Crop LAI calculation method based on multi-angle gap fraction weighting of UAV image

Abstract

The application relates to the technical field of crop yield estimation, in particular to a crop LAI calculation method based on multi-angle gap rate weighting of UAV images, which comprises the following steps: acquiring high-overlap UAV oblique images of a planting area and extracting grid units; performing aerial triangulation on the UAV oblique images and generating DSM, DEM and DOM; extracting multi-angle images of the grid units, calculating image gap rates based on super-green index binarization; optimizing the directly extracted multi-angle gap rates based on a multi-view image gap rate weighting algorithm; and calculating the crop LAI by using the weighted and optimized multi-angle gap rates based on the Beer-Lambert law. The application does not depend on or depends less on measured data, overcomes the uncertainty of scale conversion, fully taps the potential of UAV remote sensing, has high accuracy and great practical application value.

Description

Technical Field

[0001] This application relates to the field of crop yield estimation technology, specifically to a method for calculating crop LAI based on multi-angle gap rate weighting of UAV images. Background Technology

[0002] Leaf Area Index (LAI) refers to half the total leaf area of ​​plants per unit land area. It is an important indicator describing the structure of vegetation canopy and is closely related to physiological activities such as photosynthesis, transpiration, and respiration. As one of the core parameters reflecting vegetation growth status, pathological diagnosis, and crop yield prediction, and especially for flue-cured tobacco, a high-value-added special economic crop whose yield depends almost entirely on leaf weight, accurate extraction of LAI is crucial for flue-cured tobacco production. Therefore, accurately obtaining vegetation LAI has always been a research focus.

[0003] Traditional methods for measuring leaf area index (LAI) are mainly divided into two categories: direct methods and indirect methods, but they are essentially point-based LAI measurement methods. Direct measurement methods are destructive, labor-intensive, and time-consuming. Indirect methods often result in random data, easily overlooking the spatial heterogeneity of vegetation and leading to uncertainty in data accuracy. Furthermore, when quadrat size varies, directly substituting LAI data from smaller quadrats for larger ones introduces significant errors. In addition, data collection is often affected by environmental factors, making it difficult for staff to access the data, and the selected sampling points may not be representative of the entire plot. Therefore, both direct and indirect methods face the problem of enormous workload when acquiring large-scale, spatially continuous LAI data, making it difficult to meet the practical application needs such as large-area vegetation growth monitoring.

[0004] Typically, acquiring large-area LAI data relies on remote sensing technology. Remote sensing-based LAI estimation methods can be divided into empirical model methods and physical model methods. However, these methods are highly dependent on ground-based measured data; the establishment and validation of these models require a sufficient number of high-precision ground samples, making them difficult, time-consuming, and costly. Furthermore, traditional ground-based measurement methods often pre-define the size of the measurement area unit (e.g., 1m×1m, 2m×2m, ...). However, the resolution of satellite remote sensing technology is variable and cannot be directly matched to fixed area units, requiring scale transformation to achieve data scale matching. However, scale transformation may introduce more uncertainties, significantly affecting the accuracy of LAI inversion.

[0005] With the numerous advantages of low-altitude remote sensing technology represented by unmanned aerial vehicles (UAVs), it has been applied to some extent in the field of LAI data acquisition. However, most of the current LAI acquisition methods are similar to traditional remote sensing inversion methods, which means that most research has not yet fully explored the potential of UAV remote sensing technology. Summary of the Invention

[0006] This application provides a method for calculating crop LAI based on multi-angle gap rate weighting of UAV images, in order to solve or partially solve the problems mentioned in the background art.

[0007] This application provides a method for calculating crop LAI based on multi-angle gap rate weighting of UAV images, including the following steps: S100: Acquire UAV tilted images of the planting area with high overlap and extract grid cells; S200: Perform aerial triangulation and generate DSM, DEM, and DOM from UAV oblique images; S300: Extract multi-angle images of grid cells and calculate image gap rate based on super-green index binarization; S400: Optimizes the directly extracted multi-angle gap rate based on a multi-view image gap rate weighting algorithm; S500: Based on the Beer-Lambert law, it calculates crop LAI using weighted optimization of multi-angle gap ratio.

[0008] Preferably, in step S100, during the crop field growth period, a true-color image of the measured farmland is acquired based on an optical camera mounted on a UAV platform. The relative flight altitude of the UAV to the farmland ground is controlled within 50 meters to ensure that the acquired image has a preset tilt angle. In addition, the forward overlap rate and lateral overlap rate of the image are ensured to be higher than 80% and 70%, respectively.

[0009] Preferably, in step S100, the planting area to be measured is divided into spatially continuous grid cells according to a preset size, and each grid cell is regarded as a basic LAI calculation unit.

[0010] Preferably, in step S200, by performing aerial triangulation and generating DSM, DEM, and DOM on the UAV oblique image, the precise interior orientation elements, exterior orientation elements, plane coordinates (X,Y) of the grid cell center point, and elevation value Z of the UAV image are extracted. These parameters are the basis for extracting multi-angle information from the UAV oblique image and can be obtained through the motion recovery structure SfM and multi-view stereo matching (MVS) methods.

[0011] Preferably, in step S300, after obtaining the interior and exterior orientation elements of each UAV image, the center point coordinates of a certain grid cell are marked as (X,Y,Z). Using the collinearity equation in photogrammetry, all UAV images of the complete image of that grid cell can be searched, which is called a multi-angle image. The collinearity equation describes the geometric relationship between the image point, the camera center, and the object point on the same straight line. Using the frame coordinate system as the image-side coordinate system, its expression is as follows: (1) In the formula, ( ) are the coordinates of the image point, ( () represents the coordinates of the principal point in the interior orientation elements. The camera's principal distance (focal length), () represents the coordinates of the photography center. () represents the coordinates of the object point. , , ( =1,2,3) are the nine elements of the rotation matrix derived from the three attitude angles in the image exterior orientation elements; The image coordinates corresponding to the four corner points of a grid cell are judged one by one. If all four image points fall within the range of a UAV image, the UAV image is cropped according to the polygon formed by the image coordinates corresponding to the four corner points. The new image cropped is recorded as the multi-view image of the grid cell. If not, the next image is judged until all UAV images are judged. After acquiring multi-angle images of any grid cell, the center of the grid cell is taken as the origin of the coordinate system. The angle between the line connecting the UAV photography center and the origin (Obj-UAV) and the Z-axis is the observation zenith angle. The angle between the projection of Obj-UAV onto the XY plane and the X-axis is the observation azimuth angle.

[0012] Preferably, in step S300, the specific method for calculating the image gap rate based on the super-green index binarization is as follows: First, calculate the super-green index of the optical image; Then, the histogram of the ultrafiltration exponential feature image is calculated; Next, the minimum inter-class variance method is used to determine the classification threshold of the feature image; Finally, the excess green index feature image is binarized, where the excess green index ExG (Excess Green) of a visible light RGB image is defined as: ExG=2G R B(2) In the formula, G is the green band value, R is the red band gray value, and B is the blue band gray value. For any image in a multi-angle image, the pixels are divided into two categories, leaf and non-leaf, by image binarization classification. The gap ratio g of the image is defined as the proportion of non-leaf pixels to the total number of pixels.

[0013] Preferably, in step S400, the specific steps of the multi-view image gap rate weighted calculation method are as follows: Based on the proximity principle, it is assumed that the vegetation canopy is at the observed zenith angle. If the structures within the range are identical, then the observed zenith angle... Corresponding gap ratio The value can be represented as The average of all known gap ratios within the range is calculated using the following formula: (3) In the formula, This represents the neighborhood range, with values ​​of 1, 3, 5, and 7 respectively. n represents the number of known gap ratios within the neighborhood. Represents the first term within the neighborhood. The gap rate.

[0014] Preferably, in step S500, the step of calculating the crop LAI is as follows: S501: Introducing a multi-angle expression of the Beer-Lambert law, by obtaining the canopy interstitial ratio at multiple zenith angles, the LAI is expressed as an integral of multiple zenith angle directions, with the following expression: (4) In the formula, To observe the zenith angle The average value of the gap ratio in all observed azimuth directions; S502: Discretize formula (4) and... It is expressed as a weighted sum of porosity at a finite number of observed zenith angles, i.e.: (5) In the formula, (unit: The observed zenith angle (range of values) Discretized angular intervals, It is the first time that the zenith angle has been observed. A segmented ring It is the number of dividing rings. It is to observe the zenith angle. The corresponding gap ratio, These are the weights of each segmented ring.

[0015] Preferably, in step S502, in the extreme case where n=1, In step S500, the absolute value of the observed zenith angle needs to be taken.

[0016] Compared with the prior art, the beneficial effects of this application are as follows: This application does not rely on or relies less on ground-measured LAI data when calculating the area region LAI, overcoming the uncertainty of scale transformation caused by the inability to directly obtain area region LAI through point-by-point LAI measurement and the calculation of area region LAI through multi-point LAI. It fully explores the potential of UAV remote sensing, directly obtains area region LAI with high accuracy, and has great practical application value. Attached Figure Description

[0017] The present application will be further described below with reference to the accompanying drawings and embodiments.

[0018] Figure 1 This is a schematic diagram of the observation geometry of the grid cells in this application. Figure 2 This is a schematic diagram of the multi-angle gap ratio distribution of this application. Figure 3 This application provides the study area location and experimental data for embodiments of the present application. Figure 4 The diagram shows the DOM (a) of a cornfield plot in this application and the LAI inversion result (b) obtained using the method of this application. Figure 5 The diagram shows the DOM (a) of a flue-cured tobacco field in this application embodiment and the LAI inversion result (b) obtained using the method of this patent. Figure 6 This is a scatter plot showing the LAI measured by the method of this application and the actual LAI of corn in an embodiment of this application. Figure 7 This is a scatter plot showing the LAI measured by the method of this application and the actual LAI of flue-cured tobacco in the embodiments of this application. Detailed Implementation

[0019] The specification and claims use certain terms to refer to specific components. Those skilled in the art will understand that hardware manufacturers may use different names to refer to the same component. This specification and claims do not distinguish components based on differences in name, but rather on differences in function. The term "comprising" throughout the specification and claims is an open-ended term and should be interpreted as "comprising but not limited to." "Approximately" means that within an acceptable margin of error, those skilled in the art can solve the technical problem and substantially achieve the technical effect within a certain margin of error.

[0020] In the description of this application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "horizontal", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0021] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0022] like Figures 1 to 2 As shown, this application provides a method for calculating crop LAI based on multi-angle gap rate weighting of UAV images, including the following steps: S100: Acquire UAV tilted images of the planting area with high overlap and extract grid cells; S200: Perform aerial triangulation and generate DSM, DEM, and DOM from UAV oblique images; S300: Extract multi-angle images of grid cells and calculate image gap rate based on super-green index binarization; S400: Optimizes the directly extracted multi-angle gap rate based on a multi-view image gap rate weighting algorithm; S500: Based on the Beer-Lambert law, it calculates crop LAI using weighted optimization of multi-angle gap ratio.

[0023] Acquiring UAV oblique images of planting areas with high overlap involves steps such as hardware equipment (including UAV and optical camera) preparation, airspace application, flight route planning and design, aerial photography, and data copying and storage.

[0024] Specifically, in step S100, during the crop growth period in the field, a UAV platform equipped with an optical camera acquires true-color images of the farmland being measured. During flight path planning and design, the relative flight altitude of the UAV to the farmland ground is controlled within 50 meters to ensure high image clarity. With the assistance of the gimbal, the acquired images are ensured to have a preset tilt angle (typically 30°). In addition, the forward overlap rate and lateral overlap rate of the images are ensured to be higher than 80% and 70%, respectively. Such a high image overlap rate can ensure that the same area of ​​the farmland being measured is often repeatedly observed and photographed by the aerial camera from different perspectives, thereby forming multi-angle images of the same area.

[0025] Based on the above method, multi-angle images of the planting area are obtained. The measured planting area is divided into spatially continuous grid units according to a preset size. Each grid unit is regarded as a basic LAI calculation unit. In the following steps, the corresponding multi-angle images are extracted for each grid unit.

[0026] Specifically, in step S200, by performing aerial triangulation and generating DSM, DEM, and DOM from the UAV oblique imagery, precise interior orientation elements (or interior parameters), exterior orientation elements, planar coordinates (X, Y) of the grid cell center points, and elevation values ​​Z are extracted from the UAV imagery. These parameters are the basis for extracting multi-angle information from the UAV oblique imagery and can be obtained through Structure from Motion (SfM) and Multi-View Stereo (MVS) methods. Specific steps include: S201: Import Image and POS Data: Automatically reads the position of each image (GPS / PPK post-processing dynamic differential positioning / RTK real-time dynamic differential positioning), camera attitude angle, and camera intrinsic parameters. If the UAV is equipped with a high-precision POS (Position and Orientation System), aerial triangulation convergence is faster and more accurate. S202: Feature Extraction and Matching: Extract image feature points; perform feature point matching between overlapping images, find corresponding points (connection points), and remove erroneous matching points through random sampling consensus algorithm to improve matching accuracy, construct geometric relationships between images, estimate camera interior and exterior orientation elements, and generate sparse point cloud. This step yields a free network, and the accuracy only satisfies relative position, not absolute geographic accuracy. S203: Regional network adjustment: Based on POS or control points, perform absolute orientation, incorporate the free network into the real geographic coordinate system, and use bundle adjustment to solve for the precise exterior orientation elements (position + attitude angle) of each image, and optimize the image intrinsic parameters; S204: Dense Reconstruction: Depth map estimation is performed using precise image intrinsic / exterior orientation elements and sparse point clouds as guidance, through multi-view depth... Figure 1 Consistency screening, noise reduction, and hole filling ultimately generate a uniform dense point cloud; S205: Generation of DSM, DEM, and DOM: Generate a digital surface model (DSM): interpolate the point cloud into a meshed DSM that includes the heights of all surface features (buildings, trees, etc.); Generate a digital elevation model (DEM): Filter out non-ground points from the DSM to generate a DEM that reflects only the terrain surface; Generating DOM: Based on the DEM, the center projection of each image is converted into a vertical orthographic projection, and the DOM is output through mosaicking and color balancing.

[0027] Based on DOM and DSM, the (X,Y,Z) information of any point within the planting area can be further derived, thereby meeting the extraction requirements of multi-angle image units.

[0028] In the above steps, the method used in steps S201 to S203 is motion recovery structure (SfM), and the method used in step S204 is multi-view stereo matching (MVS).

[0029] Specifically, in step S300, after obtaining the precise intrinsic parameters and exterior orientation elements of each UAV image, the center point coordinates of a certain grid cell are marked as (X,Y,Z). Using the collinearity equation in photogrammetry, all UAV images of the complete image of that grid cell can be searched, which is called "multi-angle image". The collinearity equation describes the geometric relationship between the image point, the camera center, and the object point on the same straight line. Using the frame coordinate system as the image-side coordinate system, its expression is as follows: (1) In the formula, ( ) are the coordinates of the image point, ( () represents the coordinates of the principal point in the interior orientation elements. The camera's principal distance (focal length), () represents the coordinates of the photography center. () represents the coordinates of the object point. , , ( =1, 2, 3) are the nine elements of the rotation matrix derived from the three attitude angles in the image exterior orientation elements. From formula (1), it can be seen that when the principal distance... Point of object ( ) and the photographic center of each UAV image ( Given the information, the coordinates of the image point corresponding to each object point can be obtained. ).

[0030] The image coordinates corresponding to the four corner points of a grid cell are judged one by one. If all four image points fall within the range of a UAV image, the UAV image is cropped according to the polygon formed by the image coordinates corresponding to the four corner points. The new image cropped is recorded as the multi-view image of the grid cell. If not, the next image is judged until all UAV images are judged. The multi-view images of all grid cells can be obtained by following the above process.

[0031] After acquiring multi-angle images of any grid cell, the corresponding observation zenith angle and observation azimuth angle can be calculated based on the geometric relationship between the image's center position and the grid cell's center position. The geometric relationship is as follows: Figure 1 As shown, with the center of the grid cell as the origin, the angle between the line connecting the UAV's photography center and the origin (Obj-UAV) and the Z-axis is the observation zenith angle, and the angle between the projection of Obj-UAV onto the XY plane and the X-axis is the observation azimuth angle. Figure 1 In this context, Obj represents the geographic coordinates of the center of the grid cell. Location of the drone photography center Represents the observed zenith angle. The azimuth angle represents the observation angle, and h is the altitude of the UAV relative to the ground.

[0032] Specifically, in step S300, the method for calculating the image gap rate based on the super-green index binarization is as follows: First, calculate the super-green index of the optical image; Then, the histogram of the ultrafiltration exponential feature image is calculated; Next, the minimum inter-class variance method is used to determine the classification threshold of the feature image; Finally, the excess green index feature image is binarized, where the excess green index ExG (Excess Green) of a visible light RGB image is defined as: ExG=2G R B(2) In the formula, G is the green band value, R is the red band gray value, and B is the blue band gray value.

[0033] For any image in a multi-angle image, the gap ratio corresponding to the image is further calculated by image binarization classification. Specifically, for an image, the pixels are divided into two categories: "leaves" and "non-leaves". This can effectively remove the interference of branches, soil or background components. The gap ratio g of the image is defined as the proportion of non-leaves pixels to the total number of pixels. The Super Green Index ExG has the advantages of highlighting green, suppressing soil / buildings, and improving the separation of vegetation and background.

[0034] The previous step yielded the gap ratio value for each image in a multi-angle image of a single grid cell. However, the spatial distribution between the center point of the grid cell and the corresponding photographic center point of the multi-angle image is discrete, such as... Figure 2 As shown, Figure 2 The result is a schematic diagram of the spatial distribution of the multi-angle image locations extracted using the method in step S300. Figure 2 , Figure 2 The center position is the center point of a certain grid cell. The red pentagram represents the location of the center point. The distance from any point to the origin represents the size of the observed zenith angle, ranging from 0 to 90°. Starting from the right half-axis of the horizontal axis, the radius of counterclockwise rotation and the angle between it and the observed azimuth angle are the observed azimuth angle, ranging from 0 to 360°. The red dot represents the position of the sun.

[0035] from Figure 2 As can be seen, the directly extracted multi-angle gap ratio exhibits significant spatial dispersion, which means that effective data may be lacking at some observation zenith angles, thus limiting the completeness and usability of the method in practical applications. Even if observation data exists within certain zenith angle ranges, its quantity is often limited, and its distribution in the observation azimuth angle is difficult to maintain uniformity, resulting in insufficient data coverage within the four quadrants. Such non-uniform distribution leads to incomplete sampling of the directional information of the vegetation canopy, which may introduce systematic bias and reduce the stability and accuracy of gap ratio inversion. To improve the applicability of the method and the reliability of the results, it is necessary to optimize the originally extracted discretized multi-angle gap ratio.

[0036] Therefore, this application proposes a multi-view image gap rate weighted calculation method in step S400. This method is based on the proximity principle and assumes that the vegetation canopy is at the observation zenith angle. If the structures within the range are identical, then the observed zenith angle... Corresponding gap ratio The value can be represented as The average of all known gap ratios within the range is calculated using the following formula: (3) In the formula, This represents the neighborhood range, with values ​​of 1, 3, 5, and 7 respectively. n represents the number of known gap ratios within the neighborhood. Represents the first term within the neighborhood. The gap rate.

[0037] Based on the weighted calculation of gap ratio in multi-view images, the gap ratio of a certain grid cell within any zenith angle range can be obtained. Substituting this gap ratio into the Beer-Lambert law formula, the leaf area index (LAI) can be further calculated. Specifically, in step S500, the steps for calculating the LAI of crops are as follows: S501: Introducing a multi-angle expression of the Beer-Lambert law, by obtaining the canopy interstitial ratio at multiple zenith angles, the LAI is expressed as an integral of multiple zenith angle directions, with the following expression: (4) In the formula, To observe the zenith angle The average value of the gap ratio in all observed azimuth directions; S502: Discretize formula (4) and... It is expressed as a weighted sum of porosity at a finite number of observed zenith angles, i.e.: (5) In the formula, (unit: The observed zenith angle (range of values) Discretized angular intervals, It is the first time that the zenith angle has been observed. A segmented ring It is the number of dividing rings. It is to observe the zenith angle. The corresponding gap ratio, These are the weights of each segmented ring.

[0038] In step S502, in the extreme case of n=1, ,in addition, Figure 2 The observed zenith angle shown is for illustrative purposes only and is not included in the calculations used in this application. Figure 2 The observed zenith angle needs to be taken as an absolute value.

[0039] The technical solution of this application is tested and verified by referring to specific embodiments below.

[0040] Example 1 First, the general overview of the study area is defined. Located in Fuguanzhuang Town, Yishui County, Linyi City, Shandong Province (36°04′N, 118°56′E), the study area is situated in the hilly region of the Yimeng Mountains in southern central Shandong, covering a total area of ​​approximately 110 square kilometers and comprising over 50 administrative villages. The region has a warm temperate monsoon climate with four distinct seasons: hot and rainy summers, and cold and dry winters. The average annual temperature is around 13–14℃, annual precipitation is approximately 800–900 mm, and annual sunshine hours exceed 2300 hours, indicating abundant solar resources.

[0041] The favorable climate and complex topography of Fuguanzhuang Town have resulted in diverse vegetation types, primarily consisting of farmland landscapes where corn, tobacco, and other grains and cash crops are grown, while also preserving some natural vegetation and planted forests. The relatively flat terrain, abundant crop types, and significant spatial variability provide an ideal experimental environment for conducting research on UAV multi-angle observation and LAI measurement methods.

[0042] The drone data was acquired on August 21, 2025. The sky was clear with some clouds and no strong winds, allowing the drone to conduct stable data collection. The site conditions at the sample plot are as follows: Figure 3 As shown.

[0043] L2 RGB imagery was acquired using a 350 RTK drone at an altitude of 30m and a ground resolution of 1cm / pixel. Each image covered an area of ​​approximately 25m × 25m, with a forward overlap of 80% and a lateral overlap of 70%. Due to the L2 camera's 84° field of view, the maximum zenith angle of data acquired using traditional orthophoto flight was only 42°, and the data was limited. To obtain a larger quantity and wider range of multi-angle observation data to meet experimental needs, oblique photography was used for RGB image acquisition. The gimbal was tilted by 30°, and oblique images of the study area were acquired using a tic-tac-toe flight pattern. This method allowed for a maximum zenith angle of 72° in the acquired images.

[0044] The measured LAI (Label Area Index) was obtained from 16:00 to 17:00 on the same day the UAV flight ended. 1m × 1m quadrats were set up in corn and tobacco fields. LAI sampling was performed using the LAI-2200C instrument, and GPS was used to measure the location of the quadrats, obtaining the geographic coordinates of the four corner points. The corn and tobacco quadrats were randomly and evenly distributed within the fields to ensure that the obtained LAI values ​​reflected the crop growth status of the entire farmland. During the measurement process, a 180° shading cap was used to avoid errors caused by the LAI-2200C instrument's measurement range exceeding the quadrat boundaries.

[0045] The thematic maps of the spatial distribution of LAI (Laminated Area Index) for maize and flue-cured tobacco predicted using the method proposed in this application are as follows: Figure 4 , Figure 5As shown, the spatial resolution of the image in the thematic map is 1.0m, which is consistent with the grid cell size of the LAI measured in the field.

[0046] Depend on Figure 4 It can be seen that the LAI inversion value of maize is distributed between 0.00 and 6.32, and the LAI inversion value of flue-cured tobacco is distributed between 0.00 and 7.12. The visual effect of the two LAI spatial distribution thematic maps shows that the size of LAI is completely consistent with the spatial distribution trend of crop growth.

[0047] To verify the effectiveness of the method in this application, Figure 6 , Figure 7 The comparison between LAI data of maize and tobacco measured using the method proposed in this patent and field-measured LAI data is presented. Figure 6 , Figure 7 The horizontal axis represents the LAI value measured in the field, the vertical axis represents the LAI value measured by the method proposed in this invention, and the black line represents the 1:1 line.

[0048] The experiment used root mean square error (RMSE) and relative root mean square error (rRMSE) as accuracy indicators to quantitatively evaluate the difference between LAI measurement results and reference data, in order to achieve the research objective of analyzing the effectiveness and accuracy of the method. The calculation formulas for both are as follows: (6) In the formula, Represents the measured value of LAI. The predicted value represents the method of this invention. Represents the number of samples. This represents the average value of the measured LAI values. The value ranges from 0.00 to positive infinity; the smaller the value, the smaller the error between the calculated result and the true value. This represents the percentage of error in the measured data, ranging from 0.00% to 100.00%. The smaller the value, the smaller the error.

[0049] Experimental results show that the LAI measured by this method for both corn and tobacco exhibits a consistent trend with the field measurement obtained using the LAI-22004C instrument, demonstrating a high correlation. The RMSE values ​​for corn and flue-cured tobacco are 0.46 and 0.60 m² / m², respectively, while the rRMSE values ​​for corn and tobacco are 17.16% and 18.51%, respectively. This indicates that the LAI measurement method proposed in this patent application has high accuracy and significant practical application value.

[0050] The embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of this application.

Claims

1. A method for calculating crop LAI based on multi-angle gap rate weighting of UAV images, characterized in that, Includes the following steps: S100: Acquire UAV tilted images of the planting area with high overlap and extract grid cells; S200: Perform aerial triangulation and generate DSM, DEM, and DOM from UAV oblique images; S300: Extract multi-angle images of grid cells and calculate image gap rate based on super-green index binarization; S400: Optimizes the directly extracted multi-angle gap rate based on a multi-view image gap rate weighting algorithm; S500: Based on the Beer-Lambert law, it calculates crop LAI using weighted optimization of multi-angle gap ratio.

2. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 1, characterized in that: In step S100, during the crop field growth period, a UAV platform equipped with an optical camera acquires true-color images of the measured farmland. The relative flight altitude of the UAV to the farmland ground is controlled within 50 meters to ensure that the acquired images have a preset tilt angle. In addition, the forward overlap rate and lateral overlap rate of the images are ensured to be higher than 80% and 70%, respectively.

3. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 1, characterized in that: In step S100, the planting area to be measured is divided into spatially continuous grid cells according to a preset size, and each grid cell is regarded as a basic LAI calculation unit.

4. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 1, characterized in that: In step S200, by performing aerial triangulation and generating DSM, DEM, and DOM on the UAV oblique image, the precise interior orientation elements, exterior orientation elements, plane coordinates (X,Y) of the grid cell center point, and elevation value Z of the UAV image are extracted. These parameters are the basis for extracting multi-angle information from the UAV oblique image and can be obtained through the motion recovery structure SfM and multi-view stereo matching MVS method.

5. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 4, characterized in that: In step S300, after obtaining the interior and exterior orientation elements of each UAV image, the center point coordinates of a certain grid cell are marked as (X,Y,Z). Using the collinearity equation in photogrammetry, all UAV images of the complete image of that grid cell can be searched, which is called a multi-angle image. The collinearity equation describes the geometric relationship between the image point, the camera center, and the object point on the same straight line. Using the frame coordinate system as the image-side coordinate system, its expression is as follows: （1） In the formula, ( ) are the coordinates of the image point, ( () represents the coordinates of the principal point in the interior orientation elements. The camera's principal distance (focal length), () represents the coordinates of the photography center. () represents the coordinates of the object point. , , ( =1,2,3) are the nine elements of the rotation matrix derived from the three attitude angles in the image exterior orientation elements; The image coordinates corresponding to the four corner points of a grid cell are judged one by one. If all four image points fall within the range of a UAV image, the UAV image is cropped according to the polygon formed by the image coordinates corresponding to the four corner points. The new image cropped is recorded as the multi-view image of the grid cell. If not, the next image is judged until all UAV images are judged. After acquiring multi-angle images of any grid cell, the center of the grid cell is taken as the origin of the coordinate system. The angle between the line connecting the UAV photography center and the origin (Obj-UAV) and the Z-axis is the observation zenith angle. The angle between the projection of Obj-UAV onto the XY plane and the X-axis is the observation azimuth angle.

6. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 1, characterized in that: In step S300, the specific method for calculating the image gap rate based on the super-green index binarization is as follows: First, calculate the super-green index of the optical image; Then, the histogram of the ultrafiltration exponential feature image is calculated; Next, the minimum inter-class variance method is used to determine the classification threshold of the feature image; Finally, the excess green index feature image is binarized, where the excess green index ExG (Excess Green) of a visible light RGB image is defined as: ExG=2G R B(2) In the formula, G is the green band value, R is the red band gray value, and B is the blue band gray value. For any image in a multi-angle image, the pixels are divided into two categories, leaf and non-leaf, by image binarization classification. The gap ratio g of the image is defined as the proportion of non-leaf pixels to the total number of pixels.

7. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 1, characterized in that: In step S400, the specific steps of the multi-view image gap rate weighted calculation method are as follows: Based on the proximity principle, it is assumed that the vegetation canopy is at the observed zenith angle. If the structures within the range are identical, then the observed zenith angle... Corresponding gap ratio The value can be represented as The average of all known gap ratios within the range is calculated using the following formula: （3） In the formula, This represents the neighborhood range, with values ​​of 1, 3, 5, and 7 respectively. n represents the number of known gap ratios within the neighborhood. Represents the first term within the neighborhood. The gap rate.

8. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 1, characterized in that: In step S500, the steps for calculating the crop LAI are as follows: S501: Introducing a multi-angle expression of the Beer-Lambert law, by obtaining the canopy interstitial ratio at multiple zenith angles, the LAI is expressed as an integral of multiple zenith angle directions, with the following expression: （4） In the formula, To observe the zenith angle The average value of the gap ratio in all observed azimuth directions; S502: Discretize formula (4) and... It is expressed as a weighted sum of porosity at a finite number of observed zenith angles, i.e.: （5） In the formula, (unit: The observed zenith angle (range of values) Discretized angular intervals, It is the first time that the zenith angle has been observed. A segmented ring It is the number of dividing rings. It is to observe the zenith angle. The corresponding gap ratio, These are the weights of each segmented ring.

9. The method for calculating crop LAI based on multi-angle gap rate weighting of UAV images according to claim 8, characterized in that: In step S502, in the extreme case where n=1, In step S500, the absolute value of the observed zenith angle needs to be taken.

Images