Unmanned ship dynamic illumination correction method and system based on residual diffusion model

Abstract

The invention discloses an unmanned ship dynamic illumination correction method and system based on a residual diffusion model, and the method comprises the steps: 1, obtaining the navigation image of an unmanned ship and the illumination condition data of the unmanned ship; s2, extracting illumination characteristics by using an illumination sensing module; s3, performing illumination correction based on a residual diffusion model, wherein the illumination correction specifically comprises the following two processing steps: a forward diffusion process: simulating dynamic distribution of illumination interference; in the reverse diffusion process, illumination interference is removed step by step, and a corrected image is generated; s4, outputting a corrected image for target identification and path planning of the unmanned ship; the invention discloses an unmanned ship dynamic illumination correction system based on a residual diffusion model. The system comprises a data acquisition module; an illumination sensing module; a residual diffusion illumination correction module; an output and application module; the method has the advantages that the definition and the detail visibility are improved by introducing the illumination sensing module and the residual diffusion correction model.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing for unmanned ships, and in particular to a method and system for dynamic illumination correction of unmanned ships based on a residual diffusion model. Background Art

[0002] Intelligent Shipping Systems (ISS) are integrated systems that utilize advanced sensor technology, artificial intelligence (AI) algorithms, communication networks, and automation technologies to intelligently monitor, navigate, and manage the navigation status of unmanned vessels. They are widely used in marine transportation, offshore patrols, environmental monitoring, and search and rescue missions, improving navigation efficiency and safety through intelligent means.

[0003] Computer vision (CV), a key branch of artificial intelligence, aims to empower computers to process, analyze, and understand images and videos. Combining image processing, pattern recognition, deep learning, and multi-sensor fusion, CV has become a key enabler for intelligent unmanned vessel systems, providing crucial technical support for target detection, path planning, and dynamic obstacle avoidance.

[0004] In unmanned vessel vision systems, the complex navigation environment and drastic changes in lighting conditions significantly impact image quality. Strong reflections, shadows, and insufficient lighting can degrade the quality of images captured by the camera. These issues are exacerbated in dynamic lighting conditions, leading to reduced image contrast, color distortion, and loss of target information. This not only significantly reduces the accuracy of target detection and path planning, but can also threaten the safety of the unmanned vessel.

[0005] Existing image correction methods, mostly based on traditional image enhancement techniques or simple denoising algorithms, struggle to fully address dynamic lighting issues. To address these issues, designing a robust illumination correction method that can both improve image quality under strong reflections and shadows while meeting the real-time processing requirements of unmanned vessels has become a hot topic and a challenge in current research. Summary of the Invention

[0006] The purpose of the present invention is to overcome the above-mentioned defects in the prior art and provide a dynamic illumination correction method and system for unmanned ships based on the residual diffusion model. By introducing an illumination perception module and a residual diffusion correction model, it can effectively cope with lighting conditions such as strong reflection, shadow coverage and uneven illumination, and significantly improve the clarity and detail visibility of unmanned ship images.

[0007] To achieve the above object, the present invention is implemented through the following two aspects:

[0008] In a first aspect, the present invention provides a method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model, which comprises the following steps:

[0009] S1: Acquire the unmanned ship navigation image and its lighting condition data;

[0010] S2: Use the illumination perception module to extract illumination characteristics, including illumination intensity, shadow distribution, and reflection area;

[0011] S3: Perform illumination correction based on the residual diffusion model, which includes the following two processing steps:

[0012] Forward diffusion process: simulates the dynamic distribution of light interference;

[0013] Reverse diffusion process: gradually removes light interference and generates a corrected image;

[0014] S4: Output the corrected image for target recognition and path planning of the unmanned vessel.

[0015] Preferably, S2 includes the following sub-steps:

[0016] S21. Construction and Training of a Multi-Task Deep Learning Model: Build a multi-task deep learning model that includes multiple task branches for light intensity extraction, shadow area detection, and reflection area recognition.

[0017] S22. Light Intensity Extraction: Extract the intensity distribution of light in the image and generate a brightness map to reflect the overall lighting conditions.

[0018] S23. Shadow Region Detection: Leverage the shadow detection branch of a multi-task deep learning model to identify shadow regions in an image and extract their boundaries, shape, and coverage.

[0019] S24. Reflective Area Recognition: Using the reflective area recognition branch of a multi-task deep learning model, we locate highly reflective areas in an image and distinguish between reflective and non-reflective areas.

[0020] S25. Integrate the light intensity, shadow distribution, and reflection area characteristics output by the multi-task deep learning model to generate a light characteristic description to provide input for subsequent light correction.

[0021] Preferably, in step S3, the residual diffusion model models and corrects illumination interference by adding illumination conditions, wherein the specific processing steps of the forward diffusion process are: when adding noise to the image, the illumination distribution characteristics are used as conditional input, and the dynamic illumination characteristics are combined to affect the distribution of noise. The noise generation formula is:

[0022]

[0023] Among them, x t : target image or image at the current moment; x t-1 : The image at the previous moment, which serves as the basis for the current image; l: Lighting conditions, which adjust the lighting effects of the image; X res ; Residual, the difference between the target image and the degraded image; β t : diffusion coefficient, controlling the intensity of noise and residual; I: unit matrix, representing the covariance of noise; N(μ,σ 2 ): Gaussian distribution, used to model the probability distribution of the current image, where the mean Variance σ 2 =β t ·I.

[0024] Preferably, in step S3, the residual diffusion model is used to model and correct illumination interference by adding illumination conditions, wherein the specific processing steps of the reverse diffusion process are: reverse diffusion restores the target image by gradually removing noise, and the residual is used to guide the denoising direction in this process, accelerate and optimize image restoration, and the reverse diffusion formula is:

[0025] x t-1 =x t -∈θ(x t ,t,l)+λ·X res

[0026] Among them, x t : The image state at the current time step t; x t-1 : image state at time step t-1;

[0027] ∈θ(x t ,t,l): conditional denoising network, predicting noise; λ: weight, controlling the strength of residual correction; X res : Residual, the difference between the target image and the degraded image.

[0028] Preferably, an illumination condition embedding module and a residual diffusion fusion module are added to the network structure of the residual diffusion model, which includes the following contents:

[0029] Lighting condition embedding module: Builds global and local embedding mechanisms to integrate the lighting distribution feature l into the diffusion model network; global embedding: uses a multi-layer perceptron (MLP) to globally encode the lighting characteristics and generate a lighting condition vector; local embedding: uses a convolutional network to locally encode the lighting characteristics and generate a lighting condition feature map of the same size as the input image;

[0030] Residual Diffusion Fusion Module: The encoder and decoder parts of the diffusion model are adjusted to introduce dynamic guidance of the residual information Xres. In the encoder, the illumination condition features and residual information are fused with the image features through feature splicing or additive fusion. A residual feature module is added to the decoder to dynamically adjust the denoising direction and enhance the image correction effect.

[0031] Preferably, the loss function of the residual diffusion model includes the following parts:

[0032] Denoising reconstruction loss: Optimize denoising accuracy by minimizing the error between the noise predicted by the conditional denoising network and the actual noise. The formula is: L denoise =||∈θ(x t ,t,l)-∈|| 2

[0033] Among them, ∈θ(x t ,t,l) is the output noise prediction of the denoising network, ∈ is the real noise;

[0034] Light correction loss: Optimize the corrected image according to the light distribution characteristics so that its light characteristics are consistent with the target. The formula is: L light =||f(x output )-f(x input )|| 2

[0035] Among them, f(x) is the illumination feature extraction function (illumination mean), x output and x input are the corrected and original images respectively;

[0036] Detail enhancement loss: Through gradient or edge constraints, the detail recovery effect of the corrected image is enhanced. The formula is:

[0037] in, represents the gradient operator, is the real target image;

[0038] Total loss function: Combining the above three loss functions, the weight parameters λ1, λ2, and λ3 are used to balance the effects of different losses, and is defined as: L = λ1L denoise +λ2L light +λ3L detail .

[0039] In a second aspect, the present invention provides a dynamic illumination correction system for an unmanned vessel based on a residual diffusion model, comprising:

[0040] A data acquisition module is used to obtain image data of the unmanned ship under different lighting conditions;

[0041] Light perception module, used to obtain the lighting characteristics of the input image;

[0042] The residual diffusion illumination correction module dynamically corrects illumination based on the residual diffusion model and generates a corrected image;

[0043] Output and application module, outputs corrected images to support target recognition and path planning of unmanned vessels;

[0044] The above modules complete the logical link in a progressive relationship.

[0045] Preferably, the light perception module further includes a preprocessing unit for performing preliminary processing on the received image data.

[0046] Preferably, the residual diffuse illumination correction module further includes a post-processing unit for post-processing the corrected image and then outputting the post-processed image to the output and application module.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] By introducing an illumination perception module and a residual diffusion model, this technical solution can accurately simulate the dynamic distribution of illumination interference and effectively remove it. It can effectively cope with complex lighting conditions such as strong reflections, shadow coverage, and uneven illumination, significantly improving the illumination uniformity and detail clarity of images.

[0049] Furthermore, by extracting illumination characteristics through a multi-task deep learning model, the adaptability of the correction process to complex lighting scenes is enhanced; at the same time, the residual diffusion mechanism is utilized to optimize the efficiency and effect of correction. The corrected image shows higher accuracy and robustness in subsequent unmanned ship target recognition and path planning tasks, and is suitable for a variety of complex navigation environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0051] Figure 1 This is a flow diagram of a method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model provided in the first embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the framework of a dynamic illumination correction system for an unmanned vessel based on a residual diffusion model provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solution in this embodiment of the present invention in conjunction with the drawings in this embodiment of the present invention. Obviously, the embodiment described is only one embodiment of the present invention, not all embodiments of the present invention. Based on this embodiment of the present invention, all other embodiments of the present invention obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0054] Example 1:

[0055] See also Figure 1 , Embodiment 1 of the present invention provides a method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model, which includes the following steps:

[0056] S1: Acquire the unmanned ship navigation image and its lighting condition data;

[0057] Among them, the real-time images of the unmanned ship under different lighting conditions (strong light, shadow, high reflection, etc.) are obtained through the camera equipment, and the relevant parameters of the lighting environment are recorded;

[0058] S2: Use the illumination perception module to extract illumination characteristics from the unmanned ship navigation image and generate illumination descriptions. The illumination characteristics include illumination intensity, shadow distribution, reflection area and other characteristics.

[0059] Wherein, the S2 includes the following sub-steps:

[0060] S21. Construction and Training of Multi-Task Deep Learning Models: Build a multi-task deep learning model that includes multiple task branches, such as light intensity extraction, shadow area detection, and reflection area recognition.

[0061] S22. Light Intensity Extraction: Extract the intensity distribution of light in the image and generate a brightness map to reflect the overall lighting conditions.

[0062] S23. Shadow Region Detection: Leverage the shadow detection branch of a multi-task deep learning model to identify shadow regions in an image and extract their boundaries, shape, and coverage.

[0063] S24. Reflective Area Recognition: Using the reflective area recognition branch of a multi-task deep learning model, we locate highly reflective areas in an image and distinguish between reflective and non-reflective areas.

[0064] S25. Integrate the light intensity, shadow distribution, and reflection area characteristics output by the multi-task deep learning model to generate a light characteristic description vector, which provides input for subsequent light correction.

[0065] Furthermore, the multi-task deep learning model: by sharing the feature extraction layer, jointly training the light intensity, shadow distribution and reflection area detection tasks, the robustness of light perception is enhanced.

[0066] S3: Perform illumination correction based on the residual diffusion model, which includes the following two processing steps:

[0067] Forward diffusion process: simulates the dynamic distribution of light interference;

[0068] Reverse diffusion process: gradually removes light interference and generates a corrected image;

[0069] In step S3,

[0070] The residual diffusion model is used to model and correct illumination interference by adding illumination conditions.

[0071] The specific processing steps of the forward diffusion process are as follows: when adding noise to the image, the illumination distribution characteristics are used as conditional input, and the dynamic illumination characteristics are combined to affect the distribution of noise. The noise generation formula is:

[0072]

[0073] Among them, x t : target image or image at the current moment; x t-1 : The image at the previous moment, which serves as the basis for the current image; l: Lighting conditions, which adjust the lighting effects of the image; X res ; Residual, the difference between the target image and the degraded image; β t : diffusion coefficient, controlling the intensity of noise and residual; I: unit matrix, representing the covariance of noise; N(μ,σ 2 ): Gaussian distribution, used to model the probability distribution of the current image, where the mean Variance σ 2 =β t I;

[0074] The specific processing steps of the reverse diffusion process are as follows: reverse diffusion restores the target image by gradually removing noise. The residual is used to guide the denoising direction in this process, accelerate and optimize image restoration, and the reverse diffusion formula is:

[0075] x t-1 =x t -∈θ(x t ,t,l)+λ·X res

[0076] Among them, x t : The image state at the current time step t; x t-1 : image state at time step t-1; ∈θ(x t,t,l): conditional denoising network, predicting noise; λ: weight, controlling the strength of residual correction; X res : Residual, the difference between the target image and the degraded image.

[0077] In addition, an illumination condition embedding module and a residual diffusion fusion module are added to the network structure of the residual diffusion model, which includes the following contents:

[0078] Lighting condition embedding module: Builds global and local embedding mechanisms to integrate the lighting distribution feature l into the diffusion model network; global embedding: uses a multi-layer perceptron (MLP) to globally encode the lighting characteristics and generate a lighting condition vector; local embedding: uses a convolutional network to locally encode the lighting characteristics and generate a lighting condition feature map of the same size as the input image;

[0079] Residual Diffusion Fusion Module: Adjusts the encoder and decoder parts of the diffusion model, introduces dynamic guidance of residual information Xres, and fuses illumination condition features and residual information with image features through feature concatenation or addition in the encoder. A residual feature module is added to the decoder to dynamically adjust the denoising direction and enhance the image correction effect.

[0080] Furthermore, the loss function of the residual diffusion model includes the following parts:

[0081] Denoising reconstruction loss: Optimize denoising accuracy by minimizing the error between the noise predicted by the conditional denoising network and the actual noise. The formula is: L denoise =||∈θ(x t ,t,l)-∈|| 2

[0082] Among them, ∈θ(x t ,t,l) is the output noise prediction of the denoising network, ∈ is the real noise;

[0083] Light correction loss: Optimize the corrected image according to the light distribution characteristics so that its light characteristics are consistent with the target. The formula is: L light =||f(x output )-f(x input )|| 2

[0084] Among them, f(x) is the illumination feature extraction function (illumination mean), x output and x input are the corrected and original images respectively;

[0085] Detail enhancement loss: Through gradient or edge constraints, the detail recovery effect of the corrected image is enhanced. The formula is:

[0086] in, represents the gradient operator, is the real target image;

[0087] Total loss function: Combining the above three loss functions, the weight parameters λ1, λ2, and λ3 are used to balance the effects of different losses, and is defined as: L = λ1L denoise +λ2L light +λ3L detail .

[0088] S4: Output the corrected image and perform post-processing on the corrected image, including contrast adjustment and detail addition, to finally output a high-quality image. The high-quality image is used for applications such as target recognition and path planning for unmanned ships.

[0089] Unmanned vessel navigation images are acquired in dynamic environments and may be subject to interference from sea surface fluctuations, varying lighting conditions (such as daytime and nighttime, and changeable weather), and other natural environmental factors. Outputting high-quality images is precisely the solution to these problems. Deep learning-based object detection algorithms (such as YOLO and Faster R-CNN) can be used to identify target objects in images. High-quality images are input and the object detection algorithm identifies surrounding objects (such as buoys, marine obstacles, and other ships). High-quality images help these algorithms perform classification and positioning more accurately, thereby avoiding collisions and navigation errors. Based on the target recognition results of high-quality images, path planning algorithms (such as A*, Dijkstra algorithm, and RRT) can calculate the optimal navigation path in real time. Unmanned vessels can continuously adjust their paths in complex environments to avoid dynamic obstacles, such as other ships or waves.

[0090] The first embodiment of the present invention is a dynamic illumination correction method for an unmanned ship based on a residual diffusion model, which has the following advantages: by introducing an illumination perception module and a residual diffusion model, this technical solution can accurately simulate the dynamic distribution of illumination interference and effectively remove illumination interference, and can effectively cope with complex illumination conditions such as strong reflection, shadow coverage and uneven illumination, and significantly improve the illumination uniformity and detail clarity of the image; further, by extracting illumination characteristics through a multi-task deep learning model, the adaptability of the correction process to complex illumination scenes is enhanced; at the same time, the residual diffusion mechanism is utilized to optimize the efficiency and effect of the correction. After the corrected image is generated into a high-quality image through post-processing, the high-quality image exhibits higher accuracy and robustness in unmanned ship target recognition and path planning tasks, and is suitable for a variety of complex navigation environments.

[0091] Example 2:

[0092] See also Figure 2A second embodiment of the present invention provides an unmanned vessel dynamic illumination correction system based on a residual diffusion model, comprising:

[0093] Data acquisition module, used to obtain image data of the unmanned ship under different lighting conditions and record the ambient lighting characteristics;

[0094] The illumination perception module is used to obtain the illumination characteristics of the input image, including illumination intensity, shadow distribution, and reflection area identification, providing the necessary illumination distribution data information for the subsequent residual diffusion model;

[0095] The residual diffusion illumination correction module dynamically corrects illumination based on the residual diffusion model and generates a corrected image;

[0096] Output and application module, outputs corrected images to support applications such as target recognition and path planning;

[0097] The above modules complete the logical link in a progressive relationship.

[0098] The data acquisition module further includes a pre-processing unit for performing preliminary processing on the received image data.

[0099] The residual diffusion illumination correction module further includes a post-processing unit for performing post-processing on the corrected image, including contrast adjustment and detail enhancement, so that the image quality of the corrected image is enhanced and optimized, and then outputting the corrected image to the output and application module.

[0100] Among them, the lighting perception module uses a deep learning model to extract light intensity, shadow distribution and reflection area characteristics; the output and application module supports real-time correction and dynamic update.

[0101] A second embodiment of the present invention is an unmanned ship dynamic illumination correction system based on a residual diffusion model. By introducing an illumination perception module and a residual diffusion illumination correction module, it can effectively cope with lighting conditions such as strong reflection, shadow coverage, and uneven illumination, and significantly improve the clarity and detail visibility of the unmanned ship image.

[0102] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for dynamic illumination correction of unmanned vessels based on residual diffusion model, characterized in that: It includes the following steps: S1: Acquire the unmanned ship navigation image and its lighting condition data; S2: Use the illumination perception module to extract illumination characteristics, including illumination intensity, shadow distribution, and reflection area; S3: Perform illumination correction based on the residual diffusion model, which includes the following two processing steps: Forward diffusion process: simulates the dynamic distribution of light interference; Reverse diffusion process: gradually removes light interference and generates a corrected image; S4: Output the corrected image for target recognition and path planning of the unmanned vessel.

2. The method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model according to claim 1, characterized in that: The S2 includes the following sub-steps: S21. Construction and Training of a Multi-Task Deep Learning Model: Build a multi-task deep learning model that includes multiple task branches for light intensity extraction, shadow area detection, and reflection area recognition. S22. Light Intensity Extraction: Extract the intensity distribution of light in the image and generate a brightness map to reflect the overall lighting conditions. S23. Shadow Area Detection: Utilizes the shadow detection branch of a multi-task deep learning model. Identify shadow areas in images and extract the boundaries, shapes, and coverage of shadows; S24. Reflective Area Recognition: Using the reflective area recognition branch of a multi-task deep learning model, we locate highly reflective areas in an image and distinguish between reflective and non-reflective areas. S25. Integrate the light intensity, shadow distribution, and reflection area characteristics output by the multi-task deep learning model to generate a light characteristic description to provide input for subsequent light correction.

3. The method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model according to claim 1, characterized in that: In step S3, the residual diffusion model is used to model and correct illumination interference by adding illumination conditions. The specific processing steps of the forward diffusion process are as follows: when adding noise to the image, the illumination distribution characteristics are used as conditional input, and the dynamic illumination characteristics are combined to affect the distribution of noise. The noise generation formula is: Among them, x t : target image or image at the current moment; x t-1 : The image at the previous moment, which serves as the basis for the current image; l: Lighting conditions, which adjust the lighting effects of the image; X res ; Residual, the difference between the target image and the degraded image; β t : diffusion coefficient, controlling the intensity of noise and residual; I: unit matrix, representing the covariance of noise; N(μ,σ 2 ): Gaussian distribution, used to model the probability distribution of the current image, where the mean Variance σ 2 =β t ·I.

4. The method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model according to claim 1, characterized in that: In step S3, the residual diffusion model is used to model and correct illumination interference by adding illumination conditions. The specific processing steps of the reverse diffusion process are as follows: reverse diffusion gradually removes noise to restore the target image. The residual is used to guide the denoising direction in this process, accelerate and optimize image restoration, and the reverse diffusion formula is: x t-1 =x t -∈θ(x t ,t,l)+λ·X res Among them, x t : The image state at the current time step t; x t-1 : image state at time step t-1; ∈θ(x t ,t,l): conditional denoising network, predicting noise; λ: weight, controlling the strength of residual correction; X res : Residual, the difference between the target image and the degraded image.

5. The method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model according to claim 1, characterized in that: The network structure of the residual diffusion model adds an illumination condition embedding module and a residual diffusion fusion module, which includes the following contents: Lighting condition embedding module: Builds global and local embedding mechanisms to integrate the lighting distribution feature l into the diffusion model network; global embedding: uses a multi-layer perceptron (MLP) to globally encode the lighting characteristics and generate a lighting condition vector; local embedding: uses a convolutional network to locally encode the lighting characteristics and generate a lighting condition feature map of the same size as the input image; Residual Diffusion Fusion Module: The encoder and decoder parts of the diffusion model are adjusted to introduce dynamic guidance of the residual information Xres. In the encoder, the illumination condition features and residual information are fused with the image features through feature splicing or additive fusion. A residual feature module is added to the decoder to dynamically adjust the denoising direction and enhance the image correction effect.

6. A method for dynamic illumination correction of an unmanned vessel based on a residual diffusion model according to claim 1, characterized in that: The loss function of the residual diffusion model includes the following parts: Denoising reconstruction loss: Optimize denoising accuracy by minimizing the error between the noise predicted by the conditional denoising network and the actual noise. The formula is: L denoise =||∈θ(x t ,t,l)-∈|| 2 Among them, ∈θ(x t ,t,l) is the output noise prediction of the denoising network, ∈ is the real noise; Light correction loss: Optimize the corrected image according to the light distribution characteristics so that its light characteristics are consistent with the target. The formula is: L light =||f(x output )-f(x input )|| 2 Among them, f(x) is the illumination feature extraction function (illumination mean), x output and x input are the corrected and original images respectively; Detail enhancement loss: Through gradient or edge constraints, the detail recovery effect of the corrected image is enhanced. The formula is: in, represents the gradient operator, is the real target image; Total loss function: Combining the above three loss functions, the weight parameters λ1, λ2, and λ3 are used to balance the effects of different losses, and is defined as: L = λ1L denoise +λ2L light +λ3L detail .

7. A dynamic illumination correction system for unmanned vessels based on a residual diffusion model, characterized in that: include: A data acquisition module is used to obtain image data of the unmanned ship under different lighting conditions; Light perception module, used to obtain the lighting characteristics of the input image; The residual diffusion illumination correction module dynamically corrects illumination based on the residual diffusion model and generates a corrected image; Output and application module, outputs corrected images to support target recognition and path planning of unmanned vessels; The above modules complete the logical link in a progressive relationship.

8. The unmanned vessel dynamic illumination correction system based on residual diffusion model according to claim 7, characterized in that: The light sensing module further includes a pre-processing unit for performing preliminary processing on the received image data.

9. The unmanned vessel dynamic illumination correction system based on the residual diffusion model according to any one of claim 7, characterized in that: The residual diffusion illumination correction module further includes a post-processing unit for post-processing the corrected image and then outputting the post-processed image to the output and application module.

Images