Low-brightness image security enhancement method and device based on deep learning

Abstract

The invention discloses a low-brightness image security enhancement method and device based on deep learning. The method comprises the following steps: firstly, collecting a large number of low-brightness images covering different scenes, illumination conditions and object contents, and preprocessing the low-brightness images; then a deep learning model is constructed, the model comprises an image light simulation module and an image enhancement module, and the image light simulation module is used for simulating the influence of a light source on an image based on low-brightness image data, calculating scattering, reflection and shadow effects after interaction of light and the image, generating light simulation data and enhancing the light simulation data; the image enhancement module is used for adjusting related parameters for enhancement processing according to the light simulation data; then training the constructed deep learning model by using the preprocessed low-light image data; and finally, enhancing a to-be-processed low-light image by using the trained deep learning model. According to the method, the identification accuracy and the application efficiency of the model can be remarkably improved, the image enhancement quality is improved, and the enhanced image shows loyalty physical rationality and detail authenticity.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and more specifically, to a low-light image enhancement security method and device based on deep learning. Background Art

[0002] Low-light images refer to images taken in an environment with insufficient lighting. Due to insufficient lighting, low-light images usually have problems such as insufficient feature information, loss of detail data, color distortion, noise interference, etc., which are easy to be maliciously interpreted, forged and spread, and pose threats and safety hazards. Low-light image enhancement technology can significantly improve the quality and visibility of images under low-light conditions. It is particularly important for security monitoring systems. After enhancement, it can improve the accuracy of target recognition, reduce false alarms and missed alarms of security monitoring systems, and enhance the efficiency and safety reliability of the system. In the field of autonomous driving, low-light image enhancement technology can improve the vehicle's intelligent driving system at night or in low light conditions, so as to better identify pedestrians, obstacles, road signs, etc., thereby improving the safety of autonomous driving. Low-light image enhancement technology can improve image quality, make images more realistic and credible, and prevent misunderstandings and wrong decisions caused by image quality problems. At the same time, it reduces the risk of malicious tampering of images. The enhanced image features are obvious and more difficult to be tampered with without being discovered. Any slight change may affect the overall quality of the image. Low-light image enhancement technology can improve the visual effects of images and extract more feature detail information, enhance the authenticity and credibility of images, and better serve subsequent image processing tasks such as segmentation, detection, tracking, etc. These tasks also play an important role in information security and cyberspace security.

[0003] In the field of low-light image enhancement, traditional methods and deep learning-based methods are covered. Traditional methods include those based on histogram equalization, such as brightness-preserving dynamic histogram equalization, which avoids visual quality degradation by maintaining average brightness, and intelligent contrast enhancement technology based on traditional histogram equalization algorithm, which partitions and processes enhanced images; probability-based methods have also attracted much attention, such as multi-scale fusion dehazing models. However, traditional methods have limitations such as manual parameter adjustment, possible color distortion or over-enhancement, amplified noise, and poor processing of motion blur. In recent years, low-light image enhancement based on deep learning has developed rapidly, but many low-light image enhancement security methods based on deep learning have problems such as high training cost, slow processing speed, and limited detail processing effect. Summary of the invention

[0004] In order to solve the problem of limited detail processing effect in the prior art, the present invention proposes a low-light image enhancement security method and device based on deep learning.

[0005] In order to solve the above technical problems, the first aspect of the present invention provides a low-light image enhancement security method based on deep learning, comprising: Collect a large number of low-light images covering different scenes, lighting conditions and object contents, and pre-process the collected low-light images; Constructing a deep learning model, the deep learning model includes an image light simulation module and an image enhancement module, wherein the image light simulation module is used to simulate the influence of the light source on the image based on low-brightness image data, calculate the scattering, reflection and shadow effects after the interaction between light and image, and generate light simulation data; the image enhancement module is used to adjust relevant parameters for enhancement processing according to the light simulation data, and output the enhanced image; the model adopts a combination of mean square error and structural similarity loss function to guide the model to focus on image structure and detail preservation while improving brightness, and introduces a perceptual loss function to ensure that the enhanced image is similar to the reference image in high-level semantic features; The constructed deep learning model is trained using the preprocessed low-brightness image data; Use the trained deep learning model to enhance the low-light image to be processed.

[0006] In one embodiment, preprocessing the original low-brightness image data includes: Coordinate calibration, resolution adjustment and format conversion are performed on the original low-light image to generate standardized pre-processed low-light image data.

[0007] In one embodiment, the preprocessed low-brightness image data is used to train the constructed deep learning model, including: Construct a training dataset based on the preprocessed low-brightness image data; The training data set is input into the constructed deep learning model, and the model parameters are optimized by calculating the difference between the model prediction and the actual expectation. In each iteration, the gradient is calculated according to the loss function, and then the weights and bias of the network are updated using the back propagation algorithm. The loss function is constructed based on the mean square error, structural similarity loss function, and perceptual loss function.

[0008] In one implementation, the model uses a combination of mean square error and structural similarity loss functions to guide the model to focus on image structure and detail preservation while improving brightness, including: The mean square error loss function is used to measure the difference between the enhanced image and the reference normal illumination image:

[0009] in, is the total number of image pixels, is the i-th pixel of the input low-light image, is the i-th pixel of the enhanced image output by the model, is the mean square error loss function; The structural similarity loss function is introduced to measure the structural similarity between two images:

[0010] in, is the input low-light image, is the enhanced image output by the model, It is a function to measure the structural similarity between two images. The closer the value is to 1, the more similar the image structures are, the smaller the image distortion is, and the more similar and reliable the enhancement effect is. is the structural similarity loss function; The perceptual loss function is:

[0011] in, is the number of feature layers, Respectively l The number of channels, height and width of the layer feature map, is the reference image, is the enhanced image output by type, It is l The perceptual features of the enhanced image, It is l The perceptual features of the reference images, Indicates the number of channels, height, and width of the image.

[0012] In one embodiment, the image enhancement module includes a parameter adjustment submodule, a detail enhancement submodule and an image output submodule, wherein the parameter adjustment submodule is used to analyze the color distribution and light intensity in the image based on the light simulation data, adjust the parameters one by one to match the predetermined visual effect, and adjust the brightness level and color saturation to generate image parameter optimization data; the detail enhancement submodule is used to use the image parameter optimization data to perform hierarchical separation, sharpness and contrast adjustment on the details in the image, and optimize the key visual elements in the image; the image output submodule is used to adjust the resolution and perform color correction through the detail-optimized image, and output the enhanced image.

[0013] In one embodiment, the detail enhancement submodule is specifically used for: Use the gradient information of the image to construct a detail feature map; Based on the detail feature map, a multi-scale processing method is used to enhance the details. For each scale of the detail feature map, a detail enhancement filter is designed for processing. A noise suppression mechanism is introduced in the detail enhancement process. The degree of detail enhancement is adaptively adjusted according to the local variance or noise estimation value of the image. The feature maps of different scales after detail enhancement are fused to obtain the detail enhancement result; The detail enhancement result is fused with the image after brightness and color adjustment to obtain the final enhanced image.

[0014] In one embodiment, the method also includes: performing content review on the enhanced image, evaluating its authenticity and credibility by analyzing the feature distribution visibility, detail rationality, and distortion of the enhanced image, and generating image data with anti-counterfeiting properties by embedding digital watermarks and performing image content monitoring.

[0015] Based on the same inventive concept, the second aspect of the present invention provides a low-light image enhancement security device based on deep learning, comprising: The data collection and preprocessing module is used to collect a large number of low-light images covering different scenes, lighting conditions and object contents, and preprocess the collected low-light images; A model building module is used to build a deep learning model, wherein the deep learning model includes an image light simulation module and an image enhancement module, wherein the image light simulation module is used to simulate the influence of the light source on the image based on low-brightness image data, calculate the scattering, reflection and shadow effects after the interaction between light and image, and generate light simulation data; the image enhancement module is used to adjust relevant parameters for enhancement processing according to the light simulation data, and output the enhanced image; the image enhancement module adopts a combination of mean square error and structural similarity loss function to guide the model to focus on image structure and detail retention while improving brightness, and introduces a perceptual loss function to ensure that the enhanced image is similar to the reference image in high-level semantic features; A training module is used to train the constructed deep learning model using the preprocessed low-brightness image data; The image processing module is used to enhance the low-light image to be processed using the trained deep learning model.

[0016] Based on the same inventive concept, the third aspect of the present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the deep learning-based low-light image enhancement security method described in the first aspect is implemented.

[0017] Based on the same inventive concept, the fourth aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the low-light image enhancement security method based on deep learning described in the first aspect is implemented.

[0018] Compared with the prior art, the advantages and beneficial technical effects of the present invention are as follows: The present invention provides a low-light image enhancement security method based on deep learning, and constructs a low-light image enhancement model based on deep learning. Through deep learning, a large amount of low-light image data can be processed to realize the calibration transformation of image content, so that the model can recognize the key feature structure in the image, thereby significantly improving the recognition accuracy and application efficiency of the model. Combining ray tracing technology and deep learning neural renderer, the physical propagation process of light is simulated to improve the quality of image enhancement. Using the low-light image enhancement algorithm of the image enhancement module, the enhanced image shows faithful physical rationality and detail authenticity, and the effect of image enhancement is improved. The method is applicable to general objects, endowed with multi-purpose applications, and provides users with a low-light image enhancement experience.

[0019] Furthermore, through artificial intelligence technology, the optimized images are subjected to credible analysis and content review to ensure the authenticity and credibility of the images, digital watermarks are embedded and image monitoring is performed to ensure the security and traceability of the image content, effectively integrating the accuracy, authenticity, credibility and security of image enhancement, and comprehensively improving the reliability and performance of low-light image enhancement during use. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 It is a flowchart of a low-light image enhancement security method based on deep learning disclosed in an embodiment of the present invention; Figure 2 It is a framework structure diagram of a low-light image enhancement safety device based on disclosed deep learning in an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a computer-readable storage medium provided in an embodiment of the present invention; Figure 4 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention; Figure 5 2 is a structural diagram of the artificial intelligence security and image trust module in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The model is constructed using a deep convolutional neural network and trained based on a large number of low-light image data sets. A combination of mean square error and structural similarity loss functions is used to guide the model to focus on image structure and detail preservation while improving brightness. A perceptual loss function is introduced to ensure that the enhanced image is similar to the reference image in high-level semantic features to ensure consistency of visual effects. A step-by-step optimization strategy is adopted to fine-tune parameters based on evaluation indicators to avoid over-adjustment and improve image quality and model stability. Images and feature maps of different resolutions are obtained through multi-scale processing methods such as Gaussian pyramid decomposition. Feature maps are processed using detail enhancement filters to enhance high-frequency detail information. A noise suppression mechanism is introduced to adaptively adjust the degree of detail enhancement based on local variance to avoid noise amplification. Adaptive strategies and other deep learning technologies can also be combined to further optimize details. In terms of security and reliability, it is easy to cause problems such as enhanced effects that do not conform to physical laws and enhanced details that are false and unreliable in enhanced images, which cannot guarantee the security and reliability of enhanced images. Low-light image enhancement technology still faces many challenges, and more innovative methods are needed to improve the quality, efficiency, and authenticity of low-light image enhancement.

[0023] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0024] Embodiment 1 This embodiment discloses a low-light image enhancement security method based on deep learning, see Figure 1 ,include: S1: Collect a large number of low-light images covering different scenes, lighting conditions and object contents, and pre-process the collected low-light images; S2: Construct a deep learning model, which includes an image light simulation module and an image enhancement module, wherein the image light simulation module is used to simulate the influence of the light source on the image based on low-brightness image data, calculate the scattering, reflection and shadow effects after the interaction between light and image, and generate light simulation data; the image enhancement module is used to adjust relevant parameters for enhancement processing according to the light simulation data, and output the enhanced image. The model adopts a combination of mean square error and structural similarity loss function to guide the model to focus on image structure and detail retention while improving brightness, and introduces a perceptual loss function to ensure that the enhanced image is similar to the reference image in high-level semantic features; S3: Use the preprocessed low-brightness image data to train the constructed deep learning model; S4: Use the trained deep learning model to enhance the low-light image to be processed.

[0025] Specifically, S1 is the collection and preprocessing of low-light images, S2 is the construction of the model, S3 is the training of the model, and S4 is the testing or application of the model.

[0026] Normalizing low-light image data focuses on standardizing image resolution and unifying its data format to ensure data consistency and compatibility for subsequent processing. The trained deep model covers key factors such as the specific type of activation function, the weight distribution within the model, and the number of training cycles. These factors together determine the performance of the model and its ability to handle low-light images. Image feature data involves the topological architecture of the reconstructed grid, the grid density distribution, and the precise coordinate information of key feature points. These features are essential for accurately understanding and analyzing image structures. Light simulation data includes parameter settings for the location of the light source, the intensity level of the light, and the angular distribution of reflected and scattered light. By accurately simulating the light effect, it provides a more realistic lighting environment reference for image enhancement. The optimized and enhanced image clarifies the image resolution specifications, dynamic range expansion, and color correction related parameters, thereby significantly improving the image's visual effect. Image credibility mainly considers the image's distortion, the rationality of the details, and the visibility of the feature distribution, so as to comprehensively evaluate the authenticity and credibility of the image.

[0027] In one embodiment, the method further includes S5: performing content review on the enhanced image, evaluating its authenticity and credibility by analyzing the feature distribution visibility, detail rationality, and distortion of the enhanced image, and generating image data with anti-counterfeiting characteristics by embedding digital watermarks and performing image content monitoring.

[0028] In one embodiment, preprocessing the original low-brightness image data includes: Coordinate calibration, resolution adjustment and format conversion are performed on the original low-light image to generate standardized pre-processed low-light image data.

[0029] Specifically, the preprocessing process can be achieved in the following ways: Process the initial low-brightness image data, perform coordinate consistency check and alignment, and generate aligned image data; Detecting and adjusting the image pixel resolution of the aligned image data to generate calibrated image data; Based on the calibrated image data, a data format conversion operation is performed to convert the image data into a unified format to generate standardized pre-processed data.

[0030] In the specific implementation process, processing the initial low-brightness image data and performing coordinate consistency check and alignment can be achieved in the following ways: First, key point detection is performed to extract significant feature points from the image. These feature points are then carefully compared with reference points in a preset standard model library. The key point coordinates are accurately optimized using the least squares method, which involves calculating the deviation between the data points and the reference points, and removing or adjusting those points whose deviations exceed the preset error threshold. This measure ensures high standards in accuracy and consistency of the output aligned images. At the same time, a detailed error log is established to record the accuracy details and correction strategies adopted during each alignment process, so as to continuously analyze and improve future alignment work. Through this series of rigorous operations, the image data finally generated achieves a high degree of alignment accuracy.

[0031] The detection and adjustment of the image pixel resolution of the aligned image data can be achieved in the following ways: Based on the image data after coordinate alignment, each pixel is analyzed and compared with its neighboring pixels in terms of grayscale difference and color to evaluate the necessity of resolution adjustment. By referring to the preset resolution standard, the difference between the current image resolution and the standard resolution is identified, and the most appropriate resolution configuration is selected for adjustment. The image resolution is precisely adjusted using efficient interpolation or downsampling technology, and it is optimized iteratively until the goal of minimizing the average resolution error is achieved. After the calibration is completed, the changes of each pixel before and after the resolution adjustment are recorded in detail, and a calibration report and image data containing comparative data are generated.

[0032] Based on the calibrated image data, a data format conversion operation is performed to convert the image data into a unified format to generate standardized pre-processed data, which can be achieved in the following manner: Based on the calibrated image data, evaluate the calibration of the image data and select JPEG or PNG according to the needs. For images with large grayscale, high resolution, and large image files, select the JPEG format, compress the data during the conversion process, and then optimize the processing speed and storage requirements, and finally generate a unified image format to facilitate subsequent image processing and analysis. For images with small grayscale, low resolution, and small image files, select the PNG format, which supports transparency operations and losslessly processes the data during the conversion process, which is suitable for most situations. Finally, a unified image format is generated to facilitate subsequent image processing and analysis.

[0033] In one embodiment, the preprocessed low-brightness image data is used to train the constructed deep learning model, including: Construct a training dataset based on the preprocessed low-brightness image data; The training data set is input into the constructed deep learning model, and the model parameters are optimized by calculating the difference between the model prediction and the actual expectation. In each iteration, the gradient is calculated according to the loss function, and then the weights and bias of the network are updated using the back propagation algorithm. The loss function is constructed based on the mean square error, structural similarity loss function, and perceptual loss function.

[0034] Specifically, S3 builds a training data set based on the preprocessed low-light image data, trains the model, and trains the model to learn the features and structure of the low-light image, and finally obtains a trained deep learning model. The specific process includes the following: Based on the training data set, the network layer structure and activation function are selected, the input and output nodes and the number of network layers are defined, and the initialization deep learning model framework is generated; using the initialization deep learning model framework, low-brightness image data is input for feature extraction, the structural information and texture features of the image are analyzed, the model parameters are optimized, and a feature optimization model is generated; then based on the feature optimization model, multiple rounds of training are performed, the learning rate and loss function are adjusted, and the generalization ability of the model is optimized using a batch training method to generate a trained deep learning model, wherein in each round of training, the loss is calculated based on the difference between the current noise prediction and the actual noise to guide the model to learn how to process noise terms at different time steps, and the noise terms are used to reflect the noise accumulation effect based on the time step.

[0035] Among them, based on the normalized low-brightness image data, the network layer structure and activation function are selected, the number of network layers and output nodes are defined, and the specific process of generating the initialization deep learning model framework is as follows: Generate an initialized deep learning model framework based on normalized low-brightness image data. This process includes data feature analysis, network structure selection, and network parameter initialization. First, by statistically analyzing the distribution uniformity and feature density in the low-brightness image data, select the appropriate network layer structure, such as deciding how many convolutional layers to use and the number of filters in each layer. For the activation function, select Leaky ReLU according to the data characteristics to increase the ability to calculate nonlinear processing, or Log-Sigmoid to achieve the probability distribution of the output layer. Set the width and depth of the network to achieve the expected processing speed and recognition accuracy, and generate an initialized deep learning model framework.

[0036] The specific process of initializing the deep learning model framework, inputting low-brightness image data for feature extraction, analyzing the image's structural information and texture features, optimizing the model parameters, and generating a feature optimization model is as follows: The deep learning model framework is initialized, and the processed low-brightness image data (pre-processed low-brightness image data) is input to extract features layer by layer through the network layer. The convolution layer extracts local features of the image through filters, and the pooling layer reduces the spatial dimension of the features and reduces the amount of calculation. In order to stabilize the learning process, a normalization layer may be connected after each convolution layer. By iterating and adjusting the network parameters, the model parameters including bias and weight are optimized. In this process, the loss function is used to evaluate the difference between the actual results and the predicted results. The mean absolute percentage (MAPE) is used as the loss function. The loss function expression is: ,in is the sample size, It is The actual value of the samples, It is The prediction value of each sample is obtained. The loss is optimized by gradient descent method to improve the model's fitting accuracy to the data and generate a feature learning model.

[0037] Based on the feature optimization model, multiple rounds of training are performed, the loss function and learning rate are adjusted, and the generalization ability of the model is optimized using the batch training method. The specific process of generating the trained deep learning model is as follows: Based on the feature optimization model, multiple rounds of training are set to meet the learning needs of different stages. Adjust the loss function and learning rate to improve the training process and the generalization ability of the model. For example, a higher learning rate may be used for rapid convergence in the early stage, and the learning rate may be gradually reduced as the training progresses to refine the learning effect. Using the batch training method, a batch of data is used in each iteration to reduce memory usage, while the statistical characteristics of batch data are used to stabilize the gradient estimation. As the number of training rounds increases, the loss of the model on the training set may continue to decrease, but the loss on the validation set begins to rise after reaching a minimum value. At this time, the Early Stopping technology is introduced to prevent overfitting. Through repeated training, adjustment of network parameters, and improvement of network structure, a trained deep learning model is finally generated. This model can effectively process and explain the complex structure of low-brightness image data.

[0038] In one implementation, the model uses a combination of mean square error and structural similarity loss functions to guide the model to focus on image structure and detail preservation while improving brightness, including: The mean square error loss function is used to measure the difference between the enhanced image and the reference normal illumination image:

[0039] in, is the total number of image pixels, is the i-th pixel of the input low-light image, is the i-th pixel of the enhanced image output by the model, is the mean square error loss function; The structural similarity loss function is introduced to measure the structural similarity between two images:

[0040] in, is the input low-light image, is the enhanced image output by the model, It is a function to measure the structural similarity between two images. The closer the value is to 1, the more similar the image structures are, the smaller the image distortion is, and the more similar and reliable the enhancement effect is. is the structural similarity loss function; The perceptual loss function is:

[0041] in, is the number of feature layers, Respectively l The number of channels, height and width of the layer feature map, is the reference image, is the enhanced image output by type, It is l The perceptual features of the enhanced image, It is l The perceptual features of the reference images, Indicates the number of channels, height, and width of the image.

[0042] Specifically, before training the model for low-light image enhancement, a series of preparations need to be done to build an effective training environment. First, a large number of low-light images covering different scenes, lighting conditions, and object contents are collected as training data sets to ensure that the model can learn image features and enhancement modes in various low-light situations.

[0043] For the model architecture, a deep convolutional neural network is used, which contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer is used to extract local features of the image, such as capturing details such as edges and textures in low-light images through convolution kernels of different sizes. As the network depth increases, more abstract global features such as object shape and scene structure are gradually learned. The pooling layer plays a downsampling role, reducing the amount of data and retaining key information, improving computational efficiency. The fully connected layer is used to integrate the previously extracted features and finally generate an enhanced image.

[0044] During the training process, in order to guide model learning, define a suitable loss function. is the i-th pixel of the input low-light image, is the i-th pixel of the image after model enhancement. The mean square error loss function is used to measure the difference between the enhanced image and the reference normal illumination image.

[0045] In order to better preserve image details, the structural similarity (SSIM) loss function is introduced , SSIM is used to measure the structural similarity between two images. The closer its value is to 1, the more similar the image structures are, the smaller the image distortion is, and the more similar and reliable the enhancement effect is. By minimizing , so that the model can focus on maintaining image structure and details while improving brightness.

[0046] In order to ensure that the enhanced image is visually similar to the reference image, a perceptual loss function is introduced. For the image The extracted perceptual features (which can be extracted through a pre-trained neural network) minimize the perceptual loss, making the enhanced image close to the reference image in terms of high-level semantic features, thereby ensuring the consistency of the visual effect.

[0047] The model parameters are optimized by calculating the difference between the model prediction and the actual expectation. In each iteration, the gradient is calculated according to the loss function, and then the weights and biases of the network are updated using the backpropagation algorithm. For example, for the weights of the convolutional layer , the update formula is ,in is the learning rate, which determines the step size of each update. is the updated weight, is the loss function of the model. Through multiple iterative training, the model gradually learns the best mode for low-light image enhancement and can generate high-quality, detailed, and appropriately bright enhanced images based on the input low-light images.

[0048] In one embodiment, the processing flow of the image light simulation module includes: Based on low-brightness image data, the light source is configured by optimizing the BRDF model to simulate the effect of natural light or indoor lighting, adjust the light source position and intensity, and set a variety of lighting angles and color temperatures to generate light source configuration data. The BRDF model is a bidirectional reflectance distribution function model; Using the light source configuration data, analyzing the interaction between light and the surface of the low-brightness object, calculating the reflection and scattering effects, adjusting the interactive behavior of light, simulating the performance of light, and generating light image interaction data; Through the light-image interaction data, the shadow area between the light source and the image object is located, the intensity and extension of the shadow are calculated, the blurriness and boundary of the shadow are adjusted, the light-dark contrast in the image is optimized, and the light simulation data is obtained.

[0049] By optimizing the BRDF model to configure the light source to simulate the effect of natural light or indoor lighting, adjusting the light source position and intensity, and setting a variety of lighting angles and color temperatures, the specific process of generating light source configuration data is as follows: First, the BRDF model is used to simulate different types of light effects, such as natural light. According to the needs of the model, the position of the light source is adjusted to match the actual environment of the scene, such as the position of vehicles or pedestrians. In addition, the intensity of the light source is adjusted to match the changes in natural light at different times. The light effect is further refined by changing the color temperature and light angle, such as the sunset color temperature of the sun. The precise setting of these parameters ensures that the simulated light can realistically reflect the lighting conditions in the real world and generate detailed light configuration data.

[0050] The optimized BRDF model is according to the formula:

[0051] Calculate the relationship between incident light and reflected light from various angles to simulate natural lighting effects; Among them, the function Represents the bidirectional reflectance distribution function, which is used to calculate the image in the direction of the incident light and reflected light direction The light reflectivity under is the incident light direction, is the direction of reflected light, is the incident angle, is the observation angle, is the specular highlight attenuation factor, according to the observation angle Adjust the intensity of specular reflection so that the highlight effect weakens as the viewing angle increases. is the light intensity per unit incident direction, indicating the incident direction The light energy received per unit area is is the light radiation intensity in the unit reflection direction, indicating the reflection direction The brightness per unit area, is the diffuse reflection coefficient, which indicates the proportion of the diffuse reflection part. is the cosine of the angle and is used to calculate the effective component of the light incident on the surface.

[0052] It should be noted that for a three-dimensional image, it can be understood as being mapped into a three-dimensional object, and the material properties must be considered; the present invention processes a two-dimensional image, focusing on the quality processing of the image itself, and does not map it to three dimensions. Therefore, optimizing the BRDF model does not require considering the material.

[0053] During the specific implementation process, the execution process is as follows: Determine the angle parameters and : (Incident angle) is determined by the angle between the incident light direction and the normal. In actual calculations, It can be obtained by taking the arc cosine of the dot product of the light source direction vector and the normal vector: That middle, is the unit normal vector, is the unit incident ray vector.

[0054] (Observation angle) is determined by the angle between the direction of the reflected light and the direction of the observer. , It can be obtained by taking the arc cosine of the dot product of the reflected light direction vector and the viewing direction vector: in, is the reflection vector, is the observation vector.

[0055] Specular reflection coefficient and diffuse reflectance Determination of: These coefficients are usually preset according to the optical and physical properties of specific objects in the image, which can be obtained from a database or determined through experimental measurements. They reflect the ability of different materials to absorb and reflect light.

[0056] Light intensity and light radiation intensity Calculation: At a given incident direction The light intensity received by the surface of an object is usually obtained by measuring the actual lighting environment or ray tracing. In a given reflection direction The intensity of light radiation reflected from the surface of an object, usually simulated in rendering software through ray tracing algorithms.

[0057] The final BRDF value is calculated by inserting all the parameter values ​​calculated above into the formula, so as to comprehensively consider the influence of light angle and light intensity on the reflection characteristics. Used to accurately simulate and enhance the visual appearance of objects in low-light images under different lighting conditions. This includes how light is absorbed, scattered, and reflected by the surface of an object, enabling the generation of realistic enhanced images, which are widely used in security monitoring, autonomous driving, target recognition, and any application that requires precise lighting.

[0058] The specific process of using light source configuration data, analyzing the interaction between light and the surface of low-light objects, calculating reflection and scattering effects, adjusting the interactive behavior of light, simulating the performance of light, and generating light image interaction data is as follows: Analyze the interaction mechanism between light and the surface of image objects. This module is the first to use the bidirectional reflectance distribution function model to accurately calculate the scattering and reflection of light emitted from the light source when it hits the surface of the image object.

[0059] The light source configuration data is used to analyze the interaction mechanism between light and the surface of the image object. This module first uses the bidirectional reflectance distribution function model to calculate the scattering and reflection effects of light emitted from the light source when it encounters the surface of the image object. Specifically, when calculating the reflection intensity, the incident angle is taken into account. And the reflection angle , the calculation formula is: reflection intensity = light source intensity ×BRDF(θ, )×cos(θ) , where light source intensity represents the initial intensity of the light emitted by the light source. BRDF(θ, ) is the reflection coefficient, cos(θ) Represents the effect of the incident angle on the intensity of light. The scattering effect depends on the scattering coefficient and roughness of the object surface. The calculation formula is: scattered light intensity = light source intensity × scattering coefficient × (1 / 2π), where the scattering coefficient represents the scattering ability of the object surface, and (1 / 2π) is the normalized coefficient of omnidirectional scattering. Through the above series of calculations, this module can effectively simulate the specific performance of light in a variety of environments. The light environment interaction data generated in this way can not only accurately present the physical interaction between light and objects, but also can be well adapted to the actual needs of various visual scenes.

[0060] The specific process of locating the shadow area between the light source and the image object through the light image interaction data, calculating the intensity and extension of the shadow, adjusting the blur and boundary of the shadow, optimizing the light-dark contrast in the image, and obtaining the light simulation data is as follows: Through the interaction data of light and objects in the image, the shadow area caused by the obstruction between the light source and the object is located. Ray tracing technology is used to calculate the straight line path from the light source to each part of the object, and detect whether these paths are blocked by other objects, so as to determine the shadow formation area. The intensity and extension of the shadow are calculated according to the distance and angle of the light source, and the "shadow attenuation factor" is used to adjust the blurriness of the shadow edge. This factor is adjusted as a function of the distance from the light source to the shadow area. The specific calculation formula is: = Shadow intensity / base light intensity, where is the distance from the light source to the point where the shadow is generated, is the attenuation coefficient. This optimizes the light-dark contrast in the image and ultimately obtains accurate light simulation data.

[0061] In one embodiment, the image enhancement module includes a parameter adjustment submodule, a detail enhancement submodule and an image output submodule, wherein the parameter adjustment submodule is used to analyze the color distribution and light intensity in the image based on the light simulation data, adjust the parameters one by one to match the predetermined visual effect, and adjust the brightness level and color saturation to generate image parameter optimization data; the detail enhancement submodule is used to use the image parameter optimization data to perform hierarchical separation, sharpness and contrast adjustment on the details in the image, and optimize the key visual elements in the image; the image output submodule is used to adjust the resolution and perform color correction through the detail-optimized image, and output the enhanced image.

[0062] In the specific implementation process, the parameter adjustment submodule is based on an in-depth analysis of the characteristics of low-light images. First, the input low-light image is statistically analyzed for light intensity and color distribution, and the image's average brightness, brightness histogram, and the mean and variance of each color channel (such as RGB channel) are calculated. Based on these statistical data, the image's brightness and color saturation are adjusted. For brightness adjustment, the overall brightness is increased through linear or nonlinear transformations based on the difference between the image's average brightness and the target brightness. For example, suppose For the image at coordinates The pixel value at , the adjusted brightness , in is the brightness adjustment factor calculated based on the average brightness. If the average brightness is low, For color saturation adjustment, based on the statistical information of the color channels, the channels with lower saturation are enhanced. For example, if the red channel has a lower mean and a smaller variance, the color saturation can be increased by increasing the weight of the red channel. Let the adjusted pixel value be ,in is the saturation adjustment factor, refers to the values ​​of the red channel, green channel, and blue channel. is the adjusted value of the red channel.

[0063] In one embodiment, the detail enhancement submodule is specifically used for: Use the gradient information of the image to construct a detail feature map; Based on the detail feature map, a multi-scale processing method is used to enhance the details. For each scale of the detail feature map, a detail enhancement filter is designed for processing. A noise suppression mechanism is introduced in the detail enhancement process. The degree of detail enhancement is adaptively adjusted according to the local variance or noise estimation value of the image. The feature maps of different scales after detail enhancement are fused to obtain the detail enhancement result; The detail enhancement result is fused with the image after brightness and color adjustment to obtain the final enhanced image.

[0064] In the specific implementation process, the detail enhancement submodule uses image parameter optimization data to perform hierarchical separation of details in the image, adjust sharpness and contrast, and optimize key visual elements in the image, including edges and textures. The specific process of obtaining a detail-optimized image is as follows: The detail enhancement submodule aims to further improve the clarity and richness of the image. First, the gradient information of the image is used to construct the detail feature map. For the original low-light image, calculate its horizontal and vertical gradients and , for example, by calculating the Sobel operator Then, the gradient amplitude As a detail feature map, this feature map reflects detail information such as edges and textures in the image.

[0065] Based on the detail feature map, a multi-scale processing method is used to enhance the details. The original low-brightness image and the detail feature map are downsampled and upsampled at different scales to obtain multiple images and feature map representations of different resolutions. For example, a low-resolution image is obtained by decomposing the Gaussian pyramid. ( Represents different scales) and the corresponding detail feature maps .

[0066] For each scale detail feature map , design a detail enhancement filter for processing. A simple method is to use a Laplace filter , clearer and sharper.

[0067] In order to avoid noise amplification, a noise suppression mechanism is introduced in the detail enhancement process. The degree of detail enhancement is adaptively adjusted according to the local variance or noise estimation value of the image. For example, in areas with large noise, the weight of the detail enhancement filter is reduced and the amplitude of detail enhancement is reduced, thereby maintaining the smoothness of the image while enhancing the details.

[0068] The feature maps of different scales after detail enhancement are fused to obtain the final detail enhancement result. The fusion process can adopt the weighted summation method to assign different weights according to the importance of each scale feature map. For example, a larger weight is given to the feature map of lower scale (higher resolution) because it contains more detail information, while a smaller weight is given to the feature map of higher scale (lower resolution) to maintain overall consistency. Let the fused detail enhanced image be ,but ,in For the The weight of the scale, and , After detail enhancement Feature maps of different scales, is the total number of scales.

[0069] Finally, the detail enhancement result is fused with the image after brightness and color adjustment to obtain the final enhanced image. , then the final enhanced image ,in It is a weight parameter to control the degree of detail enhancement. By adjusting it properly, the effects of brightness, color and detail enhancement can be balanced to obtain a low-light image enhancement result with good visual effects and rich details.

[0070] The specific process of adjusting the resolution and performing color correction on the detail-optimized image to output the enhanced image is as follows: The final resolution adjustment and color correction are performed on the detail-optimized image to adapt to different output requirements and display devices. Color correction is based on the standard configuration in the color management system, adjusting each color channel to achieve real-world color performance. Resolution adjustment is based on the needs of the output media, and the image resolution is adjusted through sampling and resampling technology to ensure that the image maintains the best quality at various sizes. The output optimized rendered image therefore has high-quality clarity and color accuracy.

[0071] In one embodiment, the GPU cost is reduced by using Retinex-Net (an image network for low-light image enhancement). By optimizing the image processing flow in the enhancement task, Retinex-Net is end-to-end trainable and can directly learn from the input low-light image to the output enhanced image without complex manual feature engineering and adjustment of intermediate steps. Retinex-Net uses a multi-scale method to estimate the brightness of the image and can analyze and process the illumination information of the image at different scales. Images taken in low-light environments often contain more noise. Retinex-Net can learn to predict the noise level in the image and denoise the reflectivity image through the reflectivity component denoising network, thereby effectively reducing the noise in the image. Retinex-Net mainly includes Decom-Net (decomposition network) and Enhance-Net (enhancement network). Decom-Net can decompose the input low-light image into two parts: reflectivity image and illumination image. The reflectivity image contains the inherent properties of the image and is independent of the lighting conditions, while the illumination image reflects the illumination information in the image. Through this decomposition, the composition of the image can be better understood, providing a basis for subsequent enhancement processing. Enhance-Net can enhance the illumination component based on the estimated illumination image using methods such as adaptive mapping curves, thereby increasing the overall brightness of the image, improving the visibility of low-light images, making the details in the image more visible, and better restoring the color information in the image to avoid problems such as color distortion. This model is suitable for fields that require high credibility and effect of low-light image enhancement, such as security monitoring, autonomous driving, and target recognition. Compared with most low-light image enhancement deep learning models, the Retinex-Net model is more efficient in resource usage. This model reduces dependence on high-performance computing resources and significantly reduces costs and energy consumption.

[0072] In one embodiment, the method also includes: performing content review on the enhanced image, evaluating its authenticity and credibility by analyzing the feature distribution visibility, detail rationality, and distortion of the enhanced image, and generating image data with anti-counterfeiting properties by embedding digital watermarks and performing image content monitoring.

[0073] Specifically, the specific implementation process of the artificial intelligence security and image trust module is as follows: Based on the enhanced image, the feature information of the image is screened to identify the visibility of feature distribution, rationality of details, and distortion, and the generated analysis results are obtained to measure the authenticity and credibility of the enhanced image.

[0074] Based on the analysis results, check the feature information identified in the image, review it according to the pre-trained data or preset artificial intelligence model, delete images that do not meet the standards, and obtain the review confirmation results.

[0075] Through review and confirmation of the results, a digital watermark is inserted into the image, and a monitoring program is set up to track the distribution and use of the image, ensuring that the source and distribution records of the image can be queried to obtain anti-counterfeiting image data.

[0076] Among them, the feature information of the image is screened, the visibility of feature distribution, the rationality of details, and the distortion are identified, the analysis results are generated, and the authenticity and credibility of the enhanced image are measured. This can be achieved through the following process: By analyzing each pixel one by one through multi-level image scanning technology, the system can fully cover and scan the entire image data and screen the feature information of each area. In this process, the data of each area is accurately analyzed and recorded, and the system uses advanced algorithms to identify the visibility, rationality of details, and distortion of the feature distribution of the area. This sophisticated scanning process ensures the comprehensive recognition of feature information. Subsequently, the system further subdivides and reviews these features, and finally forms a comprehensive credibility analysis result.

[0077] According to the analysis results, check the feature information identified in the image, review it according to the pre-trained data or preset artificial intelligence model, delete the images that do not meet the standards, and obtain the review confirmation results. This can be achieved through the following process: After obtaining the preliminary credibility analysis results, the review system begins to perform a comprehensive inspection of the enhanced image information. During this process, the system compares each identified feature information in the image with the feature standards published within the organization. Through detailed comparison and verification work, it ensures that the feature information in the image meets the specified standards and the image enhancement effect is authentic and reliable. This series of precise comparison and review operations allows images that do not meet the regulations to be identified and removed, thereby ensuring the purity of the image content. Through this meticulous inspection process, the accuracy and reliability of the image review and confirmation results are ultimately ensured.

[0078] After reviewing and confirming the results, insert a digital watermark into the image, set up a monitoring program to track the distribution and use of the image, and ensure that the source and distribution records of the image can be queried. The specific process of obtaining anti-counterfeiting image data is as follows: According to the results of the review and confirmation, the digital watermark is immediately embedded in the enhanced image. This operation is completed through specially designed encoding technology to ensure that the pattern and embedding position of the digital watermark are accurately calculated, which not only ensures the security of the image, but also ensures that the image quality is not affected. At the same time, the system sets up an image monitoring program to monitor the use of the image in real time. The program can record the access and use details of each image to ensure that the source and use of the image can be traced and verified. Through this series of operations, a complete image enhancement anti-counterfeiting data system is formed.

[0079] Embodiment 2 Based on the same inventive concept, this embodiment discloses a low-light image enhancement security device based on deep learning, see Figure 2 ,include: The data collection and preprocessing module 201 is used to collect a large number of low-brightness images covering different scenes, lighting conditions and object contents, and preprocess the collected low-brightness images; A model building module 202 is used to build a deep learning model, wherein the deep learning model includes an image light simulation module and an image enhancement module, wherein the image light simulation module is used to simulate the influence of the light source on the image based on low-brightness image data, calculate the scattering, reflection and shadow effects after the interaction between the light and the image, and generate light simulation data; the image enhancement module is used to adjust relevant parameters for enhancement processing according to the light simulation data, and output the enhanced image; the image enhancement module uses a combination of mean square error and structural similarity loss function to guide the model to focus on image structure and detail retention while improving brightness, and introduces a perceptual loss function to ensure that the enhanced image is similar to the reference image in terms of high-level semantic features; A training module 203 is used to train the constructed deep learning model using the pre-processed low-brightness image data; The image processing module 204 is used to enhance the low-brightness image to be processed by using the trained deep learning model.

[0080] It should be noted that the image processing module 204 is a specific application or test of the model, including two modules: image light simulation and image enhancement. The specific functions of the two modules are the same as the two modules in the model building module 202.

[0081] In one embodiment, the device also includes an artificial intelligence security and image credibility module 205, which is used to perform content review on the enhanced image, evaluate its authenticity and credibility by analyzing the feature distribution visibility, detail rationality, and distortion of the enhanced image, and generate image data with anti-counterfeiting characteristics by embedding digital watermarks and performing image content monitoring.

[0082] In one embodiment, see Figure 5 , the artificial intelligence security and image trust module 205 includes: The credible analysis submodule 2051 is used to screen the feature information of the image based on the enhanced image, identify the visibility of feature distribution, rationality of details, and distortion, generate analysis results, and measure the authenticity and credibility of the enhanced image.

[0083] The content review submodule 2052 is used to check the feature information identified in the image based on the analysis results, review it according to pre-trained data or a preset artificial intelligence model, delete images that do not meet the standards, and obtain review confirmation results.

[0084] The anti-counterfeiting coding submodule 2053 is used to insert a digital watermark into the image through review and confirmation results, set up a monitoring program to track the distribution and use of the image, ensure that the source and distribution records of the image can be queried, and obtain anti-counterfeiting image data.

[0085] Since the device introduced in the second embodiment of the present invention is a device used to implement the low-light image enhancement security method based on deep learning in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, the person skilled in the art can understand the specific structure and deformation of the device, so it is not repeated here. All devices used in the method in the first embodiment of the present invention belong to the scope of protection of the present invention.

[0086] Embodiment 3 Based on the same inventive concept, see Figure 3 The present invention further provides a computer-readable storage medium 300 on which a computer program 311 is stored. When the program is executed by a processor, the method described in the first embodiment is implemented.

[0087] Since the computer-readable storage medium introduced in the third embodiment of the present invention is the computer-readable storage medium used to implement the low-light image enhancement security method based on deep learning in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, those skilled in the art can understand the specific structure and deformation of the computer-readable storage medium, so it is not repeated here. All computer-readable storage media used in the method of the first embodiment of the present invention belong to the scope of protection of the present invention.

[0088] Embodiment 4 Based on the same inventive concept, see Figure 4 The present invention further provides a computer device, including a memory 401, a processor 402, and a computer program 403 stored in the memory and executable on the processor, wherein the processor implements the method described in Embodiment 1 when executing the program.

[0089] Since the computer device introduced in the fourth embodiment of the present invention is a computer device used to implement the low-light image enhancement security method based on deep learning in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, the person skilled in the art can understand the specific structure and deformation of the computer device, so it is not repeated here. All computer devices used in the method of the first embodiment of the present invention belong to the scope of protection of the present invention.

[0090] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0091] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0092] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic creative concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if these modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A low-light image enhancement security method based on deep learning, characterized in that: include: Collect a large number of low-light images covering different scenes, lighting conditions and object contents, and pre-process the collected low-light images; Constructing a deep learning model, the deep learning model includes an image light simulation module and an image enhancement module, wherein the image light simulation module is used to simulate the influence of the light source on the image based on low-brightness image data, calculate the scattering, reflection and shadow effects after the interaction between light and image, and generate light simulation data; the image enhancement module is used to adjust relevant parameters for enhancement processing according to the light simulation data, and output the enhanced image; the model adopts a combination of mean square error and structural similarity loss function to guide the model to focus on image structure and detail preservation while improving brightness, and introduces a perceptual loss function to ensure that the enhanced image is similar to the reference image in high-level semantic features; The constructed deep learning model is trained using the preprocessed low-brightness image data; Use the trained deep learning model to enhance the low-light image to be processed.

2. The low-light image enhancement security method based on deep learning as claimed in claim 1, characterized in that: Preprocessing of the original low-light image data includes: Coordinate calibration, resolution adjustment and format conversion are performed on the original low-light image to generate standardized pre-processed low-light image data.

3. The low-light image enhancement security method based on deep learning as claimed in claim 1, characterized in that: The preprocessed low-light image data is used to train the constructed deep learning model, including: Construct a training dataset based on the preprocessed low-brightness image data; The training data set is input into the constructed deep learning model, and the model parameters are optimized by calculating the difference between the model prediction and the actual expectation. In each iteration, the gradient is calculated according to the loss function, and then the weights and bias of the network are updated using the back propagation algorithm. The loss function is constructed based on the mean square error, structural similarity loss function, and perceptual loss function.

4. The low-light image enhancement security method based on deep learning as claimed in claim 1, characterized in that: The model uses a combination of mean square error and structural similarity loss functions to guide the model to focus on image structure and detail preservation while improving brightness, including: The mean square error loss function is used to measure the difference between the enhanced image and the reference normal illumination image: in, is the total number of image pixels, is the i-th pixel of the input low-light image, is the i-th pixel of the enhanced image output by the model, is the mean square error loss function; The structural similarity loss function is introduced to measure the structural similarity between two images: in, is the input low-light image, is the enhanced image output by the model, It is a function to measure the structural similarity between two images. The closer the value is to 1, the more similar the image structures are, the smaller the image distortion is, and the more similar and reliable the enhancement effect is. is the structural similarity loss function; The perceptual loss function is: in, is the number of feature layers, Respectively l The number of channels, height and width of the layer feature map, is the reference image, is the enhanced image output by type, It is l The perceptual features of the enhanced image, It is l The perceptual features of the reference images, Indicates the number of channels, height, and width of the image.

5. The low-light image enhancement security method based on deep learning as claimed in claim 1, characterized in that: The image enhancement module includes a parameter adjustment submodule, a detail enhancement submodule and an image output submodule, wherein the parameter adjustment submodule is used to analyze the color distribution and light intensity in the image based on the light simulation data, adjust the parameters one by one to match the predetermined visual effect, and adjust the brightness level and color saturation to generate image parameter optimization data; the detail enhancement submodule is used to use the image parameter optimization data to perform hierarchical separation of details in the image, adjust the sharpness and contrast, and optimize the key visual elements in the image; the image output submodule is used to adjust the resolution and perform color correction through the detail-optimized image, and output the enhanced image.

6. The low-light image enhancement security method based on deep learning as claimed in claim 5, characterized in that: The detail enhancement submodule is specifically used for: Use the gradient information of the image to construct a detail feature map; Based on the detail feature map, a multi-scale processing method is used to enhance the details. For each scale of the detail feature map, a detail enhancement filter is designed for processing. A noise suppression mechanism is introduced in the detail enhancement process. The degree of detail enhancement is adaptively adjusted according to the local variance or noise estimation value of the image. The feature maps of different scales after detail enhancement are fused to obtain the detail enhancement result; The detail enhancement result is fused with the image after brightness and color adjustment to obtain the final enhanced image.

7. The low-light image enhancement security method based on deep learning as claimed in claim 1, characterized in that: The method also includes: conducting content review on the enhanced image, evaluating its authenticity and credibility by analyzing the feature distribution visibility, detail rationality, and distortion of the enhanced image, and generating image data with anti-counterfeiting characteristics by embedding digital watermarks and performing image content monitoring.

8. A low-light image enhancement safety device based on deep learning, characterized in that: include: The data collection and preprocessing module is used to collect a large number of low-light images covering different scenes, lighting conditions and object contents, and preprocess the collected low-light images; A model building module is used to build a deep learning model, wherein the deep learning model includes an image light simulation module and an image enhancement module, wherein the image light simulation module is used to simulate the influence of the light source on the image based on low-brightness image data, calculate the scattering, reflection and shadow effects after the interaction between light and image, and generate light simulation data; the image enhancement module is used to adjust relevant parameters for enhancement processing according to the light simulation data, and output the enhanced image; the image enhancement module adopts a combination of mean square error and structural similarity loss function to guide the model to focus on image structure and detail retention while improving brightness, and introduces a perceptual loss function to ensure that the enhanced image is similar to the reference image in high-level semantic features; A training module is used to train the constructed deep learning model using the preprocessed low-brightness image data; The image processing module is used to enhance the low-light image to be processed using the trained deep learning model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the low-light image enhancement security method based on deep learning as described in any one of claims 1 to 7 is implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the low-light image enhancement security method based on deep learning is implemented as described in any one of claims 1 to 7.

Images