CN111398562B - Farmland soil organic carbon content data acquisition method based on TM image assistance - Google Patents

Farmland soil organic carbon content data acquisition method based on TM image assistance Download PDF

Info

Publication number
CN111398562B
CN111398562B CN202010169862.2A CN202010169862A CN111398562B CN 111398562 B CN111398562 B CN 111398562B CN 202010169862 A CN202010169862 A CN 202010169862A CN 111398562 B CN111398562 B CN 111398562B
Authority
CN
China
Prior art keywords
organic carbon
carbon density
target area
soil
soil organic
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010169862.2A
Other languages
Chinese (zh)
Other versions
CN111398562A (en
Inventor
赵占辉
张宏敏
鲁春阳
李军杰
李会杰
杨锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Henan University of Urban Construction
Original Assignee
Henan University of Urban Construction
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Henan University of Urban Construction filed Critical Henan University of Urban Construction
Priority to CN202010169862.2A priority Critical patent/CN111398562B/en
Publication of CN111398562A publication Critical patent/CN111398562A/en
Application granted granted Critical
Publication of CN111398562B publication Critical patent/CN111398562B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/24—Earth materials
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/178—Methods for obtaining spatial resolution of the property being measured
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1789—Time resolved
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1793—Remote sensing

Landscapes

  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Pathology (AREA)
  • Engineering & Computer Science (AREA)
  • Environmental & Geological Engineering (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • Geology (AREA)
  • Remote Sensing (AREA)
  • Food Science & Technology (AREA)
  • Medicinal Chemistry (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

The invention relates to a farmland soil organic carbon content data acquisition method based on TM image assistance, which adopts a brand-new design method, and can acquire more sampling point data after interpolation and resampling analysis by adopting a small amount of sample data, namely quickly acquire soil organic carbon density data with high area resolution, wherein the acquired data has the advantage of high reliability; in conclusion, the design method greatly reduces the time and economic cost required by data acquisition, can provide data support for the analysis of big data of digital agriculture, and fills up the technical blank that the data acquisition is difficult at present.

Description

Farmland soil organic carbon content data acquisition method based on TM image assistance
Technical Field
The invention relates to a farmland soil organic carbon content data acquisition method based on TM image assistance, and belongs to the technical field of digital agricultural soil quality monitoring.
Background
Because the soil morphology and the evolution process are very complex, quantitative description of the soil morphology and the soil property is difficult to carry out, and the quantitative description of spatial variation characteristics, spatial correlation and the like is more complex, so that how to quickly acquire the required key information of the soil becomes a hot problem for researchers.
In the field of soil monitoring, the current large-scale sampling and investigation process is very complicated, a sample needs to be collected at a specific position of a sampling point and is brought back to a laboratory for further analysis, a large amount of manpower and time are consumed, and the manpower and time costs are greatly increased along with the increase of the investigation range and the number of sampling points, so that the large-scale soil quality investigation work is usually carried out once every 10 years or more than 10 years, 1-3 years are required for completing the large-scale investigation, the data acquisition efficiency is very low, has strong dependence on sample sampling and laboratory analysis, has the defects of time and labor waste and high cost, limits the practicability due to low working efficiency, is difficult to meet the modern agricultural requirement of rapid development, particularly, with the development of digital agriculture, the capability of acquiring agricultural big data is seriously insufficient, and the research and development of a technology capable of rapidly acquiring agricultural soil attribute data are urgently needed.
With the rise of remote sensing technology, powerful data support is provided for soil general survey and dynamic monitoring, and although the remote sensing data cannot directly detect the spatial distribution of soil physicochemical properties (mechanical composition, humidity, organic carbon content and the like) and only can acquire objective radiation information on the soil surface, the spatial and temporal differentiation characteristics on the soil organic carbon can be inverted by means of corresponding index conversion simulation. In addition, the remote sensing data has the advantages of high space-time resolution, high efficiency of obtaining mutual data and short period, and can effectively supplement the defect of insufficient space-time resolution of the traditional direct inventory method. Therefore, the direct survey and estimation method can acquire first-hand soil information data, but the space-time resolution is limited, and the current high space-time resolution data requirement is difficult to meet. According to the method for indirectly estimating the organic carbon change of the soil based on the carbon cycle process model, the simulation precision is limited by the parameter complexity, and the estimation precision of the organic carbon of the simulated soil is further influenced. If the remote sensing high space-time resolution characteristics can be combined with the simulation of the carbon cycle process of the ecosystem and the actually measured data, the accuracy of the estimated data can be ensured, and the estimation result of the soil organic carbon has higher space-time precision. The method effectively integrates multi-source data, and comprehensively evaluates the change rule of the organic carbon in the soil on a time-space scale, and becomes a hot point of current academic research.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a farmland soil organic carbon content data acquisition method based on TM image assistance, which can quickly acquire soil organic carbon density data with high areal resolution and improve the working efficiency.
The invention adopts the following technical scheme for solving the technical problems: the invention designs a farmland soil organic carbon content data acquisition method based on TM image assistance, which is used for acquiring soil organic carbon density of each position in a target area, and comprises the following steps:
step A, respectively aiming at each sample point with preset quantity and preset distribution in a target area, obtaining the soil organic carbon density of the sample point position, and then entering step B;
b, grading the soil organic carbon density value range according to preset intervals to obtain all organic carbon density grade intervals covering the soil organic carbon density value range, respectively corresponding all the organic carbon density grade intervals to preset different image gray values one by one, and then entering the step C;
step C, obtaining a TM remote sensing image of the target area, further obtaining DN values of positions of sample points in the target area, and then entering step D;
step D, aiming at each sample point position in the target area, training to obtain a target area soil organic carbon density TM remote sensing image inversion model by taking the sample point position DN value as an independent variable and the sample point position soil organic carbon density as a dependent variable, and then entering the step E;
e, applying a target area soil organic carbon density TM remote sensing image inversion model, carrying out operation processing on the TM remote sensing image of the target area, respectively corresponding the organic carbon density grade intervals to preset different image gray values one by one in combination, obtaining a target area soil organic carbon density TM remote sensing image inversion gray map, and then entering the step F;
f, setting a preset number of sampling points aiming at cultivated land soil area distribution in the target area, wherein the number of the sampling points is larger than that of the sample points, the overall distribution area of each sampling point covers all the target area to form a sampling point pattern layer of the target area, and then entering the step G;
g, superposing an inversion gray-scale map of the TM remote sensing image of the soil organic carbon density of the target area and a sampling point layer of the target area to obtain the soil organic carbon density of each sampling point position in the sampling point layer of the target area, and entering the step H;
and H, carrying out soil organic carbon density interpolation processing on the target area according to the soil organic carbon density of each sampling point position in the target area, namely obtaining the soil organic carbon density of each position in the target area.
As a preferred technical scheme of the invention: in the step A, respectively aiming at each sample point, a circular area is constructed by taking the position of the sample point as the center of a circle and taking the preset distance as the radius, the positions of the sample points with the preset number are randomly selected in the circular area, the soil organic carbon density of each sample point position is respectively obtained, and the average value of the soil organic carbon density of each sample point position is taken as the soil organic carbon density of the sample point position; and further obtaining the soil organic carbon density of each sample point position in the target area.
As a preferred technical scheme of the invention: and collecting the soil organic carbon density between the soil surface of each position in the target area and the depth of 20cm below the soil surface as the soil organic carbon density of each position in the target area.
As a preferred technical scheme of the invention: and in the step B, grading the soil organic carbon density value range according to a preset interval of 1t C/ha to obtain each organic carbon density grade interval covering the soil organic carbon density value range.
As a preferred technical scheme of the invention: in the step G, superposing an inversion gray map of the TM remote sensing image of the soil organic carbon density in the target area and a sampling point map layer of the target area, firstly, obtaining the gray value of each sampling point position in the sampling point map layer of the target area, then, respectively presetting the one-to-one correspondence relationship between different image gray values according to each organic carbon density level interval, and if the gray value of the sampling point position is equal to any one of the image gray values respectively corresponding to each organic carbon density level interval, selecting the average value between the minimum value and the maximum value in the organic carbon density level intervals corresponding to the corresponding image gray value as the organic carbon density of the sampling point position; if the gray value of the sampling point position is equal to the transition value between the gray values of the images respectively corresponding to the two organic carbon density grade intervals, selecting the average value between the minimum value and the maximum value in the corresponding large organic carbon density grade interval as the organic carbon density of the sampling point position; and thus, the organic carbon density of the soil at each sampling point position in the sampling point layer of the target area is obtained.
As a preferred technical scheme of the invention: and the linear distance between the sampling point and the non-cultivated land feature space around the sampling point is not less than a preset distance threshold.
Compared with the prior art, the farmland soil organic carbon content data acquisition method based on TM image assistance has the following technical effects by adopting the technical scheme:
according to the farmland soil organic carbon content data acquisition method based on TM image assistance, a brand new design method is adopted, interpolation and resampling analysis are carried out on a small amount of sample data, more sampling point data can be obtained, namely soil organic carbon density data with high area resolution can be quickly obtained, the obtained data has the advantage of high reliability, the contradiction between cost and the number of sample points can be solved, namely high-density sample point data can be obtained through TM image assistance by using a small amount of sample data, the cost is reduced, meanwhile, the sufficient number of sample point data can be obtained, the spatial resolution can be quickly and greatly improved, and the work efficiency of data acquisition is greatly improved; in conclusion, the design method greatly reduces the time and economic cost required by data acquisition, can provide data support for the analysis of big data of digital agriculture, and fills up the technical blank that the data acquisition is difficult at present.
Drawings
FIG. 1 is a schematic flow chart of a method for obtaining organic carbon content data of farmland soil based on TM image assistance in the design of the invention;
FIG. 2 is a sample point spatial distribution characteristic in an embodiment of the present invention;
FIG. 3 is a schematic diagram of Principal Component Analysis (PCA) lithotripsy of TM images in an embodiment of the present invention;
FIG. 4 is a gray scale map of inversion of soil organic carbon density TM remote sensing images in an embodiment of the present invention;
FIG. 5 is a graph showing the spatial distribution of sampling points in an embodiment of the present invention;
FIG. 6 is a schematic diagram showing the distribution of sampling points of an inversion graph of soil organic carbon density based on TM remote sensing images in the design and application embodiment of the present invention;
FIG. 7 is a schematic diagram of the spatial distribution of the organic carbon density in the farmland soil in an embodiment of the invention;
FIG. 8 is a schematic diagram showing the spatial distribution of organic carbon density in soil of county farmland based on survey data according to an embodiment of the present invention;
FIG. 9 is a schematic diagram of the model estimation result and the actually measured 1:1 line fitting result in the design application embodiment of the present invention.
Detailed Description
The following description will explain embodiments of the present invention in further detail with reference to the accompanying drawings.
The invention designs a farmland soil organic carbon content data acquisition method based on TM image assistance, which is used for acquiring soil organic carbon density of each position in a target area, and specifically executes the following steps A to H as shown in figure 1.
And step A, respectively aiming at each sample point with preset quantity and preset distribution in the target area, obtaining the soil organic carbon density of the sample point position, and then entering the step B.
Because the soil type, the cultivation system, the land utilization type and the climate characteristic in the range of 100 meters of the peripheral diameter of the sample point are consistent with the attribute information of the sample point, the sample point can represent the fertility characteristic of the peripheral soil; therefore, in practical application, in the step a, for each sample point in the target area, which is in the preset number and in the preset distribution, a circular area is respectively constructed for each sample point, with the position of the sample point as the center of a circle and the radius of 100 meters, the positions of the sample points in the preset number are randomly selected in the circular area, the soil organic carbon density of each sample point position is respectively obtained, and the average value of the soil organic carbon density of each sample point position is used as the soil organic carbon density of the sample point position; and further obtaining the soil organic carbon density of each sample point position in the target area.
For the collection of the soil organic carbon density, in practical application, the soil organic carbon density between the soil surface and the depth of 20cm below the soil surface at each position in the target area is collected as the soil organic carbon density at each position in the target area.
Step B, carrying out grade division according to preset intervals aiming at the soil organic carbon density value range to obtain each organic carbon density grade interval covering the soil organic carbon density value range, wherein in practical application, the grade division can be designed according to the preset interval of 1t C/ha to obtain each organic carbon density grade interval covering the soil organic carbon density value range; and respectively corresponding the organic carbon density grade intervals to preset different image gray values one by one, and then entering the step C.
In practical applications, for example, the density grade interval of each organic carbon is obtained as follows:
0～1t C/ha，1～2t C/ha，2～3t C/ha，3～4t C/ha，4～5t C/ha，5～6t C/ha，6～7t C/ha，7～8t C/ha，8～9t C/ha，9～10t C/ha，10～11t C/ha，11～12t C/ha，12～13t C/ha，13～14t C/ha，14～15t C/ha，15～16t C/ha，16～17t C/ha，17～18t C/ha，18～19t C/ha，19～20t C/ha，20～21t C/ha，21～22t C/ha，22～23t C/ha，23～24t C/ha，24～25t C/ha，25～26t C/ha，26～27t C/ha，27～28t C/ha，28～29t C/ha，29～30t C/ha，30～31t C/ha，31～32t C/ha，32～33t C/ha，33～34t C/ha，34～35t C/ha，35～36t C/ha，36～37t C/ha，37～38t C/ha，38～39t C/ha，39～40t C/ha，40～41t C/ha，41～42t C/ha，42～43t C/ha，43～44t C/ha，44～45t C/ha，45～46t C/ha，46～47t C/ha，47～48t C/ha，48～49t C/ha，49～50t C/ha。
and C, obtaining a TM remote sensing image of the target area, further obtaining DN values of positions of sample points in the target area, and then entering the step D.
And D, aiming at each sample point position in the target area, training to obtain a target area soil organic carbon density TM remote sensing image inversion model by taking the DN value of the sample point position as an independent variable and the soil organic carbon density of the sample point position as a dependent variable, and then entering the step E.
And E, applying the soil organic carbon density TM remote sensing image inversion model of the target area, carrying out operation processing on the TM remote sensing image of the target area, combining all organic carbon density grade intervals and presetting different image gray values to be in one-to-one correspondence to obtain an organic carbon density TM remote sensing image inversion gray map of the soil of the target area, and then entering the step F.
And F, setting a preset number of sampling points according to the distribution of cultivated land soil areas in the target area, wherein the linear distance between each sampling point and the non-cultivated land feature space around the sampling point is not lower than a preset distance threshold, in practical application, the linear distance between each sampling point and the non-cultivated land feature space around the sampling point is designed and limited to be not lower than 50 meters, the number of the sampling points is greater than the number of the sample points, for example, the number of the sampling points applied is designed to be 5 times of the number of the sample points, and the whole distribution area of each sampling point covers all the target area to form a sampling point pattern layer of the target area, and then entering the step G.
And G, superposing the soil organic carbon density TM remote sensing image inversion gray level map of the target area and the sampling point map layer of the target area to obtain the soil organic carbon density of each sampling point in the sampling point map layer of the target area, and then entering the step H.
In practical application, in the step G, for obtaining the organic carbon density of the soil at each sampling point position, the specific application method is as follows: superposing an inversion gray map of the TM remote sensing image of the soil organic carbon density in the target area with a sampling point map layer of the target area, firstly obtaining gray values of sampling points in the sampling point map layer of the target area, then respectively presetting one-to-one correspondence between different image gray values according to each organic carbon density grade interval, and if the gray value of the sampling point is equal to any one of the image gray values respectively corresponding to each organic carbon density grade interval, selecting an average value between a minimum value and a maximum value in the organic carbon density grade interval corresponding to the corresponding image gray value as the organic carbon density of the sampling point; if the gray value of the sampling point position is equal to the transition value between the gray values of the images respectively corresponding to the two organic carbon density grade intervals, selecting the average value between the minimum value and the maximum value in the corresponding large organic carbon density grade interval as the organic carbon density of the sampling point position; and thus, the organic carbon density of the soil at each sampling point position in the sampling point layer of the target area is obtained.
And H, carrying out soil organic carbon density interpolation processing on the target area according to the soil organic carbon density of each sampling point position in the target area, namely obtaining the soil organic carbon density of each position in the target area.
The method for acquiring the organic carbon content data of farmland soil based on TM image assistance is applied to practice, a hill county area is selected as an example, 30 field investigation sample points are collected as shown in figure 2, and TM remote sensing images shot by Landsat remote sensing satellites in the sampling time period of the area are acquired.
After preprocessing is carried out on the 1-7 wave bands of the TM remote sensing image, DN values of all wave bands at corresponding positions of the investigation sample are obtained, and the DN value statistical result is shown in the following table 1.
Figure BDA0002408800800000061
TABLE 1
Wherein, the change ranges of the DN values of the TM1-7 wave bands are respectively 17, 21, 40, 14, 71, 17 and 73, and the variances are respectively 4.26, 5.54, 10.54, 3.94, 17.21, 4.43 and 17.99. Therefore, the dispersion degree of the TM3, 5 and 7 wave bands is high, the difference of the reaction soil is more obvious, the carried information is richer, and the spatial variation of the organic carbon density of the reaction soil can be more sensitively reflected.
The problem that image information of each wave band is easy to generate multi-element collinearity cannot be thoroughly solved, and the caused direct result is that a plurality of variable information are excessively overlapped to cause the error of an analysis result to be increased, so that the spatial variability of the organic carbon in the farmland soil is difficult to accurately estimate. The multiple variables (namely, the data of the multiple wave bands) which are mutually related are concentrated into single or multiple variables which are independent of each other as less as possible by adopting a statistical method, and the result is that the multiple wave bands are replaced by the wave bands as less as possible, and simultaneously, the information as much as possible is contained for estimating the organic carbon density of the farmland soil. The specific algorithm is as follows: by means of orthogonal transformation, each wave band is used as a relevant component, each component random vector is transformed into an irrelevant new random vector, the essence is that a covariance matrix of the random vector is transformed into a diagonal matrix, an original coordinate system is transformed into a new orthogonal coordinate system geometrically, the new orthogonal coordinate system points to a plurality of orthogonal directions where sample points are most spread, then the multidimensional variable system is subjected to dimension reduction processing, the multidimensional variable system can be transformed into a low-dimensional variable system with higher precision, a proper value function is constructed, and the low-dimensional system is further transformed into a one-dimensional system. Finally, one or more variables with large variance can comprehensively reflect main information contained in multi-band variables of the multi-TM remote sensing image. The principal component analysis was performed for each band, and the analysis results are shown in table 2 below.
Figure BDA0002408800800000062
Figure BDA0002408800800000071
TABLE 2
Wherein, 7 wave bands are divided into 7 components, the table 2 shows that 7 principal components can respectively explain 74.85, 15.21, 5.86, 2.23, 0.91, 0.65 and 0.30 percent of the total variance, the 1 st principal component and the 2 nd principal component can explain 90.06 percent of the total variance, and the lithograph obtained by principal component analysis shows that the variation range of the characteristic value of the 1 st principal component and the 2 nd principal component is large, and the variation of the 3 rd principal component, the 4 th principal component, the 5 th principal component, the 6 th principal component and the 7 th principal component is not obvious, so the 2 principal components can be obtained by principal component analysis, and 90.06 percent of the total variance is explained. Since the TM1, 2, 3, 4, 5, 6, and 7 band feature vectors in the 1 st principal component are 0.94, 0.97, 0.98, -0.06, 0.94, 0.80, and 0.96, respectively, and the other 6 bands are greater than 0.8 except for the small TM4 band feature vector, the TM1, 2, 3, 5, 6, and 7 all contribute greatly to the 1 st principal component. Similarly, TM4 contributes most to the second principal component, and the other bands are smaller. In summary, 7 wave bands of the TM remote sensing image contain rich information, and the establishment of an organic carbon density estimation equation of farmland soil in a research area by adopting DN value data of the 7 wave bands of the TM image has certain rationality.
Based on the analysis results, TM remote sensingThe 7 bands of the image all contain information closely related to the organic carbon density of the soil in the research area, and the DN values of the bands are subjected to curve evaluation with the organic carbon density respectively in the embodiment, so that the optimal linear or nonlinear correlation relationship between the DN values of the bands and the organic carbon density is established. The 8 candidate model equations listed in this embodiment basically cover possible linear or nonlinear relations, specifically, linear functions, logarithmic functions, inverse functions, quadratic functions, cubic functions, composite functions, power functions and exponential functions are used as candidate models to perform curve evaluation, the TM remote sensing image DN value is used as an independent variable, the soil organic carbon density is used as a dependent variable, and according to goodness-of-fit inspection, the R value is used 2 The significance test (T test) of the statistic and regression coefficient is used as a judgment basis to explore a relation model between TM image DN value and organic carbon density, a curve evaluation curve candidate model and an equation are shown in the following table 3, wherein x is an independent variable, y is a dependent variable, and beta is 0 Is a constant number, beta 1 、β 2 、β 3 Are regression coefficients.
Figure BDA0002408800800000081
TABLE 3
The evaluation results of the single-waveband DN values and soil organic carbon density curves are shown in table 4 below, where "x" and "x" indicate that the significance levels reached p <0.05 and 0.01, respectively.
Figure BDA0002408800800000082
Figure BDA0002408800800000091
TABLE 4
After 7 wave bands of TM remote sensing image are fitted through 8 candidate model function equations, R 2 The variation ranges are respectively 0.23-0.24, 0.24-0.25, 0.29-0.31, 0.23-0.29, 0.47-0.62, 0.25-0.26 and 0.29-0.34, the significance level reaches significant or extremely significant level, and the candidate model R 2 The statistic floating range is small, and meanwhile, the relation model equation of DN values of all wave bands and soil organic carbon density is considered to be simplified as much as possible. In summary, in this embodiment, DN values of all bands of TM images in the research area have a linear correlation with the density of organic carbon in soil.
In order to further simplify the prediction equation, dimension reduction processing is performed on the 7 wave bands of the TM remote sensing image to obtain 2 principal components based on table 2, and the principal component factor load matrix is shown in table 5 below.
Figure BDA0002408800800000101
TABLE 5
Two principal component score equations can thus be derived:
principal component 1 ═ 0.94 × B 1 +0.97*B 2 +0.98*B 3 -0.06*B 4 +0.94*B 5 +0.80*B 6 +0.96*B 7 (1)
Principal component 2 ═ 0.05 × B 1 -0.09*B 2 -0.10*B 3 +0.99*B 4 +0.21*B 5 +0.12*B 6 +0.00*B 7 (2)
B 1 、B 2 、B 3 、B 4 、B 5 、B 6 、B 7 Respectively, the TM1-7 band gray scale values (DN values).
Taking the principal component 1 and the principal component 2 as independent variables, taking soil organic carbon density (SOIL) as dependent variable, and performing 2-element linear regression estimation on the independent variable and the dependent variable, wherein the regression equation is as follows:
SOCD 0.04 main component 1+0.62 main component 2-46.13 (3)
From equations (1), (2), (3) we can derive:
SOCD＝0.004*B 1 -0.018*B 2 -0.027*B 3 +0.617*B 4 +0.165*B 5 +0.106*B 6 +0.037*B 7 -46.13(4)
equation (4) is the inversion model of the TM remote sensing image of the farmland soil in the study, and the model is used to perform band operation on each band of the TM remote sensing image to obtain the inversion gray level map of the TM remote sensing image of the soil organic carbon density, as shown in fig. 4.
In addition to the original 30 sample points, the number of the sample points is increased to 150, which is equivalent to five times of the original number of the sample points, and the sample points are used as 150 sample points to form a sample point pattern layer, as shown in fig. 5. By means of an ArcGIS platform space analysis tool, superposing an inversion gray graph and a sampling point graph layer based on a soil organic carbon density TM remote sensing image, as shown in FIG. 6, extracting the 150 sampling point TM remote sensing image inversion gray values, wherein the extracted attribute values are actually remote sensing inversion surface soil organic carbon density estimation values; and finally, performing soil organic carbon density interpolation processing on the target area according to the soil organic carbon density of each sampling point position, namely obtaining the soil organic carbon density of each position in the target area, as shown in fig. 7.
In order to compare the data quality obtained by the conventional sampling method and the method, for the present embodiment, interpolation analysis is performed based on the survey data, as shown in fig. 8. The result shows that compared with the survey data interpolation result, the data spatial interpolation result obtained by the method is finer, the resolution ratio is higher, and the spatial variation characteristic is more prominent.
In conclusion, the actual measurement method can accurately obtain the spatial variation characteristics and the carbon reserves of the organic carbon density of the surface soil of the farmland, but the fineness degree of the spatial differentiation characteristics is limited by the number of sampling points, and the method has the disadvantages of low working efficiency and high cost.
The organic carbon density of farmland soil in a research area has strong spatial heterogeneity, and an estimation result obtained by a soil organic carbon density TM remote sensing image inversion model also needs to have the capability of expressing dependent variable spatial variation. Maximum value (X) of estimation result max ) Minimum value (X) min ) And extreme difference (X) max -X min ) Very close to the actual measurement results, the coefficients of variation (C.V.) of the estimated and actual measurement results are 0.17 and 0.20, respectively, which indicates that the model estimation results have certain spatial variation expression capability, as shown in table 6 below.
Estimation method Maximum value of Minimum value Extreme difference Coefficient of variation MAPE RMSR
Interpolation method for actual measurement 36.28 18.34 17.94 0.20 0.17 5.06
The method of the invention 42.65 14.59 28.06 0.20 4.90 5.87
TABLE 6
The stability and the prediction precision of the model are important for evaluating the quality of the modelThe indexes and the stability of the model can be tested by the goodness of fit of the predicted variable and the measured variable 2 Size inspection of (2), R 2 The closer to 1, the more stable the model, whereas R 2 The closer to 0, the more unstable the model, the larger the potential prediction error. The fitting degree of the model estimation result and the actual measurement result is 0.7186 as shown in FIG. 9. Therefore, the TM image inversion estimation model has high precision and certain reliability.
According to the method for acquiring the farmland soil organic carbon content data based on TM image assistance, a brand new design method is adopted, interpolation and resampling analysis are carried out on a small amount of sample data, more sampling point data can be acquired, namely soil organic carbon density data with high area resolution can be acquired quickly, the acquired data has the advantage of high reliability, the contradiction between cost and the number of sample points can be solved, namely high-density sample point data can be acquired through TM image assistance by using a small amount of sample data, the cost is reduced, meanwhile, the sufficient number of sample point data can be acquired, the spatial resolution can be improved quickly and greatly, and the work efficiency of acquiring the data is improved greatly; in conclusion, the design method greatly reduces the time and economic cost required by data acquisition, can provide data support for the analysis of big data of digital agriculture, and fills up the technical blank that the data acquisition is difficult at present.
The embodiments of the present invention have been described in detail with reference to the drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.

Claims (4)

1. A farmland soil organic carbon content data acquisition method based on TM image assistance is used for acquiring soil organic carbon density of each position in a target area, and is characterized by comprising the following steps:
step A, respectively aiming at each sample point with preset quantity and preset distribution in a target area, obtaining the soil organic carbon density of the sample point position, and then entering step B;
in the step a, respectively aiming at each sample point, a circular area is constructed by taking the position of the sample point as the center of a circle and taking the preset distance as the radius, a preset number of sample point positions are randomly selected in the circular area, the soil organic carbon density of each sample point position is respectively obtained, and the average value of the soil organic carbon density of each sample point position is taken as the soil organic carbon density of the sample point position; further obtaining the organic carbon density of soil at each sample point position in the target area;
b, carrying out grade division according to preset intervals aiming at the soil organic carbon density value range to obtain all organic carbon density grade intervals covering the soil organic carbon density value range, respectively corresponding all organic carbon density grade intervals to preset different image gray values one by one, and then entering the step C;
step C, obtaining a TM remote sensing image of the target area, further obtaining DN values of positions of sample points in the target area, and then entering step D;
step D, aiming at each sample point position in the target area, training to obtain a target area soil organic carbon density TM remote sensing image inversion model by taking the sample point position DN value as an independent variable and the sample point position soil organic carbon density as a dependent variable, and then entering the step E;
e, applying a target area soil organic carbon density TM remote sensing image inversion model, carrying out operation processing on the TM remote sensing image of the target area, combining all organic carbon density grade intervals and presetting different image gray values to be in one-to-one correspondence with each other, obtaining a target area soil organic carbon density TM remote sensing image inversion gray map, and then entering the step F;
f, setting a preset number of sampling points according to the cultivated land soil area distribution in the target area, wherein the number of the sampling points is larger than that of the sample points, the whole distribution area of each sampling point covers the whole target area to form a target area sampling point pattern layer, and then entering the step G;
g, superposing the soil organic carbon density TM remote sensing image inversion gray level map of the target area and the sampling point map layer of the target area to obtain the soil organic carbon density of each sampling point in the sampling point map layer of the target area, and entering the step H;
in the step G, superposing the target area soil organic carbon density TM remote sensing image inversion gray scale map and the target area sampling point map layer, firstly obtaining the gray scale value of each sampling point position in the target area sampling point map layer, then respectively presetting a one-to-one correspondence relationship between different image gray scale values according to each organic carbon density level interval, and if the gray scale value of the sampling point position is equal to any one of the image gray scale values respectively corresponding to each organic carbon density level interval, selecting the average value between the minimum value and the maximum value in the organic carbon density level interval corresponding to the corresponding image gray scale value as the organic carbon density of the sampling point position; if the gray value of the sampling point position is equal to the transition value between the gray values of the images corresponding to the two organic carbon density grade intervals, selecting the average value between the minimum value and the maximum value in the corresponding large organic carbon density grade intervals as the organic carbon density of the sampling point position; thus obtaining the organic carbon density of soil at each sampling point position in the sampling point layer of the target area;
and H, carrying out soil organic carbon density interpolation processing on the target area according to the soil organic carbon density of each sampling point position in the target area, namely obtaining the soil organic carbon density of each position in the target area.
2. The farmland soil organic carbon content data acquisition method based on TM image assistance as claimed in claim 1, wherein: and collecting the soil organic carbon density between the soil surface of each position in the target area and the depth of 20cm below the soil surface as the soil organic carbon density of each position in the target area.
3. The farmland soil organic carbon content data acquisition method based on TM image assistance as claimed in claim 1, wherein: and in the step B, grading the soil organic carbon density value range according to a preset interval of 1t C/ha to obtain each organic carbon density grade interval covering the soil organic carbon density value range.
4. The farmland soil organic carbon content data acquisition method based on TM image assistance as claimed in claim 1, wherein: and the linear distance between the sampling point and the non-cultivated land feature space around the sampling point is not less than a preset distance threshold.
CN202010169862.2A 2020-03-12 2020-03-12 Farmland soil organic carbon content data acquisition method based on TM image assistance Active CN111398562B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010169862.2A CN111398562B (en) 2020-03-12 2020-03-12 Farmland soil organic carbon content data acquisition method based on TM image assistance

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010169862.2A CN111398562B (en) 2020-03-12 2020-03-12 Farmland soil organic carbon content data acquisition method based on TM image assistance

Publications (2)

Publication Number Publication Date
CN111398562A CN111398562A (en) 2020-07-10
CN111398562B true CN111398562B (en) 2022-09-09

Family

ID=71434218

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010169862.2A Active CN111398562B (en) 2020-03-12 2020-03-12 Farmland soil organic carbon content data acquisition method based on TM image assistance

Country Status (1)

Country Link
CN (1) CN111398562B (en)

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107194054B (en) * 2017-05-16 2019-10-11 武汉大学 A kind of kriging analysis model building method for taking SOCD special heterogeneity into account
CN109933873B (en) * 2018-11-02 2022-07-01 中国林业科学研究院资源信息研究所 Method for remote sensing estimation of organic carbon density of soil in desertification region in high space

Also Published As

Publication number Publication date
CN111398562A (en) 2020-07-10

Similar Documents

Publication Publication Date Title
Stehman et al. Key issues in rigorous accuracy assessment of land cover products
Sheidai Karkaj et al. Establishing a suitable soil quality index for semi-arid rangeland ecosystems in northwest of Iran
CN109541172A (en) The calculation method and device of soil attribute value
Merdun Self-organizing map artificial neural network application in multidimensional soil data analysis
CN108764527B (en) Screening method for soil organic carbon library time-space dynamic prediction optimal environment variables
CN116148438B (en) Soil mineral content prediction method based on machine learning
CN116933034A (en) Method, system, electronic equipment and medium for determining carbon reserve change of ecosystem
CN115345069A (en) Lake water volume estimation method based on maximum water depth record and machine learning
CN115147721A (en) Remote sensing inversion method and device for forest accumulation
CN111488910B (en) Farmland soil organic carbon space-time data acquisition method
CN110362911A (en) A kind of agent model selection method of Design-Oriented process
CN111398562B (en) Farmland soil organic carbon content data acquisition method based on TM image assistance
CN116863341B (en) Crop classification and identification method and system based on time sequence satellite remote sensing image
Zhai et al. Wet aggregate stability predicting of soil in multiple land-uses based on support vector machine
CN115797501A (en) Time-series forest age mapping method combining forest disturbance and recovery events
CN116183868A (en) Remote sensing estimation method and system for organic carbon in soil of complex ecological system
CN113538239B (en) Interpolation method based on space-time autoregressive neural network model
CN113011086B (en) Estimation method of forest biomass based on GA-SVR algorithm
CN103955953B (en) A kind of landform towards digital soil drawing works in coordination with variable choosing method
CN111398561B (en) Farmland soil organic carbon space-time data acquisition method based on multi-source data assistance
Shareef et al. Integrating of GIS and fuzzy multi-criteria method to evaluate land degradation and their impact on the urban growth of Kirkuk city, Iraq
Habibie et al. Drought Estimation from Vegetation Phenology Analysis of Maize in Indonesia Using Deep Learning Algorithm
Atzberger et al. Portability of neural nets modelling regional winter crop acreages using AVHRR time series
Venkatesan et al. Machine learning models using paprika leaf growth forecast based on environmental and energy data
CN116778228A (en) Method for monitoring organic carbon in soil based on combination of FOD and optimal wave band

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant