Three-dimensional fluorescence-based method and apparatus for detecting crop growth conditions

Abstract

The application provides a crop growth condition detection method and device based on three-dimensional fluorescence, comprising: making a linear laser and an imaging area of a line-scan hyperspectral camera spatially coincide at an excitation area of a crop canopy surface; controlling a UAV to fly according to a preset multi-angle flight route planning strategy, and collecting multi-angle fluorescence images of the crop canopy under each preset observation angle; performing geometric registration on the multi-angle fluorescence images according to attitude data and position data of the UAV to obtain a multi-angle fluorescence data set aligned in a geometric space; performing inversion solving according to the data set and a fluorescence radiation transmission forward model to obtain an estimated value of a three-dimensional fluorescence body space distribution; performing downward attenuation correction on the estimated value to obtain three-dimensional fluorescence body data, extracting three-dimensional fluorescence feature parameters representing a crop growth condition from the three-dimensional fluorescence body data, and determining the crop growth condition. The three-dimensional reconstruction of crop canopy fluorescence information can be performed based on multi-angle fluorescence images, and three-dimensional comprehensive evaluation of the crop growth condition can be realized.

Description

Technical Field

[0001] This invention relates to the fields of agricultural information technology, UAV remote sensing and spectral detection technology, and in particular to a method and device for detecting crop growth status based on three-dimensional fluorescence. Background Technology

[0002] Three-dimensional phenotypic information of crops is an important indicator for evaluating crop growth status, predicting yield, and assessing stress response. Traditional methods for acquiring three-dimensional crop information mainly rely on active three-dimensional measurement technologies such as lidar or structured light cameras. These technologies can obtain precise three-dimensional structures of the crop canopy, but cannot directly obtain physiological and biochemical information within the crop. On the other hand, multispectral fluorescence imaging technology can sensitively reflect physiological information such as photosynthetic efficiency, chlorophyll content, and secondary metabolites in crops. However, existing fluorescence imaging technologies are mainly limited to two-dimensional planar imaging and cannot obtain the distribution characteristics of fluorescence information in the three-dimensional space of the crop canopy.

[0003] Integrating fluorescence information with three-dimensional structural information to acquire three-dimensional volumetric fluorescence data of crop canopy is of great significance for comprehensively evaluating crop growth status. For example, differences in light interception at different canopy depths lead to significant vertical gradients in fluorescence emission intensity, and plants under disease stress also exhibit spatial differentiation in fluorescence responses across different canopy layers. This three-dimensional fluorescence information cannot be captured by two-dimensional fluorescence imaging.

[0004] Existing UAV-borne hyperspectral imaging systems mostly employ a pushbroom array approach, acquiring two-dimensional planar images of the crop canopy. These systems typically use surface light sources such as pulsed LED arrays or halogen lamps as fluorescence excitation sources. Surface light sources disperse the excitation light over a large area, resulting in low excitation energy density per unit area and a low signal-to-noise ratio for the fluorescence signal. Furthermore, surface light source systems are power-intensive and bulky, making them unsuitable for the low-power and miniaturization requirements of UAV platforms. Current three-dimensional fluorescence reconstruction methods based on the combination of pushbroom array imaging and surface light sources typically use lidar or motion-restoring structure algorithms to first acquire a three-dimensional point cloud of the crop surface, then simply overlay or map the two-dimensional fluorescence spectral image onto the outer surface of the three-dimensional point cloud as a texture. This method is essentially a post-processing visual fusion of two-dimensional geometry and planar spectrum, failing to reflect the complex collision, absorption, and re-emission scattering mechanisms of fluorescent photons within the plant leaf community, unable to penetrate the canopy interior, and lacking the spatial decoupling capability for downward attenuation of actively excited light and upward absorption of fluorescence. Therefore, it cannot achieve quantitative inversion of fluorescence radiance at the true three-dimensional system level within the plant canopy.

[0005] Furthermore, due to the lack of true three-dimensional fluorophore data, existing crop growth assessments and disease classifications still heavily rely on two-dimensional planar remote sensing vegetation indices, such as the Normalized Difference Vegetation Index (NDVI) and the Sun-Induced Chlorophyll Fluorescence Index (SCI), lacking quantitative evaluation methods for crop growth based on true three-dimensional fluorescence physiological and biochemical characteristics. These planar indices are highly susceptible to non-physiological interference from canopy three-dimensional geometry, such as leaf tilt distribution, planting density, and background soil residuals.

[0006] Therefore, how to efficiently acquire multi-angle fluorescence images of crop canopy on a drone platform, and how to perform three-dimensional reconstruction of crop canopy fluorescence information based on multi-angle fluorescence images to achieve a three-dimensional comprehensive evaluation of crop growth status are key technical problems that urgently need to be solved in this field. Summary of the Invention

[0007] This invention provides a method and apparatus for detecting crop growth status based on three-dimensional fluorescence, which solves the technical problem of how to efficiently acquire multi-angle fluorescence images of crop canopy on a UAV platform, and perform three-dimensional reconstruction of crop canopy fluorescence information based on multi-angle fluorescence images, so as to realize a three-dimensional comprehensive evaluation of crop growth status.

[0008] This invention provides a method for detecting crop growth status based on three-dimensional fluorescence, comprising the following steps: Adjust the relative geometric position between the linear laser on the drone and the line-scan hyperspectral camera so that the excitation area of ​​the linear laser on the crop canopy surface coincides with the imaging area of ​​the line-scan hyperspectral camera in space. The drone is controlled to fly according to a preset multi-angle flight path planning strategy. When the excitation area and the imaging area are spatially overlapped, the line-scan hyperspectral camera is used to acquire multi-angle fluorescence images of the crop canopy at each preset observation angle, and the attitude data and position data of the drone corresponding to each observation angle are acquired simultaneously. Geometric registration is performed on the multi-angle fluorescence images based on the pose data and position data to obtain a multi-angle fluorescence dataset aligned in geometric space. Based on the multi-angle fluorescence dataset and the forward fluorescence radiative transfer model, the estimated value of the three-dimensional phosphor spatial distribution is obtained by inversion solution; The estimated spatial distribution of the three-dimensional phosphors is corrected for downward attenuation based on the propagation path of the excitation light within the canopy to obtain the three-dimensional phosphor data. Three-dimensional fluorescence characteristic parameters characterizing crop growth status are extracted from the three-dimensional fluorescence data, and crop growth status is determined based on the three-dimensional fluorescence characteristic parameters.

[0009] In one embodiment, adjusting the relative geometric position between the linear laser mounted on the UAV and the line-scan hyperspectral camera, so that the excitation area of ​​the linear laser on the crop canopy surface spatially coincides with the imaging area of ​​the line-scan hyperspectral camera, includes: Adjust the relative geometric position between the linear laser and the line-scan hyperspectral camera so that the projection linewidth of the linear laser on the ground is equal to the coverage linewidth of the instantaneous field of view of the line-scan hyperspectral camera on the ground.

[0010] In one embodiment, the multi-angle route planning strategy includes: Set a set of observation angles that includes the nadir observation angle and multiple tilt observation angles; Multiple main flight routes are planned in the target area. The spacing between adjacent main flight routes is determined based on the preset lateral overlap rate of the flight routes and the effective ground observation swath width under the maximum tilt observation angle. Multiple repeated flights are performed along each main flight route, with each flight using a different observation angle from the set of observation angles. Additional data collection is performed by adding cross-validation routes that form a preset angle with the main route.

[0011] In one embodiment, the step of geometrically registering the multi-angle fluorescence images based on the pose data and position data to obtain a geometrically aligned multi-angle fluorescence dataset includes: Based on the location data and the rotation matrix composed of the heading angle, pitch angle and roll angle in the attitude data, a projection mapping function between the fluorescence image pixel coordinates and the ground geographic coordinates at each observation angle is established. Based on the ground three-dimensional geographic coordinates of the fluorescence control points deployed in the target area and the actual observed pixel coordinates in the fluorescence image, a reprojection error loss function is constructed. The reprojection error loss function is minimized by an optimization algorithm to obtain the corrected UAV position coordinates and rotation matrix. The updated projection mapping function is obtained by updating the projection mapping function based on the corrected UAV position coordinates and rotation matrix; Extraction and matching of canopy-named feature points from fluorescence images under different observation angles yielded multiple pairs of canopy-named feature point pairs. The transformation matrix was solved by minimizing the nonlinear distance residuals of all canopy-named feature point pairs. The fluorescence images at all observation angles are resampled to a common ground geographic grid coordinate system using the updated projection mapping function and the transformation matrix to obtain the geometrically aligned multi-angle fluorescence dataset.

[0012] In one embodiment, the step of obtaining the estimated three-dimensional phosphor spatial distribution based on the multi-angle fluorescence dataset and the forward fluorescence radiative transfer model includes: The three-dimensional object space of the crop canopy is discretized into a three-dimensional voxel grid composed of multiple voxels; The fluorescence radiative transport forward model is discretized and accumulated along a discretized three-dimensional voxel grid by ray tracing, so that the canopy surface fluorescence radiance for each observation angle and each fluorescence band in the multi-angle fluorescence dataset is expressed as a weighted sum of the fluorescence intensities of each voxel, thus obtaining the discretized radiative transport equation corresponding to each observation angle and each fluorescence band. Based on the discretized radiative transfer equations for all observation angles and all fluorescence bands, a linear matrix equation for multi-angle, multi-band joint observation is constructed. Construct an objective function based on the linear matrix equation, and solve the objective function to obtain an estimate of the spatial distribution of the three-dimensional phosphor.

[0013] In one embodiment, the step of performing downward attenuation correction on the estimated spatial distribution of the three-dimensional phosphor based on the propagation path of the excitation light within the canopy to obtain three-dimensional phosphor data includes: For each voxel, the actual excitation illuminance received by the voxel is determined based on the reference excitation illuminance at the top of the canopy, the extinction coefficient of the upper voxels through which the excitation light passes before reaching the voxel, and the geometric distance through each upper voxel. The corrected fluorescence intensity is determined based on the fluorescence intensity of the voxels in the estimated spatial distribution of the three-dimensional phosphor and the actual excitation illuminance received by the voxels; the corrected fluorescence intensities of all voxels constitute the three-dimensional phosphor data.

[0014] In one embodiment, the three-dimensional fluorescence characteristic parameters include at least one of the following: canopy fluorescence vertical distribution index, total three-dimensional fluorescence, fluorescence three-dimensional spatial heterogeneity index, canopy deep fluorescence ratio, and canopy fluorescence structure centroid index; The canopy fluorescence vertical distribution index includes the average fluorescence intensity of each spatial level. The average fluorescence intensity of each spatial level is obtained by dividing the crop canopy into multiple spatial levels at equal intervals along the vertical height direction and averaging the corrected fluorescence intensity of voxels belonging to the same spatial level. The total amount of three-dimensional fluorescence is determined based on the corrected fluorescence intensity of each voxel in the far-infrared band and the physical volume of each voxel. The fluorescence three-dimensional spatial heterogeneity index is determined based on the ratio of the standard deviation of the corrected fluorescence intensity of each voxel in the far-red band to the global average corrected fluorescence intensity. The canopy deep fluorescence ratio is determined based on the ratio of the corrected fluorescence intensity of deep voxels in the red band to the corrected fluorescence intensity in the far-red band. The deep voxels are voxels in a spatial region at a height less than or equal to a preset threshold from the ground. The centroid index of the canopy fluorescence structure is determined based on the corrected fluorescence intensity of each voxel in the far-infrared band and the absolute height coordinates of each voxel.

[0015] In one embodiment, the crop growth status includes a crop growth status level. Determining the crop growth status based on the three-dimensional fluorescence feature parameters includes: The three-dimensional fluorescence feature parameters are normalized. The three-dimensional fluorescence growth index was determined based on the normalized total three-dimensional fluorescence, the normalized three-dimensional fluorescence spatial heterogeneity index, the normalized canopy deep fluorescence ratio, and the normalized canopy fluorescence structure centroid index. The growth status evaluation vector is composed of the three-dimensional fluorescence growth index and the ratio of fluorescence in the canopy depth. The Euclidean distance between the growth status evaluation vector and the cluster center vectors corresponding to multiple preset growth status levels is calculated. The crop growth status level is determined according to the minimum distance principle.

[0016] In one embodiment, the method further includes the following steps during the acquisition of multi-angle fluorescence images: If the UAV changes its flight altitude or the gimbal tilt angle changes, the UAV's flight altitude and the inertial stabilized gimbal tilt angle are acquired in real time, and the driving current of the linear laser is adjusted according to the flight altitude, the tilt angle, and the reference flight altitude.

[0017] The present invention also provides a crop growth status detection device based on three-dimensional fluorescence, comprising: A linear laser emits excitation light, which is shaped into a linear beam by a laser beam rectifier and projected onto the crop canopy, whereby the linear beam forms an excitation region on the surface of the crop canopy. A line-scan hyperspectral camera, the line-scan hyperspectral camera having an imaging area, the line-scan hyperspectral camera being used to acquire multi-angle fluorescence images of the crop canopy generated under the excitation light; An attitude measurement module is used to acquire the attitude and position data of the UAV in real time. An inertial stabilizing gimbal is used to support the linear laser and the line-scanning hyperspectral camera, and supports adjusting the observation angle within a preset angle range; A central control processing module is electrically connected to the linear laser, the line-scanning hyperspectral camera, and the attitude measurement module. The central control processing module is used to control the timing synchronization of the linear laser and the line-scanning hyperspectral camera, and to execute the crop growth status detection method based on three-dimensional fluorescence as described above. A data storage module, electrically connected to the central control and processing module, is used to store the multi-angle fluorescence image, the attitude data, the position data, and the three-dimensional phosphor data; A wireless communication module, which is electrically connected to the central control processing module, is used to realize data transmission between the UAV and the ground station; The power management module provides power to the linear laser, the line-scan hyperspectral camera, the attitude measurement module, the central control processing module, the data storage module, the wireless communication module, and the inertial stabilization gimbal.

[0018] The present invention provides a method and apparatus for detecting crop growth status based on three-dimensional fluorescence. By adjusting the relative geometric position between a linear laser mounted on a UAV and a line-scan hyperspectral camera, the excitation area of ​​the linear laser on the crop canopy surface and the imaging area of ​​the line-scan hyperspectral camera are spatially overlapped. This concentrates the excitation light energy within the camera's imaging area, improving energy utilization and enabling the acquisition of high signal-to-noise ratio fluorescence signals even under UAV payload constraints. By controlling the UAV to fly according to a preset multi-angle flight path planning strategy, and with the excitation and imaging areas spatially overlapping, the line-scan hyperspectral camera acquires multi-angle fluorescence images of the crop canopy at various preset observation angles. This allows a single line-scan hyperspectral camera to acquire fluorescence images of different vertical layers of the canopy and different observation perspectives, achieving efficient acquisition of multi-angle fluorescence images of the crop canopy on a UAV platform. By performing geometric registration of the multi-angle fluorescence images based on attitude and position data, high-precision geometric constraints are provided for subsequent three-dimensional reconstruction. By inverting and solving a multi-angle fluorescence dataset and a forward fluorescence radiative transport model, the estimated spatial distribution of three-dimensional phosphors is obtained. This decomposes the fluorescence radiance observed on the canopy surface into fluorescence emission intensities at various spatial locations within the canopy, enabling three-dimensional reconstruction of crop canopy fluorescence information based on multi-angle fluorescence images. Downward attenuation correction is applied to the estimated spatial distribution of the three-dimensional phosphors based on the propagation path of the excitation light within the canopy. This ensures that the downward attenuation caused by leaf absorption and scattering as the excitation light penetrates deeper into the canopy is compensated for voxel-by-voxel, eliminating surface shading caused by the vegetation canopy geometry and accurately characterizing the fluorescence emission intensity of deep voxels. By extracting three-dimensional fluorescence characteristic parameters representing crop growth status from the three-dimensional phosphor data and determining the crop growth status, a comprehensive three-dimensional evaluation of crop growth status is achieved. This allows the evaluation of crop growth status to move from a two-dimensional plane to three-dimensional space, accurately characterizing the physiological and biochemical heterogeneity within the crop canopy. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the overall framework of the crop growth status detection device based on three-dimensional fluorescence provided by the present invention.

[0021] Figure 2 This is a flowchart illustrating the crop growth status detection method based on three-dimensional fluorescence provided by the present invention.

[0022] Figure 3 This is a schematic diagram of the geometric relationship between linear laser excitation and line scan acquisition provided by the present invention.

[0023] Reference numerals: 1. Linear laser; 101. Laser body; 102. Laser beam rectifier; 2. Linear scan hyperspectral camera; 201. Camera body; 202. Camera lens; 3. Attitude measurement module; 4. Inertial stabilized gimbal; 5. Central control and processing module; 6. Data storage module; 7. Wireless communication module; 8. Power management module; 9. Laser transmission fiber; 10. Quadcopter UAV. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] The following is combined Figures 1 to 3 The present invention describes a method and apparatus for detecting crop growth status based on three-dimensional fluorescence.

[0026] This invention provides a crop growth status detection device based on three-dimensional fluorescence, referring to... Figure 1 As shown, it includes: a linear laser 1, a line-scanning hyperspectral camera 2, an attitude measurement module 3, an inertial stabilizing gimbal 4, a central control and processing module 5, a data storage module 6, a wireless communication module 7, and a power management module 8.

[0027] The linear laser 1 emits excitation light, which is shaped into a linear beam by the laser beam rectifier 102 and projected onto the crop canopy. The linear beam forms an excitation region on the surface of the crop canopy. Specifically, the linear laser 1 is a high-power linear semiconductor laser used to emit excitation light. The excitation light emitted by the linear laser 1 is transmitted to the laser beam rectifier 102 via the laser transmission fiber 9. The laser beam rectifier 102 shapes the excitation light into a linear beam and projects it onto the crop canopy, forming a linear excitation region on the surface of the crop canopy. This invention uses a linear laser 1 as the fluorescence excitation source, which, compared to traditional surface light source solutions, has higher energy utilization and better excitation uniformity, and can still obtain a high signal-to-noise ratio fluorescence signal under the limited payload conditions of UAVs.

[0028] The line-scan hyperspectral camera 2 has an imaging area, which is used to acquire multi-angle fluorescence images of the crop canopy generated under the excitation light. The imaging area of ​​the line-scan hyperspectral camera 2 precisely coincides with the excitation area of ​​the linear laser 1 on the surface of the crop canopy, ensuring a precise spatial correspondence between the excitation area and the imaging area. The line-scan hyperspectral camera 2 adopts a line-scan acquisition mode, which, together with the linear excitation source of the linear laser 1, achieves precise spatial matching between the excitation light and the acquisition field of view. The spectral range of the line-scan hyperspectral camera 2 is 400 nm to 800 nm, which can cover the main emission bands of plant chlorophyll fluorescence.

[0029] The attitude measurement module 3 is used to acquire the attitude and position data of the UAV in real time. The attitude data includes the UAV's heading angle, pitch angle, and roll angle, while the position data includes the UAV's three-dimensional coordinates in object space. The attitude measurement module 3 provides accurate attitude and position data for geometric correction and three-dimensional reconstruction of multi-angle fluorescence images.

[0030] The inertial stabilization gimbal 4 supports the linear laser 1 and the line-scanning hyperspectral camera 2, and allows adjustment of the observation angle within a preset angle range. Specifically, the inertial stabilization gimbal 4 is a three-axis inertial stabilization gimbal, used to support the linear laser 1 and the line-scanning hyperspectral camera 2, isolate the drone's flight vibrations, and ensure image quality. The inertial stabilization gimbal 4 supports tilt functionality, allowing adjustment of the observation angle within a preset angle range, thereby enabling the acquisition of multi-angle fluorescence images of the crop canopy.

[0031] The central control processing module 5 is electrically connected to the linear laser 1, the line-scan hyperspectral camera 2, and the attitude measurement module 3. The central control processing module 5 controls the timing synchronization of the linear laser 1 and the line-scan hyperspectral camera 2, and executes the crop growth status detection method based on three-dimensional fluorescence as described in this invention. Specifically, the central control processing module 5 uses an embedded GPU computing platform and is electrically connected to the linear laser 1, the line-scan hyperspectral camera 2, and the attitude measurement module 3. The central control processing module 5 controls the timing synchronization of the linear laser 1 and the line-scan hyperspectral camera 2, preprocesses the fluorescence images acquired by the line-scan hyperspectral camera 2, and executes the crop growth status detection method based on three-dimensional fluorescence as described in this invention, including steps such as geometric registration of multi-angle fluorescence images, inversion solution based on the fluorescence radiative transport forward model, downlink attenuation correction, extraction of three-dimensional fluorescence feature parameters, and determination of crop growth status. The central control processing module 5 is also responsible for communication control with the ground station.

[0032] The data storage module 6 is electrically connected to the central control and processing module 5 and is used to store the multi-angle fluorescence image, the attitude data, the position data, and the three-dimensional phosphor data. The wireless communication module 7 is electrically connected to the central control processing module 5 and is used to realize data transmission between the UAV and the ground station. The power management module 8 supplies power to the linear laser 1, the line-scan hyperspectral camera 2, the attitude measurement module 3, the central control processing module 5, the data storage module 6, the wireless communication module 7, and the inertial stabilization gimbal 4.

[0033] The crop growth status detection method based on three-dimensional fluorescence provided by the present invention will be described below. The crop growth status detection method based on three-dimensional fluorescence described below can be referred to in correspondence with the crop growth status detection device based on three-dimensional fluorescence described above.

[0034] Figure 2 This is a flowchart illustrating the crop growth status detection method based on three-dimensional fluorescence provided by the present invention, as shown below. Figure 2 As shown, the method includes the following: S110. Adjust the relative geometric position between the linear laser 1 and the line-scan hyperspectral camera 2 carried by the UAV so that the excitation area of ​​the linear laser 1 on the surface of the crop canopy coincides with the imaging area of ​​the line-scan hyperspectral camera 2 in space.

[0035] In this invention, the excitation light emitted by the linear laser 1 is shaped into a linear beam by the laser beam rectifier 102 and projected onto the surface of the crop canopy to form an excitation region. The line-scanning hyperspectral camera 2 has an imaging region, which is the strip area of ​​the ground covered by the camera's instantaneous field of view. To fully utilize the excitation light energy and improve the signal-to-noise ratio of the fluorescence signal, this invention adjusts the relative geometric position between the linear laser 1 and the line-scanning hyperspectral camera 2 to make the excitation region and the imaging region precisely coincide in space.

[0036] S120. Control the UAV to fly according to the preset multi-angle flight path planning strategy. When the excitation area and the imaging area are spatially overlapped, the line-scan hyperspectral camera 2 collects multi-angle fluorescence images of the crop canopy at each preset observation angle, and simultaneously acquires the attitude data and position data of the UAV corresponding to each observation angle.

[0037] In this invention, a multi-angle flight path planning strategy is used to achieve efficient acquisition of multi-angle fluorescence images of crop canopies. This strategy sets up a set of observation angles, including nadir observation angles and multiple tilt observation angles. Multiple main flight paths are planned within the target area, and multiple repeated flights are performed along each main flight path. Each flight uses a different observation angle from the set of observation angles, thereby completing a full-coverage scan of the target area at each observation angle. Through this strategy, the UAV can quickly acquire multi-angle fluorescence images of the crop canopy at different tilt observation angles in a single or a small number of flights.

[0038] With the excitation and imaging regions spatially overlapping, the UAV flies according to a pre-defined multi-angle flight path planning strategy. For each pre-defined observation angle in the set of observation angles, the UAV performs an independent flight, during which the line-scan hyperspectral camera 2 acquires a fluorescence image of the crop canopy at that observation angle. Each flight uses a different observation angle, thereby obtaining fluorescence images corresponding to each pre-defined observation angle, which together constitute a multi-angle fluorescence image. During the acquisition process, the attitude data and position data of the UAV corresponding to each observation angle are acquired simultaneously. Among them, the attitude data includes the UAV's heading angle, pitch angle, and roll angle, which are provided by the inertial measurement unit; the position data includes the UAV's three-dimensional coordinates in the object space, which are provided by the Real-Time Kinematic Global Positioning System (RTK-GPS). The inertial measurement unit and RTK-GPS are integrated into the attitude measurement module 3.

[0039] S130. Geometric registration is performed on the multi-angle fluorescence image based on the pose data and position data to obtain a multi-angle fluorescence dataset aligned in geometric space.

[0040] Fluorescence images acquired from different observation angles exhibit perspective distortion and geometric distortion. This invention performs geometric registration on multi-angle fluorescence images based on the UAV attitude and position data, unifying the fluorescence images from different observation angles into a common ground geographic grid coordinate system, resulting in a geometrically aligned multi-angle fluorescence dataset. After geometric registration, the fluorescence images from each angle have a consistent spatial coordinate reference, eliminating the geometric distortion caused by angle changes.

[0041] S140. Based on the multi-angle fluorescence dataset and the forward fluorescence radiative transfer model, perform inversion to obtain the estimated value of the three-dimensional phosphor spatial distribution.

[0042] In this invention, a forward fluorescence radiative transport model is used to describe the relationship between the fluorescence radiance received by the line-scan hyperspectral camera 2 on the canopy surface and the fluorescence emission intensity of each voxel within the canopy. This forward model is used for subsequent inversion solutions.

[0043] The inversion solution is the process of calculating the fluorescence emission intensity of each voxel within the canopy using a forward fluorescence radiative transport model based on the multi-angle fluorescence dataset. The multi-angle fluorescence dataset consists of fluorescence images observed from different angles, with pixel values ​​in each image representing the fluorescence radiance of the canopy surface. This invention discretizes the three-dimensional object space of the crop canopy into multiple voxels, and performs ray tracing accumulation along the discretized voxel grid using the forward fluorescence radiative transport model. This allows the fluorescence radiance of the canopy surface to be expressed as a weighted sum of the fluorescence emission intensities of each voxel, thereby establishing a mapping relationship between the fluorescence radiance of the canopy surface and the fluorescence emission intensities of each voxel within the canopy. By inverting this mapping relationship, the estimated three-dimensional spatial distribution of fluorescence emission intensity of each voxel within the canopy is obtained, i.e., the estimated three-dimensional phosphor spatial distribution.

[0044] S150. The estimated spatial distribution of the three-dimensional phosphor is corrected by downward attenuation based on the propagation path of the excitation light within the canopy to obtain the three-dimensional phosphor data.

[0045] Downward attenuation correction is a process of compensating for energy attenuation caused by leaf absorption and scattering of excitation light as it passes through the canopy. When the excitation light emitted by the linear laser 1 mounted on the UAV penetrates downwards from the top of the canopy, it loses some energy due to leaf absorption and scattering at each voxel layer it passes through, resulting in lower actual excitation light illuminance received by deeper voxels compared to surface voxels. This invention performs downward attenuation correction on the estimated spatial distribution of the three-dimensional phosphor based on the propagation path of the excitation light within the canopy. Specifically, it compensates for the fluorescence emission intensity of each voxel according to the propagation path of the excitation light, obtaining the corrected fluorescence intensity of each voxel. The corrected fluorescence intensities of all voxels constitute the three-dimensional phosphor data. After downward attenuation correction, the surface shading caused by the vegetation canopy geometry is eliminated, and the fluorescence emission intensity of deeper voxels can be accurately characterized.

[0046] S160. Extract three-dimensional fluorescence characteristic parameters that characterize crop growth status from the three-dimensional fluorescent data, and determine crop growth status based on the three-dimensional fluorescence characteristic parameters.

[0047] Three-dimensional fluorescence characteristic parameters are one or more quantitative indicators extracted from three-dimensional fluorophore data that reflect the spatial distribution characteristics of fluorescence intensity within the crop canopy. These parameters may include those reflecting the vertical distribution of fluorescence intensity within the canopy, those reflecting the quantitative relationship between crop biomass and overall photosynthetic activity, those reflecting the degree of spatial disorder in physiological activity within the canopy, those reflecting the physiological state of the deeper regions beneath the canopy, and those reflecting the variation of the center of fluorescence activity with depth within the canopy. This invention utilizes these three-dimensional fluorescence characteristic parameters to construct a comprehensive evaluation index. Based on this comprehensive evaluation index, the crop growth status level is determined, and the crop growth status level is used to characterize the crop's growth condition.

[0048] The crop growth status detection method based on three-dimensional fluorescence provided by this invention adjusts the relative geometric position between a linear laser 1 and a line-scanning hyperspectral camera 2 mounted on a UAV, so that the excitation area of ​​the linear laser 1 on the crop canopy surface and the imaging area of ​​the line-scanning hyperspectral camera 2 spatially overlap. This concentrates the excitation light energy within the camera's imaging area, improving energy utilization and enabling the acquisition of high signal-to-noise ratio fluorescence signals even under UAV payload constraints. By controlling the UAV to fly according to a preset multi-angle flight path planning strategy, and with the excitation and imaging areas spatially overlapping, the line-scanning hyperspectral camera 2 acquires multi-angle fluorescence images of the crop canopy at various preset observation angles. This allows a single line-scanning hyperspectral camera 2 to acquire fluorescence images of different vertical layers of the canopy and different observation perspectives, achieving efficient acquisition of multi-angle fluorescence images of the crop canopy on a UAV platform. By performing geometric registration of the multi-angle fluorescence images based on attitude and position data, high-precision geometric constraints are provided for subsequent three-dimensional reconstruction. By inverting and solving a multi-angle fluorescence dataset and a forward fluorescence radiative transport model, the estimated spatial distribution of three-dimensional phosphors is obtained. This decomposes the fluorescence radiance observed on the canopy surface into fluorescence emission intensities at various spatial locations within the canopy, enabling three-dimensional reconstruction of crop canopy fluorescence information based on multi-angle fluorescence images. Downward attenuation correction is applied to the estimated spatial distribution of the three-dimensional phosphors based on the propagation path of the excitation light within the canopy. This ensures that the downward attenuation caused by leaf absorption and scattering as the excitation light penetrates deeper into the canopy is compensated for voxel-by-voxel, eliminating surface shading caused by the vegetation canopy geometry and accurately characterizing the fluorescence emission intensity of deep voxels. By extracting three-dimensional fluorescence characteristic parameters representing crop growth status from the three-dimensional phosphor data and determining the crop growth status, a comprehensive three-dimensional evaluation of crop growth status is achieved. This allows the evaluation of crop growth status to move from a two-dimensional plane to three-dimensional space, accurately characterizing the physiological and biochemical heterogeneity within the crop canopy.

[0049] In one embodiment, adjusting the relative geometric position between the linear laser 1 mounted on the UAV and the line-scan hyperspectral camera 2, so that the excitation area of ​​the linear laser 1 on the crop canopy surface coincides spatially with the imaging area of ​​the line-scan hyperspectral camera 2, includes: Adjust the relative geometric position between the linear laser 1 and the line-scan hyperspectral camera 2 so that the projection linewidth of the linear laser 1 on the ground is equal to the coverage linewidth of the instantaneous field of view of the line-scan hyperspectral camera 2 on the ground.

[0050] Specifically, such as Figure 3 As shown in the schematic diagram of the geometric relationship between linear laser excitation and line-scan acquisition, the linear laser 1 includes a laser body 101 and a laser beam rectifier 102, and the line-scan hyperspectral camera 2 includes a camera body 201 and a camera lens 202. 10 is a quadcopter UAV. To establish the spatial overlap between the linear laser 1 and the line-scan hyperspectral camera 2, their relative geometric positions need to be adjusted so that the projection linewidth of the linear laser 1 on the ground is equal to the coverage linewidth corresponding to the instantaneous field of view of the line-scan hyperspectral camera 2 on the ground.

[0051] Let the flight altitude of the drone be... The linewidth of a linear laser beam projected onto the ground is denoted as . The coverage linewidth corresponding to the instantaneous field of view of the ground of the line-scan hyperspectral camera 2 is denoted as . The field of view of the line-scan hyperspectral camera 2 and the flight altitude of drones The decision is: , formula (1); By adjusting the relative geometric positions of the linear laser 1 and the line-scanning hyperspectral camera 2, the projection linewidth of the linear laser on the ground is made equal to the coverage linewidth corresponding to the instantaneous field of view of the line-scanning hyperspectral camera 2 on the ground, that is: , formula (2); Therefore, we can conclude that: , formula (3); This equation establishes the spatial overlap between the linear laser 1 and the line-scanning hyperspectral camera 2, making the excitation region and the imaging region correspond precisely in space.

[0052] After completing spatial alignment adjustments, the UAV flies at a constant speed along the flight path. To ensure continuous spatial sampling along the flight direction, the ground distance between adjacent scan lines in the flight direction is defined as the ground sampling distance in the flight direction. , Flight speed of drones Sampling period of line-scan hyperspectral camera 2 The decision is recorded as: , formula (4); To ensure continuous sampling along the flight path, it is necessary to make Not greater than the horizontal ground sampling distance ,Right now: , formula (5); The lateral ground sampling distance is the ground sampling distance along the line scan direction. The value is determined by the drone's flight altitude, the field of view of the line-scan hyperspectral camera 2, and the effective number of pixels in the line-scan direction, i.e.: , formula (6); in, Let be the effective number of pixels in the line scan direction. Solving the above formulas (4), (5), and (6) simultaneously, we obtain the constraint condition for the sampling period of the line scan hyperspectral camera 2 as follows: , formula (7); This constraint limits the upper limit of the sampling period for the line-scan hyperspectral camera 2, given the UAV's flight speed, altitude, camera field of view, and effective pixel count. By adhering to this constraint, the UAV can ensure no spatial omissions between adjacent scan lines, achieving continuous hyperspectral imaging of surface targets. This constraint is determined after the aforementioned spatial overlap adjustment, providing a basis for sampling period parameters for subsequent continuous acquisition of multi-angle fluorescence images.

[0053] This invention achieves precise spatial matching between the excitation and imaging regions by adjusting the relative geometric positions of the linear laser 1 and the line-scanning hyperspectral camera 2. This ensures that the projection linewidth of the linear laser on the ground is equal to the coverage linewidth corresponding to the instantaneous field of view of the line-scanning hyperspectral camera 2 on the ground. This concentrates the excitation light energy within the camera's imaging area, avoiding the problem of excitation light energy dispersion in traditional surface light source schemes, and improving excitation light utilization and fluorescence signal-to-noise ratio. By determining the sampling period constraints of the line-scanning hyperspectral camera 2 based on the UAV's flight altitude, camera field of view angle, and the effective number of pixels in the line-scanning direction, and by collecting data according to these constraints, it ensures that there is no spatial omission between adjacent scan lines along the flight direction. This achieves continuous hyperspectral imaging of surface targets, providing spatially continuous multi-angle fluorescence image data for subsequent 3D reconstruction.

[0054] In one embodiment, a multi-angle flight path planning strategy was designed to acquire multi-angle fluorescence images of the crop canopy. The multi-angle flight path planning strategy comprises four parts: setting the set of observation angles, main flight path planning, multi-angle repeated acquisition, and cross-validation flight path planning. Specifically, the multi-angle flight path planning strategy includes: S1001, Set up a set of observation angles that includes nadir observation angles and multiple tilt observation angles; definition A set of observation angles for: , formula (8); in, Indicates the angle of observation at the nadir. This indicates the tilt observation angle. , This is the preset number of observation angles. This represents the maximum tilt observation angle. This set of observation angles covers multiple observation directions from the nadir to the maximum tilt observation angle, providing angular parameters for subsequent acquisition of multi-angle fluorescence images.

[0055] S1002. Plan multiple main routes in the target area. The spacing between adjacent main routes is determined based on the preset lateral overlap rate of the routes and the effective ground observation swath width under the maximum tilt observation angle.

[0056] In the main flight path planning, the width of the target area perpendicular to the flight direction is set as follows: The maximum tilt observation angle is The effective ground observation swath at the maximum tilt observation angle is To ensure continuous coverage of the target area under different observation angles, the lateral overlap rate of the UAV's preset flight path must be met. Spacing between adjacent main routes The lateral overlap rate of the flight path is determined based on the effective ground observation swath and the preset flight path lateral overlap rate, i.e.: , formula (9); Number of main routes required for the target area for: , formula (10); Here, [] indicates rounding up. This yields the number of main flight paths required to cover the entire target area and the spacing between adjacent main flight paths, providing a flight path layout for subsequent multi-angle repeated data collection.

[0057] S1003. Perform multiple repeated flights along each main route, with each flight using a different observation angle from the set of observation angles.

[0058] In multi-angle repeated acquisition, for the k-th (k=1,2,…,M) main flight path, the UAV performs multiple repeated flights, and the sequence of observation angles for each flight is represented as follows: , formula (11); in, This represents the sequence of observation angles for the k-th main flight path.

[0059] When a drone makes multiple repeated flights along the same main flight path, it adjusts to the corresponding observation angle in each of the different sorties. A full-coverage scan of the target area was completed from each observation angle. Therefore, the total number of flight data collections during the main flight phase was [number missing]. for: , formula (12); in, Number of main routes The number of preset observation angles is set. In this way, a set of fluorescence images is obtained for each main flight path at each observation angle, providing multi-view observation data for subsequent three-dimensional reconstruction of multi-angle fluorescence images.

[0060] S1004. Add cross-verification routes that form a preset angle with the main route for supplementary data collection.

[0061] In cross-validation route planning, a cross-validation route is added to the main route at a predetermined angle. The angle between the cross-validation route and the main route... (That is, the aforementioned preset included angle) is: , formula (13); in, For angular deviation tolerance, the preferred option is... This refers to orthogonal cross-validation. The data obtained from the cross-validation route forms a multi-baseline, multi-angle three-dimensional observation geometric relationship with the main route data, which is used to improve the accuracy of crop canopy three-dimensional information recovery and fluorescence spatial information reconstruction.

[0062] This invention enables a single line-scan hyperspectral camera 2 to acquire fluorescence images from multiple observation angles by setting an observation angle set that includes the nadir observation angle and multiple tilt observation angles. This avoids the complexity of multi-camera systems while ensuring spatial consistency of multi-angle data. By determining the spacing and number of adjacent main flight paths based on the target area width, the effective ground observation swath width at the maximum tilt observation angle, and the preset lateral overlap rate of the flight paths, the invention ensures continuous coverage of the target area under different observation angle conditions and avoids coverage gaps caused by changes in observation angle.

[0063] By performing multiple repeated flights along each main flight path, with each flight employing different observation angles from the set of observation angles, a complete set of fluorescence images was obtained at each observation angle. This provided multi-view observation data for subsequent 3D reconstruction based on multi-angle fluorescence images. By adding cross-validation flight paths at preset angles to the main flight paths, the data acquired by the cross-validation flight paths formed a multi-baseline, multi-angle stereoscopic observation geometric relationship with the main flight path data, improving the accuracy of crop canopy 3D information recovery and fluorescence spatial information reconstruction.

[0064] In one embodiment, the method further includes the following steps during the acquisition of multi-angle fluorescence images: If the UAV changes its flight altitude or the gimbal tilt angle changes, the flight altitude of the UAV and the tilt angle of the inertial stabilized gimbal 4 are acquired in real time, and the driving current of the linear laser 1 is adjusted according to the flight altitude, the tilt angle and the reference flight altitude.

[0065] Specifically, during the acquisition of multi-angle fluorescence images, when the UAV changes its flight altitude or the gimbal tilts and the observation angle changes, the area of ​​the stripe covering the crop canopy ground by the laser changes, leading to a change in the excitation energy density per unit area and affecting the quantitative inversion of fluorescence intensity. Let the reference flight altitude be... The reference laser power is The drone's flight altitude is H, and the tilt observation angle of the inertial stabilized gimbal 4 is [missing information]. The dynamic compensation model for laser power is then: , formula (14); in, This represents the laser output power at the current observation angle, i.e., at the flight altitude H and the side-roll observation angle. Under the given conditions, the required output power value for linear laser 1. Reference laser power. This refers to the reference flight altitude. Under certain conditions, this is a reference value for the output power of linear laser 1. Reference flight altitude. It is a pre-set standard flight altitude at which the excitation energy density received per unit area of ​​the ground target is at the baseline level without power compensation.

[0066] Then the power compensation coefficient for: , formula (15); The UAV loading control system adjusts the laser drive current in real time according to the calculated power compensation coefficient, so that the laser output power reaches the laser output power required at the current observation angle, thereby maintaining a constant excitation energy density received per unit area of ​​the ground target.

[0067] This invention compensates for the changes in the coverage area of ​​the laser projection strip caused by changes in flight altitude and tilt observation angle by calculating the power compensation coefficient based on the real-time acquired UAV flight altitude and gimbal tilt observation angle and adjusting the driving current of the linear laser 1 in real time. This maintains a constant excitation energy density received per unit area of ​​the ground target and ensures the comparability of fluorescence data collected at different observation angles and flight altitudes.

[0068] In one embodiment, since fluorescence images acquired from different observation angles exhibit perspective distortion, visual differences, and geometric distortion, this invention establishes a geometric registration scheme for multi-angle fluorescence images within a unified spatial reference frame. This scheme utilizes RTK-GNSS and IMU to acquire the UAV's position and attitude data, and performs direct geolocation for each scan line. The geometric registration of multi-angle fluorescence images includes four steps: direct geolocation based on collinearity equations, regional network adjustment optimization based on fluorescence control points, nonlinear collaborative registration of corresponding points at multiple angles, and construction of a unified spatial reference frame.

[0069] The step of geometrically registering the multi-angle fluorescence images based on the pose data and position data to obtain a geometrically aligned multi-angle fluorescence dataset includes: S1301. Based on the position data and the rotation matrix composed of the heading angle, pitch angle and roll angle in the attitude data, establish a projection mapping function between the fluorescence image pixel coordinates and the ground geographic coordinates at each observation angle.

[0070] In direct geolocation based on collinearity equations, RTK-GNSS and IMU are used to acquire the UAV's position and attitude data in real time. The position data consists of the coordinates of the projection center of the line-scan hyperspectral camera 2 in the object-space (geographic) coordinate system. Let the first... At time t, the coordinates of the projection center of the line-scan hyperspectral camera 2 in the object coordinate system are: Attitude data is heading angle Pitch angle ω and roll angle κ are used to construct a rotation matrix from attitude data. .

[0071] , formula (16); in, Rotation matrix Element.

[0072] For any ground target point Its true coordinates in the object coordinate system are Its observation angle Fluorescence images acquired below The coordinates of the corresponding image point on are According to the geometric relationship of the central projection, the image point coordinates and the object coordinates satisfy the collinearity equation, which is as follows: , formula (17); , formula (18); in, For the camera's equivalent focal length, Let these be the coordinates of the camera's principal point. By solving the collinearity equations simultaneously, the pixel coordinates of the fluorescence image at each observation angle are established. With ground geographic coordinates Projection mapping function between : , formula (19); in, This provides elevation values ​​for the average elevation surface or digital elevation model of the crop canopy. This projection mapping function establishes a mapping relationship from image pixel coordinates to ground geographic coordinates, providing an initial projection mapping function for subsequent regional network adjustment and optimization.

[0073] S1302. Based on the ground three-dimensional geographic coordinates of the fluorescence control points deployed in the target area and the actual observed pixel coordinates in the fluorescence image, a reprojection error loss function is constructed. The reprojection error loss function is minimized through an optimization algorithm to obtain the corrected UAV position coordinates and rotation matrix. The projection mapping function is updated based on the corrected UAV position coordinates and rotation matrix to obtain the updated projection mapping function.

[0074] In the adjustment optimization of the regional network based on fluorescence control points, the projection mapping function is refined using fluorescence control points deployed within the target region. Let the number of fluorescence control points deployed within the target region be... The ground three-dimensional geographic coordinates of each control point are as follows: .Will Substituting into the collinearity equations (17) and (18), the predicted image coordinates of the control points on the fluorescence image are calculated. The actual observed pixel coordinates of the control points in the fluorescence image are marked as follows: Construct the reprojection error loss function for: , formula (20); in, The sum of squared errors between the predicted image point coordinates and the actual observed pixel coordinates for all control points is used. The reprojection error loss function is minimized using the Gauss-Newton algorithm and the Levenberg-Marquardt optimization algorithm to obtain the corrected UAV position coordinates and rotation matrix, which are then used as the optimized camera's precise external projection position parameters. , formula (21); in, The corrected UAV position coordinates, This is the corrected rotation matrix.

[0075] The projection mapping function is updated using the corrected UAV position coordinates and rotation matrix, i.e., using... replace This yields an updated projection mapping function that eliminates system residuals caused by RTK-GNSS and IMU synchronization errors and camera mounting eccentricity.

[0076] S1303. Extract and match canopy feature points from fluorescence images at different observation angles to obtain multiple pairs of canopy feature point pairs. Solve the transformation matrix by minimizing the nonlinear distance residuals of all canopy feature point pairs.

[0077] In multi-angle nonlinear co-registration of corresponding points, for fluorescence images under different observation angles, such as observation angle... Fluorescence image below and observation angle Fluorescence image below Extract and match to obtain For canopy-named feature point pairs: , formula (22); in, It is a set of corresponding feature point pairs in the canopy. This represents the number of pairs of feature points with the same name. and These are the pixel coordinates of the corresponding feature points in the two fluorescence images. For the first At the observation angle, the same feature points are in the same position. The pixel coordinates in the fluorescence image, For the first At the observation angle, the same feature points are in the same position. The pixel coordinates in the fluorescence image. Establish the thin-plate spline function or local homography matrix transformation relationship: , formula (23); From the perspective of observation To the observation angle The transformation matrix is ​​used to transform Pixel coordinates in fluorescence images mapped to The corresponding location in the fluorescence image. The transformation matrix is ​​solved by minimizing the sum of the nonlinear distance residuals of all pairs of corresponding feature points. This transformation matrix is ​​used to achieve pixel-level precise alignment of local details between fluorescence images viewed from different angles. That is, during the solution process, the transformation matrix is ​​used to achieve pixel-level precise alignment of local details between fluorescence images viewed from different angles. Substituting into formula (23) yields the predicted coordinates, and the predicted coordinates are compared with the actual coordinates. The geometric distance between them is used as a nonlinear distance residual. By minimizing all The transformation matrix is ​​obtained by summing the nonlinear distance residuals of pairs of feature points with the same name. .

[0078] S1304. Using the updated projection mapping function and the transformation matrix, the fluorescence images at all observation angles are resampled to a common ground geographic grid coordinate system to obtain the geometrically aligned multi-angle fluorescence dataset.

[0079] In the construction of the unified spatial reference frame, the updated projection mapping function is utilized. and transformation matrix All observation angles The fluorescence images were resampled to a common ground geographic grid coordinate system. In, that is: , formula (24); in, For the first observation angle Fluorescent images were resampled to a public ground geographic grid coordinate system. The image after; This is the resampling operation function. This represents the transformation matrix. Ultimately, a multi-angle fluorescence dataset that is geometrically aligned is formed in the object space. The images in this multi-angle fluorescence dataset all have a completely consistent spatial coordinate reference and pixel resolution.

[0080] , formula (25).

[0081] This invention establishes a projection mapping function based on position data and a rotation matrix composed of heading, pitch, and roll angles from attitude data. This function maps the pixel coordinates of fluorescence images at various observation angles to ground geographic coordinates, providing an initial geometric mapping relationship for subsequent spatial alignment. By constructing and minimizing a reprojection error loss function based on the ground 3D geographic coordinates of fluorescence control points and their actual observed pixel coordinates in the fluorescence images, the corrected position coordinates and rotation matrix are obtained and used to replace the original parameters in the projection mapping function. This eliminates system residuals caused by RTK-GNSS and IMU synchronization errors and camera installation eccentricity, improving the accuracy of the projection mapping function. Furthermore, by extracting and matching canopy-named feature points from fluorescence images at different observation angles and minimizing the sum of nonlinear distance residuals for all pairs of feature points to solve the transformation matrix, the local parallax registration residuals caused by the 3D heterogeneity of the crop canopy are eliminated, achieving pixel-level precise alignment of local details between fluorescence images from different observation angles. By using projection mapping functions and transformation matrices to resample fluorescence images from all observation angles to a public ground geographic grid coordinate system, a multi-angle fluorescence dataset aligned in geometric space is obtained. Images from each angle have a completely consistent spatial coordinate reference and pixel resolution, providing high-precision geometric constraints for subsequent 3D reconstruction.

[0082] In one embodiment, the present invention proposes a three-dimensional fluorescence reconstruction scheme for crop canopies based on a fluorescence radiative transfer model. By constructing a forward radiative transfer model and applying regularized constraints to the inverse problem, a three-dimensional quantitative inversion of fluorescence emission intensity within the canopy is achieved. The step of obtaining an estimate of the three-dimensional phosphor spatial distribution based on the multi-angle fluorescence dataset and the forward fluorescence radiative transfer model includes: S1401. Discretize the three-dimensional object space of the crop canopy into a three-dimensional voxel grid composed of multiple voxels.

[0083] The three-dimensional object space of the crop canopy to be tested is discretized into a form by... Three-dimensional voxel network composed of individual voxels : , formula (26); in, For the first Individual factors, The total number of voxels in the three-dimensional voxel network.

[0084] Each voxel Internally includes leaf area density and chlorophyll content Feature parameter vector It is used to characterize crop structure and physiological state.

[0085] , formula (27); in, For the first The feature parameter vector within an individual element, Indicates the first Leaf area density within individual units Indicates the first The chlorophyll content within an individual.

[0086] For the observation angle and specific fluorescence emission bands The line-scan hyperspectral camera 2 received the fluorescence radiation intensity of the canopy surface. It satisfies the following one-dimensional continuous radiative transport integral equation (i.e., the above fluorescence radiative transport forward model): , formula (28); in, This indicates the transmission distance along the sensor's line of sight into the canopy. It represents the maximum boundary distance that the line of sight passes through the canopy, which is the total path length that the line of sight travels from the canopy surface to the point of exiting the canopy. This is the integration variable, used to iterate from 0 to... Each transmission distance between them. for The extinction coefficient of the canopy medium is defined as: , formula (29); In the formula, Let be the blade set projection function, representing the projection at the observation angle. The average projected area of ​​the lower blade in the direction perpendicular to the line of sight; For position Leaf area density at that location Leaf single-scatter albedo; Position along the line of sight path The term representing the spontaneous emission source of fluorescence at a given location indicates the fluorescence intensity spontaneously emitted by the crop leaf under excitation light, and is expressed as: , formula (30); In the formula, For position The actual illuminance of excitation light received by crop leaves. For this position The leaves at the location Fluorescence quantum yield in the band.

[0087] The above continuous radiative transport integral equation establishes a forward relationship between the fluorescence radiance of the canopy surface and the fluorescence emission inside the canopy, providing a forward model for subsequent discretization solutions.

[0088] S1402. The forward model of fluorescence radiative transfer is discretized and accumulated along the discretized three-dimensional voxel grid by ray tracing, so that the fluorescence radiative intensity of the canopy surface for each observation angle and each fluorescence band in the multi-angle fluorescence dataset is expressed in the form of a weighted sum of the fluorescence intensity of each voxel, and the discretized radiative transfer equation corresponding to each observation angle and each fluorescence band is obtained. The continuous radiative transfer integral equation is discretized and accumulated along a discrete three-dimensional voxel network using ray tracing, and then converted into an algebraic form: , formula (31); in, For the first Voxels in the fluorescence emission band The fluorescence emission intensity to be determined is as follows. For the first Individual elements from the perspective of observation and fluorescence emission band The weighting coefficients below, their magnitudes depend on the distance the observation line passes through the first... The geometric path length of the voxels and the cumulative exponential decay along the transmission path. This algebraic form transforms the continuous integral equation into a discrete weighted sum form, allowing the fluorescence radiance of the canopy surface to be directly expressed as a weighted sum of the fluorescence emission intensities of each voxel.

[0089] , formula (32); in, This represents the solution for estimating the spatial distribution of three-dimensional phosphors. Indicates matrix transpose. To reconcile the regularization parameters of observation residuals and prior constraints, Let be the spatial continuity Laplace regularization matrix. The observation vector is composed of the pixel-level radiance of the observed fluorescence after registration for each observation angle and each fluorescence band. To preset the number of observation angles, This represents the total number of pixels in a single fluorescence image. (Number of fluorescence bands) The system transfer matrix consists of weighting coefficients. for The transpose of .

[0090] S1403. Based on the discretized radiative transfer equations for all observation angles and all fluorescence bands, construct a linear matrix equation for multi-angle, multi-band joint observation. By integrating observation image data from all observation angles and all fluorescence bands, the linear matrix equation for multi-angle, multi-band joint observation is constructed as follows: , formula (33); in, The three-dimensional phosphor vector of the canopy to be reconstructed contains The fluorescence emission intensity of the voxel; This represents the random observation noise term.

[0091] S1404. Construct an objective function based on the linear matrix equation, and solve the objective function to obtain the estimated value of the three-dimensional phosphor spatial distribution.

[0092] Due to the system transmission matrix It exhibits high ill-conditioned behavior and introduces a spatially continuous Laplace regularization matrix. To constrain the spatial smoothness of fluorescence intensity between adjacent voxels, a least-squares objective function with a regularization penalty term is constructed. : , formula (34); in, To reconcile the regularization parameters between the observation residuals and prior constraints, the partial derivative of the objective function is taken and set to zero to obtain the optimal solution for estimating the spatial distribution of the three-dimensional phosphors. The solution for estimating the spatial distribution of the three-dimensional phosphor is the estimated value of the three-dimensional spatial distribution of fluorescence emission intensity of each voxel within the canopy.

[0093] This invention discretizes the three-dimensional object space of the crop canopy into a three-dimensional voxel grid composed of multiple voxels, transforming the radiative transfer problem in continuous space into a numerical solution problem on discrete voxels. This provides a spatial discretization basis for the subsequent inversion of the three-dimensional spatial distribution of fluorescence emission intensity within the canopy. By performing ray tracing discretization and accumulation along the discretized three-dimensional voxel grid using the forward model of fluorescence radiative transfer, the fluorescence radiance of the canopy surface for each observation angle and each fluorescence band is expressed as a weighted sum of the fluorescence intensities of each voxel. This transforms the continuous integral equation into a discrete algebraic equation, allowing the fluorescence radiance observed on the canopy surface to be decomposed into the sum of the contributions of the fluorescence emission intensities of each voxel. By constructing a linear matrix equation for multi-angle, multi-band joint observation based on the discretized radiative transfer equations for all observation angles and all fluorescence bands, all observation data are integrated into a unified mathematical framework. This transforms the inversion problem into a problem of solving matrix equations, providing a mathematical foundation for reconstructing the three-dimensional fluorescence distribution using multi-angle observation information. By constructing and solving a least-squares objective function with a regularization penalty term based on the linear matrix equation, and utilizing the spatial continuity Laplace regularization constraint on the spatial smoothness of fluorescence intensity between adjacent voxels during the solution process, the instability caused by the ill-conditioned nature of the system transfer matrix is ​​overcome, and a stable quantitative inversion of the three-dimensional spatial distribution of fluorescence emission intensity of each voxel within the canopy is achieved.

[0094] In one embodiment, the step of performing downward attenuation correction on the estimated spatial distribution of the three-dimensional phosphor based on the propagation path of the excitation light within the canopy to obtain three-dimensional phosphor data includes: S1501. For each voxel, the actual excitation illuminance received by the voxel is determined based on the reference excitation illuminance at the top of the canopy, the extinction coefficient of the upper voxels through which the excitation light passes before reaching the voxel, and the geometric distance through each upper voxel.

[0095] When an external active excitation light source penetrates deep into the canopy, it experiences downward attenuation due to light absorption and scattering by the leaves, resulting in lower actual excitation illuminance received by deeper voxels compared to surface voxels. Let the reference excitation illuminance at the top of the canopy be... A variant of Beer-Lambert's law is used to describe the downward propagation of excitation light along the incident direction within the canopy. , formula (35); in, For the first The actual excitation illuminance received by the individual. The reference excitation illuminance at the top of the canopy, Indicates that the excitation light reaches the first The upper voxel set that the voxel passes through before the voxel. For the first Individual elements at excitation wavelength The extinction coefficient at the bottom, To excite light to pass through the first Geometric distance of individual elements.

[0096] S1502. Based on the fluorescence intensity of the voxels in the estimated spatial distribution of the three-dimensional phosphor and the actual excitation illuminance received by the voxels, determine the corrected fluorescence intensity; the corrected fluorescence intensity of all voxels constitutes the three-dimensional phosphor data.

[0097] After determining the actual excitation illuminance received by each voxel, the solution is estimated using the three-dimensional phosphor spatial distribution obtained above. The downward excitation energy is corrected on a voxel-by-voxel basis using the following compensation formula: , formula (36); in, For the first The corrected fluorescence intensity of a voxel represents the true fluorescence emission intensity of that voxel after eliminating the effect of downward attenuation. The third value in the three-dimensional phosphor spatial distribution estimate Fluorescence intensity of the individual voxels.

[0098] The fluorescence intensity after voxel correction constitutes the three-dimensional phosphor data. .

[0099] , formula (37); in, They are respectively number 1 to number 2. Corrected fluorescence intensity of the individual voxels.

[0100] This invention quantifies the attenuation of excitation light during its propagation within the canopy by determining the actual excitation illuminance received by each voxel based on the reference excitation illuminance at the top of the canopy, the extinction coefficient of the upper voxels through which the excitation light passes before reaching each voxel, and the geometric distance traveled through each upper voxel. This provides a quantitative basis for subsequent voxel-by-voxel compensation correction. The corrected fluorescence intensity is determined based on the fluorescence intensity of each voxel in the estimated spatial distribution of the three-dimensional phosphor and the actual excitation illuminance received by each voxel. The corrected fluorescence intensity of all voxels constitutes the three-dimensional phosphor data, eliminating inversion artifacts caused by downward attenuation due to leaf absorption and scattering as the excitation light penetrates deeper into the canopy. It also eliminates surface shading caused by the vegetation canopy geometry, allowing for accurate characterization of the fluorescence emission intensity of deep voxels. This enables precise characterization of deep physiological and biochemical heterogeneity changes within the crop canopy caused by diseases.

[0101] In one embodiment, the absolute three-dimensional fluorescence data obtained after radiative transfer inversion and downlink excitation light attenuation correction is represented as a high-dimensional matrix: , formula (38); Where M is the total number of voxels in the three-dimensional voxel network. The selected characteristic fluorescence emission band. For the first Voxels in the fluorescence emission band The corrected fluorescence intensity.

[0102] Based on the three-dimensional phosphor data The three-dimensional fluorescence feature parameters constructed and extracted by the present invention include at least one of the following: canopy fluorescence vertical distribution index, total three-dimensional fluorescence, fluorescence three-dimensional spatial heterogeneity index, canopy deep fluorescence ratio, and canopy fluorescence structure centroid index. The canopy fluorescence vertical distribution index includes the average fluorescence intensity of each spatial level. The average fluorescence intensity of each spatial level is obtained by dividing the crop canopy into multiple spatial levels at equal intervals along the vertical height direction and averaging the corrected fluorescence intensity of voxels belonging to the same spatial level. Specifically, the crop canopy is divided into equal intervals along the vertical direction. The spatial hierarchy, denoted as the first ( =1,2,.., The voxel set of the layer is The number of voxels contained in this layer is Then the first Layer-level average fluorescence intensity Defined as: , formula (39); in, For the first Corrected fluorescence intensity of voxels at 740 nm; The average fluorescence intensity across all spatial levels constitutes the characteristic vector of the vertical fluorescence distribution of the entire crop canopy. : , formula (40); The total three-dimensional fluorescence intensity is determined based on the corrected fluorescence intensity of each voxel in the far-infrared band and the physical volume of each voxel. To characterize the absolute quantitative relationship between crop biomass and overall photosynthetic activity, the total three-dimensional fluorescence emission within the canopy of the entire plant or a region is defined. satisfy: , formula (41); in, The physical volume of a single voxel.

[0103] The fluorescence three-dimensional spatial heterogeneity index is determined based on the ratio of the standard deviation of the corrected fluorescence intensity of each voxel in the far-red band to the global average corrected fluorescence intensity. Specifically, the global average fluorescence intensity of voxels in the entire crop canopy is first calculated. : , formula (42); The fluorescence three-dimensional spatial heterogeneity index Defined as the ratio of the standard deviation of fluorescence intensity of all voxels to the global mean, it is essentially the spatial variation coefficient (CV) of fluorescence data. , formula (43); This index is used to quantitatively characterize the degree of spatial disorder of physiological activity within the canopy caused by local infection of the disease.

[0104] Because the lower (deeper) part of the crop canopy is more prone to the growth of hidden diseases, let the canopy height be... Define the distance from the ground as less than or equal to a preset threshold (preferably). The deep voxel set is The canopy-deep fluorescence ratio is determined based on the ratio of the corrected fluorescence intensity of deep voxels in the red band to the corrected fluorescence intensity in the far-red band. The deep voxels are voxels located within a spatial region at a height less than or equal to a preset threshold from the ground. To eliminate the shielding effect of the canopy structure on photofluorescence, this invention extracts the ratio of deep voxels at 690 nm (red peak) to 740 nm (far-red peak) as the canopy-deep fluorescence ratio. : , formula (44); in, For the first Corrected fluorescence intensity of the voxel at 690 nm.

[0105] To quantitatively characterize the change of the centroid of fluorescence activity with depth within the canopy, a height of the centroid of fluorescence spatial distribution with a clear physical meaning is defined. The centroid index of the canopy fluorescence structure is used. This centroid index is determined based on the corrected fluorescence intensity of each voxel in the far-infrared band and the absolute height coordinates of each voxel.

[0106] , formula (45); in, For the first The absolute altitude coordinates of an individual element in a geographic coordinate system.

[0107] This invention divides the crop canopy into multiple spatial levels at equal intervals along its vertical direction and averages the corrected fluorescence intensity of voxels belonging to the same spatial level to obtain the canopy fluorescence vertical distribution index. This index reflects the distribution characteristics of fluorescence intensity in the vertical direction of the canopy and provides a quantitative indicator for assessing the differences in physiological activity at different canopy heights. By determining the total three-dimensional fluorescence based on the corrected fluorescence intensity of each voxel in the far-infrared band and the physical volume of each voxel, the absolute quantitative relationship between crop biomass and overall photosynthetic activity is quantified, providing a total indicator for assessing overall growth status. Furthermore, by determining the ratio of the standard deviation of the corrected fluorescence intensity of each voxel in the far-infrared band to the global average corrected fluorescence intensity, the fluorescence three-dimensional spatial heterogeneity index is determined, quantitatively characterizing the degree of spatial disorder of physiological activity within the canopy caused by localized disease infection. The canopy deep fluorescence ratio was determined by comparing the corrected fluorescence intensity of deep voxels in the red band with their corrected fluorescence intensity in the far-infrared band. This eliminated the shielding effect of the canopy structure on photofluorescence, enabling the characterization of the physiological state of the deep regions beneath the canopy and providing a quantitative indicator for detecting hidden diseases in the population. Furthermore, the canopy fluorescence structure centroid index was determined by comparing the corrected fluorescence intensity of each voxel in the far-infrared band with its absolute height coordinates. This quantitatively characterized the change of the fluorescence activity centroid within the canopy with depth, providing a quantitative indicator for assessing the spatial migration of the fluorescence activity centroid within the canopy.

[0108] In one embodiment, the crop growth status includes a crop growth status level. Determining the crop growth status based on the three-dimensional fluorescence feature parameters includes: S1601. Normalize the three-dimensional fluorescence feature parameters; To eliminate the influence of the dimensions of each feature parameter, the extracted feature parameters are first normalized using the range transformation method: , formula (46); in, These are the normalized feature parameter values. These are the original feature parameter values. , These are the minimum and maximum values ​​of the feature parameter, respectively. Wheat samples with different disease levels can be collected beforehand to form a training set. For each three-dimensional fluorescence feature parameter, the minimum and maximum values ​​of that parameter for all samples in the training set are calculated and denoted as [the minimum and maximum values ​​are then listed]. , .

[0109] S1602, Based on the normalized three-dimensional fluorescence total amount Normalized fluorescence three-dimensional spatial heterogeneity index Normalized canopy depth fluorescence ratio and normalized canopy fluorescence structure centroid index To determine the three-dimensional fluorescence growth index; By integrating multidimensional fluorescence characteristic parameters, a three-dimensional fluorescence growth index (3D-FGI) was constructed for comprehensively evaluating crop growth and the degree of damage caused by disasters. , formula (47); in, , , , These are the weight coefficients of the corresponding indicators, and they satisfy... The weighting coefficients are determined by training a pre-collected training set of wheat samples with different disease severities using multiple linear regression or support vector machine (SVM).

[0110] S1603, Based on the three-dimensional fluorescence growth index ratio of fluorescence to deep canopy A growth status evaluation vector is formed, and the Euclidean distance between the growth status evaluation vector and the cluster center vectors corresponding to multiple preset growth status levels is calculated. The crop growth status level is determined according to the minimum distance principle.

[0111] This invention constructs a multidimensional evaluation vector for growth, projecting the crop under test onto a multidimensional evaluation feature space composed of comprehensive indices and characteristic indicators, thereby obtaining a growth status evaluation vector. : , formula (48); Suppose the system pre-sets or obtains a prior space set of cluster center vectors corresponding to the four crop growth status levels (excellent, good, average, poor) through K-means clustering training: , formula (49); in, For the first Each growth state level corresponds to a priori space cluster center vector. For the first The coordinates of the cluster centers corresponding to each growth state level in the three-dimensional fluorescence growth index dimension. For the first The coordinates of the cluster centers corresponding to each growth state level on the dimension of the fluorescence ratio in the deep canopy.

[0112] Calculate the current growth status evaluation vector of the crop to be tested. Cluster center vectors of each prior space Euclidean distance between : , formula (50); Finally, the actual growth status level of the crop under test is determined based on the principle of minimum distance. : , formula (51); In the formula, the output is The value directly corresponds to the actual growth status of the crop, that is: the corresponding crop growth status level: , formula (52); This invention eliminates the influence of dimensions on three-dimensional fluorescence feature parameters by normalizing them, enabling comparison and fusion of feature parameters with different dimensions on the same scale, thus providing standardized input for constructing a comprehensive evaluation index. A three-dimensional fluorescence growth index is determined based on the normalized total three-dimensional fluorescence, the normalized three-dimensional fluorescence spatial heterogeneity index, the normalized canopy depth fluorescence ratio, and the normalized canopy fluorescence structure centroid index. This integrates spatial fluorescence features from multiple dimensions, constructing a comprehensive index for evaluating crop growth and damage caused by disasters. A growth status evaluation vector is formed based on the three-dimensional fluorescence growth index and the canopy depth fluorescence ratio. The Euclidean distance between this growth status evaluation vector and the cluster center vectors corresponding to multiple preset growth status levels is calculated. The crop growth status level is determined according to the minimum distance principle, achieving objective quantitative discrimination of the growth status of the crop under test. The crop under test is projected into a multi-dimensional evaluation feature space composed of the comprehensive index and feature indicators, and the growth status level is automatically determined through distance measurement.

[0113] Let the final absolute three-dimensional fluorescence data obtained after the aforementioned three-dimensional reconstruction and excitation light attenuation correction be represented as a discrete set of voxels. : , formula (53); in, For the first Individual factors The corrected fluorescence intensity, with its spatial geometric center coordinates as follows: . , , and z, respectively, are the x-axis, y-axis, and z-axis coordinates of the i-th voxel in the object space.

[0114] To transform abstract, high-dimensional discrete prime data into intuitive and readable crop physiological distribution maps, this invention designs two synergistic methods for three-dimensional representation and internal analysis visualization: The first method is based on MarchingCubes for equal-face value rendering and display: To visually demonstrate the macroscopic spatial continuity distribution characteristics of fluorescence intensity within the crop canopy, we first use discrete voxel sets... A continuous spatial fluorescence field function was constructed using a trilinear interpolation algorithm. : , formula (54); in, is a continuous spatial fluorescence field function, representing the fluorescence intensity value at any position (x,y,z) in the object space; It is a trilinear interpolation function.

[0115] Suppose that a preset fluorescence intensity threshold needs to be given according to the disease symptom. The corresponding fluorescence isosurface In object space, it is defined as the following implicit surface: , formula (55); in, It is a three-dimensional real number space.

[0116] The implicit surface defined by equation (55) is solved by voxel-level topological discretization using the MarchingCubes algorithm to extract the three-dimensional geometric model of fluorescent isosurfaces represented by triangular meshes. : , formula (56); in, It is a discrete set of vertex coordinates of a triangular mesh. For the set of topological edges, This is a set of triangular face indices, representing a face consisting of three vertices.

[0117] To enhance visual recognition and reflect disease gradient changes, this invention utilizes a pseudo-color mapping function. Color space mapping is performed on the geometric model. Let the global minimum and maximum values ​​of the entire phosphor data be respectively... and Then the vertex Mapped pseudo-color values satisfy: , formula (57); in, As vertices Spatial fluorescence field function value at; By the pseudo color value Rendered onto the corresponding mesh surface, ultimately generating a 3D fluorescence isosurface rendering model of the crop canopy. : , formula (58); The model is specifically designed to highlight the three-dimensional envelope morphology of regions with high (or low) fluorescence abnormal activity, such as wheat powdery mildew mycelial infection foci.

[0118] The second method involves drawing and analyzing slices of volume in any spatial direction: To deeply analyze the distribution of hidden fluorescence information in unexposed areas within the crop canopy, this invention designs an interactive scheme for plot rendering. A virtual plot plane with arbitrary orientation is defined in three-dimensional space. : , formula (59); in: And satisfy =1, which is the unit normal vector of the slice plane; for The three components; The intercept parameter is used to determine the plane offset distance; Let be the coordinates of any three-dimensional point on the plane. For the center coordinates of any voxel in the three-dimensional phosphor data... to the virtual slice plane absolute geometric distance Represented as: , formula (60); Set the sampling thickness parameter of the slice plane as follows When the distance between voxels satisfies the following constraint: , formula (61); The selected voxel is then determined to belong to the spatial region covered by the current slice. These selected voxels are then resampled and projected onto the slice's two-dimensional local coordinate system. In the middle, a two-dimensional distribution map of fluorescence intensity in the slice plane is formed. : , formula (62); Furthermore, by using pseudo-color lookup table mapping, a two-dimensional fine cross-sectional fluorescence distribution image containing rich gray levels or color gradients is obtained. : , formula (63); In practical engineering applications and system operation: (1) When the unit normal vector of the plane Take standard substrates respectively , and At that time, the system renders and outputs standard slice images of the crop canopy in three orthogonal directions: YZ (canopy cross section), XZ (canopy longitudinal section), and XY (canopy horizontal section). (2) When the unit normal vector Satisfying the following conditions for any azimuth angle and pitch angle When the arbitrary space is constructed: , formula (64); Using formulas (62) and (63), fluorescence distribution maps of spatial oblique sections in any direction are obtained through real-time resampling. This method eliminates the technical limitation of traditional two-dimensional remote sensing, which can only observe the surface layer, and realizes multi-angle, non-destructive three-dimensional analysis of the physiological status of the lower leaves of the lush crop canopy.

[0119] This invention constructs a continuous spatial fluorescence field function based on a discrete voxel set using a trilinear interpolation algorithm, transforming discrete voxel data into a continuous spatial fluorescence distribution field. This provides a continuously differentiable scalar field basis for isosurface extraction. By defining fluorescence isosurfaces according to a preset fluorescence intensity threshold and using the MarchingCubes algorithm to extract a three-dimensional geometric model of the fluorescence isosurfaces represented by triangular meshes, abstract fluorescence intensity data is transformed into an intuitive three-dimensional geometric surface. This allows the macroscopic spatial continuity distribution characteristics of fluorescence intensity within the canopy to be visually displayed in the form of a three-dimensional curved surface, which is used to highlight the three-dimensional envelope morphology of regions with high or low fluorescence anomalous activity.

[0120] By performing pseudo-color mapping on the vertices of triangular meshes and rendering the pseudo-color values ​​onto the corresponding mesh surfaces, a three-dimensional fluorescence isosurface rendering model is generated. This maps fluorescence intensity values ​​to intuitive color changes, enhancing visual recognition and reflecting disease gradient changes. By defining virtual slice planes in arbitrary directions in three-dimensional space and calculating the absolute geometric distance of each voxel to this plane, combined with sampling thickness parameters, voxels belonging to the spatial region covered by the slice are determined. This allows users to cut and analyze three-dimensional fluorescence data from any direction, breaking through the limitation of traditional two-dimensional remote sensing that can only observe the surface layer.

[0121] By resampling and projecting the selected voxels onto the two-dimensional local coordinate system of the slice to form a two-dimensional distribution map of fluorescence intensity on the slice plane, and then obtaining a two-dimensional fine cross-sectional fluorescence distribution image through pseudo-color lookup table mapping, the fluorescence information of any direction section in the three-dimensional volume data is transformed into a two-dimensional cross-sectional image. This enables an intuitive display of the distribution of hidden fluorescence information in non-exposed areas inside the crop canopy, and allows for multi-angle, non-destructive three-dimensional analysis of the physiological status of the lower leaves of the lush canopy.

[0122] To facilitate understanding of the technical solution of the present invention, the working process of the present invention will be described in detail.

[0123] Step 1: Plan the flight mission at the ground station, set multi-angle flight path parameters for the target area, including the set of observation angles, flight path spacing, flight altitude and flight speed, and upload the flight path file to the UAV flight control system.

[0124] Step 2: Deploy fluorescence control points within the target area and obtain the accurate three-dimensional geographic coordinates of each fluorescence control point for subsequent regional network adjustment and refinement.

[0125] Step 3: On the ground, check the relative geometric matching between the deflection angle of the projected beam of the airborne linear laser 1 and the field of view of the line-scan hyperspectral camera 2 to ensure that the linear laser 1 can accurately cover the instantaneous scan line of the camera at each tilt observation angle, and establish the spatial overlap between the excitation area and the imaging area.

[0126] Step four: Start the drone and control its flight according to the flight path file uploaded in step one. During each flight segment, the inertial stabilizing gimbal 4 switches sequentially according to a preset sequence of observation angles. At each observation angle, under the spatial overlap condition established in step three, multi-angle fluorescence images of the crop canopy are acquired using the line-scan hyperspectral camera 2. Attitude and position data corresponding to each observation angle are simultaneously acquired via the inertial measurement unit and RTK-GPS.

[0127] Step 5: Perform cross-validation flight path data acquisition. After completing the main flight path repeat flight mission, the UAV automatically adjusts its course and enters the cross-validation flight path at a preset angle to the main flight path for additional data acquisition, supplementing the acquisition with cross-validation fluorescence images and corresponding attitude and position data. The cross-validation data and the main flight path data obtained in Step 4 form a multi-baseline, multi-angle stereoscopic observation geometric relationship, which is used to eliminate redundancy in multi-period data and improve the accuracy of three-dimensional structure reconstruction.

[0128] Step six: During flight, the central control processing module 5 performs real-time streaming preprocessing on the raw fluorescence images acquired in steps four and five. This includes inverse environmental spectral correction based on synchronously acquired ambient light intensity, radiometric calibration, and coarse geometric correction based on the original position and attitude data, resulting in preprocessed fluorescence images. The preprocessed low-resolution thumbnails and flight path are transmitted in real-time to the ground station via the wireless communication module 7 for ground technicians to monitor growth and equipment status in real time.

[0129] Step 7: After completing all flight routes, export all data (including multi-angle fluorescence images, attitude data, position data, and preprocessed fluorescence images) from data storage module 6 to the computer.

[0130] Step 8: Discretize the canopy space of the crop to be tested into a three-dimensional voxel grid. Perform geometric registration based on the multi-angle fluorescence images and attitude position data exported in Step 7 to obtain a geometrically aligned multi-angle fluorescence dataset. Perform regularized inversion based on the multi-angle fluorescence dataset and the forward fluorescence radiative transfer model to obtain the estimated value of the three-dimensional phosphor spatial distribution.

[0131] Step nine: Based on the propagation path of the excitation light within the canopy, the estimated three-dimensional phosphor spatial distribution obtained in step eight is subjected to downward attenuation correction to achieve bidirectional optical path decoupling and excitation light downward attenuation correction, thereby obtaining three-dimensional phosphor data.

[0132] Step 10: Extract five core spatial physiological characteristic parameters from the three-dimensional phosphor data obtained in Step 9, including the vertical distribution index of canopy fluorescence, total three-dimensional fluorescence, three-dimensional spatial heterogeneity index of fluorescence, ratio of deep canopy fluorescence, and centroid index of canopy fluorescence structure. After normalizing each index, construct a three-dimensional fluorescence growth index using the normalized characteristic parameters, and determine the crop growth status level based on the three-dimensional fluorescence growth index and the ratio of deep canopy fluorescence.

[0133] Step 11: Based on the three-dimensional phosphor data obtained in Step 9, perform multi-mode stereoscopic visualization and interactive analysis. Select the isoplanar rendering mode to generate a three-dimensional phosphor isoplanar rendering model, or select the slice volume drawing mode to generate a cross-sectional fluorescence distribution image in any direction, so as to realize the visualization display of the three-dimensional phosphor data.

[0134] Step 12 involves organizing, storing, and backing up the flight path files, fluorescence control point data, multi-angle fluorescence images, attitude data, position data, preprocessed fluorescence images, geometrically aligned multi-angle fluorescence datasets, three-dimensional phosphor spatial distribution estimates, three-dimensional phosphor data, three-dimensional fluorescence isoplanar rendering models and cross-sectional fluorescence distribution images, as well as the crop growth status levels generated in steps one through eleven, thus completing the data archiving.

[0135] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting crop growth status based on three-dimensional fluorescence, characterized in that, include: Adjust the relative geometric position between the linear laser on the drone and the line-scan hyperspectral camera so that the excitation area of ​​the linear laser on the crop canopy surface coincides with the imaging area of ​​the line-scan hyperspectral camera in space. The drone is controlled to fly according to a preset multi-angle flight path planning strategy. When the excitation area and the imaging area are spatially overlapped, the line-scan hyperspectral camera is used to acquire multi-angle fluorescence images of the crop canopy at each preset observation angle, and the attitude data and position data of the drone corresponding to each observation angle are acquired simultaneously. Geometric registration is performed on the multi-angle fluorescence images based on the pose data and position data to obtain a multi-angle fluorescence dataset aligned in geometric space. Based on the multi-angle fluorescence dataset and the forward fluorescence radiative transfer model, the estimated value of the three-dimensional phosphor spatial distribution is obtained by inversion solution; The estimated spatial distribution of the three-dimensional phosphors is corrected for downward attenuation based on the propagation path of the excitation light within the canopy to obtain the three-dimensional phosphor data. Three-dimensional fluorescence characteristic parameters characterizing crop growth status are extracted from the three-dimensional fluorescence data, and crop growth status is determined based on the three-dimensional fluorescence characteristic parameters.

2. The crop growth status detection method based on three-dimensional fluorescence according to claim 1, characterized in that, Adjusting the relative geometric position between the linear laser mounted on the UAV and the line-scan hyperspectral camera, so that the excitation area of ​​the linear laser on the crop canopy surface spatially coincides with the imaging area of ​​the line-scan hyperspectral camera, includes: Adjust the relative geometric position between the linear laser and the line-scan hyperspectral camera so that the projection linewidth of the linear laser on the ground is equal to the coverage linewidth of the instantaneous field of view of the line-scan hyperspectral camera on the ground.

3. The crop growth status detection method based on three-dimensional fluorescence according to claim 1, characterized in that, The multi-angle route planning strategy includes: Set a set of observation angles that includes the nadir observation angle and multiple tilt observation angles; Multiple main flight routes are planned in the target area. The spacing between adjacent main flight routes is determined based on the preset lateral overlap rate of the flight routes and the effective ground observation swath width under the maximum tilt observation angle. Multiple repeated flights are performed along each main flight route, with each flight using a different observation angle from the set of observation angles. Additional data collection is performed by adding cross-validation routes that form a preset angle with the main route.

4. The crop growth status detection method based on three-dimensional fluorescence according to claim 1, characterized in that, The step of geometrically registering the multi-angle fluorescence images based on the pose data and position data to obtain a geometrically aligned multi-angle fluorescence dataset includes: Based on the location data and the rotation matrix composed of the heading angle, pitch angle and roll angle in the attitude data, a projection mapping function between the fluorescence image pixel coordinates and the ground geographic coordinates at each observation angle is established. Based on the ground three-dimensional geographic coordinates of the fluorescence control points deployed in the target area and the actual observed pixel coordinates in the fluorescence image, a reprojection error loss function is constructed. The reprojection error loss function is minimized by an optimization algorithm to obtain the corrected UAV position coordinates and rotation matrix. The updated projection mapping function is obtained by updating the projection mapping function based on the corrected UAV position coordinates and rotation matrix; Extraction and matching of canopy-named feature points from fluorescence images under different observation angles yielded multiple pairs of canopy-named feature point pairs. The transformation matrix was solved by minimizing the nonlinear distance residuals of all canopy-named feature point pairs. The fluorescence images at all observation angles are resampled to a common ground geographic grid coordinate system using the updated projection mapping function and the transformation matrix to obtain the geometrically aligned multi-angle fluorescence dataset.

5. The crop growth status detection method based on three-dimensional fluorescence according to claim 1, characterized in that, The step of obtaining the estimated three-dimensional phosphor spatial distribution based on the multi-angle fluorescence dataset and the forward fluorescence radiative transport model includes: The three-dimensional object space of the crop canopy is discretized into a three-dimensional voxel grid composed of multiple voxels; The fluorescence radiative transport forward model is discretized and accumulated along a discretized three-dimensional voxel grid by ray tracing, so that the canopy surface fluorescence radiance for each observation angle and each fluorescence band in the multi-angle fluorescence dataset is expressed as a weighted sum of the fluorescence intensities of each voxel, thus obtaining the discretized radiative transport equation corresponding to each observation angle and each fluorescence band. Based on the discretized radiative transfer equations for all observation angles and all fluorescence bands, a linear matrix equation for multi-angle, multi-band joint observation is constructed. Construct an objective function based on the linear matrix equation, and solve the objective function to obtain an estimate of the spatial distribution of the three-dimensional phosphor.

6. The crop growth status detection method based on three-dimensional fluorescence according to claim 1, characterized in that, The step of performing downward attenuation correction on the estimated spatial distribution of the three-dimensional phosphors based on the propagation path of the excitation light within the canopy to obtain three-dimensional phosphor data includes: For each voxel, the actual excitation illuminance received by the voxel is determined based on the reference excitation illuminance at the top of the canopy, the extinction coefficient of the upper voxels through which the excitation light passes before reaching the voxel, and the geometric distance through each upper voxel. The corrected fluorescence intensity is determined based on the fluorescence intensity of the voxels in the estimated spatial distribution of the three-dimensional phosphor and the actual excitation illuminance received by the voxels; the corrected fluorescence intensities of all voxels constitute the three-dimensional phosphor data.

7. The crop growth status detection method based on three-dimensional fluorescence according to claim 6, characterized in that, The three-dimensional fluorescence characteristic parameters include at least one of the following: canopy fluorescence vertical distribution index, total three-dimensional fluorescence, fluorescence three-dimensional spatial heterogeneity index, canopy deep fluorescence ratio, and canopy fluorescence structure centroid index; The canopy fluorescence vertical distribution index includes the average fluorescence intensity of each spatial level. The average fluorescence intensity of each spatial level is obtained by dividing the crop canopy into multiple spatial levels at equal intervals along the vertical height direction and averaging the corrected fluorescence intensity of voxels belonging to the same spatial level. The total amount of three-dimensional fluorescence is determined based on the corrected fluorescence intensity of each voxel in the far-infrared band and the physical volume of each voxel. The fluorescence three-dimensional spatial heterogeneity index is determined based on the ratio of the standard deviation of the corrected fluorescence intensity of each voxel in the far-red band to the global average corrected fluorescence intensity. The canopy deep fluorescence ratio is determined based on the ratio of the corrected fluorescence intensity of deep voxels in the red band to the corrected fluorescence intensity in the far-red band. The deep voxels are voxels in a spatial region at a height less than or equal to a preset threshold from the ground. The centroid index of the canopy fluorescence structure is determined based on the corrected fluorescence intensity of each voxel in the far-infrared band and the absolute height coordinates of each voxel.

8. The crop growth status detection method based on three-dimensional fluorescence according to claim 7, characterized in that, The crop growth status includes crop growth status levels. Determining crop growth status based on the three-dimensional fluorescence feature parameters includes: The three-dimensional fluorescence feature parameters are normalized. The three-dimensional fluorescence growth index was determined based on the normalized total three-dimensional fluorescence, the normalized three-dimensional fluorescence spatial heterogeneity index, the normalized canopy deep fluorescence ratio, and the normalized canopy fluorescence structure centroid index. The growth status evaluation vector is composed of the three-dimensional fluorescence growth index and the ratio of fluorescence in the canopy depth. The Euclidean distance between the growth status evaluation vector and the cluster center vectors corresponding to multiple preset growth status levels is calculated. The crop growth status level is determined according to the minimum distance principle.

9. The crop growth status detection method based on three-dimensional fluorescence according to claim 1, characterized in that, The method further includes the following steps during the acquisition of multi-angle fluorescence images: If the UAV changes its flight altitude or the gimbal tilt angle changes, the UAV's flight altitude and the inertial stabilized gimbal tilt angle are acquired in real time, and the driving current of the linear laser is adjusted according to the flight altitude, the tilt angle, and the reference flight altitude.

10. A crop growth status detection device based on three-dimensional fluorescence, characterized in that, include: A linear laser emits excitation light, which is shaped into a linear beam by a laser beam rectifier and projected onto the crop canopy. The linear beam forms an excitation region on the surface of the crop canopy. A line-scan hyperspectral camera, the line-scan hyperspectral camera having an imaging area, the line-scan hyperspectral camera being used to acquire multi-angle fluorescence images of the crop canopy generated under the excitation light; An attitude measurement module is used to acquire the attitude and position data of the UAV in real time. An inertial stabilizing gimbal is used to support the linear laser and the line-scanning hyperspectral camera, and supports adjusting the observation angle within a preset angle range. A central control processing module is electrically connected to the linear laser, the line-scanning hyperspectral camera, and the attitude measurement module. The central control processing module is used to control the timing synchronization of the linear laser and the line-scanning hyperspectral camera, and to execute the crop growth status detection method based on three-dimensional fluorescence as described in any one of claims 1 to 9. A data storage module, electrically connected to the central control and processing module, is used to store the multi-angle fluorescence image, the attitude data, the position data, and the three-dimensional phosphor data; A wireless communication module, which is electrically connected to the central control processing module, is used to realize data transmission between the UAV and the ground station; The power management module provides power to the linear laser, the line-scan hyperspectral camera, the attitude measurement module, the central control processing module, the data storage module, the wireless communication module, and the inertial stabilization gimbal.

Images