Modeling method for multi-scale fruit environment characteristic parameterized classification quantitative detection

Abstract

The invention relates to a modeling method for multi-scale fruit environment characteristic parameterization classification quantitative detection, which comprises the following steps: S10, obtaining illumination and temperature space distribution characteristics through high-precision satellite remote sensing image data, carrying out gridding processing on environment characteristics of a to-be-detected area, and obtaining a data matrix according to a solar radiation model and a surface temperature inversion model; classifying and partitioning the temperature and illumination features of the gridded orchard area through ground experiment control points; s20, establishing a multi-angle profiling calibration field through the temperature and illumination characteristics of the gridding orchard area; s30, accurate angle calibration is carried out, fruit reflectivity characteristic angles are selected, and a multi-angle polarization reflectivity database is established; and S40, acquiring a reflectivity change rule of fruits in the same region at different angles and polarization directions by using an unmanned aerial vehicle multispectral camera, extracting key parameters with obvious environment temperature and illumination characteristics, and fitting a region scale spectral reflectivity curve model and a typical environment multi-angle polarization spectral response function. The fruit partition quantitative detection precision is improved.

Description

Technical Field

[0001] The present invention specifically relates to a modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters. Background Art

[0002] There are certain differences in temperature difference and light conditions in different regions. Even for fruits on the same fruit tree, there are significant differences in the quality of different growth parts. In particular, the quality of fruits such as sugar content is greatly affected by environmental factors such as outdoor light and day-night temperature difference, and has certain regional distribution characteristics, which is also a sign of the environmental adaptability of fruit quality. How to quantitatively characterize the influence of environmental factors on the spatial spectrum of far-field fruits, quickly and accurately obtain the spectral components of fruits at different scales, and study the spatial distribution characteristics and variation laws of fruit quality affected by regional environment is of great significance for the adjustment of the fruit industry structure and the realization of high-quality and efficient development.

[0003] To solve this problem, the prior art has used unmanned aerial vehicle multi-spectral cameras for detection. However, due to the low spectral resolution of conventional fruit detection spectral models, the influence of environmental factors cannot be effectively distinguished. There are disadvantages such as low detection accuracy, large amount of data, and high cost.

[0004] In view of this, the present invention proposes a new modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters. A spectrometer with relatively low spectral resolution can be used, and a high detection accuracy can be obtained through the multi-angle polarization method, and regional evaluation can be realized without image registration. Summary of the Invention

[0005] The purpose of the present invention is to provide a modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters. Through the multi-spectral multi-angle polarization method of unmanned aerial vehicles, study the change law of the reflectance of the same fruit in different angle polarization directions, and fit the reflectance curve model to quantitatively detect the accuracy of fruit quality. Then, a polarizer is added at the detection end to change the direction for calibration at different angles of polarization. Through the method of combining environmental and spectral quantification, the calibration of a high-resolution spectrometer and the spatial calibration of a low-resolution spectral camera are realized, and the quantitative detection accuracy of spectral imaging is improved. The accuracy of regional quantitative detection of fruits is improved.

[0006] In order to achieve the above object, the technical solution adopted is as follows:

[0007] A modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters includes the following steps:

[0008] S10: Obtain the spatial distribution characteristics of light and temperature through high-precision satellite remote sensing image data, perform grid processing on the temperature and light of the area to be measured, and classify and mark the temperature and light characteristics of the grid-like orchard area based on the solar radiation model and the land surface temperature inversion model through ground experimental control points;

[0009] S20: By gridding the temperature and light characteristics of the orchard area, a multi-angle profiling calibration field with spatial coordinate information is established, and multi-scale and multi-angle quantitative detection is achieved through the transfer and interactive verification of the multi-angle reflectivity quantitative detection database and the spectral database quantitative detection model;

[0010] S30: grid classification and precise angle calibration of the spatial environment characteristics to be measured, selection of characteristic angles of fruit reflectance, and establishment of a multi-angle polarization reflectance database;

[0011] S40: Use the multispectral camera of the UAV to obtain the reflectivity variation pattern of fruits in the same area at different angles and polarization directions, obtain the fruit growth temperature and light characteristic model through the zoning classification and marking experiment, extract the key parameters with obvious environmental characteristics through the targeted reinforcement training strategy, fit the regional scale spectral reflectance curve model and the multi-angle polarization spectral response function of typical environmental characteristics, and combine spectral quantitative detection with UAV spectral image regional evaluation.

[0012] Furthermore, the modeling method first grids the characteristics of the space environment to be tested, performs angle simulation and calibration modeling indoors, and then performs verification modeling outdoors.

[0013] The spectral response function for quantitative indoor angle is as follows:

[0014] 1. Light source, fruit, detection orientation and other factors will affect the spectral characteristics of non-in-situ fruit. The spectral morphological characteristics of fruit from multiple angles and the variation law of polarized reflectance spectrum of fruit at multiple scales are studied to reveal the spectral spatial response function mechanism of non-in-situ fruit components.

[0015] 2. Quantitative description of light source, spectral coherence, illumination angle (azimuth and altitude), wavelength, and wave width. Different fruits are equivalent to scattering media with different refractive indices and particles. The horizontal and vertical spatial distribution characteristics of different structures are described by high-order optical parameters. Less simulates the reflectivity characteristics of different parameters. Through statistical correlation modeling of spectral morphological parameters at different angles, peak height, slit width, peak shape, peak area and angle, the spectral response functions of different band angles are obtained through targeted reinforcement training strategies.

[0016] 3. The quantitative relationship between the peak value of polarization degree and roughness is established based on the correlation diagram of wavelength, wave width, altitude angle, azimuth angle and polarization degree. Based on the multi-factor correlation characteristics, the optimal factor level is found, and the quantitative detection mechanism model of the prediction and calibration area is established to improve the detection accuracy of multi-scale fruits.

[0017] Furthermore, when modeling outdoors, in step S10, the temperature and light of the area to be measured need to be gridded using high-precision satellite remote sensing image data, specifically in the following steps:

[0018] 1) Solar illumination environmental factor model: Based on the coordinates and time of the area to be measured, select solar radiation remote sensing satellite data products for preprocessing such as atmospheric radiation transfer, geometric correction, and radiation correction. The digital elevation model (DEM) is used for terrain shadow and solar radiation modeling, and illumination modeling: terrain analysis + solar radiation calculation. High-precision satellite remote sensing image data is used to grid the illumination of the area to be measured, and illuminometers or solar radiometers are placed at grid nodes to collect illumination conditions, and the outdoor partially coherent light sources are quantitatively described. The main parameters are spectral coherence, illumination angle (azimuth angle and elevation angle), wavelength, and spectral width. Obtain the solar radiation model of fruit tree fruits under illumination.

[0019] 2) Surface temperature inversion model: Based on the coordinates and time of the area to be measured, select thermal infrared and meteorological satellite data remote sensing satellite data products for preprocessing such as atmospheric radiation transfer, geometric correction, and radiation correction. Temperature inversion: thermal infrared band processing + emissivity correction to obtain the temperature spatial distribution characteristics of the area to be measured. High-precision satellite remote sensing image data is used to grid the temperature of the area to be measured, and air thermometers, soil thermometers, and infrared thermal imagers are placed at grid nodes to calibrate the surface and fruit tree fruit temperatures. The main parameters are temperature and spatial position, and establish the surface temperature inversion model of fruit tree fruits.

[0020] 3) Fruit tree fruits are equivalent to scattering media with different refractive indices and particles, distributed according to different branch angles and tree shapes. The horizontal and vertical spatial distribution characteristics of different structures are described by high-order optical parameters. Less simulates a single fruit tree or a region of interest (ROI) of fruit trees with different spacings, and simulates the reflectivity characteristics of different parameters.

[0021] 4) Experimentally measure the far-field spectral characteristics of light sources and fruits such as red dates and apricots at different angles and distances, extract spectral morphological parameters such as peak height, peak width, and peak area, and establish spectral response functions at different angles. From the correlation diagrams of factors such as wavelength, spectral width, elevation angle, azimuth angle, and degree of polarization, establish the quantitative relationship between the peak value of the degree of polarization and roughness.

[0022] 5) Establish a multi-parameter BPDF quantitative detection model, invert the fruit spatial distribution function from the far-field spectrum and the component spectrum of the fruit, change the spatial distribution structure of the calibration area to reconstruct the scattering kernel function, and quantitatively detect the fruit quality in the prediction area. Compare the quantitative detection results and accuracy evaluation of the calibration and prediction areas.

[0023] Establish an angular spectral response function model, and transfer multi-angle and spectral quantitative models.

[0024] Furthermore, transfer of multi-angle and spectral quantitative models: Take the spatial azimuth distance and angle between the calibration and prediction as key parameters, fit the correlation coefficient model of angle, wavelength, and polarized reflectance data, and realize the transfer of the multi-angle reflectance database model and the spectral polarization database model.

[0025] Furthermore, the environmental characteristics of the space to be measured in step S10 include: temperature, light, orientation, and polarization.

[0026] Furthermore, in step S10: for the method of grid division of the space to be measured, the space region boundary is divided according to the diffraction angle θ = 1.22λ / D = x / y of the camera lens and the front field of view angle FOV = x / y of the camera lens; where D represents the lens diameter size, x represents the distance between two points in space, and y represents the vertical distance from the lens central axis to the surface to be measured, and the space to be measured is grid-divided by the shooting height and the field of view angle.

[0027] Furthermore, in step S40, the fitted multi-angle polarization spectral response function is f(x) =

[0028] K(K0 + K1 + k2f(θ1, θ2) + k2f(θ3, θ4)) + T(T0 + T1 + T2f(θ1, θ2) + T2f(θ3, θ4)); where the k parameter and the T parameter are respectively the coefficients of the response functions of the grid environmental scattering characteristic terms, K0 represents isotropy, K1 represents geometric structure scattering, k2 represents surface scattering, and the angles correspond to the elevation angle and the azimuth angle. Reconstruct the typical environmental characteristic spectral scattering kernel function.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] The present invention combines GF-2 satellite remote sensing images and ground interaction verification experiments to select the typical environment of the orchard in the area to be measured, classifies and marks the light and temperature fields through spatial azimuth grid, and uses the change law of the reflectance of the same fruit in different angular polarization directions by the multi-spectral multi-angle polarization of the unmanned aerial vehicle to fit the reflectance curve model to quantitatively detect the accuracy of the fruit quality. An angular profiling calibration field is established by the temperature and light characteristics of the grid orchard area, and the spectral quantitative detection is combined with the regional evaluation of the unmanned aerial vehicle spectral image.

[0031] Secondly, for the multi-spectral camera of the unmanned aerial vehicle (5 channels), a polarizer is added in front of the camera to change the polarization direction for polarization calibration of the multi-polarization direction and polarization multi-angle reflectance. The four polarization directions of the polarizer are (0p, 45p, 90p, 135p). Correspondingly, the horizontal direction (0 - 360 degrees) is rotated 8 directions (0, 45, 90, 135, 180, 225, 270, 315) every 45 degrees; 160 polarization reflectance spectral images are obtained at one flight height of the same jujube orchard; according to the FOV field of view angle of the unmanned aerial vehicle and the resolution accuracy requirements, the flight height of the unmanned aerial vehicle is increased. For example, for every 0.5 m increase in height, a group of 160 polarization spectral images are taken by the above method, and the height is increased by 1 m. Correspondingly, the elevation angle is changed in the vertical direction to obtain 320 multi-angle polarization spectral images.

[0032] Finally, extract the morphological parameters of spectral reflectance and establish a multi-angle spectral reflectance database. Mainly use the layer stacking function of ENVI software to fuse the multi-angle polarized reflection spectral images, import them in sequence according to the shooting order, expand the reflectances in different directions of the jujube orchard along the Z-axis in sequence, and obtain the multi-angle polarized reflectance curves. Select different regions in the jujube orchard such as red dates, tree branches, open spaces, etc. as ROIs, conduct regional ROI feature statistics, and obtain the spectral mean features of different regions of ROIs. Use spectral morphological feature parameters such as peak width, peak height, and peak area to quantitatively describe the spectral characteristics of regional quality, and fit the multi-angle BPDF polarized reflectance function models of the quality of fruits such as red dates moisture, SSC, etc. Realize the regional quality evaluation of fruits by unmanned aerial vehicle.

[0033] The present invention calibrates the angle distribution function of far-field spectral images to invert the quality spectrum, and realizes the calibration of a low-resolution spectral camera by a high-resolution spectrometer and a multi-angle polarization method. The present invention makes a quality space distribution map for thematic data fusion, which can be used to carry out the distribution of moisture, acidity, and sugar content of regional fruit quality. Combined with geographical information, illumination temperature, and GPS coordinates, a big data system of fruit quality distribution environment is marked and established. By means of the ground fruit quality spectrum plus the regional illumination model, the regional quality prediction and evaluation of fruits in unknown regions are realized. Through the variation law of temperature and illumination over time, a spatio-temporal evolution model of fruits is established, providing data support for improving the quality and efficiency of the fruit industry and the future layout and structural adjustment. Brief Description of the Drawings

[0034] Figure 1 It is a schematic flow chart of a modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters provided by an embodiment of the present application;

[0035] Figure 2 It is a schematic diagram of a method for gridifying a space to be measured provided by an embodiment of the present application;

[0036] Figure 3 It is a schematic diagram of establishing an angular spectral response function model provided by an embodiment of the present application;

[0037] Figure 4 It is a schematic diagram of gridifying the space of an outdoor area to be measured provided by an embodiment of the present application;

[0038] Figure 5 It is a schematic flow chart of thematic mapping and quantitative evaluation of fruit regions provided by an embodiment of the present application. Detailed Description of the Embodiment

[0039] In order to further elaborate on a modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters of the present invention and achieve the intended invention purpose, the following will, in combination with preferred embodiments, detail the specific implementation manner, structure, characteristics and functions of a modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters proposed according to the present invention. In the following description, different "one embodiment" or "embodiments" do not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0040] The following will, in combination with specific embodiments, further introduce in detail a modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters of the present invention:

[0041] In the technical solution of the present invention, through high-precision satellite remote sensing image data and grid processing of temperature and light in the area to be measured, including: classifying and partitioning the temperature and light characteristics of the grid orchard area through ground experimental control points; establishing a multi-angle conformal calibration field based on fruit tree characteristic parameters, and determining the inversion grid area through spectral multi-angle and image interactive verification; combining spectral quantitative detection with unmanned aerial vehicle spectral image area evaluation through the spectral response function of the fruit area calibration; among them, the method for grid division of the space to be measured mainly solves the problem of how to improve the detection accuracy when using a spectrometer with low spectral resolution by corresponding the image field of view angles at different distances of the spectral camera with the field of view angle of the spectrometer and mutual verification, etc. The technical solution of the present invention is as follows:

[0042] Figure 1 It is a schematic flowchart of a modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters provided for the embodiments of the present application, including the following steps:

[0043] S10: Obtain the spatial distribution characteristics of light and temperature through high-precision satellite remote sensing image data, perform grid processing on the temperature and light in the area to be measured, and classify and partition the temperature and light characteristics of the grid orchard area through ground experimental control points according to the solar radiation model and the surface temperature inversion model;

[0044] S20: Establish a multi-angle conformal calibration field with spatial coordinate information through the temperature and light characteristics of the grid orchard area, and achieve multi-scale multi-angle quantitative detection through the transfer and interactive verification of the quantitative detection database and the spectral database of multi-angle reflectance quantitative detection;

[0045] S30: Grid classification and precise angle calibration of the spatial environmental characteristics to be measured, select the fruit reflectance characteristic angle, and establish a multi-angle polarization reflectance database;

[0046] S40: Use a drone multi-spectral camera to obtain the reflectance change rules of fruits in the same area at different angles and polarization directions. Through zoned classification marking experiments, obtain the growth temperature and light characteristic models of fruits. Through a directional reinforcement training strategy, extract key parameters with obvious environmental characteristics, fit the spectral reflectance curve model at the regional scale and the multi-angle polarization spectral response function of typical environmental characteristics, and combine spectral quantitative detection with the regional evaluation of drone spectral images.

[0047] Preferably, for the modeling method, first perform grid processing on the environmental characteristics of the space to be measured, conduct angle simulation calibration modeling indoors, and then go outdoors for verification modeling and cross-validation.

[0048] Further preferably, when conducting modeling outdoors, in step S10, it is necessary to perform grid processing on the temperature and light of the area to be measured through high-precision satellite remote sensing image data, and establish an angle spectral response function model and transfer of multi-angle and spectral quantitative models.

[0049] Further preferably, for the transfer of multi-angle and spectral quantitative models: Take the azimuth distance and angle between the calibration and prediction spaces as key parameters, fit the correlation coefficient model of angle, wavelength, and polarization reflectance data, and realize the transfer of the multi-angle reflectance database model and the spectral polarization database model.

[0050] Further preferably, the environmental characteristics of the space to be measured in step S10 include: temperature, light, azimuth, and polarization.

[0051] Preferably, in step S10: For the grid division method of the space to be measured, divide the space region boundary according to the diffraction angle of the camera lens θ = 1.22λ / D = x / y and the front field of view angle FOV of the camera lens = x / y; where D represents the lens diameter size, x represents the distance between two points in space, y represents the vertical distance from the center axis of the lens to the surface to be measured, and grid the space to be measured through the shooting height and field of view angle.

[0052] Preferably, in step S40, the fitted multi-angle polarization spectral response function is f(x) =

[0053] K(K0 + K1 + k2f(θ1,θ2) + k2f(θ3,θ4)) + T(T0 + T1 + T2f(θ1,θ2) + T2f(θ3,θ4)); where the k parameter and T parameter are the coefficients of the response functions of the grid environmental scattering characteristic terms respectively, K0 represents isotropy, K1 represents geometric structure scattering, k2 represents surface scattering, and the angles correspond to the elevation angle and azimuth angle. Reconstruct the spectral scattering kernel function of typical environmental characteristics.

[0054] Example 1.

[0055] (1) Precise angle calibration of the indoor BRDF device:

[0056] The indoor BRDF device is precisely calibrated at different angles, and the characteristic angles of fruit reflectance are selected to establish an angle database. According to the angle calibration of standard plates in different detection spaces, a fruit angle reflectance model is established. BRDF collects the light intensity reflectance at different angles of red dates in the crisp, fully ripe, and white ripe stages, and adds the spectral multi-angle polarization acquisition data. A large number of experiments are performed to fit the polarization spectral reflectance at different angles to establish an angle polarization spectral quantitative detection model. The surface of red dates in the crisp and ripe stage is smooth and has obvious polarization characteristics. Red dates with different color qualities are selected for comparison, and the multi-angle spectral reflectance of red, green, and red and white red date samples is used to quantitatively describe the quality of red dates by fitting the curve equation.

[0057] The ring light source is used to light up in sequence to simulate the lighting conditions of the sun at different times, avoiding image registration and realizing a multi-angle reflectivity database of typical lighting in the area to be measured at different times.

[0058] (2) Gridded angle and azimuth calibration spectral model for indoor space to be measured:

[0059] ① Theoretical simulation calculation of different fruit angle feature models:

[0060] According to the different periods of fruit maturity, we collected tissue sections of individual fruits to measure the pore size distribution, collected information such as the shape of the fruit tree (round head type, open head type), tree height, and row spacing, and simulated with less to calculate multi-scale scattering characteristics such as scattering coefficient and absorption coefficient. We characterized the multi-angle scattering characteristics of single fruit and canopy scales, selected appropriate collection angles and gridded spaces, and the fruit canopy scale can be equivalent to the scattering medium, and the particle size and refractive index can be quantitatively described. We used less software to simulate and calculate the scattering parameters of different fruits, such as scattering characteristics at different scales, scattering coefficients and absorption coefficients of leaves, and scattering coefficients and absorption coefficients of canopies with different reflectivity angles.

[0061] ② Gridding method of the space to be measured:

[0062] The spatial resolution of the camera is affected by the lens and the detection distance and azimuth. The spatial region boundary is divided according to the diffraction angle of the camera lens θ = 1.22λ / D = x / y and the front field of view angle FOV = x / y of the camera lens, where D represents the lens diameter size, x represents the distance between two spatial points, and y represents the vertical distance from the lens center axis to the surface to be measured. The space to be measured is gridded by the shooting height and the field of view angle. Consider the optical transfer function (MTF) to describe the spatial frequencies in different directions. As the distance increases, the field of view angle FOV becomes smaller, and the spatial cut-off frequency decreases, making it impossible to obtain high-frequency detail information. By collecting data from multiple angles, frequency stitching compensates for the resolution decrease caused by the spatial cut-off frequency. The iterative algorithm for collecting data from multiple angles improves the spatial resolution. By changing different angles, the high-frequency image details at large angles are increased, and spatial frequency registration improves the spatial resolution of the images captured by the camera. By changing the spatial angle and distance in sequence, the farther the distance, the smaller the camera field of view angle, and the more corresponding frequency information is missing. By taking images at more angles, as shown in Figure 2 (B). The accuracy of radiation calibration in the gridded area is improved by iteratively shooting images from multiple angles at a long distance and cross-validating with the predicted image positions. As shown in Figure 2 (A).

[0063] ③ Multi-angle ground profiling calibration method:

[0064] Through spatial gridding processing of the predicted area, control points are selected to divide the inversion area, and a fiber optic spectrometer and a multi-spectral camera are used for fixed-point multi-angle calibration and cross-validation. Specifically, it is divided by the illumination characteristics of a single fruit. Multiple fruits are combined according to a single-row ROI to form a single-row reflectance characteristic, and multiple fruits are combined vertically to form the row-column characteristics of the overall fruit image.

[0065] The reflectance characteristics of the fruit quality in the entire area are inverted through the combination of regions of interest (ROIs) of fruits in different rows and columns. According to the structural parameters such as the growth tree shape, tree height, row spacing, and plant spacing of the fruit trees in the orchard, a multi-angle profiling ground calibration field for fruit trees is established. A multi-angle profiling calibration field is established by gridding the temperature and illumination characteristics of the orchard area. The multi-angle ground profiling calibration field is used for cross-validation of spatial spectra and images, and a multi-angle polarization spectral response function model of the fruit tree canopy is established. The spectral quantitative detection is combined with the evaluation of the spectral image area by the unmanned aerial vehicle.

[0066] The multi-angle reflectance database is corrected by the method of cross-validating the reflectance of fruits at different positions inverted from the reflectance of a single fruit, realizing the one-to-one correspondence between the spectral reflectance and the corresponding position of the spectral image. By performing spatial segmentation on fruits at different angles and distances and matching the angles and azimuths in the model library. A multi-angle spectral and image data model library is established to achieve multi-angle quantitative detection.

[0067] (3) Azimuth calibration spectral image calibration model:

[0068] The standard board establishes a calibration indoor field at different angles to carry out polarization calibration and spatial orientation calibration of the standard board. The spectrum calibrates a certain point of the image, and the geometrically calibrated image is then calibrated at different orientations through the spectral coherence law. The spectrum and the image are mutually calibrated and verified. The spectral composite image extracts spectral and angular reflectance information from the image.

[0069] Experiments with standard calibration boards at different orientations establish a multi-angle calibration database, collect experiments on the multi-angle reflectance of fruits, and interactively verify and optimize the angular orientation model to establish a multi-angle spatial orientation spectral model of fruits. Establish a spectral angular reflectance model to achieve a refined quantitative description of spatial spectral images at different orientations. Specifically:

[0070] ① Establish an angular calibration model for the standard white board

[0071] Since there are differences in the reflectance of the standard white board at different angles during the actual experimental process, which has a certain impact on the detection accuracy, it is necessary to calibrate the reflectance of the standard white board at different angles. The influence brought by the angle of the standard white board is greater than that of the fruit quality. By establishing a functional relationship of the optimal standard white board angle variable, the correlation coefficients at different standard white board angles are calculated to achieve the prediction of the placement angle of the standard white board under different detection conditions. Taking the multi-angle of red dates as an example, the standard white board angle function relationship is obtained: f(x) = a*(sin(x - pi)) + b*((x - 10)^2) + c, a = 72.28(68.65, 85.92), b = 0.0006659(0.0006585, 0.0006733), c = 68.09(58.22, 77.97). The prediction and experimental verification results

[0072] ② Establish a distance calibration model for the standard white board

[0073] The spectrum will be affected by the spectral coherence during transmission, and there are certain differences in the sample spectra at different distances. Use standard white boards at different distances from the sample to correct the spectrum and establish a standard white board correction model at different distances to achieve the calibration of the sample spectra at different distances.

[0074] (4) Establish an angular spectral response function model:

[0075] The multi-spectral camera is calibrated with multi-angle polarization outdoors, spatial position, geometry and radiation. The characteristic orientation spatial background spectrum is selected to establish the spectral response function and angle response function. Multi-angle experiments are equivalent to improving the spectral resolution and detection accuracy. When there are large differences in reflectivity at different angles, and multi-angle differences in spectrum and light intensity, the spectral camera that combines spectrum with imaging can give the spatial distribution characteristics of fruit quality. The correlation coefficients of angle, wavelength, polarization and reflectivity are analyzed and fitted into an angle function model. Standard plates are placed during imaging, and radiation calibration, geometric calibration and ambient light field calibration are performed. A refined quantitative description of ambient lighting and angle is made to construct a spatial angle orientation spectral response function. The dotted box represents the prediction area, and the solid line part is the calibration area, such as Figure 3 (A, B, C, D) as shown.

[0076] (5) Multi-scale, multi-angle and spectral quantitative detection model transfer:

[0077] On the basis of consistent fruit quality, spectral reflectance and angle, polarization, and reflectance are collected under the same conditions to establish a statistical regression relationship model. The spatial orientation distance, angle, and polarization direction are used as key parameters, and the experimental data are fitted with angle, wavelength, reflectance, and polarization feature data to establish a correlation coefficient model. Through the multi-angle reflectance quantitative detection database and the spectrum database quantitative detection model, multi-scale and multi-angle quantitative detection is achieved.

[0078] (6) Spatial grid division of the outdoor test area:

[0079] The regional orchard images were selected through satellite images, and the normalized vegetation index NDVI was used for differential processing. GF2 has 4 bands, and 4 bands were obtained together with NDVI processing, totaling 8 spectral images. Using the ENVI software function fusion, NDVI13, NDVI14, NDVI23 NDVI24 (NDVI13 means NDVI processing for bands 1 and 3 respectively) were used to obtain orchards with obvious image color differences as the study area. Figure 4 As shown in (A), the GPS coordinates of the multi-angle near-ground and multi-angle reflectivity images of the ground are obtained by UAV. Figure 4 (B) The reflectivity shown is more detailed and accurate, and a quantitative comparison calibration experiment is conducted. The ground illuminance meter and ground thermometer are used to obtain the ground temperature and the temperature field distribution experiment of the thermal imager. The thermal imager and spectrometer are calibrated for spatial division and partition, and the typical ambient temperature, light and orientation calibration model of the orchard is reconstructed.

[0080] ① Grid model of spatial temperature and illumination field in the area to be measured:

[0081] Based on the temperature and light reflectance differences at different test points in the area to be measured, and taking the difference between each point being less than a specific value as the grid size division condition according to the actual orchard situation, typical environmental control points are selected to perform grid processing on the geographical orientation, temperature, and light conditions of the area to be measured. Through the grid processing of the outdoor orchard space, the typical environmental orientation, temperature, and light characteristics of the area are extracted. Carry out multi-angle polarization orientation calibration and thermal imager temperature zone calibration, establish calibration models for different temperature ranges and gradient experiment zones, and reduce the temperature influence to improve the quantitative detection accuracy of outdoor red dates. Establish a multi-angle profiling calibration field through the temperature and light characteristics of the grid orchard area, and combine spectral quantitative detection with the evaluation of the drone spectral image area.

[0082] ② Grid model of temperature and light fields in the area to be measured at different time periods:

[0083] Since the solar altitude angle and azimuth angle change at different time periods, the above spatial grid model will change. The changed area and angle can be determined by simulating and calculating the solar altitude angle and azimuth angle at different time periods to correct the above grid space area. For example, it is about 1 degree every half hour, and at the same time, the illuminance needs to be corrected.

[0084] ③ Reconstruct the typical environmental orientation model

[0085] Through the simulation calculation of the geographical information coordinates (GPS) of orchard fruit trees to determine the geographical orientation, and the solar altitude angle and azimuth angle at different times of morning, noon, and evening, and through the reconstruction and reproduction of the typical environmental light of the fruit quality spectrum, accurately and quantitatively describe the fruit area such as geographical coordinates and growth orientation characteristic spectra. Establish a calibration spectrum model for the typical environment and fruit orientation of actual fruit trees.

[0086] (7) UAV multi-angle polarization reflectance experiment on fruits near the ground:

[0087] ① UAV calibration experiment at different orientations and multi-angle reflectance database:

[0088] The UAV multispectral camera collects at different heights of 3 meters, 5 meters, and 8 meters. At each height, the spectral camera of the UAV in the hovering mode horizontally adds a polarizer with four polarization directions (0p, 45p, 90p, 135p) and rotates 8 directions every 45 degrees (0, 45, 90, 135, 180, 225, 270, 315) to obtain multi-angle spectral images. Modify the longitude and latitude of the header file of each picture, such as the longitude and latitude coordinates of the jujube orchard, and input the longitude and latitude coordinates of the area to be measured, such as E81°31'14"N 40°31'42".

[0089] This embodiment specifically inputs and arranges in the order of the following 4 polarization directions: 0p(0, 45, 90, 135, 180, 225, 270, 315), 45p(0, 45, 90, 135, 180, 225, 270, 315), 90p(0, 45, 90, 135, 180, 225, 270, 315), 135p(0, 45, 90, 135, 180, 225, 270, 315). One scene has 4 * 8 * 5 = 160 reflected band images, and different angles and polarization azimuths are used as the abscissa of the multi-angle polarization reflectance. The Envi software synthesizes 160 reflectance files to establish a multi-angle reflectance database of fruits of different orientations with classification marks.

[0090] Finally, use the layer stacking function of the envi software to fuse the multi-angle polarization reflection spectrograms, import them in sequence according to the shooting order, expand the reflectances in different directions of the jujube orchard along the Z-axis in turn, and obtain the multi-angle polarization reflectance curves. Select different regions in the jujube orchard such as red dates, branches, open spaces, etc. as ROIs, perform regional ROI feature statistics, and obtain the spectral mean characteristics of different regions of ROIs. Use spectral morphological feature parameters such as peak width, peak height, and peak area to quantitatively describe the spectral characteristics of regional quality, and fit the multi-angle BPDF polarization reflectance function models of the quality of fruits such as red date moisture and SSC. Realize the regional quality evaluation of fruits by unmanned aerial vehicle.

[0091] ② Collection of physical and chemical indexes of artificially marked fruits and classification mark spectral measurement experiment:

[0092] Use a spectrometer to collect the spectra of in-situ artificially marked fruits, record the corresponding fruit shape characteristics such as color and texture, and measure the physical and chemical indexes such as fruit moisture and sugar content. Through the spectral database of fruits in different orientations, establish a quantitative detection model for the spectra of fruits in different orientations.

[0093] ③ Classification mark of fruit spectral reflectance and multi-angle reflectance cross-validation model

[0094] For the ROI feature regions of the multi-spectral and multi-angle images of the unmanned aerial vehicle, select red and green fruits for classification marks, such as the reflectances of red dates in green and red according to color. Further perform cross-validation of azimuth marks and single-quality mark detections to achieve quantitative characterization of fruits. Provide a reference for quality classification and pricing.

[0095] Example 2. Fruit regional thematic mapping and quantitative evaluation:

[0096] Unmanned aerial vehicle regional quantitative evaluation: Since there are cases of second and third flowering and fruiting in the orchard area, it is necessary to identify and further conduct regional quality classification and quantitative evaluation. Add geographical coordinate information of fruit orientation, multi-angle and spatial spectra through the unmanned aerial vehicle for thematic mapping, providing an important reference for improving the quality and efficiency of the fruit industry. (Such asFigure 5 Among A, B, C, D, E, F, G)

[0097] As mentioned above, it is only the preferred embodiment of the embodiments of the present invention, and does not impose any form of limitation on the embodiments of the present invention. Any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the embodiments of the present invention still fall within the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A modeling method for parametric classification and quantitative detection of multi-scale fruit environmental characteristic parameters, characterized in that Including the following steps: S10: Obtain the spatial distribution characteristics of light and temperature through high-precision satellite remote sensing image data, perform grid processing on the temperature and light of the area to be measured, and classify and mark the temperature and light characteristics of the grid-shaped orchard area according to the solar radiation model and the surface temperature inversion model through ground experimental control points; S20: Establish a multi-angle profiling calibration field with spatial coordinate information based on the temperature and light characteristics of the grid-shaped orchard area, and realize multi-scale multi-angle quantitative detection through the transfer and interactive verification of the quantitative detection models in the multi-angle reflectance quantitative detection database and the spectral database; S30: Grid classification and precise angle calibration of the spatial environmental characteristics to be measured, select the angle of the fruit reflectance characteristics, and establish a multi-angle polarization reflectance database; S40: Use a drone multi-spectral camera to obtain the variation law of the reflectance of fruits in the same area at different angles and polarization directions, obtain the fruit growth temperature and light characteristic models through the partition classification marking experiment, extract the key parameters with obvious environmental characteristics through the directional reinforcement training strategy, fit the spectral reflectance curve model at the regional scale and the multi-angle polarization spectral response function of typical environmental characteristics, and combine spectral quantitative detection with the regional evaluation of drone spectral images.

2. The modeling method according to claim 1, wherein For the modeling method, first perform grid processing on the spatial environmental characteristics to be measured, perform angle simulation calibration modeling indoors, and then perform verification modeling and interactive verification outdoors.

3. The modeling method according to claim 2, wherein When performing modeling outdoors, in step S10, it is necessary to perform grid processing on the temperature and light of the area to be measured through high-precision satellite remote sensing image data, and establish an angle spectral response function model and a multi-angle and spectral quantitative model transfer.

4. The modeling method according to claim 3, wherein For the multi-angle and spectral quantitative model transfer: Take the calibration and prediction spatial azimuth distance and angle as key parameters, fit the correlation coefficient model of the angle, wavelength, and polarization reflectance data, and realize the transfer of the multi-angle reflectance database model and the spectral polarization database model.

5. The modeling method according to claim 2, wherein The spatial environmental characteristics to be measured in step S10 include: temperature, light, azimuth, and polarization.

6. The modeling method according to claim 1, wherein In step S10: For the method of grid division of the space to be measured, divide the spatial area boundary according to the camera lens diffraction angle θ = 1.22λ / D = x / y and the front field of view angle FOV = x / y of the camera lens; where D represents the lens diameter size, x represents the distance between two points in space, y represents the vertical distance from the center axis of the lens to the surface to be measured, and grid the space to be measured through the shooting height and the field of view angle.

7. The modeling method according to claim 1, wherein In the step S40, the fitted multi-angle polarization spectral response function is f(x) = K(K0 + K1 + k2f(θ1, θ2) + k2f(θ3, θ4)) + T(T0 + T1 + T2f(θ1, θ2) + T2f(θ3, θ4)); where the k parameter and the T parameter respectively correspond to the coefficients of the response function of the gridded environmental scattering characteristic terms, K0 represents isotropy, K1 represents geometric structure scattering, k2 represents surface scattering, and the angles correspond to the altitude angle and the azimuth angle. Reconstruct the typical environmental characteristic spectral scattering kernel function.

Images