Chlorophyll content inversion method and system based on improved random linear mixed model

Abstract

The present invention relates to the field of data processing technology, and proposes a chlorophyll content inversion method and system based on an improved random linear mixed model, comprising the following steps: obtaining unmanned aerial vehicle hyperspectral data for pixel screening, combining the spectral data based on the improved random linear mixed model, and obtaining a mixed pixel spectral database of pure pixels of target crops under different backgrounds; performing similarity matching to obtain continuous hyperspectral data with the highest similarity to pixel composition in satellite multispectral data; and then reversely calculating based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area, thereby obtaining the best inversion result. The present invention utilizes unmanned aerial vehicle hyperspectral data, and reduces the interference of mixed pixels in satellite remote sensing data by optimizing the random mixed model, so as to achieve low-cost satellite data large-area, high-precision chlorophyll content inversion.

Description

Technical Field

[0001] The present invention relates to the technical field related to data processing, and in particular to a chlorophyll content inversion method and system based on an improved random linear mixed model. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] As a basic indicator of vegetation health and photosynthesis, chlorophyll plays a vital role in judging crop growth status and evaluating crop yield. Traditional methods of measuring chlorophyll content are usually based on sample collection and chemical analysis of crops in a small area. This method is not only destructive and time-consuming, but also difficult to expand the research scope on a large scale. Due to the unique spectral characteristics of vegetation in the visible / near-infrared region, the use of remote sensing technology to monitor vegetation parameters over a wide range of spatial and temporal scales has become a common choice. However, there are still certain problems.

[0004] Although satellite spectral remote sensing has been widely used in monitoring vegetation chlorophyll content, it is affected by the mutual constraints of its spatial resolution and spectral resolution. High spatial resolution satellite images often have insufficient bands, making it difficult to accurately invert vegetation chlorophyll content and the data cost is high; high spectral resolution images usually have a spatial resolution of meters, and each pixel may contain spectral information of multiple ground objects, that is, multiple mixed pixels, making it impossible to accurately invert chlorophyll content; for example, intercropping corn and soybeans is a common way to improve land utilization, which also leads to the fact that corn and soybeans grow very similarly during the critical growth period of July to September and are difficult to distinguish. When estimating chlorophyll content based on satellite images, due to the limitation of spatial resolution, existing studies often ignore the fact that satellite images usually contain not only the reflection signal of corn, but also the interference of spectral information of soybeans to form mixed pixels; UAV remote sensing flight mode is flexible and the flight altitude is controllable, but due to hardware conditions, it is impossible to monitor the chlorophyll content of vegetation over a large area.

[0005] At present, there are challenges in inverting vegetation chlorophyll content using satellite remote sensing technology. Specifically, the observed spectral reflectance is not only affected by chlorophyll, but also by canopy structure (such as leaf area index), crop intercropping and soil background. This makes it difficult to extract pure chlorophyll information from the reflectance spectrum. To overcome this problem, researchers have proposed using spectral indices (such as NDVI, EVI, etc.) to help separate chlorophyll information. However, these spectral indices are often easily affected by canopy structure and have low sensitivity to chlorophyll. In addition, although some spectral indices with higher accuracy (such as NAOC) can more accurately reflect chlorophyll content, they require more bands (spectral information at different wavelengths) to achieve this. The currently widely used open and free satellite data usually only provide limited bands and cannot provide enough information to accurately estimate chlorophyll content. This makes it difficult to improve the accuracy of chlorophyll content inversion based on traditional satellite data, limiting its practical application in large-area remote sensing monitoring. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a chlorophyll content inversion method and system based on an improved random linear mixed model. By utilizing UAV hyperspectral data and optimizing the random mixed model, the interference of mixed pixels in satellite remote sensing data can be reduced, and low-cost satellite data can be used to invert large-area and high-precision chlorophyll content.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] One or more embodiments provide a chlorophyll content inversion method based on an improved random linear mixed model, comprising the following steps:

[0009] The UAV hyperspectral data is obtained for pixel screening, and the spectral data is combined based on the improved random linear mixed model to obtain a mixed pixel spectral database of pure pixels of the target crop under different backgrounds;

[0010] The acquired satellite multispectral data is matched with the spectrum of the mixed pixel spectrum database to obtain continuous hyperspectral data with the highest similarity to the pixel composition in the satellite multispectral data;

[0011] The obtained continuous hyperspectral data are reversely calculated based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area;

[0012] The vegetation NAOC index of the target crop was calculated based on the pure pixel reflectance of the target crop, and a linear regression was performed with the measured chlorophyll value to obtain the optimal inversion result of the chlorophyll content of the target crop.

[0013] One or more embodiments provide a chlorophyll content inversion system based on an improved random linear mixed model, including:

[0014] Mixed pixel spectral database construction module: It is configured to obtain the UAV hyperspectral data for pixel screening, combine the spectral data based on the improved random linear mixed model, and obtain the mixed pixel spectral database of the pure pixels of the target crop under different backgrounds;

[0015] Matching module: configured to perform similarity matching between the acquired satellite multispectral data and the spectrum of the mixed pixel spectrum database, and obtain continuous hyperspectral data with the highest similarity to the pixel composition in the satellite multispectral data;

[0016] Vegetation spectral data calculation module: configured to reversely calculate the obtained continuous hyperspectral data based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area;

[0017] Inversion module: It is configured to calculate the vegetation NAOC index of the target crop based on the pure pixel reflectance of the target crop, and perform linear regression with the measured chlorophyll value to obtain the best inversion result of the chlorophyll content of the target crop.

[0018] An electronic device includes a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are run by the processor, the steps in the chlorophyll content inversion method based on the improved random linear mixed model are completed.

[0019] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the steps in the above-mentioned chlorophyll content inversion method based on an improved random linear mixed model are completed.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] In the present invention, by making full use of the advantages of UAV's hyperspectral data, a mixed hyperspectral database under different planting conditions of target crops is constructed. By matching the acquired satellite data, the chlorophyll content of the target crops is inverted, which solves the common mixed pixel problem when using public and low-cost satellite data to invert vegetation chlorophyll content in large-scale areas, and improves the prediction accuracy of chlorophyll content.

[0022] The advantages of the present invention and additional advantages will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention but do not constitute a limitation of the present invention.

[0024] Figure 1 is a flow chart of the inversion method of embodiment 1 of the present invention;

[0025] Figure 2 is a flow chart of the construction process of the mixed pixel spectral database B of Example 1 of the present invention;

[0026] Figure 3 is a schematic diagram of an exemplary NAOC spectral index of Embodiment 1 of the present invention;

[0027] Figure 4 It is an inversion image of NAOC and measured chlorophyll value in Example 1 of the present invention. DETAILED DESCRIPTION

[0028] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0029] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0030] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments of the present invention and the features in the embodiments can be combined with each other. The embodiments will be described in detail below in conjunction with the accompanying drawings.

[0031] Example 1

[0032] In the technical solutions disclosed in one or more embodiments, Figures 1 to 4 As shown in FIG. 1 , a chlorophyll content inversion method based on an improved random linear mixed model includes the following steps:

[0033] Step 1: Obtain UAV hyperspectral data for pixel screening, combine the spectral data based on the improved random linear mixed model, and obtain a mixed pixel spectral database B of pure pixels of the target crop under different backgrounds;

[0034] Step 2: perform similarity matching between the acquired satellite multispectral data and the spectrum of the mixed pixel spectrum database B to obtain continuous hyperspectral data with the highest similarity to the pixel composition in the satellite multispectral data;

[0035] Step 3: The obtained continuous hyperspectral data is reversely calculated based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area;

[0036] Step 4: Calculate the vegetation NAOC index of the target crop based on the pure pixel reflectance of the target crop, and perform linear regression with the measured chlorophyll value to obtain the best inversion result of the chlorophyll content of the target crop.

[0037] In this embodiment, by making full use of the advantages of the hyperspectral data of drones, a mixed hyperspectral database under different planting conditions of target crops is constructed. The chlorophyll content of the target crops is inverted by matching the acquired satellite data. This solves the common problem of mixed pixels when using public and low-cost satellite data to invert the chlorophyll content of vegetation in large-scale areas, and accurately improves the prediction accuracy of chlorophyll content.

[0038] It should be noted that the target of chlorophyll inversion in this embodiment can be any crop, such as corn, wheat, soybeans and other crops. By inverting and measuring the chlorophyll of crops, the growth of crops can be evaluated and the yield of crops can be predicted. This embodiment is mainly explained using corn as an example.

[0039] In step 1, the UAV hyperspectral image taken by the UAV is obtained, and pixels are screened according to the NDVI value (Normalized Difference Vegetation Index), and combined with the known pure soil pixel spectra in the existing database to obtain a mixed pixel spectrum database B containing pure pixels of the object to be identified;

[0040] Specifically, in this embodiment, the UAV hyperspectral images of the study area for 5 months (July, August, and September 2022, and August and September 2023) are screened by NDVI values ​​to obtain a hyperspectral database A consisting of pure corn pixels and pure soybean pixels.

[0041] In this embodiment, the object to be identified is corn, and the hyperspectral database A includes pure pixels of corn;

[0042] Furthermore, an improved random linear mixing model is proposed, which randomly linearly mixes the hyperspectral database A with different soil spectra from the ENVI database and soybean pure pixel spectra extracted from drone data in different proportions (such as 40%-100%). Thus, the spectral reflectance characteristics of corn under various soil and soybean proportions are simulated. Through the above method, 56,265 simulated corn hyperspectral data are generated as the hyperspectral database B for further analysis and research.

[0043] ENVI (Environment for Visualizing Images) is a remote sensing image processing software used to analyze geospatial data, especially satellite images and aerial images. It is mainly used in environmental monitoring, geographic information systems and scientific research. ENVI database usually refers to various geographic and image data sets stored in or with ENVI software.

[0044] Specifically, corn is used as the object to be identified, such as Figure 2 As shown in Figure 1, the construction process of the mixed pixel spectral database B is as follows:

[0045] Step 11: Filter pixels of the acquired UAV hyperspectral image according to the NDVI value to obtain a hyperspectral database A composed of mixed pixels. For example, for corn, a hyperspectral database A composed of pure corn pixels and pure soybean pixels can be obtained.

[0046] The band range of the hyperspectral instrument carried by the drone is 400-1000nm. After preprocessing such as geometric correction and atmospheric correction, the acquired data is collected in the study area for 5 months (July, August, September 2022, August, September 2023). Combined with the distribution of corn field surveys, the classification threshold NDVI (Normalized Difference Vegetation Index) value of different landforms is determined to accurately obtain the distribution of corn and soybean pixels. The NDVI formula is:

[0047] (1)

[0048] Among them, the value range of NDVI is [-1,1], -1 indicates high visible light reflectance; 0 indicates rocks or bare soil, etc.; positive values ​​indicate vegetation coverage, and it increases with the increase of coverage. By setting the NDVI threshold conditions as shown in Table 1, the pure pixels of corn and soybeans in each month are extracted based on the threshold to construct a hyperspectral database A.

[0049] Table 1 Thresholds for extracting pure pixels based on different acquisition times of UAV data

[0050]

[0051] The soil spectrum file used in this embodiment comes from the ASTE spectrum library (version 2) provided by the ENVI spectrum library. According to the main soil types in the study area, the following nine soil spectra were selected: brown fine sandy loam, brown fine sandy loam, brown to dark brown gravelly fine sandy loam, brown to dark brown sandy loam, dark brown fine sandy loam, dark yellow-brown micro-loam, gray silty clay, brown to dark brown gravelly loam, and light brown dry silty clay.

[0052] Based on the optimized random linear mixed model and the UAV hyperspectral data A, a mixed pixel spectral database B is created; it is necessary to combine the spectral samples of the hyperspectral database A and the soil spectral library. If all spectral samples are used, all possible combinations will result in a data set that is too large. This embodiment uses a random sampling method to create a mixed pixel spectral database B, and the specific description is as follows;

[0053] Step 12, grouping the spectral samples of the hyperspectral database A and the soil spectral library;

[0054] The spectral library samples were first divided into two groups: crops and soils, and the crop group was further subdivided into corn and soybean;

[0055] Among them, crop samples are divided into five categories according to the collection month, and soil samples are divided into nine categories according to the soil type; each pixel is regarded as a pure crop target or a mixed target with different soil backgrounds. Therefore, the mixed spectral data can be synthesized through the subsequent steps:

[0056] Step 13: Based on the grouping results and the NDVI thresholds of the drone hyperspectral data at different periods, a pure target pool of crops and a soil spectral background pool are constructed;

[0057] In this embodiment, the pure target pool of crops includes a pure target pool of pure corn pixels and a pure target pool of pure soybean pixels;

[0058] Specifically, according to the NDVI threshold of each period of the UAV hyperspectral data, 100 corn pixels and 20 soybean pixels were extracted each month to form a pure target pool of 500 pure corn pixels and 100 pure soybean pixels. Considering the intercropping of corn and soybean, the pixels in these target pools were used to construct the crop part of the linear mixed spectrum.

[0059] The soil spectral background pool consisted of nine soil types in the study area and was used to generate different soil components in the linear mixed spectra;

[0060] Step 14: Randomly extract a crop spectrum data from the pure target pool of the crop, and linearly combine it with each soil spectrum in the soil spectrum background pool in turn to obtain a mixed pixel and calculate the corresponding reflectance , construct mixed pixel spectral database B;

[0061] Since there are significant differences in corn and soybean endmembers between different growth stages, it is necessary to randomly select endmembers from the target pool for combination to evaluate all possible crop compositions. During each sampling process, at most one spectrum in the corn and soybean categories is extracted from the target pool at each stage, and at least one crop spectrum is obtained to exclude the bare soil situation; the extracted crop spectrum is then linearly combined with each soil spectrum in the background pool in sequence to form a mixed pixel. Previous studies have shown that in areas with vegetation coverage below 39%, the estimated chlorophyll values ​​are usually lower than the actual measured values, and the estimation accuracy is low.

[0062] Furthermore, in this embodiment, the abundance of crop spectra in the spectral library is set to 0.4 to 1, with a step size of 0.1.

[0063] The traditional linear model relies on crop hyperspectral data collected in the laboratory for simulation, which is an ideal state. UAV data is closer to satellite data. In addition, the traditional linear model only considers the impact of soil background and ignores intercropping. Intercropping, such as corn and soybeans, grows similarly during the growth period, which will affect the satellite monitoring LCC results.

[0064] Traditional mixing models are mainly used to calculate the abundance and reflectance of spectral data for numerical calculations. In this embodiment, an improved random linear mixing model is innovatively used to establish a mixed pixel spectral database B. During the database construction process, parameters such as crop abundance are changed in real time, thereby achieving data diversification of the mixed pixel spectral database.

[0065] Furthermore, the formula of the improved random linear mixed model is as follows:

[0066] (2)

[0067] in, is the reflectance of the linearly blended pixel, It is The abundance of the soil background (the proportion of pixels), Corresponding to soil reflectivity, It is The abundance of corn components, represents the reflectance of the corn spectrum selected from different target pools. Correspondingly, is the abundance of soybean ingredients, Represents the reflectance of each selected soybean spectrum.

[0068] f(t) represents the coefficient obtained by changing the spectral reflectance with time t; is the amplitude, is the set value; T is the change period of reflectivity;

[0069] This embodiment adds a time function f(t) to the random linear mixed model, considers the effect of time on reflectivity, and improves the recognition accuracy;

[0070] The spectral reflectance of the ground object changes with time. For example, the reflectance of plants changes with the growth cycle or seasonal changes. f(t) can reflect these natural cycles, such as the growth period, withering period or other dynamic changes of the ground objects. The time function f(t) is introduced into the model to capture and simulate the change characteristics related to these times. The assignment of the time function used in the model needs to be set according to the specific ground object reflectance characteristics and the law of time change. The amplitude µ represents the amplitude of the change in spectral reflectance, that is, the maximum range of the reflectance change. µs is the amplitude change of soil reflectance, which is small and takes 0.02. The amplitude change of the reflectance of the target crop µm and intercropping crop µb during the growth period is large and takes 0.3. T is the reflectance change period. The soil reflectance change period Ts takes the longest growth period of the target crop and intercropping crop. In this embodiment, 130 days is taken for corn; the reflectance change period Tb of soybeans is 120 days, and the reflectance change period Tm of corn is also 130 days. The day when the spectral reflectance of the ground object reaches its maximum value in a year can be determined by data collected by satellite. For example, if the spectral reflectance of the corresponding 5-period drone data reaches its maximum value on July 2, 2022, then .

[0071] Improved random linear mixed model by constraints: It can exclude the hyperspectral data of bare soil, thus reducing the amount of data processing in the later stage and improving the recognition efficiency. At the same time, by introducing the time function f(t), it can reflect the change of the spectral reflectance of the ground object over time, so that the model can not only explain the instantaneous performance of the spectral reflectance, but also simulate the evolution process of the ground object on a larger time scale.

[0072] For the calculation items of the target crop and its intercrops, multiple reflectance data are sampled and summed up and then averaged to improve the accuracy of the calculation; specifically, in this embodiment, the number of subgroups in the target pool Set to 1 to 5, with a step size of 1, the total number of combinations is: calculate, is the number of selected soybean subgroups. Combined with the background pool, the final generated corn hyperspectral data contains 56,265 hyperspectral data.

[0073] Among them, the subgroups in the target pool are obtained by dividing the target pool into multiple subgroups. For example, the corn target pool is obtained by five phases of imaging. That is 5;

[0074] In step 2, the acquired satellite multispectral data is matched with the spectrum of the mixed pixel spectrum database B for spectral similarity, so as to obtain continuous hyperspectral data with the highest similarity to the pixels covering the field measured points in the satellite multispectral data;

[0075] In this embodiment, the hyperspectral data is the spectral data acquired by the drone; the spectral data acquired by the satellite is multispectral data;

[0076] The hyperspectral data of the mixed pixel spectral database B is mixed pixel data obtained by linearly combining one or two crop spectra with soil spectral backgrounds in different proportions; that is, the proportion of pure pixels of the target crop to be identified in each mixed pixel data is known; thus, the proportion of pure pixels of the target crop to be identified in the satellite spectral data can be determined by similarity matching;

[0077] In some embodiments, the spectral similarity matching may utilize cosine similarity or a dynamic time warping (DTW) method.

[0078] Specifically, the Dynamic Time Warping (DTW) method is used to first align time series or waveform data (such as spectral curves) to calculate the similarity of spectral data. DTW can handle the problem of matching asynchronous spectral curves.

[0079] Specifically, cosine similarity is used for spectral matching, which is measured by the cosine of the angle between two vectors to determine whether the two vectors point in roughly the same direction. The formula is as follows:

[0080] (3)

[0081] Where ||x|| is the Euclidean norm of vector x, defined as ;

[0082] Using the cosine similarity method, the satellite matches the spectrum with the greatest similarity from the spectral library, counts the proportion of crops in the pixels represented by each mixed spectrum, and compares it with the actual crop proportion. The comparison results show that the accuracy of the cosine similarity reaches 80%, and the data with the maximum similarity above 0.98 reaches 84%.

[0083] Since satellite data is multispectral data, most of which are mixed pixels, it is impossible to obtain accurate target crop distribution information. In step 2 of this embodiment, by matching the satellite's multispectral data with the hyperspectral data obtained by a drone of a known proportion, the drone hyperspectral data with the highest similarity to the satellite multispectral data pixels can be found, that is, the proportion of pure pixels of the target crop in the satellite multispectral data and the drone hyperspectral data with the highest similarity is basically the same; since the drone hyperspectral data is simple in structure, the reflectance of the pure pixel of the target crop can be obtained by reverse calculation based on the drone hyperspectral data in step 3, that is, the reflectance of the pure pixel of the target crop corresponding to the pixel in the satellite shooting area, and the chlorophyll content of the target crop can be further determined;

[0084] In this embodiment, reverse calculation is performed based on the hyperspectral data in the mixed pixel spectral database B constructed by the UAV. Since the mixed pixel combination composition in the mixed pixel spectral database B is known, the pure pixel spectral information of the target crop can be obtained, which greatly improves the accuracy of the calculation.

[0085] In this embodiment, the hyperspectral data of the UAV is used to calculate the content of the target crop in the satellite multispectral data, which reduces the amount and difficulty of calculation and improves the accuracy of the calculation.

[0086] Step 3: The obtained continuous hyperspectral data is reversely calculated based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area;

[0087] The improved random linear mixed model in this step is the above formula (2), where the reflectance of the mixed pixel of each continuous hyperspectral data is known, the pixel composition is known, and the abundance of each pixel ( and ) is known, so the reflectance of the target crop can be calculated;

[0088] In step 4, the chlorophyll content is obtained by inversion;

[0089] The Normalized Area Over Reflectance Curve (NAOC) is an overall spectral index that includes multiple spectral bands. The calculation of the NAOC index is as follows:

[0090] (4)

[0091] in, ρ is the reflectivity, λ is the wavelength, is the maximum far-infrared reflectivity, and the wavelengths "a" and "b" correspond to wavelengths of 643nm and 795nm respectively. Equal to the reflectivity at wavelength b. Represents the area of ​​a rectangular part, such as Figure 3 As shown, the integral of the curve (light-colored area) Divide by the area of ​​the rectangular part and subtract from 1 to get the area under the curve value normalized to 1 (dark part);

[0092] The process of calculating the NAOC index for the matched hyperspectral data and performing linear regression inversion between the obtained vegetation index and the chlorophyll value measured in the field includes the following steps:

[0093] Step 41, according to the obtained pure pixel reflectance of the target crop in the satellite pixel area, the NAOC index value is calculated according to the NAOC formula, and the calculation formula is formula (4);

[0094] Step 42: linearly fit the obtained NAOC index value to the chlorophyll content (abbreviated as LCC) of the field measured point, and obtain the linear fitting relationship between LCC and NAOC.

[0095] Taking corn as an example, calculate the fitting R 2 The result is 0.782, indicating that the random mixture model optimized by combined UAV hyperspectral data is more accurate than the publicly available MuSyQ maize LCC Sentinel-2 satellite inversion product (R 2 =0.40) has been greatly improved, which verifies the feasibility of the method.

[0096] The NAOC index is calculated for the matched hyperspectral data, and the obtained vegetation index is linearly regressed with the chlorophyll value measured in the field. The training set results are as follows: Figure 4 Although there is a certain degree of dispersion, the correlation coefficient is still high compared with the satellite inversion model and product data of the same period (R 2 =0.782). It can be seen that the present embodiment can effectively improve the inversion accuracy of crop chlorophyll content by combining UAV and satellite data.

[0097] Through experiments, the results show that the correlation coefficient between the crop NAOC value and chlorophyll after accurate vegetation information extraction is very high, R 2is 0.782. Therefore, this embodiment combines the UAV hyperspectral data of the research area, targets common corn planting methods (monoculture, soybean intercropping), and creates a corn hyperspectral database B with different backgrounds based on the proposed optimized random mixing model. The accuracy of crop chlorophyll content inversion from satellite remote sensing data is improved by the spectral similarity matching method. This solves the problems of few satellite bands, low resolution, and intercropping faced in crop chlorophyll content monitoring, and provides a relatively reliable and easy-to-promote method for low-cost, large-area, and high-precision satellite remote sensing data chlorophyll content inversion.

[0098] The solution proposed in this embodiment provides a new technical means for satellite large-scale, high-precision leaf chlorophyll content inversion. The technology is also replicable and extensible to other crop types with complex planting structures, providing strong technical support for farmland ecosystems to respond to global changes, sustainable development, and implement a new round of 100 billion jin of grain production capacity increase.

[0099] Example 2

[0100] Based on Example 1, this embodiment provides a chlorophyll content inversion system based on an improved random linear mixed model, including:

[0101] Mixed pixel spectral database construction module: It is configured to obtain the UAV hyperspectral data for pixel screening, combine the spectral data based on the improved random linear mixed model, and obtain the mixed pixel spectral database of the pure pixels of the target crop under different backgrounds;

[0102] Matching module: configured to perform similarity matching between the acquired satellite multispectral data and the spectrum of the mixed pixel spectrum database, and obtain continuous hyperspectral data with the highest similarity to the pixel composition in the satellite multispectral data;

[0103] Vegetation spectral data calculation module: configured to reversely calculate the obtained continuous hyperspectral data based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area;

[0104] Inversion module: It is configured to calculate the vegetation NAOC index of the target crop based on the pure pixel reflectance of the target crop, and perform linear regression with the measured chlorophyll value to obtain the best inversion result of the chlorophyll content of the target crop.

[0105] It should be noted here that the various modules in this embodiment correspond one-to-one to the various steps in Example 1, and the specific implementation process is the same, which will not be repeated here.

[0106] Example 3

[0107] This embodiment provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps in the chlorophyll content inversion method based on the improved random linear mixed model of Embodiment 1 are completed.

[0108] Example 4

[0109] This embodiment provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps in the chlorophyll content inversion method based on the improved random linear mixed model of Embodiment 1 are completed.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. The chlorophyll content inversion method based on the improved random linear mixed model is characterized by: The steps include: The UAV hyperspectral data is obtained for pixel screening, and the spectral data is combined based on the improved random linear mixed model to obtain a mixed pixel spectral database of pure pixels of the target crop under different backgrounds; The formula of the improved random linear mixed model is as follows: in, is the reflectance of the mixed pixel, It is The abundance of the soil background, Corresponding to soil reflectivity, It is The abundance of corn components, represents the reflectance of maize spectra selected from different target pools; is the abundance of soybean ingredients, represents the reflectance of each selected soybean spectrum; is the number of subgroups in the target pool; is the number of selected soybean subgroups; f(t) represents the coefficient obtained by changing the spectral reflectance with time t; is the amplitude, is the set value; T is the change period of reflectivity, Indicates the day when the spectral reflectance of the object reaches its maximum value in a year; The acquired satellite multispectral data is matched with the spectrum of the mixed pixel spectrum database to obtain continuous hyperspectral data with the highest similarity to the pixel composition in the satellite multispectral data; The obtained continuous hyperspectral data are reversely calculated based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area; The vegetation NAOC index of the target crop was calculated based on the pure pixel reflectance of the target crop, and a linear regression was performed with the measured chlorophyll value to obtain the optimal inversion result of the chlorophyll content of the target crop.

2. The chlorophyll content inversion method based on the improved random linear mixed model according to claim 1, characterized in that: The method for constructing a mixed pixel spectral database comprises the following steps: The acquired UAV hyperspectral images are screened by pixels according to NDVI values ​​to obtain a hyperspectral database A consisting of mixed pixels; Group the spectral samples of the hyperspectral database A and the soil spectral library; Based on the grouping results and the NDVI thresholds of the drone hyperspectral data at different periods, a pure target pool of crops and a soil spectral background pool were constructed; A type of crop spectral data is randomly extracted from the pure target pool of crops, and linearly combined with each soil spectrum in the soil spectral background pool using an improved random linear mixed model to form a mixed pixel, and then a mixed pixel spectral database is constructed.

3. The chlorophyll content inversion method based on the improved random linear mixed model according to claim 1, characterized in that: The spectral similarity matching adopts cosine similarity.

4. The chlorophyll content inversion method based on the improved random linear mixed model according to claim 1, characterized in that: Spectral similarity matching uses dynamic time warping method.

5. The chlorophyll content inversion method based on the improved random linear mixed model according to claim 1, characterized in that: The process of linear regression inversion of chlorophyll values ​​includes the following steps: According to the matched pure pixel hyperspectral curve of the target crop, the NAOC index value is calculated; The obtained NAOC index value was linearly fitted with the chlorophyll content of the field measured points to obtain the linear fitting formula of chlorophyll content and NAOC index value.

6. The chlorophyll content inversion system based on the improved random linear mixed model is characterized by: include: Mixed pixel spectral database construction module: It is configured to obtain the UAV hyperspectral data for pixel screening, combine the spectral data based on the improved random linear mixed model, and obtain the mixed pixel spectral database of the pure pixels of the target crop under different backgrounds; The formula of the improved random linear mixed model is as follows: in, is the reflectance of the mixed pixel, It is The abundance of the soil background, Corresponding to soil reflectivity, It is The abundance of corn components, represents the reflectance of maize spectra selected from different target pools; is the abundance of soybean ingredients, represents the reflectance of each selected soybean spectrum; is the number of subgroups in the target pool; is the number of selected soybean subgroups; f(t) represents the coefficient obtained by changing the spectral reflectance with time t; is the amplitude, is the set value; T is the change period of reflectivity, Indicates the day when the spectral reflectance of the object reaches its maximum value in a year; Matching module: configured to perform similarity matching between the acquired satellite multispectral data and the spectrum of the mixed pixel spectrum database, and obtain continuous hyperspectral data with the highest similarity to the pixel composition in the satellite multispectral data; Vegetation spectral data calculation module: configured to reversely calculate the obtained continuous hyperspectral data based on the improved random linear mixed model to obtain the pure pixel reflectance of the target crop in the satellite pixel area; Inversion module: It is configured to calculate the vegetation NAOC index of the target crop based on the pure pixel reflectance of the target crop, and perform linear regression with the measured chlorophyll value to obtain the best inversion result of the chlorophyll content of the target crop.

7. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps in the chlorophyll content inversion method based on the improved random linear mixed model described in any one of claims 1 to 5 are completed.

8. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps in the chlorophyll content inversion method based on an improved random linear mixed model as described in any one of claims 1 to 5.

Images