Underwater target detection correction method and system based on unmanned aerial vehicle-mounted hyperspectrum

Abstract

The invention discloses an underwater target detection and correction method and system based on unmanned aerial vehicle-mounted hyperspectrum, and the method innovatively integrates the water quality parameter and optical characteristic quantitative modeling and wave active offset technology, and forms a dual anti-interference mechanism. On one hand, the water quality parameters are converted into core optical characteristics influencing transmission of light in a water body, and a quantitative model is constructed to accurately describe propagation differences of light under different water qualities; and on the other hand, spectrum signal fluctuation and position offset caused by waves can be actively counteracted by generating suppression waves with opposite phases, the signal-to-noise ratio loss can be greatly reduced under the dual effects, the inherent reflection spectrum characteristics of the underwater target are completely reserved, and a high-fidelity data basis is provided for subsequent recognition.

Description

Technical Field

[0001] This invention relates to the field of underwater target detection and correction technology, specifically to an underwater target detection and correction method and system based on UAV-borne hyperspectral imaging. Background Technology

[0002] Underwater optical detection technology is a technology that uses optical sensors to capture light signals reflected from underwater targets and combines them with spectral analysis and other methods to achieve functions such as underwater target identification, topographic mapping, and ecological monitoring. It is widely used in marine engineering, underwater security, seabed resource development and other fields, and is an important supporting technology for underwater environmental perception.

[0003] However, this technology faces a core bottleneck: natural water bodies, such as lakes and oceans, are not stable optical media. Their optical properties are frequently affected by factors such as suspended particles, dissolved substances, and colored soluble organic matter, leading to varying degrees of light attenuation and scattering during transmission. This optical variability caused by water quality directly affects the authenticity and accuracy of the target's reflectance spectrum, thus causing the following three main problems with detection performance:

[0004] First, the signal attenuation is severe. The target reflected light is continuously absorbed and scattered by the water body during transmission to the sensor, resulting in a significant decrease in signal strength and signal-to-noise ratio, making it difficult to reliably capture effective optical information.

[0005] Secondly, spectral confusion is exacerbated. Different targets, such as reefs and mines, have inherently distorted reflectance spectra due to interference from water color and turbidity. They may exhibit similar apparent optical characteristics in specific spectral bands, leading to target misjudgment.

[0006] Third, it has poor cross-domain adaptability. Detection algorithms optimized for the optical conditions of a specific body of water are often difficult to transfer directly to other environments due to the spatial heterogeneity of water optical parameters, thus limiting the generalization ability of the technology. Secondly, wave motion, as a dynamic source of interference, further exacerbates the instability of the optical environment. Waves stir up the water, causing an imbalance in the distribution of suspended particles and resuspending bottom sediments, altering the spatiotemporal distribution of turbidity in the water. At the same time, it causes water surface fluctuations, underwater turbulence, and relative positional shifts between the sensor and the target, resulting in random changes in the light transmission path and dynamic fluctuations in the incident angle.

[0007] Therefore, how to effectively suppress the interference of water optical properties on detection signals and overcome the limitations of existing detection methods in adapting to water dynamics has become a key issue that urgently needs to be addressed to improve the accuracy and robustness of underwater optical detection. Summary of the Invention

[0008] This invention provides a method and system for underwater target detection and correction based on UAV-borne hyperspectral imaging, which effectively suppresses the interference of water optical properties on the detection signal and overcomes the limitations of existing detection methods and systems in terms of dynamic adaptation to water.

[0009] This invention provides the following technical solution: a method for underwater target detection and correction based on UAV-borne hyperspectral imaging, comprising: S1, apparent reflectance acquisition: acquiring the apparent reflectance of the target water area and pre-setting water depth parameters; S2, water quality parameter calculation: calculating the water quality parameters of the target water area; S3, apparent reflectance correction: based on the water quality parameters of the target water area and a pre-set water area-coefficient mapping library, acquiring the coefficients corresponding to the target water area, the coefficients being obtained by simultaneously solving the measured intrinsic optical parameter coefficients and measured water quality parameters of the same water area as the target water area, and correcting the apparent reflectance by combining the pre-set water depth parameters, the intrinsic optical parameters and coefficients of the standard water body; S4, wave interference suppression: decomposing the wave frequency components of the target water area to generate a suppression wave with opposite phase, superimposing the suppression wave with the water wave to cancel wave interference, and outputting the corrected spectral data.

[0010] This innovative underwater target detection and correction method based on UAV-borne hyperspectral imaging integrates quantitative modeling of water quality parameters and optical properties with active wave cancellation technology to form a dual anti-interference mechanism. By combining the correlation between water quality parameters and key optical property parameters, the weights of hyperspectral bands are dynamically optimized. On the one hand, water quality parameters are transformed into core optical properties that affect light transmission in water, and a quantitative model is constructed to accurately characterize the differences in light propagation under different water qualities. On the other hand, by generating suppression waves with opposite phases, the spectral signal fluctuations and position shifts caused by waves can be actively canceled. Under the dual effect, the signal-to-noise ratio loss can be significantly reduced, and the inherent reflectance spectral characteristics of underwater targets can be completely preserved, providing a high-fidelity data foundation for subsequent identification.

[0011] As an optional scheme of the underwater target detection and correction method based on UAV-borne hyperspectral imaging described in this invention, the water quality parameters include CODmn, CHL-a, NH3-N, and SS. The inherent optical parameters of the water body include absorption coefficient, backscattering coefficient, and attenuation coefficient. The standard water body is obtained by uniformly mixing water samples from several sea areas including the target water area. The construction of the water area-coefficient mapping library includes: S3.1, measuring CODmn, CHL-a, NH3-N, and SS of any water area in the four sea areas; S3.2, measuring the absorption coefficient, backscattering coefficient, and attenuation coefficient of any water area in the four sea areas; S3.3, calculating the fitting coefficients of the absorption coefficient and various water quality parameters by combining the measured absorption coefficient with its corresponding CODmn, CHL-a, NH3-N, and SS. The calculation process includes: ,in, The measured absorption coefficient is... These are the absorption constants of pure water. From the table, we can find that C1 is the measured CODmn, C5 is the measured CHL-a, C6 is the measured NH3-N, and C7 is the measured SS. This is the reference wavelength, typically 440nm, p1, p2, d1, s d n1, n2 and s n These are the fitting coefficients that need to be solved. It is related to chlorophyll a absorption by phytoplankton and can be measured experimentally. It is a spectral vector curve, which needs to be determined through experimental data. For each wavelength λ, a regression equation is established, and the nonlinear least squares method is used to fit the curve for all wavelengths. And the global parameter p2. In practical applications, The spectral shape of phytoplankton is generally considered to be relatively stable and can be used as a basis for scaling with a small amount of measurement data, without the need for independent fitting at each wavelength. It is related to the absorption of CODmn by colored dissolved organic matter and can be experimentally measured. Here, d1 and s d It is a constant. s d CODmn is the spectral slope of the absorption, describing the rate at which absorption decays with increasing wavelength. A nonlinear regression method is used to adjust the values ​​of d1 and sd so that the formula predicts the absorption at different wavelengths. It best matches the experimental measurements. Non-algal particulate matter absorption, related to suspended solids and ammonia nitrogen, can be experimentally measured, n1, n2 and s n All are constants.

[0012] S3.4. The measured backscattering coefficient is combined with its corresponding CHL-a and SS to calculate the fitting coefficients between the backscattering coefficient and various water quality parameters. The calculation process includes: ,in, The measured backscattering coefficient is... This is the backscattering constant of pure water. From the table, C5 is the measured CHL-a, C7 is the measured SS, and r1, r2, and r3 are the fitting coefficients to be solved. S3.5. The attenuation coefficient is calculated by simultaneously solving the attenuation coefficient with its corresponding measured absorption coefficient and measured backscattering coefficient to obtain the fitting coefficients between the attenuation coefficient and the water quality parameters. The calculation process includes: ,in, The attenuation coefficient is... and These are the measured absorption coefficient and backscattering coefficient, respectively. and S3.6 The fitting coefficients to be solved are defined in section S3.6. A mapping relationship is established by associating the fitting coefficients corresponding to any water body with the absorption coefficients and backscattering coefficients of the standard water body, forming a corresponding water body-coefficient mapping library.

[0013] As an optional scheme of the underwater target detection and correction method based on UAV-borne hyperspectral imaging described in this invention, the apparent reflectivity correction by combining preset water depth parameters, inherent optical parameters of a standard water body, and coefficients includes: obtaining the coefficients corresponding to the target water area, and solving the coefficients and water quality parameters of the target water area to obtain the simultaneous absorption coefficient, simultaneous backscattering coefficient, and simultaneous attenuation coefficient of the target water area. , among which, among which This is the corrected apparent reflectance. It is the apparent reflectance of the target water body. Let λ be the simultaneous absorption coefficient of the target water body at wavelength λ. Let λ be the backscattering coefficient of the target water body at wavelength λ. Let be the simultaneous attenuation coefficient of the target water body at wavelength λ. λ represents the measured absorption coefficient of a standard water body at a wavelength of λ. λ is the measured backscattering coefficient of a standard water body at wavelength λ, and z is the preset water depth parameter.

[0014] As an optional scheme of the underwater target detection and correction method based on UAV-borne hyperspectral imaging described in this invention, the suppression wave generation includes: S4.1, replacing the water surface height with the hyperspectral DN value, and modeling its variation with spatial points as a spatial-wavelength two-dimensional function of the water surface height, wherein the spatial-wavelength two-dimensional function of the water surface height is: Where (x,y) represents spatial coordinates, y is the push-broom direction, which implies time information. Hyperspectral imaging uses line push-broom imaging. At a frame rate of 50, the time of the first row of data is 0.02s, and the time of the 50th row is 1s; λ is the wavelength dimension of the hyperspectral image. This represents the actual reflectivity information; N represents the total number of basic sine waves that make up complex water ripples. The amplitude of the nth component is represented by , and the intensity of the wave is represented by . and It is a wave number vector, representing the direction and magnitude of wave propagation in space, corresponding to the wave number in the x and y directions respectively, and is related to the wavelength; S4.2. Perform a two-dimensional Fourier transform on a fixed wavelength to analyze its frequency components; S4.3. Obtain pixels whose hyperspectral DN values ​​deviate from the average value by more than a threshold; S4.4. Generate a wave with opposite phase and the same amplitude as a suppression wave based on each pixel.

[0015] A system applying any of the above-mentioned underwater target detection and correction methods based on UAV-borne hyperspectral imaging includes:

[0016] Apparent reflectance acquisition module: acquires the apparent reflectance of the target water area and presets water depth parameters;

[0017] Water quality parameter calculation module: Calculates the water quality parameters of the target water area;

[0018] Apparent reflectance correction module: Based on the water quality parameters of the target water area and the preset water area-coefficient mapping library, obtain the coefficients corresponding to the target water area. The coefficients are obtained by solving the measured inherent optical parameter coefficients and measured water quality parameters of the same water area as the target water area. The apparent reflectance is corrected by combining the preset water depth parameters, the inherent optical parameters and coefficients of the standard water body.

[0019] Wave interference suppression module: After decomposing the wave frequency components of the target water area, it generates a suppression wave with opposite phase. The suppression wave is superimposed with the water wave to cancel out the wave interference and output the corrected spectral data.

[0020] The present invention has the following beneficial effects:

[0021] 1. This innovative underwater target detection and correction method based on UAV-borne hyperspectral imaging integrates quantitative modeling of water quality parameters and optical properties with active wave cancellation technology, forming a dual anti-interference mechanism. By combining the correlation between seven water quality parameters (CODmn, CHL-a, NH3-N, and SS) and key optical property parameters (absorption coefficient a(λ) and backscattering coefficient b(λ), the hyperspectral band weights are dynamically optimized. On the one hand, this method transforms water quality parameters into core optical properties such as absorption and scattering that affect light propagation in water, constructing a quantitative model to accurately characterize the differences in light propagation under different water qualities. On the other hand, by decomposing wave frequency components through Fourier transform and generating suppressor waves with opposite phases, it can actively cancel the spectral signal fluctuations and positional shifts caused by waves. This dual effect significantly reduces the signal-to-noise ratio loss, fully preserving the inherent reflectance spectral characteristics of underwater targets and providing a high-fidelity data foundation for subsequent identification.

[0022] 2. This underwater target detection and correction method based on UAV-borne hyperspectral imaging overcomes scenario limitations compared to traditional fixed-band, single-parameter compensation methods through two major design improvements. First, based on measured data of water quality parameters and inherent optical parameters from multiple sea areas including the Bohai Sea, Yellow Sea, East China Sea, and South China Sea, it fits the specific optical model coefficients for each sea area and constructs a model mapping library. During actual measurements, matching coefficients can be called according to the target water area to adapt to the differences in water quality across different sea areas. Second, by presetting a water depth parameter z (e.g., 7.5 meters for mines and 1.5 meters for underwater lightning protection networks), and dynamically adjusting the reflectivity correction formula based on water quality parameters, it can adapt to the depth distribution characteristics of different targets and cope with the spatiotemporal changes in water quality within the same sea area, ensuring stable accuracy in nearshore, offshore, and different target detection scenarios.

[0023] 3. The optical characteristic correction method in this UAV-based hyperspectral underwater target detection correction method utilizes a well-defined quantization formula to automatically normalize reflectance for different water depths (z) and water quality vectors (C). For waters with complex water quality and severe optical attenuation, this formula can accurately eliminate optical interference from water quality parameters. For waters with relatively stable water quality, data normalization can also be achieved through the standard water quality vector (C0). Simultaneously, during wave suppression, the dominant interference mode is screened through a signal strength threshold to ensure controllable cancellation effect. The final output correction data can be directly used for spectral feature extraction and localization of various target types, fully meeting the stringent requirements for detection accuracy in underwater security, resource exploration, and other scenarios. Attached Figure Description

[0024] Figure 1 This is the original hyperspectral image before correction in Embodiment 1 of the present invention.

[0025] Figure 2 This is a comparison diagram of reflectivity before and after the correction in Embodiment 1 of the present invention.

[0026] Figure 3 for Figure 1 A schematic diagram showing the suppression of waves using the method of Example 1.

[0027] Figure 4 for Figure 1 A schematic diagram of the recognition results after correction using the method of Example 1. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Example 1

[0030] Please see Figures 1 to 4 One of the underwater target detection and correction methods based on UAV-borne hyperspectral imaging includes:

[0031] S1. Apparent reflectance acquisition: through unmanned aerial vehicle (UAV) systems and auxiliary means, such as... Figure 1 The system captures hyperspectral images to obtain the apparent reflectance of the target water area, providing complete raw data for subsequent calibration. The apparent reflectance can be obtained by acquiring raw spectral data of the target water area using a hyperspectral sensor mounted on a UAV. The reflectance can be calculated by mounting a zenith optical module on the UAV system or placing a reference plate with known reflectance on the ground before takeoff, and then comparing it with the raw spectral data. A water depth parameter z is then preset. When setting the water depth parameter z, the characteristics of the target are considered. For example, the typical deployment depth of mines is 5-10 meters, so z = 7.5 meters is set; underwater lightning protection networks are mostly distributed at depths of 0-3 meters, so z = 1.5 meters is set; for other targets, the z value is determined based on prior depth information.

[0032] S2. Water quality parameter calculation: Calculate the water quality parameters of the target water area. The water quality parameters of the target water area can be calculated through spectral data, such as based on empirical methods or machine learning methods, or by sending multi-point sampling to the laboratory for analysis and calculation.

[0033] S3. Apparent Reflectance Correction: Based on the water quality parameters of the target water area and the preset water area-coefficient mapping library, the coefficients corresponding to the target water area are obtained. The coefficients are obtained by solving the measured intrinsic optical parameter coefficients and measured water quality parameters of the same water area as the target water area. The apparent reflectance is corrected by combining the preset water depth parameters, the intrinsic optical parameters and coefficients of the standard water body.

[0034] In this embodiment, the standard water body was obtained by uniformly mixing multiple water samples collected from four sea areas: Bohai Sea, Yellow Sea, East China Sea, and South China Sea. The target water body was one of the four sea areas: Bohai Sea, Yellow Sea, East China Sea, and South China Sea. The water quality parameters CODmn, CHL-a, NH3-N, and SS of each standard water body were measured under laboratory conditions, as well as the inherent optical parameters of the water body, absorption coefficient, backscattering coefficient, and attenuation coefficient, to obtain the basic measured data for modeling.

[0035] The construction of the pre-defined water area-coefficient mapping library includes:

[0036] S3.1. The CODmn, CHL-a, NH3-N and SS of any water body in the four major sea areas were measured. The water quality parameters were obtained by chemical analysis using high-precision instruments in the laboratory to obtain the concentration of the four core water quality indicators. The water quality parameter vector C = [C1, C5, C6, C7]T = [CODmn, CHL-a, NH3-N, SS]ᵀ was constructed.

[0037] S3.2. Measure the absorption coefficient, backscattering coefficient, and attenuation coefficient of any water body in the four major sea areas.

[0038] absorption coefficient The actual measurements included: measurements obtained using a water absorption-attenuation measuring instrument. Data. It can be accessed through... The formula is used for calculation, where L is the path length of light in the sample. It is the initial light intensity. It is the light intensity collected by the reflecting cavity.

[0039] Backscattering coefficient Actual measurements included direct measurements using professional backscattering measurement equipment, such as WET Labs' BB series sensors.

[0040] Attenuation coefficient The actual measurements include: values ​​obtained using an indoor hyperspectral testing device based on the Lambert-Beer law. These values ​​are then analyzed using the formula... Where L is the path length of light in the sample. It is the initial light intensity. It is the intensity of transmitted light after passing through the water sample.

[0041] Multiple sets of actual measurements , and Substitute the corresponding water quality parameters into the theoretical model, and fit the optical model coefficients specific to the sea area using methods such as the least squares method.

[0042] Specifically, it includes:

[0043] S3.3. The measured absorption coefficient is combined with its corresponding CODmn, CHL-a, NH3-N, and SS to calculate the fitting coefficients between the absorption coefficient and various water quality parameters. The combined calculation includes:

[0044]

[0045] in, The measured absorption coefficient is... These are the absorption constants of pure water. From the table, we can find that C1 is the measured CODmn, C5 is the measured CHL-a, C6 is the measured NH3-N, and C7 is the measured SS. This is the reference wavelength, typically 440nm, p1, p2, d1, s d n1, n2 and s n These are the fitting coefficients that need to be solved. The fitting coefficients for the corresponding water area are obtained by solving for the measured absorption coefficients and water quality parameters of multiple sets of samples from the same water area.

[0046] S3.4. The measured backscattering coefficient, its corresponding CHL-a, and SS are combined to calculate the fitting coefficients between the backscattering coefficient and the various water quality parameters. The combined calculation includes:

[0047]

[0048] in, The measured backscattering coefficient is... The backscattering constant of pure water is given by the table. C5 is the measured CHL-a, C7 is the measured SS, and r1, r2, and r3 are the fitting coefficients to be solved. The fitting coefficients for the corresponding water area are obtained by solving the measured backscattering coefficients and water quality parameters of multiple samples from the same water area.

[0049] S3.5. The attenuation coefficient is combined with its corresponding measured absorption coefficient and measured backscattering coefficient to calculate the fitting coefficients between the attenuation coefficient and the water quality parameters. This combined calculation includes:

[0050]

[0051] in, The attenuation coefficient is... and These are the measured absorption coefficient and backscattering coefficient, respectively. and These are the fitting coefficients that need to be solved. The fitting coefficients for the corresponding water area are obtained by solving for the measured attenuation coefficients and water quality parameters corresponding to multiple sets of samples from the same water area.

[0052] S3.6. Establish a mapping relationship between the fitting coefficients corresponding to any water body and the absorption coefficients and backscattering coefficients of the standard water body to form a corresponding water body-coefficient mapping library. Define the baseline concentration vector of the standard water body as C0 = [C01,C05,C06,C07]ᵀ, and obtain standard water body samples by uniformly mixing all samples. Obtain the baseline concentration vector C0 and the absorption coefficients of the standard water body by measuring the standard water body samples in the laboratory. and backscattering coefficient The specific coefficients {p1,p2,d1,s} obtained by fitting the data to each sea area d ,n1,n2,s n The functions r1, r2, r3, m(λ), n(λ)} are stored in association with the sea area and simultaneously associated with standard water body parameters. and This forms a complete model mapping library for use during actual measurement and calibration.

[0053] The apparent reflectivity correction, which combines preset water depth parameters, inherent optical parameters of a standard water body, and coefficients, includes: obtaining the coefficients corresponding to the target water body; and solving for the simultaneous absorption coefficient, simultaneous backscattering coefficient, and simultaneous attenuation coefficient of the target water body by combining the coefficients with the water quality parameters of the target water body.

[0054]

[0055] Among them, This is the corrected apparent reflectance. It is the apparent reflectance of the target water body. Let λ be the simultaneous absorption coefficient of the target water body at wavelength λ. Let λ be the backscattering coefficient of the target water body at wavelength λ. Let be the simultaneous attenuation coefficient of the target water body at wavelength λ. λ represents the measured absorption coefficient of a standard water body at a wavelength of λ. λ is the measured backscattering coefficient of a standard water body at wavelength λ, and z is the preset water depth parameter.

[0056] S4. Wave Interference Suppression: After decomposing the wave frequency components of the target water area, a suppression wave with opposite phase is generated. The suppression wave is then superimposed with the water wave to cancel out wave interference, and the corrected spectral data is output. The generation of the suppression wave includes:

[0057] S4.1. Replace the water surface height with the hyperspectral DN value, and model its variation with spatial points as a two-dimensional spatial-wavelength function of the water surface height. The two-dimensional spatial-wavelength function of the water surface height is:

[0058] Where (x,y) represents spatial coordinates, y is the push-broom direction, which implies time information. Hyperspectral imaging uses line push-broom imaging. At a frame rate of 50, the time of the first row of data is 0.02s, and the time of the 50th row is 1s; λ is the wavelength dimension of the hyperspectral image. This represents the actual reflectivity information; N represents the total number of basic sine waves that make up complex water ripples. The amplitude of the nth component is represented by , and the intensity of the wave is represented by . and It is a wave number vector, representing the direction and magnitude of wave propagation in space, corresponding to the wave number in the x and y directions respectively, and is related to the wavelength; Let n be the phase of the nth component, representing the initial offset of the wave.

[0059] S4.2 Perform a two-dimensional Fourier transform on a fixed wavelength and analyze its frequency components, including:

[0060]

[0061] in, Let i be the complex-valued function after Fourier transform, and i be the imaginary unit. In practical calculations, the Discrete Fourier Transform or the Fast Fourier Transform is used.

[0062] S4.3. Obtain pixels whose hyperspectral DN values ​​deviate from the average value by more than a threshold, and identify the dominant mode with the highest energy, i.e., the principal component, including:

[0063]

[0064] The formula above represents a set containing M elements, each of which is a complete set of wave characteristic parameters for a hyperspectral pixel. M is the number of dominant components, determined by the signal intensity threshold for each band of the hyperspectral image. For each band λ, the average intensity of the DN values ​​of all pixels is calculated, and then a threshold is set based on the degree of deviation from the average intensity; for example, only points with energy higher than twice the average energy are selected. Only pixels whose DN values ​​deviate from the average by more than the threshold are considered as dominant modes that need to be suppressed; therefore, M represents the number of pixels that need to be processed in that band.

[0065] S4.4. Generate a wave with opposite phase and the same amplitude for each pixel as a suppression wave. After superimposing the suppression wave with the water wave to cancel out wave interference, output the corrected spectral data.

[0066] The suppressed wave is:

[0067]

[0068] The suppression wave and the water wave superimpose and cancel each other out:

[0069]

[0070] The main components were completely neutralized, leaving only unidentified and untreated minor components.

[0071] Finally, the corrected core data is standardized and output, and detection applications are carried out in combination with the target reflectivity characteristics. By setting the z-value and correction data at the corresponding depth, the spectral features of each target can be extracted simultaneously, realizing the classification and positioning of multiple types of targets, and fully meeting the actual needs of underwater target search and detection.

[0072] This application can be applied to the military field, mainly addressing two pain points: anti-counterfeiting identification and environmental interference. In terms of anti-counterfeiting, underwater environments naturally possess camouflage characteristics, and hyperspectral imaging inherently has significant advantages in terms of materials and anti-counterfeiting. Regarding environmental interference, differences in water bodies and sea surface waves have a significant impact. For example, if a large amount of effort is spent training and optimizing an identification model in the East China Sea, the model may not be universally applicable in the South China Sea due to differences in water quality, requiring the investment of data to remodel and optimize. Water quality also changes in different seasons due to ocean current variations, and the differences in water bodies are mainly concentrated in their inherent optical parameters.

[0073] Existing theories suggest an intrinsic link between inherent optical parameters and water pollution, such as water quality parameter concentrations, and these parameters are typically applied to civilian water quality monitoring and large-area satellite water quality inversion. This proposed solution, however, employs a reverse derivation method. It uses water quality parameter concentrations retrieved via hyperspectral inversion to infer the inherent optical parameters, correcting for differences in various water bodies. This approach represents a significant innovation in its application logic.

[0074] The second environmental interference is the significant impact of sea waves. Maritime weather is highly variable, and waves also reflect tidal changes. Eliminating wave interference is an urgent need at the application level. Hyperspectral data can be decomposed into grayscale images wavelength by wavelength. In pushbroom imaging mode, spatial dimensions implicitly contain temporal information. By processing the image using Fourier transform algorithms, wave interference can be minimized. De-wave processing of hyperspectral images is also a significant innovation in both technology and application.

[0075] Example 2

[0076] A system for underwater target detection and correction based on UAV-borne hyperspectral imaging, as described in Example 1, includes:

[0077] Apparent reflectance acquisition module: acquires the apparent reflectance of the target water area and presets water depth parameters;

[0078] Water quality parameter calculation module: Calculates the water quality parameters of the target water area;

[0079] Apparent reflectance correction module: Based on the water quality parameters of the target water area and the preset water area-coefficient mapping library, obtain the coefficients corresponding to the target water area. The coefficients are obtained by solving the measured inherent optical parameter coefficients and measured water quality parameters of the same water area as the target water area. The apparent reflectance is corrected by combining the preset water depth parameters, the inherent optical parameters and coefficients of the standard water body.

[0080] Wave interference suppression module: After decomposing the wave frequency components of the target water area, it generates a suppression wave with opposite phase. The suppression wave is superimposed with the water wave to cancel out the wave interference and output the corrected spectral data.

[0081] This system addresses the challenges posed by the complex and variable optical properties of natural water bodies due to factors such as suspended matter and chemical substances. Furthermore, waves not only agitate the water and exacerbate the dynamic fluctuations in optical parameters but also cause surface undulations and underwater turbulence. This results in uncertain attenuation and scattering of light signals during transmission in the water, leading to distortion of the target reflection spectrum, reduced signal-to-noise ratio, confusion of spectral characteristics of different targets, and the inability of existing methods to dynamically adapt to different aquatic environments, ultimately affecting the accuracy of underwater target identification and detection.

[0082] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0083] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for unmanned aerial hyperspectral-based underwater target detection and correction, characterized in that, The method comprises the following steps: S1, apparent reflectance acquisition: acquiring the apparent reflectance of the target water area, and presetting a water depth parameter; S2, water quality parameter calculation: calculating the water quality parameters of the target water area; S3, apparent reflectance correction: based on the water quality parameters of the target water area and the preset water area-coefficient mapping library, obtaining the corresponding coefficient of the target water area, the coefficient being obtained by simultaneously solving the measured inherent optical parameter coefficient and the measured water quality parameter of the same water area as the target water area, and combining the preset water depth parameter, the water body inherent optical parameter of the standard water body and the coefficient to correct the apparent reflectance; S4, wave interference suppression: decomposing the wave frequency component of the target water area to generate a phase-opposite suppression wave, superimposing the suppression wave and the water wave to offset the wave interference, and then outputting the corrected spectral data. 2.The unmanned aerial hyperspectral-based underwater target detection correction method according to claim 1, characterized in that: The water quality parameters include CODmn, CHL-a, NH3-N and SS, the water body inherent optical parameters include absorption coefficient, backscattering coefficient and attenuation coefficient, and the standard water body is obtained by uniformly mixing water samples of several sea areas containing the target water area.

3. The UAV-based hyperspectral underwater target detection correction method according to claim 1 or 2, characterized in that: The construction of the water area-coefficient mapping library comprises the following steps: S3.1, measuring CODmn, CHL-a, NH3-N and SS of any water area in the four sea areas; S3.2, measuring absorption coefficient, backscattering coefficient and attenuation coefficient of any water area in the four sea areas; S3.3, calculating the fitting coefficients of absorption coefficient and water quality parameters by simultaneously calculating the measured absorption coefficient and its corresponding CODmn, CHL-a, NH3-N and SS; S3.4, calculating the fitting coefficients of backscattering coefficient and water quality parameters by simultaneously calculating the measured backscattering coefficient and its corresponding CHL-a and SS; S3.5, calculating the fitting coefficients of attenuation coefficient and water quality parameters by simultaneously calculating the measured absorption coefficient and the measured backscattering coefficient corresponding to the attenuation coefficient; S3.6, establishing a mapping relationship by associating the fitting coefficients of any water area with the absorption coefficient and the backscattering coefficient of the standard water body to form a corresponding water area-coefficient mapping library.

4. The UAV-based hyperspectral underwater target detection correction method of claim 3, wherein: The simultaneous calculation of the measured absorption coefficient and its corresponding CODmn, CHL-a, NH3-N and SS comprises: ， wherein, is the measured absorption coefficient, is the pure water absorption constant, which can be obtained from the table, C1 is the measured CODmn, C5 is the measured CHL-a, C6 is the measured NH3-N, C7 is the measured SS, is the reference wavelength, usually 440 nm, p1, p2, d1, s d , n1, n2 and s n are the fitting coefficients to be solved.

5. The UAV-based hyperspectral detection method for underwater targets according to claim 3, characterized in that: The simultaneous calculation of the measured backscattering coefficient and its corresponding CHL-a and SS comprises: ， wherein, is the measured backscattering coefficient, is the pure water backscattering constant, which can be obtained from a table, C5 is the measured CHL-a, C7 is the measured SS, and r1, r2 and r3 are the fitting coefficients to be solved.

6. The UAV-borne hyperspectral-based underwater target detection correction method according to claim 3, characterized in that: The simultaneous calculation of the attenuation coefficient and the measured absorption coefficient and the measured backscattering coefficient corresponding to the attenuation coefficient comprises: ， wherein is the attenuation coefficient, and are the measured absorption and backscatter coefficients, respectively, and are the fitting coefficients to be solved for.

7. The UAV-borne hyperspectral-based underwater target detection correction method according to claim 4, characterized in that: The correction of the apparent reflectance by combining the preset water depth parameter, the water body inherent optical parameter of the standard water body and the coefficient comprises: obtaining the simultaneous absorption coefficient, the simultaneous backscattering coefficient and the simultaneous attenuation coefficient of the target water area by simultaneously solving the coefficient and the water quality parameters of the target water area, ， wherein, is the corrected apparent reflectance, is the apparent reflectance of the target water body, is the concurrent absorption coefficient of the target water body at wavelength λ; is the backscattering coefficient of the target water body at wavelength λ; is the concurrent attenuation coefficient of the target water body at wavelength λ, is the measured absorption coefficient of the standard water body at wavelength λ; is the measured backscattering coefficient of the standard water body at wavelength λ, and z is a preset water depth parameter. 8.The UAV-borne hyperspectral-based underwater target detection correction method of claim 1, wherein: The generation of the suppression wave comprises: S4.1, replacing the water surface height with the hyperspectral DN value, and modeling the water surface height as a spatial-wavelength two-dimensional function with the spatial point changing; S4.2, performing two-dimensional Fourier transform on the fixed wavelength to analyze the frequency component; S4.3, obtaining the pixel points whose hyperspectral DN value deviates from the average value by more than a threshold value; S4.4, generating a wave with opposite phase and same amplitude as the suppression wave according to each pixel point.

9. The UAV-borne hyperspectral-based underwater target detection correction method according to claim 8, characterized in that: The spatial-wavelength two-dimensional function of the water surface height is: where (x, y) represents spatial coordinates, y is the push-broom direction, which implies time information, hyperspectral uses line push-broom imaging, and in the case of 50 frame frequencies, the time of the first row of data is 0.02 s, and the time of the 50th row is 1 s; λ is the wavelength dimension of the hyperspectral image; represents the real reflectivity information; N represents the total number of basic sinusoidal waves constituting the complex water ripple; represents the amplitude of the nth component, and represents the intensity of the wave; and is a wave vector, which represents a vector of the propagation direction and size of the wave in space, and corresponds to the wave number in the x and y directions, respectively, which is related to the wavelength; is the phase of the nth component, which represents the initial offset of the wave.

10. A system for applying the method for calibrating underwater target detection based on unmanned aerial hyperspectral imaging according to any one of claims 1-9, characterized in that it comprises: The method comprises the following steps: An apparent reflectance acquisition module: acquiring the apparent reflectance of the target water area and presetting a water depth parameter; A water quality parameter calculation module: calculating the water quality parameter of the target water area; An apparent reflectance correction module: based on the water quality parameter of the target water area and the preset mapping library of water area-coefficients, obtaining the corresponding coefficient of the target water area, the coefficient being obtained by simultaneously solving the measured inherent optical parameter coefficient and the measured water quality parameter of the same water area as the target water area, and correcting the apparent reflectance in combination with the preset water depth parameter, the water inherent optical parameter of the standard water body and the coefficient; A wave interference suppression module: decomposing the wave frequency component of the target water area to generate a phase-opposed suppression wave, superimposing the suppression wave and the water wave to offset the wave interference and output the corrected spectral data.

Images