for estimating tobacco leaf weight based on canopy reflectance spectrum provided by the application;

[0039] Figure 2 is the canopy reflectance spectrum of 350-1000nm of tobacco plants measured by the ground object spectrometer in the field provided by the application;

[0040] Figure 3 is the scatter plot of the results of estimating tobacco leaf weight by using the canopy reflectance spectrum of 350-1000nm full spectral band provided by the application;

[0041] Figure 4 is the scatter plot of the results of estimating tobacco leaf weight by using the canopy reflectance spectrum of 500-780nm green light, red light and red edge region provided by the application;

[0042] Figure 5 is the scatter plot of the results of estimating tobacco leaf weight by using the canopy reflectance spectrum of 500-660nm green light and red light region provided by the application;

[0043] Figure 6 is the second flowchart of the method for estimating tobacco leaf weight based on canopy reflectance spectrum provided by the application;

[0044] Figure 7 is the structural schematic diagram of the device for estimating tobacco leaf weight based on canopy reflectance spectrum provided by the application;

[0045] Figure 8Fig. 1 is a structural schematic diagram of an electronic device provided by the present application. DETAILED DESCRIPTION

[0046] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0047] Vegetation fixes solar energy through photosynthesis, and constitutes an important way of material and energy circulation of an ecological system.

[0048] Leaves are the main organs for vegetation to carry out photosynthesis, and are particularly important for the yield of tobacco leaves and tea leaves. Conventional methods for obtaining the weight of tobacco leaves mainly include two kinds. One is to collect tobacco leaf samples in the field, and obtain the weight of the tobacco leaves by weighing. The other is to measure the agronomic parameters of tobacco plants in the field, such as the length and width of the leaves, and then estimate the weight of the tobacco leaves by the quantitative relationship between the agronomic parameters and the weight of the leaves.

[0049] The method of directly sampling and weighing destroys the normal growth of tobacco plants, and is not suitable for large-scale investigation of the weight of tobacco leaves. In the method of determining the weight of leaves by agronomic parameters, the process of measuring in the field is time-consuming and laborious, and the efficiency is low, which is not suitable for large-scale investigation of the weight of tobacco leaves.

[0050] Spectral technology has the characteristics of rapidity and non-destructiveness. The canopy reflectance spectrum of tobacco plants contains various information of tobacco plants, such as canopy structure, leaf pigment and leaf water content, and can be used to estimate the weight of tobacco leaves.

[0051] In addition, hyperspectral remote sensing technology combines remote sensing technology and spectral technology, and can obtain canopy reflectance spectra of tobacco plants at different scales by virtue of the multi-platform advantage of remote sensing technology, which has a significant advantage in large-scale and periodic monitoring of the weight of tobacco leaves. The development of canopy reflectance spectrum-based estimation of the weight of tobacco leaves has broad application prospects.

[0052] The embodiments of the present application will be described below. Figures 1-8 The embodiments of the present application provide a method and device for estimating the weight of tobacco leaves based on canopy reflectance spectrum.

[0053] The execution subject of the method for estimating leaf weight of tobacco plant based on canopy reflectance spectrum provided in the embodiments of the present application can be an electronic device or software or a functional module or a functional entity capable of implementing the method for estimating leaf weight of tobacco plant based on canopy reflectance spectrum in the electronic device. The electronic device in the embodiments of the present application includes but is not limited to a server. It should be noted that the above execution subject does not constitute a limitation on the present application.

[0054] Figure 1 is one of the flowcharts of the method for estimating leaf weight of tobacco plant based on canopy reflectance spectrum provided in the present application, as shown in Figure 1 includes but is not limited to the following steps:

[0055] Firstly, in step S1, a target key spectrum of a tobacco plant to be measured is acquired; the target key spectrum is determined based on a variety, a growth stage and a canopy reflectance spectrum of the tobacco plant to be measured.

[0056] The canopy reflectance spectrum is related to the variety, the leaf weight and other properties of the tobacco plant. The leaf weight and the canopy reflectance spectrum response are different for different growth stages of the tobacco plant. According to the canopy reflectance spectrum response of the tobacco plant, the green light, the red light and the red edge region are determined as important spectral intervals for estimating the leaf weight. The target key spectrum in the important spectral intervals is determined in the canopy reflectance spectrum of the tobacco plant to be measured.

[0057] Optionally, before the band selection in the canopy reflectance spectrum, the method further includes:

[0058] acquiring initial canopy spectral data of the tobacco plant to be measured;

[0059] performing noise reduction processing on the initial canopy spectral data to generate denoised spectral data;

[0060] performing baseline correction on the denoised spectral data to generate the canopy reflectance spectrum of the tobacco plant to be measured.

[0061] Specifically, the noise filtering on the initial canopy spectral data can be implemented by using a wavelet transform-based adaptive threshold denoising algorithm, and the denoised spectral data is obtained after the noise reduction. The baseline correction algorithm based on asymmetric least squares can be used to remove the interference of fluorescence background on spectral analysis.

[0062] In addition, the spectral data can be subjected to standard normalization processing to eliminate the influence of noise and data dimension on spectral analysis.

[0063] The effect of the algorithm tends to converge to a better effect, which facilitates improving the accuracy and stability of modeling.

[0064] Optionally, the acquiring of the target key spectrum of the tobacco plant to be measured includes:

[0065] determining a spectral range of a plurality of important spectral intervals for leaf weight estimation based on a spectral response of a tobacco plant canopy and a canopy reflectance spectrum of the tobacco plant to be measured;

[0066] combining the plurality of important spectral intervals to obtain a key spectral interval based on a variety and a growth stage of the tobacco plant to be measured;

[0067] performing band selection in the canopy reflectance spectrum using the spectral range of the important spectral intervals and the key spectral interval to extract the target key spectrum.

[0068] The canopy reflectance spectrum of the tobacco plant can cover a visible near-infrared and short-wave infrared (VNIR-SWIR) interval, and the spectral range is 400-2500 nm.

[0069] The canopy reflectance spectrum of the tobacco plant in the VNIR-SWIR interval is related to the variety, leaf weight and other attributes of the tobacco plant, and the canopy reflectance spectral response of the leaf weight is different for different growth stages of the tobacco plant. According to the canopy reflectance spectral response of the tobacco plant, the green light, red light and red edge region are determined as important spectral intervals for estimating the leaf weight.

[0070] Figure 2 The present application provides a field object spectrometer for measuring the canopy reflectance spectrum of 350-1000 nm of the tobacco plant in situ in the field, as shown in Figure 2 The abscissa is the wavelength (Wavelength) in nanometers (nm), and the ordinate is the reflectance (Reflectance) in percentage (%). Since different spectral measurement instruments can collect different spectral bands, the spectral range of green light is mainly 500-600 nm, the spectral range of red light is mainly 600-680 nm, and the spectral range of red edge is mainly 680-780 nm. Therefore, the spectral range of each important spectral interval can be adjusted appropriately according to the wavelength setting of the used spectrometer to ensure that the collected canopy reflectance spectrum can contain the spectral band of the important spectral interval.

[0071] For example, the range of green light is 500-600 nm, and the green light range of a certain sensor is 540-570 nm. At this time, the band of 540-570 nm can be selected.

[0072] For another example, the spectral measurement instrument can use an ASD Field Spec 3 field spectrometer, the spectral range of which is 350-2500 nm, and the spectral range of green light, red light and red edge region is included, so the spectral range of green light is determined as 500-600 nm, and the spectral range of red light is determined as 600-660 nm;

[0073] The green light, red light and red edge spectral intervals are combined, and according to the type and growth stage of the tobacco plant, a combination with the best estimation result is selected as the key spectrum for estimating the leaf weight.

[0074] For example, K326 tobacco plants and Yunyan-87 tobacco plants are widely planted in a certain region. The K326 tobacco plant prefers to select the green light and red light interval for estimating the leaf weight, and then selects the green light and red edge spectral interval. The Yunyan-87 tobacco plant prefers to select the green light and red edge spectral interval for estimating the leaf weight, and then selects the green light and red light interval. The tobacco plant prefers to select the green light and red edge spectral interval for estimating the leaf weight at the topping stage, and prefers to select the red light and red edge spectral interval for estimating the leaf weight at the mature stage of the lower leaves.

[0075] The spectral combination of green light, red light and red edge includes three combinations of green light and red light, green light and red edge, and red light and red edge.

[0076] In combination with the spectral range and the spectral combination, the reflectance spectrum of the key interval is extracted from the canopy reflectance spectrum data of the to-be-measured tobacco plant measured by the spectral measurement instrument as the target key spectrum.

[0077] Specifically, according to the variety and growth stage of the to-be-measured tobacco plant, the combination of green light and red light is determined as the key spectral interval, and then the green light reflectance spectrum of 500-600 nm and the red light reflectance spectrum of 600-660 nm are extracted from the canopy reflectance spectrum measured by the ASD spectrometer as the target key spectrum for estimating the leaf weight.

[0078] According to the tobacco leaf weight estimation method based on the canopy reflectance spectrum provided by the application, the variety and growth stage of the tobacco plant are combined with the canopy reflectance spectrum collected by the collection instrument to obtain the target key spectrum with strong correlation with the leaf weight of the tobacco plant, and then provide a basis for accurate estimation of the canopy leaf weight.

[0079] Optionally, the canopy spectral response of the tobacco plant and the canopy reflectance spectrum of the to-be-measured tobacco plant determine the spectral range of a plurality of important spectral intervals for estimating the leaf weight, including:

[0080] According to the canopy spectral response of the tobacco plant, a plurality of important spectral intervals are determined;

[0081] The spectral range of the plurality of important spectral intervals is determined in the canopy reflectance spectrum.

[0082] Considering that there is high correlation between adjacent bands of hyperspectral data, and that the spectral resolution of hyperspectral data is high and the number of bands is large, in order to reduce the spectral data redundancy and model complexity, a band selection algorithm can be used to select bands in the canopy reflectance spectrum in the key spectral range.

[0083] For example, a genetic algorithm (GA) is selected for band selection, and a partial least squares regression (PLSR) is used for model construction. In the case of determining green light and red light as key bands, in the GA-PLSR modeling process, the bands of green light and red light are first input into the GA program for band selection, and the selected bands are input into the PLSR for modeling to construct an initial model.

[0084] Optionally, before inputting the target key spectrum into the tobacco plant leaf weight estimation model, the method further comprises:

[0085] According to the variety and growth stage of each sample tobacco plant, the sample key spectrum and the sample leaf weight of the plurality of sample tobacco plants are obtained;

[0086] The initial model is trained using the sample key spectrum and the sample leaf weight of the plurality of sample tobacco plants, to obtain the tobacco plant leaf weight estimation model.

[0087] Specifically, in the tobacco plant leaf weight estimation area, representative tobacco plants are selected as sample tobacco plants for canopy reflectance spectrum data collection and leaf weight measurement;

[0088] The canopy reflectance spectrum of the tobacco plants and the selected sample tobacco plants in the tobacco plant leaf weight estimation area is collected using a ground object spectrometer or a hyperspectral remote sensing camera.

[0089] The leaves of the selected sample tobacco plants are collected and weighed to obtain the fresh leaf weight as the sample leaf weight.

[0090] For example, the tobacco plants in a tobacco plantation in a certain place can be taken as the research object. The test field of a tobacco plant variety in the tobacco plantation has 4 fertilization levels, and the test field is divided into 4 levels according to the fertilization amount, and there are 12 plots in total. Three representative tobacco plants in each plot are selected as sample tobacco plants, and a total of 36 sample tobacco plants are determined.

[0091] The canopy reflectance spectrum of the tobacco plants is measured using an ASD Field Spec 3 ground object spectrometer to obtain the canopy reflectance spectrum of the 36 samples and the remaining tobacco plants to be estimated. Then, according to the variety and growth stage of the sample tobacco plants, the key spectral range of the sample tobacco plants is determined, and according to the spectral range and the key spectral range of the sample tobacco plants, the sample key spectrum is determined in the collected canopy reflectance spectrum of the tobacco plants.

[0092] The selected 36 samples were collected from the leaves of the tobacco plants. The collected tobacco leaves were stored in plastic bags, transported to the laboratory in an incubator, and weighed using an electronic balance on the same day. The fresh weight of each tobacco leaf was obtained, and the measured weight was used as the sample leaf weight.

[0093] The initial model can be constructed based on curve fitting, multiple regression, machine learning, or deep learning algorithms.

[0094] For example, the sample tobacco plant is K326, and the field tobacco plant is topped in early July. After topping, the tobacco leaves begin to mature, and the canopy reflectance spectrum measurement and leaf sampling can be selected after the tobacco plant is topped.

[0095] From the three combinations of green light, red light, and red edge, the green light and red light combination is selected as the key spectral interval for estimating the leaf weight of K326 tobacco plants at the topping stage.

[0096] Optionally, the sample key spectrum and sample leaf weight of the plurality of sample tobacco plants are used to train the initial model to obtain the tobacco leaf weight estimation model, comprising:

[0097] The combination of the sample key spectrum and the sample leaf weight of each sample tobacco plant is used as a sample to obtain a plurality of samples.

[0098] Based on the plurality of samples, a training set and a validation set are constructed;

[0099] The training samples in the training set are input into the initial model to obtain a trained initial model;

[0100] The validation samples in the validation set are input into the trained initial model to obtain the predicted weight corresponding to the validation samples output by the trained initial model;

[0101] The loss value is calculated using a preset loss function according to the predicted weight and the sample leaf weight of the validation samples;

[0102] In the case where the loss value is less than a preset threshold, or the number of training times of the initial model reaches a preset number, the tobacco leaf weight estimation model is determined.

[0103] Each sample key spectrum and the corresponding sample leaf weight of each sample key spectrum are combined as a sample, and a plurality of samples can be obtained.

[0104] The same variety and growth stage of the tobacco plant are used as a sample set.

[0105] The initial model is trained by using the training sample in each training set respectively, and the tobacco plant leaf weight estimation model corresponding to each variety and growth stage is obtained. The models are combined to form a model library. When the leaf weight of a to-be-tested tobacco plant needs to be estimated, the corresponding tobacco plant leaf weight estimation model can be selected from the model library according to the variety and growth stage of the to-be-tested tobacco plant, and then the target key spectrum of the to-be-tested tobacco plant is input into the model.

[0106] The initial model is trained by using the training sample in each training set respectively, and the tobacco plant leaf weight estimation model corresponding to each variety and growth stage is obtained. The models are combined to form a model library. When the leaf weight of a to-be-tested tobacco plant needs to be estimated, the corresponding tobacco plant leaf weight estimation model can be selected from the model library according to the variety and growth stage of the to-be-tested tobacco plant, and then the target key spectrum of the to-be-tested tobacco plant is input into the model.

[0107] For example, after obtaining a sample set composed of multiple samples of variety A in growth stage B, the sample set is divided into a training set and a validation set.

[0108] For any training sample in the training set, the sample key spectrum and sample leaf weight in the training sample are input into the initial model to train the model, and the predicted weight corresponding to the training sample is output, wherein the predicted weight refers to the predicted weight corresponding to different key spectra for the training sample.

[0109] After the training samples in the training set train the initial model, the trained initial model is obtained. The validation samples in the validation set are input into the trained initial model in turn, and the predicted weight output by the trained initial model is obtained.

[0110] On this basis, the loss value is calculated according to the predicted weight and the sample leaf weight of the validation sample in the validation set by using a preset loss function.

[0111] The preset loss function can be a root mean square error (RMSE) function, and correspondingly, the loss value is RMSE. In other embodiments, the preset loss function can be set according to actual needs, which is not limited here.

[0112] After the loss value is calculated, the current training process is completed, the model parameters in the initial model are updated, and then the next round of training is performed. During the training process, if the loss value calculated for a certain trained model is less than a preset threshold, the initial model training is completed.

[0113] Alternatively, when the number of times of training the initial model reaches a preset number of times, the trained model with the smallest loss value is selected as the tobacco plant leaf weight estimation model. The preset number of times can be a preset number of iterations.

[0114] Specifically, the whole iteration process is mainly driven by GA, GA selects a batch of wave bands, corresponding to the construction sample, PLSR constructs the corresponding initial model, uses the batch of samples to train and verify the initial model, and uses the loss function to calculate the loss value; GA performs wave band selection again, PLSR constructs the initial model again, calculates the loss value again, and continues iteration. Until the preset number of iterations is reached, then the smallest value of RMSE is selected to correspond to the model, as the final tobacco plant leaf weight estimation model of A variety in B growth stage.

[0115] According to the tobacco plant leaf weight estimation method based on canopy reflectance spectrum provided by the application, the tobacco plant leaf weight estimation model is constructed, thereby providing a basis for accurate prediction of the tobacco plant leaf weight.

[0116] Further, in step S2, the target key spectrum is input into the tobacco plant leaf weight estimation model to obtain the leaf weight of the tobacco plant to be measured output by the tobacco plant leaf weight estimation model; the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights.

[0117] The target key spectrum of the tobacco plant to be measured is input into the tobacco plant leaf weight estimation model, and the tobacco plant leaf weight estimation model calculates the leaf weight of the tobacco plant to be measured according to the target key spectrum, thereby obtaining and outputting the leaf weight of the tobacco plant to be measured.

[0118] Figure 3 is a scatter plot of the results of the tobacco plant leaf weight estimation using the full spectrum of 350-1000nm canopy reflectance spectrum provided by the application, Figure 4 is a scatter plot of the results of the tobacco plant leaf weight estimation using the green light, red light and red edge region canopy reflectance spectrum of 500-780nm provided by the application, Figure 5 is a scatter plot of the results of the tobacco plant leaf weight estimation using the green light and red light region canopy reflectance spectrum of 500-660nm provided by the application, and Figures 3-5 In the figure, the abscissa is the measured fresh weight (Measured fresh weight), the unit is gram (g); the ordinate is the estimated fresh weight (Estimated fresh weight) output by the tobacco plant leaf weight estimation model, the unit is gram (g); in an ideal case, the values of the abscissa and the ordinate are the same, the RMSE is the root mean square error, the smaller the RMSE, the stronger the estimation ability of the model; the R 2 (Coefficient of determination) is the determination coefficient, R 2 closer to 1, the stronger the estimation ability of the model.

[0119] The method for estimating tobacco leaf weight based on canopy reflectance spectrum provided by the application realizes estimation of tobacco leaf weight through a small number of key spectral bands, reduces the complexity of the model, improves prediction accuracy, and can realize measurement of canopy reflectance spectrum of a large range of tobacco plants according to the non-contact characteristics of spectral technology combined with different load platforms, so as to quickly obtain the leaf weight of a large range of tobacco plants.

[0120] The application further provides a server, which is provided with a processor; and further comprises a memory and a program or instruction stored on the memory and capable of running on the processor, and the program or instruction is executed by the processor to perform the method for estimating tobacco leaf weight based on canopy reflectance spectrum according to any one of the above embodiments.

[0121] The server provided by the application realizes estimation of tobacco leaf weight through a small number of key spectral bands, reduces the complexity of the model, improves prediction accuracy, and can realize measurement of canopy reflectance spectrum of a large range of tobacco plants according to the non-contact characteristics of spectral technology combined with different load platforms, so as to quickly obtain the leaf weight of a large range of tobacco plants.

[0122] Figure 6 is a flowchart of the method for estimating tobacco leaf weight provided by the application, as shown in Figure 6 , comprising:

[0123] First, the canopy reflectance spectrum of sample tobacco plants and the sample leaf weight are obtained;

[0124] Second, the key spectral interval is determined according to the variety and growth stage of the sample tobacco plants;

[0125] Then, the sample key spectrum is extracted according to the key spectral interval;

[0126] Then, the sample key spectrum and the combination of the sample key spectrum and the sample leaf weight corresponding to the sample key spectrum are used to train the estimation model, and a tobacco leaf weight estimation model is obtained;

[0127] Specifically, for example, after a sample set composed of multiple samples of A variety in B growth stage is obtained, the sample set is divided into a training set and a validation set;

[0128] For any training sample in the training set, the sample key spectrum and the sample leaf weight in the training sample are simultaneously input into an initial model to train the model, and the predicted weight corresponding to the training sample is output, wherein the predicted weight refers to the predicted weight corresponding to different key spectrums of the training sample.

[0129] After the training samples in the training set are used to train the initial model, a trained model is obtained, and the samples in the validation set are sequentially input into the trained model to obtain the predicted weight output by the trained model.

[0130] On this basis, the loss value is calculated according to the predicted weight of the sample in the verification set and the sample leaf weight.

[0131] After the loss value is calculated, the current round of training process is completed, the model parameters in the initial model are updated, and then the next round of training is performed. During the training process, if the loss value calculated for a certain training model is less than a preset threshold, the initial model training is completed.

[0132] Alternatively, in the case where the number of training of the initial model reaches a preset number, the model with the smallest loss value is selected as the tobacco plant leaf weight estimation model. The preset number can be a preset iteration number.

[0133] Specifically, the entire iteration process is mainly driven by GA, GA selects a batch of bands, corresponding to the construction sample, PLSR constructs the corresponding initial model, and trains and verifies the initial model using the batch of samples, and calculates the loss value using the loss function; GA performs band selection again, PLSR constructs the initial model again, calculates the loss value again, and continues iteration. Until a preset iteration number is reached, then the model corresponding to the smallest value in the RMSE is selected as the final model of the tobacco plant leaf weight estimation model of A variety at B growth stage.

[0134] Finally, the target key spectrum of the to-be-tested tobacco plant is input into the tobacco plant leaf weight estimation model to obtain the leaf weight estimation result output by the tobacco plant leaf weight estimation model.

[0135] The tobacco plant leaf weight estimation device based on canopy reflectance spectrum provided by the present application is described below. The tobacco plant leaf weight estimation device based on canopy reflectance spectrum described below can be mutually corresponding and referenced with the tobacco plant leaf weight estimation method based on canopy reflectance spectrum described above.

[0136] Figure 7 is a structural schematic diagram of the tobacco plant leaf weight estimation device based on canopy reflectance spectrum provided by the present application, as Figure 7 shown,

[0137] The acquisition module 701 is configured to acquire a target key spectrum of a to-be-tested tobacco plant; the target key spectrum is determined based on a variety, a growth stage and a canopy reflectance spectrum of the to-be-tested tobacco plant;

[0138] The input module 702 is configured to input the target key spectrum into a tobacco plant leaf weight estimation model to obtain a leaf weight of the to-be-tested tobacco plant output by the tobacco plant leaf weight estimation model; the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights.

[0139] During operation of the device, the acquisition module 701 acquires a target key spectrum of a to-be-tested tobacco plant; the target key spectrum is determined based on a variety, a growth stage and a canopy reflectance spectrum of the to-be-tested tobacco plant; the input module 702 inputs the target key spectrum into a tobacco plant leaf weight estimation model to acquire a leaf weight of the to-be-tested tobacco plant output by the tobacco plant leaf weight estimation model; and the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights.

[0140] The device for estimating a tobacco plant leaf weight based on a canopy reflectance spectrum provided by the application can realize estimation of a tobacco plant leaf weight through a small amount of key spectral bands, reduce complexity of a model, improve prediction accuracy, and realize measurement of a canopy reflectance spectrum of a large range of tobacco plants according to a non-contact feature of spectral technology combined with different load platforms, so as to quickly acquire a large range of tobacco plant leaf weights.

[0141] Figure 8 is a structural schematic diagram of an electronic device provided by the application, as shown in Figure 8 The electronic device can include a processor 810, a communications interface 820, a memory 830 and a communications bus 840, wherein the processor 810, the communications interface 820 and the memory 830 complete mutual communication through the communications bus 840. The processor 810 can call a logical instruction in the memory 830 to execute a method for estimating a tobacco plant leaf weight based on a canopy reflectance spectrum, which includes: acquiring a target key spectrum of a to-be-tested tobacco plant; the target key spectrum is determined based on a variety, a growth stage and a canopy reflectance spectrum of the to-be-tested tobacco plant; inputting the target key spectrum into a tobacco plant leaf weight estimation model to acquire a leaf weight of the to-be-tested tobacco plant output by the tobacco plant leaf weight estimation model; and the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights.

[0142] Further, the logic instructions in the memory 830 described above can be implemented by a software functional unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or partially contribute to the prior art, or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0143] In another aspect, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program is executed by a processor, so that the computer can execute the method for estimating the leaf weight of a tobacco plant based on the canopy reflectance spectrum, which comprises the following steps: obtaining a target key spectrum of a to-be-tested tobacco plant; the target key spectrum is determined based on the variety, growth stage and canopy reflectance spectrum of the to-be-tested tobacco plant; inputting the target key spectrum into a tobacco plant leaf weight estimation model to obtain the leaf weight of the to-be-tested tobacco plant output by the tobacco plant leaf weight estimation model; and the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights.

[0144] In another aspect, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program is executed by a processor, so that the computer can execute the method for estimating the leaf weight of a tobacco plant based on the canopy reflectance spectrum, which comprises the following steps: obtaining a target key spectrum of a to-be-tested tobacco plant; the target key spectrum is determined based on the variety, growth stage and canopy reflectance spectrum of the to-be-tested tobacco plant; inputting the target key spectrum into a tobacco plant leaf weight estimation model to obtain the leaf weight of the to-be-tested tobacco plant output by the tobacco plant leaf weight estimation model; and the tobacco plant leaf weight estimation model is obtained by training a plurality of sample key spectra with sample leaf weights.

[0145] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0146] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0147] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for estimating the weight of tobacco leaves based on canopy reflectance spectra, characterized in that, include: Obtain the target key spectrum of the tobacco plant to be tested; The target key spectrum is determined based on the variety, growth stage, and canopy reflectance spectrum of the tobacco plant to be tested; The target key spectrum is input into the tobacco plant leaf weight estimation model to obtain the leaf weight of the tobacco plant to be tested, which is output by the tobacco plant leaf weight estimation model. The tobacco plant leaf weight estimation model is obtained by training multiple sample key spectra with sample leaf weights. The acquisition of the target key spectrum of the tobacco plant to be tested includes: Based on the spectral response of the tobacco plant canopy and the canopy reflectance spectrum of the tobacco plant under test, the spectral range of several important spectral intervals for leaf weight estimation is determined. Based on the variety and growth stage of the tobacco plants to be tested, the multiple important spectral intervals are combined, and the combination with the best estimation result is selected as the key spectral interval. By utilizing the spectral range of the important spectral intervals and the key spectral intervals, band selection is performed on the canopy reflectance spectrum of the tobacco plant under test measured from the spectral measuring instrument to extract the target key spectrum.

2. The method for estimating tobacco leaf weight based on canopy reflectance spectrum according to claim 1, characterized in that, The spectral range for determining multiple important spectral intervals for leaf weight estimation based on the canopy spectral response of the tobacco plant and the canopy reflectance spectrum of the tobacco plant under test includes: Based on the canopy spectral response of tobacco plants, several important spectral ranges were determined; The spectral ranges of the plurality of important spectral intervals are determined in the canopy reflectance spectrum.

3. The method for estimating tobacco leaf weight based on canopy reflectance spectrum according to claim 1, characterized in that, Before inputting the target key spectrum into the tobacco leaf weight estimation model, the method further includes: Based on the variety and growth stage of each sample tobacco plant, the key spectra and leaf weights of multiple sample tobacco plants were obtained. The initial model was trained using the key spectra and leaf weights of the multiple sample tobacco plants to obtain the tobacco plant leaf weight estimation model.

4. The method for estimating tobacco leaf weight based on canopy reflectance spectrum according to claim 3, characterized in that, The process of training an initial model using key spectra and leaf weights of multiple sample tobacco plants to obtain a tobacco plant leaf weight estimation model includes: The combination of the key spectrum of each sample tobacco plant and the weight of the sample leaf is used as a sample to obtain multiple samples; Based on the aforementioned samples, a training set and a validation set are constructed; The training samples in the training set are input into the initial model to obtain the trained initial model; Input the validation samples in the validation set into the trained initial model to obtain the predicted weight corresponding to the validation samples output by the trained initial model; The loss value is calculated based on the predicted weight of the verification sample and the weight of the sample blade using a preset loss function; If the loss value is less than a preset threshold, or if the number of training iterations of the initial model reaches a preset number, the tobacco leaf weight estimation model is determined.

5. The method for estimating tobacco leaf weight based on canopy reflectance spectrum according to any one of claims 1-4, characterized in that, Before band selection in the canopy reflectance spectrum, the following is also included: Obtain the initial canopy spectral data of the tobacco plant to be tested; The initial canopy spectral data is denoised to generate denoised spectral data; Baseline correction is performed on the noise-reduced spectral data to generate the canopy reflectance spectrum of the tobacco plant to be tested.

6. A server, characterized in that, The server is equipped with a processor; it also includes a memory and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, perform the tobacco leaf weight estimation method based on canopy reflectance spectrum as described in any one of claims 1-5.

7. A device for estimating the weight of tobacco leaves based on canopy reflectance spectrum, characterized in that, include: The acquisition module is used to acquire the target key spectrum of the tobacco plant to be tested; The target key spectrum is determined based on the variety, growth stage, and canopy reflectance spectrum of the tobacco plant to be tested; The input module is used to input the target key spectrum into the tobacco plant leaf weight estimation model to obtain the leaf weight of the tobacco plant to be tested output by the tobacco plant leaf weight estimation model; the tobacco plant leaf weight estimation model is obtained by training multiple sample key spectra with sample leaf weights. The acquisition of the target key spectrum of the tobacco plant to be tested includes: Based on the spectral response of the tobacco plant canopy and the canopy reflectance spectrum of the tobacco plant under test, the spectral range of several important spectral intervals for leaf weight estimation is determined. Based on the variety and growth stage of the tobacco plants to be tested, the multiple important spectral intervals are combined, and the combination with the best estimation result is selected as the key spectral interval. By utilizing the spectral range of the important spectral intervals and the key spectral intervals, band selection is performed on the canopy reflectance spectrum of the tobacco plant under test measured from the spectral measuring instrument to extract the target key spectrum.

8. 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, it implements the tobacco leaf weight estimation method based on canopy reflectance spectrum as described in any one of claims 1-5.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the tobacco leaf weight estimation method based on canopy reflectance spectrum as described in any one of claims 1-5.

Images