Machine learning based vineyard soil organic matter identification analysis method and system

Abstract

The application discloses a kind of vineyard soil organic matter identification analysis method and system based on machine learning, method includes: collecting the environmental data of vineyard, remote sensing data and organic matter actual measurement data, remote sensing data is preprocessed, and preprocessed image is obtained;Feature extraction is carried out to preprocessed image, and spectral feature is obtained;Initial organic matter estimation model is constructed, and initial organic matter estimation model is trained by environmental data, spectral feature and organic matter actual measurement, and organic matter estimation model is obtained;Based on organic matter evaluation model, organic matter evaluation analysis is carried out to vineyard soil, and vineyard organic matter distribution result is obtained.The application can realize the rapid real-time monitoring of the organic matter content of large-area vineyard soil, and provide timely soil quality information for agricultural production.

Description

Technical Field

[0001] This invention belongs to the field of grape cultivation technology, and in particular relates to a method and system for identifying and analyzing soil organic matter in vineyards based on machine learning. Background Technology

[0002] Organic matter plays a vital role in viticulture. It not only improves soil structure, increases permeability and water retention, but also provides abundant nutrients for grapevines, promoting healthy root development. Simultaneously, organic matter enhances soil biological activity, increases microbial diversity, and aids in disease control and soil chemical improvement. Furthermore, it reduces soil erosion and compaction, creating a favorable soil environment for grape growth. Most importantly, organic matter improves grape quality, enhancing fruit flavor and aroma, while also increasing resistance to adverse conditions and reducing pesticide use, making it crucial for producing safe and healthy grapes.

[0003] In modern precision agriculture, the monitoring and management of soil organic matter is of great significance for improving grape yield and quality. Traditional methods for detecting soil organic matter in vineyards usually rely on laboratory chemical analysis. While these methods are accurate, they are time-consuming, costly, and cannot achieve large-scale real-time monitoring. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a machine learning-based method and system for identifying and analyzing organic matter in vineyard soil, thereby resolving the issues present in the prior art.

[0005] To achieve the above objectives, this invention provides a machine learning-based method for identifying and analyzing organic matter in vineyard soil, comprising the following steps:

[0006] Collect environmental data, remote sensing data, and actual organic matter measurement data of the vineyard; preprocess the remote sensing data to obtain a preprocessed image.

[0007] Feature extraction is performed on the preprocessed image to obtain spectral features;

[0008] An initial organic matter estimation model is constructed, and the initial organic matter estimation model is trained using environmental data, spectral characteristics, and actual organic matter measurements to obtain an organic matter estimation model.

[0009] Based on the organic matter assessment model, an organic matter assessment and analysis was conducted on the vineyard soil to obtain the results of organic matter distribution in the vineyard.

[0010] Preferably, the method for preprocessing the remote sensing data includes:

[0011] Remove cloud shadows from remote sensing images;

[0012] Radiometric calibration is performed on the remote sensing image after removing cloud shadows to obtain the apparent radiance satellite image, and the apparent reflectance corresponding to each pixel in the apparent radiance satellite image is calculated.

[0013] Atmospherically absorbed and intrinsic atmospheric reflectance are corrected for satellite images that exhibit reflectivity to obtain atmospherically corrected images.

[0014] The equivalent average background reflectance of each pixel in the atmospheric corrected image is calculated, and the satellite image is corrected pixel by pixel based on the equivalent average background reflectance to obtain the final corrected image.

[0015] Image enhancement processing is performed based on the final corrected image, and the enhanced image is verified using manually sampled data.

[0016] Preferably, the method for feature extraction of the preprocessed image includes:

[0017] The preprocessed image is denoised, and then the spectrum of all pixels is averaged to obtain the average spectrum of the image.

[0018] By performing a first-order differential on the spectrum of each pixel, steep changes in the spectral curve can be obtained;

[0019] The second derivative of the spectrum of each pixel is taken to obtain the bending points in the spectral curve;

[0020] The coefficients of the tassel transformation are calculated based on the average spectrum and its first and second derivatives, and the coefficients of the tassel transformation are used as spectral features.

[0021] Preferably, the initial organic matter estimation model is a network model based on support vector machines.

[0022] Preferably, the method for obtaining the organic matter distribution results of the vineyard includes: stitching together images of the vineyard, and then analyzing the stitched images using an organic matter estimation model to obtain the organic matter distribution results of the vineyard.

[0023] Preferably, the method for stitching together vineyard images includes:

[0024] Search for identical lines in two images to be stitched together as stitching lines; segment the image along the stitching lines to obtain images to be stitched together; stitch the two large-area images to be stitched together to obtain an initial stitched image; eliminate the stitching seams in the initial stitched image using a uniform light and color method.

[0025] This invention also provides a machine learning-based system for identifying and analyzing organic matter in vineyard soil, comprising:

[0026] The data acquisition module is used to collect environmental data, remote sensing data, and actual measurement data of organic matter in the vineyard, and to preprocess the remote sensing data to obtain a preprocessed image.

[0027] The feature extraction module, connected to the data acquisition module, is used to extract features from the preprocessed image to obtain spectral features;

[0028] The model building module is connected to the data acquisition module and the feature extraction module respectively, and is used to build an initial organic matter estimation model. The initial organic matter estimation model is trained by environmental data, spectral features and actual organic matter measurements to obtain an organic matter estimation model.

[0029] The assessment module, connected to the model building module, is used to conduct organic matter assessment and analysis on vineyard soil based on the organic matter assessment model, and obtain the distribution results of organic matter in the vineyard.

[0030] Preferably, the data acquisition module includes an acquisition submodule and a preprocessing submodule;

[0031] The acquisition submodule is used to collect environmental data, remote sensing data and actual measurement data of organic matter in the vineyard;

[0032] The preprocessing submodule is used to preprocess the remote sensing data.

[0033] Preferably, the preprocessing submodule includes a cloud shadow removal unit, a radiometric calibration unit, an atmospheric correction unit, a proximity effect correction unit, an image enhancement unit, and a verification unit;

[0034] The cloud shadow removal unit is used to remove cloud shadows in remote sensing images;

[0035] The radiometric calibration unit is used to perform radiometric calibration on remote sensing images with cloud shadows removed to obtain apparent radiance satellite images, and to calculate the apparent reflectance corresponding to each pixel in the apparent radiance satellite images.

[0036] The atmospheric correction unit is used to correct atmospheric absorption and intrinsic atmospheric reflectivity in satellite images that reflect reflectivity, thereby obtaining atmospherically corrected images.

[0037] The proximity effect correction unit is used to calculate the equivalent average background reflectance corresponding to each pixel in the atmospheric corrected image, and to perform proximity effect correction on the satellite image pixel by pixel based on the equivalent average background reflectance to obtain the final corrected image.

[0038] The image enhancement unit is used to perform image enhancement processing based on the final corrected image;

[0039] The verification unit is used to verify the enhanced image using manually sampled data.

[0040] Preferably, the evaluation module includes a splicing submodule and a recognition submodule;

[0041] The stitching submodule is used to stitch together vineyard images to obtain a stitched image;

[0042] The identification submodule is used to analyze the spliced ​​image through an organic matter estimation model to obtain the distribution results of organic matter in the vineyard;

[0043] The splicing submodule includes a splicing line acquisition unit, an image segmentation unit, a splicing unit, and a splicing seam elimination unit;

[0044] The splicing line acquisition unit is used to search for the same line in two images to be spliced ​​as the splicing line;

[0045] The image segmentation unit is used to segment the image along the splicing line to obtain the image to be spliced.

[0046] The stitching unit is used to stitch two large-area images to be stitched together to obtain an initial stitched image;

[0047] The seam elimination unit is used to eliminate seams in the initial stitched image by using a uniform light and color method.

[0048] Compared with the prior art, the present invention has the following advantages and technical effects:

[0049] This invention discloses a machine learning-based method and system for identifying and analyzing soil organic matter in vineyards. The method includes: collecting environmental data, remote sensing data, and actual organic matter measurement data from the vineyard; preprocessing the remote sensing data to obtain a preprocessed image; extracting features from the preprocessed image to obtain spectral features; constructing an initial organic matter estimation model; training the initial organic matter estimation model using environmental data, spectral features, and actual organic matter measurements to obtain the final organic matter estimation model; and performing an organic matter assessment and analysis on the vineyard soil based on the organic matter assessment model to obtain the organic matter distribution results. This invention enables rapid, real-time monitoring of soil organic matter content in large-area vineyards, providing timely soil quality information for agricultural production. The organic matter distribution results provided by this invention can assist agricultural producers in soil management and fertilization decisions, improving vineyard production efficiency and fruit quality. Attached Figure Description

[0050] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0051] Figure 1This is a flowchart of the method for identifying and analyzing organic matter in vineyard soil according to an embodiment of the present invention. Detailed Implementation

[0052] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0053] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0054] Example 1

[0055] like Figure 1 As shown, this embodiment provides a machine learning-based method for identifying and analyzing organic matter in vineyard soil, including the following steps:

[0056] Collect environmental data, remote sensing data, and actual organic matter measurement data of the vineyard; preprocess the remote sensing data to obtain preprocessed images;

[0057] Feature extraction is performed on the preprocessed image to obtain spectral features;

[0058] An initial organic matter estimation model was constructed, and the model was trained using environmental data, spectral characteristics, and actual organic matter measurements to obtain the final organic matter estimation model.

[0059] Organic matter was assessed and analyzed in vineyard soil based on an organic matter assessment model to obtain the distribution results of organic matter in the vineyard.

[0060] Furthermore, methods for preprocessing remote sensing data include:

[0061] Remove cloud shadows from remote sensing images;

[0062] Radiometric calibration is performed on the remote sensing image after removing cloud shadows to obtain the apparent radiance satellite image, and the apparent reflectance corresponding to each pixel in the apparent radiance satellite image is calculated.

[0063] Atmospherically absorbed and intrinsic atmospheric reflectance are corrected for satellite images that exhibit reflectivity to obtain atmospherically corrected images.

[0064] The equivalent average background reflectance of each pixel in the atmospheric corrected image is calculated, and the satellite image is corrected pixel by pixel based on the equivalent average background reflectance to obtain the final corrected image.

[0065] Image enhancement processing is performed based on the final corrected image, and the enhanced image is verified using manually sampled data.

[0066] The verification method involves PSF detection of the image: a direct detection method based on the physical definition of the point spread function. A parametric point spread model is used to fit the surface of the target response data of each sub-pixel to determine the peak position of each target response. Then, the sub-pixel interpolated point spread response is obtained by position registration of the sub-pixel target response values ​​or count values ​​of a 3*3 array with non-integer pixel intervals to reduce the influence of system sampling effects and random noise. The point spread function of the optical remote sensing satellite imaging system is obtained by fitting a parametric Gaussian model. The obtained system point spread function is subjected to discrete Fourier transform, modulus taking, and normalization to obtain the modulation transfer function of the optical remote sensing satellite imaging system. The image quality is evaluated by the modulation transfer function.

[0067] Furthermore, methods for feature extraction from preprocessed images include:

[0068] The preprocessed image is denoised, and then the spectrum of all pixels is averaged to obtain the average spectrum of the image.

[0069] By performing a first-order differential on the spectrum of each pixel, steep changes in the spectral curve can be obtained;

[0070] The second derivative of the spectrum of each pixel is taken to obtain the bending points in the spectral curve;

[0071] The coefficients of the tasseled cap transformation are calculated based on the average spectrum and its first and second derivatives, and the coefficients of the tasseled cap transformation are used as spectral features.

[0072] Furthermore, the initial organic matter estimation model is a network model based on support vector machines.

[0073] Furthermore, methods for obtaining the distribution results of organic matter in vineyards include: stitching together images of vineyards, and then analyzing the stitched images using an organic matter estimation model to obtain the distribution results of organic matter in vineyards.

[0074] Furthermore, methods for stitching together vineyard images include:

[0075] In the image stitching process, the first step is to identify identical lines in the two images to be stitched together; these lines will serve as the basis for subsequent stitching. Next, the images are divided along these lines to obtain two image portions to be stitched together. Then, these two large image portions are stitched together to form a complete initial stitched image. Finally, to make the stitched image visually more natural and coherent, a uniform lighting and color matching method is used to eliminate any seams that may exist in the initial stitched images, resulting in a seamless and natural stitching effect.

[0076] This embodiment also proposes a machine learning-based system for identifying and analyzing soil organic matter in vineyards, including:

[0077] The data acquisition module is used to collect environmental data, remote sensing data, and actual organic matter measurement data of the vineyard, and to preprocess the remote sensing data to obtain preprocessed images.

[0078] The feature extraction module, connected to the data acquisition module, is used to extract features from the preprocessed image to obtain spectral features;

[0079] The model building module is connected to the data acquisition module and the feature extraction module respectively. It is used to build an initial organic matter estimation model. The initial organic matter estimation model is trained by environmental data, spectral features and actual organic matter measurements to obtain the organic matter estimation model.

[0080] The assessment module, connected to the model building module, is used to conduct organic matter assessment and analysis of vineyard soil based on the organic matter assessment model, and to obtain the results of organic matter distribution in the vineyard.

[0081] Furthermore, the data acquisition module includes an acquisition submodule and a preprocessing submodule;

[0082] The data acquisition submodule is used to collect environmental data, remote sensing data, and actual measurement data of organic matter in the vineyard.

[0083] The preprocessing submodule is used to preprocess remote sensing data.

[0084] Furthermore, the preprocessing submodule includes a cloud shadow removal unit, a radiometric calibration unit, an atmospheric correction unit, a proximity effect correction unit, an image enhancement unit, and a verification unit;

[0085] The cloud shadow removal unit is used to remove cloud shadows from remotely sensed images;

[0086] The radiometric calibration unit is used to perform radiometric calibration on remote sensing images after removing cloud shadows to obtain satellite images with apparent radiance, and to calculate the apparent reflectance of each pixel in the satellite images with apparent radiance.

[0087] The atmospheric correction unit is used to correct atmospheric absorption and intrinsic atmospheric reflectance in satellite images that reflect reflectance, thereby obtaining atmospherically corrected images.

[0088] The proximity effect correction unit is used to calculate the equivalent average background reflectance of each pixel in the atmospheric corrected image, and to perform proximity effect correction on the satellite image pixel by pixel based on the equivalent average background reflectance to obtain the final corrected image.

[0089] The image enhancement unit is used to perform image enhancement processing based on the final corrected image;

[0090] The verification unit is used to verify the enhanced image using manually sampled data.

[0091] Furthermore, the evaluation module includes a splicing submodule and a recognition submodule;

[0092] The stitching submodule is used to stitch together vineyard images to obtain a stitched image;

[0093] The identification submodule is used to analyze the stitched images using an organic matter estimation model to obtain the distribution results of organic matter in the vineyard;

[0094] The stitching submodule includes a stitching line acquisition unit, an image segmentation unit, a stitching unit, and a stitching seam elimination unit;

[0095] The splicing line acquisition unit is used to search for identical lines from two images to be spliced ​​as splicing lines;

[0096] The image segmentation unit is used to segment the image along the stitching line to obtain the image to be stitched;

[0097] The stitching unit is used to stitch together two large-area images to obtain an initial stitched image;

[0098] The seam elimination unit is used to eliminate seams in the initial stitched image by using a uniform light and color method.

[0099] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A machine learning-based method for identifying and analyzing organic matter in vineyard soil, characterized in that, Includes the following steps: Collect environmental data, remote sensing data, and actual organic matter measurement data of the vineyard; preprocess the remote sensing data to obtain a preprocessed image. The method for preprocessing the remote sensing data includes: Remove cloud shadows from remote sensing images; Radiometric calibration is performed on the remote sensing image after removing cloud shadows to obtain the apparent radiance satellite image, and the apparent reflectance corresponding to each pixel in the apparent radiance satellite image is calculated. Atmospherically absorbed and intrinsic atmospheric reflectance are corrected for satellite images that exhibit reflectivity to obtain atmospherically corrected images. The equivalent average background reflectance of each pixel in the atmospheric corrected image is calculated, and the satellite image is corrected pixel by pixel based on the equivalent average background reflectance to obtain the final corrected image. Image enhancement processing is performed based on the final corrected image, and the enhanced image is verified using manually sampled data; The verification method involves PSF detection of the image: a direct detection method based on the physical definition of the point spread function is used. A parametric point spread model is employed to fit the surface of each sub-pixel target response data to determine the peak position of each target response. Then, the point spread response of sub-pixel interpolation is obtained by position registration of the sub-pixel target response values ​​or count values ​​of a 3*3 array with non-integer pixel intervals. The point spread function of the optical remote sensing satellite imaging system is obtained by fitting a parametric Gaussian model. The obtained system point spread function is subjected to discrete Fourier transform, and the modulus transfer function of the optical remote sensing satellite imaging system is obtained by taking the modulus and normalizing it. The image quality is evaluated by the modulation transfer function. Feature extraction is performed on the preprocessed image to obtain spectral features; The method for feature extraction of the preprocessed image includes: The preprocessed image is denoised, and then the spectrum of all pixels is averaged to obtain the average spectrum of the image. By performing a first-order differential on the spectrum of each pixel, steep changes in the spectral curve can be obtained; The second derivative of the spectrum of each pixel is taken to obtain the bending points in the spectral curve; The coefficients of the tassel transformation are calculated based on the average spectrum and its first and second derivatives, and the coefficients of the tassel transformation are used as spectral features. An initial organic matter estimation model is constructed, and the initial organic matter estimation model is trained using environmental data, spectral characteristics, and actual organic matter measurements to obtain an organic matter estimation model. Based on the organic matter estimation model, an organic matter assessment and analysis was conducted on the vineyard soil to obtain the results of organic matter distribution in the vineyard. The initial organic matter estimation model is a network model based on support vector machines; Methods for obtaining the distribution results of organic matter in vineyards include: stitching together images of vineyards, and then analyzing the stitched images using an organic matter estimation model to obtain the distribution results of organic matter in vineyards; The method for stitching together vineyard images includes: Search for identical lines in two images to be stitched together as stitching lines; segment the image along the stitching lines to obtain images to be stitched together; stitch the two large-area images to be stitched together to obtain an initial stitched image; eliminate the stitching seams in the initial stitched image using a uniform light and color method.

2. A machine learning-based system for identifying and analyzing organic matter in vineyard soil, characterized in that, include: The data acquisition module is used to collect environmental data, remote sensing data, and actual measurement data of organic matter in the vineyard, and to preprocess the remote sensing data to obtain a preprocessed image. The data acquisition module includes an acquisition submodule and a preprocessing submodule; The acquisition submodule is used to collect environmental data, remote sensing data and actual measurement data of organic matter in the vineyard; The preprocessing submodule is used to preprocess the remote sensing data; The preprocessing submodule includes a cloud shadow removal unit, a radiometric calibration unit, an atmospheric correction unit, a proximity effect correction unit, an image enhancement unit, and a verification unit; The cloud shadow removal unit is used to remove cloud shadows in remote sensing images; The radiometric calibration unit is used to perform radiometric calibration on remote sensing images with cloud shadows removed to obtain apparent radiance satellite images, and to calculate the apparent reflectance corresponding to each pixel in the apparent radiance satellite images. The atmospheric correction unit is used to correct atmospheric absorption and intrinsic atmospheric reflectivity in satellite images that reflect reflectivity, thereby obtaining atmospherically corrected images. The proximity effect correction unit is used to calculate the equivalent average background reflectance corresponding to each pixel in the atmospheric corrected image, and to perform proximity effect correction on the satellite image pixel by pixel based on the equivalent average background reflectance to obtain the final corrected image. The image enhancement unit is used to perform image enhancement processing based on the final corrected image; The verification unit is used to verify the enhanced image using manually sampled data; The verification method involves PSF detection of the image: a direct detection method based on the physical definition of the point spread function is used. A parametric point spread model is employed to fit the surface of each sub-pixel target response data to determine the peak position of each target response. Then, the point spread response of sub-pixel interpolation is obtained by position registration of the sub-pixel target response values ​​or count values ​​of a 3*3 array with non-integer pixel intervals. The point spread function of the optical remote sensing satellite imaging system is obtained by fitting a parametric Gaussian model. The obtained system point spread function is subjected to discrete Fourier transform, and the modulus transfer function of the optical remote sensing satellite imaging system is obtained by taking the modulus and normalizing it. The image quality is evaluated by the modulation transfer function. The feature extraction module, connected to the data acquisition module, is used to extract features from the preprocessed image to obtain spectral features; The method for feature extraction of the preprocessed image includes: The preprocessed image is denoised, and then the spectrum of all pixels is averaged to obtain the average spectrum of the image. By performing a first-order differential on the spectrum of each pixel, steep changes in the spectral curve can be obtained; The second derivative of the spectrum of each pixel is taken to obtain the bending points in the spectral curve; The coefficients of the tassel transformation are calculated based on the average spectrum and its first and second derivatives, and the coefficients of the tassel transformation are used as spectral features. The model building module is connected to the data acquisition module and the feature extraction module respectively, and is used to build an initial organic matter estimation model. The initial organic matter estimation model is trained by environmental data, spectral features and actual organic matter measurements to obtain an organic matter estimation model. Based on the organic matter estimation model, an organic matter assessment and analysis was conducted on the vineyard soil to obtain the results of organic matter distribution in the vineyard. The initial organic matter estimation model is a network model based on support vector machines; The assessment module, connected to the model building module, is used to conduct organic matter assessment and analysis on vineyard soil based on the organic matter estimation model, and to obtain the distribution results of organic matter in the vineyard. The evaluation module includes a splicing submodule and a recognition submodule; The stitching submodule is used to stitch together vineyard images to obtain a stitched image; The identification submodule is used to analyze the spliced ​​image through an organic matter estimation model to obtain the distribution results of organic matter in the vineyard; The splicing submodule includes a splicing line acquisition unit, an image segmentation unit, a splicing unit, and a splicing seam elimination unit; The splicing line acquisition unit is used to search for the same line in two images to be spliced ​​as the splicing line; The image segmentation unit is used to segment the image along the splicing line to obtain the image to be spliced. The stitching unit is used to stitch two large-area images to be stitched together to obtain an initial stitched image; The seam elimination unit is used to eliminate seams in the initial stitched image by using a uniform light and color method.

Images