Unmanned aerial vehicle lai inversion method and device based on high-precision canopy positioning

Abstract

The application discloses a UAV LAI inversion method based on high-precision canopy positioning, which comprises the following steps: acquiring UAV images of a research area, collecting ground LAI measured values, and recording coordinates of LAI measuring points by using a handheld RTK measuring instrument; taking the coordinates of the LAI measuring points as the center coordinates, extracting pixel points of a view buffer from the orthographic image, and extracting preselected characteristic values of the LAI measuring points from the UAV image characteristics of the pixel points in the view buffer, so as to construct a sample set of high-precision spatial matching between the LAI measured values and the preselected characteristic values of the LAI measuring points; establishing a target inversion model based on the sample set, and applying the target inversion model to the research area to calculate the LAI values pixel by pixel and generate a pixel-level LAI inversion map. The application can improve the spatial registration accuracy of the ground samples and the UAV images, reduce the sample mismatching error, and obtain pixel-level LAI distribution results on continuous space.

Description

Technical Field

[0001] This invention relates to the fields of agricultural remote sensing monitoring and precision agriculture technology, specifically to a UAV LAI inversion method based on high-precision canopy positioning, and also to a UAV LAI inversion device based on high-precision canopy positioning, and a UAV fine inversion of field crop leaf area index (LAI) based on high-precision positioning and pixel-by-pixel inversion using a handheld RTK measuring instrument. Background Technology

[0002] Leaf Area Index (LAI) is an important parameter reflecting crop canopy structure, growth status, and light utilization capacity, and it has significant application value in crop growth monitoring, yield estimation, nutrient diagnosis, and precision agriculture management. Current methods for obtaining LAI mainly include manual sampling and ground observation using plant canopy analyzers (such as the LAI-2200C). While these methods can obtain relatively reliable observation data, they suffer from low measurement efficiency, limited spatial coverage, and difficulty in reflecting continuous spatial distribution in the field. With the development of UAV remote sensing technology, using UAVs equipped with multispectral sensors to acquire crop canopy imagery and combining it with ground-measured data to establish inversion models has become an important means of rapid LAI monitoring. UAV multispectral imagery has advantages such as high resolution, flexible acquisition, and high coverage efficiency, providing rich spectral and spatial structure information for crop LAI inversion.

[0003] However, existing LAI retrieval methods based on UAV imagery still have shortcomings. Current methods typically establish a correspondence between the LAI measurement points of the canopy analyzer and the pixels in the UAV imagery by roughly estimating the location. While this establishes a correspondence between measurement points and image pixels, the principle of the canopy analyzer is to calculate the LAI value using spectral information intercepted by vegetation leaves within its field of view. Its field of view involves tens of thousands or even millions of image pixels, making it difficult to establish a good correspondence between the spectral observation value of a single pixel and the LAI value observed on the ground by the canopy analyzer. Especially for field crops with clearly defined rows and ridges and strong spatial heterogeneity in canopy structure, the shift of the ground LAI measurement point along the row direction can significantly alter the canopy structure within the field of view centered on that measurement point. This results in the spectral or vegetation index features extracted from a single pixel not directly corresponding to the ground LAI observation value, leading to problems such as insufficient representativeness of the corresponding pixels, increased mismatch errors, and decreased model stability. Furthermore, under manual observation conditions, canopy analyzers typically require the use of a canopy cap to alter the effective field of view and eliminate interference from the observer and direct sunlight, serving as an auxiliary constraint. Therefore, it is necessary to propose a method that uses a handheld RTK to accurately record the location of LAI (Laboratory Area Indicator) measurement points, determines the field of view using plant height and the canopy analyzer's field of view angle, and achieves pixel-by-pixel UAV-based fine inversion of LAI in field crops within the field of view. Summary of the Invention

[0004] The purpose of this invention is to solve the problems of inaccurate spatial correspondence between the field of view of the ground canopy analyzer and the features of UAV images, large pixel sample mismatch errors, and difficulty in achieving fine LAI inversion in the prior art. This invention provides a UAV LAI inversion method based on high-precision canopy positioning, and also provides UAV LAI inversion equipment based on high-precision canopy positioning.

[0005] This invention uses the precise location of the LAI measurement point recorded by the handheld RTK instrument as the spatial anchor point for constructing the field of view. It calculates the field of view radius corresponding to the measurement point by combining the crop height and the instrument's field of view angle. It constructs a corresponding pixel buffer with the coordinates recorded by the handheld RTK instrument as the center, and establishes an inversion model based on the vegetation index features within the buffer, thereby realizing the fine inversion of LAI values ​​within the pixel-by-pixel field of view.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The UAV LAI inversion method based on high-precision canopy positioning includes the following steps:

[0008] Step 1: Acquire UAV images of the study area during the same crop growth period; for each pre-set LAI measurement point in the study area: use a plant canopy analyzer to perform ground LAI measurement to obtain the measured LAI value, and record the coordinates of the LAI measurement point with a handheld RTK measuring instrument;

[0009] Step 2: The UAV imagery is preprocessed to obtain orthophotos, and the orthophotos and LAI measurement points are located in the same reference coordinate system;

[0010] Step 3: Generate a digital surface model from the orthophoto. and digital elevation model And calculate the crop canopy height. ;

[0011] Step 4: Based on crop canopy height The center angle of the field of view of the observation ring of the plant canopy analyzer selected during ground-based LAI measurements. The radius of the preselected field of view buffer corresponding to each observation ring is determined; the directional angle range of the preselected field of view buffer is determined from the actual observation direction angle range during ground LAI measurement; the radius and directional angle range of the preselected field of view buffer constitute the preselected field of view buffer parameter group.

[0012] Step 5: Using the coordinates of the LAI measurement points as the coordinates of the center of the preselected field of view buffer, extract the pixels of the preselected field of view buffer from the orthophoto based on the parameter group of the preselected field of view buffer, and extract the preselected feature values ​​of the LAI measurement points from the UAV image features of the pixels in the preselected field of view buffer. Construct a sample set from the LAI measured values ​​and the preselected feature values ​​of the LAI measurement points.

[0013] Step 6: Based on the sample set and the parameter groups of each pre-selected view buffer, establish the corresponding LAI inversion model, and determine the target inversion model and the corresponding view buffer parameter group according to the preset evaluation index;

[0014] Step 7: For each pixel of the orthophoto image within the study area, calculate the LAI generation value pixel by pixel based on the target inversion model and the view buffer parameter set to generate the LAI inversion map.

[0015] As described above, the preprocessing includes radiometric calibration, geometric correction, image stitching, orthorectification, and spatial registration.

[0016] The radius of the preselected view buffer as described above is calculated based on the following formula:

[0017] ,

[0018] in, The radius of the preselected view buffer. The height of the crop canopy. The center angle of the field of view of the observation ring.

[0019] As mentioned above, the directional angle range of the preselected view buffer is equal to the actual observation directional angle range of the ground-based LAI measurement;

[0020] When the plant canopy analyzer is equipped with a cap to block the field of view in non-target directions during ground-based LAI measurements, the pre-selected field of view buffer does not cover the cap.

[0021] When there are no directional constraints during ground-based LAI measurements, the directional angle range of the pre-selected view buffer is 360°.

[0022] As described above, step 5 specifically includes the following steps:

[0023] Step 5.1: Extract UAV image features of each pixel in the study area from the orthophoto, and select pre-selected features from the UAV image features;

[0024] Step 5.2: For each LAI measurement point, with the LAI measurement point as the center, delineate the corresponding preselected view buffer in the orthophoto according to the parameter group of each preselected view buffer; extract the values ​​of the preselected features of each pixel in the preselected view buffer in the orthophoto; take the average value of the values ​​of the preselected features of each pixel in the same preselected view buffer under the same preselected feature as the preselected feature value of the corresponding LAI measurement point under the corresponding preselected feature and the preselected view buffer;

[0025] Step 5.3: Combine different preselected view buffer parameter groups and different preselected feature types in pairs to obtain different preselected buffer-feature setting combinations;

[0026] For each preselected buffer-feature setting combination: the ground-based measured LAI value and the preselected feature value of the LAI measurement point are paired to form LAI data pairs, and the LAI data pairs of all LAI measurement points in the study area constitute the corresponding sample groups; the sample groups corresponding to all preselected buffer-feature setting combinations constitute the sample set.

[0027] The pre-selected features mentioned above in step 5 are one or more of the following UAV image features:

[0028] Difference vegetation index (DVI), super green index (ExG), super green and super red difference index (ExGR), green leaf index (GLI), normalized difference vegetation index (GNDVI), normalized red edge difference index (NDRE), normalized vegetation index (NDVI), normalized green and red difference index (NGRDI), optimized soil-regulated vegetation index (OSAVI), and visible light atmospheric impedance vegetation index (VARI).

[0029] As described above, step 6 specifically includes the following steps:

[0030] A LAI inversion model is constructed to fit the mapping relationship between the pre-selected feature values ​​of LAI measurement points and the measured LAI values. The model is fitted using sample groups corresponding to different pre-selected buffer-feature setting combinations to determine the specific parameters of the corresponding LAI inversion model, which are then used as LAI inversion models to be screened. Based on preset evaluation indicators, a target inversion model is selected from the LAI inversion models to be screened, and the pre-selected buffer-feature setting combination corresponding to the target inversion model is the preferred buffer-feature setting combination.

[0031] The evaluation index is one or more of the following: coefficient of determination R², root mean square error RMSE, and mean absolute error MAE.

[0032] As described above, step 7 specifically includes the following steps:

[0033] Each pixel in the orthophoto image within the study area is taken as the current processing point. For each current processing point: the corresponding view buffer is determined by the view buffer parameter group in the preferred buffer-feature setting combination; the preferred feature is the feature type in the preferred buffer-feature setting combination, and the values ​​of the preferred features of each pixel in the view buffer of the corresponding orthophoto image are extracted; the average value of the preferred features of all pixels in the corresponding view buffer is calculated; the average value of the preferred features corresponding to the current processing point is input into the target inversion model to obtain the corresponding LAI generation value.

[0034] Based on the relative spatial position of each pixel, the LAI generated values ​​are arranged in a raster image to generate an LAI inversion map.

[0035] A computer device includes a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements steps 2 to 7 of any of the UAV LAI inversion methods based on high-precision canopy positioning.

[0036] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements steps 2 to 7 of any of the UAV LAI inversion methods based on high-precision canopy positioning.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This invention focuses on the precise correspondence between LAI (Layered Area) observations from the LAI-2200C plant canopy analyzer and the corresponding vegetation features from a UAV within the same field of view. The core function of the handheld RTK (Real-Time Kinematic) is to accurately record the spatial location of LAI measurement points and use this location as the spatial anchor point for subsequent field-of-view construction. Plant height information is not used as a dependent variable in LAI inversion modeling, but rather to construct the field-of-view radii corresponding to the five observation loops of the LAI-2200C, thereby constructing a ring-of-view buffer that perfectly matches ground observations. Existing technologies obtain image ranges matching the LAI-2200C observation scale through resampling, and then use resampling processing for LAI inversion to set the size of the five observation loops. This results in an overly coarse spatial correspondence, and the center position of the circular convolution kernel used may deviate significantly from the actual midpoint of the canopy analyzer's observations. Furthermore, existing technologies based on resampling significantly reduce the spatial resolution of the inversion results. Compared to existing technologies, this invention effectively reduces the mismatch problem of pixel samples within the field of view caused by the deviation of the measurement center position of the sample points. In particular, it reduces feature distortion caused by significant changes in the canopy structure within the buffer zone due to the offset of the measurement points in the row direction. This effectively improves the representativeness, consistency, and reliability of the inversion model of the samples, and ensures pixel-level LAI inversion results from UAV images. This invention utilizes the LAI measurement results of the canopy analyzer and the corresponding vegetation index within the pixel field of view buffer to establish an inversion model, realizing fine inversion of LAI within the field of view of UAV images pixel by pixel. This results in a refined spatial distribution map of LAI for field crops, more accurately characterizing the crop canopy structure and the spatial heterogeneity of LAI. Attached Figure Description

[0039] Figure 1 This is a flowchart of the present invention.

[0040] Figure 2 This is a schematic diagram of LAI measurement using a plant canopy analyzer based on a handheld RTK measurement device. The center angle of the field of view of the observation ring.

[0041] Figure 3 This image shows the orthophoto of rice during its growth period from an UAV orthophoto and the spatial distribution of canopy height. (a) is the orthophoto of rice during its growth period from the UAV orthophoto; (b) is the spatial distribution of canopy height.

[0042] Figure 4 A schematic diagram of the field of view buffers corresponding to the five observation rings of the LAI-2200C plant canopy analyzer; (a) schematic diagram of the 270° field of view buffer; (b) schematic diagram of the 360° field of view buffer. The radius of the preselected view buffer.

[0043] Figure 5 This is a graph showing the RMSE differences of LAI inversion results for various vegetation indices under different viewpoint buffers. Among them, NDRE is the Normalized Red Edge Difference Index, GNDVI is the Normalized Green Light Difference Vegetation Index, DVI is the Difference Vegetation Index, NDVI is the Normalized Vegetation Index, OSAVI is the Optimized Soil-Regulated Vegetation Index, NGRDI is the Normalized Green-Red Difference Index, VARI is the Visible Light Atmospheric Impedance Vegetation Index, ExG is the Supergreen Index, ExGR is the Supergreen-Superred Difference Index, and GLI is the Green Leaf Index.

[0044] Figure 6 This is a pixel-by-pixel LAI inversion result based on UAV imagery. Detailed Implementation

[0045] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to embodiments. The embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0046] Example 1:

[0047] Step 1: Select the study area and set up LAI measurement points within the study area; during the same crop growth stage (the booting stage is selected in this example): acquire UAV images of the study area; for each LAI measurement point within the study area, use a plant canopy analyzer to perform ground LAI measurements to obtain the measured LAI values, and simultaneously record the spatial coordinates of the center of each LAI measurement point using a handheld RTK measuring instrument as the coordinates of the LAI measurement points.

[0048] like Figure 1As shown, UAV imagery acquisition and concurrent ground-based LAI (Latency Area Index) measurements were conducted within the study area. The UAV, equipped with a multispectral sensor, performed low-altitude aerial photography of the study area, acquiring UAV images of the area. These images included multispectral and visible light images, with the multispectral images encompassing at least green, red, red-edge, and near-infrared bands. LAI measurement points were simultaneously deployed on the ground within the study area. At each LAI measurement point, a plant canopy analyzer was used to measure LAI, and a handheld RTK (Real-Time Kinematic) measuring instrument was used to simultaneously record the spatial coordinates of the LAI measurement point center. In this embodiment, the plant canopy analyzer used was model LAI-2200C. The handheld RTK measuring instrument provided centimeter-level positioning accuracy to improve the spatial registration accuracy between the ground-based LAI measurement point locations and the corresponding pixels in the UAV images, ensuring an accurate spatial correspondence between the ground-based measured samples and the UAV images. It also served as a spatial anchor point for subsequent field-of-view buffer construction. Since the present invention requires the construction of a field-of-view buffer corresponding to the observation field of the plant canopy analyzer with the LAI measuring point as the center, the positional accuracy of the LAI measuring point will directly affect the representativeness of the canopy structure and its corresponding UAV image features within the field-of-view buffer.

[0049] Step 2: Preprocess the UAV imagery to obtain orthophotos. The orthophotos and LAI measurement points are located in the same reference coordinate system. The orthophotos include multispectral reflectance images and orthophoto visible light images.

[0050] The preprocessing of UAV imagery includes radiometric calibration, geometric correction, image stitching, orthorectification, and spatial registration, ensuring that the preprocessed orthorectified imagery and LAI measurement points are aligned within a unified coordinate reference frame. In some implementations, aerial survey control points are deployed, and their spatial coordinates are recorded using a handheld RTK measuring instrument to assist in the geometric correction of the UAV imagery, ensuring that multiple orthorectified images can be overlapped. The main function of the handheld RTK measuring instrument in this invention is to accurately acquire the coordinates of the LAI measurement points and, based on this, determine the spatial reference constructed using a pixel-by-pixel view buffer.

[0051] Step 3: Generate a digital surface model based on orthophotos and digital elevation model Calculate the crop canopy height using the following formula. :

[0052] ,

[0053] Among them, digital surface model Digital elevation model (DEM) includes elevation information of crop canopy and land surface. This only includes bare ground elevation information. The orthophoto imagery of rice in the booting stage and the spatial distribution of canopy height in the study area of ​​this embodiment are shown below. Figure 2 As shown, this can serve as an important basis for subsequently determining the feature extraction range (and the view buffer).

[0054] Step 4: Based on crop canopy height And the center angle of the field of view of the observation ring of the plant canopy analyzer selected during ground-based LAI measurements. The radius of the pre-selected view buffer zone corresponding to each observation ring is determined; the directional angle range of the pre-selected view buffer zone is determined based on the actual observation directional angle range during ground LAI measurement. The radius and directional angle range of the pre-selected view buffer zone constitute the pre-selected view buffer zone parameter group; the view buffer zone is the pixel range in the orthophoto with the buffer anchor point as the center, one view buffer zone corresponds to one buffer anchor point, and the view buffer zone is used for subsequent steps to extract UAV image features; for the buffer anchor point selected in subsequent step 5, it is the LAI measurement point; for the buffer anchor point selected in step 7, it is each pixel point in the orthophoto. Specifically:

[0055] (1) Determine the radius of the preselected view buffer zone:

[0056] In this embodiment, the LAI-2200C plant canopy analyzer performs LAI observations from an upward perspective, reflecting the comprehensive response of the canopy structure within a certain field of view above the LAI measurement point. UAV imagery, on the other hand, performs observations from a downward perspective, with its feature extraction range representing the ground projection area. To match the image feature extraction range of the UAV imagery with the ground observation support domain, this invention uses the precisely recorded LAI measurement point location by a handheld RTK measuring instrument as the center, combined with the crop canopy height... and the center angle of the field of view of the observation ring The radius of the view buffer used for feature extraction is calculated according to the following formula:

[0057] ,

[0058] in, The radius of the preselected view buffer. The height of the crop canopy. The center angle of the field of view of the observation ring is shown. In this embodiment, the center angles of the five annular fields of view of the plant canopy analyzer are 7°, 23°, 38°, 53°, and 68°, respectively, and the corresponding field of view buffers are as follows: Figure 3 As shown.

[0059] (2) Determine the directional angle range of the pre-selected view buffer:

[0060] In this embodiment, the directional angle range of the preselected view buffer is equal to the actual observation directional angle range of the ground LAI measurement in step 1.

[0061] When the plant canopy analyzer is equipped with a cover to shield the field of view from non-target directions during ground-based LAI measurements, the pre-selected field of view buffer is a directional field of view buffer (the pre-selected field of view buffer is a sector centered on the buffer anchor point). The pre-selected field of view buffer is determined based on the actual observation direction angle range in step 1 of the ground-based LAI measurement, ensuring that the pre-selected field of view buffer does not cover the cover, thus reducing the impact of non-target interference on image feature extraction. For example, in actual ground-based LAI measurements, the canopy behind the person may obscure the measurement results. Therefore, this cover is used to shield the interference signal in the area behind the person, and the corresponding actual observation direction angle range does not include the angle range of the tester holding the plant canopy analyzer.

[0062] When no directional constraints are set during ground-based LAI measurement (i.e., no cover cap is set), the constructed pre-selected view buffer is an omnidirectional view buffer with a directional angle range of 360° (the pre-selected view buffer is a circle with the buffer anchor point as the center).

[0063] In this field, when using the LAI2200C plant canopy analyzer for measurements, the observation rings can be manually turned off to reduce errors caused by inappropriate viewing angles. For example, on slopes, if it is difficult to ensure strict geometric consistency between the upper and lower observations, the outermost two or three observation rings can be masked during the analysis to reduce geometric errors caused by the slope.

[0064] By constructing a view buffer, the features extracted from UAV images can more realistically represent the canopy structure range corresponding to the ground LAI measurement, reducing sample mismatch error.

[0065] Step 5: Using the coordinates of the LAI measurement points obtained from the handheld RTK measuring instrument as the coordinates of the center of the pre-selected view buffer, and based on the pre-selected view buffer parameter set, select pixels within the pre-selected view buffer from the orthophoto. Extract the pre-selected feature values ​​of the LAI measurement points from the UAV image features of the pixels within the pre-selected view buffer. Construct a sample set using the corresponding measured LAI values ​​and the pre-selected feature values ​​of the LAI measurement points, thereby establishing the correlation between the measured ground LAI values ​​and the corresponding UAV image features within the view buffer. Specifically, this includes the following steps:

[0066] Step 5.1: Extract UAV image features of each pixel in the study area from the orthophoto. Select pre-selected features from the UAV image features, with the number of types of pre-selected features being ≥1. UAV image features include spectral features and texture features. In this embodiment, vegetation index features are selected from spectral features. Vegetation index features calculated from the multispectral reflectance image in the orthophoto include: Difference Vegetation Index (DVI), Supergreen Index (ExG), Supergreen-Superred Difference Index (ExGR), Green Leaf Index (GLI), Normalized Difference Vegetation Index (GNDVI), Normalized Red Edge Difference Index (NDRE), Normalized Vegetation Index (NDVI), Normalized Green-Red Difference Index (NGRDI), Optimized Soil-Regulated Vegetation Index (OSAVI), and Visible Light Atmospheric Impedance Vegetation Index (VARI). The pre-selected features are one or more of the above vegetation indices.

[0067] Step 5.2: For each LAI measurement point, based on different pre-selected view buffer parameter groups, select the corresponding pre-selected view buffer and calculate the corresponding pre-selected feature value for the LAI measurement point:

[0068] For each LAI measurement point, a preselected view buffer is defined in the corresponding orthophoto based on the LAI measurement point as the center and the parameter group of each preselected view buffer. The values ​​of the preselected features of each pixel in the preselected view buffer in the orthophoto are extracted. The average value of the values ​​of the preselected features of each pixel in the same preselected view buffer under the same preselected feature is taken as the preselected feature value of the LAI measurement point under the corresponding preselected feature and the preselected view buffer.

[0069] Step 5.3: Combine the different preselected view buffer parameter groups from Step 4 with the different preselected feature types from Step 5.1 in pairs to obtain different preselected buffer-feature setting combinations.

[0070] For each pre-selected buffer-feature setting combination: the ground-based measured LAI value and the pre-selected feature value of the LAI measurement point are paired to form LAI data pairs, and the LAI data pairs of all LAI measurement points in the study area constitute the corresponding sample groups; the sample groups corresponding to all pre-selected buffer-feature setting combinations constitute the sample set; thus forming multiple sets of high-precision spatial matching sample sets.

[0071] Step 6: Based on the sample set and the parameter sets of each pre-selected view buffer, establish the corresponding LAI inversion model, and determine the target inversion model and the corresponding view buffer parameter set according to the preset evaluation indicators. The specific process includes:

[0072] A LAI inversion model is constructed to fit the mapping relationship between the pre-selected feature values ​​of LAI measurement points and the measured LAI values. The model is fitted using sample groups corresponding to different pre-selected buffer-feature setting combinations to determine the specific parameters of the corresponding LAI inversion model, which are then used as LAI inversion models to be screened. Based on preset evaluation indicators, a target inversion model is selected from the LAI inversion models to be screened. The pre-selected buffer-feature setting combination corresponding to the target inversion model is the preferred buffer-feature setting combination.

[0073] In this embodiment, the LAI inversion model adopts a linear regression model to fit the mapping relationship between the pre-selected feature values ​​of LAI measurement points and the measured values ​​of LAI. Other models suitable for quantitative inversion can also be used.

[0074] For LAI inversion models established under different pre-selected buffer-feature settings, evaluation metrics (one or more of the following metrics: coefficient of determination R², root mean square error RMSE, and mean absolute error MAE) are used to evaluate the accuracy of the inversion models. When multiple evaluation metrics are used, the priority from high to low is: coefficient of determination R², root mean square error RMSE, and mean absolute error MAE. Figure 4 The diagram shows the mean root mean square error (RMSE) of each vegetation index under the multi-scale buffer. By comparing the model error under different conditions, the pre-selected buffer-feature setting combination with the best accuracy is selected as the preferred buffer-feature setting combination, and the corresponding LAI inversion model to be screened is used as the final target inversion model.

[0075] Step 7: Apply the target inversion model to all pixels to be inverted in the UAV imagery of the study area. For each pixel in the orthophoto imagery of the study area, calculate the LAI generation value pixel by pixel based on the target inversion model and the view buffer parameter set to generate an LAI inversion map, thereby obtaining the pixel-level LAI spatial distribution result. Specifically, this includes the following steps:

[0076] Each pixel in the orthophoto of the study area is sequentially used as the current processing point. For each current processing point: the corresponding view buffer is determined by the view buffer parameter group in the preferred buffer-feature setting combination; the preferred feature is the feature type in the preferred buffer-feature setting combination, and the values ​​of the preferred features of each pixel in the view buffer of the corresponding orthophoto are extracted; the average value of the preferred features of all pixels in the corresponding view buffer is calculated; the average value of the preferred features corresponding to the current processing point is input into the target inversion model to obtain the corresponding LAI generated value. Based on the relative spatial position of each pixel, the LAI generated values ​​are arranged in raster form to generate a pixel-by-pixel optimal LAI inversion map to represent the continuous spatial distribution of crop LAI in the study area. Figure 5The image shown is the pixel-by-pixel optimal LAI inversion map, reflecting the spatial distribution effect obtained by this invention through high-precision positioning with a handheld RTK measuring instrument, sample matching under field-of-view constraints, and pixel-by-pixel inversion. This result can be used for crop growth monitoring, field difference analysis, yield assessment, and precision agricultural management.

[0077] In this embodiment, high-precision positioning using a handheld RTK measuring instrument is used to improve the spatial registration accuracy between ground sample points and UAV image pixels; canopy height and field of view parameters are used together to determine the image feature extraction range that matches the ground observation field of view; by comparing models under different scales and different shaped field of view buffer conditions, the optimal inversion model is selected; finally, pixel-by-pixel fine inversion of crop LAI in the study area is achieved. For those skilled in the art, adjustments can be made to the sampling point layout, feature extraction method, and inversion model form without departing from the principle of this invention, and all such adjustments should fall within the protection scope of this invention.

[0078] Example 2

[0079] A UAV-based LAI inversion device using high-precision positioning is used to implement the UAV-based LAI inversion method using high-precision canopy positioning in Example 1, including:

[0080] The preprocessing module is used to implement step 2 in Example 1.

[0081] The crop canopy height calculation module is used to implement step 3 in Example 1.

[0082] The preselected view buffer parameter group construction module is used to implement step 4 in Example 1.

[0083] The sample set construction module is used to implement step 5 in Example 1.

[0084] The filtering module is used to implement step 6 in Example 1.

[0085] The application module is used to implement step 7 in embodiment 1.

[0086] Example 3

[0087] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform steps 2 to 7 of the above embodiment 1.

[0088] Example 4

[0089] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements steps 2 to 7 of Embodiment 1 above.

[0090] Example 5

[0091] A computer program product includes a computer program that, when executed by a processor, implements steps 2 to 7 of Embodiment 1 described above.

[0092] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for LAI inversion of unmanned aerial vehicles (UAVs) based on high-precision canopy positioning, characterized in that, Includes the following steps: Step 1: Acquire UAV images of the study area during the same crop growth period; for each pre-set LAI measurement point in the study area: use a plant canopy analyzer to perform ground LAI measurement to obtain the measured LAI value, and record the coordinates of the LAI measurement point with a handheld RTK measuring instrument; Step 2: The UAV imagery is preprocessed to obtain orthophotos, and the orthophotos and LAI measurement points are located in the same reference coordinate system; Step 3: Generate a digital surface model from the orthophoto. and digital elevation model And calculate the crop canopy height. ; Step 4: Based on crop canopy height The center angle of the field of view of the observation ring of the plant canopy analyzer selected during ground-based LAI measurements. The radius of the preselected field of view buffer corresponding to each observation ring is determined; the directional angle range of the preselected field of view buffer is determined from the actual observation direction angle range during ground LAI measurement; the radius and directional angle range of the preselected field of view buffer constitute the preselected field of view buffer parameter group. Step 5: Using the coordinates of the LAI measurement points as the coordinates of the center of the preselected field of view buffer, based on the parameter group of the preselected field of view buffer, extract the pixels of the preselected field of view buffer from the orthophoto, and extract the preselected feature values ​​of the LAI measurement points from the UAV image features of the pixels in the preselected field of view buffer. Construct a sample set from the LAI measured values ​​and the preselected feature values ​​of the LAI measurement points. Step 6: Based on the sample set and the parameter groups of each pre-selected view buffer, establish the corresponding LAI inversion model, and determine the target inversion model and the corresponding view buffer parameter group according to the preset evaluation index; Step 7: For each pixel of the orthophoto image within the study area, calculate the LAI generation value pixel by pixel based on the target inversion model and the view buffer parameter set to generate the LAI inversion map; Step 5 specifically includes the following steps: Step 5.1: Extract UAV image features of each pixel in the study area from the orthophoto, and select pre-selected features from the UAV image features; Step 5.2: For each LAI measurement point, with the LAI measurement point as the center, delineate the corresponding preselected view buffer in the orthophoto according to the parameter group of each preselected view buffer; extract the values ​​of the preselected features of each pixel in the preselected view buffer in the orthophoto; take the average value of the values ​​of the preselected features of each pixel in the same preselected view buffer under the same preselected feature as the preselected feature value of the corresponding LAI measurement point under the corresponding preselected feature and the preselected view buffer; Step 5.3: Combine different preselected view buffer parameter groups and different preselected feature types in pairs to obtain different preselected buffer-feature setting combinations; For each preselected buffer-feature setting combination: the ground-based measured LAI value and the preselected feature value of the LAI measurement point are paired to form LAI data pairs, and the LAI data pairs of all LAI measurement points in the study area constitute the corresponding sample groups; the sample groups corresponding to all preselected buffer-feature setting combinations constitute the sample set.

2. The UAV LAI inversion method based on high-precision canopy positioning according to claim 1, characterized in that, The preprocessing includes: radiometric calibration, geometric correction, image stitching, orthorectification, and spatial registration.

3. The UAV LAI inversion method based on high-precision canopy positioning according to claim 1, characterized in that, The radius of the preselected view buffer is calculated based on the following formula: ， in, The radius of the preselected view buffer. The height of the crop canopy. The center angle of the field of view of the observation ring.

4. The UAV LAI inversion method based on high-precision canopy positioning according to claim 1, characterized in that, The directional angle range of the preselected field of view buffer is equal to the actual observation directional angle range of the ground-based LAI measurement; When the plant canopy analyzer is equipped with a cover cap to block the field of view in non-target directions during ground LAI measurement, the preselected field of view buffer does not cover the cover cap. When there are no directional constraints during ground-based LAI measurements, the directional angle range of the pre-selected view buffer is 360°.

5. The UAV LAI inversion method based on high-precision canopy positioning according to claim 1, characterized in that, The pre-selected features in step 5 are one or more of the following UAV image features: Difference vegetation index (DVI), super green index (ExG), super green and super red difference index (ExGR), green leaf index (GLI), normalized difference vegetation index (GNDVI), normalized red edge difference index (NDRE), normalized vegetation index (NDVI), normalized green and red difference index (NGRDI), optimized soil-regulated vegetation index (OSAVI), and visible light atmospheric impedance vegetation index (VARI).

6. The UAV LAI inversion method based on high-precision canopy positioning according to claim 1, characterized in that, Step 6 specifically includes the following steps: A LAI inversion model is constructed to fit the mapping relationship between the pre-selected feature values ​​of LAI measurement points and the measured LAI values. The model is fitted using sample groups corresponding to different pre-selected buffer-feature setting combinations to determine the specific parameters of the corresponding LAI inversion model, which are then used as LAI inversion models to be screened. Based on preset evaluation indicators, a target inversion model is selected from the LAI inversion models to be screened, and the pre-selected buffer-feature setting combination corresponding to the target inversion model is the preferred buffer-feature setting combination. The evaluation index is one or more of the following: coefficient of determination R², root mean square error RMSE, and mean absolute error MAE.

7. The UAV LAI inversion method based on high-precision canopy positioning according to claim 1, characterized in that, Step 7 specifically includes the following steps: Each pixel in the orthophoto image within the study area is taken as the current processing point. For each current processing point: the corresponding view buffer is determined by the view buffer parameter group in the preferred buffer-feature setting combination; the preferred feature is the feature type in the preferred buffer-feature setting combination, and the values ​​of the preferred features of each pixel in the view buffer of the corresponding orthophoto image are extracted; the average value of the preferred features of all pixels in the corresponding view buffer is calculated; the average value of the preferred features corresponding to the current processing point is input into the target inversion model to obtain the corresponding LAI generation value. Based on the relative spatial position of each pixel, the generated LAI values ​​are arranged in a raster image to generate an LAI inversion map.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements steps 2 to 7 of the UAV LAI inversion method based on high-precision canopy positioning as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements steps 2 to 7 of the UAV LAI inversion method based on high-precision canopy positioning as described in any one of claims 1 to 7.

Images