CN115495988B - Soil remote sensing inversion method based on optimal time window selection - Google Patents
Soil remote sensing inversion method based on optimal time window selection Download PDFInfo
- Publication number
- CN115495988B CN115495988B CN202211189137.7A CN202211189137A CN115495988B CN 115495988 B CN115495988 B CN 115495988B CN 202211189137 A CN202211189137 A CN 202211189137A CN 115495988 B CN115495988 B CN 115495988B
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- soil
- target
- month
- value
- 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
Links
- 239000002689 soil Substances 0.000 title claims abstract description 149
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000011160 research Methods 0.000 claims abstract description 52
- 238000011156 evaluation Methods 0.000 claims description 36
- 238000005070 sampling Methods 0.000 claims description 36
- 238000012545 processing Methods 0.000 claims description 14
- 238000012937 correction Methods 0.000 claims description 12
- 238000010801 machine learning Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 10
- 230000005855 radiation Effects 0.000 claims description 9
- 238000002310 reflectometry Methods 0.000 claims description 6
- 230000008878 coupling Effects 0.000 claims description 4
- 238000010168 coupling process Methods 0.000 claims description 4
- 238000005859 coupling reaction Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 5
- 239000010902 straw Substances 0.000 abstract description 5
- 238000013461 design Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004181 pedogenesis Methods 0.000 description 2
- 239000004016 soil organic matter Substances 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Analytical Chemistry (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a soil remote sensing inversion method based on optimal time window selection, which comprises the steps of firstly, selecting optimal years to reduce possible influence of factors such as climate among different years on soil attribute remote sensing inversion accuracy; secondly, determining the optimal month jointly by combining the crop growth condition reflected by the regional vegetation index and the accuracy result of inversion of soil attributes of different months, and reducing the influence of crop and crop straw coverage; finally, determining an optimal time window for inversion of the regional soil attribute by integrating the optimal year and the optimal month, and obtaining the target soil attribute in the target research region by using a target prediction model in the optimal time window; the remote sensing image in the optimal time window is designed and utilized, so that the influence of factors such as climate, crop coverage and the like can be effectively overcome, the accuracy of the remote sensing inversion of the soil property is effectively improved, and the practical application of the remote sensing inversion technology of the soil property is promoted.
Description
Technical Field
The invention relates to a soil remote sensing inversion method based on optimal time window selection, and belongs to the technical field of soil attribute detection.
Background
The soil has multiple functions, not only has production function, but also has important ecological and environmental functions, and the protection of soil resources is important for national grain safety. Currently, there are some unreasonable links and measures in the utilization of soil resources, and effective monitoring of soil resources is a precondition for realizing sustainable utilization and protection of soil resources. However, the soil has strong space-time variability, and is comprehensively influenced by factors such as matrix, climate, topography, biology and the like in the long forming process, so that the soil has strong regional characteristics.
In recent years, digital soil mapping technology which indirectly reflects soil properties by comprehensively utilizing related parameters in the soil formation process has become an important method for regional soil property mapping. As related research is advanced, the research is gradually discovered to face a certain challenge in mechanism interpretation and application in areas with small variation of related parameters in the soil formation process. The remote sensing technology has the characteristics of strong timeliness and large coverage, and has been applied to a certain degree in soil attribute inversion and mapping in recent years.
Compared with the spectrum of green crops and crop straws, the soil has unique spectral curve characteristics, and the soil attribute information is reflected by remote sensing by utilizing the soil spectral information. The soil and the soil surface of farmland are inevitably influenced by covering factors such as crops, crop straws and the like, so that the accuracy of soil remote sensing inversion is possibly reduced; on the other hand, the soil itself state such as its moisture difference may also affect inversion accuracy. Eliminating or reducing the impact of these factors on inversion accuracy is a key element in soil remote sensing applications.
In recent years, many soil remote sensing inversion applications select remote sensing image time windows only according to regional crop growth characteristics, and the method has a certain effect on month selection, but often lacks objective basis on year selection, so that inversion accuracy has larger uncertainty. The remote sensing inversion application of the regional soil represented by image synthesis has larger prediction precision difference in different regions and also has certain challenges in the application of the region with shorter bare soil time window.
Disclosure of Invention
The invention aims to solve the technical problem of providing a soil remote sensing inversion method based on optimal time window selection, which can effectively and accurately realize rapid acquisition of regional soil properties and improve actual working efficiency.
The invention adopts the following technical scheme for solving the technical problems: the invention designs a soil remote sensing inversion method based on optimal time window selection, which comprises the following steps of A to E, obtaining a target prediction model of target soil properties of a target research area, and applying the target prediction model to obtain the target soil properties in the target research area;
step A, obtaining remote sensing images which respectively correspond to at least two months and cover a target research area, wherein each month is at least one scene remote sensing image, and a data value corresponding to a target soil attribute at each sampling position is preset in the target research area, then performing atmospheric correction processing on each scene remote sensing image, and then entering the step B;
step B, respectively aiming at each scene remote sensing image, based on the remote sensing image and a data value corresponding to the target soil attribute at each preset sampling position in a target research area, applying a preset machine learning algorithm, and training to obtain a first prediction model taking the remote sensing image value at each sampling position in the target research area as input and the data value corresponding to the target soil attribute at each sampling position as output; obtaining a first prediction model corresponding to each scene remote sensing image respectively, and then entering the step C;
step C, based on the evaluation results of the preset target evaluation indexes corresponding to the first prediction models, carrying out statistics summarization on the first prediction models corresponding to different years, determining the optimal years of soil attribute remote sensing inversion, and then entering the step D;
step D, based on the evaluation results of the preset target evaluation indexes corresponding to the first prediction models, coupling the normalized vegetation indexes NDVI, carrying out statistics summarization on the first prediction models corresponding to different months, determining the optimal month of soil attribute remote sensing inversion, and then entering the step E;
and E, taking the optimal year and the optimal month of the soil attribute remote sensing inversion as an optimal time window of the soil attribute remote sensing inversion, and training to obtain a target prediction model which takes the remote sensing image value of each sampling position in the target research area as input and the data value of each sampling position corresponding to the target soil attribute as output based on each scene remote sensing image in the target research area and the data value of each sampling position corresponding to the target soil attribute in the target research area in the optimal time window of the soil attribute remote sensing inversion by applying a preset machine learning algorithm.
As a preferred technical scheme of the invention: in the step A, the following operations are executed for each scene remote sensing image respectively, and the atmosphere correction processing update is carried out;
the operation is as follows: for each pixel of the remote sensing image, firstly performing radiation calibration processing according to L=DN, so as to obtain radiation brightness data L, and then performing atmospheric correction processing on the remote sensing image by using a FLAASH model according to the radiation brightness data L; thereby realizing the atmosphere correction processing updating of the remote sensing image; wherein DN represents the pixel value of the remote sensing image, and alpha and beta are the conversion coefficients attached to the remote sensing image respectively.
As a preferred technical scheme of the invention: in the step C, R of the coefficient evaluation index is correspondingly determined based on each first prediction model 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 And the year corresponding to the first prediction model to which the value and the minimum RMSE value belong together is used as the optimal year for the soil attribute remote sensing inversion.
As a preferred technical scheme of the invention: the step D comprises the following steps D1 to D4;
step D1, corresponding to the determination coefficient evaluation index R based on each first prediction model 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 D2, taking the month corresponding to the first prediction model to which the value and the minimum RMSE value belong together as the month to be matched, and then entering into the step D2;
step D2, aiming at each scene remote sensing image, respectively, according to the following formula:
NDVI=(NIR-Red)/(NIR+Red)
obtaining the NDVI value of each pixel in a target research area in the remote sensing image, obtaining an average value as the NDVI value corresponding to the target research area in the remote sensing image, further obtaining the NDVI value corresponding to each scene remote sensing image respectively, and then entering the step D3; wherein NIR represents the reflectivity of the near infrared band in the remote sensing image, and Red represents the reflectivity of the Red band in the remote sensing image;
step D3, respectively aiming at different months, calculating and obtaining average values of NDVI values corresponding to all the remote sensing images in the month, taking the average values as the NDVI values corresponding to the month, further obtaining the NDVI values corresponding to different months, and then entering the step D4;
step D4, if the month corresponding to the minimum NDVI value is the same month as the month to be matched, the month is used as the optimal month for the remote sensing inversion of the soil attribute; and if the month corresponding to the minimum NDVI value is different from the month to be matched, selecting the month corresponding to the minimum NDVI value and the adjacent next month thereof as the optimal month for remote sensing inversion of the soil attribute based on the closed loop sequence of the end-to-end month.
As a preferred technical scheme of the invention: the step E comprises the following steps E1 to E4;
e1, taking the optimal year and the optimal month of the soil attribute remote sensing inversion as an optimal time window of the soil attribute remote sensing inversion, and taking a first prediction model corresponding to the remote sensing image of the scene in the step B as a target prediction model if the number of the remote sensing images contained in the optimal time window of the soil attribute remote sensing inversion is one scene; if the number of the remote sensing images contained in the soil attribute remote sensing inversion optimal time window is larger than one scene, entering a step E2;
e2, carrying out remote sensing inversion on each scene remote sensing image in an optimal time window based on soil attributes to obtain each combination respectively comprising at least two scene remote sensing images, and then entering a step E3;
e3, respectively aiming at each combination, based on each scene remote sensing image in the combination and the data value of each sampling position corresponding to the target soil attribute in the target research area, applying a preset machine learning algorithm, and training to obtain a second prediction model taking the remote sensing image value of each sampling position in the target research area as input and the data value of each sampling position corresponding to the target soil attribute as output; obtaining second prediction models corresponding to the combinations respectively, and then entering a step E4;
and E4, determining an optimal second prediction model as a target prediction model based on the evaluation results of the second prediction models corresponding to the preset target evaluation indexes respectively.
As a preferred technical scheme of the invention: in the step E4, R of the coefficient evaluation index is determined based on the second prediction models 2 Value, valueRMSE value of root mean square error evaluation index, and statistics of maximum R 2 And a second prediction model which the value and the minimum RMSE value jointly belong to is taken as a target prediction model.
As a preferred technical scheme of the invention: the preset machine learning algorithm is an SVM algorithm.
Compared with the prior art, the soil remote sensing inversion method based on optimal time window selection has the following technical effects:
according to the soil remote sensing inversion method based on optimal time window selection, firstly, the optimal year is selected, so that the possible influence of factors such as weather among different years on soil attribute remote sensing inversion accuracy is reduced; secondly, determining the optimal month jointly by combining the crop growth condition reflected by the regional vegetation index and the accuracy result of inversion of soil attributes of different months, and reducing the influence of crop and crop straw coverage; finally, determining an optimal time window for inversion of the regional soil attribute by integrating the optimal year and the optimal month, and obtaining the target soil attribute in the target research region by using a target prediction model in the optimal time window; the remote sensing image in the optimal time window is designed and utilized, so that the influence of factors such as climate, crop coverage and the like can be effectively overcome, the accuracy of the remote sensing inversion of the soil property is effectively improved, and the practical application of the remote sensing inversion technology of the soil property is promoted.
Drawings
FIG. 1 is a flow chart of a soil remote sensing inversion method based on optimal time window selection designed by the invention;
FIG. 2 is a spatial distribution diagram of preset sampling locations within a target study area in accordance with an embodiment of the present invention;
FIG. 3 is a plot of soil organic matter prediction results obtained in a design embodiment of the present invention;
FIG. 4 is a spatial prediction diagram of soil organic matters obtained in a design example of the present invention.
Detailed Description
The following describes the embodiments of the present invention in further detail with reference to the drawings.
The soil surface coverings in different areas and the state of the soil have a certain time variation rule, the rule is reflected on the annual change, namely, the difference exists between months, meanwhile, certain variation exists between years, and the influence of interference factors on the soil spectrum can be reduced by selecting effective years and month time windows, so that the inversion precision is improved. Therefore, the method for realizing the remote sensing inversion of the soil attribute by searching the image in the optimal time window can rapidly and accurately acquire the soil attribute information.
In practical application, as shown in fig. 1, the invention designs a soil remote sensing inversion method based on optimal time window selection, and a target prediction model of target soil properties of a target research area is obtained according to the following steps A to E, and the target prediction model is applied to obtain the target soil properties in the target research area.
And A, obtaining remote sensing images which respectively correspond to each month of at least two years and cover a target research area, wherein each month is at least one scene remote sensing image, presetting data values of each sampling position corresponding to target soil attributes in the target research area, then respectively aiming at each scene remote sensing image, performing the following operations, performing atmosphere correction processing update, and then entering the step B.
The operation is as follows: for each pixel of the remote sensing image, firstly performing radiation calibration processing according to L=DN, so as to obtain radiation brightness data L, and then performing atmospheric correction processing on the remote sensing image by using a FLAASH model according to the radiation brightness data L; thereby realizing the atmosphere correction processing updating of the remote sensing image; wherein DN represents the pixel value of the remote sensing image, and alpha and beta are the conversion coefficients attached to the remote sensing image respectively.
In practical implementation, step A is to select a remote sensing image source with a re-returning interval of not less than 16 days, such as MODIS, landsat, sentinel-2, for obtaining a remote sensing image, so as to ensure that the remote sensing image can be obtained every month.
Step B, respectively aiming at each scene remote sensing image, based on the remote sensing image and the data value of each sampling position corresponding to the target soil attribute preset in the target research area, applying a machine learning algorithm such as an SVM algorithm, and training to obtain a first prediction model taking the remote sensing image value of each sampling position in the target research area as input and the data value of each sampling position corresponding to the target soil attribute as output; and C, obtaining a first prediction model corresponding to each scene remote sensing image respectively, and then entering the step C.
Step C, corresponding to the determination coefficient evaluation index R based on each first prediction model 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 And D, taking the year corresponding to the first prediction model to which the value and the minimum RMSE value belong together as the optimal year for the soil attribute remote sensing inversion, and then entering the step D.
And D, based on the evaluation results of the preset target evaluation indexes corresponding to the first prediction models, coupling the normalized vegetation indexes NDVI according to the following steps D1 to D4, carrying out statistics summarization on the first prediction models corresponding to different months, determining the optimal month of soil attribute remote sensing inversion, and then entering the step E.
Step D1, corresponding to the determination coefficient evaluation index R based on each first prediction model 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 D2, taking the month corresponding to the first prediction model to which the value and the minimum RMSE value belong together as the month to be matched, and then entering into the step D2;
step D2, aiming at each scene remote sensing image, respectively, according to the following formula:
NDVI=(NIR-Red)/(NIR+Red)
obtaining the NDVI value of each pixel in a target research area in the remote sensing image, obtaining an average value as the NDVI value corresponding to the target research area in the remote sensing image, further obtaining the NDVI value corresponding to each scene remote sensing image respectively, and then entering the step D3; wherein NIR represents the reflectivity of the near infrared band in the remote sensing image, and Red represents the reflectivity of the Red band in the remote sensing image;
step D3, respectively aiming at different months, calculating and obtaining average values of NDVI values corresponding to all the remote sensing images in the month, taking the average values as the NDVI values corresponding to the month, further obtaining the NDVI values corresponding to different months, and then entering the step D4;
step D4, if the month corresponding to the minimum NDVI value is the same month as the month to be matched, the month is used as the optimal month for the remote sensing inversion of the soil attribute; and if the month corresponding to the minimum NDVI value is different from the month to be matched, selecting the month corresponding to the minimum NDVI value and the adjacent next month thereof as the optimal month for remote sensing inversion of the soil attribute based on the closed loop sequence of the end-to-end month.
And E, taking the optimal year and the optimal month of the soil attribute remote sensing inversion as an optimal time window of the soil attribute remote sensing inversion, and training to obtain a target prediction model which takes the remote sensing image value of each sampling position in the target research area as input and the data value of each sampling position corresponding to the target soil attribute as output based on each scene remote sensing image in the target research area and the data value of each sampling position corresponding to the target soil attribute in the target research area in the optimal time window of the soil attribute remote sensing inversion by applying a preset machine learning algorithm such as an SVM algorithm.
In practical applications, the specific design of the step E is as follows steps E1 to E4 to obtain the target prediction model.
E1, taking the optimal year and the optimal month of the soil attribute remote sensing inversion as an optimal time window of the soil attribute remote sensing inversion, and taking a first prediction model corresponding to the remote sensing image of the scene in the step B as a target prediction model if the number of the remote sensing images contained in the optimal time window of the soil attribute remote sensing inversion is one scene; if the number of the remote sensing images contained in the soil attribute remote sensing inversion optimal time window is larger than one scene, entering a step E2;
e2, carrying out remote sensing inversion on each scene remote sensing image in an optimal time window based on soil attributes to obtain each combination respectively comprising at least two scene remote sensing images, and then entering a step E3;
e3, respectively aiming at each combination, based on each scene remote sensing image in the combination and the data value of each sampling position corresponding to the target soil attribute in the target research area, applying a preset machine learning algorithm such as an SVM algorithm, and training to obtain a second prediction model taking the remote sensing image value of each sampling position in the target research area as input and the data value of each sampling position corresponding to the target soil attribute as output; obtaining second prediction models corresponding to the combinations respectively, and then entering a step E4;
step E4, based on each second prediction model, respectively corresponding R of the decision coefficient evaluation index 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 And a second prediction model which the value and the minimum RMSE value jointly belong to is taken as a target prediction model.
The soil remote sensing inversion method based on the optimal time window selection is applied to the actual situation, the soil remote sensing inversion method is implemented by taking farmland soil organic matters as inversion objects and long-time-sequence MODIS remote sensing images as image sources in the target research area of Henan province and Share county, and the steps A to E are executed according to the specific embodiment of the invention as shown in FIG. 1.
Step A. Selecting an MODIS image (MOD 09A 1) with 500m spatial resolution synthesized in 8 days as a remote sensing image source to ensure that the remote sensing image can be obtained every month, wherein the image has 7 wave bands (wave band 1:620-670 nm, wave band 2:841-876 nm, wave band 3:459-479 nm, wave band 4:545-565 nm, wave band 5:1230-1250 nm, wave band 6:1628-1652 nm and wave band 7:2105-2155 nm), acquiring 644 remote sensing images covering 1 month in Qingzhu county to 2021 month in total from 2008, and presetting sampling positions in a target research area according to the figure 2 to obtain data values of soil organic matters corresponding to the preset sampling positions in the target research area, and performing atmospheric correction treatment on the remote sensing images, and then entering the step B.
Step B, regarding 644-scene MODIS remote sensing images, respectively aiming at each remote sensing image, based on the remote sensing images and data values corresponding to soil organic matters at preset sampling positions in a target research area, applying an SVM algorithm, and training to obtain a first prediction model taking the remote sensing image values of each sampling position in the target research area as input and the data values corresponding to target soil attributes of each sampling position as output; and obtaining 644-view MODIS remote sensing images respectively corresponding to the first prediction models, and then entering the step C.
And C, carrying out statistics summarization on the first prediction models corresponding to different years based on the evaluation results of the preset target evaluation indexes corresponding to the first prediction models, determining the optimal years of the soil attribute remote sensing inversion, and then entering the step D.
The above step C is specifically as follows in the embodiment:
step C1, predicting soil organic matters of the 644-view remote sensing image obtained by calculation, namely R 2 Summarizing the values of the RMSE according to different years from 2008 to 2021, and then entering step C2;
step C2. R in statistics of different years 2 The maximum value is 0.31, the rmse minimum value is 0.25%, and the year is 2010, so the optimal year is predicted by remote sensing with 2010 as the soil property, and then the step D is entered.
And D, based on the evaluation results of the preset target evaluation indexes corresponding to the first prediction models, coupling the normalized vegetation indexes NDVI according to the following steps D1 to D4, carrying out statistics summarization on the first prediction models corresponding to different months, determining the optimal month of the remote sensing inversion of the soil organic matters, and then entering the step E.
The above step D is specifically as follows in the embodiment:
step D1, summarizing and counting R according to different months, wherein the calculated soil organic matter prediction result of the 644-view remote sensing image 2 With the RMSE results, the maximum R is counted 2 D2, taking the month corresponding to the first prediction model to which the value and the minimum RMSE value belong together as the month to be matched, and then entering into the step D2;
step D2, aiming at all 644-view remote sensing images, the following formula is adopted:
NDVI=(NIR-Red)/(NIR+Red)
calculating to obtain the NDVI value of each pixel in the Zhuixian in the remote sensing image, obtaining an average value as the NDVI value corresponding to the target research area in the remote sensing image, further obtaining the NDVI value respectively corresponding to each scene of the remote sensing image, and then entering the step D3;
step D3, respectively aiming at different months, calculating and obtaining average values of NDVI values corresponding to all the remote sensing images in the month, taking the average values as the NDVI values corresponding to the month, further obtaining the NDVI values corresponding to different months, and then entering the step D4;
step D4, if the month corresponding to the minimum NDVI value is the same month as the month to be matched, the month is used as the optimal month for the remote sensing inversion of the soil attribute; and if the month corresponding to the minimum NDVI value is different from the month to be matched, selecting the month corresponding to the minimum NDVI value and the adjacent next month thereof together as the optimal month for remote sensing inversion of the soil attribute based on the closed loop sequence of the end-to-end month, and then entering the step E.
And E, determining an optimal time window of remote sensing inversion by combining the determined optimal year and the optimal month, constructing a target prediction model of the image in the optimal time window, and generating a soil attribute parameter map in a target research area.
In the embodiment, according to step E1, the optimal year 2010 and the optimal month 10 month and 11 month are comprehensively determined, and the optimal time window for remote sensing inversion of the soil attribute is determined to be 10 month and 11 month in 2010, and the number of remote sensing images in the window is larger than one scene; the method comprises the steps of selecting and integrating two remote sensing images to establish a target prediction model, namely, according to a step E2, obtaining each combination respectively comprising at least two remote sensing images; obtaining second prediction models corresponding to the combinations according to the step E3, and finally adopting a Leave-one-out cross validation method to use R according to the step E4 2 Evaluating model accuracy together with two indices of RMSE, where R 2 The larger the value is, the smaller the RMSE value is, the higher the model precision is, and the optimal result of the model is R 2 =0.45, rmse=0.23%, a scatter plot of the model predictions is shown in fig. 3, with the maximum R being the final statistic 2 And a second prediction model which the value and the minimum RMSE value jointly belong to is taken as a target prediction model.
In practical application, after the target prediction model is obtained, the target prediction model is applied to obtain soil organic matters in the target research area, and the result is shown in fig. 4.
In the application of the designed soil remote sensing inversion method based on the optimal time window selection, the possible influence of factors such as weather among different years on the soil attribute remote sensing inversion accuracy is reduced by selecting the optimal years; secondly, determining the optimal month jointly by combining the crop growth condition reflected by the regional vegetation index and the accuracy result of inversion of soil attributes of different months, and reducing the influence of crop and crop straw coverage; finally, determining an optimal time window for inversion of the regional soil attribute by integrating the optimal year and the optimal month, and obtaining the target soil attribute in the target research region by using a target prediction model in the optimal time window; the remote sensing image in the optimal time window is designed and utilized, so that the influence of factors such as climate, crop coverage and the like can be effectively overcome, the accuracy of the remote sensing inversion of the soil property is effectively improved, and the practical application of the remote sensing inversion technology of the soil property is promoted.
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 spirit of the present invention.
Claims (6)
1. A soil remote sensing inversion method based on optimal time window selection is characterized in that: the method comprises the following steps of A to E, obtaining a target prediction model of target soil properties of a target research area, and applying the target prediction model to obtain the target soil properties of the target research area;
step A, obtaining remote sensing images which respectively correspond to at least two months and cover a target research area, wherein each month is at least one scene remote sensing image, and a data value corresponding to a target soil attribute at each sampling position is preset in the target research area, then performing atmospheric correction processing on each scene remote sensing image, and then entering the step B;
step B, respectively aiming at each scene remote sensing image, based on the remote sensing image and a data value corresponding to the target soil attribute at each preset sampling position in a target research area, applying a preset machine learning algorithm, and training to obtain a first prediction model taking the remote sensing image value at each sampling position in the target research area as input and the data value corresponding to the target soil attribute at each sampling position as output; obtaining a first prediction model corresponding to each scene remote sensing image respectively, and then entering the step C;
step C, based on the evaluation results of the preset target evaluation indexes corresponding to the first prediction models, carrying out statistics summarization on the first prediction models corresponding to different years, determining the optimal years of soil attribute remote sensing inversion, and then entering the step D;
step D, based on the evaluation results of the preset target evaluation indexes corresponding to the first prediction models, coupling the normalized vegetation indexes NDVI according to the following steps D1 to D4, carrying out statistics summarization on the first prediction models corresponding to different months, determining the optimal month of soil attribute remote sensing inversion, and then entering the step E;
step D1, corresponding to the determination coefficient evaluation index R based on each first prediction model 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 D2, taking the month corresponding to the first prediction model to which the value and the minimum RMSE value belong together as the month to be matched, and then entering into the step D2;
step D2, aiming at each scene remote sensing image, respectively, according to the following formula:
NDVI=(NIR-Red)/(NIR+Red)
obtaining the NDVI value of each pixel in a target research area in the remote sensing image, obtaining an average value as the NDVI value corresponding to the target research area in the remote sensing image, further obtaining the NDVI value corresponding to each scene remote sensing image respectively, and then entering the step D3; wherein NIR represents the reflectivity of the near infrared band in the remote sensing image, and Red represents the reflectivity of the Red band in the remote sensing image;
step D3, respectively aiming at different months, calculating and obtaining average values of NDVI values corresponding to all the remote sensing images in the month, taking the average values as the NDVI values corresponding to the month, further obtaining the NDVI values corresponding to different months, and then entering the step D4;
step D4, if the month corresponding to the minimum NDVI value is the same month as the month to be matched, the month is used as the optimal month for the remote sensing inversion of the soil attribute; if the month corresponding to the minimum NDVI value is different from the month to be matched, selecting the month corresponding to the minimum NDVI value and the adjacent next month thereof as the optimal month for remote sensing inversion of soil attribute based on the closed loop sequence of the end-to-end month;
and E, taking the optimal year and the optimal month of the soil attribute remote sensing inversion as an optimal time window of the soil attribute remote sensing inversion, and training to obtain a target prediction model which takes the remote sensing image value of each sampling position in the target research area as input and the data value of each sampling position corresponding to the target soil attribute as output based on each scene remote sensing image in the target research area and the data value of each sampling position corresponding to the target soil attribute in the target research area in the optimal time window of the soil attribute remote sensing inversion by applying a preset machine learning algorithm.
2. The soil remote sensing inversion method based on optimal time window selection according to claim 1, wherein the soil remote sensing inversion method comprises the following steps: in the step A, the following operations are executed for each scene remote sensing image respectively, and the atmosphere correction processing update is carried out;
the operation is as follows: for each pixel of the remote sensing image, firstly performing radiation calibration processing according to L=DN, so as to obtain radiation brightness data L, and then performing atmospheric correction processing on the remote sensing image by using a FLAASH model according to the radiation brightness data L; thereby realizing the atmosphere correction processing updating of the remote sensing image; wherein DN represents the pixel value of the remote sensing image, and alpha and beta are the conversion coefficients attached to the remote sensing image respectively.
3. The soil remote sensing inversion method based on optimal time window selection according to claim 1, wherein the soil remote sensing inversion method comprises the following steps: in the step C, R of the coefficient evaluation index is correspondingly determined based on each first prediction model 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 And the year corresponding to the first prediction model to which the value and the minimum RMSE value belong together is used as the optimal year for the soil attribute remote sensing inversion.
4. The soil remote sensing inversion method based on optimal time window selection according to claim 1, wherein the soil remote sensing inversion method comprises the following steps: the step E comprises the following steps E1 to E4;
e1, taking the optimal year and the optimal month of the soil attribute remote sensing inversion as an optimal time window of the soil attribute remote sensing inversion, and taking a first prediction model corresponding to the remote sensing image of the scene in the step B as a target prediction model if the number of the remote sensing images contained in the optimal time window of the soil attribute remote sensing inversion is one scene; if the number of the remote sensing images contained in the soil attribute remote sensing inversion optimal time window is larger than one scene, entering a step E2;
e2, carrying out remote sensing inversion on each scene remote sensing image in an optimal time window based on soil attributes to obtain each combination respectively comprising at least two scene remote sensing images, and then entering a step E3;
e3, respectively aiming at each combination, based on each scene remote sensing image in the combination and the data value of each sampling position corresponding to the target soil attribute in the target research area, applying a preset machine learning algorithm, and training to obtain a second prediction model taking the remote sensing image value of each sampling position in the target research area as input and the data value of each sampling position corresponding to the target soil attribute as output; obtaining second prediction models corresponding to the combinations respectively, and then entering a step E4;
and E4, determining an optimal second prediction model as a target prediction model based on the evaluation results of the second prediction models corresponding to the preset target evaluation indexes respectively.
5. The soil remote sensing inversion method based on optimal time window selection according to claim 4, wherein the soil remote sensing inversion method comprises the following steps: in the step E4, R of the coefficient evaluation index is determined based on the second prediction models 2 The value, and the RMSE value of the root mean square error evaluation index, the maximum R is counted 2 And a second prediction model which the value and the minimum RMSE value jointly belong to is taken as a target prediction model.
6. A soil remote sensing inversion method based on optimal time window selection according to any one of claims 1 to 5, wherein: the preset machine learning algorithm is an SVM algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211189137.7A CN115495988B (en) | 2022-09-28 | 2022-09-28 | Soil remote sensing inversion method based on optimal time window selection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211189137.7A CN115495988B (en) | 2022-09-28 | 2022-09-28 | Soil remote sensing inversion method based on optimal time window selection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115495988A CN115495988A (en) | 2022-12-20 |
| CN115495988B true CN115495988B (en) | 2023-06-13 |
Family
ID=84472812
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211189137.7A Active CN115495988B (en) | 2022-09-28 | 2022-09-28 | Soil remote sensing inversion method based on optimal time window selection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115495988B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110046415A (en) * | 2019-04-08 | 2019-07-23 | 中国科学院南京地理与湖泊研究所 | A kind of soil organic matter content remote sensing dynamic playback method of space-time fining |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106767687B (en) * | 2017-02-22 | 2019-05-28 | 河海大学 | A method of utilizing remote sensing moisture measurement beach elevation |
| CN110222656B (en) * | 2019-06-11 | 2020-05-05 | 成都理工大学 | Ground vegetation ecological water quantitative inversion method based on remote sensing technology |
| CN111678866B (en) * | 2020-05-28 | 2021-08-17 | 电子科技大学 | Soil water content inversion method for multi-model ensemble learning |
-
2022
- 2022-09-28 CN CN202211189137.7A patent/CN115495988B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110046415A (en) * | 2019-04-08 | 2019-07-23 | 中国科学院南京地理与湖泊研究所 | A kind of soil organic matter content remote sensing dynamic playback method of space-time fining |
Non-Patent Citations (2)
| Title |
|---|
| The simultaneous prediction of soil properties and vegetation coverage from Vis-NIR Hyperspectral data with a one-dimensional convolutional neural network:a labortatory simulation study;Changkun Wang et al.;REMOTE SENSING;第14卷(第2期);全文 * |
| 基于遥感的广东省陆河县水土流失敏感性分析;李新尧等;亚热带水土保持;第33卷(第3期);16-20 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115495988A (en) | 2022-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11751500B2 (en) | Computer-implemented methods, computer readable medium and systems for a precision agriculture platform that detects soil conditions | |
| US10838936B2 (en) | Computer-implemented methods, computer readable medium and systems for generating an orchard data model for a precision agriculture platform | |
| CN109359623B (en) | Hyperspectral image migration classification method based on depth joint distribution adaptive network | |
| US10748081B2 (en) | Computer-implemented methods, computer readable medium and systems for a precision agriculture platform that identifies generic anomalies in crops | |
| AU2021282487A1 (en) | Forecasting national crop yield during the growing season | |
| CN108985959B (en) | Wheat powdery mildew remote sensing monitoring method based on surface temperature inversion technology | |
| US10395355B2 (en) | Computer-implemented methods, computer readable medium and systems for a precision agriculture platform | |
| US10410334B2 (en) | Computer-implemented methods, computer readable medium and systems for a precision agriculture platform with a satellite data model | |
| WO2021007352A1 (en) | Crop yield forecasting models | |
| US10943173B2 (en) | Computer-implemented methods, computer readable medium and systems for generating a satellite data model for a precision agriculture platform | |
| CN110579186B (en) | Crop growth monitoring method based on inversion of leaf area index by inverse Gaussian process | |
| CA2981473A1 (en) | Forecasting national crop yield during the growing season | |
| CN114663489A (en) | Crop leaf area index remote sensing inversion method and system under constraint of space-time characteristics of land blocks | |
| CN102222238A (en) | Automatic simulation method of natural-color products of high-space-resolution remote sensing images | |
| CN116310840A (en) | Winter wheat remote sensing identification method integrating multiple key weather period spectral features | |
| CN103186893A (en) | Universal high-resolution remote sensing image fusion method | |
| Yuan et al. | Research on rice leaf area index estimation based on fusion of texture and spectral information | |
| Kim et al. | Growth monitoring of field-grown onion and garlic by CIE L* a* b* color space and region-based crop segmentation of UAV RGB images | |
| CN115495988B (en) | Soil remote sensing inversion method based on optimal time window selection | |
| CN109086661A (en) | A kind of crops relative radiometric normalization method and device | |
| Zou et al. | Combining spectral and texture feature of UAV image with plant height to improve LAI estimation of winter wheat at jointing stage | |
| Gautam et al. | Image processing techniques and neural network models for predicting plant nitrate using aerial images | |
| Sun et al. | Potential of multi-seasonal vegetation indices to predict rice yield from UAV multispectral observations | |
| Abdalla et al. | Color Consistency of UAV Imagery using Multi-Channel CNN-based Image-to-Image Regression and Residual Learning | |
| CN111695533B (en) | Remote sensing mapping method for automatically monitoring planting areas of winter wheat and summer corn year by year |
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 |