Soil organic matter content monitoring method and system based on drone and hyperspectral

Abstract

The present invention discloses a soil organic matter content monitoring method and system based on unmanned aerial vehicles and hyperspectral, and relates to the technical field of soil organic matter content monitoring, in order to solve the problem of inaccurate monitoring results when a hyperspectral camera monitors the organic matter content of soil. The present invention utilizes computers and image processing technology to quickly complete the processing and analysis of a large amount of image data, and realizes real-time or rapid monitoring of the soil in the monitoring area. By comparing statistical indicators such as standard deviation and correlation, characteristic bands that are significantly different from the spectral characteristics of the monitored soil are screened out. This method not only reduces redundant information, but also highlights the most critical characteristic information for soil classification and identification. The principal component analysis method is used to perform dimensionality reduction processing on the modeling data set, and the most important characteristic vectors are extracted as the principal components, thereby improving the calculation efficiency and prediction accuracy of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of soil organic matter content monitoring, and in particular to a soil organic matter content monitoring method and system based on unmanned aerial vehicle and hyperspectral. Background Art

[0002] Soil organic matter content monitoring refers to the regular or irregular measurement and analysis of the total amount or relative content of organic matter in the soil.

[0003] The Chinese patent with announcement number CN114019082B discloses a soil organic matter content monitoring method and system, which mainly constructs a SOM regression estimation model by using a variable space iterative shrinkage algorithm combined with an extreme random tree, and then uses the constructed SOM regression estimation model to obtain the organic matter content of each pixel point on the soil hyperspectral image, and inverts to obtain an organic matter content distribution map, which can improve the timeliness and accuracy of soil organic matter content monitoring while reducing the monitoring workload and research costs, thereby realizing the rapid monitoring of crops and natural resources by drones at a customized spatiotemporal scale, and providing data support for precision agriculture. Although the above patent solves the problem of soil organic matter content monitoring, there are still the following problems in actual operation:

[0004] 1. Before monitoring the organic matter content in the soil, the drone and camera equipment were not further inspected, resulting in damage to the equipment and inability to monitor the soil.

[0005] 2. After the hyperspectral camera captures the soil image, it does not perform more sophisticated image processing and spectral analysis, resulting in the inability to obtain the soil characteristic bands in the captured image.

[0006] 3. The soil organic matter content data obtained by the hyperspectral camera was not further analyzed and compared, resulting in inaccurate monitoring results. Summary of the invention

[0007] The purpose of the present invention is to provide a soil organic matter content monitoring method and system based on unmanned aerial vehicles and hyperspectral. By using computers and image processing technology, a large amount of image data processing and analysis can be completed quickly, and real-time or rapid monitoring of the soil in the monitoring area can be achieved. By comparing statistical indicators such as standard deviation and correlation, characteristic bands with significant differences in the spectral characteristics of the monitored soil are screened out. This method not only reduces redundant information, but also highlights the most critical characteristic information for soil classification and identification. The principal component analysis method is used to reduce the dimension of the modeling data set, and the most important eigenvectors are extracted as the principal components, which improves the calculation efficiency and prediction accuracy of the model. The partial least squares regression method is used to establish a prediction model between organic matter content and spectral reflectance, which can handle the complex relationship between independent variables and dependent variables and can solve the problems in the prior art.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] The soil organic matter content monitoring method based on drones and hyperspectral includes the following steps:

[0010] S1: Confirmation of drone and camera equipment: Confirm the type of drone and camera, and debug the drone and camera before monitoring;

[0011] S2: Camera soil image capture: The UAV flies along a specified path in the monitoring area, and the camera captures images in the monitoring area. The captured images are annotated with information, and the target soil image is obtained after the information annotation is completed;

[0012] S3: Capture image preprocessing: perform image preprocessing on the target soil image, extract characteristic bands after image preprocessing, and mark the characteristic bands extracted from the target soil image as standard soil images;

[0013] S4: Preprocessing image model construction: construct an organic matter content prediction model for the standard soil image, and obtain the target soil model after the model is constructed;

[0014] S5: Construct model organic matter content distribution: extract soil organic matter content data in the target soil model, compare the extracted soil organic matter content data with the soil in the monitoring area, and display the soil organic matter content distribution of the comparison data.

[0015] Preferably, the drone type and camera type are confirmed in S1, and the drone and camera are debugged before monitoring, including:

[0016] The drone type is a multi-rotor drone, the camera type is a hyperspectral camera, and the multi-rotor drone and the hyperspectral camera are compatible;

[0017] Before using drones and cameras to monitor soil organic matter content, the drones are first subjected to flight inspections, flight tests, and flight parameter adjustments;

[0018] Then conduct connection test, parameter setting and calibration of the camera;

[0019] After the drone and camera are debugged, they are ready to monitor the soil organic matter content.

[0020] Preferably, in S2, the drone flies along a prescribed path in the monitoring area, the camera captures images in the monitoring area, and the captured images are annotated with information, including:

[0021] The UAV platform confirms the path data of the monitoring area and imports the confirmed path data into the UAV. The UAV then flies in the monitoring area according to the path data. The flight path of the UAV is planned according to the topography, landform and soil characteristics of the monitoring area.

[0022] When the drone flies over the monitoring area, the hyperspectral camera captures the soil in the monitoring area;

[0023] The hyperspectral camera automatically captures soil images of the monitoring area, and at the same time, it can be set to continuously capture or timed capture images according to monitoring requirements;

[0024] Finally, the captured soil images are annotated with coordinate system information, and the target soil image is obtained after the coordinate system information is annotated.

[0025] Preferably, the setting of continuous or timed image capture according to monitoring requirements also includes:

[0026] Extracting the image capture position corresponding to the current image capture;

[0027] Real-time monitoring of the distance between the current UAV travel position and the image capture position;

[0028] Compare the distance between the current UAV travel position and the image capture position with a preset first distance threshold;

[0029] When the distance between the current UAV travel position and the image capture position reaches a preset first distance threshold, extracting a capture mode of the current image capture, wherein the capture mode includes continuous capture and timed capture;

[0030] When the capture mode of the current image capture is continuous capture, the first speed reduction adjustment is performed, wherein the target flight speed of the drone corresponding to the first speed reduction adjustment is obtained by the following formula:

[0031]

[0032] Among them, V m Indicates the target flight speed of the drone. At the same time, after calculation, V m Less than or equal to V min When V m 1.63V min ; V0 represents the current speed of the drone; T c Indicates the preset capture and collection interval duration; T d Indicates the image capture time interval corresponding to continuous capture; V min Indicates the preset minimum flight speed to ensure the UAV's travel efficiency;

[0033] Adjusting the flight speed of the UAV according to the target flight speed;

[0034] When the drone reaches the image capturing position, a secondary speed adjustment is performed on the traveling speed of the drone during the image capturing process according to the current image capturing mode of the drone.

[0035] Preferably, when the drone reaches the image capturing position, a secondary speed adjustment is performed on the traveling speed of the drone during the image capturing process according to the current image capturing mode of the drone, including:

[0036] Real-time monitoring of the drone’s movement location;

[0037] When the traveling position of the drone reaches the image capturing position, extracting the image capturing time interval corresponding to the continuous image capturing;

[0038] The speed adjustment coefficient is obtained according to the image capture time interval combined with the current traveling speed of the UAV; wherein the speed adjustment coefficient is obtained by the following formula:

[0039]

[0040] Where, v represents the speed adjustment coefficient; T d Indicates the image capture time interval corresponding to continuous capture; V m Indicates the target flight speed of the drone; L min and L max Respectively represent the minimum distance and the maximum distance corresponding to the reference distance value range between the positions corresponding to each two adjacent capture actions in the preset continuous capture mode;

[0041] The speed adjustment coefficient is used to perform secondary speed adjustment on the UAV, wherein the adjusted travel speed corresponding to the secondary speed adjustment is obtained by the following formula:

[0042]

[0043] Among them, V t Indicates the adjusted travel speed corresponding to the secondary speed adjustment; V m Indicates the target flight speed of the UAV, V min It indicates the preset minimum flight speed to ensure the UAV's travel efficiency; v indicates the speed adjustment coefficient.

[0044] Preferably, the target soil image is preprocessed in S3, and characteristic bands are extracted after the image preprocessing, including:

[0045] The image preprocessing of the target soil image is to sequentially perform radiometric conversion, geometric correction, atmospheric correction and reflectivity conversion on the target soil image;

[0046] The image preprocessing process is as follows:

[0047] The target soil image is converted into a radiance, the radiance conversion is to obtain pixel values ​​in the image using a linear formula, and the pixel values ​​are converted into radiance values, wherein the radiance values ​​include a gain parameter and an offset parameter;

[0048] After the radiometric conversion is completed, geometric correction is performed. The geometric correction includes rough geometric correction and fine geometric correction. The rough geometric correction is to correct the geometric distortion in the target soil image, and the fine geometric correction is to correct the target soil image using the ground control points in the monitoring area.

[0049] After the geometric correction is completed, atmospheric correction is performed. Atmospheric correction is to correct the target soil image using the 6S model, where the 6S model is retrieved from the database;

[0050] After the atmospheric correction, reflectivity conversion is performed, and the reflectivity conversion is to convert the radiant brightness value after the atmospheric correction into a reflectivity value;

[0051] Finally, the target soil image is obtained after image preprocessing.

[0052] Preferably, the target soil image is preprocessed in S3, and characteristic bands are extracted after the image preprocessing, further comprising:

[0053] The spectral characteristics of the soil in the monitoring area are retrieved from the database, and the reflection and absorption characteristics of the soil in the monitoring area in different bands are obtained according to the spectral characteristics;

[0054] Convert the target soil image after image preprocessing into band data;

[0055] Extract characteristic bands from the converted band data;

[0056] Among them, the characteristic band refers to the data whose standard deviation and correlation are not within the comparison range when the band data is compared with the reflection and absorption characteristic data of the monitored soil in different bands;

[0057] The characteristic band data in the acquired target soil image is labeled as a standard soil image.

[0058] Preferably, constructing an organic matter content prediction model for the standard soil image in S4 includes:

[0059] Obtain the actual organic matter content sampling data of the soil in the monitoring area from the database;

[0060] Match the characteristic band data in the standard soil image with the actual organic matter content sampling data to form a modeling data set after data matching;

[0061] The prediction model between actual organic matter content and spectral reflectance was established using principal component analysis combined with partial least squares regression and modeling data sets;

[0062] Among them, the covariance matrix of the modeling data set is first calculated, and the calculated covariance matrix is ​​subjected to eigendecomposition, and the eigenvalues ​​and eigenvectors are obtained after the eigendecomposition;

[0063] Select the eigenvectors corresponding to several largest eigenvalues ​​as principal components, project the modeling data set into the principal components, and obtain the reduced-dimensional modeling data set matrix after the projection is completed.

[0064] Preferably, the step of constructing an organic matter content prediction model for the standard soil image in S4 further includes:

[0065] Extract the first pair of components from the organic matter content sampling data and the characteristic band data respectively;

[0066] The first pair of extracted components are regressed through the regression model;

[0067] Among them, when the first pair of components is regressed through the regression model, if the regression accuracy is not within the standard range, the second pair of components are extracted from the organic matter content sampling data and the characteristic band data respectively, and the regression is performed again;

[0068] The soil content prediction model was obtained based on the regression model of the modeling dataset matrix and components.

[0069] The soil organic matter content monitoring system based on drones and hyperspectral includes:

[0070] Soil organic matter content distribution display module is used for:

[0071] Divide the modeling dataset into training and testing sets;

[0072] The training set is imported into the soil content prediction model for training, and after the training is completed, a trained soil content prediction model is obtained;

[0073] The test set is input into the trained soil content prediction model to obtain the predicted value of soil organic matter content, and the predicted value is compared with the actual soil organic matter content value;

[0074] The comparison results are matched according to the coordinate system information, and the comparison results and the monitored organic matter content data in each coordinate system of the monitoring area are displayed on the display terminal in the form of a chart.

[0075] Compared with the prior art, the present invention has the following beneficial effects:

[0076] 1. The soil organic matter content monitoring method and system based on drone and hyperspectral provided by the present invention can greatly reduce the risks during the flight by conducting detailed drone flight inspection, testing and parameter adjustment before monitoring. The camera is connected and tested, and parameters are set and calibrated to ensure that the camera can work stably during the flight. The combination of drone and hyperspectral camera realizes the automatic image capture function, and continuous capture or timed capture can be set according to monitoring requirements, which reduces manual intervention and improves work efficiency.

[0077] 2. The soil organic matter content monitoring method and system based on drone and hyperspectral provided by the present invention can quickly complete the processing and analysis of a large amount of image data by using computers and image processing technology, realize real-time or rapid monitoring of the soil in the monitoring area, and screen out characteristic bands with significant differences from the spectral characteristics of the monitored soil by comparing statistical indicators such as standard deviation and correlation. This method not only reduces redundant information, but also highlights the characteristic information that is most critical for soil classification and identification.

[0078] 3. The soil organic matter content monitoring method and system based on drone and hyperspectral provided by the present invention uses principal component analysis to reduce the dimension of the modeling data set, extracts the most important eigenvector as the principal component, and improves the calculation efficiency and prediction accuracy of the model. At the same time, feature extraction helps to reveal the intrinsic connection and law between soil organic matter content and spectral reflectance, and uses partial least squares regression to establish a prediction model between organic matter content and spectral reflectance, which can handle the complex relationship between independent variables and dependent variables, including linear and nonlinear relationships. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is a schematic diagram of the soil organic matter content monitoring steps of the present invention;

[0080] Figure 2 It is a schematic diagram of the soil organic matter content monitoring process of the present invention. DETAILED DESCRIPTION

[0081] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0082] In order to solve the problem in the prior art that the drone and camera equipment are not further inspected before the soil organic matter content is monitored, which results in equipment damage and inability to monitor the soil, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0083] The soil organic matter content monitoring method based on drones and hyperspectral includes the following steps:

[0084] S1: Confirmation of drone and camera equipment: Confirm the type of drone and camera, and debug the drone and camera before monitoring;

[0085] Among them, detailed drone flight inspection, testing and parameter adjustment before monitoring can greatly reduce the risks during the flight and improve the safety of operations;

[0086] S2: Camera soil image capture: The UAV flies along a specified path in the monitoring area, and the camera captures images in the monitoring area. The captured images are annotated with information, and the target soil image is obtained after the information annotation is completed;

[0087] The combination of drones and hyperspectral cameras realizes the automatic image capture function, which can set continuous capture or timed capture according to monitoring needs, reducing manual intervention and improving work efficiency.

[0088] S3: Capture image preprocessing: perform image preprocessing on the target soil image, extract characteristic bands after image preprocessing, and mark the characteristic bands extracted from the target soil image as standard soil images;

[0089] Among them, the use of computers and image processing technology can quickly complete the processing and analysis of a large amount of image data, realizing real-time or rapid monitoring of the soil in the monitoring area;

[0090] S4: Preprocessing image model construction: construct an organic matter content prediction model for the standard soil image, and obtain the target soil model after the model is constructed;

[0091] Among them, by predicting the soil organic matter content in different regions, more accurate and effective agricultural management measures such as fertilization, irrigation and crop planting can be formulated to improve agricultural production efficiency and output;

[0092] S5: Constructing the model organic matter content distribution: extracting the soil organic matter content data in the target soil model, comparing the extracted soil organic matter content data with the soil in the monitoring area, and displaying the soil organic matter content distribution of the comparison data;

[0093] Among them, the charts can clearly display key information such as the distribution of soil organic matter content in different monitoring areas, the difference between predicted values ​​and actual values, etc., which helps researchers or decision makers to quickly grasp the soil quality status.

[0094] Confirm the drone type and camera type in S1, and debug the drone and camera before monitoring, including:

[0095] The drone type is a multi-rotor drone, the camera type is a hyperspectral camera, and the multi-rotor drone and the hyperspectral camera are compatible;

[0096] Before using drones and cameras to monitor soil organic matter content, the drones are first subjected to flight inspections, flight tests, and flight parameter adjustments;

[0097] Then conduct connection test, parameter setting and calibration of the camera;

[0098] After the drone and camera are debugged, they are ready to monitor the soil organic matter content.

[0099] Specifically, the hyperspectral camera can capture the reflectance of the soil surface in multiple continuous and narrow spectral bands, which makes the monitoring of soil organic matter content more accurate. Hyperspectral data can provide rich spectral information, which helps to distinguish different components in the soil, including organic matter. Multi-rotor drones have flexible flight capabilities and high spatial resolution, and can quickly cover large areas and improve monitoring efficiency. Its stable flight platform also ensures that the camera can obtain high-quality data. By making the multi-rotor drone compatible with the hyperspectral camera, seamless integration between the two is ensured. This compatibility enables the drone to stably carry the camera for flight operations, and the camera can accurately record the required spectral information. Detailed drone flight inspection, testing and parameter adjustment before monitoring can greatly reduce the risks during the flight process and improve the safety of operations. The connection test, parameter setting and calibration of the camera ensure that the camera can work stably during the flight and provide accurate and reliable data.

[0100] In S2, the drone flies along a specified path in the monitoring area, and the camera captures images in the monitoring area, and annotates the captured images with information, including:

[0101] The UAV platform confirms the path data of the monitoring area and imports the confirmed path data into the UAV. The UAV then flies in the monitoring area according to the path data. The flight path of the UAV is planned according to the topography, landform and soil characteristics of the monitoring area.

[0102] When the drone flies over the monitoring area, the hyperspectral camera captures the soil in the monitoring area;

[0103] The hyperspectral camera automatically captures soil images of the monitoring area, and at the same time, it can be set to continuously capture or timed capture images according to monitoring requirements;

[0104] Finally, the captured soil images are annotated with coordinate system information, and the target soil image is obtained after the coordinate system information is annotated.

[0105] Specifically, drones can fly quickly and flexibly in the monitoring area according to a preset path, covering a wide range of hard-to-reach ground areas. This capability greatly improves the efficiency of data collection, especially in environments with complex terrain, inaccessible or dangerous environments. Hyperspectral cameras can capture high-resolution images of the soil and provide rich spectral information. This information is crucial for applications such as soil type identification, quantitative analysis of soil composition, and soil pollution detection, ensuring high accuracy and reliability of data. The combination of drones and hyperspectral cameras realizes automated image capture, which can set continuous capture or timed capture according to monitoring needs, reducing manual intervention and improving work efficiency. Drones can transmit image data in real time, allowing monitoring work to respond quickly to environmental changes.

[0106] Specifically, the setting of continuous or timed image capture according to monitoring requirements also includes:

[0107] Extracting the image capture position corresponding to the current image capture;

[0108] Real-time monitoring of the distance between the current UAV travel position and the image capture position;

[0109] Compare the distance between the current UAV travel position and the image capture position with a preset first distance threshold;

[0110] When the distance between the current UAV travel position and the image capture position reaches a preset first distance threshold, extracting a capture mode of the current image capture, wherein the capture mode includes continuous capture and timed capture;

[0111] When the capture mode of the current image capture is continuous capture, the first speed reduction adjustment is performed, wherein the target flight speed of the drone corresponding to the first speed reduction adjustment is obtained by the following formula:

[0112]

[0113] Among them, V m Indicates the target flight speed of the drone. At the same time, after calculation, V m Less than or equal to V min When V m 1.63V min; V0 represents the current speed of the drone; T c Indicates the preset capture and collection interval duration; T d Indicates the image capture time interval corresponding to continuous capture; V min Indicates the preset minimum flight speed to ensure the UAV's travel efficiency;

[0114] Adjusting the flight speed of the UAV according to the target flight speed;

[0115] When the drone reaches the image capturing position, a secondary speed adjustment is performed on the traveling speed of the drone during the image capturing process according to the current image capturing mode of the drone.

[0116] The technical effect of the above technical solution is: by real-time monitoring the distance between the UAV and the preset image capture position, and when the preset first distance threshold is reached, the speed is adjusted accordingly according to the capture mode (continuous capture or timed capture), thereby ensuring that the UAV can accurately capture images at the specified position, thereby improving the accuracy and reliability of image capture.

[0117] When approaching the image capture position, the first deceleration adjustment allows the drone to approach the target at a speed that is neither too fast nor too slow, avoiding the problem of missing the capture opportunity due to too fast speed or wasting energy and time due to too slow speed. At the same time, by setting the minimum flight speed V to ensure the drone's travel efficiency min , and calculated the target flight speed V m Lower than V min When the voltage is adjusted to 1.63V min , ensuring that the drone maintains efficient movement while also ensuring sufficient energy reserves and stability.

[0118] The technical solution supports two modes: continuous capture and timed capture. According to different mission requirements, the appropriate capture mode can be flexibly selected. This flexibility enables the drone to perform at its best in a variety of application scenarios.

[0119] When the drone reaches the image capture position, a secondary speed adjustment is performed according to the current capture mode, which further improves the accuracy and efficiency of image capture. Through intelligent speed adjustment strategies, drones can automatically adjust their speed at different capture stages to adapt to different capture requirements and environmental changes. Through accurate and efficient image capture and intelligent speed adjustment, drones can better complete image capture tasks and improve user experience. Whether it is scientific research, surveying and mapping, environmental monitoring or other fields that require image capture, they can all benefit from it.

[0120] In summary, this technical solution significantly improves the efficiency and accuracy of UAVs in image capture tasks through precise position monitoring, intelligent speed adjustment, and flexible capture mode support, and has important application value and technical significance.

[0121] Specifically, when the drone reaches the image capturing position, a secondary speed adjustment is performed on the traveling speed of the drone during the image capturing process according to the current image capturing mode of the drone, including:

[0122] Real-time monitoring of the drone’s movement location;

[0123] When the traveling position of the drone reaches the image capturing position, extracting the image capturing time interval corresponding to the continuous image capturing;

[0124] The speed adjustment coefficient is obtained according to the image capture time interval combined with the current traveling speed of the UAV; wherein the speed adjustment coefficient is obtained by the following formula:

[0125]

[0126] Where, v represents the speed adjustment coefficient; T d Indicates the image capture time interval corresponding to continuous capture; V m Indicates the target flight speed of the drone; L min and L max Respectively represent the minimum distance and the maximum distance corresponding to the reference distance value range between the positions corresponding to each two adjacent capture actions in the preset continuous capture mode;

[0127] The speed adjustment coefficient is used to perform secondary speed adjustment on the UAV, wherein the adjusted travel speed corresponding to the secondary speed adjustment is obtained by the following formula:

[0128]

[0129] Among them, V t Indicates the adjusted travel speed corresponding to the secondary speed adjustment; V m Indicates the target flight speed of the UAV, V min It indicates the preset minimum flight speed to ensure the UAV's travel efficiency; v indicates the speed adjustment coefficient.

[0130] The technical effect of the above technical solution is: when the UAV reaches the image capture position, a secondary speed adjustment is performed according to the specific requirements of the current capture mode (continuous capture or timed capture). This adjustment ensures that the UAV can move at the most suitable speed during the capture process, thereby reducing the problem of image blur caused by too fast speed or low capture efficiency caused by too slow speed, and improving the quality and stability of image capture. By introducing the speed adjustment coefficient v and combining the image capture time interval T d 、The target flight speed V of the UAV m And the preset reference distance value range (L min and L max ), the technical solution can more accurately control the speed of the drone. This optimization allows the drone to maintain efficient capture while avoiding unnecessary energy consumption and improving resource utilization efficiency.

[0131] The technical solution takes into account the reference distance value range (L min and L max ), which enables the drone to flexibly adjust its speed according to the actual capture requirements and environmental conditions. This enhanced adaptability enables the drone to maintain good capture results in a variety of complex environments. By setting the minimum flight speed V to ensure the drone's travel efficiency min , and ensure that the adjusted travel speed V t Not less than V min The technical solution effectively avoids the possibility of a drone crash or other safety issues caused by a low speed. This safety protection measure improves the reliability and service life of the drone. Through accurate, flexible and efficient secondary speed adjustment strategies, drones can better meet users' needs for image capture quality and efficiency. Whether it is scientific research, surveying and mapping, environmental monitoring or other fields that require high-precision image capture, users can get better experience and results.

[0132] In summary, this technical solution achieves efficient, high-quality and safe capture effects in the image capture process through real-time monitoring, precise calculation and flexible adjustment of the drone's travel speed, and has important application value and technical significance.

[0133] In order to solve the problem in the prior art that the hyperspectral camera does not perform more sophisticated image processing and spectral analysis after capturing the soil image, which results in the inability to obtain the soil characteristic bands in the captured image, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0134] The target soil image is preprocessed in S3, and characteristic bands are extracted after image preprocessing, including:

[0135] The image preprocessing of the target soil image is to sequentially perform radiometric conversion, geometric correction, atmospheric correction and reflectivity conversion on the target soil image;

[0136] The image preprocessing process is as follows:

[0137] The target soil image is converted into a radiance, the radiance conversion is to obtain pixel values ​​in the image using a linear formula, and the pixel values ​​are converted into radiance values, wherein the radiance values ​​include a gain parameter and an offset parameter;

[0138] After the radiometric conversion is completed, geometric correction is performed. The geometric correction includes rough geometric correction and fine geometric correction. The rough geometric correction is to correct the geometric distortion in the target soil image, and the fine geometric correction is to correct the target soil image using the ground control points in the monitoring area.

[0139] After the geometric correction is completed, atmospheric correction is performed. Atmospheric correction is to correct the target soil image using the 6S model, where the 6S model is retrieved from the database;

[0140] After the atmospheric correction, reflectivity conversion is performed, and the reflectivity conversion is to convert the radiant brightness value after the atmospheric correction into a reflectivity value;

[0141] Finally, the target soil image is obtained after image preprocessing.

[0142] The spectral characteristics of the soil in the monitoring area are retrieved from the database, and the reflection and absorption characteristics of the soil in the monitoring area in different bands are obtained according to the spectral characteristics;

[0143] Convert the target soil image after image preprocessing into band data;

[0144] Extract characteristic bands from the converted band data;

[0145] Among them, the characteristic band refers to the data whose standard deviation and correlation are not within the comparison range when the band data is compared with the reflection and absorption characteristic data of the monitored soil in different bands;

[0146] The characteristic band data in the acquired target soil image is labeled as a standard soil image.

[0147] Specifically, by converting image pixel values ​​into radiation brightness values ​​through linear formulas and taking into account gain parameters and offset parameters, the impact of sensor characteristics on data can be reduced, thereby improving data accuracy. Atmospheric correction methods such as the 6S model can be used to eliminate the impact of atmospheric absorption, scattering and other factors on image data, making subsequent analysis more accurate. Geometric coarse correction and geometric fine correction can be used to eliminate geometric distortions in images, such as tilt and distortion, making images more consistent with actual geographic features and improving the visual quality and geographic accuracy of images. By converting radiation brightness values ​​into reflectance values, image information is closer to the actual reflectance characteristics of the surface, facilitating subsequent spectral analysis and application. Reflectance images directly reflect the reflectance characteristics of the surface and are closely related to environmental parameters such as organic matter content, soil moisture, and vegetation cover, making it easy to establish a quantitative relationship model between these parameters and spectral characteristics. By retrieving known spectral characteristic data from a database and comparing them with the band data in the image, the characteristic bands of the target soil can be scientifically identified. This analysis method based on spectral characteristics improves the accuracy of soil classification and identification. By using computers and image processing technology, it can quickly complete the processing and analysis of a large amount of image data, realize real-time or rapid monitoring of soil in the monitoring area, and screen out characteristic bands that are significantly different from the spectral characteristics of the monitored soil by comparing statistical indicators such as standard deviation and correlation. This method not only reduces redundant information, but also highlights the most critical characteristic information for soil classification and identification. By accurately analyzing the spectral characteristics of soil, we can better understand the quality and health of soil, providing a scientific basis for precision agriculture and environmental protection.

[0148] In order to solve the problem that the soil organic matter content data obtained by the hyperspectral camera is not further analyzed and compared with the organic matter content in the existing technology, resulting in inaccurate monitoring results, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0149] The organic matter content prediction model of the standard soil image in S4 is constructed, including:

[0150] Obtain the actual organic matter content sampling data of the soil in the monitoring area from the database;

[0151] Match the characteristic band data in the standard soil image with the actual organic matter content sampling data to form a modeling data set after data matching;

[0152] The prediction model between actual organic matter content and spectral reflectance was established using principal component analysis combined with partial least squares regression and modeling data sets;

[0153] Among them, the covariance matrix of the modeling data set is first calculated, and the calculated covariance matrix is ​​subjected to eigendecomposition, and the eigenvalues ​​and eigenvectors are obtained after the eigendecomposition;

[0154] Select the eigenvectors corresponding to several largest eigenvalues ​​as principal components, project the modeling data set into the principal components, and obtain the reduced-dimensional modeling data set matrix after the projection is completed.

[0155] Extract the first pair of components from the organic matter content sampling data and the characteristic band data respectively;

[0156] The first pair of extracted components are regressed through the regression model;

[0157] Among them, when the first pair of components is regressed through the regression model, if the regression accuracy is not within the standard range, the second pair of components are extracted from the organic matter content sampling data and the characteristic band data respectively, and the regression is performed again;

[0158] The soil content prediction model was obtained based on the regression model of the modeling dataset matrix and components.

[0159] Specifically, the principal component analysis method is used to reduce the dimension of the modeling data set, and the most important eigenvectors are extracted as the principal components. This process reduces the redundancy and noise of the data, and improves the computational efficiency and prediction accuracy of the model. At the same time, feature extraction helps to reveal the intrinsic connection and law between soil organic matter content and spectral reflectance. The partial least squares regression method is used to establish a prediction model between organic matter content and spectral reflectance. This method can handle the complex relationship between independent variables and dependent variables, including linear and nonlinear relationships. In addition, the construction process of the regression model is flexible and can be adjusted and optimized according to the actual situation of the data to improve the prediction accuracy and generalization ability of the model, by iteratively extracting components and performing regression until the regression accuracy reaches the standard range. This iterative process ensures the prediction accuracy and stability of the model, avoids the problem of model failure caused by poor single regression results, and the modeling process can be realized by computers and automated software, reducing manual intervention and errors, and improving work efficiency and accuracy. The constructed organic matter content prediction model can provide a scientific basis and decision support for soil management and agricultural production. By predicting the soil organic matter content in different regions, more precise and effective agricultural management measures such as fertilization, irrigation and crop planting can be formulated to improve agricultural production efficiency and output.

[0160] The soil organic matter content monitoring system based on drones and hyperspectral includes:

[0161] Soil organic matter content distribution display module is used for:

[0162] Divide the modeling dataset into training and testing sets;

[0163] The training set is imported into the soil content prediction model for training, and after the training is completed, a trained soil content prediction model is obtained;

[0164] The test set is input into the trained soil content prediction model to obtain the predicted value of soil organic matter content, and the predicted value is compared with the actual soil organic matter content value;

[0165] The comparison results are matched according to the coordinate system information, and the comparison results and the monitored organic matter content data in each coordinate system of the monitoring area are displayed on the display terminal in the form of a chart.

[0166] Specifically, by dividing the modeling data set into a training set and a test set, it can be ensured that the model can also show good performance on unseen data. This division helps prevent model overfitting, improves the generalization ability of the model, and makes the model more reliable in practical applications. Using the test set to verify the trained model can intuitively see the prediction effect of the model. By comparing the predicted value with the actual value, the error rate, accuracy and other evaluation indicators can be calculated to comprehensively evaluate the performance of the model. The comparison results and monitoring data are displayed in the form of charts on the display terminal, which can make the results more intuitive and easy to understand. The chart can clearly show the distribution of soil organic matter content in different monitoring areas, the difference between predicted values ​​and actual values ​​and other key information, which helps researchers or decision makers quickly grasp the soil quality status and make corresponding decisions. The whole process is based on data analysis and model prediction, realizing the transformation from data to knowledge. By deeply exploring the data characteristics of soil organic matter content, the laws and trends of soil quality changes can be revealed.

[0167] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0168] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. The soil organic matter content monitoring method based on drone and hyperspectral is characterized by: The steps include: S1: Confirmation of drone and camera equipment: Confirm the type of drone and camera, and debug the drone and camera before monitoring; S2: Camera soil image capture: The UAV flies along a specified path in the monitoring area, and the camera captures images in the monitoring area. The captured images are annotated with information, and the target soil image is obtained after the information annotation is completed; S3: Capture image preprocessing: perform image preprocessing on the target soil image, extract characteristic bands after image preprocessing, and mark the characteristic bands extracted from the target soil image as a standard soil image; S4: Preprocessing image model construction: construct an organic matter content prediction model for the standard soil image, and obtain the target soil model after the model is constructed; S5: Constructing the model organic matter content distribution: extracting the soil organic matter content data in the target soil model, comparing the extracted soil organic matter content data with the soil in the monitoring area, and displaying the soil organic matter content distribution of the comparison data; In S2, the drone flies along a specified path in the monitoring area, and the camera captures images in the monitoring area, and annotates the captured images with information, including: The UAV platform confirms the path data of the monitoring area and imports the confirmed path data into the UAV. The UAV then flies in the monitoring area according to the path data. The flight path of the UAV is planned according to the topography, landform and soil characteristics of the monitoring area. When the drone flies over the monitoring area, the hyperspectral camera captures the soil in the monitoring area; The hyperspectral camera automatically captures soil images of the monitoring area, and at the same time, it can be set to continuously capture or timed capture images according to monitoring requirements; Finally, the captured soil images are annotated with coordinate system information, and after the coordinate system information is annotated, the target soil image is obtained; The settings for continuous or timed image capture according to monitoring requirements include: Extracting the image capture position corresponding to the current image capture; Real-time monitoring of the distance between the current UAV travel position and the image capture position; Compare the distance between the current UAV travel position and the image capture position with a preset first distance threshold; When the distance between the current UAV travel position and the image capture position reaches a preset first distance threshold, extracting a capture mode of the current image capture, wherein the capture mode includes continuous capture and timed capture; When the capture mode of the current image capture is continuous capture, the first speed reduction adjustment is performed, wherein the target flight speed of the drone corresponding to the first speed reduction adjustment is obtained by the following formula: Among them, V m Indicates the target flight speed of the drone. At the same time, after calculation, V m Less than or equal to V min When V m is 1.63Vmin; V0 represents the current speed of the drone; T c Indicates the preset capture and collection interval duration; T d Indicates the image capture time interval corresponding to continuous capture; V min Indicates the preset minimum flight speed to ensure the UAV's travel efficiency; Adjusting the flight speed of the UAV according to the target flight speed; When the drone reaches the image capturing position, a secondary speed adjustment is performed on the traveling speed of the drone during the image capturing process according to the current image capturing mode of the drone.

2. The soil organic matter content monitoring method based on drone and hyperspectral according to claim 1 is characterized by: Confirm the drone type and camera type in S1, and debug the drone and camera before monitoring, including: The drone type is a multi-rotor drone, the camera type is a hyperspectral camera, and the multi-rotor drone and the hyperspectral camera are compatible; Before using drones and cameras to monitor soil organic matter content, the drones are first subjected to flight inspections, flight tests, and flight parameter adjustments; Then conduct connection test, parameter setting and calibration of the camera; After the drone and camera are debugged, they are ready to monitor the soil organic matter content.

3. The soil organic matter content monitoring method based on drone and hyperspectral according to claim 2 is characterized in that: When the drone reaches the image capturing position, a secondary speed adjustment is performed on the traveling speed of the drone during the image capturing process according to the current image capturing mode of the drone, including: Real-time monitoring of the drone’s location; When the traveling position of the drone reaches the image capturing position, extracting the image capturing time interval corresponding to the continuous image capturing; The speed adjustment coefficient is obtained according to the image capture time interval combined with the current traveling speed of the UAV; wherein the speed adjustment coefficient is obtained by the following formula: Where, v represents the speed adjustment coefficient; T d Indicates the image capture time interval corresponding to continuous capture; V m Indicates the target flight speed of the drone; L min and L max Respectively represent the minimum distance and the maximum distance corresponding to the reference distance value range between the positions corresponding to each two adjacent capture actions in the preset continuous capture mode; The speed adjustment coefficient is used to perform secondary speed adjustment on the UAV, wherein the adjusted travel speed corresponding to the secondary speed adjustment is obtained by the following formula: Among them, V t Indicates the adjusted travel speed corresponding to the secondary speed adjustment; V m Indicates the target flight speed of the UAV, V min It indicates the preset minimum flight speed to ensure the UAV's travel efficiency; v indicates the speed adjustment coefficient.

4. The soil organic matter content monitoring method based on drone and hyperspectral according to claim 1 is characterized in that: The target soil image is preprocessed in S3, and characteristic bands are extracted after image preprocessing, including: The image preprocessing of the target soil image is to sequentially perform radiometric conversion, geometric correction, atmospheric correction and reflectivity conversion on the target soil image; The image preprocessing process is as follows: The target soil image is converted into a radiance, the radiance conversion is to obtain pixel values ​​in the image using a linear formula, and the pixel values ​​are converted into radiance values, wherein the radiance values ​​include a gain parameter and an offset parameter; After the radiometric conversion is completed, geometric correction is performed. The geometric correction includes rough geometric correction and fine geometric correction. The rough geometric correction is to correct the geometric distortion in the target soil image, and the fine geometric correction is to correct the target soil image using the ground control points in the monitoring area. After the geometric correction is completed, atmospheric correction is performed. Atmospheric correction is to correct the target soil image using the 6S model, where the 6S model is retrieved from the database; After the atmospheric correction, reflectivity conversion is performed, and the reflectivity conversion is to convert the radiance value after the atmospheric correction into the reflectivity value; Finally, the target soil image is obtained after image preprocessing.

5. The soil organic matter content monitoring method based on drone and hyperspectral according to claim 1 is characterized in that: In S3, the target soil image is preprocessed, and characteristic bands are extracted after the image preprocessing, which also includes: The spectral characteristics of the soil in the monitoring area are retrieved from the database, and the reflection and absorption characteristics of the soil in the monitoring area in different bands are obtained according to the spectral characteristics; Convert the target soil image after image preprocessing into band data; Extract characteristic bands from the converted band data; Among them, the characteristic band refers to the data whose standard deviation and correlation are not within the comparison range when the band data is compared with the reflection and absorption characteristic data of the monitored soil in different bands; The characteristic band data in the acquired target soil image is labeled as a standard soil image.

6. The soil organic matter content monitoring method based on drone and hyperspectral according to claim 5 is characterized in that: The organic matter content prediction model of the standard soil image in S4 is constructed, including: Obtain the actual organic matter content sampling data of the soil in the monitoring area from the database; Match the characteristic band data in the standard soil image with the actual organic matter content sampling data to form a modeling data set after data matching; The principal component analysis method combined with partial least squares regression and modeling data set was used to establish a prediction model between actual organic matter content and spectral reflectance; Among them, the covariance matrix of the modeling data set is first calculated, and the calculated covariance matrix is ​​subjected to eigendecomposition, and the eigenvalues ​​and eigenvectors are obtained after the eigendecomposition; Select the eigenvectors corresponding to several largest eigenvalues ​​as principal components, project the modeling data set into the principal components, and obtain the reduced-dimensional modeling data set matrix after the projection is completed.

7. The soil organic matter content monitoring method based on drone and hyperspectral according to claim 6 is characterized in that: The construction of the organic matter content prediction model for the standard soil image in S4 also includes: Extract the first pair of components from the organic matter content sampling data and the characteristic band data respectively; The first pair of extracted components are regressed through the regression model; Among them, when the first pair of components is regressed through the regression model, if the regression accuracy is not within the standard range, the second pair of components are extracted from the organic matter content sampling data and the characteristic band data respectively, and the regression is performed again; The soil content prediction model was obtained based on the regression model of the modeling dataset matrix and components.

8. The soil organic matter content monitoring system based on drone and hyperspectrum is applied in the soil organic matter content monitoring method based on drone and hyperspectrum as claimed in claim 7, characterized in that: include: Soil organic matter content distribution display module is used for: Divide the modeling dataset into training and testing sets; The training set is imported into the soil content prediction model for training, and after the training is completed, a trained soil content prediction model is obtained; The test set is input into the trained soil content prediction model to obtain the predicted value of soil organic matter content, and the predicted value is compared with the actual soil organic matter content value; The comparison results are matched according to the coordinate system information, and the comparison results and the monitored organic matter content data in each coordinate system of the monitoring area are displayed on the display terminal in the form of a chart.

Images