Unmanned aerial vehicle spectral image BRDF rapid acquisition and adaptive generalization modeling method

Abstract

The invention relates to a modeling method, in particular to an unmanned aerial vehicle spectral image BRDF rapid acquisition and adaptive generalization modeling method. The method comprises the following steps: 1, determining a research area, obtaining multi-angle BRDF data of the research area, and obtaining illumination change data at the same time; 2, preprocessing the acquired multi-angle BRDF data in combination with illumination change data to obtain preprocessed multi-angle images, splicing the multi-angle images to generate a panorama of the research area, and acquiring a digital surface model graph at the same time; 3, carrying out pixel-by-pixel classification on the research area based on the panorama to obtain a ground feature category map; calculating the gradient and the slope direction of the research area pixel by pixel based on the digital surface model graph; 4, correcting a preset BRDF model by using the acquired illumination change data, gradient and slope direction; and 5, self-adaptive generalization modeling is carried out. According to the method, high-efficiency BRDF acquisition can be realized, and the influence of regional change, space environment and topographic factors on BRDF modeling precision can be eliminated.

Description

Technical Field

[0001] The present invention relates to a modeling method, and in particular to a method for rapid acquisition and adaptive generalization modeling of BRDF of unmanned aerial vehicle (UAV) spectral images. Background Art

[0002] Rapid acquisition systems and adaptive generalized modeling of BRDF (bidirectional reflectance distribution function) data from unmanned aerial vehicle (UAV) spectral imagery are important research areas in remote sensing quantification, offering significant technical advantages and application potential in real-world scenarios. Traditional BRDF acquisition methods often rely on fixed ground equipment, resulting in low acquisition efficiency, poor scene adaptability, and insufficient dynamic monitoring capabilities, limiting their practical application. Unmanned aerial vehicle (UAV) platforms offer a novel solution for BRDF acquisition. First, UAVs' maneuverability enables multi-angle observation in complex terrain and hard-to-reach areas, significantly improving data collection coverage and efficiency. Second, UAVs equipped with hyperspectral or multispectral sensors, combined with push-broom or frame-type imaging techniques, can rapidly acquire high-precision surface optical property data, meeting the needs of real-time monitoring in dynamic scenes. Furthermore, UAV platforms equipped with GPS and inertial navigation systems enable high-precision spatial positioning and attitude control, ensuring the consistency and accuracy of multi-angle data. Leveraging these advantages, designing a general, efficient UAV BRDF acquisition solution for complex real-world scenarios is a pressing issue.

[0003] In terms of BRDF modeling, there is a lack of systematic specifications for the scale of the modeling area. Moreover, when the scale changes, the spatial environmental factors and terrain factors in the area become more complex, and the applicability of the obtained BRDF distribution in the new area will deteriorate, thus limiting the application of this technology in large-scale areas. This requires solving the problem of high-precision modeling of BRDF when the area changes, and at the same time correcting the spatial factors that affect the accuracy of BRDF during the modeling process to ensure that when BRDF is applied in a large area, the obtained BRDF distribution can best reflect the real reflectivity characteristics of the ground objects. Summary of the Invention

[0004] The purpose of the present invention is to solve the technical problems of low efficiency of existing BRDF acquisition, poor adaptability of BRDF distribution when the region changes, and the influence of spatial environment and terrain factors on the accuracy of BRDF modeling, and to provide a method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A method for fast acquisition and adaptive generalization modeling of BRDF of UAV spectral images is special in that it includes the following steps:

[0007] Step 1: Determine the study area, obtain multi-angle BRDF data of the study area, and obtain illumination change data at the same time;

[0008] Step 2: Preprocess the acquired multi-angle BRDF data in combination with the illumination change data to obtain preprocessed multi-angle images. Then, a panoramic view of the study area is generated based on the multi-angle images, and a digital surface model is obtained at the same time.

[0009] Step 3: Classify the study area pixel by pixel based on the panoramic image to obtain a feature classification map; at the same time, calculate the slope and aspect of the study area pixel by pixel based on the digital surface model map;

[0010] Step 4: Use the acquired illumination change data, slope and aspect to calibrate the preset BRDF model;

[0011] Step 5: Adaptive generalization modeling;

[0012] Step 5.1, perform BRDF modeling on the 1×1 pixel in the study area based on the corrected BRDF model to obtain the 1×1 pixel BRDF model;

[0013] Step 5.2: Based on the 1×1 pixel BRDF model, the feature classification map, slope, and aspect, calculate the total gradient through the adaptive diffusion window and determine the diffusion direction of the 1×1 pixel BRDF model. Determine whether the total gradient meets the threshold condition. If so, diffuse the 1×1 pixel BRDF model along the diffusion direction to 2×2 pixels to obtain the 2×2 pixel BRDF model and execute step 5.3. If not, execute step 5.4.

[0014] Step 5.3: Based on the 2×2 pixel BRDF model, obtain the 3×3 pixel BRDF model in the same way as step 5.2; and so on, until the total gradient does not meet the threshold condition and the diffusion is terminated;

[0015] Step 5.4: Obtain the BRDF model of n×n pixels when the diffusion is terminated, where n is an integer greater than or equal to 1, to complete the adaptive generalization modeling.

[0016] Furthermore, in step 3, the slope and aspect are calculated using the following formula:

[0017]

[0018] Where: Slope is the slope, Z x is the rate of change of elevation in the east-west direction, Z y is the rate of elevation change in the north-south direction, and Aspect is the slope direction.

[0019] Furthermore, in step 4, the preset BRDF model is corrected using the following formula:

[0020] R Corrected (θ i ,θ v ,φ,λ)=I(t,λ)·T(slope,aspect)·R(θ i ,θ v ,φ,λ);

[0021] Where: θ i ,θ v is the solar zenith angle and the observation zenith angle, φ is the relative azimuth angle between the solar azimuth angle and the observation azimuth angle, λ is the corresponding band, t is the corresponding time, R(θ i ,θ v ,φ,λ) is the original BRDF model, I(t,λ) is the illumination correction factor, T(slope,aspect) is the terrain correction factor, R Corrected (θ i ,θ v ,φ,λ) is the corrected BRDF model.

[0022] Furthermore, in step 5, the calculation equation of the total gradient is:

[0023]

[0024] Where: G Total is the total gradient, G C is the category matrix, G S is the slope matrix, G A is the aspect matrix, i, j and k represent the weights of the three matrices respectively;

[0025] The threshold condition Dec(G Total )for:

[0026]

[0027] Where: w is the threshold.

[0028] Furthermore, step one is specifically as follows: determine the study area, use a hyperspectral or multispectral camera on a drone, and use a multi-rectangular nested flight method to obtain multi-angle BRDF data of the study area, while using downlink irradiance monitoring equipment to obtain light change data.

[0029] Furthermore, in step 2, the preprocessing includes atmospheric correction, radiation correction and geometric correction.

[0030] Furthermore, step 2 also includes: performing pixel-by-pixel registration of the acquired multi-angle images with the panoramic image.

[0031] Furthermore, step three also includes: matching the illumination change data in the time dimension and the spectrum dimension, ensuring that the illumination change data are at the same moment in the time dimension and the same wavelength band in the spectrum dimension.

[0032] Furthermore, in step three, SVM is used to classify the study area pixel by pixel based on the panoramic image using a machine learning algorithm to obtain a feature category map.

[0033] The beneficial effects of the present invention are:

[0034] The present invention is based on unmanned aerial vehicle (UAV) and irradiance monitoring equipment. It addresses the problems of low efficiency of traditional BRDF acquisition, the impact of spatial environment and terrain factors on BRDF modeling accuracy, and poor adaptability of BRDF models due to regional scale changes. It realizes the rapid acquisition of multi-angle information of the target area, and simultaneously corrects the BRDF model using illumination changes and terrain distribution to obtain a more accurate BRDF distribution of ground objects.

[0035] At the same time, based on the window-adaptive generalization theory, the BRDF distribution of a single pixel is generalized to the largest optimal observation area for the same type of ground feature. This invention leverages the advantages of drones, such as flexibility, complex scene acquisition, wide coverage, and high acquisition efficiency, to provide rich data for BRDF modeling. The modeling layer fully considers spatial environmental factors and terrain factors to ensure high model accuracy. At the same time, the regional adaptive generalization theory further generalizes the BRDF distribution of a single pixel, significantly promoting the application of high-precision BRDF models across large areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of an embodiment of the method for rapid acquisition and adaptive generalized modeling of BRDF based on UAV spectral images of the present invention;

[0037] Figure 2 a is a schematic diagram of a nested flight acquisition scheme carried out by a UAV in an embodiment of the present invention, and b is a schematic diagram of a multi-rectangle nested flight scheme;

[0038] Figure 3 a is the category map of the study area in the embodiment of the present invention, b is the DSM of the study area generated in the embodiment of the present invention, c is the slope map of the study area in the embodiment of the present invention, and d is the aspect map of the study area in the embodiment of the present invention. DETAILED DESCRIPTION

[0039] To make the objectives, advantages, and features of the present invention more clear, the following is a further detailed description of a method for rapid acquisition and adaptive generalization modeling of BRDF for drone spectral images proposed by the present invention, in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent based on the following specific embodiments.

[0040] See also Figure 1 This embodiment provides a method for fast acquisition and adaptive generalization modeling of BRDF of UAV spectral images, and the specific implementation method includes the following steps:

[0041] (1) Determine the research area, use a drone-mounted high / multispectral image, and use a multi-rectangular nested flight plan to obtain multi-angle BRDF data of the area, such as Figure 2 As shown, downlink irradiance monitoring equipment is used simultaneously to monitor solar illumination changes. The landforms in the study area in this example are primarily grassland, water, bare land, trees, and concrete roads. The average surface elevation is approximately 400.0 m, with regional elevations ranging from 392.2 to 415.0 m, and an elevation difference of 22.8 m.

[0042] (2) The multi-angle BRDF data obtained are combined with the illumination change data to obtain pre-processed multi-angle images through atmospheric correction, radiation correction, and geometric correction. The panoramic view of the study area is generated based on the multi-angle image stitching, and the DSM (digital surface model) is generated at the same time. Figure 3 Finally, each multi-angle image is registered pixel by pixel with the panoramic image based on the characteristics of the ground features such as camera pose and observation geometry.

[0043] (3) Using machine learning algorithms such as SVM (support vector machine) and random forest to classify the study area pixel by pixel based on the panoramic image, the ground feature category map is obtained. Specifically, according to the distribution characteristics of the ground features in the study area, the categories are divided into grassland, water, trees, bare land, roads, and artificially paved 0.4 reflectivity standard targets, such as Figure 3 As shown in a. The acquired illumination change data is strictly matched in the time dimension and the spectral dimension, ensuring that the time dimension is at the same moment and the spectral dimension is at the same band. Using the generated DSM map, the slope (Slope) and aspect (Aspect) of the study area are calculated pixel by pixel as shown in Figure 3 c. Figure 3 d, the calculation formula is as follows:

[0044]

[0045] Where Z x is the rate of change of elevation in the east-west direction, Z y is the rate of change of elevation in the north-south direction.

[0046] (4) Use the acquired illumination changes and slope and aspect to calibrate the BRDF model. Specifically, in the time and spectral dimensions, illumination correction matches the illumination changes acquired by the downlink irradiance measurement device with the multi-angle images acquired by the drone's high / multi-spectral imagery to ensure that they are at the same time and in the same wavelength band. The slope and aspect correction is the correction of solar geometry and observation geometry. The values ​​of the two corrections are determined by the actual acquired data, and the calculation formula is as follows:

[0047] R Corrected (θ i ,θ v ,φ,λ)=I(t,λ)·T(slope,aspect)·R(θ i ,θ v ,φ,λ)

[0048] Where θ i ,θ v is the solar zenith angle and the observation zenith angle, φ is the relative azimuth angle between the solar azimuth angle and the observation azimuth angle, λ is the corresponding band, t is the corresponding time, R(θ i ,θ v ,φ,λ) is the original BRDF model, I(t,λ) is the illumination correction factor, T(slope,aspect) is the terrain correction factor, R Corrected (θ i ,θ v ,φ,λ) is the corrected BRDF model.

[0049] (5) Adaptive generalization modeling

[0050] Step 5.1: Perform BRDF modeling on the 1×1 pixel in the study area based on the corrected BRDF model to obtain a 1×1 pixel BRDF model.

[0051] Step 5.2: Based on the 1×1 pixel BRDF model, the feature category map, slope, and aspect, calculate the total gradient through the adaptive diffusion window to determine the diffusion direction of the 1×1 pixel BRDF model; determine whether the total gradient meets the threshold condition. If so, diffuse the 1×1 pixel BRDF model to 2×2 pixels along the diffusion direction to obtain the 2×2 pixel BRDF model, and execute step 5.3; if not, execute step 5.4.

[0052] Step 5.3: Based on the 2×2 pixel BRDF model, obtain the 3×3 pixel BRDF model in the same way as step 5.2; and so on, until the total gradient does not meet the threshold condition and the diffusion is terminated.

[0053] Step 5.4: Obtain the BRDF model of n×n pixels when the diffusion is terminated, where n is an integer greater than or equal to 1, to complete the adaptive generalization modeling.

[0054] The specific theory of adaptive diffusion window is as follows:

[0055] For a pixel's category C(x,y), the distribution of its surrounding pixels is as follows:

[0056]

[0057] Similarly, the slope S(x, y) and the slope A(x, y) of the pixel have this distribution. The gradient of the central pixel C(x, y) of the category matrix diffused in each direction is calculated as follows:

[0058]

[0059] In the x-direction, there are x-up and x-down. In the y-direction, there are y-left and y-right. From the top left corner to the bottom right corner, defined as d1, the directions are d1-up-left and d1-down-right. From the bottom left corner to the top right corner, defined as d2, the directions are d2-down-left and d2-up-right. As the region dimension increases, the same method is used to calculate the gradient from the central region to the surrounding pixels.

[0060] Similarly, the slope matrix G S , the aspect matrix G A Similar gradient calculations are also performed, and the final gradient sum is as follows, where i, j, and k represent the weights of the three gradients respectively.

[0061]

[0062] Whether the model will eventually spread to higher latitudes is expressed as follows: where w represents the threshold, and the threshold is set based on the fitting error.

[0063]

[0064] The diffusion direction is selected according to the size of the total gradient. For example, there are two types of up and down in the x direction, two types of left and right in the y direction, two types of upper left and lower right in the d1 direction, and two types of lower left and upper right in the d2 direction.

[0065] For example, the current area contains 1×1 pixels. When selecting a direction, the category gradient, slope gradient, and direction gradient of this 1×1 area in all surrounding directions are calculated. Then, the sum of the gradients in all directions is sorted, and finally the ones with the smallest gradient change values ​​are selected as the diffusion directions. The number of directions selected is based on ensuring that the area after diffusion is an n×n square area.

Claims

1. A method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images, characterized by: The following steps are involved: Step 1: Determine the study area, obtain multi-angle BRDF data of the study area, and obtain illumination change data at the same time; Step 2: Preprocess the acquired multi-angle BRDF data in combination with the illumination change data to obtain preprocessed multi-angle images. Then, a panoramic view of the study area is generated based on the multi-angle images, and a digital surface model is obtained at the same time. Step 3: Classify the study area pixel by pixel based on the panoramic image to obtain a feature classification map; at the same time, calculate the slope and aspect of the study area pixel by pixel based on the digital surface model map; Step 4: Use the acquired illumination change data, slope and aspect to calibrate the preset BRDF model; Step 5: Adaptive generalization modeling; Step 5.1, perform BRDF modeling on the 1×1 pixel in the study area based on the corrected BRDF model to obtain the 1×1 pixel BRDF model; Step 5.2: Based on the 1×1 pixel BRDF model, the feature classification map, slope, and aspect, calculate the total gradient through the adaptive diffusion window and determine the diffusion direction of the 1×1 pixel BRDF model. Determine whether the total gradient meets the threshold condition. If so, diffuse the 1×1 pixel BRDF model along the diffusion direction to 2×2 pixels to obtain the 2×2 pixel BRDF model and execute step 5.

3. If not, execute step 5.

4. Step 5.3: Based on the 2×2 pixel BRDF model, obtain the 3×3 pixel BRDF model in the same way as step 5.2; and so on, until the total gradient does not meet the threshold condition and the diffusion is terminated; Step 5.4: Obtain the BRDF model of n×n pixels when the diffusion is terminated, where n is an integer greater than or equal to 1, to complete the adaptive generalization modeling.

2. The method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images according to claim 1 is characterized in that: In step 3, the slope and aspect are calculated using the following formula: Where: Slope is the slope, Z x is the rate of change of elevation in the east-west direction, Z y is the rate of elevation change in the north-south direction, and Aspect is the slope direction.

3. The method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images according to claim 2 is characterized in that: In step 4, the preset BRDF model is corrected using the following formula: R Corrected (i i ,i v ,φ,λ)=I(t,λ)·T(slope,aspect)·R(θ i ,i v ,f,l); Where: θ i ,θ v is the solar zenith angle and the observation zenith angle, φ is the relative azimuth angle between the solar azimuth angle and the observation azimuth angle, λ is the corresponding band, t is the corresponding time, R(θ i ,θ v ,φ,λ) is the original BRDF model, I(t,λ) is the illumination correction factor, T(slope,aspect) is the terrain correction factor, R Corrected (θ i ,θ v ,φ,λ) is the corrected BRDF model.

4. The method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images according to claim 3 is characterized in that: In step 5, the calculation equation of the total gradient is: Where: G Total is the total gradient, G C is the category matrix, G S is the slope matrix, G A is the aspect matrix, i, j and k represent the weights of the three matrices respectively; The threshold condition Dec(G Total )for: Where: w is the threshold.

5. The method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images according to any one of claims 1 to 4, characterized in that: Step 1 is as follows: Determine the study area, use a hyperspectral or multispectral camera on a drone, and use a multi-rectangular nested flight method to obtain multi-angle BRDF data of the study area. At the same time, use downlink irradiance monitoring equipment to obtain light change data.

6. The method for rapid acquisition and adaptive generalization modeling of BRDF of unmanned aerial vehicle spectral images according to claim 5 is characterized by: In step 2, the preprocessing includes atmospheric correction, radiation correction and geometric correction.

7. The method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images according to claim 6 is characterized in that: Step 2 also includes: The acquired multi-angle images are registered pixel by pixel with the panoramic image.

8. The method for rapid acquisition and adaptive generalization modeling of BRDF of UAV spectral images according to claim 7 is characterized in that: Step three also includes: Match the illumination change data in the time dimension and the spectral dimension, ensuring that they are at the same moment in the time dimension and the same band in the spectral dimension.

9. The method for rapid acquisition and adaptive generalization modeling of BRDF of unmanned aerial vehicle spectral images according to claim 8, characterized in that: In step three, SVM is used to classify the study area pixel by pixel based on the panoramic image using a machine learning algorithm to obtain a feature category map.

Images