A hyperspectral remote sensing oil spill monitoring and analysis method and monitoring platform

Abstract

The present application relates to the field of material composition detection, and specifically discloses a hyperspectral remote sensing oil spill monitoring and analysis method and a monitoring platform, the monitoring platform including a hyperspectral sensor, a positioning system, a posture control system, and a carrier, wherein the posture control system is used to adjust the shooting angle of the hyperspectral sensor. When collecting data, the hyperspectral sensor is mounted on the carrier, and spectral collection is performed on detection sites set in the inland target area from different angles, hyperspectral images under different shooting angles are extracted, and hyperspectral data under multiple angles are comprehensively analyzed. The composition of the oil spill and the content of different components are determined according to the spectral characteristics of the oil spill in different bands, the thickness of the oil spill is estimated according to the spectral reflectance characteristics under multiple angles, and the total amount of oil spill is calculated according to the oil spill thickness and the oil spill area. The technical solution of the present application analyzes the inland detection sites based on the hyperspectral data under multiple angles, which can improve the accuracy of oil spill component detection, coverage and oil spill thickness calculation.

Description

Technical Field

[0001] The present application relates to the fields of environmental monitoring and remote sensing technology, and in particular to a hyperspectral remote sensing oil spill monitoring and analysis method and a monitoring platform. Background Art

[0002] Oil spills in inland waterways are characterized by suddenness, high dynamism, and low identification accuracy. They pose serious risks to socioeconomic development and the ecological and environmental safety of inland waterways. Emergency monitoring capabilities are severely limited, particularly in complex inland waterway environments. Effective emergency monitoring technologies are urgently needed for accurate monitoring. The extent of an oil spill, which is related to its spatial distribution and volume, is crucial for on-site disposal. The type of spilled oil, which influences source identification, penalties, and pollution cleanup, is crucial for pollution management. Currently, passive and active optical remote sensing are important tools for oil spill monitoring. Passive optical remote sensing, with its advantages of high spectral and spatial resolution, has been the most widely researched and applied in this field. Passive optical remote sensing monitoring of surface oil spills in inland waterways is susceptible to flares and can be difficult to effectively distinguish between different types of light oil. Furthermore, most optical remote sensing monitoring of surface oil spills in inland waterways relies on single-dimensional data, such as spectral, polarization, or infrared data. The detection and identification accuracy required for precise monitoring applications is insufficient. Summary of the Invention

[0003] The purpose of this application is to extract and analyze multi-dimensional optical features, build a detection and recognition model, and achieve accurate monitoring of oil spill range and thickness and effective identification of different types of light oil.

[0004] The technical solution of this application is to provide a hyperspectral remote sensing oil spill monitoring and analysis method, which includes:

[0005] Step 1: Use a hyperspectral imager to collect hyperspectral images of the detection site of the target area from a vertical perspective, a first oblique perspective, and a second oblique perspective, wherein the shooting positions of the hyperspectral imager at the three perspectives are in the same plane and are equidistant from the target area;

[0006] Step 2: preprocessing the collected hyperspectral image and performing spectral correction on the preprocessed hyperspectral image;

[0007] Step 3: extract endmembers from all spectral data contained in the hyperspectral image, establish a mixture model based on the extracted endmembers, and unmix the mixture model to obtain the components and content ratios corresponding to all pixels in the hyperspectral image;

[0008] Step 4: extracting reflectance spectra from the hyperspectral image, establishing an oil film thickness calculation model based on the reflectance spectra at multiple viewing angles, and estimating the oil film thickness according to the oil film thickness calculation model;

[0009] Step 5: Calculate the oil spill volume in the shooting area corresponding to the hyperspectral image based on the composition and content ratio of all pixels in the hyperspectral image at a vertical viewing angle, as well as the corresponding oil film thickness. Calculate the total oil spill volume based on the oil spill volume of all shooting areas in the target area.

[0010] Step 6: Based on the endmember information of all vertical viewing angle hyperspectral images taken in the target area, all pixels with oil spills are found to obtain the total number of pixels in the oil spill area. The total area of ​​the oil spill area in the target area is calculated based on the total number of pixels in the oil spill area.

[0011] Furthermore, step 1 specifically includes:

[0012] Using a hyperspectral imager carried by an unmanned aerial vehicle to collect data from a predetermined height directly above the detection site, a first hyperspectral image is obtained at a vertical viewing angle, wherein the optical axis of the hyperspectral imager is vertically downward and perpendicular to the surface where the detection site is located;

[0013] Using a hyperspectral imager to collect data from the same height on one side of the original position of the UAV, a second hyperspectral image is obtained at the first oblique viewing angle, wherein the angle between the optical axis of the hyperspectral imager and the horizontal direction is 30° to 60°;

[0014] A hyperspectral imager is used to collect data from the same height on the other side of the original position of the UAV to obtain a third hyperspectral image at a second oblique viewing angle, wherein the angle between the optical axis of the hyperspectral imager and the horizontal direction is 30° to 60°, and the optical axis of the hyperspectral imager at the first oblique viewing angle and the optical axis of the hyperspectral imager at the second oblique viewing angle are symmetrical about the vertical line of the detection site.

[0015] Furthermore, step 3 specifically includes:

[0016] Step 3.1, extract endmembers from the spectral data of all pixels in the hyperspectral image and establish a mixture model;

[0017] Spectral features, including absorption peaks, reflection peaks, and spectral slopes, are extracted from the spectral data of each pixel. The extracted spectral features are matched with the known oil component spectral database to obtain the corresponding end members. The mixed model is established as follows:

[0018] R=f(M,A)+N

[0019] Where R is the original observation data matrix, M is the spectrum matrix of the endmembers in each pixel, where each column represents the spectrum of an endmember, A is the abundance matrix of different endmembers in each pixel, and N is the noise vector;

[0020] In step 3.2, the mixed model is unmixed to obtain the endmember spectrum matrix and endmember abundance matrix corresponding to all pixels in the hyperspectral image, and the oil spill components and contents in the target area are statistically calculated as follows:

[0021] Initialize the spectral matrix M and the abundance matrix A, and update the spectral matrix M and the abundance matrix A through iterative optimization so that their product approaches the original observation data matrix R. Set the stopping condition, which is to stop the iteration when the maximum number of iterations is reached, and update the endmember spectral matrix M:

[0022]

[0023] Update the abundance matrix A:

[0024]

[0025] Where the symbol ⊙ represents point-by-point multiplication of elements, the symbol ÷ represents point-by-point division of elements, and E is a matrix whose internal values ​​are all 1. Based on the iterative spectral matrix M and abundance matrix A, the oil spill components and the content ratios of different components at each pixel are obtained. The subscript T represents the matrix transpose.

[0026] Furthermore, step 2 specifically includes:

[0027] Filtering the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image collected by the hyperspectral imager to remove noise in the images;

[0028] Using water bodies from other areas without oil pollution as the control group, a point was selected in the control area as the control point. The reflectance of the control point was measured using a hyperspectral imager. The reflectance was used for spectral correction to eliminate the influence of ambient light and sensor, ensuring the consistency and accuracy of the data. The details are as follows:

[0029]

[0030] Where R raw (i, j, λ) is the reflectance of the original hyperspectral image, R ref (λ) is the reflectance of the hyperspectral image of the control group, R cal (i, j, λ) is the reflectance of the corrected hyperspectral image, i and j represent the abscissa of the corrected point and ordinate λ is the wavelength.

[0031] Furthermore, step 4 specifically includes:

[0032] The pixel point at the geometric center of the hyperspectral image is taken as its center point, and the reflectance spectrum of the center points of the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image is obtained, which is calculated as R n(λ), where n represents the labels of different hyperspectral images corresponding to the same pixel, n = 1, 2, 3;

[0033] The reflectivity model of the oil film thickness is established based on the reflectivity spectra of the center points of the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image:

[0034] R n (λ)=R ref (λ)·e -2α(λ)·d

[0035] Where R ref (λ) is the reflectivity of the control group hyperspectral image, α is the absorption coefficient of the oil film at wavelength λ, and d is the thickness of the oil film. The reflectivity model of the oil film thickness is optimized, and the reflectivity of different shooting angles is integrated to obtain the oil film thickness calculation model:

[0036]

[0037] Set the initial oil film thickness d0, substitute d0 into the formula and use the gradient descent method to perform optimization iterative calculations. Set the stopping condition to be that the change in d value is less than the set threshold, and get the final oil film thickness d k+1 , traverse the oil film thickness of all detection points in the target area.

[0038] Furthermore, step 5 specifically includes:

[0039] The oil film thickness at the center point of the first hyperspectral image is taken as the oil film thickness of the entire shooting area corresponding to it. The oil spill volume of the area is calculated based on the area of ​​the shooting area and the corresponding oil film thickness. The oil spill volume of all shooting areas included in the target area is accumulated to obtain the total oil spill volume, which is as follows:

[0040]

[0041] Where, L oil is the total oil spill volume in the target area, S x is the area of ​​the shooting area corresponding to the first hyperspectral image, d x is the oil film thickness at the center point of the first hyperspectral image, X is the total number of all shooting areas included in the target area, and x is the label of the shooting area included in the target area;

[0042] The shooting area is divided into small blocks in equal proportion according to the position of the pixel points in the hyperspectral image. Each pixel point corresponds to a small block. According to the area S of the shooting area, x Calculate the area s of the small area block, and calculate the oil film thickness d of the corresponding shooting area according to the area s of the small area block x, the oil spill components and content ratios corresponding to the pixel points, and the content of different oil spill components in the shooting area are calculated as follows:

[0043]

[0044] Where, O y is the content of a certain oil spill component in the shooting area, a y is the content ratio of a certain oil spill component in the pixel, y is the label of a certain oil spill component, C is the total number of all small area blocks in the shooting area, and c is the label of the small area blocks in the shooting area.

[0045] Furthermore, step 6 specifically includes:

[0046] All pixels with end members in the first hyperspectral image are used as pixels of the oil spill area. The first hyperspectral images corresponding to all shooting areas in the target area are traversed in the same way to calculate the total number of pixels in the oil spill area in the target area. The oil spill area is calculated based on the total number of pixels in the oil spill area and the area of ​​the small area blocks corresponding to the pixels, as follows:

[0047] S oil =P·s

[0048] Where S oil is the oil spill area of ​​the target area, and P is the total number of pixels in the oil spill area.

[0049] The present application also provides a hyperspectral remote sensing oil spill monitoring platform, which includes: a hyperspectral imager, a positioning system, a posture control system, a carrier and a host computer;

[0050] The hyperspectral imager is used to collect hyperspectral images of the target area from different shooting angles; the positioning system is used to position the hyperspectral imager so that the position of the hyperspectral imager can be adjusted through the position information fed back by the positioning system when collecting hyperspectral images; the attitude control system is used to adjust the shooting angle of the hyperspectral imager to achieve multi-angle shooting of the target area; the carrier is used to carry the hyperspectral imager for data collection; the host computer is used to display, store and analyze the hyperspectral images collected by the hyperspectral imager.

[0051] Furthermore, the hyperspectral remote sensing oil spill monitoring platform is used to execute the method of any one of claims 1-7.

[0052] The beneficial effects of this application are:

[0053] The technical solution in the present application collects hyperspectral images of the target area from multiple shooting angles, establishes a hybrid model based on the hyperspectral images at different shooting angles, calculates the oil spill components and the content ratio of each component, extracts the reflectance spectra at different shooting angles from the hyperspectral image data, establishes an oil film thickness calculation model, and estimates the oil film thickness. Compared with the detection based on single-dimensional spectral data in the existing technology, the technical solution in the present application fuses and calculates the data at different shooting angles, and the obtained data such as oil spill components, oil content and oil film thickness are more accurate. By fusing data at different shooting angles, the influence of occlusion and flare caused by a single perspective can be eliminated, the accuracy and reliability of spectral data can be improved, and more accurate spectral data can be provided for the detection of oil spill components and the content of each component. At the same time, more comprehensive oil film thickness information can be provided, reducing errors caused by local thickness changes or the influence of lighting angles, further improving detection accuracy, and providing a scientific basis and reference data for environmental monitoring and governance. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The advantages of the above and / or additional aspects of the present application will become apparent and readily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0055] Figure 1 is a schematic flow chart of a hyperspectral remote sensing oil spill monitoring and analysis method according to one embodiment of the present application;

[0056] Figure 2 1 is a reflectance spectrum diagram of different oil products according to an example of the present application. DETAILED DESCRIPTION

[0057] In order to more clearly understand the above-mentioned objectives, features and advantages of the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features therein can be combined with each other in the absence of conflict.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present application is not limited to the specific embodiments disclosed below.

[0059] like Figure 1 As shown, this embodiment provides a hyperspectral remote sensing oil spill monitoring platform, which is used to detect oil spill areas in inland rivers and includes a hyperspectral imager, a positioning system, a posture control system, a carrier, and a host computer.

[0060] A hyperspectral imager is used to capture hyperspectral images of a target area from different viewing angles. The hyperspectral imager includes a hyperspectral sensor, an optical system (e.g., a system consisting of lenses, filters, and other components), a data processing unit, a storage unit, and a display device. The optical system is used to collect light from the entire target area scene and focus it onto the hyperspectral sensor; the hyperspectral sensor is used to obtain hyperspectral data from the entire target area scene through the optical system; the data processing unit is used to perform feature extraction, classification, and other operations on the raw hyperspectral data, ultimately converting it into a hyperspectral image; the storage unit is used to store data within the instrument; and the display device is used to display the hyperspectral image. The hyperspectral imager in this embodiment is conventional in the art and will not be described in detail here.

[0061] Each pixel in a hyperspectral image has corresponding spectral information, that is, it can simultaneously provide position information and spectral information corresponding to the position; wherein, the position of each pixel in the image corresponds to a specific point in the target area scene being photographed, and each pixel includes a complete spectrum (that is, includes a complete spectral characteristic curve). Different substances reflect, absorb or emit light differently at different wavelengths, so the spectrum is a mixed spectrum, reflecting the result of mixing the spectra of different substances. Subsequent spectral analysis needs to first demix the mixed spectrum to extract the pure spectral characteristics of each substance.

[0062] In this embodiment, the airborne hyperspectral imager is a passive sensor that does not need to emit wavelength-adjustable light when photographing the target area. It uses natural light from the sun or other ground light sources to illuminate the target area for imaging. These natural light sources provide sufficient spectral information. The hyperspectral imager can generate a hyperspectral image by receiving the reflected light from the target area. The hyperspectral imager mainly receives visible light and near-infrared light. Specifically, it can receive spectral information from the visible light wavelength (about 400 nanometers) to the near-infrared light wavelength (about 2500 nanometers). Different wavelengths of light in this wavelength range have different energies and penetration capabilities, which enables the hyperspectral imager to capture the detailed spectral characteristics reflected from the surface of the object.

[0063] The positioning system is used to position the hyperspectral imager so that the position of the hyperspectral imager can be adjusted through the position information fed back by the positioning system when collecting hyperspectral images.

[0064] The attitude control system is used to adjust the shooting angle of the hyperspectral imager to achieve multi-angle shooting of the target area.

[0065] The carrier is used to carry the hyperspectral imager for data collection. In this embodiment, the carrier can be divided into a road-based carrier and an air-based carrier. The road-based carrier is such as an unmanned vehicle, and the air-based carrier is such as a drone.

[0066] The host computer is used to display, store and analyze the hyperspectral images collected by the hyperspectral imager.

[0067] like Figure 1 As shown, this embodiment provides a hyperspectral remote sensing oil spill monitoring and analysis method, which is used to collect and analyze spectral data from detection sites set in a target inland river area to obtain information on the composition, thickness, and area of ​​the oil spill. The method includes the following steps:

[0068] Step 1: Use a hyperspectral imager to collect hyperspectral images of the detection site of the target area from a vertical perspective, a first oblique perspective, and a second oblique perspective, respectively. The shooting positions of the hyperspectral imager at the three perspectives are in the same plane and are equidistant from the target area.

[0069] The hyperspectral imager, positioning system, and attitude control system are installed on the carrier UAV, and the hyperspectral imager is connected to the attitude control system.

[0070] Based on the data fed back by the positioning system, the drone is controlled to reach directly above the detection point set in the target area. The attitude control system is used to adjust the acquisition angle of the hyperspectral imager so that its optical axis is vertically downward and perpendicular to the surface where the detection point is located. The hyperspectral imager is used to collect data from a predetermined height directly above the detection point, and the first hyperspectral image under the vertical perspective is obtained and uploaded to the host computer.

[0071] Based on the data fed back by the positioning system, the drone is controlled to translate a predetermined distance in any direction, wherein the angle between the straight line between the current position of the drone and the detection position and the horizontal direction is 30° to 60°. The predetermined distance can be set by a predetermined height so that the position of the drone after translation meets the condition of an angle of 30° to 60° with the horizontal direction. The attitude control system is used to adjust the acquisition angle of the hyperspectral imager so that its optical axis is aligned with the detection position for data collection, and a second hyperspectral image under the first inclined perspective is obtained and uploaded to the host computer.

[0072] Based on the data fed back by the positioning system, the drone is controlled to translate in the original direction by twice the predetermined distance, wherein the angle formed by the drone with the horizontal direction after the second movement is equal to the angle after the first movement, satisfying the condition that the angle with the horizontal direction is 30° to 60°. The attitude control system is used to adjust the acquisition angle of the hyperspectral imager so that its optical axis is aligned with the detection site for data collection, and the third hyperspectral image under the second inclined perspective is obtained and uploaded to the host computer.

[0073] Step 2: preprocess the collected hyperspectral image and perform spectral correction on the preprocessed hyperspectral image.

[0074] The first hyperspectral image, the second hyperspectral image, and the third hyperspectral image collected by the hyperspectral imager are filtered to remove noise in the images. In this embodiment, the preprocessing is denoising, and the denoising operation can be performed by applying existing filters or other denoising algorithms.

[0075] Using water bodies from other areas without oil pollution as the control group, a point was selected in the control area as the control point. The reflectance of the control point was measured using a hyperspectral imager. The reflectance was used for spectral correction to eliminate the influence of ambient light and sensor, ensuring the consistency and accuracy of the data. The details are as follows:

[0076]

[0077] Where R raw (i, j, λ) is the reflectance of the original hyperspectral image, R ref (λ) is the reflectance of the hyperspectral image of the control group, R cal (i, j, λ) is the reflectance of the corrected hyperspectral image, i and j represent the abscissa of the corrected point, and ordinate λ is the wavelength. The coordinates in the above formula are the coordinates in the hyperspectral image, corresponding to the points in the actual target area.

[0078] In this embodiment, when measuring the reflectivity of the control site, it is necessary to ensure that the acquisition time is within the same time period as the acquisition time of the test site, and the lighting conditions are roughly the same to reduce errors. Because the water in the control area is unpolluted, the reflectivity of each site in this area is very close or identical. Therefore, a single site can be selected as the control site, and the effect of the shooting direction on the reflectivity of the control site can be ignored.

[0079] Step 3: Extract endmembers (endmembers are the spectra of a pure material, such as diesel) from all spectral data contained in the hyperspectral image, establish a mixing model based on the extracted endmembers, unmix the mixing model, and obtain the components and content ratios corresponding to all pixels in the hyperspectral image.

[0080] Step 3.1: extract endmembers from the spectral data of all pixels in the hyperspectral image and establish a mixture model.

[0081] Spectral features are extracted from the spectral data of each pixel, including the absorption peak (the wavelength location of the absorption peak), the reflection peak (absorption intensity), and the spectral slope (the slope in different wavelength ranges). The extracted spectral features are matched with a known database of oil component spectra to obtain the corresponding endmembers. In this embodiment, in addition to the above-mentioned spectral feature analysis method, other algorithms such as spectral angle mapping and extreme spectrum extraction algorithms can also be used for endmember extraction. The hybrid model is established as follows:

[0082] R=f(M,A)+N

[0083] Where R is the original observation data matrix (and the observed hyperspectral data matrix, representing the spectral data of each pixel), M is the spectral matrix of the endmembers in each pixel (including all the endmembers in the hyperspectral image), where each column represents the spectrum of an endmember, A is the abundance matrix of different endmembers in each pixel (and the matrix of content ratios), and N is the noise vector.

[0084] In this embodiment, the spectrum received by the hyperspectral imager corresponds to the result of mixing the reflection spectra of different substances in the instantaneous field of view on the water surface of the target area, where the spectrum of each pixel point is a mixed spectrum, reflecting the spectral information of the mixture of different components. It is necessary to decompose the mixed spectrum of all pixel points into spectra of multiple different end members (i.e., pure substance spectra), and calculate their corresponding abundance (content ratio) to finally obtain the oil spill components and content in the water body of the target area.

[0085] In step 3.2, the mixed model is unmixed to obtain the endmember spectrum matrix and endmember abundance matrix corresponding to all pixels in the hyperspectral image, and the oil spill composition and content in the target area are statistically calculated.

[0086] Initialize the spectral matrix M and the abundance matrix A, update the spectral matrix M and the abundance matrix A through iterative optimization method, so that their product approaches the original observation data matrix R, and set the stopping condition, which is to stop the iteration when the maximum number of iterations is reached.

[0087] Update the endmember spectral matrix M:

[0088]

[0089] Update the abundance matrix A:

[0090]

[0091] Where the symbol ⊙ represents point-by-point multiplication of elements, the symbol ÷ represents point-by-point division of elements, and E is a matrix whose internal values ​​are all 1. Based on the iterative spectral matrix M and abundance matrix A, the endmembers and their corresponding abundances are obtained for each pixel in the hyperspectral image. This means that the oil spill components and their content ratios at each pixel are obtained. The subscript T represents the matrix transpose.

[0092] Step 4: extract the reflectance spectrum from the hyperspectral image, establish an oil film thickness calculation model based on the reflectance spectra of multiple viewing angles, and estimate the oil film thickness according to the oil film thickness calculation model.

[0093] Oil spills primarily exist in the form of an oil film within the core area, representing the bulk of the spill and posing the greatest environmental risk. For inland river oil spill response, simply determining the location and distribution of the spill is insufficient. Only by accurately determining the thickness of the oil film on the water surface and the total volume of oil spilled can more effective response measures be formulated to minimize the damage. Oil films of varying thickness exhibit specific reflection and absorption characteristics at specific wavelengths.

[0094] In this embodiment, the spectral data structure of the hyperspectral image is a three-dimensional data structure, represented by the dimensions (i, j, λ), where (i, j) is the coordinate of the pixel point and λ is the wavelength. Each pixel corresponds to a reflectance spectrum, i.e., a complete spectral curve. The corresponding λ of the spectral curve can range from 400nm to 1000nm. The horizontal axis of the spectral curve is the wavelength λ, and the vertical axis is the reflectance value. The horizontal axis corresponds to several wavelength channels, each of which corresponds to a reflectance value. Reflectance spectra are required for oil film thickness analysis.

[0095] The pixel point at the geometric center of the hyperspectral image is taken as its center point, and the reflectance spectrum of the center points of the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image is obtained, which is calculated as R n (λ), where n represents the labels of different hyperspectral images corresponding to the same pixel, n = 1, 2, 3, such as R1(λ) is the reflectance spectrum of the center point of the first hyperspectral image.

[0096] The reflectivity model of the oil film thickness is established based on the reflectivity spectra of the center points of the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image:

[0097] R n (λ)=R ref (λ)·e -2α(λ)·d

[0098] Where R ref (λ) is the reflectance of the hyperspectral image of the control group, α is the absorption coefficient of the oil film at wavelength λ, and d is the thickness of the oil film.

[0099] The least squares method is used to optimize the reflectivity model of the oil film thickness mentioned above. By combining the reflectivity of different shooting angles, a more accurate oil film thickness calculation model is obtained:

[0100]

[0101] Set the initial oil film thickness d0, substitute d0 into the above formula and use the gradient descent method to perform iterative optimization. For the optimization of the oil film thickness d, set the stopping condition to be that the change of the d value is less than the set threshold, as follows:

[0102] Set the initial oil film thickness d0 = 1mm, the learning rate η = 0.01, and the threshold value to 0.001, and calculate the current oil film thickness d k The reflectivity model is:

[0103]

[0104] Calculate the gradient of the loss function:

[0105]

[0106] Update oil film thickness:

[0107]

[0108] When |d k+1 -d k When |<0.001mm, stop the iteration and get the final oil film thickness d k+1 , calculate the oil film thickness d at all detection points in the target area according to the above method k+1 .

[0109] Using an oil film sample of known thickness to verify the accuracy of the computational model and assess the model's errors and deviations, the model parameters can be appropriately adjusted to improve computational accuracy. In this embodiment, the center point of the hyperspectral image is the projection point of the hyperspectral imager's optical axis on the water surface. For a specific detection point in the target area, the center points of the first, second, and third hyperspectral images all correspond to that detection point, differing only in the acquisition perspective.

[0110] Step 5: Calculate the amount of oil spilled in the shooting area corresponding to the first hyperspectral image based on the composition and content ratio corresponding to all pixels in the hyperspectral image under the vertical viewing angle, as well as the oil film thickness at the center point. Calculate the total amount of oil spilled based on the amount of oil spilled in all shooting areas in the target area.

[0111] The oil film thickness at the center point of the first hyperspectral image is taken as the oil film thickness of the entire shooting area corresponding to it. The oil spill volume of the area is calculated based on the area of ​​the shooting area and the corresponding oil film thickness. The oil spill volume of all shooting areas included in the target area is accumulated to obtain the total oil spill volume, which is as follows:

[0112]

[0113] Where, L oil is the total oil spill volume in the target area, S x is the area of ​​the shooting area corresponding to the first hyperspectral image, d xis the oil film thickness at the center point of the first hyperspectral image, X is the total number of all shooting areas included in the target area, and x is the label of the shooting area included in the target area. For example, when x is 1, (S1·d1) represents the oil spill volume of the shooting area with label 1.

[0114] The shooting area is divided into small blocks in equal proportion according to the position of the pixel points in the hyperspectral image. Each pixel point corresponds to a small block. According to the area S of the shooting area, x Calculate the area s of the small area block, and calculate the oil film thickness d of the corresponding shooting area according to the area s of the small area block x , the oil spill components and content ratios corresponding to the pixel points (the oil spill components and content ratios corresponding to the pixel points can be obtained through the spectrum matrix M and the abundance matrix A), and the content of different oil spill components in the shooting area is calculated as follows:

[0115]

[0116] Where, O y is the content of a certain oil spill component in the shooting area, a y is the content ratio of a certain oil spill component in the pixel, y is the label of a certain oil spill component, C is the total number of all small area blocks in the shooting area, and c is the label of the small area blocks in the shooting area.

[0117] In step 6, based on the end-member information of the hyperspectral images at all vertical viewing angles in the target area obtained in step 3, all pixels with oil spills are found to obtain the total number of pixels in the oil spill area. The total area of ​​the oil spill area in the target area is calculated based on the total number of pixels in the oil spill area.

[0118] All pixels with end members (i.e., oil spill components) in the first hyperspectral image are used as pixels in the oil spill area. The first hyperspectral images corresponding to all shooting areas in the target area are traversed in the same way to calculate the total number of pixels in the oil spill area in the target area. The oil spill area is calculated based on the total number of pixels in the oil spill area and the area of ​​the small area blocks corresponding to the pixels, as follows:

[0119] S oil =P·s

[0120] Where S oil is the oil spill area of ​​the target area, and P is the total number of pixels in the oil spill area.

[0121] In this embodiment, the target area refers to the area that needs to be detected in the inland water area, and the shooting area refers to dividing the target area into multiple smaller areas. One shooting area corresponds to shooting a set of hyperspectral images (including hyperspectral images collected at the vertical perspective, the first oblique perspective, and the second oblique perspective), and one shooting area corresponds to one detection site.

[0122] In Example 1, a certain amount of water is taken from a sample pool, and five typical inland water oil spill oils are put into the sample pool as pollutants to form a sewage mixture, simulating the sewage in the inland water oil spill area. The oil products include crude oil, fuel oil, palm oil, No. 0 diesel, and No. 95 gasoline; the attitude control system is used to adjust the acquisition angle of the hyperspectral imager so that its optical axis is vertically downward and perpendicular to the surface where the detection site is located, and the hyperspectral imager is used to collect data from a predetermined height directly above the detection site to obtain a first hyperspectral image under a vertical viewing angle; the position of the drone and the shooting angle of the hyperspectral imager are adjusted to collect data from a position with an elevation angle of 45 degrees on one side of the detection site to obtain a second hyperspectral image under a first oblique viewing angle; the position of the detection site is adjusted to collect data from a position with an elevation angle of 45 degrees on the other side of the detection site to obtain a third hyperspectral image under a second oblique viewing angle; the sewage mixture is observed at different observation angles to obtain hyperspectral images of the sewage mixture to capture the spectral reflectance characteristics of the oil film at different wavelengths, and the reflectance spectrum of the collected image is compared with the spectrum map in the database. The reflectance spectrum in the database is as follows Figure 2 As shown, the component content ratio and oil film thickness of each oil product are calculated according to the method in steps 3 to 4, and the obtained data are compared with the oil content put into water in the experiment and the actual measured oil film thickness. The calculation error does not exceed 5%.

[0123] The steps in this application can be adjusted in order, combined, and deleted according to actual needs.

[0124] The units in the device of the present application can be combined, divided and deleted according to actual needs.

[0125] Although the present application is disclosed in detail with reference to the accompanying drawings, it should be understood that these descriptions are merely exemplary and are not intended to limit the application of the present application. The scope of protection of the present application is defined by the appended claims and may include various modifications, alterations and equivalents made to the invention without departing from the scope and spirit of the present application.

Claims

1. A hyperspectral remote sensing oil spill monitoring and analysis method, characterized in that: The hyperspectral remote sensing oil spill monitoring and analysis method comprises: Step 1: Use a hyperspectral imager to collect hyperspectral images of the detection site of the target area from a vertical perspective, a first oblique perspective, and a second oblique perspective, wherein the shooting positions of the hyperspectral imager at the three perspectives are in the same plane and are at equal distances from the target area, specifically including: Use the hyperspectral imager on the drone to collect data from a predetermined height directly above the detection site to obtain a first hyperspectral image at a vertical viewing angle, wherein the optical axis of the hyperspectral imager is vertically downward and perpendicular to the surface where the detection site is located; use the hyperspectral imager to collect data from the same height on one side of the original position of the drone to obtain a second hyperspectral image at a first oblique viewing angle, wherein the angle between the optical axis of the hyperspectral imager and the horizontal direction is 30° to 60°; use the hyperspectral imager to collect data from the same height on the other side of the original position of the drone to obtain a third hyperspectral image at a second oblique viewing angle, wherein the angle between the optical axis of the hyperspectral imager and the horizontal direction is 30° to 60°, and the optical axis of the hyperspectral imager at the first oblique viewing angle and the optical axis of the hyperspectral imager at the second oblique viewing angle are symmetrical about the vertical line of the detection site; Step 2: preprocessing the collected hyperspectral image and performing spectral correction on the preprocessed hyperspectral image; Step 3: extract endmembers from all spectral data contained in the hyperspectral image, establish a mixture model based on the extracted endmembers, and unmix the mixture model to obtain the components and content ratios corresponding to all pixels in the hyperspectral image, specifically including: Step 3.1, extract endmembers from the spectral data of all pixels in the hyperspectral image and establish a mixture model; Spectral features, including absorption peaks, reflection peaks, and spectral slopes, are extracted from the spectral data of each pixel. The extracted spectral features are matched with the known oil component spectral database to obtain the corresponding end members. The mixed model is established as follows: ； Where R is the original observation data matrix, M is the spectrum matrix of the endmembers in each pixel, where each column represents the spectrum of an endmember, A is the abundance matrix of different endmembers in each pixel, and N is the noise vector; In step 3.2, the mixed model is unmixed to obtain the endmember spectrum matrix and endmember abundance matrix corresponding to all pixels in the hyperspectral image, and the oil spill components and contents in the target area are statistically calculated as follows: Initialize the spectral matrix M and the abundance matrix A, and update the spectral matrix M and the abundance matrix A through iterative optimization so that their product approaches the original observation data matrix R. Set the stopping condition, which is to stop the iteration when the maximum number of iterations is reached, and update the endmember spectral matrix M: ； Update the abundance matrix A: ； In the formula, the symbol Indicates element-wise multiplication, the symbol It represents the point-by-point division of elements. E is a matrix whose internal values ​​are all 1. Based on the iterative spectral matrix M and abundance matrix A, the oil spill components and the content ratios of different components at each pixel are obtained. The subscript T represents the matrix transpose. Step 4: extracting reflectance spectra from the hyperspectral image, establishing an oil film thickness calculation model based on the reflectance spectra from multiple viewing angles, and estimating the oil film thickness based on the oil film thickness calculation model, specifically including: The pixel point at the geometric center of the hyperspectral image is taken as its center point, and the reflectance spectra of the center points of the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image are obtained, which is calculated as , where n represents the labels of different hyperspectral images corresponding to the same pixel, n=1,2,3; The reflectivity model of the oil film thickness is established based on the reflectivity spectra of the center points of the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image: ； Where, is the reflectance of the hyperspectral image of the control group, is the absorption coefficient of the oil film at wavelength λ, and d is the thickness of the oil film. The reflectivity model of the oil film thickness is optimized, and the reflectivity of different shooting angles is integrated to obtain the calculation model of the oil film thickness: ； Set the initial oil film thickness d0, substitute d0 into the formula and use the gradient descent method to perform optimized iterative calculations. Set the stopping condition to be that the change in d value is less than the set threshold, and get the final oil film thickness , traverse the oil film thickness of all detection sites in the target area; Step 5: Calculate the oil spill volume in the shooting area corresponding to the hyperspectral image based on the composition and content ratio of all pixels in the hyperspectral image at a vertical viewing angle, as well as the corresponding oil film thickness. Calculate the total oil spill volume based on the oil spill volume of all shooting areas in the target area. Step 6: Based on the endmember information of all vertical viewing angle hyperspectral images taken in the target area, all pixels with oil spills are found to obtain the total number of pixels in the oil spill area. The total area of ​​the oil spill area in the target area is calculated based on the total number of pixels in the oil spill area.

2. The hyperspectral remote sensing oil spill monitoring and analysis method according to claim 1, wherein: The step 2 specifically includes: Filtering the first hyperspectral image, the second hyperspectral image, and the third hyperspectral image collected by the hyperspectral imager to remove noise in the images; Using water bodies from other areas without oil pollution as the control group, a point was selected in the control area as the control point. The reflectance of the control point was measured using a hyperspectral imager. The reflectance was used for spectral correction to eliminate the influence of ambient light and sensor, ensuring the consistency and accuracy of the data. The details are as follows: ； Where, is the reflectance of the original hyperspectral image, is the reflectance of the hyperspectral image of the control group, is the reflectance of the corrected hyperspectral image, i and j represent the horizontal and vertical coordinates of the corrected point is the wavelength.

3. The hyperspectral remote sensing oil spill monitoring and analysis method according to claim 1, wherein: The step 5 specifically includes: The oil film thickness at the center point of the first hyperspectral image is taken as the oil film thickness of the entire shooting area corresponding to it. The oil spill volume of the area is calculated based on the area of ​​the shooting area and the corresponding oil film thickness. The oil spill volume of all shooting areas included in the target area is accumulated to obtain the total oil spill volume, which is as follows: ； Where, is the total oil spill volume in the target area, is the area of ​​the shooting area corresponding to the first hyperspectral image, is the oil film thickness at the center point of the first hyperspectral image, X is the total number of all shooting areas included in the target area, and x is the label of the shooting area included in the target area; The shooting area is divided into small blocks in equal proportion according to the position of the pixel points in the hyperspectral image. Each pixel point corresponds to a small block. Calculate the area s of the small area block, and calculate the oil film thickness of the corresponding shooting area according to the area s of the small area block , the oil spill components and content ratios corresponding to the pixel points, and the content of different oil spill components in the shooting area are calculated as follows: ； Where, is the content of a certain oil spill component in the shooting area, is the content ratio of a certain oil spill component in the pixel, y is the label of a certain oil spill component, C is the total number of all small area blocks in the shooting area, and c is the label of the small area blocks in the shooting area.

4. The hyperspectral remote sensing oil spill monitoring and analysis method according to claim 3, wherein: The step 6 specifically includes: All pixels with end members in the first hyperspectral image are used as pixels of the oil spill area. The first hyperspectral images corresponding to all shooting areas in the target area are traversed in the same way to calculate the total number of pixels in the oil spill area in the target area. The oil spill area is calculated based on the total number of pixels in the oil spill area and the area of ​​the small area blocks corresponding to the pixels, as follows: ； Where, is the oil spill area of ​​the target area, and P is the total number of pixels in the oil spill area.

5. A hyperspectral remote sensing oil spill monitoring platform for executing the hyperspectral remote sensing oil spill monitoring and analysis method according to any one of claims 1 to 4, characterized in that: The hyperspectral remote sensing oil spill monitoring platform includes a hyperspectral imager, a positioning system, a posture control system, a carrier and a host computer; The hyperspectral imager is used to collect hyperspectral images of the target area from different shooting angles; the positioning system is used to position the hyperspectral imager so that the position of the hyperspectral imager can be adjusted according to the position information fed back by the positioning system when collecting the hyperspectral image; the attitude control system is used to adjust the shooting angle of the hyperspectral imager to achieve multi-angle shooting of the target area; the carrier is used to carry the hyperspectral imager for data collection; The host computer is used to display, store and analyze the hyperspectral images collected by the hyperspectral imager.

Images