Fine surface albedo acquisition method and system for hyper-spectrum unmanned aerial vehicle remote sensing

Abstract

The invention discloses a fine earth surface albedo acquisition method and system for hyper-spectrum unmanned aerial vehicle remote sensing, and the method comprises the steps: shooting an earth surface feature image corresponding to each pixel through a visible imaging lens fixed on an unmanned aerial vehicle load, and carrying out the preprocessing, adaptive cutting and semantic segmentation of each earth surface feature image; and according to a semantic segmentation result and in combination with the earth surface albedo of each earth surface type, carrying out weighted calculation on the earth surface feature image to obtain a fine earth surface albedo of the earth surface feature image. Therefore, the earth surface albedo corresponding to each pixel observed by the unmanned aerial vehicle can be quickly and accurately obtained.

Description

Technical Field

[0001] The present invention belongs to the technical field of environmental remote sensing, and in particular relates to a method and system for obtaining fine surface albedo for hyperspectral unmanned aerial vehicle remote sensing. Background Art

[0002] Satellite remote sensing has a large temporal and spatial resolution, and vehicle-mounted monitoring instruments cannot enter the factory area for monitoring. UAV spectral remote sensing can overcome technical problems such as low temporal and spatial resolution, monitoring blind spots, and difficulty in identifying pollution sources.

[0003] A major technical challenge for UAV spectral remote sensing in atmospheric monitoring is the rapid changes in the surface. Changes in surface type directly affect surface reflectivity, which is a key parameter in remote sensing data processing. Different surface types (such as water bodies, forests, grasslands, urban buildings, etc.) have different reflective characteristics. If the surface type changes rapidly during the flight of the UAV, the change in reflectivity will cause significant changes in the spectral data received by the remote sensing sensor.

[0004] Surface albedo is a key parameter for calculating the atmospheric mass factor (AMF) in the radiation transfer model. Accurate surface albedo is particularly important for processing spectral data. If the surface albedo is not correctly corrected, the radiation calculation in the radiation transfer model will be affected, resulting in errors in key parameters such as the radiation output and the atmospheric mass factor (AMF) of the radiation transfer model. This error will directly affect the accuracy of environmental monitoring data such as pollutant concentration and gas distribution. Summary of the invention

[0005] In view of the above, an object of the present invention is to provide a method and system for obtaining fine surface albedo for hyperspectral UAV remote sensing, so as to quickly and accurately obtain the surface albedo corresponding to each pixel observed by the UAV.

[0006] To achieve the above-mentioned purpose of the invention, an embodiment provides a method for obtaining fine surface albedo for hyperspectral UAV remote sensing, comprising the following steps:

[0007] The visible imaging lens fixed on the UAV payload captures the surface feature image corresponding to each pixel, and performs preprocessing, adaptive cropping and semantic segmentation on each surface feature image;

[0008] According to the semantic segmentation results and combined with the surface albedo of each surface type, a weighted calculation of the surface feature image is performed to obtain the refined surface albedo of the surface feature image.

[0009] Preferably, the visible imaging lens is fixed on the same axis as the telescope module on the UAV payload in a coaxial fixing manner, so that the center of the photo taken by the visible imaging lens coincides with the center of the telescope module, while ensuring that the telescope elevation angle and the visible imaging lens rotate synchronously, ensuring that each surface feature image taken by the visible imaging lens corresponds one-to-one to the pixel of each elevation angle of the telescope.

[0010] Preferably, each surface feature image is preprocessed, including:

[0011] The surface feature image is geometrically corrected based on the random sampling consistency algorithm and the automatic matching algorithm of homologous points to eliminate the influence of geometric deformation on the surface albedo in the surface feature image.

[0012] Preferably, each surface feature image is adaptively cropped, including:

[0013] Field of view angle θ using visible imaging lens camera and the field of view angle θ of the telescope module telescope The image width W of the visible imaging lens is calculated by the ratio of camera and the telescope's image width W telescope The ratio is:

[0014]

[0015] The surface feature images are adaptively cropped based on the width ratio to ensure that each image taken by the visible imaging lens fully corresponds to each pixel of the telescope.

[0016] Preferably, the field of view angle θ of the telescope module is telescope It is generally a fixed value, and its size depends on the optical fiber and optical lens used. It can be seen that the field of view angle θ of the imaging lens camera Calculated by:

[0017]

[0018] Among them, h is the height of the drone from the ground, α is the inclination angle between the telescope and the horizontal direction, and W pixel and H pixel are the pixel width and height of the surface feature image taken, W image and H image are the width and height of the camera sensor, θ horizonrtal and θ vertical Represents the field of view of the visible imaging lens in the horizontal and vertical directions. Since the change in the field of view caused by the change in the focal length of the visible imaging lens is linear, θ is selected. horizonrtal As the field of view of the visible imaging lens, denoted by θ camera .

[0019] Preferably, semantic segmentation is performed on each surface feature image, including:

[0020] A deep convolutional neural network model is used to segment the adaptively cropped surface feature image, and the model output is the segmentation mask M. i (x,y), where M i (x,y) represents the surface type i at pixel (x,y).

[0021] Preferably, the deep convolutional neural network model adopts a SegNet model.

[0022] Preferably, weighted calculation of the surface feature image is performed based on the semantic segmentation result and combined with the surface albedo of each surface type, including:

[0023] For each pixel, use A i Represents the surface albedo of surface type i, based on the surface albedo A i The weighted calculation of the surface feature image is expressed as:

[0024]

[0025] Among them, S pixel represents the actual ground area of ​​each pixel, n represents the total number of surface types, Albedo represents the weighted fine surface albedo, S i Represents the area of ​​surface type i, and the calculation formula is:

[0026]

[0027] Where W and H are the width and height of the surface feature image, respectively.

[0028] Preferably, the method further comprises: obtaining fine surface albedo of different central wavelengths through conversion coefficients, comprising:

[0029]

[0030] Albedo'=f×Albedo

[0031] Where f is the conversion coefficient, k is an exponential variation coefficient related to the type or material of the feature, and the coefficient k determines the rate of change of albedo with wavelength, reflecting the reflection characteristics of the surface material at different wavelengths; λ1 and λ2 are two central wavelengths, Albedo' represents the fine surface albedo at the central wavelength λ2, and Albedo represents the fine surface albedo at the central wavelength λ1.

[0032] To achieve the above-mentioned purpose of the invention, the embodiment also provides a fine surface albedo acquisition system for hyperspectral UAV remote sensing, including a UAV payload with a visible imaging lens and a telescope module, and a computing module.

[0033] The UAV payload captures the surface feature image corresponding to each pixel through the visible imaging lens on it;

[0034] The calculation module performs preprocessing, adaptive cropping and semantic segmentation on each surface feature image taken, and then performs weighted calculation of the surface feature image according to the semantic segmentation result and in combination with the surface albedo of each surface type to obtain the fine surface albedo of the surface feature image.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention adopts a combination of software and hardware. For the surface feature image corresponding to each pixel captured by the hardware visible imaging lens, the surface feature image is preprocessed, adaptively cropped and semantically segmented through a specific software algorithm. Then, weighted calculation of the surface feature image is performed according to the semantic segmentation result and in combination with the surface albedo of each surface type to obtain the fine surface albedo of the surface feature image. In this way, the surface albedo of each observation pixel in the entire target area can be finely portrayed. Compared with the traditional single surface albedo, the observation result will be more accurate and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 is a flow chart of a method for obtaining fine surface albedo for hyperspectral UAV remote sensing provided by an embodiment;

[0039] Figure 2 It is a schematic diagram of an image of a surface feature captured by a drone observation and a visible imaging lens provided in an embodiment;

[0040] Figure 3 is a schematic diagram of adaptive cropping provided by an embodiment;

[0041] Figure 4 The embodiment of the present invention provides a surface albedo result of a corresponding pixel obtained from a single surface feature image. DETAILED DESCRIPTION

[0042] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0043] As shown in the figure, the embodiment provides a method for obtaining fine surface albedo for hyperspectral UAV remote sensing, including the following steps:

[0044] S1, uses the visible imaging lens fixed on the UAV payload to capture the surface feature image corresponding to each pixel, and performs preprocessing, adaptive cropping and semantic segmentation on each surface feature image.

[0045] In the embodiment, the visible imaging lens and the telescope module of the drone payload are fixed on the same axis, and a coaxial fixing method is adopted to ensure that the center of the photo taken by the visible imaging lens coincides with the center of the telescope module, and at the same time ensure that the telescope elevation angle and the visible imaging lens rotate synchronously, which can ensure that each surface feature image taken by the visible imaging lens corresponds to a pixel at each elevation angle of the telescope. In order to ensure that the drone observation results reach meter-level resolution, the drone flies at an altitude of 100-120 meters, so the field of view of the telescope needs to be controlled within 5°.

[0046] During the flight of the drone, the drone will shake, and when the telescope elevation angle changes, in order to ensure the clarity of the surface feature image taken by the visible imaging lens, the visible imaging lens is automatically zoomed, so the field of view of the visible imaging lens changes. The visible imaging lens can be calculated based on the surface feature image taken. The calculation method is as follows:

[0047]

[0048] The height of the drone from the ground is h, the inclination angle of the telescope is α (the angle with the horizontal direction), and the pixel width and height of the surface feature image taken are W respectively. pixel and H pixel , the width and height of the camera sensor are W image and H image θ horizonrtal and θ vertical Represents the field of view of the visible imaging lens in the horizontal and vertical directions. Since the change in the field of view caused by the change in the focal length of the visible imaging lens is linear, θ is selected horizonrtal As the field of view of the visible imaging lens, it is denoted as θ in the following text. camera .

[0049] After obtaining the surface feature image, first, each surface feature image is preprocessed, including: for each surface feature image captured by the visible imaging lens, the photo is geometrically corrected based on the RANSAC (random sampling consensus algorithm) and the automatic matching algorithm of homologous points to eliminate the influence of geometric deformation on the surface albedo.

[0050] Each preprocessed surface feature image is then adaptively cropped to ensure that it fully matches the pixel size, including: Figure 2 As shown, the field of view angle θ of the visible imaging lens is camera and the field of view angle θ of the telescope module telescope The ratio of the image width of the visible imaging lens to the image width of the telescope is calculated by the ratio of the image width of the visible imaging lens to the image width of the telescope, where the field of view angle of the telescope module is θ telescope It is a fixed value, and its size depends on the optical fiber and optical lens used, namely:

[0051]

[0052] The surface feature images are adaptively cropped based on the width ratio to ensure that each image taken by the visible imaging lens fully corresponds to each pixel of the telescope.

[0053] Finally, the surface feature image after adaptive cropping is subjected to semantic segmentation of surface types based on machine learning. In order to perform semantic segmentation of surface types in detail, a deep convolutional neural network (SegNet) model is introduced to segment the surface feature image after adaptive cropping into different surface types i (i = 1, 2, 3...n), such as grassland, bare soil, vegetation, cement pavement, asphalt pavement, etc. The model output is the segmentation mask M i (x,y), where M i (x,y) represents the surface type i at pixel (x,y). The area S of each surface type i i It can be calculated by the following formula:

[0054]

[0055] Among them, S pixel is the actual ground area of ​​each pixel, W and H are the width and height of the surface feature image respectively.

[0056] S2, weighted calculation of the surface feature image is performed according to the semantic segmentation result and combined with the surface albedo of each surface type to obtain the refined surface albedo of the surface feature image.

[0057] In the embodiment, for each pixel, use A iIt represents the surface albedo of surface type i (the surface albedo will change with the wavelength. During the calculation process, the surface albedo value corresponding to the band is selected according to the inversion band required by the target gas in the hyperspectral inversion process). Table 1 shows the common surface types and their surface albedo. The weighted surface albedo of different surface types is used to calculate the refined surface albedo Albedo:

[0058]

[0059] Table 1. Common surface types and their surface albedo

[0060]

[0061]

[0062] The surface albedo will change with wavelength. The central wavelength of the required band is used in the calculation process. The conversion coefficient between surface albedo and wavelength is as follows:

[0063]

[0064] Where f is the conversion coefficient, k is an exponential variation coefficient related to the type of land feature or material, which determines the rate of change of albedo with wavelength. The exponential variation coefficient k is usually obtained by fitting experimental data, reflecting the reflection characteristics of the surface material at different wavelengths; λ1 and λ2 are two central wavelengths. The k values ​​of common surface types are: green vegetation k value range: 0.02–0.05; water k value range: 0.001–0.005; soil k value range: 0.01–0.03.

[0065] Therefore, the fine surface albedo at different central wavelengths is obtained through the conversion coefficient:

[0066] Albedo'=f×Albedo (7)

[0067] Wherein, Albedo' represents the fine surface albedo at the central wavelength λ2, and Albedo represents the fine surface albedo at the central wavelength λ1.

[0068] Based on the above method, the embodiment also provides an experimental example, in which a specific observation site is selected, which is a certain Environmental Protection Energy Co., Ltd. This site is a centralized garbage processing area. Taking the UAV hyperspectral remote sensing technology as an example, the process of UAV hyperspectral remote sensing observation is selected to illustrate the complete process of the method of the present invention.

[0069] In the experimental example, the drone's flight altitude is 110 meters, and the telescope's elevation angle (the angle between the telescope and the horizontal direction is 0°) is set to: 8°, 10°, 12°, 14°, 17°, 22°, 32°, 52°. The uneven setting of the elevation angle is to ensure the uniformity of the ground pixels. The drone has a total of 8 horizontal hovering points, and the telescope rotates once at each hovering point according to the elevation angle sequence. Therefore, there are a total of 64 pixels for the entire observation area.

[0070] Then, adaptive cropping and fine surface albedo calculation are performed for each surface feature image collected by the visible imaging lens. The detailed process is as follows: Figure 3 shown.

[0071] The main surface types in the observation scene are grassland, bare soil, cement surface, broad-leaved trees, bushes, and asphalt pavement. Combining Table 1 with formulas (5), (6), and (7), the surface albedo of the above surface types in the ultraviolet band can be obtained (here the surface albedo at the central wavelength λ = 354nm in the NO2 inversion band of 338-370nm is calculated), as shown in Table 2.

[0072] Table 2 Surface albedo of main surface types in the observation area at λ = 354nm

[0073]

[0074] Combining formulas (1-7) and Table 2, we can get the fine surface albedo value of the observation area, such as Figure 4 As shown, the surface type in the visible observation area is varied, and the refined surface albedo is crucial to the result, which is more accurate than a uniform surface.

[0075] An embodiment also provides a fine surface albedo acquisition system for hyperspectral UAV remote sensing, including a UAV payload with a visible imaging lens and a telescope module, and a computing module, wherein the UAV payload captures a surface feature image corresponding to each pixel through the visible imaging lens thereon; the computing module preprocesses, adaptively crops, and semantically segments each captured surface feature image, and then performs weighted calculation of the surface feature image based on the semantic segmentation result and in combination with the surface albedo of each surface type to obtain the fine surface albedo of the surface feature image.

[0076] In the embodiment system, the specific implementation methods of the drone payload and the computing module are detailed in the above embodiment method and will not be repeated here.

[0077] The method and system provided in the above embodiment, as a new solution for calculating the surface albedo of a single pixel, can finely characterize the surface albedo of each observation pixel in the entire target area for UAV hyperspectral observation. Compared with the traditional single surface albedo, the observation results will be more accurate and reliable.

[0078] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for obtaining fine surface albedo for hyperspectral UAV remote sensing, characterized in that: The following steps are involved: The visible imaging lens fixed on the UAV payload captures the surface feature image corresponding to each pixel, and performs preprocessing, adaptive cropping and semantic segmentation on each surface feature image; According to the semantic segmentation results and combined with the surface albedo of each surface type, a weighted calculation of the surface feature image is performed to obtain the refined surface albedo of the surface feature image.

2. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 1 is characterized in that: The visible imaging lens is fixed on the same axis as the telescope module on the UAV payload in a coaxial fixing manner, so that the center of the photo taken by the visible imaging lens coincides with the center of the telescope module, while ensuring that the telescope elevation angle and the visible imaging lens rotate synchronously, ensuring that each surface feature image taken by the visible imaging lens corresponds one-to-one to the pixel of each elevation angle of the telescope.

3. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 1, characterized in that: Preprocess each surface feature image, including: The surface feature image is geometrically corrected based on the random sampling consistency algorithm and the automatic matching algorithm of homologous points to eliminate the influence of geometric deformation on the surface albedo in the surface feature image.

4. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 1, characterized in that: Adaptively crop each surface feature image, including: Field of view angle θ using visible imaging lens camera and the field of view angle θ of the telescope module telescope The image width W of the visible imaging lens is calculated by the ratio of camera and the telescope's image width W telescope The ratio is: The surface feature images are adaptively cropped based on the width ratio to ensure that each image taken by the visible imaging lens fully corresponds to each pixel of the telescope.

5. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 4 is characterized in that: Field of view angle θ of the telescope module telescope is a fixed value, and its size depends on the optical fiber and optical lens used. It can be seen that the field of view angle θ of the imaging lens camera Calculated by: Among them, h is the height of the drone from the ground, α is the inclination angle between the telescope and the horizontal direction, and W pixel and H pixel are the pixel width and height of the surface feature image taken, W image and H image are the width and height of the camera sensor, θ horizonrtal and θ vertical Represents the field of view of the visible imaging lens in the horizontal and vertical directions. Since the change in the field of view caused by the change in the focal length of the visible imaging lens is linear, θ is selected. horizonrtal As the field of view of the visible imaging lens, denoted by θ camera .

6. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 1, characterized in that: Semantic segmentation is performed on each surface feature image, including: A deep convolutional neural network model is used to segment the adaptively cropped surface feature image, and the model output is the segmentation mask M. i (x,y), where M i (x,y) represents the surface type i at pixel (x,y).

7. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 6, characterized in that: The deep convolutional neural network model adopts the SegNet model.

8. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 1, characterized in that: Based on the semantic segmentation results and combined with the surface albedo of each surface type, weighted calculation of surface feature images is performed, including: For each pixel, use A i Represents the surface albedo of surface type i, based on the surface albedo A i The weighted calculation of the surface feature image is expressed as: Among them, S pixel represents the actual ground area of ​​each pixel, n represents the total number of surface types, Albedo represents the weighted fine surface albedo, S i Represents the area of ​​surface type i, and the calculation formula is: Where W and H are the width and height of the surface feature image, respectively.

9. The method for obtaining fine surface albedo for hyperspectral UAV remote sensing according to claim 1, characterized in that: Also includes: The detailed surface albedo at different central wavelengths is obtained through conversion coefficients, including: Albedo'=f×Albedo Where f is the conversion coefficient, k is an exponential variation coefficient related to the type or material of the feature, and the coefficient k determines the rate of change of albedo with wavelength, reflecting the reflection characteristics of the surface material at different wavelengths; λ1 and λ2 are two central wavelengths, Albedo' represents the fine surface albedo at the central wavelength λ2, and Albedo represents the fine surface albedo at the central wavelength λ1.

10. A fine surface albedo acquisition system for hyperspectral UAV remote sensing, characterized in that: Includes a drone payload with a visible imaging lens and telescope module, as well as a computing module, The UAV payload captures the surface feature image corresponding to each pixel through the visible imaging lens on it; The calculation module performs preprocessing, adaptive cropping and semantic segmentation on each surface feature image taken, and then performs weighted calculation of the surface feature image according to the semantic segmentation result and in combination with the surface albedo of each surface type to obtain the fine surface albedo of the surface feature image.

Images