A dynamic fusion correction method for multivariate data in a high-throughput plant phenotyping platform in field environments

Abstract

The present invention relates to a multi-source data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment, which solves the defect that it is difficult to realize multi-environment data fusion correction compared with the prior art. The present invention comprises the following steps: acquiring environmental data and crop phenotypic data; performing wind speed and wind direction correction; performing light correction; performing temperature correction; and data fusion correction. The present invention corrects and fuses multi-source data through a real-time algorithm, thereby realizing high-precision multi-level dynamic fusion correction of phenotypic data.

Description

Technical Field

[0001] The invention relates to the technical field of plant protection data correction and fusion, and in particular to a multivariate data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment. Background Art

[0002] The refined management of modern agriculture increasingly relies on the high-precision collection of phenotypic data, especially in complex field environments. The integration of multiple sensors such as lidar, depth cameras, thermal imaging cameras, multispectral cameras, and high-definition cameras through track-based phenotyping platforms to obtain multi-dimensional phenotypic information of crops has become the mainstream means of field crop monitoring. The application of this technology can not only fully reflect the growth status of crops, but also provide a scientific basis for precise fertilization, irrigation, and pest and disease control.

[0003] The complexity of the field environment means that crop posture changes are affected by a variety of factors, including wind, light, temperature and humidity. These environmental factors will significantly interfere with the sensor data collection process, thereby reducing the accuracy of multi-sensor fusion data. For example, under strong wind conditions, the swaying of crops will cause deviations in the three-dimensional point cloud data collected by lidar and depth cameras, while changes in light intensity may cause the image quality of multi-spectral cameras to deteriorate and affect the stability of spectral reflectance data. In addition, changes in temperature and humidity will affect the transpiration of crops, thereby changing the posture of leaves. These factors are superimposed on each other, causing significant interference in the collection process of crop phenotypic information, especially in the fusion process of three-dimensional point cloud data, thermal imaging data and spectral data, the error is more obvious.

[0004] Most existing correction methods compensate for a single environmental factor under static conditions and cannot cope with the interference of multiple factors in complex field environments. Although these technologies have achieved certain results in controlled environments such as laboratories or greenhouses, they are inadequate in dynamic, real field environments and cannot meet the agricultural sector's demand for high-precision, multi-dimensional crop phenotypic data. For example, existing solutions simply correct the static characteristics of crops, ignoring the combined effects of wind speed, light intensity, temperature and humidity, which greatly reduces the accuracy and reliability of data collection.

[0005] Therefore, in order to meet the needs of precision agriculture, it is urgent to develop an innovative dynamic correction model that can correct the interference of crop posture changes on multi-sensor data collection under the combined effect of various environmental factors, and ensure that high-precision multi-dimensional phenotypic data can still be obtained under complex field conditions. Summary of the invention

[0006] The purpose of the present invention is to solve the defect that it is difficult to realize multi-environment data fusion correction in the prior art, and to provide a multi-data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment to solve the above problem.

[0007] In order to achieve the above object, the technical solution of the present invention is as follows:

[0008] A multi-data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment comprises the following steps:

[0009] Acquisition of environmental data and crop phenotypic data: Acquisition of environmental data and crop phenotypic data;

[0010] Correction of wind speed and direction: The method combining fluid mechanics equations with the blade flexible structure dynamics model is used to accurately calculate the influence of wind speed and direction on the posture of crop leaves. By establishing a physical model and considering the effect of aerodynamic forces, the displacement and deformation of the blades under the action of wind are dynamically corrected, and the three-dimensional point cloud data is corrected.

[0011] Light correction: By introducing a light response biological model and combining it with data from a real-time light sensor, the posture of crop leaves and the exposure and white balance settings of the image are dynamically adjusted to achieve correction under lighting conditions;

[0012] Temperature correction: Based on the physical mechanism of stomatal opening and closing, a temperature and humidity regulation model was constructed to simulate the response of stomatal pores under different temperature and humidity conditions, and then deduce the change of leaf temperature, so as to achieve temperature correction;

[0013] Data fusion and correction: Perform high-precision data fusion and correction to generate three-dimensional crop point clouds and phenotypic data without external environmental interference, and realize dynamic fusion and correction of multivariate data.

[0014] The acquisition of the environmental data and the crop phenotype data comprises the following steps:

[0015] Environmental data, including wind speed, wind direction, light intensity, temperature and humidity, are acquired in real time through environmental monitoring sensors, and used as external disturbance inputs in the correction model to determine the model's dynamic adjustment of crop posture;

[0016] Crop phenotyping data is collected through multiple sensors on a plant phenotyping platform in a field environment, including data from lidar, depth cameras, thermal imaging cameras, multispectral cameras, and high-definition cameras.

[0017] The wind speed and wind direction correction comprises the following steps:

[0018] The swing and deformation process of the blade is described by the flexible body dynamics equation:

[0019] As a flexible body, the blade is affected by wind speed and direction, generating positive pressure and shear force. According to the structural characteristics and force conditions of the blade, the correction model uses the flexible body dynamics equation to simulate the swing and deformation process of the blade. The deformation of the blade is approximated by a one-dimensional bending beam model, as shown below:

[0020]

[0021] Where y(x,t) is the deformation of the blade, x is the coordinate of the blade along its length, t is time, E is the elastic modulus of the blade material, I is the moment of inertia of the blade section, c is the damping coefficient, which is used to represent dissipative effects such as air resistance, and f(x,t) is the external force of the wind on the blade as a function of wind speed and direction;

[0022] Combining fluid mechanics, wind speed and wind direction data, we can calculate the effect of wind on the blades:

[0023] The real-time data obtained by the anemometer and wind vane: wind speed v and the angle θ between the blade and the wind direction, combined with the wind force formula in fluid mechanics, the correction model calculates the force of wind on the blade; the magnitude of the force is proportional to the square of the wind speed, and the change of wind direction changes the angle at which the blade is subjected to force. The wind force is calculated using the wind force formula in fluid mechanics;

[0024] Assuming that the wind speed v acts on the blade and the wind force F acts on the unit area of ​​the blade, the expression is as follows:

[0025]

[0026] Where ρ is the air density, C D is the drag coefficient of the blade, which depends on the shape of the blade and the wind direction, A is the windward area of ​​the blade, v is the wind speed, which changes with time t, and the square of the wind speed is proportional to the force;

[0027] In addition, combined with the influence of wind direction, the windward area A of the blade depends on the angle θ between the blade and the wind direction. The specific relationship is as follows:

[0028] A=A0·cos(θ),

[0029] Where A0 is the actual area of ​​the blade, θ is the angle between the wind direction and the blade normal, which is measured in real time by the wind vane. When θ = 0, the wind is perpendicular to the blade, and when θ = 90°, the blade is parallel to the wind.

[0030] f(x, t) is the local external force on the blade at a specific position x and time t. In order to derive the local external force f(x, t) from the overall wind force F, the geometric shape, position and force distribution of the blade are considered. It is assumed that the wind force F acts evenly on the surface of the blade, and the wind force density, i.e., force / unit length, is used to represent the external force at any position on the blade. Because the wind force F changes with time t, it is a function of time. The total length of the blade is L, and f(x, t) is defined as the wind force density, i.e., the force per unit length, as shown below:

[0031]

[0032] Dynamic adjustment of blade posture and point cloud data correction:

[0033] According to the stress of the blades under different wind speeds and directions, the displacement y(x,t) of the blades under the action of wind is solved through the flexible body dynamics equation, and the posture of the blades is dynamically adjusted according to the displacement distribution at different times t and spatial coordinates x; combined with the three-dimensional point cloud data collected by the lidar and depth camera of the high-throughput plant phenotyping platform, the three-dimensional point cloud of the crop is corrected according to the impact of wind on the crop posture, and the blade offset caused by wind speed is dynamically compensated to restore the true three-dimensional shape under windless conditions;

[0034] Assuming the collected point cloud coordinates are (X, Y, Z), the corrected coordinates after wind action are as follows:

[0035] X ′ =X+y x (x,t),Y ′ =Y+y y (x,t),Z ′ =Z+y z (x,t);

[0036] Among them, y x (x,t),y y (x,t),y z (x, t) is the displacement of the blade in the X, Y, and Z axes calculated by the mechanical equation;

[0037] Finally, the correction model restores the three-dimensional geometric structure of the blade under windless conditions (X ′ ,Y ′ ,Z ′ ).

[0038] The illumination correction comprises the following steps:

[0039] Build a light response model:

[0040] Plant leaves are phototropic, and when the light intensity and angle change, the leaves rotate to optimize their photosynthetic efficiency;

[0041] The correction model is based on the photosynthetic response characteristics of plants, and the light response formula is as follows:

[0042] θ leaf =θ0+f(N,α)

[0043] Among them, θ leaf is the angle of the blade relative to the horizontal plane, θ0 is the initial angle, that is, the blade posture under no light or uniform light conditions, and N is the current light intensity, which is obtained in real time by the light intensity sensor in units of α is the incident angle of light, that is, the angle between the light and the surface of the leaf. f(N,α) is the light response function of the leaf, which is used to describe the rotation characteristics of the leaf under changing light. It is expressed as follows:

[0044] f(N,α)=k1·log(N)+k2·sin(α)

[0045] Among them, k1 and k2 are empirical parameters that depend on the plant species and environment;

[0046] Exposure and white balance adjustment:

[0047] The light intensity sensor monitors the light intensity in real time and automatically adjusts the camera's exposure time and white balance settings to ensure that the images captured by the multispectral and high-definition cameras under different lighting conditions have consistent brightness and color. The following formula is used for automatic adjustment:

[0048]

[0049] Among them, the exposure time t exp The adjustment is dynamically adjusted according to the light intensity N to ensure consistent imaging brightness, where k exp is a constant that depends on the specific settings of the camera. The adjustment of white balance is based on the color temperature T under different light sources. light Correction is performed. The color temperature is related to the light intensity N and the light direction α. ​​W is the adjustment coefficient of white balance. Adjust the white balance of the camera under different lighting environments. ref The reference white light source has a color temperature of 6500K. light is the color temperature of the current environment, acquired by the sensor in real time;

[0050] Dynamic correction of spectral reflectance data:

[0051] By real-time monitoring of the ambient light intensity N and adjusting the exposure time t exp, ensuring consistency under different lighting conditions. Then, the uncorrected reflectivity is corrected according to the lighting conditions and the impact of the incident angle α to obtain the dynamically corrected wavelength λ i The reflectivity under as follows:

[0052]

[0053] in, is the original wavelength λ i The reflectance under the condition is obtained by the multispectral sensor of the phenotyping platform; g(N,α) is a correction function that describes the influence of light intensity and incident angle on reflectance;

[0054] Finally, by correcting the reflectivity of all wavelengths, the dynamically corrected spectral reflectance data D is obtained. spectral as follows:

[0055]

[0056] Photodynamic correction of RGB color information:

[0057] The raw color data obtained by the RGB camera contains the intensity values ​​of each pixel in the red, green, and blue channels, expressed as follows:

[0058] d color ={R,G,B}

[0059] Through white balance correction, the influence of color temperature under different light sources is corrected; by adjusting the exposure, the consistency of image brightness under different lighting conditions is ensured. The final corrected RGB color data is as follows:

[0060]

[0061] Blade attitude correction:

[0062] According to the light response characteristics of the leaves, the calibration model adjusts the posture of the leaves and simulates the change of leaf angle under the change of light intensity;

[0063] Combining the image data collected by the multispectral camera and the high-definition camera, the correction model compensates for the imaging error caused by uneven illumination through posture adjustment, and the angle Δθ of the blade rotation leaf The formula is as follows:

[0064] Δθ leaf =f(N,α)

[0065] Then, according to the rotation changes of the blades, the calibration model adjusts the three-dimensional coordinates of the blades;

[0066] Get the geometric data obtained by the wind correction model (X ′,Y ′ ,Z ′ ), the position after light model correction is obtained by transforming the following rotation matrix, which is used to adjust the three-dimensional position of the blade. According to the angle change of the blade, the spatial position of the blade is corrected to obtain the geometric structure data D after double correction by wind correction and light correction. geometry ,

[0067]

[0068] The temperature correction comprises the following steps:

[0069] Establish a temperature and humidity regulation model based on the stomatal regulation model:

[0070] Based on the physical mechanism of stomatal opening and closing, a temperature and humidity regulation model was constructed to simulate the response of stomata under different temperature and humidity conditions, and then deduce the changes in leaf temperature.

[0071] The temperature and humidity adjustment model is described as follows:

[0072]

[0073] Where, ΔT stomata is the leaf temperature change caused by the opening and closing of stomata. T and H are the current ambient temperature and humidity, respectively, which are obtained through the temperature and humidity sensor. opt and H opt It is the optimal temperature and humidity for the plant to achieve optimal stomata opening and closing. max , T min , H max and H min are the temperature and humidity extremes when the stomata are fully open or closed, respectively. h1 and h2 are parameters determined based on plant physiological characteristics, indicating the weight of the effect of stomatal regulation on temperature.

[0074] Dynamic correction of crop temperature in multi-source data:

[0075] The correction model corrects the temperature data collected by the phenotypic platform sensor based on the real-time data of the temperature and humidity sensor. The temperature correction process formula is as follows:

[0076] D thermal =T sensor +ΔT stomata

[0077] Among them, T sensor is the initial temperature data collected by the thermal infrared sensor of the phenotyping platform, D thermal To correct the temperature data for temperature and humidity, by introducing ΔT stomata, compensates for the leaf temperature changes caused by the opening and closing of stomata, ensuring that the temperature measured by the sensor is more consistent with the actual temperature conditions of the crop.

[0078] The fusion correction of the data comprises the following steps:

[0079] Re-integration based on corrected data:

[0080] The fusion of multi-source data is represented as weighted average or optimized combination. The goal of the fusion is to generate a unified three-dimensional point cloud data. At the same time, each point cloud point also contains temperature and spectral reflectance information.

[0081] The fusion process is as follows:

[0082] D fused =α·D geometry +β·D spectral +γ·D thermal +δ·D color

[0083] Among them, D fused It is the fused 3D point cloud data, including geometry, spectrum, temperature and color data; D geometry It is the geometric structure data after double correction by wind correction and light correction; D spectral is the spectral reflectance data after dynamic correction; D thermal It is the temperature data corrected by temperature and humidity; D color is the corrected RGB color data; α, β, γ, δ are the weight coefficients of each sensor data, which are adjusted based on the reliability of the data or environmental conditions;

[0084] Optimize the weight coefficients of each part: Adaptive weight adjustment based on environmental factors, automatically adjust according to environmental conditions under different environmental conditions.

[0085] Beneficial Effects

[0086] Compared with the prior art, the multi-source data dynamic fusion correction method of the present invention for a high-throughput plant phenotyping platform in a field environment corrects and fuses multi-source data through a real-time algorithm, thereby achieving high-precision multi-level dynamic fusion correction of phenotypic data.

[0087] The present invention firstly uses the principles of physics to correct wind speed and light based on fluid mechanics and optical properties, and dynamically adjusts the three-dimensional posture of crops. At the same time, combined with biological knowledge, the influence of temperature and humidity is modeled through the theory of stomatal regulation, so as to accurately deduce the changes in crop surface temperature.

[0088] The present invention not only improves the accuracy of multi-sensor data fusion, but also significantly enhances the adaptability to crop phenotypic changes in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 is a method sequence diagram of the present invention;

[0090] Figure 2 It is a dynamic correction flow chart of the present invention. DETAILED DESCRIPTION

[0091] In order to have a further understanding and recognition of the structural features and the effects achieved by the present invention, a preferred embodiment and accompanying drawings are used for detailed description as follows:

[0092] For high-throughput plant phenotyping platforms in field environments, existing technologies lack dynamic correction mechanisms and can usually only handle static conditions or compensation for a single factor, and cannot cope with the dynamic changes of multiple factors in the field environment, such as wind, light, temperature and humidity; at the same time, existing technologies are relatively simple in handling multiple environmental factors, and cannot comprehensively consider the impact of different environmental factors on crop phenotypic data; in addition, the correction model is often oversimplified and fails to accurately describe the flexible structure of crop leaves and their nonlinear response in a dynamic environment, resulting in low data correction accuracy; in terms of multi-sensor data fusion, the existing technology processing method is relatively rough and fails to fully cope with the impact of environmental interference on the accuracy of multi-source data fusion. Therefore, existing technologies are difficult to meet the demand for high-precision, multi-dimensional crop phenotypic data in precision agriculture.

[0093] like Figure 1 and Figure 2 As shown, the multi-data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment described in the present invention comprises the following steps:

[0094] The first step is to obtain environmental data and crop phenotypic data: obtain environmental data and crop phenotypic data.

[0095] (1) Environmental data, including wind speed, wind direction, light intensity, temperature and humidity, are acquired in real time through environmental monitoring sensors, and used as external disturbance inputs in the correction model to determine the model's dynamic adjustment of crop posture.

[0096] (2) Collect crop phenotyping data through multiple sensors on a plant phenotyping platform in a field environment, including data from lidar, depth cameras, thermal imaging cameras, multispectral cameras, and high-definition cameras.

[0097] The technical difficulty of dynamic fusion correction of multivariate data lies in how to effectively deal with the variable external factors in the field environment, such as wind, light, temperature and humidity, which will act on crops at the same time, resulting in deviations in the data collected by different sensors. Accurate correction requires comprehensive consideration of these dynamic environmental interferences and real-time correction of crop postures through complex physical models. In addition, data fusion also needs to deal with the heterogeneity of multi-source data from lidar, depth cameras, spectral sensors, etc. to ensure that the fused data is highly consistent. The technical significance lies in improving the accuracy and reliability of crop phenotypic data through dynamic correction and data fusion, assisting precision agriculture, and achieving efficient crop management and precise water and nutrition monitoring.

[0098] The second step is to correct the wind speed and direction: a method combining fluid mechanics equations with the blade flexible structure dynamics model is used to accurately calculate the impact of wind speed and direction on the posture of crop leaves; by establishing a physical model and considering the effect of aerodynamic forces, the displacement and deformation of the blades under the action of wind are dynamically corrected to correct the three-dimensional point cloud data.

[0099] (1) The swing and deformation process of the blade is described by the flexible body dynamics equation:

[0100] As a typical flexible body, the blade will be subjected to complex positive pressure and shear force under the combined effect of wind speed and wind direction. The positive pressure mainly acts on the surface of the blade, while the shear force acts on the edge of the blade and its force direction. Due to the flexible characteristics of the blade structure, it will not only produce obvious swing under the action of wind force, but also local bending and deformation. The correction model combines the mechanical properties of the blade and accurately simulates the dynamic response process of the blade under the action of wind force through the flexible body dynamics equation. In order to better describe the complex deformation behavior of the blade, the correction model uses a one-dimensional bending beam model for approximation. The model can capture the bending deformation of the blade at different positions and its dynamic changes over time, so as to accurately estimate the overall swing and local deformation of the blade under the action of wind force. This mechanical model not only takes into account the influence of wind speed and wind direction on the blade, but also reflects the comprehensive effect of factors such as blade material properties, geometric shape and force distribution. The model is shown below:

[0101]

[0102] Where y(x,t) is the deformation of the blade, x is the coordinate of the blade along its length, t is time, E is the elastic modulus of the blade material, I is the moment of inertia of the blade section, c is the damping coefficient, which is used to account for dissipative effects such as air resistance, and f(x,t) is the external force of the wind on the blade as a function of wind speed and direction.

[0103] (2) Combine fluid mechanics, wind speed and wind direction data to calculate the effect of wind on the blades:

[0104] Real-time data obtained by anemometers and wind vanes, including wind speed and the angle between the blade and wind direction, provide key external environmental inputs for the correction model. Wind speed reflects the intensity of air flow, while the angle between the blade and wind direction determines the direction and distribution of wind force on the blade surface. Combined with the wind force formula in fluid mechanics, the correction model can calculate the force exerted on the blade by wind speed and wind direction. The magnitude of wind force is proportional to the square of wind speed, and a small change in wind speed will lead to a large change in force. In addition, changes in wind direction will adjust the force angle of the blade and affect the distribution pattern of wind force. Through the wind force formula, the correction model comprehensively considers the changes in wind speed, wind direction and windward area of ​​the blade, and dynamically calculates the normal pressure and shear force exerted by wind force on the blade. This calculation can not only reflect the instantaneous change of wind force intensity, but also accurately describe the adjustment of blade posture under different wind directions, providing a reliable mechanical basis for further blade swing and deformation correction.

[0105] Assuming that the wind speed v acts on the blade and the wind force F acts on the unit area of ​​the blade, the expression is as follows:

[0106]

[0107] Where ρ is the air density, C D is the drag coefficient of the blade, which depends on the shape of the blade and the wind direction, A is the windward area of ​​the blade, v is the wind speed, which changes with time t, and the square of the wind speed is proportional to the force;

[0108] In addition, combined with the influence of wind direction, the windward area A of the blade depends on the angle θ between the blade and the wind direction. The specific relationship is as follows:

[0109] A=A0·cos(θ),

[0110] Where A0 is the actual area of ​​the blade, θ is the angle between the wind direction and the blade normal, which is measured in real time by the wind vane. When θ = 0, the wind is perpendicular to the blade, and when θ = 90°, the blade is parallel to the wind.

[0111] f(x, t) is the local external force on the blade at a specific position x and time t. In order to derive the local external force f(x, t) from the overall wind force F, the geometric shape, position and force distribution of the blade are considered. It is assumed that the wind force F acts evenly on the surface of the blade, and the wind force density, i.e., force / unit length, is used to represent the external force at any position on the blade. Because the wind force F changes with time t, it is a function of time. The total length of the blade is L, and f(x, t) is defined as the wind force density, i.e., the force per unit length, as shown below:

[0112]

[0113] (3) Dynamic adjustment of blade posture and point cloud data correction:

[0114] According to the stress of the blades under different wind speeds and directions, the displacement y(x,t) of the blades under the action of wind is solved through the flexible body dynamics equation, and the posture of the blades is dynamically adjusted according to the displacement distribution at different times t and spatial coordinates x; combined with the three-dimensional point cloud data collected by the lidar and depth camera of the high-throughput plant phenotyping platform, the three-dimensional point cloud of the crop is corrected according to the impact of wind on the crop posture, and the blade offset caused by wind speed is dynamically compensated to restore the true three-dimensional shape under windless conditions;

[0115] Assuming the collected point cloud coordinates are (X, Y, Z), the corrected coordinates after wind action are as follows:

[0116] X ′ =X+y x (x,t),Y ′ =Y+y y (x,t),Z ′ =Z+y z (x,t);

[0117] Among them, y x (x,t),y y (x,t),y z (x, t) is the displacement of the blade in the X, Y, and Z axes calculated by the mechanical equation;

[0118] Finally, the correction model restores the three-dimensional geometric structure of the blade under windless conditions (X ′ ,Y ′ ,Z ′ ).

[0119] The third step is to perform light correction: by introducing a light response biological model and combining it with data from real-time light sensors, the posture of crop leaves as well as the exposure and white balance settings of the image are dynamically adjusted to achieve correction under lighting conditions.

[0120] (1) Establishing a light response model:

[0121] When establishing a light response model, the focus is on simulating the natural response behavior of plant leaves to changes in light intensity and angle. Plant leaves have obvious phototropism. When the light intensity increases or the incident angle of light changes, the leaves will rotate by adjusting their posture to expose more leaf area to light to maximize the efficiency of photosynthesis. This physiological mechanism of phototropism is not only an important means for plants to adapt to environmental changes, but also the key to improving their energy utilization efficiency. The light response model dynamically simulates how leaves adjust their posture under different lighting conditions by considering light intensity, incident angle and the physiological characteristics of leaves. The model can accurately predict the rotation angle and speed of leaves when the light changes, thereby optimizing photosynthetic efficiency and providing a basis for subsequent data correction.

[0122] The correction model is based on the photosynthetic response characteristics of plants, and the light response formula is as follows:

[0123] θ leaf =θ0+f(N,α)

[0124] Among them, θ leaf is the angle of the blade relative to the horizontal plane, θ0 is the initial angle, that is, the blade posture under no light or uniform light conditions, and N is the current light intensity, which is obtained in real time by the light intensity sensor in units of α is the incident angle of light, that is, the angle between the light and the surface of the leaf. f(N,α) is the light response function of the leaf, which is used to describe the rotation characteristics of the leaf under changing light. It is expressed as follows:

[0125] f(N,α)=k1·log(N)+k2·sin(α)

[0126] Among them, k1 and k2 are empirical parameters that depend on the plant species and environment.

[0127] (2) Exposure and white balance adjustment:

[0128] By monitoring the light intensity in real time through the light intensity sensor, the system can accurately capture changes in ambient lighting and automatically adjust the camera's exposure time and white balance settings. This process is critical because different lighting conditions can lead to significant differences in image brightness and color, which can affect the accuracy and consistency of subsequent data analysis. The mechanism for automatically adjusting the exposure time enables the camera sensor to capture the best image under changing lighting conditions by setting the appropriate exposure time, avoiding problems with overexposure or underexposure of the image. At the same time, the adjustment of white balance ensures that the colors of images acquired under different light sources present a true and natural effect, eliminating color casts caused by changes in the color temperature of the light source. The specific process is automatically adjusted through the following formula,

[0129]

[0130] Among them, the exposure time t exp The adjustment is dynamically adjusted according to the light intensity N to ensure consistent imaging brightness, where k exp is a constant that depends on the specific settings of the camera. The adjustment of white balance is based on the color temperature T under different light sources. light Correction is performed. The color temperature is related to the light intensity N and the light direction α. ​​W is the adjustment coefficient of white balance. Adjust the white balance of the camera under different lighting environments. ref The reference white light source has a color temperature of 6500K. light It is the color temperature of the current environment, acquired by the sensor in real time.

[0131] (3) Dynamic correction of spectral reflectance data:

[0132] On the basis of ensuring image consistency, the uncorrected spectral reflectance data is dynamically corrected according to the influence of lighting conditions and incident angle. The uncorrected reflectance is significantly affected by the ambient light intensity and the incident angle of the leaves, which may cause the measurement results to deviate from the actual situation. The system establishes a lighting model and dynamically adjusts the reflectance in combination with real-time monitoring data to eliminate errors caused by changes in lighting conditions. Finally, after this dynamic correction process, the system outputs reflectance data at the corrected wavelength to ensure that these data can accurately reflect the spectral characteristics of plant leaves under specific lighting conditions. By real-time monitoring of the ambient light intensity N and adjusting the exposure time t exp , ensuring consistency under different lighting conditions. Then, the uncorrected reflectivity is corrected according to the lighting conditions and the impact of the incident angle α to obtain the dynamically corrected wavelength λ i The reflectivity under as follows:

[0133]

[0134] in, is the original wavelength λ i The reflectance under the condition is obtained by the multispectral sensor of the phenotyping platform; g(N,α) is a correction function that describes the influence of light intensity and incident angle on reflectance;

[0135] Finally, by correcting the reflectivity of all wavelengths, the dynamically corrected spectral reflectance data D is obtained. spectral as follows:

[0136]

[0137] (4) Dynamic correction of RGB color information:

[0138] Dynamic correction of RGB color information is crucial, mainly because different light sources have different color temperatures, which will cause color casts in the captured images, thus affecting the accuracy of the colors. Color casts make the colors in the image inconsistent with the actual situation, which in turn affects the assessment of the health of the plant. Dynamic correction can adapt to changes in light sources by adjusting white balance and exposure settings in real time, eliminating brightness and color differences caused by uneven lighting or intensity changes, thereby ensuring that the captured images can truly present the color characteristics of the plants under different lighting conditions, improving the accuracy and reliability of data analysis.

[0139] The raw color data obtained by the RGB camera contains the intensity values ​​of each pixel in the red, green, and blue channels, expressed as follows:

[0140] d color ={R,G,B}

[0141] Through white balance correction, the influence of color temperature under different light sources is corrected; by adjusting the exposure, the consistency of image brightness under different lighting conditions is ensured. The final corrected RGB color data is as follows:

[0142]

[0143] (5) Blade posture correction:

[0144] According to the light response characteristics of the leaves, the correction model can effectively adjust the posture of the leaves to adapt to changes in light intensity. Plant leaves will dynamically rotate according to the intensity and angle of the ambient light to optimize photosynthetic efficiency. In this process, the correction model simulates the adjustment of leaf angles under light changes by analyzing real-time light data and the physiological characteristics of the leaves. This adjustment not only helps to improve the photosynthetic efficiency of the leaves, but also provides a basis for subsequent data correction.

[0145] Combining the image data collected by the multispectral camera and the high-definition camera, the correction model compensates for the imaging error caused by uneven illumination through posture adjustment, and the angle Δθ of the blade rotation leaf The formula is as follows:

[0146] Δθ leaf =f(N,α)

[0147] Then, according to the rotation changes of the blades, the calibration model adjusts the three-dimensional coordinates of the blades;

[0148] Get the geometric data obtained by the wind correction model (X ′ ,Y ′ ,Z ′), the position after light model correction is obtained by transforming the following rotation matrix, which is used to adjust the three-dimensional position of the blade. According to the angle change of the blade, the spatial position of the blade is corrected to obtain the geometric structure data D after double correction by wind correction and light correction. geometry ,

[0149]

[0150] The fourth step is to perform temperature correction: Based on the physical mechanism of stomatal opening and closing, a temperature and humidity regulation model is constructed to simulate the response of stomata under different temperature and humidity conditions, and then deduce the changes in leaf temperature to achieve temperature correction.

[0151] (1) Establish a temperature and humidity regulation model based on the stomatal regulation model:

[0152] The temperature and humidity regulation model established based on the stomatal regulation model aims to deeply understand the physiological responses of plants under different temperature and humidity conditions. The model is based on the physical mechanism of stomatal opening and closing, and considers how the dynamic regulation of stomata affects water evaporation and gas exchange, thereby affecting the temperature and overall physiological state of the leaves. Specifically, the degree of opening and closing of stomata is affected by factors such as ambient temperature, humidity, and light intensity. When the temperature rises or the humidity decreases, the stomata will open moderately to promote the absorption of carbon dioxide and the evaporation of water; in high humidity or low temperature environments, the stomata may be closed to reduce water loss.

[0153] In this framework, the temperature and humidity regulation model can simulate the temperature changes of leaves under different climatic conditions by quantitatively describing the effect of stomatal opening and closing on leaf temperature. The model takes into account the response time of stomatal opening and closing and the physiological characteristics under different environmental conditions, and uses relevant physical equations to derive the temperature changes of leaves. The temperature and humidity regulation model is described as follows:

[0154]

[0155] Where, ΔT stomata is the leaf temperature change caused by the opening and closing of stomata. T and H are the current ambient temperature and humidity, respectively, which are obtained through the temperature and humidity sensor. opt and H opt It is the optimal temperature and humidity for the plant to achieve optimal stomata opening and closing. max , T min , H max and H min are the temperature and humidity extremes when the stomata are fully open or closed, respectively. h1 and h2 are parameters determined based on the physiological characteristics of the plant, representing the weight of the influence of stomatal regulation on temperature.

[0156] (2) Dynamic correction of crop temperature in multi-source data:

[0157] The correction model corrects the temperature data collected by the phenotyping platform sensor based on the real-time data of the temperature and humidity sensor to ensure that the acquired temperature data can more accurately reflect the actual physiological state of the plant. By monitoring the temperature and humidity of the environment in real time, the model can dynamically adjust the temperature value recorded by the sensor to eliminate errors caused by environmental changes. This process is particularly important because environmental factors have a direct impact on the growth and health of plants, and inaccurate temperature data may lead to incorrect assessments of plant water stress and physiological status.

[0158] Specifically, the correction process uses the real-time data provided by the temperature and humidity sensors to establish a mathematical model to link the degree of stomatal opening and closing with the ambient temperature and humidity, and derive the effect of stomatal regulation on leaf temperature. The following is a formula for the temperature correction process that can quantitatively describe this correction process:

[0159] D thermal =T sensor +ΔT stomata

[0160] Among them, T sensor is the initial temperature data collected by the thermal infrared sensor of the phenotyping platform, D thermal To correct the temperature data for temperature and humidity, by introducing ΔT stomata , compensates for the leaf temperature changes caused by the opening and closing of stomata, ensuring that the temperature measured by the sensor is more consistent with the actual temperature conditions of the crop.

[0161] The fifth step is data fusion and correction: perform high-precision data fusion and correction to generate three-dimensional crop point clouds and phenotypic data without external environmental interference, and realize dynamic fusion and correction of multivariate data.

[0162] (1) Re-integration based on corrected data:

[0163] The multi-dimensional data (geometry, spectral reflectance, temperature, color, etc.) of different sensors are combined to generate high-precision crop phenotypic data, ensuring that the true morphology and physical characteristics of the crop can be accurately presented. The data fusion process not only combines the multi-dimensional information from multiple sensors, but also combines the environmentally corrected data to ensure that different sensors can still output accurate crop phenotypic data under external disturbances. The weighted fusion algorithm ensures the comprehensive utilization of the data from each sensor. The fusion process is as follows:

[0164] D fused =α·D geometry +β·D spectral +γ·D thermal +δ·D color

[0165] Among them, D fused It is the fused 3D point cloud data, including geometry, spectrum, temperature and color data; D geometry It is the geometric structure data after double correction by wind correction and light correction; D spectral is the spectral reflectance data after dynamic correction; D thermal It is the temperature data corrected by temperature and humidity; D color is the corrected RGB color data; α, β, γ, δ are the weight coefficients of each sensor data, which are adjusted based on the reliability of the data or environmental conditions.

[0166] (2) Optimize the weight coefficients of each part: Adaptive weight adjustment based on environmental factors, automatically adjust according to environmental conditions under different environmental conditions.

[0167] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A multivariate data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment, characterized in that: The following steps are involved: 11) Acquisition of environmental data and crop phenotypic data: Acquisition of environmental data and crop phenotypic data; 12) Correction of wind speed and direction: The method combining fluid mechanics equations with the dynamics model of the flexible structure of the blade is used to accurately calculate the influence of wind speed and direction on the posture of crop leaves; By establishing a physical model and considering the effect of aerodynamic forces, the displacement and deformation of the blades under the action of wind are dynamically corrected, and the three-dimensional point cloud data is corrected; 13) Perform light correction: By introducing a light response biological model and combining data from real-time light sensors, the posture of crop leaves and the exposure and white balance settings of the image are dynamically adjusted to achieve correction under lighting conditions; The illumination correction comprises the following steps: 131) Establishing a light response model: Plant leaves are phototropic, and when the light intensity and angle change, the leaves rotate to optimize their photosynthetic efficiency; The correction model is based on the photosynthetic response characteristics of plants, and the light response formula is as follows: i leaf =θ0+f(N,α) Among them, θ leaf is the angle of the blade relative to the horizontal plane, θ0 is the initial angle, that is, the blade posture under no light or uniform light conditions, and N is the current light intensity, which is obtained in real time by the light intensity sensor in units of α is the incident angle of light, that is, the angle between the light and the surface of the leaf. f(N,α) is the light response function of the leaf, which is used to describe the rotation characteristics of the leaf under changing light. It is expressed as follows: f(N,α)=k1·log(N)+k2·sin(α) Among them, k1 and k2 are empirical parameters that depend on the plant species and environment; 132) Exposure and white balance adjustment: The light intensity sensor monitors the light intensity in real time and automatically adjusts the camera's exposure time and white balance settings to ensure that the images captured by the multispectral and high-definition cameras under different lighting conditions have consistent brightness and color. The following formula is used for automatic adjustment: Among them, the exposure time t exp The adjustment is dynamically adjusted according to the light intensity N to ensure consistent imaging brightness, where k exp is a constant that depends on the specific settings of the camera. The adjustment of white balance is based on the color temperature T under different light sources. light Correction is performed. The color temperature is related to the light intensity N and the light direction α. ​​W is the adjustment coefficient of white balance. Adjust the white balance of the camera under different lighting environments. ref The reference white light source has a color temperature of 6500K. light is the color temperature of the current environment, acquired by the sensor in real time; 133) Dynamic correction of spectral reflectance data: By real-time monitoring of the ambient light intensity N and adjusting the exposure time t exp , ensuring consistency under different lighting conditions. Then, the uncorrected reflectivity is corrected according to the lighting conditions and the impact of the incident angle α to obtain the dynamically corrected wavelength λ i The reflectivity under as follows: in, is the original wavelength λ i The reflectance under the condition is obtained by the multispectral sensor of the phenotyping platform; g(N,α) is a correction function that describes the influence of light intensity and incident angle on reflectance; Finally, by correcting the reflectivity of all wavelengths, the dynamically corrected spectral reflectance data D is obtained. spectral as follows: 134) Light dynamic correction of RGB color information: The raw color data obtained by the RGB camera contains the intensity values ​​of each pixel in the red, green, and blue channels, expressed as follows: d color ={R,G,B} Through white balance correction, the influence of color temperature under different light sources is corrected; by adjusting the exposure, the consistency of image brightness under different lighting conditions is ensured. The final corrected RGB color data is as follows: 135) Blade posture correction: According to the light response characteristics of the leaves, the calibration model adjusts the posture of the leaves and simulates the change of leaf angle under the change of light intensity; Combining the image data collected by the multispectral camera and the high-definition camera, the correction model compensates for the imaging error caused by uneven illumination through posture adjustment, and the angle Δθ of the blade rotation leaf The formula is as follows: Dth leaf =f(N,α) Then, according to the rotation changes of the blades, the calibration model adjusts the three-dimensional coordinates of the blades; The geometric data (X′, Y′, Z′) obtained by the wind correction model is obtained. The position after the light model correction is transformed by the following rotation matrix. This rotation matrix is ​​used to adjust the three-dimensional position of the blade. According to the angle change of the blade, the spatial position of the blade is corrected to obtain the geometric structure data D after double correction by wind correction and light correction. geometry , 14) Temperature correction: Based on the physical mechanism of stomatal opening and closing, a temperature and humidity regulation model was constructed to simulate the response of stomatal pores under different temperature and humidity conditions, and then deduce the change of leaf temperature, thereby achieving temperature correction; The temperature correction comprises the following steps: 141) Establish a temperature and humidity regulation model based on the stomatal regulation model: Based on the physical mechanism of stomatal opening and closing, a temperature and humidity regulation model was constructed to simulate the response of stomata under different temperature and humidity conditions, and then deduce the changes in leaf temperature. The temperature and humidity adjustment model is described as follows: Where, ΔT stomata is the leaf temperature change caused by the opening and closing of stomata. T and H are the current ambient temperature and humidity, respectively, which are obtained through the temperature and humidity sensor. opt and H opt It is the optimal temperature and humidity for the plant to achieve optimal stomata opening and closing. max , T min , H max and H min are the temperature and humidity extremes when the stomata are fully open or closed, respectively. h1 and h2 are parameters determined based on plant physiological characteristics, indicating the weight of the effect of stomatal regulation on temperature. 142) Dynamic correction of crop temperature in multi-source data: The correction model corrects the temperature data collected by the phenotypic platform sensor based on the real-time data of the temperature and humidity sensor. The temperature correction process formula is as follows: D thermal =T sensor +ΔT stomata Among them, T sensor is the initial temperature data collected by the thermal infrared sensor of the phenotyping platform, D thermal To correct the temperature data for temperature and humidity, by introducing ΔT stomata , compensates for the leaf temperature changes caused by the opening and closing of stomata, ensuring that the temperature measured by the sensor is more consistent with the actual temperature conditions of the crop; 15) Data fusion correction: Perform high-precision data fusion and correction to generate three-dimensional crop point clouds and phenotypic data without external environmental interference, and realize dynamic fusion correction of multivariate data; The fusion correction of the data comprises the following steps: 151) Re-integration based on corrected data: The fusion of multi-source data is represented as weighted average or optimized combination. The goal of the fusion is to generate a unified three-dimensional point cloud data. At the same time, each point cloud point also contains temperature and spectral reflectance information. The fusion process is as follows: D fused =α·D geometry +β·D spectral +γ·D thermal +δ·D color Among them, D fused It is the fused 3D point cloud data, including geometry, spectrum, temperature and color data; D geometry It is the geometric structure data after double correction by wind correction and light correction; D spectral is the spectral reflectance data after dynamic correction; D thermal It is the temperature data corrected by temperature and humidity; D color is the corrected RGB color data; α, β, γ, δ are the weight coefficients of each sensor data, which are adjusted based on the reliability of the data or environmental conditions; 152) Optimize the weight coefficients of each part: Adaptive weight adjustment based on environmental factors, automatically adjust according to environmental conditions under different environmental conditions.

2. The multivariate data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment according to claim 1, characterized in that: The acquisition of the environmental data and the crop phenotype data comprises the following steps: 21) Real-time environmental data, including wind speed, wind direction, light intensity, temperature and humidity, are acquired through environmental monitoring sensors, which are used as external disturbance inputs in the correction model to determine the dynamic adjustment of the model to the crop posture; 22) Collect crop phenotyping data through multiple sensors on a plant phenotyping platform in a field environment, which includes data from lidar, depth camera, thermal imaging camera, multispectral camera and high-definition camera.

3. The multivariate data dynamic fusion correction method for a high-throughput plant phenotyping platform in a field environment according to claim 1, characterized in that: The wind speed and wind direction correction comprises the following steps: 31) The swing and deformation process of the blade is described by the flexible body dynamics equation: As a flexible body, the blade is affected by wind speed and direction, generating positive pressure and shear force. According to the structural characteristics and force conditions of the blade, the correction model uses the flexible body dynamics equation to simulate the swing and deformation process of the blade. The deformation of the blade is approximated by a one-dimensional bending beam model, as shown below: Where y(x,t) is the deformation of the blade, x is the coordinate of the blade along its length, t is time, E is the elastic modulus of the blade material, I is the moment of inertia of the blade section, c is the damping coefficient, which is used to represent dissipative effects such as air resistance, and f(x,t) is the external force of the wind on the blade as a function of wind speed and direction; 32) Combine fluid mechanics, wind speed and wind direction data to calculate the effect of wind on the blades: The real-time data obtained by the anemometer and wind vane: wind speed v and the angle θ between the blade and the wind direction, combined with the wind force formula in fluid mechanics, the correction model calculates the force of wind on the blade; the magnitude of the force is proportional to the square of the wind speed, and the change of wind direction changes the angle at which the blade is subjected to force. The wind force is calculated using the wind force formula in fluid mechanics; Assuming that the wind speed v acts on the blade and the wind force F acts on the unit area of ​​the blade, the expression is as follows: Where ρ is the air density, C D is the drag coefficient of the blade, which depends on the shape of the blade and the wind direction, A is the windward area of ​​the blade, v is the wind speed, which changes with time t, and the square of the wind speed is proportional to the force; In addition, combined with the influence of wind direction, the windward area A of the blade depends on the angle θ between the blade and the wind direction. The specific relationship is as follows: A=A0·cos(θ), Where A0 is the actual area of ​​the blade, θ is the angle between the wind direction and the blade normal, which is measured in real time by the wind vane. When θ = 0, the wind is perpendicular to the blade, and when θ = 90°, the blade is parallel to the wind. f(x, t) is the local external force on the blade at a specific position x and time t. In order to derive the local external force f(x, t) from the overall wind force F, the geometric shape, position and force distribution of the blade are considered. It is assumed that the wind force F acts evenly on the surface of the blade, and the wind force density, i.e., force / unit length, is used to represent the external force at any position on the blade. Because the wind force F changes with time t, it is a function of time. The total length of the blade is L, and f(x, t) is defined as the wind force density, i.e., the force per unit length, as shown below: 33) Dynamic adjustment of blade posture and point cloud data correction: According to the stress of the blades under different wind speeds and directions, the displacement y(x,t) of the blades under the action of wind is solved through the flexible body dynamics equation, and the posture of the blades is dynamically adjusted according to the displacement distribution at different times t and spatial coordinates x; combined with the three-dimensional point cloud data collected by the lidar and depth camera of the high-throughput plant phenotyping platform, the three-dimensional point cloud of the crop is corrected according to the impact of wind on the crop posture, and the blade offset caused by wind speed is dynamically compensated to restore the true three-dimensional shape under windless conditions; Assuming the collected point cloud coordinates are (X, Y, Z), the corrected coordinates after wind action are as follows: X′=X+y x (x,t),Y′=Y+y y (x,t),Z′=Z+ y z (x,t); Among them, y x (x,t),y y (x,t),y z (x, t) is the displacement of the blade in the X, Y, and Z axes calculated by the mechanical equation; Finally, the correction model restores the three-dimensional geometric structure (X′, Y′, Z′) of the blade under windless conditions.

Images