WO2020063518A1

WO2020063518A1 - Soil moisture detection method and apparatus based on random forest regression algorithm, and electronic device

Info

Publication number: WO2020063518A1
Application number: PCT/CN2019/107236
Authority: WO
Inventors: 荆文龙; 周成虎; 姚凌; 杨骥
Original assignee: 广州地理研究所
Priority date: 2018-09-30
Filing date: 2019-09-23
Publication date: 2020-04-02
Also published as: CN109522516B; CN109522516A

Abstract

A soil moisture detection method and apparatus based on a random forest regression algorithm, and an electronic device. The soil moisture detection method based on a random forest regression algorithm comprises the following steps: establishing and training an optimal random forest algorithm model according to a vegetation index, ground surface temperature, albedo, digital elevation model and ground surface coverage of a sample area at a second resolution and according to satellite remote sensing soil moisture data of the sample area; and inputting a vegetation index, ground surface temperature, albedo, digital elevation model and ground surface coverage of a first area to be detected at the second resolution into the optimal random forest algorithm model to acquire the detected soil moisture of the first area to be detected. By means of the soil moisture detection method based on a random forest regression algorithm, soil moisture data of an unknown area that lacks satellite remote sensing soil moisture data can be detected.

Description

Method, device and electronic equipment for detecting soil moisture based on random forest regression algorithm

Technical field

The invention relates to the field of geographic information technology, in particular to a method, a device and an electronic device for detecting soil moisture based on a random forest regression algorithm.

Background technique

Soil moisture is an important basic parameter for climate, hydrology, ecology, and agriculture research. It directly controls the transport and balance of water and heat between the landing site and the atmosphere. At present, remote sensing technology can obtain regional large-scale information on changes in terrestrial soil moisture and apply it to various fields such as terrestrial hydrological research, detection of floods and droughts, assessment of crop growth trends, and research on natural and ecological environments. Limitations of technology, there are also a large number of unknown areas where satellite remote sensing soil moisture data cannot be obtained.

Summary of the Invention

Based on this, an object of the present invention is to provide a soil moisture detection method based on a random forest regression algorithm, which can detect soil moisture data in an unknown region where satellite remote sensing soil moisture data is missing.

The present invention is implemented by the following scheme:

A method for detecting soil moisture based on a random forest regression algorithm includes the following steps:

Obtain the vegetation index, surface temperature, albedo, digital elevation model and surface coverage of the sample area;

Resampling the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage at the first resolution of the sample area to the same second resolution as the satellite remote sensing soil moisture data of the sample area, where the first resolution is greater than Two resolution

The optimal random forest algorithm model is established and trained according to the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the sample area, and satellite remote sensing soil moisture data of the sample area, wherein the said The vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the sample area are used as input samples for the training optimal random forest algorithm model, and satellite remote sensing soil moisture data of the sample area is used as the input sample. Output samples for training the optimal random forest algorithm model;

Obtain the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first area to be detected;

Resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover at the first resolution of the first area to be detected to the same second resolution as the satellite remote sensing soil moisture data of the first area to be detected;

The vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the first area to be detected are input into the optimal random forest algorithm model to obtain the soil detection humidity of the first area to be detected.

The method for detecting soil moisture based on a random forest regression algorithm in the present invention uses a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample region having satellite remote sensing soil moisture data, and a remote sensing soil moisture of the sample region Based on the data, an optimal random forest algorithm model is established. Using this model, the soil moisture data in the missing regions can be calculated, which can make up for the shortcomings of satellite remote sensing monitoring and improve the soil moisture data.

In one embodiment, the optimal random forest is established and trained according to the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the sample area, and satellite remote sensing soil moisture data of the sample area. Before the algorithm model, the following steps are also included:

In the sample area, an area where the surface temperature during the day is less than the first temperature and / or the area of the water body is greater than the first percentage is removed.

Areas where the surface temperature is too low during the day may be snow-covered areas, and areas with a water area greater than the first percentage may be rivers, lakes, and seas. Therefore, removing the above areas can more accurately obtain soil moisture data.

In one embodiment, the method further includes the following steps:

Acquiring the first-resolution vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second region to be detected;

The vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution of the second area to be detected are input into the optimal random forest algorithm model to obtain the soil moisture of the first resolution of the second area to be detected ;

Residual correction is performed on the soil detection humidity of the first resolution of the second to-be-detected area to obtain the spatially down-scale soil humidity of the second to-be-detected area.

Using the optimal random forest algorithm model, the soil detection humidity with a resolution up to the first resolution can be obtained, and the residual error correction can be performed to obtain a more accurate soil detection humidity at the first resolution, that is, the spatial down-scale soil humidity.

In one embodiment, performing residual correction on the soil detection humidity of the first resolution of the second area to be detected to obtain the spatially down-scaled soil humidity of the second area to be detected specifically includes the following steps:

Resampling the soil detection humidity of the first resolution in the second area to be detected to the second resolution;

Calculating the difference between the satellite remote sensing soil moisture data in the second to-be-detected area and the soil moisture in the second to-be-detected area resampled to the second resolution to obtain a first residual error;

Interpolating the first residual space to a first resolution to obtain a second residual at a first resolution;

Add the soil moisture at the first resolution of the second area to be detected to the second residual to obtain the spatially down-scaled soil moisture at the second area to be detected.

In one embodiment, the first resolution is 0.05 ° * 0.05 °, and the second resolution is 0.25 ° * 0.25 °.

Further, the present invention also provides a soil moisture detection device based on a random forest regression algorithm, which is characterized in that it includes:

A first data acquisition module, configured to obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample area;

A first resampling module for resampling the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage at the first resolution of the sample region to the same second resolution as the satellite remote sensing soil moisture data of the sample region, The first resolution is greater than the second resolution;

Random forest training module, used to establish and train the optimal random forest according to the second resolution vegetation index, surface temperature, albedo, digital elevation model and surface coverage of the sample area, and satellite remote sensing soil moisture data of the sample area. An algorithm model, wherein the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the sample area are used as input samples for training the optimal random forest algorithm model, and the satellites of the sample area Remote sensing soil moisture data as output samples of the trained optimal random forest algorithm model;

A second data acquisition module, configured to obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of the first area to be detected;

The second resampling module is used for resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution of the first area to be detected to be the same as the satellite remote sensing soil moisture data of the first area to be detected Second resolution

A first soil moisture acquisition module, configured to input a vegetation index, a surface temperature, an albedo, a digital elevation model, and a surface coverage of the second resolution of the first area to be detected into the optimal random forest algorithm model to obtain the first to be detected The soil of the area is tested for humidity.

The soil moisture detection device based on the random forest regression algorithm according to the present invention uses the vegetation index, surface temperature, albedo, digital elevation model and ground cover of a sample area with satellite remote sensing soil moisture data, and remotely sensed soil moisture in the sample area. Based on the data, an optimal random forest algorithm model is established. Using this model, the soil moisture data in the missing regions can be calculated, which can make up for the shortcomings of satellite remote sensing monitoring and improve the soil moisture data.

In one embodiment, the method further includes:

The removing module is used to remove an area in the sample area where the surface temperature during the day is less than the first temperature and / or the area of the water body is greater than 20%.

In one embodiment, the method further includes:

A third data acquisition module, configured to obtain a first-resolution vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second region to be detected;

A second soil moisture acquisition module, configured to input the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first resolution of the second area to be detected into the optimal random forest algorithm model to obtain the second to be detected The first resolution soil moisture detection in the region;

The residual error correction module is configured to perform residual error correction on the soil detection humidity of the first resolution of the second to-be-detected region to obtain the spatial down-scale soil humidity of the second to-be-detected region.

Further, the present invention also provides a computer-readable medium having stored thereon a computer program that, when executed by a processor, implements any one of the soil moisture detection methods based on the random forest regression algorithm described above.

Further, the present invention further provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable by the processor. When the processor executes the computer program, the processor implements the above-mentioned Any soil moisture detection method based on random forest regression algorithm.

For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic flowchart of a soil moisture detection method based on a random forest regression algorithm in an embodiment; FIG.

FIG. 2 is a schematic diagram of a spatial downscaling soil moisture process in an embodiment; FIG.

3 is a schematic diagram of a residual error correction process in an embodiment;

4 is a schematic flowchart of a soil moisture detection method based on a random forest regression algorithm in an embodiment;

5 is a schematic flowchart of a soil moisture detection method based on a random forest regression algorithm in an embodiment;

6 is a schematic structural diagram of a soil moisture detection device based on a random forest regression algorithm in an embodiment;

FIG. 7 is a schematic structural diagram of an electronic device in an embodiment.

detailed description

Referring to FIG. 1, in one embodiment, a soil moisture detection method based on a random forest regression algorithm includes the following steps:

Step S10: Obtain the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the sample area.

The sample area is an area having satellite remote sensing soil moisture data, and the size of the sample area is an integer multiple of the resolution of the satellite remote sensing soil moisture data. The surface temperature is an average value of the surface temperature within a day, and the daytime surface temperature and the nighttime surface temperature are obtained by a satellite sensor modis (moderate-resolution imaging spectrometer). The Digital Elevation Model (DEM), referred to as DEM, is a digital simulation of the ground terrain through limited terrain elevation data, that is, a digital representation of the terrain surface shape, which is a set of ordered numerical arrays that represent the ground elevation A solid ground model that can be obtained from satellite remote sensing data. The vegetation index is a vegetation coverage index formed by combining satellite visible light and near-infrared bands according to the spectral characteristics of vegetation, and qualitatively and quantitatively evaluates vegetation coverage and its growth vigor. The value of the vegetation index is usually -1 to 1. In snow-covered, water and desert areas, the vegetation index is usually a constant less than zero. The albedo is the ratio of the surface's reflected flux of incident solar radiation to the incident solar radiation, which can be obtained from a variety of satellite remote sensing data. Land cover (LUCC, Land-Use, Land-Cover, Change) includes land use and land cover, etc., which can be obtained through satellite remote sensing data.

Step S20: Resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover at the first resolution of the sample area to the same second resolution as the satellite remote sensing soil moisture data of the sample area, where the first resolution The rate is greater than the second resolution.

Wherein, in this embodiment, the first resolution of the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage is 0.05 ° * 0.05 °, and the second resolution of the satellite remote sensing precipitation data is 0.25 ° * 0.25 °. Therefore, the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution need to be resampled to the second resolution.

Step S30: Establish and train an optimal random forest algorithm model based on the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the second resolution of the sample area, and satellite remote sensing soil moisture data of the sample area. The vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the second resolution of the sample area are used as input samples for the training optimal random forest algorithm model, and satellite remote sensing soil moisture data of the sample area As an output sample of the training optimal random forest algorithm model.

The satellite remote sensing soil moisture data in the sample area is soil moisture data monitored by satellite remote sensing in the sample area. The optimal random forest algorithm model is that after repeated training, the soil moisture data calculation error reaches a minimum. Random forest algorithm model.

Step S40: Obtain the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first area to be detected.

In this embodiment, the first area to be detected is an area without soil moisture data of a satellite remote sensing monitoring area, and the above data and the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the sample area can be obtained through the same Way to get.

Step S50: Resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover at the first resolution of the first region to be detected to the same second resolution as the satellite remote sensing soil moisture data of the first region to be detected .

Step S60: The vegetation index, the surface temperature, the albedo, the digital elevation model, and the ground cover of the second resolution of the first area to be detected are input into the optimal random forest algorithm model to obtain the soil detection humidity of the first area to be detected.

The method for detecting soil moisture based on a random forest regression algorithm in the present invention uses a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample region having satellite remote sensing soil moisture data, and remotely sensed soil moisture in a sample region. Based on the data, an optimal random forest algorithm model is established. Using this model, the soil moisture data in the missing regions can be calculated, which can make up for the shortcomings of satellite remote sensing monitoring and improve the soil moisture data.

In one embodiment, in order to obtain the soil moisture data of the sample area more accurately, the method further includes the following steps:

Step S2A: Remove an area in the sample area where the surface temperature during the day is less than the first temperature and / or the area of the water body is greater than the first percentage.

Areas where the surface temperature is too low during the day may be snow-covered areas, and areas with a water area greater than the first percentage may be rivers, lakes, and seas. Therefore, removing the above areas can more accurately obtain soil moisture data. In this embodiment, the first temperature is 0 ° C, and the first percentage is 20%.

Referring to FIG. 2, in an embodiment, the method further includes the following steps:

Step S70: Obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a surface coverage of the first resolution of the second region to be detected.

The second area to be detected is an area having satellite remote sensing soil moisture data, and the size of the sample area is an integer multiple of the resolution of the satellite remote sensing soil moisture data.

Step S80: input the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first resolution of the second region to be detected into the optimal random forest algorithm model to obtain the first resolution of the second region to be detected The soil is tested for humidity.

Step S90: Perform residual error correction on the detected soil humidity of the first resolution in the second to-be-detected area to obtain the spatial down-scale soil humidity of the second to-be-detected area.

The residual is the difference between the soil moisture data of the second area to be detected measured by satellite remote sensing and the soil moisture data of the second area to be detected obtained from the optimal random forest algorithm model. Residual correction is performed on the precipitation data of the second to-be-detected area to obtain more accurate soil moisture data of the second to-be-detected area with the first resolution.

Referring to FIG. 3, in an embodiment, step S90 specifically includes the following steps:

Step S91: Resampling the soil detection humidity of the first resolution in the second region to be detected to the second resolution.

Step S92: Calculate the difference between the satellite remote sensing soil moisture data in the second area to be detected and the soil moisture detected in the second area to be resampled to the second resolution to obtain the first residual error.

Step S93: The first residual space is interpolated to a first resolution to obtain a second residual of the first resolution.

Step S94: Adding the soil moisture at the first resolution of the second area to be detected to the second residual to obtain the spatially downscaled soil moisture at the second area to be detected.

Due to the characteristics of soil moisture monitoring, there must be some soil moisture that cannot be effectively indicated by the optimal random forest algorithm model, that is, residuals. The present invention minimizes the model error by downscaling this part of the residual and adding it to the soil moisture simulation value with a resolution of 0.05 ° * 0.05 °. Since the satellite remote sensing soil moisture data resolution of the second area to be detected is 0.25 ° * 0.25 °, the satellite remote sensing soil moisture data value of the second area to be detected can be subtracted from the second resolution of the second resolution area of the second area to be detected after resampling. The soil moisture data calculated the residuals at a resolution of 0.25 ° * 0.25 °. Due to the randomness of the residuals, this embodiment downscales the residuals by spatial interpolation. In this embodiment, the thin-plate spline function interpolation method (Thin-plate Spline) is used to interpolate the residuals. To get the best downscaling results.

In one embodiment, the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first resolution of the sample area are resampled to the same second resolution as the satellite remote sensing soil moisture data of the sample area, which is It is realized by calculating the average value of all the first resolution pixels in the range of the second resolution pixels, that is, calculating the average value of each 0.05 ° * 0.05 ° pixels in the range of 0.25 ° * 0.25 ° pixels.

In one embodiment, the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first resolution of the first region to be detected are resampled to the same as the satellite remote sensing soil moisture data of the first region to be detected. The second resolution is achieved by calculating the average value of all the first resolution pixels in the range of the second resolution pixels, that is, calculating each of the pixels in the range of 0.25 ° * 0.25 ° 0.05% * 0.05 ° The average value of the cells.

Please refer to FIG. 4. In a specific embodiment, the soil moisture detection method based on the random forest regression algorithm of the present invention includes the following steps:

Step S401: Obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample area.

Step S402: Resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution of the sample area to the second resolution.

Step S403: Establish the original sample set S according to the vegetation index, the surface temperature, the albedo, the digital elevation model and the surface coverage of the sample region and the satellite remote sensing soil moisture data of the sample region.

Step S404: Extract k training sample sets from the original sample set S by the Bootstrap method.

Step S405: learning the k training sets to generate k decision tree models. In the decision tree generation process, there are a total of 4 input variables, and n variables are randomly selected from the four variables. Each internal node is split using the optimal splitting method on the n characteristic variables, and the value of n is in a random forest. The formation of the model is a constant constant.

Step S406: Combine the results of the k decision trees, and repeatedly train to form the optimal random forest algorithm model.

Step S407: Obtain the vegetation index, the surface temperature, the albedo, the digital elevation model, and the surface coverage of the first area to be detected.

Step S408: Resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first area to be detected to the second resolution.

Step S409: The vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first sampled area to be resampled are input into the optimal random forest algorithm model to obtain the soil detection of the first area to be detected at the second resolution. humidity.

Please refer to FIG. 5. In a specific embodiment, the soil moisture detection method based on the random forest regression algorithm of the present invention includes the following steps:

Step S501: Obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample area.

Step S502: Resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution of the sample area to the second resolution.

Step S503: Establish the original sample set S according to the vegetation index, the surface temperature, the albedo, the digital elevation model, and the surface coverage of the sample region and the satellite remote sensing soil moisture data of the sample region.

Step S504: Extract k training sample sets from the original sample set S by the Bootstrap method.

Step S505: learning the k training sets to generate k decision tree models. In the decision tree generation process, there are a total of 4 input variables, and n variables are randomly selected from the four variables. Each internal node is split using the optimal splitting method on the n characteristic variables, and the value of n is in a random forest. The formation of the model is a constant constant.

Step S506: Combine the results of the k decision trees, and repeatedly train to form the optimal random forest algorithm model.

Step S507: Obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of the first resolution of the second region to be detected.

Step S508: input the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first resolution of the second region to be detected into the optimal random forest algorithm model to obtain the first resolution of the second region to be detected Soil moisture data.

Step S509: Resampling the soil moisture data of the first resolution to the second resolution in the second region to be detected.

Step S510: Calculate a difference between satellite remote sensing soil moisture data in the second area to be detected and soil moisture data in the second resolution of the second area to be detected after resampling to obtain a first residual error.

Step S511: Interpolate the first residual space to a first resolution to obtain a second residual of the first resolution.

Step S512: Add the soil moisture data of the first resolution of the second region to be detected to the second residual to obtain the spatially down-scaled soil moisture data of the second region to be detected.

The method for detecting soil moisture based on a random forest regression algorithm in the present invention uses a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample region having satellite remote sensing soil moisture data, and remotely sensed soil moisture in a sample region. Data, establish the optimal random forest algorithm model, and use this model to calculate the soil moisture data in the missing regions, which can make up for the shortcomings of satellite remote sensing monitoring and improve the soil moisture data; at the same time, use this model to calculate the second to be detected Regional high-resolution soil moisture data, and resampling the soil moisture data, and performing residual correction with satellite remote sensing soil moisture data to obtain highly accurate spatial down-scale soil moisture data.

Please refer to FIG. 6. In one embodiment, the soil moisture detection device 600 based on the random forest regression algorithm of the present invention includes:

A first data acquisition module 601, configured to obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample area;

A first resampling module 602, configured to resample the first resolution vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the sample area to the same second resolution as the satellite remote sensing soil moisture data of the sample area Where the first resolution is greater than the second resolution;

The random forest training module 603 is used to establish and train the optimal random according to the vegetation index, surface temperature, albedo, digital elevation model and surface coverage of the second resolution of the sample area, and satellite remote sensing soil moisture data of the sample area. A forest algorithm model, wherein the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the second resolution of the sample area are used as input samples for training the optimal random forest algorithm model. Satellite remote sensing soil moisture data as output samples of the training optimal random forest algorithm model;

A second data acquisition module 604, configured to obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of the first area to be detected;

The second resampling module 605 is configured to resample the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution of the first area to be detected to satellite remote sensing soil moisture data of the first area to be detected Same second resolution

A first soil moisture acquisition module 606 is configured to input a vegetation index, a surface temperature, an albedo, a digital elevation model, and a surface coverage of the second resolution of the first area to be detected into the optimal random forest algorithm model to obtain the first The soil in the test area is tested for humidity.

In one embodiment, the method further includes:

The removing module 607 is configured to remove an area in the sample area where the surface temperature during the day is less than the first temperature and / or the area of the water body is greater than 20%.

In one embodiment, the method further includes:

A third data acquisition module 608, configured to acquire a first resolution vegetation index, a surface temperature, an albedo, a digital elevation model, and a surface coverage of the second region to be detected;

A second soil moisture acquisition module 609 is configured to input a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of the first resolution of the second region to be detected into the optimal random forest algorithm model to obtain a second Soil detection humidity of the first resolution in the detection area;

The residual error correction module 610 is configured to perform residual error correction on the soil detection humidity of the first resolution of the second region to be detected to obtain the spatially down-scaled soil humidity of the second region to be detected.

In one embodiment, the residual correction module 610 includes:

A third resampling unit 6101, configured to resample the soil detection humidity of the first resolution to the second resolution in the second region to be detected;

A first residual acquisition unit 6102 is configured to calculate a difference between the satellite remote sensing soil moisture data in the second area to be detected and the soil humidity detected in the second area to be resampled to the second resolution to obtain the first residual ;

A first residual acquisition unit 6103, configured to interpolate the first residual space to a first resolution to obtain a second residual at a first resolution;

The down-scale soil moisture data acquisition unit 6104 is configured to add the first-resolution soil detection humidity in the second region to be detected and the second residual to obtain spatial down-scale soil humidity data in the second region to be detected.

In one embodiment, the first resampling module 602 includes a first resolution calculation unit 6021, configured to calculate a vegetation index and a surface temperature of a sample region of all the first resolution pixels in a range of the second resolution pixels. , Albedo, digital elevation model, and surface coverage averages, that is, the average of each 0.05 ° * 0.05 ° pixel in the range of 0.25 ° * 0.25 ° pixels.

In one embodiment, the second resampling module 605 includes a first resolution calculation unit 6051, configured to calculate the vegetation index of the first area to be detected in all the first resolution pixels in the range of the second resolution pixels, The surface temperature, albedo, digital elevation model, and average surface coverage are used to calculate the average of each 0.05 ° * 0.05 ° pixel in the range of 0.25 ° * 0.25 ° pixels.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and are not repeated here.

The soil moisture detection device based on the random forest regression algorithm according to the present invention uses the vegetation index, surface temperature, albedo, digital elevation model and ground cover of a sample area with satellite remote sensing soil moisture data, and remotely sensed soil moisture in the sample area. Data, establish the optimal random forest algorithm model, and use this model to calculate the soil moisture data in the missing regions, which can make up for the shortcomings of satellite remote sensing monitoring and improve the soil moisture data; at the same time, use this model to calculate the second to be detected Regional high-resolution soil moisture data, and resampling the soil moisture data, and performing residual correction with satellite remote sensing soil moisture data to obtain highly accurate spatial down-scale soil moisture data.

The present invention also provides a computer-readable medium on which a computer program is stored. When the computer program is executed by a processor, the method for detecting soil moisture based on a random forest regression algorithm in any of the above embodiments is implemented.

Referring to FIG. 7, in an embodiment, the electronic device 70 of the present invention includes a memory 71 and a processor 72, and a computer program stored in the memory 71 and executable by the processor 72. The processor 72. When the computer program is executed, a soil moisture detection method based on a random forest regression algorithm in any of the above embodiments is implemented.

In this embodiment, the controller 72 may be one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), and field programmable Gate array (FPGA), controller, microcontroller, microprocessor, or other electronic component. The storage medium 71 may take the form of a computer program product implemented on one or more storage media (including, but not limited to, magnetic disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media includes permanent and non-permanent, removable and non-removable media, and information can be stored by any method or technology. Information may be computer-readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, Magnetic tape cartridges, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media may be used to store information that can be accessed by computing devices.

In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods may be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner. For example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, which may be electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objective of the solution of this embodiment.

The above-mentioned embodiments only express several implementation manners of the present invention, and their descriptions are more specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be noted that, for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can be made, which all belong to the protection scope of the present invention.

Claims

A method for detecting soil moisture based on a random forest regression algorithm is characterized in that it includes the following steps:

Obtain the vegetation index, surface temperature, albedo, digital elevation model and surface coverage of the sample area;

Resampling the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage at the first resolution of the sample area to the same second resolution as the satellite remote sensing soil moisture data of the sample area, where the first resolution is greater than Two resolution

The optimal random forest algorithm model is established and trained according to the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the sample area, and satellite remote sensing soil moisture data of the sample area, wherein the said The vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the sample area are used as input samples for the training optimal random forest algorithm model, and satellite remote sensing soil moisture data of the sample area is used as the input sample. Output samples for training the optimal random forest algorithm model;

Obtain the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first area to be detected;

Resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover at the first resolution of the first area to be detected to the same second resolution as the satellite remote sensing soil moisture data of the first area to be detected;

The vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second resolution of the first area to be detected are input into the optimal random forest algorithm model to obtain the soil detection humidity of the first area to be detected.
The method for detecting soil moisture based on a stochastic forest regression algorithm according to claim 1, characterized in that, according to the second resolution vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the sample area, and the sample Regional satellite remote sensing of soil moisture data, before establishing and training the optimal random forest algorithm model, the following steps are also included:

In the sample area, an area where the surface temperature during the day is less than the first temperature and / or the area of the water body is greater than the first percentage is removed.
The method for detecting soil moisture based on a random forest regression algorithm according to any one of claims 1 or 2, further comprising the following steps:

Acquiring the first-resolution vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second region to be detected;

The vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution of the second area to be detected are input into the optimal random forest algorithm model to obtain the soil moisture of the first resolution of the second area to be detected ;

Residual correction is performed on the soil detection humidity of the first resolution of the second to-be-detected area to obtain the spatially down-scale soil humidity of the second to-be-detected area.
The method for detecting soil moisture based on a random forest regression algorithm according to claim 3, characterized in that residual correction is performed on the soil detection humidity of the first resolution of the second area to be detected to obtain the information of the second area to be detected. The spatial down-scale soil moisture includes the following steps:

Resampling the soil detection humidity of the first resolution in the second area to be detected to the second resolution;

Calculating the difference between the satellite remote sensing soil moisture data in the second to-be-detected area and the soil moisture in the second to-be-detected area resampled to the second resolution to obtain a first residual error;

Interpolating the first residual space to a first resolution to obtain a second residual at a first resolution;

Add the soil moisture at the first resolution of the second area to be detected to the second residual to obtain the spatially down-scaled soil moisture at the second area to be detected.
The method for detecting soil moisture based on a random forest regression algorithm according to claim 1, characterized in that:

The first resolution is 0.05 ° * 0.05 °, and the second resolution is 0.25 ° * 0.25 °.
A soil moisture detection device based on a random forest regression algorithm, which is characterized in that it includes:

A first data acquisition module, configured to obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of a sample area;

A first resampling module for resampling the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage at the first resolution of the sample region to the same second resolution as the satellite remote sensing soil moisture data of the sample region, The first resolution is greater than the second resolution;

Random forest training module, used to establish and train the optimal random forest according to the second resolution vegetation index, surface temperature, albedo, digital elevation model and surface coverage of the sample area, and satellite remote sensing soil moisture data of the sample area. Algorithm model

A second data acquisition module, configured to obtain a vegetation index, a surface temperature, an albedo, a digital elevation model, and a ground cover of the first area to be detected;

The second resampling module is used for resampling the vegetation index, surface temperature, albedo, digital elevation model, and ground cover of the first resolution of the first area to be detected to be the same as the satellite remote sensing soil moisture data of the first area to be detected Second resolution

A first soil moisture acquisition module, configured to input a vegetation index, a surface temperature, an albedo, a digital elevation model, and a surface coverage of the second resolution of the first area to be detected into the optimal random forest algorithm model to obtain the first to be detected The soil of the area is tested for humidity.
The soil moisture detection device based on the random forest regression algorithm according to claim 6, further comprising:

The removing module is used to remove an area in the sample area where the surface temperature during the day is less than the first temperature and / or the area of the water body is greater than 20%.
The soil moisture detection device based on a random forest regression algorithm according to any one of claims 6 or 7, further comprising:

A third data acquisition module, configured to obtain a first-resolution vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the second region to be detected;

A second soil moisture acquisition module, configured to input the vegetation index, surface temperature, albedo, digital elevation model, and surface coverage of the first resolution of the second area to be detected into the optimal random forest algorithm model to obtain the second to be detected The first resolution soil moisture detection in the region;

The residual error correction module is configured to perform residual error correction on the soil detection humidity of the first resolution of the second to-be-detected region to obtain the spatial down-scale soil humidity of the second to-be-detected region.
A computer-readable medium having a computer program stored thereon is characterized by:

When the computer program is executed by a processor, the method for detecting soil moisture based on a random forest regression algorithm according to any one of claims 1 to 5 is implemented.
An electronic device includes a memory, a processor, and a computer program stored in the memory and executable by the processor, and is characterized by:

When the processor executes the computer program, the method for detecting soil moisture based on a random forest regression algorithm according to any one of claims 1 to 5 is implemented.