System and method for multispectral online analysis of soil dryness and humidity based on unmanned aerial vehicle

Abstract

The invention relates to the field of agricultural intelligent monitoring, in particular to a system and method for multispectral online analysis of soil dryness and humidity based on an unmanned aerial vehicle, and the system comprises an unmanned aerial vehicle collection module, a multispectral data processing module, an analysis and modeling module and a data storage and display module. The unmanned aerial vehicle acquisition module is responsible for efficiently acquiring multispectral image data, the multispectral data processing module is responsible for performing radiation correction, geometric correction, noise filtering and spectral feature extraction on the acquired data, and the analysis and modeling module processes the multispectral data in real time and predicts soil humidity. The data storage and display module is used for storing historical data, and the soil humidity distribution condition is displayed in real time through a visual interface, so that the problem that key information of soil humidity is still difficult to efficiently and accurately extract from spectral data and accurate prediction of the soil humidity through multispectral data at present is solved. And the problem of how to maintain relatively high prediction precision under different region and environment conditions is solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent agricultural monitoring, and in particular to a system and method for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles. Background Art

[0002] Soil dryness and wetness is one of the important parameters in agricultural production, which directly affects crop growth and water management. Traditional soil moisture detection methods mainly include manual sampling and fixed sensor methods. Manual sampling usually relies on manual sampling and obtains soil moisture data through laboratory analysis or field equipment measurement. Although this method can provide more accurate results, it has disadvantages such as small coverage, long sampling time, and high cost. It is difficult to meet the real-time monitoring needs of large-scale farmland. Fixed soil moisture sensors are used to continuously monitor soil moisture at specific locations, but their limitations are narrow coverage, complex equipment maintenance, and high cost.

[0003] With the advancement of agricultural technology, multispectral imaging technology has been introduced into soil moisture monitoring. Multispectral imaging evaluates soil moisture by analyzing the spectral reflectance characteristics of different bands. This method has the advantages of non-contact, wide coverage and strong real-time performance. Multispectral sensors can reflect the moisture status of the soil surface by capturing information in the visible and near-infrared bands. However, the spectral characteristics of the soil are affected by many factors, such as soil type, vegetation coverage, lighting conditions and surface roughness, resulting in a more complex relationship between soil moisture and spectral reflectance. Existing technologies still find it difficult to efficiently and accurately extract key information about soil moisture from spectral data, especially when environmental factors change greatly, resulting in large errors.

[0004] The rapid development of drone technology has provided a new means for agricultural monitoring. Drones equipped with multispectral sensors can quickly cover large areas and obtain soil moisture data through high-resolution images. Compared with traditional methods, drones can significantly reduce monitoring costs and improve data collection efficiency, but they also face challenges such as high data processing complexity and poor environmental adaptability. Existing drone remote sensing technology mainly focuses on image acquisition and processing, but how to accurately predict soil moisture through multispectral data and how to maintain high prediction accuracy under different regions and environmental conditions are still important issues facing current technology.

[0005] In summary, the present invention provides a method for online analysis of soil dryness and wetness based on unmanned aerial vehicle multi-spectrum to solve the problems raised in the above background technology. Summary of the invention

[0006] The present invention provides a system and method for online analysis of soil dryness and moisture based on unmanned aerial vehicle multispectral technology, which combines the advantages of high maneuverability of unmanned aerial vehicles and multispectral imaging technology, can perform real-time soil moisture analysis on large areas of farmland, and provide precise irrigation management and agricultural resource optimization solutions.

[0007] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a system and method for online analysis of soil dryness and moisture based on unmanned aerial vehicle multispectral, including a unmanned aerial vehicle acquisition module, a multispectral data processing module, an analysis and modeling module, and a data storage and display module. The unmanned aerial vehicle acquisition module is responsible for the efficient acquisition of multispectral image data, and has autonomous flight and path planning capabilities. The multispectral data processing module is responsible for radiation correction, geometric correction, noise filtering and spectral feature extraction of the collected data. The analysis and modeling module processes multispectral data in real time and predicts soil moisture. The data storage and display module is used to store historical data and display the soil moisture distribution in real time through a visual interface to provide decision support.

[0008] It is further defined that a method for a system for online analysis of soil dryness and moisture based on unmanned aerial vehicle multispectral is specifically performed as follows: Step S1, a drone equipped with a multispectral sensor is used to perform a flight scan of a target area to collect multispectral image data, wherein the multispectral image data includes blue light, green light, red light, near infrared and other bands;

[0009] Step S2, performing data preprocessing on the collected multispectral image data to extract spectral characteristic parameters related to soil moisture;

[0010] Step S3, humidity prediction based on machine learning: input the extracted spectral features into the machine learning model to predict the dryness and wetness of the soil;

[0011] Step S4, real-time result display and feedback optimization: Generate a visual image of soil moisture distribution through the data processing system, and optimize the calculation model based on the field calibration data to improve the prediction accuracy.

[0012] It is further defined that in step S1, the multispectral sensor carried by the drone has multiple bands ranging from visible light to near infrared, and its sampling accuracy is not less than 5nm wavelength resolution.

[0013] It is further defined that the data preprocessing in step S2 specifically includes the following steps:

[0014] Step S2.1, performing radiation correction to remove the influence of light intensity changes;

[0015] Step S2.2, perform geometric correction to align the multispectral image data to a unified geographic reference coordinate system;

[0016] Step S2.3, performing noise filtering to remove high-frequency noise introduced by equipment or environmental factors;

[0017] The spectral characteristic parameters include the normalized difference vegetation index NDVI and the surface reflectance ρ λ The normalized difference vegetation index NDVI is used to reflect the impact of vegetation coverage on humidity, and the surface reflectivity ρ λ Used to quantify the spectral response of soil in different bands.

[0018] It is further defined that the training data of the machine learning model described in step S3 is derived from the data obtained in the following steps:

[0019] Step S3.1, calculate the soil moisture index SMI, consider the multi-band reflectance and nonlinear relationship, use the deep learning model to dynamically adjust the weights, and extract the spectral feature parameters of the multispectral image;

[0020] Step S3.2, the measured soil moisture data of ground sampling is compared with the predicted value, and the mean square error (MSE) is selected as the loss function to measure the error between the predicted soil moisture value and the true value;

[0021] Step S3.3, record and save each monitoring data to form a traceable data source.

[0022] It is further defined that the real-time results in step S3 are displayed and feedback optimized. The real-time result display uses the Kriging interpolation method to generate a spatial distribution map based on discrete points. The interpolation weights are optimized by adding environmental information to make the humidity estimation more consistent with the actual terrain and environmental changes. The humidity data records finally generated are stored in a time series. The time axis can intuitively reflect the soil moisture change trend in the collection area, assisting users in making decisions.

[0023] It is further defined that the calculation formula of the normalized difference vegetation index NDVI is:

[0024]

[0025] Where NIR represents the reflectivity in the near-infrared band, and R represents the reflectivity in the red band;

[0026] Surface reflectivity ρ λ The calculation formula is:

[0027]

[0028] Where L λ is the observed radiance, L λ,atm is the upward radiation brightness caused by atmospheric scattering, which can be estimated by the atmospheric correction model. s,λis the incident radiation flux density of the sun in this band, θ s is the solar zenith angle in radians, which can be calculated from the geographical location and observation time. λ,down is the atmospheric downward transmittance, T λ,up is the upward air permeability of the atmosphere.

[0029] The resulting data are further limited and divided into three levels according to different degrees: soil with water layer, moist soil but without water layer, and relatively dry soil. Different colors are assigned to distinguish the degree of soil dryness and wetness in the target area, and superimposed on the satellite map of the web page to provide users with more intuitive irrigation and water use decisions.

[0030] Further limiting, the spectral characteristic parameters extracted from multispectral images are used to calculate the soil moisture index SMI. Considering the multi-band reflectivity and nonlinear relationship, the calculation formula is as follows:

[0031]

[0032] where ρ NIR represents the reflectivity in the near-infrared band, ρ Red represents the reflectivity of the red light band, ρ Green represents the reflectance of the green light band, and a, b, c, and γ are dynamic weight parameters obtained through model training.

[0033] The measured soil moisture data of ground sampling is compared with the predicted value, and the mean square error (MSE) is selected as the loss function to measure the error between the predicted soil moisture value and the true value. The calculation formula is:

[0034]

[0035] where y i Indicates the true soil moisture value, Represents the model prediction value.

[0036] The beneficial effects of adopting the above technical solution are:

[0037] 1. Installing multispectral sensors on drones is a way to use drones to quickly and cheaply cover a large area of ​​farmland, solving the spatial limitations of soil moisture monitoring in traditional methods.

[0038] 2. The present invention combines multispectral imaging with machine learning algorithms, which can extract accurate soil moisture information from complex spectral data and maintain high prediction accuracy under different environmental conditions.

[0039] 3. The system is real-time and can provide instant feedback for agricultural irrigation and water resources management, support precision agriculture and intelligent decision-making. The method and system have broad application prospects and can be widely used in agricultural irrigation management, environmental monitoring, ecological protection, agricultural disaster warning and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is the overall control flow chart of the present invention;

[0041] Figure 2 It is a functional diagram of each module in the present invention. DETAILED DESCRIPTION

[0042] The specific implementation methods of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, with the aim of helping those skilled in the art to have a more complete, accurate and in-depth understanding of the concept and technical solution of the present invention and facilitating its implementation.

[0043] like Figure 1 , Figure 2 As shown, the present invention is a system and method for online analysis of soil dryness and moisture based on unmanned aerial vehicle multispectral technology, which combines the high maneuverability of unmanned aerial vehicles with the advantages of multispectral imaging technology, can perform real-time soil moisture analysis on large areas of farmland, and provide accurate irrigation management and agricultural resource optimization solutions.

[0044] The technical solution adopted by the present invention is: a system and method for online analysis of soil dryness and moisture based on unmanned aerial vehicle multispectral, including an unmanned aerial vehicle acquisition module, a multispectral data processing module, an analysis and modeling module, and a data storage and display module. The unmanned aerial vehicle acquisition module is responsible for the efficient acquisition of multispectral image data and has autonomous flight and path planning capabilities. The multispectral data processing module is responsible for radiation correction, geometric correction, noise filtering and spectral feature extraction of the collected data. The analysis and modeling module processes multispectral data in real time and predicts soil moisture. The data storage and display module is used to store historical data and display the soil moisture distribution in real time through a visual interface to provide decision support. The specific steps are as follows:

[0045] Step S1, using a drone equipped with a multispectral sensor to perform a flight scan of the target area to collect multispectral image data, which includes blue light, green light, red light, near infrared and other bands.

[0046] Carry out autonomous operation, send operation instructions to the aircraft through remote instructions, and automatically collect multispectral image data of the target area. This step sends the planned route file of the area to be flown to the drone connected to the platform through the remote platform. The drone receives the flight instruction and automatically performs the flight mission according to the preset flight path. During the execution of the mission, the drone collects high-altitude ground multispectral image data through automatic photo taking instructions. The overlap rate of adjacent photographed images is set between 60% and 80%. Then, automatic image transmission is performed to upload the collected image data to the cloud. Using the Internet of Things communication technology, the pictures carrying multispectral information and location information are transmitted to the remote cloud platform in real time and synchronously. The cloud platform receives and saves the data, which is convenient for further processing of the collected data.

[0047] Step S2.1, performing radiation correction to remove the influence of light intensity changes;

[0048] Step S2.2, perform geometric correction to align the multispectral image data to a unified geographic reference coordinate system;

[0049] Step S2.3, perform noise filtering to remove high-frequency noise introduced by equipment or environmental factors.

[0050] This involves image registration, image fusion and color correction. Images taken by drones at different positions and angles may have problems such as position offset, rotation or scale change. Image registration technology can align these images to a unified coordinate system so that they can be correctly stitched together. Registration technology includes methods such as feature extraction, feature matching and geometric transformation. Images taken by drones usually have overlapping areas. Image fusion technology can seamlessly fuse these overlapping areas so that the stitched images look natural and continuous. Fusion methods include multiple exposure synthesis, multiple resolution fusion and other technologies. Images taken at different times or under different lighting conditions may have color differences. Color correction technology can perform color correction on images so that the stitched images remain consistent in color, improving the look and feel and usability.

[0051] In step S3, the humidity prediction based on machine learning is to input the extracted spectral features into the machine learning model to predict the dryness and wetness of the soil, including the normalized difference vegetation index NDVI, the soil conditioning index SI and the surface reflectance ρ λ These characteristics can reflect the changes in soil moisture. The Normalized Difference Vegetation Index (NDVI) is used to reflect the impact of vegetation coverage on humidity. The calculation formula is as follows:

[0052]

[0053] Where NIR represents the reflectivity in the near-infrared band, and R represents the reflectivity in the red band;

[0054] Surface reflectivity ρ λ Used to quantify the spectral response of soil in different bands, the calculation formula is as follows:

[0055]

[0056] Where L λ is the observed radiance, L λ,atm is the upward radiation brightness caused by atmospheric scattering, which can be estimated by the atmospheric correction model (such as 6S or MODTRAN). s,λ is the incident radiation flux density of the sun in this band (solar irradiance, unit W / ㎡·μm), θ s is the solar zenith angle in radians, which can be calculated from the geographical location and observation time. λ,down is the atmospheric downward transmittance, T λ,up is the atmospheric upward air permeability;

[0057] Step S4, real-time result display and feedback optimization: Generate a visual image of soil moisture distribution through the data processing system, and optimize the calculation model based on the field calibration data to improve the prediction accuracy.

[0058] In step S3, the training data of the machine learning model comes from the data obtained in the following steps:

[0059] Step S3.1, calculate the soil moisture index SMI, consider the multi-band reflectivity and nonlinear relationship, use the deep learning model to dynamically adjust the weight, and calculate the soil moisture index SMI based on the spectral feature parameters extracted from the multispectral image. Consider the multi-band reflectivity and nonlinear relationship, use the deep learning model to dynamically adjust the weight. The calculation formula is as follows:

[0060]

[0061] where ρ NIR represents the reflectivity in the near-infrared band, ρ Red represents the reflectivity of the red light band, ρ Green represents the reflectivity of the green light band, a, b, c, and γ are dynamic weight parameters obtained through model training;

[0062] Step S3.2, the measured soil moisture data of ground sampling is compared with the predicted value, and the mean square error (MSE) is selected as the loss function to measure the error between the predicted soil moisture value and the true value. The calculation formula is:

[0063]

[0064] where y i Indicates the true soil moisture value, represents the model prediction value;

[0065] Step S3.3, record and save each monitoring data to form a traceable data source.

[0066] The real-time results are mapped using the Kriging interpolation method. The interpolation weights are optimized by adding environmental information (soil type, terrain factors) to make the humidity estimation more consistent with the actual terrain and environmental changes. The final optimization formula is as follows:

[0067]

[0068] Among them, f(x i ,y i ) is the environmental factor impact function, and the environmental impact functions in different regions are different.

[0069] The visualization image in step S4 displays the soil moisture distribution in real time and provides decision support. The generated result data is divided into three levels according to different degrees: soil with water layer, moist soil but without water layer, and relatively dry soil. Different colors are given to distinguish the degree of soil dryness and wetness in the target area, and superimposed on the satellite map of the web page, which can provide users with more intuitive irrigation and water use decisions.

[0070] The present invention is described above by way of example in conjunction with the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention; or the above-mentioned concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A system for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles, characterized by: It includes a drone acquisition module, a multispectral data processing module, an analysis and modeling module, and a data storage and display module. The drone acquisition module is responsible for the efficient acquisition of multispectral image data and has autonomous flight and path planning capabilities. The multispectral data processing module is responsible for radiation correction, geometric correction, noise filtering and spectral feature extraction of the collected data. The analysis and modeling module processes multispectral data in real time and predicts soil moisture. The data storage and display module is used to store historical data and display soil moisture distribution in real time through a visual interface to provide decision support.

2. The method of a system for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles according to claim 1 is characterized by: The specific steps are as follows: Step S1, using a drone equipped with a multispectral sensor to perform a flight scan of a target area to collect multispectral image data, wherein the multispectral image data includes blue light, green light, red light, near infrared and other bands; Step S2, performing data preprocessing on the collected multispectral image data to extract spectral characteristic parameters related to soil moisture; Step S3, humidity prediction based on machine learning: input the extracted spectral features into the machine learning model to predict the dryness and wetness of the soil; Step S4, real-time result display and feedback optimization: Generate a visual image of soil moisture distribution through the data processing system, and optimize the calculation model based on the field calibration data to improve the prediction accuracy.

3. The method for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles according to claim 2 is characterized in that: In step S1, the multispectral sensor carried by the drone has multiple bands ranging from visible light to near infrared, and its sampling accuracy is not less than 5nm wavelength resolution.

4. The method for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles according to claim 2 is characterized in that: The data preprocessing in step S2 specifically includes the following steps: Step S2.1, performing radiation correction to remove the influence of light intensity changes; Step S2.2, perform geometric correction to align the multispectral image data to a unified geographic reference coordinate system; Step S2.3, performing noise filtering to remove high-frequency noise introduced by equipment or environmental factors; The spectral characteristic parameters include the normalized difference vegetation index NDVI and the surface reflectance ρ λ The normalized difference vegetation index NDVI is used to reflect the impact of vegetation coverage on humidity, and the surface reflectivity ρ λ Used to quantify the spectral response of soil in different bands.

5. The method for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles according to claim 2 is characterized in that: The training data of the machine learning model described in step S3 is derived from the data obtained in the following steps: Step S3.1, calculate the soil moisture index SMI, consider the multi-band reflectance and nonlinear relationship, use the deep learning model to dynamically adjust the weights, and extract the spectral feature parameters of the multispectral image; Step S3.2, the measured soil moisture data of ground sampling is compared with the predicted value, and the mean square error (MSE) is selected as the loss function to measure the error between the predicted soil moisture value and the true value; Step S3.3, record and save each monitoring data to form a traceable data source.

6. The method for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles according to claim 4 is characterized in that: The real-time results in step S3 are displayed and feedback optimized. The real-time result display uses the Kriging interpolation method to generate a spatial distribution map based on discrete points. The interpolation weights are optimized by adding environmental information to make the humidity estimation more consistent with the actual terrain and environmental changes. The humidity data records finally generated are stored in a time series. The time axis can intuitively reflect the soil moisture change trend in the collection area, assisting users in making decisions.

7. The method for online soil dryness and moisture analysis based on multi-spectral analysis of unmanned aerial vehicles according to claim 4 is characterized in that: The calculation formula of the normalized difference vegetation index NDVI is: Where NIR represents the reflectivity in the near-infrared band, and R represents the reflectivity in the red band; Surface reflectivity ρ λ The calculation formula is: Where L λ is the observed radiance, L λ,atm is the upward radiation brightness caused by atmospheric scattering, which can be estimated by the atmospheric correction model. s,λ is the incident radiation flux density of the sun in this band, θ s is the solar zenith angle in radians, which can be calculated from the geographical location and observation time. λ,down is the atmospheric downward transmittance, T λ,up is the upward air permeability of the atmosphere.

8. The method for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles according to claim 6 is characterized in that: The generated result data are divided into three levels according to different degrees: soil with water layer, moist soil but without water layer, and relatively dry soil. Different colors are assigned to distinguish the degree of soil dryness and wetness in the target area, and superimposed on the satellite map of the web page to provide users with more intuitive irrigation and water use decisions.

9. The method for online analysis of soil dryness and moisture based on multi-spectral analysis of unmanned aerial vehicles according to claim 5 is characterized in that: The spectral characteristic parameters extracted from multispectral images are used to calculate the soil moisture index SMI. Considering the multi-band reflectivity and nonlinear relationship, the calculation formula is as follows: where ρ NIR represents the reflectivity in the near-infrared band, ρ Red represents the reflectivity of the red light band, ρ Green represents the reflectance of the green light band, and a, b, c, and γ are dynamic weight parameters obtained through model training. The measured soil moisture data of ground sampling is compared with the predicted value, and the mean square error (MSE) is selected as the loss function to measure the error between the predicted soil moisture value and the true value. The calculation formula is: where y i Indicates the true soil moisture value, Represents the model prediction value.

Images