Method for monitoring growth state of crops in real time

Abstract

The invention relates to the technical field of spectral analysis, in particular to a crop growth state real-time monitoring method, which comprises the following steps of: acquiring hyperspectral data and temperature data consisting of all wavebands and reflection intensities at each sampling moment; acquiring a spectrum curve at each sampling moment and a progressive line of each wave band; calculating an asymptotic line noise evaluation coefficient, and further obtaining an asymptotic line difference coefficient; calculating a spectrum progressive temperature drift coefficient at each sampling moment, forming a temperature drift verification sequence at each sampling moment according to the hyperspectral data and the temperature data, and further obtaining a temperature drift verification factor at each sampling moment; calculating a noise temperature drift verification coefficient, and obtaining a self-adaptive wavelet threshold value of each decomposition layer number at each sampling moment; and denoising the hyperspectral data at each sampling moment according to a self-adaptive wavelet threshold to obtain the growth condition of the crops. The invention aims to solve the problems of incomplete noise reduction and excessive smoothness of signals caused by a fixed threshold in traditional wavelet threshold filtering.

Description

Technical Field

[0001] This application relates to the technical field of spectral analysis, and particularly to a method for real-time monitoring of the growth status of crops. Background Art

[0002] With global climate change and population growth, agricultural production faces unprecedented challenges. To ensure food security and improve crop yield and quality, it has become crucial to monitor the growth status of crops in real time. Currently, the methods for monitoring crop growth mainly rely on sensor spectral monitoring technology. Due to the complexity and diversity of the farmland environment, the data collected by sensors often contain noise and missing values, which affect the accuracy of the monitoring results.

[0003] Spectral analysis can more accurately reflect the data information of various components in crop growth, so it is often used to detect the growth of crops. Since the sensor is affected by temperature changes, the crop information obtained by detection will have a temperature drift phenomenon, and the spectral information of the crop is affected by noise due to temperature differences. If the growth of the crop is directly detected, it is easy to cause misdetection. Therefore, it is necessary to perform data denoising on the spectral information of the crop. Wavelet threshold denoising has been widely used in data denoising due to its advantages of simple algorithm, small computational amount and easy implementation. However, the fixed threshold selection usually makes the denoised signal have phenomena such as edge blurring, over-smoothing and local oscillation, and cannot achieve an ideal denoising effect. To solve the above problems, the present invention proposes a method for real-time monitoring of the growth status of crops, aiming to adaptively adjust the threshold in wavelet denoising by using the temperature drift situation and data signals of the spectral data of crop growth, and improve the accuracy of crop growth monitoring. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method for real-time monitoring of the growth status of crops to solve the existing problems.

[0005] The method for real-time monitoring of the growth status of crops of the present invention adopts the following technical solution: An embodiment of the present invention provides a method for real-time monitoring of the growth status of crops, and the method includes the following steps: Obtain hyperspectral data and temperature data composed of all bands and reflection intensities at each sampling moment; Obtain the spectral curves at each sampling moment based on the hyperspectral data; obtain the asymptotic line functions for each band in the spectral curves; obtain the asymptotic line noise evaluation coefficients at each sampling moment based on the hyperspectral data and the asymptotic line functions; obtain the asymptotic line difference coefficients at each sampling moment based on the asymptotic line functions; obtain the spectral asymptotic temperature drift coefficients at each sampling moment based on the asymptotic line difference coefficients and the asymptotic line noise evaluation coefficients; form the temperature drift verification sequences at each sampling moment based on the hyperspectral data and the temperature data; obtain the temperature drift verification factors at each sampling moment based on the temperature drift verification sequences; obtain the noise temperature drift calibration coefficients at each sampling moment based on the temperature drift verification factors and the hyperspectral data; obtain the adaptive wavelet thresholds for each decomposition level at each sampling moment based on the noise temperature drift calibration coefficients; Use the adaptive wavelet threshold as the threshold in the wavelet soft threshold function for wavelet denoising, and perform wavelet denoising on the hyperspectral data of the crops at each sampling moment, and calculate the cosine similarity between the denoised hyperspectral data and the hyperspectral data in the database to obtain the growth conditions of the crops.

[0006] Further, the obtaining of the spectral curves at each sampling moment based on the hyperspectral data includes: Use the least squares method to fit the hyperspectral data at each sampling moment to obtain the spectral curves at each sampling moment.

[0007] Further, the obtaining of the asymptotic line functions for each band in the spectral curves includes: For the spectral curves at each sampling moment, make a tangent line at the point where each band is located, and use the tangent line as the asymptotic line function for each band in the spectral curves.

[0008] Further, the obtaining of the asymptotic line noise evaluation coefficients at each sampling moment includes: For each sampling moment, calculate the absolute value of the difference between the reflection intensity corresponding to the (i - 1)-th band in the hyperspectral data and the function value of the asymptotic line function of the i-th band at the (i - 1)-th band as the first absolute value of the difference, calculate the absolute value of the difference between the reflection intensity corresponding to the (i + 1)-th band in the hyperspectral data and the function value of the asymptotic line function of the i-th band at the (i + 1)-th band as the second absolute value of the difference, calculate the sum value of the first absolute value of the difference and the second absolute value of the difference, and use the sum value of all such sum values for all bands at each sampling moment as the asymptotic line noise evaluation coefficient at each sampling moment.

[0009] Further, the obtaining of the asymptotic line difference coefficients at each sampling moment includes: Calculate the absolute value of the difference between the asymptotic line function values of each sampling moment and its previous sampling moment for each band, calculate the integral value of the absolute value of the difference from the adjacent previous band to the subsequent band for each band, and use the mean value of the integral values for all bands as the asymptotic line difference coefficient at each sampling moment.

[0010] Further, the obtaining of the spectral gradual temperature drift coefficients at each sampling moment includes: For each sampling moment, calculate the product of the first preset weight adjustment factor and the asymptote noise evaluation coefficient as the first product, calculate the result of the exponential function with the natural constant as the base and the absolute value of the difference between the temperature data at each sampling moment and the previous sampling moment as the exponent, calculate the product of the second preset weight adjustment factor, the asymptote difference coefficient and the calculation result as the second product, and take the sum of the first product and the second product as the spectral gradual temperature drift coefficient at each sampling moment.

[0011] Further, the forming of the temperature drift verification sequence at each sampling moment according to the hyperspectral data and the temperature data includes: Obtain the limited range within plus or minus 10% of the temperature data at each sampling moment, and arrange the hyperspectral data at the sampling moments where all the temperature data within the limited range are located in the order of the sampling moments to form the temperature drift verification sequence.

[0012] Further, the obtaining of the temperature drift verification factor at each sampling moment includes: Calculate the absolute value of the difference between the spectral gradual temperature drift coefficient of each hyperspectral data in the temperature drift verification sequence at each sampling moment and the spectral gradual temperature drift coefficient at each sampling moment, and calculate the sum of the absolute values of the differences of all the hyperspectral data in the temperature drift verification sequence; Calculate the result of the exponential function with the natural constant as the base and the spectral gradual temperature drift coefficient at each moment as the exponent, and calculate the product of the sum and the calculation result as the temperature drift verification factor at each sampling moment.

[0013] Further, the obtaining of the noise temperature drift calibration coefficient at each sampling moment includes: For each sampling moment, calculate the difference between the reflection intensity of the hyperspectral data in the i-th band and the mean value of the reflection intensities of all the hyperspectral data in the i-th band in the temperature drift verification sequence, calculate the mean value of the differences in all the bands, and take the product of the temperature drift verification factor and the mean value as the noise temperature drift calibration coefficient at each sampling moment.

[0014] Further, the obtaining of the adaptive wavelet threshold at each sampling moment and each decomposition level includes: For each sampling moment, use wavelet decomposition to process the hyperspectral data to obtain the total number of decomposition levels and all the wavelet coefficients at each decomposition level; Calculate the mean square deviation of all the wavelet coefficients at each decomposition level at the sampling moment, calculate the square root of the noise temperature drift calibration coefficient at the sampling moment, and calculate the product of the mean square deviation and the square root; Calculate the sum of the number 1 and the total number of decomposition levels, and calculate the result of the logarithmic function with the sum as the true number and the natural constant as the base; Calculate the ratio of the product to the calculation result as the adaptive wavelet threshold for each decomposition level at each sampling moment.

[0015] The present invention has at least the following beneficial effects: The present invention measures the interference of noise by using the asymptote approximation according to the information of each reflection band in the hyperspectral data, thereby obtaining the asymptote noise evaluation coefficient to measure the interference of noise on the data. In addition, the spectral asymptote temperature drift coefficient is obtained by combining the temperature difference to measure the influence of temperature drift on each sampling moment, and the temperature drift verification sequence is obtained by limiting the temperature range. The information in the temperature drift sequence is used to verify the noise and temperature drift to obtain the noise temperature drift calibration coefficient, which solves the problems of incomplete noise reduction and over-smoothing of the signal caused by the fixed threshold in the traditional wavelet threshold filtering, improves the data quality of the hyperspectral, can more accurately reflect the influence of environmental noise and temperature drift phenomena, and finally adjusts the adaptive wavelet threshold according to different wavelet hierarchical levels, improving the detection effect of the growth trend of crops. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a flowchart of a method for real-time monitoring of the growth state of a crop provided by the present invention; Figure 2 It is a flowchart of data noise reduction. Detailed Embodiments

[0018] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following will describe in detail the specific embodiments, structures, features and effects of a method for real-time monitoring of the growth state of a crop proposed according to the present invention in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field of the present invention.

[0020] The following will specifically describe the specific solution of a method for real-time monitoring of the growth state of a crop provided by the present invention in combination with the drawings.

[0021] A method for real-time monitoring of the growth status of crops provided by an embodiment of the present invention. Specifically, the following method for real-time monitoring of the growth status of crops is provided. Please refer to Figure 1 , the method includes the following steps: Step S001, obtain the spectral data of the growth of rice through a spectral analyzer, and collect hyperspectral data and temperature data.

[0022] The growth stages of rice are generally divided into the seedling stage, tillering stage, jointing stage, booting stage, heading stage and maturity stage. Different soil and water conditions are required in different growth stages. Therefore, it is necessary to monitor the growth of rice in real time, and adjust the organic matter, trace elements and water in the soil according to the growth of rice to meet the healthy growth of rice. The spectral information during the growth detection of rice is easily interfered by the cold region environmental conditions. Therefore, it is necessary to perform adaptive noise reduction on the obtained spectral data. The specific processing steps are as Figure 2 shown.

[0023] To obtain the growth of rice, this embodiment uses a Hyperion 3000 spectral analyzer to sample the hyperspectral data of rice stems in a cold region experimental paddy field. Since the growth of rice does not change significantly in a short period of time, the sampling interval in this embodiment is set to 30 minutes, and the hyperspectral data at a single sampling moment is denoted as , where represents the hyperspectral data of cold region rice at the kth sampling moment, represents the reflection intensity of the i-th band at the kth sampling moment, represents the number of bands collected by the spectral analyzer. In addition, to measure the influence of the temperature spectrometer working, equal-interval sampling is performed with the rice hyperspectral data, and the temperature data at the kth sampling moment is .

[0024] Step S002, construct a curve asymptote according to the reflection intensity information of each band in the cold region rice hyperspectral data, thereby obtaining an asymptote noise evaluation coefficient, combine the temperature difference between two adjacent sampling moments and the asymptote integral difference of the corresponding band to obtain a spectral asymptote temperature drift coefficient, obtain a temperature drift verification sequence by limiting the temperature range, use the high-light reflection information of each band in the temperature drift verification sequence to obtain a noise temperature drift calibration coefficient, and finally obtain an adaptive level wavelet threshold for different wavelet decomposition levels.

[0025] Due to factors such as redundant data bands collected by the hyperspectral analyzer and large special day-night temperature differences in cold regions, the hyperspectral data of rice collected by the spectral analyzer cannot directly reflect the true state of rice and contains interference from many environmental noises. Therefore, it is necessary to perform noise reduction processing on the collected spectral data. Wavelet denoising has the advantage of good noise reduction effect and is widely used. However, the selection of the threshold in wavelet denoising has a great impact on the signal denoising effect. A smaller threshold will result in incomplete signal denoising and residual noise; a larger threshold will oversmooth the effective signal and cause signal distortion.

[0026] The interference of environmental noise is random, so the degree of interference of environmental noise at each sampling moment is different. Therefore, it is necessary to evaluate the noise interference situation based on the hyperspectral data of rice stems in cold regions at each sampling moment.

[0027] First, in this embodiment, the hyperspectral data at the k-th sampling moment is fitted into a continuous spectral curve . Since the nonlinear least squares method is a well-known technology, it will not be elaborated in this embodiment.

[0028] In the continuous spectral curve , taking the coordinates of each band as the tangent point and the tangent line corresponding to the band as the direction, the band asymptote function is constructed accordingly. For example, for the coordinates of the i-th band, in the corresponding continuous spectral curve , the tangent direction is constructed, and thus the asymptote function of the current band can be obtained.

[0029] Asymptotes are constructed for each band at the current sampling moment , and the asymptote noise evaluation coefficient is constructed: In the formula, represents the asymptote noise evaluation coefficient of the hyperspectral data of rice in cold regions at the k-th sampling moment; represents the number of bands collected by the spectral analyzer, represents the reflection intensities of the i-1 and i+1 bands in the hyperspectral data at the k-th sampling moment, respectively represent the function values of the asymptote function of the i-th band in the hyperspectral data at the k-th sampling moment at the i-1 and i+1 bands.

[0030] Formula logic: When the influence of the noise signal is small at the current sampling moment, the continuous hyperspectral curve can be well fitted through non - linear fitting, and there will be no large difference transformation in the reflection intensities of the two adjacent bands of the i - th band. Thus, the difference between the reflection intensity value of the same band and the asymptote function value is small, and finally the asymptote noise evaluation coefficient at the current sampling moment is obtained. The value is small. On the contrary, when the noise interference is severe at the current moment, there will be many outliers, and the value of the abnormal reflection intensity deviates far from the asymptote. Thus, the obtained asymptote noise evaluation coefficient increases.

[0031] Through the asymptote noise evaluation coefficient the noise interference situation of the hyperspectral data at the current sampling moment can be reflected. The noise interference situations at different sampling moments are different. At the same time, the special environmental conditions in cold regions need to be considered. In cold regions, the temperature difference between day and night is large. Although it can effectively promote the accumulation of rice starch, it also brings difficulties to the growth detection of cold - region rice. Under different temperature conditions, there will be certain differences in the reflection intensities of the hyperspectral data of rice in the corresponding bands. This phenomenon is called temperature drift. Thus, by combining the temperature change difference suitable for different samplings and the asymptote of the spectral curve, the spectral asymptote temperature - drift coefficient can be obtained: In the formula, represents the spectral asymptote temperature - drift coefficient at the k - th sampling moment, and represent the weight adjustment factors, which are set to 0.4 and 0.6 respectively according to experience, represents the asymptote noise evaluation coefficient of the hyperspectral data of cold - region rice at the k - th sampling moment, and represent the temperature values at the k - th sampling moment and the (k - 1) - th sampling moment respectively, and e is the natural constant; represents the asymptote difference coefficient at the k - th sampling moment, represents the number of bands collected by the spectral analyzer, and represent the spectral curve asymptote functions at the k - th sampling moment and the (k - 1) - th sampling moment at the i - th reflection band respectively.

[0032] Formula logic: Due to the harsh living environment of cold-region rice, the growth degree of cold-region rice between two sampling time periods is small, so the difference between the hyperspectral data obtained at two sampling moments is small. If the noise of the hyperspectral data collected at this time is large, there will be a large difference between the hyperspectral data obtained at two sampling moments. From this, it can be obtained that the larger the difference between the asymptotes of the single-band curves corresponding to two adjacent sampling moments, at this time, the integral value of the function obtained by subtracting the asymptote functions of the two sampling moments between the two sampling moments is large, and the asymptote difference coefficient is obtained. The larger the value, and at the same time the temperature will also affect the larger the environmental noise, making the asymptote noise evaluation coefficient have a large value, and finally make the spectral asymptote temperature drift coefficient increase in value. On the contrary, when the environmental temperature changes little, the difference between the asymptote function values is small, making the spectral asymptote temperature drift coefficient decrease in value.

[0033] The influence of the temperature drift phenomenon caused by temperature change on the hyperspectral data is reflected by the spectral asymptote temperature drift coefficient , but this process is obtained through the asymptotes of the reflection intensities of each band of the spectral curve, and there may be certain errors. Therefore, it is necessary to correct it by combining the real hyperspectral data.

[0034] In a one-day sampling cycle, within the range of plus or minus 10% of the temperature data at the k-th sampling moment, the hyperspectral data obtained at the sampling moments where the temperature data within the limited range is located are acquired, and the hyperspectral data are arranged in the order of sampling moments to form a temperature drift verification sequence.

[0035] Furthermore, in this embodiment, the interference situation of the cold-region rice hyperspectral data by the environment will be analyzed, and the noise temperature drift calibration coefficient of each sampling moment is constructed by combining the data distribution in the temperature drift verification sequence and the spectral asymptote temperature drift coefficient of each sampling moment. The specific expression is: In the formula, represents the noise temperature drift calibration coefficient at the k-th sampling moment, represents the temperature drift verification factor at the k-th sampling moment; represents the total number of bands collected by the spectral analyzer, is the number of hyperspectral data in the temperature drift verification sequence at the k-th sampling moment, represents the reflection intensity of the i-th band in the cold-region rice hyperspectral data at the k-th sampling moment, represents the mean value of the reflection intensities of all hyperspectral data in the temperature drift verification sequence at the current sampling moment in the i-th band. represents the spectral progressive temperature drift coefficient at the k-th sampling moment, represents the spectral progressive temperature drift coefficient of the j-th hyperspectral data in the temperature drift verification sequence obtained at the k-th sampling moment.

[0036] Formula logic: Taking the temperature at the k-th sampling moment as the standard, in the temperature drift verification sequence screened by limiting the temperature range, although there are certain differences in temperature, theoretically the temperature drift of hyperspectral data is small. Thus, when the current sampling moment is greatly affected by environmental interference, there are large differences in the reflection intensity between each band at the current sampling moment and the temperature drift verification sequence, that is , in addition, when calculating the spectral progressive temperature drift coefficient , considering the asymptotic line noise evaluation coefficient , thus has a large difference from the spectral progressive temperature drift coefficient in the temperature drift verification sequence, which makes the temperature drift verification factor have a large value. On the contrary, if the cold region rice is less affected by environmental noise at the current sampling time, finally has a small value.

[0037] By traversing the hyperspectral data of each sampling moment, the noise temperature drift calibration coefficient corresponding to the corresponding moment can be obtained, which can reflect the interference of the hyperspectral data of cold region rice at the current sampling moment by the environment. By analyzing the distribution characteristics of the noise in the data, the interference of the noise in different hierarchical wavelet coefficients is different, and it shows that as the number of wavelet decomposition layers increases, the interference of the noise also decreases rapidly.

[0038] Therefore, in this embodiment, for the hyperspectral data of each sampling moment, wavelet decomposition is performed by the wavelet decomposition algorithm, the sym2 wavelet basis is used, and the number of wavelet decomposition layers is set to 5. Among them, the wavelet decomposition algorithm is a well-known technology in the art and will not be elaborated here.

[0039] Combined with the noise temperature drift calibration coefficient at the current sampling moment, an adaptive wavelet threshold is constructed for different decomposition layers: In the formula, represents the adaptive wavelet threshold at the m-th decomposition layer after wavelet decomposition of the hyperspectral data at the k-th sampling moment; represents the mean square deviation of the wavelet coefficients at the m-th decomposition layer after wavelet decomposition of the hyperspectral data at the k-th sampling moment, represents the noise temperature drift calibration coefficient at the k-th sampling moment, and m represents the m-th decomposition layer in the wavelet decomposition.

[0040] Formula logic: When the hyperspectral data at the current sampling moment is greatly affected by environmental noise, the obtained noise temperature drift calibration coefficient The value is large, and the mean square error of wavelet coefficients at different levels is obtained through wavelet decomposition is large, and the finally obtained adaptive wavelet threshold is large. When m corresponds to the first layer, at this time , for the first layer the value is large, and the noise is mainly concentrated in the first layer. As the wavelet decomposition level increases, the adaptive hierarchical wavelet threshold will relatively decrease, so the noise suppression effect is better when the noise interference is large. On the contrary, when the noise interference is small, the obtained adaptive hierarchical wavelet threshold will be small, and more signal detail information can be retained at this time.

[0041] Among them, in this embodiment, the Haar wavelet basis function is used to perform wavelet decomposition on the hyperspectral data. The implementer can select other wavelet basis functions according to the actual situation; after the hyperspectral data is wavelet decomposed, multiple layers of wavelet coefficients will be obtained. Each layer of wavelet coefficients represents the information of the hyperspectral data in different frequency ranges. The specific decomposition process and wavelet coefficients of wavelet decomposition are well-known technologies to those skilled in the art and will not be elaborated here.

[0042] Thus, the adaptive wavelet threshold is obtained.

[0043] Step S003, perform wavelet threshold denoising on the hyperspectral data of the cold region rice stem height at each sampling moment according to the adaptive hierarchical wavelet threshold, and compare it with the hyperspectral data in the database to obtain the growth situation of the current rice.

[0044] According to Step S002, the adaptive wavelet threshold can be obtained according to the noise interference situation of the hyperspectral data at each sampling moment , and the adaptive wavelet threshold is used as the threshold in the wavelet soft threshold function in wavelet denoising to obtain the improved wavelet soft threshold function. Among them, the wavelet soft threshold function is well-known to those skilled in the art and will not be elaborated here.

[0045] According to the improved wavelet soft threshold function, wavelet denoising is used to denoise the obtained hyperspectral data. Among them, the process of wavelet denoising is: wavelet decomposition, threshold denoising, and signal reconstruction. In this embodiment, after wavelet decomposition of the collected hyperspectral data, the improved wavelet soft threshold function is used to denoise the wavelet coefficients, and then the denoised wavelet coefficients are inversely transformed to obtain the denoised hyperspectral data.

[0046] The specific denoising process is a well-known technology to those skilled in the art and will not be elaborated here.

[0047] Thus, the denoised hyperspectral data of cold region rice is obtained .

[0048] The hyperspectral data Match the hyperspectral data filtered at each sampling moment with the standard hyperspectral data of rice at the seedling stage, tillering stage, jointing stage, booting stage, heading stage and maturity stage in the database. Specifically, calculate the cosine similarity between the hyperspectral data of the rice growth stage at each sampling moment and the hyperspectral data in the database, and perform similarity ranking. Among them, the calculation method of cosine similarity is a well-known technology and will not be elaborated here.

[0049] During the sampling period from zero to twenty-four hours in a day, count the maximum value of the similarity ranking of the growth stage at each sampling moment and mark it. Thus, the stage with the highest similarity occurrence rate within the one-day sampling period is the growth stage of the current rice. Provide the analysis data to the technicians of cold-region rice cultivation technology to provide data support for their rice cultivation work.

[0050] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the advantages or disadvantages of the embodiments. And the above specific embodiments of this specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0051] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between each embodiment, reference can be made to each other. Each embodiment focuses on the differences from other embodiments.

[0052] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; modifying the technical solutions recorded in the foregoing embodiments, or equivalently replacing some of the technical features, does not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of each embodiment of the present application, and should all be included within the protection scope of the present application.

Claims

1. A method for real-time monitoring of the growth state of crops, characterized in that, The method includes the following steps: Obtain hyperspectral data and temperature data composed of all bands and reflection intensities at each sampling moment; Obtain the spectral curve at each sampling moment according to the hyperspectral data; obtain the asymptote function of each band in the spectral curve; obtain the asymptote noise evaluation coefficient at each sampling moment according to the hyperspectral data and the asymptote function; obtain the asymptote difference coefficient at each sampling moment according to the asymptote function; obtain the spectral asymptote temperature drift coefficient at each sampling moment according to the asymptote difference coefficient and the asymptote noise evaluation coefficient; form the temperature drift verification sequence at each sampling moment according to the hyperspectral data and the temperature data; obtain the temperature drift verification factor at each sampling moment according to the temperature drift verification sequence; obtain the noise temperature drift calibration coefficient at each sampling moment according to the temperature drift verification factor and the hyperspectral data; obtain the adaptive wavelet threshold of each decomposition level at each sampling moment according to the noise temperature drift calibration coefficient; Take the adaptive wavelet threshold as the threshold in the wavelet soft threshold function in wavelet denoising, use wavelet denoising to denoise the hyperspectral data of the crops at each sampling moment, and calculate the cosine similarity between the denoised hyperspectral data and the hyperspectral data in the database to obtain the growth condition of the crops.

2. The real-time monitoring method for the growth state of a crop according to claim 1, characterized in that The obtaining of the spectral curve at each sampling moment according to the hyperspectral data includes: Use the least squares method to fit the hyperspectral data at each sampling moment to obtain the spectral curve at each sampling moment.

3. The real-time monitoring method for the growth state of a crop according to claim 1, characterized in that The obtaining of the asymptote function of each band in the spectral curve includes: For the spectral curve at each sampling moment, make a tangent line at the point where each band is located, and take the tangent line as the asymptote function of each band in the spectral curve.

4. The real-time monitoring method for the growth state of a crop according to claim 1, characterized in that, The obtaining of the asymptote noise evaluation coefficient at each sampling moment includes: For each sampling moment, calculate the absolute value of the difference between the reflection intensity corresponding to the (i - 1)-th band in the hyperspectral data and the function value of the asymptote function of the i-th band at the (i - 1)-th band as the first absolute value of the difference, calculate the absolute value of the difference between the reflection intensity corresponding to the (i + 1)-th band in the hyperspectral data and the function value of the asymptote function of the i-th band at the (i + 1)-th band as the second absolute value of the difference, calculate the sum value of the first absolute value of the difference and the second absolute value of the difference, and take the sum value of all such sum values of all bands at each sampling moment as the asymptote noise evaluation coefficient at each sampling moment.

5. The real-time monitoring method for the growth state of a crop according to claim 1, wherein, The obtaining of the asymptote difference coefficient at each sampling moment includes: Calculate the absolute value of the difference between the asymptote function values of each sampling moment and its previous sampling moment at each band, calculate the integral value of the absolute value of the difference from the adjacent previous band to the next band at each band, and take the mean value of the integral values of all bands as the asymptote difference coefficient at each sampling moment.

6. The real-time monitoring method for the growth state of a crop according to claim 1, characterized in that, The obtaining of the spectral asymptote temperature drift coefficient at each sampling moment includes: For each sampling moment, calculate the product of the first preset weight adjustment factor and the asymptote noise evaluation coefficient as the first product, calculate the result of the exponential function with the natural constant as the base and the absolute value of the difference between the temperature data at each sampling moment and the temperature data at the previous sampling moment as the exponent, calculate the product of the second preset weight adjustment factor, the asymptote difference coefficient and the calculation result as the second product, and take the sum of the first product and the second product as the spectral asymptotic temperature drift coefficient at each sampling moment.

7. The real-time monitoring method for the growth state of a crop according to claim 1, characterized in that, The composition of the temperature drift verification sequence at each sampling moment according to the hyperspectral data and the temperature data includes: Obtain the limited range within plus or minus 10% of the temperature data at each sampling moment. For the hyperspectral data at the sampling moments where all the temperature data are within the limited range, arrange them in the order of the sampling moments to form the temperature drift verification sequence.

8. The real-time monitoring method for the growth state of a crop according to claim 1, characterized in that, The obtaining of the temperature drift verification factor at each sampling moment includes: Calculate the absolute value of the difference between the spectral asymptotic temperature drift coefficient of each hyperspectral data in the temperature drift verification sequence at each sampling moment and the spectral asymptotic temperature drift coefficient at each sampling moment, and calculate the sum of the absolute values of the differences of all the hyperspectral data in the temperature drift verification sequence; Calculate the result of the exponential function with the natural constant as the base and the spectral asymptotic temperature drift coefficient at each moment as the exponent, and calculate the product of the sum value and the calculation result as the temperature drift verification factor at each sampling moment.

9. A real-time monitoring method for the growth state of crops according to claim 1, characterized in that, The obtaining of the noise temperature drift calibration coefficient at each sampling moment includes: For each sampling moment, calculate the difference between the reflection intensity of the hyperspectral data in the i-th band and the mean value of the reflection intensities of all the hyperspectral data in the i-th band in the temperature drift verification sequence, calculate the mean value of the differences of all the bands, and take the product of the temperature drift verification factor and the mean value as the noise temperature drift calibration coefficient at each sampling moment.

10. A real-time monitoring method for the growth state of crops according to claim 1, characterized in that, The obtaining of the adaptive wavelet threshold at each decomposition level at each sampling moment includes: For each sampling moment, use wavelet decomposition to process the hyperspectral data to obtain the total number of decomposition levels and all the wavelet coefficients at each decomposition level; Calculate the mean square deviation of all the wavelet coefficients at each decomposition level at the sampling moment, calculate the square root of the noise temperature drift calibration coefficient at the sampling moment, and calculate the product of the mean square deviation and the square root; Calculate the sum value of the number 1 and the total number of decomposition levels, and calculate the result of the logarithmic function with the sum value as the true number and the natural constant as the base; Calculate the ratio of the product to the calculation result as the adaptive wavelet threshold at each decomposition level at each sampling moment.

Images