Wavelength-driven underwater image enhancement and super-resolution method and device

Through the wavelength-driven method, adaptive wavelength light source and multi-wavelength image acquisition, combined with spectral analysis and deep learning, the environmental adaptability and noise problems of underwater image enhancement and super-resolution methods in the prior art are solved, achieving high Quality image processing and detail recovery.

CN119027354AInactive Publication Date: 2024-11-26CHANGZHOU ZHIGAN TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411125115.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure CN119027354A_ABST
    Figure CN119027354A_ABST
Patent Text Reader

Abstract

The invention discloses a wavelength-driven underwater image enhancement and super-resolution method and device, and belongs to the technical field of image processing. The method comprises the following steps: S1, initial environment monitoring and sensor data acquisition are carried out, and adaptive wavelength selection is carried out; s2, real-time data monitoring and multi-wavelength image acquisition are carried out; s3, performing spectral analysis and adaptive correction on the acquired multi-wavelength images according to different wavelengths to sequentially obtain adaptive initial images with different wavelengths; s4, performing dynamic denoising and preprocessing on the initial images with different wavelengths in sequence; s5, super-resolution processing is carried out on the image after preprocessing and dynamic denoising; s6, synthesizing the images with different wavelengths; and S7, outputting the synthesized image, and feeding back image parameters to the step S1. Through an adaptive wavelength driving algorithm, real-time image enhancement and super-resolution processing are realized, and the real-time requirements of underwater detection, monitoring and navigation are met.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a wavelength-driven underwater image enhancement and super-resolution method and device. Background Art

[0002] Wavelength-driven is a method that uses the wavelength information of light to process and optimize images. Especially in the context of underwater image processing, wavelength-driven technology enhances image quality and resolution by analyzing and utilizing the characteristics of light of different wavelengths in underwater environments. The core idea of ​​this method is to optimize images based on the propagation characteristics of light in water.

[0003] Current underwater image enhancement and super-resolution methods mainly include the following mainstream technologies: image processing technology, physical model method, deep learning method and optical compensation method. Although the current underwater image enhancement and super-resolution methods have solved the underwater image quality problem to a certain extent, each method has its limitations. Image processing technology cannot effectively handle complex color distortion and scattering problems. In a high-noise environment, the effect is limited and detail information is easily lost. The generalization ability of deep learning methods for different underwater environments has certain limitations and needs to be adjusted and optimized for different scenes. The optical compensation method has high requirements for the selection and configuration of light sources and filters, and different environments require different configurations. In addition, these methods are too dependent on special equipment, and there are relatively large errors in color correction, and the resolution and environmental adaptability are not high. Summary of the invention

[0004] The purpose of the present invention is to provide a wavelength-driven underwater image enhancement and super-resolution method and device to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solution: a wavelength-driven underwater image enhancement and super-resolution method, the method comprising the following steps:

[0006] S1, initial environment monitoring and sensor data collection, adaptive wavelength selection through adaptive wavelength light source;

[0007] S2, real-time data monitoring and multi-wavelength image acquisition;

[0008] S3, performing spectral analysis and adaptive correction on the collected multi-wavelength images according to different wavelengths, and obtaining adaptive initial images with different wavelengths in turn;

[0009] S4, performing dynamic denoising and preprocessing on initial images with different wavelengths in sequence;

[0010] S5, performing super-resolution processing on the image that has undergone preprocessing and dynamic denoising;

[0011] S6, synthesizing multiple images with different wavelengths;

[0012] S7, output the synthesized image, and feed back the image parameters to step S1;

[0013] The sensors include a turbidity sensor, a light intensity sensor and a temperature sensor; the turbidity sensor is used to measure the concentration of suspended particles in the water to determine the turbidity of the water; the light intensity sensor is used to measure the underwater light intensity to determine the current light conditions; the temperature sensor is used to measure the water temperature;

[0014] The adaptive wavelength light source achieves real-time wavelength selection by adjusting the intensity of LED light sources with different wavelengths.

[0015] In step S1,

[0016] The adaptive wavelength selection is a method for dynamically adjusting the wavelength combination of the light source according to the real-time underwater environmental parameters; the method aims to optimize the acquisition quality of underwater images, reduce the absorption and scattering effects of light in water by selecting the best wavelength combination, and improve the clarity and contrast of the image; the specific steps are:

[0017] S2-1. Preliminary selection of wavelengths: Based on the existing underwater spectral characteristic data, a set of preliminary wavelengths is selected, including multiple wavelengths in the visible light range;

[0018] S2-2, parameter model: Establish a model based on environmental parameters to calculate the effectiveness of each wavelength in the current environment. The model takes into account the light absorption coefficient a(λ,t) and scattering coefficient b(λ,I), which change with the wavelength and underwater environmental parameters; the attenuation formula of light in water is:

[0019] I(λ,d)=I0(λ)·e -(a(λ,t)+b(λ,I)·d

[0020] Where I(λ,d) is the intensity of light with wavelength λ at light distance d; I0(λ) is the initial light intensity; a(λ,t is the absorption coefficient of wavelength λ at turbidity t; b(λ,I) is the scattering coefficient of wavelength λ at light intensity I; d is the distance the light propagates;

[0021] The absorption coefficient a(λ,t) and scattering coefficient b(λ,I) are calculated as:

[0022]

[0023] where a0 and b0 are the reference absorption and scattering coefficients at wavelength λ; k a and k bis a constant related to the environmental parameters turbidity t and light intensity I, k a =0.2, k b =0.1;

[0024] S2-3. Calculation of the best wavelength combination: The optimization algorithm is used to find the wavelength combination that makes the light propagation effect in water optimal under the current environmental parameters; the objective function is to analyze the received synthetic image and establish a comprehensive evaluation index of image quality, including image contrast and clarity:

[0025] Qualityλ)=∑(Contrast(p)+Sharpness(p))

[0026] Among them, Contrast(p) and Sharpness(p) represent the contrast and sharpness of the image corresponding to the wavelength λ, respectively;

[0027] The optimal wavelength combination is to find the wavelength combination that makes the light propagation effect in water optimal under the current environmental parameters through an optimization algorithm;

[0028] The underwater environmental parameters include turbidity and light intensity.

[0029] In step S3-2,

[0030] The contrast of the image uses the mean contrast, specifically:

[0031]

[0032] Where z is the total number of pixels in the image; is the mean value of the image pixels; p n is the nth pixel value in the image;

[0033] The clarity of the image is calculated using the mean of the image gradient; specifically:

[0034] Calculate the gradient of the image: Use the Sobel operator to calculate the gradient of the image in the x and y directions;

[0035] G x (x,y)=p(x+1,y-1)+2p(x+1,y)+p(x+1,y+1)-(p(x-1,y-1)

[0036] +2p(x-1,y)+p(x-1,y+1))

[0037] G y (x,y)=p(x-1,y+1)+2p(x,y+1)+p(x+1,y+1)-(p(x-1,y-1)

[0038] +2p(x,y-1)+p(x+1,y-1))

[0039] Among them, p(x,y) is the pixel value of the image at position (x,y), G x (x,y) and G y (x, y) are the gradients of the image in the x and y directions at the (x, y) position respectively;

[0040] Calculate the magnitude of the gradient: Combine the gradients in the x-direction and the y-direction to calculate the gradient magnitude of each pixel;

[0041]

[0042] Where G(x,y) is the magnitude of the gradient at position (x,y);

[0043] Calculate the mean of the gradient amplitude: average the gradient amplitudes of all pixels to get the image clarity index; the larger the gradient amplitude, the sharper the image edge, that is, the clearer it is;

[0044]

[0045] Where M and N are the width and height of the image respectively.

[0046] In step S2,

[0047] Multi-wavelength image acquisition is the process of using a multi-spectral camera and an adaptive wavelength light source to illuminate underwater target objects with light of different wavelengths to collect multiple image data. The specific steps of multi-wavelength image acquisition are:

[0048] S4-1. Setting the wavelength of the light source: adjusting the output wavelength and intensity of each LED light source of the adaptive wavelength light source according to the selected optimal wavelength combination;

[0049] S4-2, start the multispectral camera: start the multispectral camera and configure the camera parameters so that it can capture images at different wavelengths;

[0050] S4-3, collecting images wavelength by wavelength: for each selected wavelength, adjusting the light source to emit light of the wavelength; collecting images at the wavelength and saving them in a memory;

[0051] S4-4, switch to the next wavelength, and repeat the above process until the image acquisition of all wavelengths is completed.

[0052] In step S3,

[0053] Spectral analysis is based on the physical properties of light propagation in water. Light of different wavelengths has different absorption and scattering coefficients in water. By analyzing these coefficients, we can understand the propagation of light in water and provide data support for subsequent adaptive correction.

[0054] The adaptive correction includes adaptive color correction and scattering compensation;

[0055] Color correction is to compensate for the absorption differences of different wavelengths of light in water, making the color of the image closer to the real situation; the correction formula is as follows:

[0056] I'(λ)=I(λ)·e (a (λ)+b(λ) )·d

[0057] Where, I'(λ) is the corrected light intensity; I(λ) is the initial light intensity;

[0058] Scattering compensation is to reduce the impact of scattering in water on the image and increase the clarity of the image; the scattering compensation formula is as follows:

[0059]

[0060] Among them, I comp (λ) is the light intensity after scattering compensation; k s is the scattering coefficient;

[0061] An adaptively corrected initial image p(λ) with a wavelength of λ is obtained through adaptive correction.

[0062] In step S4,

[0063] Dynamic denoising is to remove the noise in the above initial image p(λ);

[0064] The dynamic denoising adopts the mean filtering method to remove high-frequency noise by calculating the mean of each pixel neighborhood in the image; the formula is as follows:

[0065]

[0066] Where p'((x,y)|λ) is the pixel value of the image at position (x,y) with a wavelength of λ and after denoising; p((x+i,y+j)|λ) is the pixel value of the initial image at position (x,y) with a wavelength of λ and after adaptive correction; Q is the neighborhood of position (x,y);

[0067] The image preprocessing includes image normalization and image sharpening;

[0068] The formula for image normalization is as follows:

[0069]

[0070] Among them, p" (λ) is the image with wavelength λ and normalized; p' (λ) is the image with wavelength λ and denoised; p' min(λ) and p' max (λ) are the minimum and maximum pixel values ​​in p'(λ), respectively;

[0071] Image sharpening is to enhance the edges in the image to improve the details and clarity of the image; using the Laplace operator; the formula is as follows:

[0072] P(λ)=p”(λ)+α·Δp”(λ)

[0073] Where p"'(λ) is the sharpened image with wavelength λ; α is the sharpening intensity coefficient; Δp"(λ) is the Laplace transform result of p"(λ);

[0074] p"'(λ) is the image with wavelength λ after denoising and preprocessing.

[0075] In step S5,

[0076] Super-resolution processing is to reconstruct low-resolution images to improve image resolution. Through machine learning methods: using convolutional neural networks, learn the mapping relationship from low resolution to super-resolution from a large amount of training data, and through multiple convolutional layers, learn the high-level features of the image to perform super-resolution processing on the image.

[0077] Input image: p''(λ)∈R M×N×C , where p''(λ) is the image with wavelength λ and after denoising and preprocessing; M, N and C are the width, height and number of channels of the image respectively, p''(λ) is an RGB image, so the number of channels C = 3;

[0078] Feature extraction: extract features from low-resolution images through convolution operations;

[0079] 3x3 convolution kernel, 3 input channels, 64 output channels;

[0080] Generate a super-resolution image P through convolution operation.

[0081] In step S6,

[0082] The image synthesis is specifically as follows: the images P(λ1), P(λ2), ..., P(λl) that have completed super-resolution processing are synthesized into an image P by weighted averaging.

[0083] P=∑(w1*P(λ1)+w2*P(λ2)+…+w l *P(λl)

[0084] Where P(λ1), P(λ2), …, P(λl) are images with wavelengths of λ1, λ2, …, λl that have undergone super-resolution processing; w1, w2, …, w lare the weights corresponding to P(λ1), P(λ2), …, P(λl).

[0085] The weight is specifically calculated as follows:

[0086] Weights w1, w2, ..., w l Adaptively select according to the contrast and clarity of each image,

[0087]

[0088] Among them, w i is the weight corresponding to the image P(λi).

[0089] A wavelength-driven underwater image enhancement and super-resolution device, characterized in that: the device is used to implement the above method, and specifically includes the following parts:

[0090] Lighting receiving sensor: It has multiple wavelengths of lighting, provides adaptive wavelength light source, and can adaptively adjust the wavelength combination according to real-time environmental parameters; it is equipped with sensors to monitor the turbidity and light conditions of the underwater environment;

[0091] Multispectral cameras: Capable of capturing image data at multiple wavelengths simultaneously and working in conjunction with adaptive wavelength light sources;

[0092] Environmental Monitoring Module: Includes multiple sensors for real-time monitoring of various parameters of the underwater environment, including turbidity and light intensity; equipped with a processor to calculate the optimal wavelength combination based on sensor data;

[0093] Image processing module: equipped with a high-speed processor and large-capacity memory for processing the collected image data; contains a spectral analysis unit for absorption and scattering analysis, as well as adaptive color correction;

[0094] Adaptive denoising and preprocessing module: dynamically adjusts the parameters of the denoising algorithm and brightness contrast adjustment algorithm according to real-time environmental parameters;

[0095] Super-resolution processing module: equipped with a deep learning processor capable of running adaptive super-resolution models; the model library has preset adaptive super-resolution algorithms to meet different application requirements;

[0096] Image synthesis and post-processing module: synthesize enhanced images of different wavelengths and perform adaptive post-processing;

[0097] Display and storage unit: super-resolution display screen for real-time viewing of enhanced images; large-capacity memory for storing processed high-quality image data;

[0098] Control interface: A human-machine interactive interface that allows users to set parameters, view real-time images, and adjust processing options; includes remote control capabilities for easy operation in different scenarios.

[0099] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0100] 1. The present invention utilizes the characteristics of light of different wavelengths to accurately correct the color distortion of underwater images. According to the different degrees of absorption and scattering of light of different wavelengths in water, the original color can be effectively restored by analyzing the information of these wavelengths;

[0101] 2. The present invention can identify and correct image blur caused by scattering through wavelength analysis, which helps to improve the clarity and contrast of the image. At the same time, by using a super-resolution algorithm and combining wavelength-driven information, a super-resolution image can be reconstructed from a low-resolution underwater image, which is very helpful for detail recognition and analysis, making the details clearer;

[0102] 3. The present invention can effectively identify and reduce noise in the image through the wavelength driving method, especially the scattering noise caused by particles and impurities in the water, so that a cleaner image can be obtained and the image quality can be improved;

[0103] 4. The present invention can adapt to different underwater environments (such as clear seawater and turbid rivers) according to the different absorption and scattering characteristics of light; the wavelength-driven method can adjust the correction parameters according to the specific environment, and has good adaptability and flexibility;

[0104] 5. The present invention realizes real-time image enhancement and super-resolution processing through an adaptive wavelength-driven algorithm to meet the real-time needs of underwater detection, monitoring and navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0106] Figure 1 It is a schematic diagram of the steps of a wavelength-driven underwater image enhancement and super-resolution method of the present invention;

[0107] Figure 2 Schematic diagram of the steps of multi-wavelength image acquisition. DETAILED DESCRIPTION

[0108] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0109] See also Figure 1-Figure 2 The present invention provides a technical solution: a wavelength-driven underwater image enhancement and super-resolution method, the method comprising the following steps:

[0110] S1, initial environment monitoring and sensor data collection, adaptive wavelength selection through adaptive wavelength light source;

[0111] S2, real-time data monitoring and multi-wavelength image acquisition;

[0112] S3, performing spectral analysis and adaptive correction on the collected multi-wavelength images according to different wavelengths, and obtaining adaptive initial images with different wavelengths in turn;

[0113] S4, performing dynamic denoising and preprocessing on initial images with different wavelengths in sequence;

[0114] S5, performing super-resolution processing on the image that has undergone preprocessing and dynamic denoising;

[0115] S6, synthesizing multiple images with different wavelengths;

[0116] S7, output the synthesized image, and feed back the image parameters to step S1;

[0117] The sensors include a turbidity sensor, a light intensity sensor and a temperature sensor; the turbidity sensor is used to measure the concentration of suspended particles in the water to determine the turbidity of the water; the light intensity sensor is used to measure the underwater light intensity to determine the current light conditions; the temperature sensor is used to measure the water temperature;

[0118] The adaptive wavelength light source achieves real-time wavelength selection by adjusting the intensity of LED light sources with different wavelengths.

[0119] In step S1,

[0120] The adaptive wavelength selection is a method for dynamically adjusting the wavelength combination of the light source according to the real-time underwater environmental parameters; the method aims to optimize the acquisition quality of underwater images, reduce the absorption and scattering effects of light in water by selecting the best wavelength combination, and improve the clarity and contrast of the image; the specific steps are:

[0121] S2-1. Preliminary selection of wavelengths: Based on the existing underwater spectral characteristic data, a set of preliminary wavelengths is selected, including multiple wavelengths in the visible light range;

[0122] S2-2, parameter model: Establish a model based on environmental parameters to calculate the effectiveness of each wavelength in the current environment. The model takes into account the light absorption coefficient a(λ,t) and scattering coefficient b(λ,I), which change with the wavelength and underwater environmental parameters; the attenuation formula of light in water is:

[0123] I(λ,d)=I0(λ)·e -(a(λ,t)+b(λ,I)·d

[0124] Where I(λ,d) is the intensity of light with wavelength λ at light distance d; I0(λ) is the initial light intensity; a(λ,t) is the absorption coefficient of wavelength λ at turbidity t; b(λ,I) is the scattering coefficient of wavelength λ at light intensity I; d is the distance the light propagates;

[0125] The absorption coefficient a(λ,t) and scattering coefficient b(λ,I) are calculated as:

[0126]

[0127] where a0 and b0 are the reference absorption and scattering coefficients at wavelength λ; k a and k b is a constant related to the environmental parameters turbidity t and light intensity I, k a =0.2, k b =0.1;

[0128] S2-3. Calculation of the best wavelength combination: The optimization algorithm is used to find the wavelength combination that makes the light propagation effect in water optimal under the current environmental parameters; the objective function is to analyze the received synthetic image and establish a comprehensive evaluation index of image quality, including image contrast and clarity:

[0129] Qualityλ)=∑(Contrast(p)+Sharpness(p))

[0130] Among them, Contrast(p) and Sharpness(p) represent the contrast and sharpness of the image corresponding to the wavelength λ, respectively;

[0131] The optimal wavelength combination is to find the wavelength combination that makes the light propagation effect in water optimal under the current environmental parameters through an optimization algorithm;

[0132] The underwater environmental parameters include turbidity and light intensity.

[0133] In step S3-2,

[0134] The contrast of the image uses the mean contrast, specifically:

[0135]

[0136] Where z is the total number of pixels in the image; is the mean value of the image pixels; p n is the nth pixel value in the image;

[0137] The clarity of the image is calculated using the mean of the image gradient; specifically:

[0138] Calculate the gradient of the image: Use the Sobel operator to calculate the gradient of the image in the x and y directions;

[0139] G x (x,y)=p(x+1,y-1)+2p(x+1,y)+p(x+1,y+1)-(p(x-1,y-1)

[0140] +2p(x-1,y)+p(x-1,y+1))

[0141] G y (x,y)=p(x-1,y+1)+2p(x,y+1)+p(x+1,y+1)-(p(x-1,y-1)

[0142] +2p(x,y-1)+p(x+1,y-1))

[0143] Among them, p(x,y) is the pixel value of the image at position (x,y), G x (x,y) and G y (x, y) are the gradients of the image in the x and y directions at the (x, y) position respectively;

[0144] Calculate the magnitude of the gradient: Combine the gradients in the x-direction and the y-direction to calculate the gradient magnitude of each pixel;

[0145]

[0146] Where G(x,y) is the magnitude of the gradient at position (x,y);

[0147] Calculate the mean of the gradient amplitude: average the gradient amplitudes of all pixels to get the image clarity index; the larger the gradient amplitude, the sharper the image edge, that is, the clearer it is;

[0148]

[0149] Where M and N are the width and height of the image respectively.

[0150] In step S2,

[0151] Multi-wavelength image acquisition is the process of using a multi-spectral camera and an adaptive wavelength light source to illuminate underwater target objects with light of different wavelengths to collect multiple image data. The specific steps of multi-wavelength image acquisition are:

[0152] S4-1. Setting the wavelength of the light source: adjusting the output wavelength and intensity of each LED light source of the adaptive wavelength light source according to the selected optimal wavelength combination;

[0153] S4-2, start the multispectral camera: start the multispectral camera and configure the camera parameters so that it can capture images at different wavelengths;

[0154] S4-3, collecting images wavelength by wavelength: for each selected wavelength, adjusting the light source to emit light of the wavelength; collecting images at the wavelength and saving them in a memory;

[0155] S4-4, switch to the next wavelength, and repeat the above process until the image acquisition of all wavelengths is completed.

[0156] In step S3,

[0157] Spectral analysis is based on the physical properties of light propagation in water. Light of different wavelengths has different absorption and scattering coefficients in water. By analyzing these coefficients, we can understand the propagation of light in water and provide data support for subsequent adaptive correction.

[0158] The adaptive correction includes adaptive color correction and scattering compensation;

[0159] Color correction is to compensate for the absorption differences of different wavelengths of light in water, making the color of the image closer to the real situation; the correction formula is as follows:

[0160] I'(λ)=I(λ)·e (a)λ)+b)λ))·d

[0161] Where, I')λ) is the corrected light intensity; I(λ) is the initial light intensity;

[0162] Scattering compensation is to reduce the impact of scattering in water on the image and increase the clarity of the image; the scattering compensation formula is as follows:

[0163]

[0164] Among them, I comp (λ) is the light intensity after scattering compensation; k s is the scattering coefficient;

[0165] An adaptively corrected initial image p(λ) with a wavelength of λ is obtained through adaptive correction.

[0166] In step S4,

[0167] Dynamic denoising is to remove the noise in the above initial image p(λ);

[0168] The dynamic denoising adopts the mean filtering method to remove high-frequency noise by calculating the mean of each pixel neighborhood in the image; the formula is as follows:

[0169]

[0170] Where p'((x,y)|λ) is the pixel value of the image at position (x,y) with a wavelength of λ and after denoising; p((x+i,y+j)|λ) is the pixel value of the initial image at position (x,y) with a wavelength of λ and after adaptive correction; Q is the neighborhood of position (x,y);

[0171] The image preprocessing includes image normalization and image sharpening;

[0172] The formula for image normalization is as follows:

[0173]

[0174] Among them, p" (λ) is the image with wavelength λ and normalized; p' (λ) is the image with wavelength λ and denoised; p' min (λ) and p' max (λ) are the minimum and maximum pixel values ​​in p'(λ), respectively;

[0175] Image sharpening is to enhance the edges in the image to improve the details and clarity of the image; using the Laplace operator; the formula is as follows:

[0176] P(λ)=p”(λ)+α·Δp”(λ)

[0177] Where p"'(λ) is the sharpened image with wavelength λ; α is the sharpening intensity coefficient; Δp"(λ) is the Laplace transform result of p"(λ);

[0178] p"'(λ) is the image with wavelength λ after denoising and preprocessing.

[0179] In step S5,

[0180] Super-resolution processing is to reconstruct low-resolution images to improve image resolution. Through machine learning methods: using convolutional neural networks, learn the mapping relationship from low resolution to super-resolution from a large amount of training data, and through multiple convolutional layers, learn the high-level features of the image to perform super-resolution processing on the image.

[0181] Input image: p''(λ)∈R M×N×C, where p''(λ) is the image with wavelength λ and after denoising and preprocessing; M, N and C are the width, height and number of channels of the image respectively, p''(λ) is an RGB image, so the number of channels C = 3;

[0182] Feature extraction: extract features from low-resolution images through convolution operations;

[0183] 3x3 convolution kernel, 3 input channels, 64 output channels;

[0184] Generate a super-resolution image P through convolution operation.

[0185] In step S6,

[0186] The image synthesis is specifically as follows: the images P(λ1), P(λ2), ..., P(λl) that have completed super-resolution processing are synthesized into an image P by weighted averaging.

[0187] P=∑(w1*P(λ1)+w2*P(λ2)+…+w l *P(λl)

[0188] Where P(λ1), P(λ2), …, P(λl) are images with wavelengths of λ1, λ2, …, λl that have undergone super-resolution processing; w1, w2, …, w l are the weights corresponding to P(λ1), P(λ2), …, P(λl).

[0189] The weight is specifically calculated as follows:

[0190] Weights w1, w2, ..., w l Adaptively select according to the contrast and clarity of each image,

[0191]

[0192] Among them, w i is the weight corresponding to the image P(λi).

[0193] A wavelength-driven underwater image enhancement and super-resolution device, characterized in that: the device is used to implement the above method, and specifically includes the following parts:

[0194] Lighting receiving sensor: It has multiple wavelengths of lighting, provides adaptive wavelength light source, and can adaptively adjust the wavelength combination according to real-time environmental parameters; it is equipped with sensors to monitor the turbidity and light conditions of the underwater environment;

[0195] Multispectral cameras: Capable of capturing image data at multiple wavelengths simultaneously and working in conjunction with adaptive wavelength light sources;

[0196] Environmental Monitoring Module: Includes multiple sensors for real-time monitoring of various parameters of the underwater environment, including turbidity and light intensity; equipped with a processor to calculate the optimal wavelength combination based on sensor data;

[0197] Image processing module: equipped with a high-speed processor and large-capacity memory for processing the collected image data; contains a spectral analysis unit for absorption and scattering analysis, as well as adaptive color correction;

[0198] Adaptive denoising and preprocessing module: dynamically adjusts the parameters of the denoising algorithm and brightness contrast adjustment algorithm according to real-time environmental parameters;

[0199] Super-resolution processing module: equipped with a deep learning processor capable of running adaptive super-resolution models; the model library has preset adaptive super-resolution algorithms to meet different application requirements;

[0200] Image synthesis and post-processing module: synthesize enhanced images of different wavelengths and perform adaptive post-processing;

[0201] Display and storage unit: super-resolution display screen for real-time viewing of enhanced images; large-capacity memory for storing processed high-quality image data;

[0202] Control interface: A human-machine interactive interface that allows users to set parameters, view real-time images, and adjust processing options; includes remote control capabilities for easy operation in different scenarios.

[0203] In an embodiment of the present invention, the environmental parameter measures:

[0204] Turbidity: t = 0.8; Light intensity: I = 300 lux; Water temperature: 15°C; Water depth d = 1;

[0205] Based on environmental monitoring data, the model parameters are:

[0206] a(450nm, 0.8)=0.15, b(450nm, 300)=0.20

[0207] a(550nm, 0.8)=0.10, b(550nm, 300)=0.18

[0208] a(650nm, 0.8)=0.08, b(650nm, 300)=0.15

[0209] Calculate the intensity attenuation of light at different wavelengths and distances:

[0210] I(450nm, 1m) = I0(450nm)*e -(0.15+0.20)*1 =I0(450nm)*e -0.35

[0211] I(550nm, 1m)=I0(450nm)*e -(0.10+0.12)*1 =I0(450nm)*e -0.22

[0212] I(650nm, 1m)=I0(450nm)*e -(0.08+0.15)*1 =I0(450nm)*e -0.23

[0213] According to the above calculation results, the wavelength (550nm) that minimizes light attenuation is selected as one of the best wavelengths. Other wavelengths are further combined through the optimization algorithm to obtain the final wavelength combination. Through the adaptive wavelength selection algorithm, the best wavelength combination is determined to be 450nm, 550nm and 650nm.

[0214] Image acquisition:

[0215] Set the light source wavelength to 450 nm, start the multispectral camera, and collect images at 450 nm.

[0216] Switch the light source wavelength to 550nm and collect images at 550nm.

[0217] Switch the light source wavelength to 650nm and collect images at 650nm.

[0218] The current underwater environment parameters are:

[0219] Turbidity t = 0.8; light intensity I = 300 lux; water depth d = 1 m;

[0220] Select wavelength λ = 550nm, the known reference absorption coefficient and scattering coefficient are:

[0221] a0=0.1;b0=0.05

[0222] Constant k a =0.2; constant k b =0.1, the initial illumination is 1;

[0223] a(550nm)=0.1*e 0.2*0.8 ≈0.116

[0224] b(550nm)=0.05*e 0.1*300 ≈1.0

[0225] I'(550nm)=1·e (0.116+1.0)·1 ≈3.053

[0226]

[0227] In our experiment we measured the attenuation of light in water, according to the Beer-Lambert law:

[0228] Assuming that we measured the initial light intensity I_0 = 100 in the experiment and the light intensity I = 60 at a distance d = 0 meters, we can calculate k s :

[0229]

[0230] This value means that in this particular body of water, the scattering coefficient of light is 0.051m -1 .

[0231] The image size is 5*5, specifically:

[0232]

[0233] Image after denoising and normalization:

[0234]

[0235] Image after image sharpening:

[0236] Sharpening strength coefficient α = 0.5:

[0237] Image sharpening using the Laplacian operator:

[0238]

[0239] Perform super-resolution processing on the pre-processed and dynamic denoising images; reconstruct the image from 5*5 to 10*10;

[0240] Synthesize multiple images with different wavelengths;

[0241]

[0242] Output the synthesized image; calculated w1=0.4,w2=0.6

[0243]

[0244] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0245] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A wavelength-driven underwater image enhancement and super-resolution method, characterized in that: The method comprises the following steps: S1, initial environment monitoring and sensor data collection, adaptive wavelength selection through adaptive wavelength light source; S2, real-time data monitoring and multi-wavelength image acquisition; S3, performing spectral analysis and adaptive correction on the collected multi-wavelength images respectively, and obtaining adaptive initial images with different wavelengths in turn; S4, performing dynamic denoising and preprocessing on initial images with different wavelengths in sequence; S5, performing super-resolution processing on the image that has undergone preprocessing and dynamic denoising; S6, synthesizing multiple images with different wavelengths; S7, output the synthesized image, and feed back the image parameters to step S1; The sensors include a turbidity sensor, a light intensity sensor and a temperature sensor; the turbidity sensor is used to measure the concentration of suspended particles in the water to determine the turbidity of the water; the light intensity sensor is used to measure the underwater light intensity to determine the current light conditions; the temperature sensor is used to measure the water temperature; The adaptive wavelength light source achieves real-time wavelength selection by adjusting the intensity of LED light sources with different wavelengths.

2. The wavelength-driven underwater image enhancement and super-resolution method according to claim 1, characterized in that: In step S1, The adaptive wavelength selection is a method for dynamically adjusting the wavelength combination of the light source according to the real-time underwater environmental parameters; the method aims to optimize the acquisition quality of underwater images, reduce the absorption and scattering effects of light in water by selecting the best wavelength combination, and improve the clarity and contrast of the image; the specific steps are: S2-1. Preliminary selection of wavelengths: Based on the existing underwater spectral characteristic data, a set of preliminary wavelengths is selected, including multiple wavelengths in the visible light range; S2-2, parameter model: Establish a model based on environmental parameters to calculate the effectiveness of each wavelength in the current environment. The model takes into account the light absorption coefficient a(λ,t) and scattering coefficient b(λ,I), which change with the wavelength and the real-time underwater environmental parameters; the attenuation formula of light in water is: I(λ,d)=I0(λ)·e -(a(λ,t)+b(λ,I)·d Where I(λ,d) is the intensity of light with wavelength λ at light distance d; I0(λ) is the initial light intensity; a(λ,t) is the absorption coefficient of wavelength λ at turbidity t; b(λ,I) is the scattering coefficient of wavelength λ at light intensity I; d is the distance the light propagates; The absorption coefficient a(λ,t) and scattering coefficient b(λ,I) are calculated as: where a0 and b0 are the reference absorption and scattering coefficients at wavelength λ; k a and k b is a constant related to the environmental parameters turbidity t and light intensity I, k a =0.2, k b =0.1; S2-3. Calculation of the best wavelength combination: The algorithm is used to find the wavelength combination that makes the light propagation effect in water optimal under the current environmental parameters; the objective function is to analyze the received synthetic image and establish a comprehensive evaluation index of image quality, including image contrast and clarity: Quality(λ)=∑(Contrast(p)+Sharpness(p)) Among them, Quality(λ) is a comprehensive evaluation index of the image quality with wavelength λ; Contrast(p) and Sharpness(p) represent the contrast and sharpness of the image corresponding to wavelength λ, respectively; The optimal wavelength combination is to find the wavelength combination that enables light to propagate the farthest in water under current environmental parameters through an algorithm; The underwater environmental parameters include turbidity and light intensity.

3. The wavelength-driven underwater image enhancement and super-resolution method according to claim 2, characterized in that: In step S2-3, The contrast of the image uses the mean contrast, specifically: Where z is the total number of pixels in the image; is the mean value of the image pixels; p n is the nth pixel value in the image; The clarity of the image is calculated using the mean of the image gradient; specifically: Calculate the gradient of the image: Use the Sobel operator to calculate the gradient of the image in the x and y directions; G x (x,y)=p(x+1,y-1)+2p(x+1,y)+p(x+1,y+1)-(p(x-1,y-1) +2p(x-1,y)+p(x-1,y+1)) G y (x,y)=p(x-1,y+1)+2p(x,y+1)+p(x+1,y+1)-(p(x-1,y-1) +2p(x,y-1)+p(x+1,y-1)) Among them, p(x,y) is the pixel value of the image at position (x,y), G x (x,y) and G y (x, y) are the gradients of the image in the x and y directions at the (x, y) position respectively; Calculate the magnitude of the gradient: Combine the gradients in the x-direction and the y-direction to calculate the gradient magnitude of each pixel; Where G(x,y) is the magnitude of the gradient at position (x,y); Calculate the mean of the gradient amplitude: average the gradient amplitudes of all pixels to get the image clarity index; the larger the gradient amplitude, the sharper the image edge, that is, the clearer it is; Where M and N are the width and height of the image respectively.

4. The wavelength-driven underwater image enhancement and super-resolution method according to claim 1, characterized in that: In step S2, Multi-wavelength image acquisition is the process of using a multi-spectral camera and an adaptive wavelength light source to illuminate underwater target objects with light of different wavelengths to collect multiple image data. The specific steps of multi-wavelength image acquisition are: S4-1. Setting the wavelength of the light source: adjusting the output wavelength and intensity of each LED light source of the adaptive wavelength light source according to the selected optimal wavelength combination; S4-2, start the multispectral camera: enable it to capture images at different wavelengths; S4-3, collecting images wavelength by wavelength: for each selected wavelength, adjusting the light source to emit light of the wavelength; collecting images at the wavelength and saving them in a memory; S4-4, switch to the next wavelength, and repeat the above process until the image acquisition of all wavelengths is completed.

5. The wavelength-driven underwater image enhancement and super-resolution method according to claim 1, characterized in that: In step S3, Spectral analysis is based on the physical properties of light propagation in water. Light of different wavelengths has different absorption and scattering coefficients in water. By analyzing these coefficients, we can understand the propagation of light in water and provide data support for subsequent adaptive correction. The adaptive correction includes adaptive color correction and scattering compensation; Adaptive color correction is to compensate for the absorption differences of different wavelengths of light in water, making the color of the image closer to the real situation; the correction formula is as follows: I′(λ)=I(λ)·e (a(λ)+b(λ)·d Where, I'(λ) is the corrected light intensity; I(λ) is the initial light intensity; Scattering compensation is to reduce the impact of scattering in water on the image and increase the clarity of the image; the scattering compensation formula is as follows: Among them, I comp (λ) is the light intensity after scattering compensation; k s is the scattering coefficient; An adaptively corrected initial image p(λ) with a wavelength of λ is obtained through adaptive correction.

6. The wavelength-driven underwater image enhancement and super-resolution method according to claim 1, characterized in that: In step S4, Dynamic denoising is to remove the noise in the above initial image p(λ); The dynamic denoising adopts the mean filtering method to remove high-frequency noise by calculating the mean of each pixel neighborhood in the image; the formula is as follows: Where p'((x,y)|λ) is the pixel value of the image at position (x,y) with a wavelength of λ and after denoising; p((x+i,y+j)|λ) is the pixel value of the initial image at position (x,y) with a wavelength of λ and after adaptive correction; Q is the neighborhood of position (x,y); The image preprocessing includes image normalization and image sharpening; The formula for image normalization is as follows: Among them, p" (λ) is the image with wavelength λ and normalized; p' (λ) is the image with wavelength λ and denoised; p' min (λ) and p' max (λ) are the minimum and maximum pixel values ​​in p'(λ), respectively; Image sharpening is to enhance the edges in the image to improve the details and clarity of the image; using the Laplace operator; the formula is as follows: P(λ)=p”(λ)+α·Δp”(λ) Where p"'(λ) is the sharpened image with wavelength λ; α is the sharpening intensity coefficient; Δp"(λ) is the Laplace transform result of p"(λ); p"'(λ) is the image with wavelength λ after denoising and preprocessing.

7. The wavelength-driven underwater image enhancement and super-resolution method according to claim 6, characterized in that: In step S5, Super-resolution processing is to reconstruct low-resolution images to improve image resolution. Through machine learning methods: using convolutional neural networks, learn the mapping relationship from low resolution to super-resolution from a large amount of training data, and through multiple convolutional layers, learn the high-level features of the image to perform super-resolution processing on the image. Input image: p''(λ)∈R M×N×C , where p"'(λ) is the image with wavelength λ and has been denoised and preprocessed; M, N and C are the width, height and number of channels of the image respectively. p"'(λ) is an RGB image, so the number of channels C = 3; Feature extraction: extract features from low-resolution images through convolution operations; 3x3 convolution kernel, 3 input channels, 64 output channels; Generate a super-resolution image P through convolution operation.

8. The wavelength-driven underwater image enhancement and super-resolution method according to claim 4, characterized in that: In step S6, The image synthesis is specifically as follows: the images P(λ1), P(λ2), ..., P(λl) that have completed super-resolution processing are synthesized into an image P by weighted averaging. P=∑(w1*P(λ1)+w2*P9λ2)+…+w l *P(λl)) Where P(λ1), P(λ2), …, P(λl) are images with wavelengths of λ1, λ2, …, λl that have undergone super-resolution processing; w1, w2, …, w l are the weights corresponding to P(λ1), P(λ2), …, P(λl).

9. The wavelength-driven underwater image enhancement and super-resolution method according to claim 8, characterized in that: The weight is specifically calculated as follows: Weights w1, w2, ..., w l Adaptively selecting according to the contrast and clarity of each image, the contrast and clarity are calculated as shown in claim 3; Among them, w i is the weight corresponding to the image P(λi).

10. A wavelength-driven underwater image enhancement and super-resolution device, characterized in that: The device is used to implement the above method, and specifically comprises the following parts: Lighting receiving sensor: It has multiple wavelengths of lighting, provides adaptive wavelength light source, and can adaptively adjust the wavelength combination according to real-time environmental parameters; it is equipped with sensors to monitor the turbidity and light conditions of the underwater environment; Multispectral cameras: Capable of capturing image data at multiple wavelengths simultaneously and working in conjunction with adaptive wavelength light sources; Environmental Monitoring Module: Includes multiple sensors for real-time monitoring of various parameters of the underwater environment, including turbidity and light intensity; equipped with a processor to calculate the optimal wavelength combination based on sensor data; Image processing module: equipped with a high-speed processor and large-capacity memory for processing the collected image data; contains a spectral analysis unit for absorption and scattering analysis, as well as adaptive color correction; Adaptive denoising and preprocessing module: dynamically adjusts the parameters of the denoising algorithm and brightness contrast adjustment algorithm according to real-time environmental parameters; Super-resolution processing module: equipped with a deep learning processor capable of running adaptive super-resolution models; the model library has preset adaptive super-resolution algorithms to meet different application requirements; Image synthesis and post-processing module: synthesize enhanced images of different wavelengths and perform adaptive post-processing; Display and storage unit: super-resolution display screen for real-time viewing of enhanced images; large-capacity memory for storing processed high-quality image data; Control interface: A human-machine interactive interface that allows users to set parameters, view real-time images, and adjust processing options; includes remote control capabilities for easy operation in different scenarios.

Citation Information

Cited By

  • Underwater facility defect geometric feature measurement method, system, medium and equipment

    CN121346656A