An intelligent bare soil remote sensing identification method, system and device

Abstract

The application provides a kind of intelligent bare soil remote sensing identification method, system and device, comprising the following steps: S1, obtain the remote sensing image containing target area within a certain time;S2, the standardization processing is carried out to remote sensing image, and the same standardization image of dimension, spatial resolution and wave band configuration is obtained;S3, the identification index of each pixel in standardization image is calculated according to reflectivity data;S4, the time-space adaptive factor is calculated according to the time information and spatial information in metadata;S5, according to identification index, time-space adaptive factor and preset identification condition, each pixel is identified to bare soil;S6, according to bare soil identification result, the bare soil mask synthesis is carried out to remote sensing image, and the bare soil image is obtained.The application can realize high-precision identification of bare soil in different periods in specific area.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing identification technology, specifically relating to an intelligent remote sensing identification method, system, and device for bare soil. Background Technology

[0002] Current remote sensing technology for identifying bare soil has the following problems: Traditional methods require users to register accounts with multiple data providers, such as ESA, NASA, and the Chinese Academy of Sciences, apply for data access permissions, and manually download image data. This entire process can take days or even weeks. The downloaded raw data also requires complex preprocessing steps, including atmospheric correction, geometric correction, and radiometric calibration. These tasks require specialized remote sensing processing software and extensive technical experience. Furthermore, traditional local processing lacks the large-scale data management and preprocessing capabilities of cloud platforms, requiring users to manage and back up terabyte-level data themselves.

[0003] Second, current identification methods primarily rely on the NDVI (Normalized Difference Vegetation Index) for bare soil identification, using a fixed threshold to distinguish between vegetated and bare soil areas. However, this single-index method is prone to misclassification in complex terrain environments, particularly in scenarios with sparse vegetation in arid regions, mixed pixels in suburban areas, and crop rotation in farmland. The fixed-threshold method cannot adapt to the influence of different geographical environments and seasonal variations; the applicability of the same NDVI threshold varies greatly across different latitudes, seasons, and climatic conditions. A more serious problem is that during multi-temporal image synthesis, differences in spatial resolution, projection coordinate systems, and data formats between images acquired at different times can lead to shape inconsistencies in the image arrays during overlay calculations. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent remote sensing identification method, system, and device for bare soil, which can achieve high-precision identification of bare soil in a specific area at different times.

[0005] This invention relates to an intelligent remote sensing identification method for bare soil, comprising the following steps: S1. Acquire remote sensing images of the target area within a certain time period. The remote sensing images contain reflectance data and metadata for 6 bands. S2. Standardize the remote sensing images to obtain standardized images with the same dimension, spatial resolution and band configuration. S3. Calculate the recognition index of each pixel in the standardized image based on the reflectance data. The recognition index includes: vegetation index, bare soil index, soil-adjusted vegetation index, normalized water body index, normalized differential water body index, brightness index and spectral ratio filter index. S4. Extract month information from the time information in the metadata to determine the seasonal adjustment factor, extract latitude information from the spatial information in the metadata, and calculate the spatiotemporal adaptive factor based on the seasonal adjustment factor, latitude information, and the basic threshold determined based on the large sample vegetation index. The spatiotemporal adaptive factor is used to determine the upper limit of the threshold of the vegetation index in the identification conditions. S5. Perform bare soil identification on each pixel based on the identification indicators, spatiotemporal adaptive factors and preset identification conditions; S6. Based on the bare soil identification results, perform bare soil masking synthesis on the remote sensing images to obtain bare soil images.

[0006] Furthermore, S2 includes the following steps: S201. Obtain the dimensions and number of bands of the remote sensing image; S202. Calculate the frequency of target area combinations in remote sensing images; S203. Select the most frequently combined remote sensing imagery as the standard template; S204. Process non-standard images into standardized images with the same dimensions, spatial resolution, and band configuration as the standard template, based on the dimensions and number of bands.

[0007] Furthermore, S204 includes the following steps: S20401. Calculate the scaling ratio of each non-standard image relative to the standard template and scale it accordingly; S20402. Perform resampling interpolation based on the pixel position; S20403. Process the standardized image according to the number of bands. When the number of target bands is greater than the number of current bands, fill in the gaps by copying bands with similar spectra. When the number of target bands is less than the number of current bands, select the first N bands to retain according to their importance.

[0008] Furthermore, the formula for calculating the scaling ratio is: s x =w t / w c s y =h t / h c s x w is the scaling factor for the horizontal axis. t To standardize the width of the image, w c s is the width of the current image. y h is the scaling factor for the vertical axis. t For the height of the standardized image, h c The current image height; The formula for calculating resampling interpolation is: f(x,y)=f(0,0)×(1-x)×(1-y)+f(1,0)×x×(1-y)+f(0,1)×(1-x)×y+f(1,1)×x×y, where x and y represent pixel coordinates.

[0009] Furthermore, the formula for calculating the vegetation index is: NDVI=(NIR-RED) / (NIR+RED+0.0001); The formula for calculating the bare soil index is: BSI=(SWIR1+RED-NIR-BLUE) / (SWIR1+RED+NIR+BLUE+0.0001); The formula for calculating the soil-adjusted vegetation index is: SAVI=1.5×(NIR-RED) / (NIR+RED+0.5+0.0001); The formula for calculating the normalized water index is: NDWI=(GREEN-NIR) / (GREEN+NIR+0.0001); The formula for calculating the Normalized Difference Water Index is: MNDWI=(GREEN-SWIR1) / (GREEN+SWIR1+0.0001); The formula for calculating the luminance index is: Brightness=(BLUE+GREEN+RED) / 3; The formula for calculating the spectral ratio filter index is: Ratio = RED / NIR; Wherein, NIR is the reflectance of near-infrared light, RED is the reflectance of red light, SWIR1 and SWIR2 are the reflectances of short-wave infrared light, BLUE is the reflectance of blue light, and GREEN is the reflectance of green light.

[0010] Furthermore, S4 includes the following steps: S401. Extract month information based on time information, and extract latitude information of the target area center based on location information; S402. Determine the seasonal adjustment factor based on the monthly information; S403. Calculate the spatiotemporal adaptive factor based on the seasonal adjustment factor and latitude information.

[0011] Furthermore, the expression for the seasonal adjustment factor is: ; The formula for calculating the latitude adjustment factor is: α φ=-0.002×φ+0.1; φ is the dimension; The formula for calculating the spatiotemporal adaptive factor is: ; Among them, NDVI b,t The base threshold.

[0012] Furthermore, the preset recognition conditions in S5 are: ; ; ; ; Among them, C main As the core identification condition, C water C is a condition for water body identification. brightness For brightness recognition conditions, C spectral For infrared identification conditions, NDVI is the vegetation index, BSI is the bare soil index, SAVI is the soil-adjusted vegetation index, NDWI is the normalized water index, MNDWI is the normalized differential water index, Brightness is the brightness index, and Ratio is the spectral ratio filter index.

[0013] The present invention also provides an intelligent bare soil remote sensing identification system, comprising: The acquisition module is used to acquire remote sensing images containing the target area within a certain time period. The remote sensing images contain reflectance data and metadata for 6 bands. The preprocessing module is used to standardize remote sensing images to obtain standardized images with the same dimension, spatial resolution, and band configuration. The index calculation module is used to calculate the recognition index of each pixel in the standardized image based on reflectance data. The recognition index includes: vegetation index, bare soil index, soil-adjusted vegetation index, normalized water body index, normalized differential water body index, brightness index, and spectral ratio filter index. The factor calculation module is used to calculate the spatiotemporal adaptive factor based on the time and spatial information in the metadata. The identification module is used to identify bare soil for each pixel based on identification indicators, spatiotemporal adaptive factors, and preset identification conditions. The synthesis module is used to synthesize bare soil images by performing bare soil masking based on the bare soil identification results, thereby obtaining bare soil images.

[0014] The present invention also provides an intelligent bare soil remote sensing identification device, including a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the steps of the above method.

[0015] The beneficial effects of this invention are: (1) Significantly improve data acquisition efficiency: Remote sensing images can be directly connected to the cloud, shortening the traditional data acquisition process that takes several days or even weeks to minutes. Users do not need to register accounts and apply for permissions on multiple platforms, which greatly reduces the technical threshold and time cost.

[0016] (2) Solving the compatibility problem of multi-source images: Through standardized processing technology, the shape inconsistency error of images of different time phases and different resolutions in the synthesis process is effectively solved, ensuring the feasibility of multi-time phase image superposition analysis and improving the processing success rate to over 95%.

[0017] (3) Improve the accuracy of bare soil identification: The traditional single NDVI index method is replaced by a comprehensive discrimination of eight spectral indices. Combined with a multi-level logical judgment mechanism, the accuracy of bare soil identification is increased from 75-80% of the traditional method to 90-95%, which significantly reduces the misclassification rate.

[0018] (4) Enhanced environmental adaptability: The introduction of a spatiotemporal adaptive factor enables dynamic adjustment of the identification threshold, which can automatically adapt to environmental changes at different latitudes (0°-60°) and in different seasons, overcoming the limitations of the fixed threshold method and improving the universality and robustness of the method.

[0019] (5) Improve the level of automation: The entire process is automated, from data acquisition, preprocessing, index calculation to bare soil identification. This reduces the reliance on professional remote sensing software and technical experience, enabling non-professional users to perform high-precision bare soil identification.

[0020] (6) Optimize computational efficiency: By selecting the most frequent image combination as the standard template, 70-85% of remote sensing images do not require geometric transformation. Combined with a multi-threaded concurrent processing mechanism, the overall processing efficiency is improved by 3-5 times. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method in this invention; Figure 2 This is a display image of multi-band remote sensing imagery; Figure 3 A comparison chart showing the results of bare soil identification in different regions and at different time periods. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0023] like Figures 1 to 3As shown, the present invention provides an intelligent remote sensing identification method for bare soil, comprising the following steps: S1. Acquire remote sensing images containing the target area within a certain time period. The remote sensing images include reflectance data and metadata for six bands. Specifically, S1 includes the following steps: S101. Establish cloud connection and authentication. First, the GEE cloud data acquisition module establishes a secure connection with the GoogleEarthEngine platform. The API authentication management unit supports two authentication methods: (1) Service account key authentication: The system reads the RSA-2048-bit key file stored locally and uses the google.auth.service_account.Credentials.from_service_account_file() function to establish authentication. The authentication token is valid for T=3600 seconds, and the system sets the automatic refresh threshold to T×0.9=3240 seconds; (2) OAuth2.0 authentication: Access tokens are obtained through user interactive authorization. Access token refresh follows the formula Token_new=f(client_id,client_secret,refresh_token,grant_type). The system uses a heartbeat detection mechanism to check the connection status every 300 seconds to ensure the connection stability of long-term tasks.

[0024] S102. Constructing image query conditions. The intelligent image query unit receives the target area parameters and time range parameters input by the user and automatically constructs complex multi-dimensional filtering conditions.

[0025] (1) Setting the spatial range: The system converts the boundary coordinates (λmin, λmax, φmin, φmax) of the target area specified by the user into a Geometry object of GEE, and filters the image within the spatial range through the geometric intersection determination algorithm Intersection=Geometry_ROI∩Geometry_Image.

[0026] (2) Set time window: Convert the user-specified time range [t1,t2] into a GEE DateRange object, and execute the time filtering condition t1≤Image_Date≤t2, with time precision accurate to the day.

[0027] (3) Setting quality control: The system automatically sets the cloud coverage threshold α=20% and filters high-quality images by calculating Cloud_Coverage=(N_cloud_pixels / N_total_pixels)×100%.

[0028] (4) Overall quality score: The candidate images are ranked using the quality score function Q=w1×(100-Cloud%)+w2×(Signal_Noise_Ratio)+w3×(Geometric_Accuracy), where the weight coefficients w1=0.5, w2=0.3, w3=0.2.

[0029] S103. Execute batch download task. The image download unit performs multi-threaded batch download based on the query results. Specifically, (1) Band selection strategy: The system prioritizes downloading six key bands: B2 (blue light 490nm), B3 (green light 560nm), B4 (red light 665nm), B8 (near infrared 842nm), B11 (shortwave infrared 1610nm), and B12 (shortwave infrared 2190nm). These band combinations can support complete multispectral analysis.

[0030] (2) Concurrency control mechanism: The number of concurrent download threads is dynamically adjusted by the formula N=min(CPU_cores×2,Available_Bandwidth / Min_Bandwidth_Per_Thread) to avoid system resource overload.

[0031] (3) Download quality assurance: Each download task is checked with MD5: MD5_check=hash(downloaded_bytes)==expected_MD5, and the function of resuming interrupted download is supported to ensure data integrity.

[0032] (4) Metadata extraction: Simultaneously download the image metadata file MTD_MSIL2A.xml, and extract QUANTIFICATION_VALUE, RADIO_ADD_OFFSET, and solar zenith angle θ. s These key parameters prepare for subsequent calculations of reflectivity for each band. The calculation formulas are existing formulas and will not be elaborated upon. It should be noted that for Sentinel-2 satellites, they can directly output the reflectivity for each band, so users do not need to calculate it based on metadata.

[0033] Output a standardized Sentinel-2 multispectral image file set, containing reflectance data for 6 bands and complete metadata information, stored in GeoTIFF format, with WGS84 / UTM projection as the geographic coordinate system.

[0034] S2. Standardize the remote sensing imagery to obtain a standardized imagery with the same latitude, spatial resolution, and band configuration. Specifically, S2 includes the following steps: S201. Obtain the dimensions and number of bands of the remote sensing image. For the remote sensing image, use the rasterio.open() function to read all image files one by one, obtain the dimensions (height, width) through dataset.shape, obtain the number of bands through dataset.count, and construct the image attribute list image_inventory=[{'filename':fname,'bands':bands,'height':height,'width':width,'dtype':dtype}forfnameinfile_list].

[0035] S202. Calculate the frequency of target area combinations in remote sensing imagery. Create a statistical table of target area shape distribution and use collections.Counter() to count the frequency of shape combinations formed by each stitching method.

[0036] S203. Select the most frequently occurring combination of remote sensing images as the standard template. Calculate the probability of occurrence of each stitching method and select the most common array shape as the standard template. This selection strategy enables 70-85% of remote sensing images to be processed without geometric transformation, significantly improving processing efficiency.

[0037] S204. Process the non-standard imagery into a standardized imagery with the same dimensions, spatial resolution, and band configuration as the standard template, based on the dimensions and number of bands. For images with mismatched spatial dimensions, perform adaptive resampling. Specifically, S204 includes the following steps: S20401. Calculate and scale each non-standard image relative to the standard template. The formula for calculating the scaling ratio is: s x =w t / w c s y =h t / h c s x w is the scaling factor for the horizontal axis. t To standardize the width of the image, w c s is the width of the current image. y h is the scaling factor for the vertical axis. t For the height of the standardized image, h c This is the height of the current image. Scale the remaining images according to the scaling ratio. S20402. Perform resampling interpolation based on pixel position. Calculate the spectral values ​​at the edges using a bilinear interpolation algorithm. The formula for resampling interpolation is: f(x,y)=f(0,0)×(1-x)×(1-y)+f(1,0)×x×(1-y)+f(0,1)×(1-x)×y+f(1,1)×x×y; x and y represent pixel coordinates. Resampling is used to maintain the continuity of spectral features.

[0038] S20403. Standardized imagery is processed according to the number of bands, specifically using two processing strategies: (1) Band completion strategy: When the number of target bands is greater than the number of current bands, the most similar band is selected for copying using the spectral similarity calculation formula spectral_similarity=1-abs(mean(band_i)-mean(band_j)) / (mean(band_i)+mean(band_j)). mean(band_i) represents the average value of all pixel values ​​of the target band, and mean(band_j) represents the average value of all pixel values ​​of the current band.

[0039] (2) Band clipping strategy: When the number of target bands is less than the number of current bands, the bands are sorted from largest to smallest according to their importance weight band_importance=[0.8,1.0,1.0,1.2,0.9,0.9], and the top N key bands are retained.

[0040] In this step, the data types of the affected values ​​are standardized: they are uniformly converted to the float32 data type to avoid differences in numerical precision; all NoData values ​​are standardized to NaN to establish a consistent invalid value identification system.

[0041] S3. Calculate the recognition index for each pixel in the standardized image based on reflectance data. The recognition index includes: vegetation index, bare soil index, soil-adjusted vegetation index, normalized water index, normalized difference water index, brightness index, and spectral ratio filter index. Specifically, the formula for calculating the vegetation index is: NDVI = (NIR - RED) / (NIR + RED + 0.0001); The NDVI value of healthy vegetation is usually in the range of 0.3-0.8, while the NDVI value of bare soil and sparse vegetation is usually in the range of -0.1-0.3.

[0042] The formula for calculating the bare soil index is: BSI = (SWIR1 + RED - NIR - BLUE) / (SWIR1 + RED + NIR + BLUE + 0.0001); This index is specifically designed to highlight the spectral characteristics of bare soil. The BSI value of bare soil areas is usually positive and greater than 0.1, while the BSI value of vegetated areas is usually negative or close to zero.

[0043] The formula for calculating the soil-adjusted vegetation index is: SAVI = 1.5 × (NIR - RED) / (NIR + RED + 0.5 + 0.0001); The soil-adjusted vegetation index is set to 0.5, which is suitable for medium vegetation cover conditions. This index has better linear response characteristics in sparse vegetation areas.

[0044] The formula for calculating the normalized water index is: NDWI=(GREEN-NIR) / (GREEN+NIR+0.0001); The formula for calculating the Normalized Difference Water Index is: MNDWI=(GREEN-SWIR1) / (GREEN+SWIR1+0.0001); The formula for calculating the luminance index is: Brightness=(BLUE+GREEN+RED) / 3; The formula for calculating the spectral ratio filter index is: Ratio = RED / NIR; Wherein, NIR is the reflectance of near-infrared light, RED is the reflectance of red light, SWIR1 and SWIR2 are the reflectances of short-wave infrared light, BLUE is the reflectance of blue light, and GREEN is the reflectance of green light.

[0045] S4. Extract month information from the time information in the metadata to determine the seasonal adjustment factor, and extract latitude information from the spatial information in the metadata. Calculate the spatiotemporal adaptive factor based on the seasonal adjustment factor, latitude information, and a base threshold determined based on a large sample vegetation index. The spatiotemporal adaptive factor is used to determine the upper limit of the vegetation index threshold in the identification criteria. Specifically, S4 includes the following steps: S401. Parse the time information and date from the metadata, use the datetime.strptime() function to extract the month information, and extract the latitude information of the target area center based on the location information.

[0046] S402. Determine the seasonal adjustment factor based on the month information. Specifically, the expression for the seasonal adjustment factor is: The seasonal adjustment factor is 0.8 in spring to avoid misjudging newly emerging vegetation, and 1.2 in autumn to reduce disturbance from withered vegetation.

[0047] S403. Calculate the spatiotemporal adaptive factor based on the seasonal adjustment factor and latitude information. Specifically, the formula for calculating the latitude adjustment factor is: α φ=-0.002×φ+0.1; φ is the dimension. This formula takes into account the latitudinal zonation of vegetation distribution.

[0048] The formula for calculating the spatiotemporal adaptive factor is: Among them, NDVI b,t The basic threshold is determined based on large sample statistics and is generally set to 0.2.

[0049] S5. Bare soil identification is performed on each pixel based on the identification indicators, spatiotemporal adaptive factors, and preset identification conditions. The preset identification conditions integrate multi-condition logical judgment criteria to achieve rigorous bare soil identification. Specifically, the preset identification conditions in S5 are as follows: Core identification criteria: ; Water body identification conditions: ; Brightness recognition conditions; ; Infrared recognition conditions: .

[0050] The above identification conditions are applied using the bare soil discrimination function final_bare_soil=Cmain&Cwater&Cbrightness&Cspectral. Only pixels that pass the judgment at all levels are finally marked as bare soil. This multi-level judgment mechanism significantly improves the reliability and accuracy of the identification results.

[0051] After recognition, morphological optimization is performed to improve the spatial continuity and accuracy of the recognition results. First, a morphological opening operation is performed using a 3×3 pixel structuring element B. Where ⊖ represents the etching operation, , ⊕ indicates an expansion operation, A is a binary image of bare soil masked, and B is a 3×3 structuring element matrix. A combination of erosion and dilation operations effectively removes isolated noise points and small protrusions. Morphological closing operations are then performed. Small cavities within bare soil areas were filled using a reverse process of expansion followed by erosion, repairing identification breaks caused by localized shading or small vegetation patches. Finally, connectivity analysis was performed using the 8-neighborhood connectivity criterion. Where M(x,y) is the binary mask after morphological processing, and Ck is the k-th connected component. The area of ​​each connected component is calculated. According to the judgment conditions Fragmented areas smaller than 500 square meters are removed. These small patches are usually pseudo-bare soil areas caused by edge effects or mixed pixels. This processing step further improves the reliability and practicality of the identification results.

[0052] S6. Based on the bare soil identification results, the remote sensing images are composited using a bare soil mask to obtain bare soil images. In this step, time-series statistical analysis based on bare soil frequency is used to analyze persistent bare soil areas. Different composite algorithms based on different values ​​are used to synthesize bare soil images. For example, median composite is used to suppress the influence of outliers, or mean composite is used to provide stable statistical characteristics. A quality assessment is performed using identification confidence levels to generate a bare soil probability map and a deterministic map.

[0053] The present invention also provides an intelligent bare soil remote sensing identification system, comprising: The acquisition module is used to acquire remote sensing images containing the target area within a certain time period. The remote sensing images contain reflectance data and metadata for 6 bands. The preprocessing module is used to standardize remote sensing images to obtain standardized images with the same dimension, spatial resolution, and band configuration. The index calculation module is used to calculate the recognition index of each pixel in the standardized image based on reflectance data. The recognition index includes: vegetation index, bare soil index, soil-adjusted vegetation index, normalized water body index, normalized differential water body index, brightness index, and spectral ratio filter index. The factor calculation module is used to calculate the spatiotemporal adaptive factor based on the time and spatial information in the metadata. The identification module is used to identify bare soil for each pixel based on identification indicators, spatiotemporal adaptive factors, and preset identification conditions. The synthesis module is used to synthesize bare soil images by performing bare soil masking based on the bare soil identification results, thereby obtaining bare soil images.

[0054] The present invention also provides an intelligent bare soil remote sensing identification device, including a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to implement the steps of the above method.

[0055] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent remote sensing identification of bare soil, characterized in that, Includes the following steps: S1. Acquire remote sensing images containing the target area within a certain time period. The remote sensing images contain reflectance data and metadata for 6 bands. S2. Standardize the remote sensing image to obtain a standardized image with the same dimension, spatial resolution and band configuration. S3. Calculate the identification index of each pixel in the standardized image based on the reflectance data. The identification index includes: vegetation index, bare soil index, soil-adjusted vegetation index, normalized water body index, normalized differential water body index, brightness index, and spectral ratio filter index. S4. Extract month information from the time information in the metadata to determine the seasonal adjustment factor, extract latitude information from the spatial information in the metadata, and calculate the spatiotemporal adaptive factor based on the seasonal adjustment factor, latitude information and the basic threshold determined based on the large sample vegetation index. The spatiotemporal adaptive factor is used to determine the upper limit of the threshold of the vegetation index in the identification conditions. S5. Perform bare soil identification on each pixel according to the identification index, spatiotemporal adaptive factor and preset identification conditions; S6. Based on the bare soil identification results, the remote sensing image is synthesized using a bare soil mask to obtain a bare soil image.

2. The intelligent bare soil remote sensing identification method according to claim 1, characterized in that, S2 includes the following steps: S201. Obtain the dimension and number of bands of the remote sensing image; S202. Calculate the frequency of target area combinations in remote sensing images; S203. Select the most frequently combined remote sensing imagery as the standard template; S204. Based on the stated dimensions and number of bands, process the non-standard image into a standardized image with the same dimensions, spatial resolution, and band configuration as the standard template.

3. The intelligent bare soil remote sensing identification method according to claim 2, characterized in that, S204 includes the following steps: S20401. Calculate the scaling ratio of each non-standard image relative to the standard template and scale it accordingly; S20402. Perform resampling interpolation based on the pixel position; S20403. Process the standardized image according to the number of bands. When the number of target bands is greater than the number of current bands, fill in the gaps by copying bands with similar spectra. When the number of target bands is less than the number of current bands, select the first N bands to retain according to their importance.

4. The intelligent bare soil remote sensing identification method according to claim 3, characterized in that, The formula for calculating the scaling ratio is: s x =w t / w c s y =h t / h c s x w is the scaling factor for the horizontal axis. t To standardize the width of the image, w c s is the width of the current image. y h is the scaling factor for the vertical axis. t For the height of the standardized image, h c The current image height; The formula for calculating the resampling interpolation is: f(x,y)=f(0,0)×(1-x)×(1-y)+f(1,0)×x×(1-y)+f(0,1)×(1-x)×y+f(1,1)×x×y, where x and y represent pixel coordinates.

5. The intelligent bare soil remote sensing identification method according to claim 1, characterized in that, The formula for calculating the vegetation index is as follows: NDVI=(NIR-RED) / (NIR+RED+0.0001); The formula for calculating the bare soil index is as follows: BSI=(SWIR1+RED-NIR-BLUE) / (SWIR1+RED+NIR+BLUE+0.0001); The formula for calculating the soil-adjusted vegetation index is as follows: SAVI=1.5×(NIR-RED) / (NIR+RED+0.5+0.0001); The formula for calculating the normalized water index is as follows: NDWI=(GREEN-NIR) / (GREEN+NIR+0.0001); The formula for calculating the normalized difference water index is as follows: MNDWI=(GREEN-SWIR1) / (GREEN+SWIR1+0.0001); The formula for calculating the brightness index is: Brightness=(BLUE+GREEN+RED) / 3; The formula for calculating the spectral ratio filtering index is: Ratio = RED / NIR; Wherein, NIR is the reflectance of near-infrared light, RED is the reflectance of red light, SWIR1 and SWIR2 are the reflectances of short-wave infrared light, BLUE is the reflectance of blue light, and GREEN is the reflectance of green light.

6. The intelligent bare soil remote sensing identification method according to claim 1, characterized in that, S4 includes the following steps: S401. Extract month information based on time information, and extract latitude information of the target area center based on location information; S402. Determine the seasonal adjustment factor based on the month information; S403. Calculate the spatiotemporal adaptive factor based on the seasonal adjustment factor and latitude information.

7. The intelligent bare soil remote sensing identification method according to claim 6, characterized in that, The expression for the seasonal adjustment factor is: ； The formula for calculating the latitude adjustment factor is: α φ =-0.002×φ+0.1; φ is the dimension; The formula for calculating the spatiotemporal adaptive factor is as follows: ； Among them, NDVI b,t The base threshold is set at 0.

2.

8. The intelligent bare soil remote sensing identification method according to claim 1, characterized in that, The preset identification conditions in S5 are: ； ； ； ； Among them, C main As the core identification condition, C water C is a condition for water body identification. brightness For brightness recognition conditions, C spectral For infrared recognition conditions, T NDVI,d , where NDVI is the spatiotemporal adaptive factor, BSI is the bare soil index, SAVI is the soil-adjusted vegetation index, NDWI is the normalized water index, MNDWI is the normalized differential water index, Brightness is the brightness index, and Ratio is the spectral ratio filter index.

9. An intelligent bare soil remote sensing identification system, characterized in that, include: The acquisition module is used to acquire remote sensing images containing the target area within a certain time period. The remote sensing images contain reflectance data and metadata for 6 bands. The preprocessing module is used to standardize the remote sensing image to obtain a standardized image with the same dimension, spatial resolution and band configuration. The index calculation module is used to calculate the identification index of each pixel in the standardized image based on the reflectance data. The identification index includes: vegetation index, bare soil index, soil-adjusted vegetation index, normalized water body index, normalized differential water body index, brightness index, and spectral ratio filter index. The factor calculation module is used to extract month information from the time information in the metadata to determine the seasonal adjustment factor, extract latitude information from the spatial information in the metadata, and calculate the spatiotemporal adaptive factor based on the seasonal adjustment factor, latitude information and the basic threshold determined based on the large sample vegetation index. The spatiotemporal adaptive factor is used to determine the upper limit of the threshold of the vegetation index in the identification conditions. The identification module is used to identify bare soil for each pixel based on the identification indicators, spatiotemporal adaptive factors and preset identification conditions; The synthesis module is used to synthesize the remote sensing image by performing bare soil masking based on the bare soil identification results, so as to obtain a bare soil image.

10. An intelligent bare soil remote sensing identification device, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, and the processor being used to execute the computer program to implement the steps of the method of claim 1.

Images