Image enhancement method and system for intelligent network connection automobile test scene

Abstract

The invention discloses an image enhancement method and system for an intelligent network connection automobile test scene, and the method comprises the following steps: S1, obtaining original image data in the intelligent network connection automobile test scene, and carrying out the preprocessing of the original image data; s2, constructing a scene feature recognition model, performing scene dynamic feature extraction on the preprocessed image data, and determining environment category parameters of a current test scene; and S3, based on the environment category parameters, calling a corresponding scene adaptive enhancement strategy library. The invention relates to the technical field of intelligent network connection automobile testing. According to the image enhancement method and system for the intelligent networked automobile test scene, a multi-index image quality evaluation system is established, the system comprises the peak signal-to-noise ratio (PSNR), the structural similarity (SSIM) and the information entropy, quantitative evaluation is carried out after enhancement each time, enhancement parameters are automatically adjusted by taking the evaluation result as a feedback signal, a closed-loop iterative optimization process is formed, and the image enhancement efficiency is improved. And the output image is ensured to meet the quality threshold requirement.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicle testing, and more specifically, to an image enhancement method and system for intelligent connected vehicle testing scenarios. Background Technology

[0002] In the development of intelligent connected vehicles, the testing phase is crucial, directly impacting their performance and safety in real-world road environments. Image data, as a vital information source for intelligent connected vehicle testing, contains rich scene information such as road conditions, traffic signs, surrounding vehicles, and pedestrians. It plays an irreplaceable role in testing and validating the perception, decision-making, and control systems of intelligent connected vehicles.

[0003] However, in actual intelligent connected vehicle testing scenarios, the quality of the acquired raw images often fails to meet testing requirements due to various factors. For example, different weather conditions (such as rain, fog, and strong light) can lead to problems such as blurry images, low contrast, and high noise; complex road conditions and traffic environments can also make it difficult to clearly identify target objects in the images; changes in light intensity can cause images to be too bright or too dark, affecting the detail representation of the images. These low-quality images can interfere with the accurate judgment of related intelligent connected vehicle systems, reducing the reliability and effectiveness of the tests.

[0004] Currently, while some image enhancement methods exist, most lack specific design for intelligent connected vehicle testing scenarios. They often employ fixed enhancement parameters and strategies, failing to dynamically adjust according to different testing environments, resulting in poor performance when processing complex and varied intelligent connected vehicle test images. For example, using the same enhancement parameters may not effectively remove noise or improve contrast for images under different weather conditions; similarly, a fixed brightness adjustment strategy may not achieve the desired visual effect for images with varying lighting intensities. Therefore, there is an urgent need for an image enhancement method that can adapt to the characteristics of intelligent connected vehicle testing scenarios to improve image quality and provide strong support for intelligent connected vehicle testing. Summary of the Invention

[0005] The purpose of this invention is to provide an image enhancement method and system* for intelligent connected vehicle test scenarios. It solves the problems of existing image enhancement methods that use fixed enhancement parameters and strategies, which cannot be dynamically adjusted according to different test scenarios. This results in poor performance when processing complex and ever-changing intelligent connected vehicle test images. For images under different weather conditions, using the same enhancement parameters may not be effective in removing noise and improving contrast. For images with different light intensities, fixed brightness adjustment strategies may not be able to achieve the ideal visual effect.

[0006] This invention achieves the above objective through the following technical solution: an image enhancement method for intelligent connected vehicle testing scenarios, the method comprising the following steps: S1. Obtain raw image data in the intelligent connected vehicle test scenario and preprocess the raw image data; S2. Construct a scene feature recognition model, extract dynamic scene features from the preprocessed image data, and determine the environmental category parameters of the current test scene; S3. Based on the environment category parameters, call the corresponding scene adaptive enhancement strategy library and adjust the image enhancement parameters in real time through dynamic image enhancement algorithm; S4. Use the adjusted enhancement parameters to enhance the preprocessed image data and generate the enhanced image. S5. Evaluate the quality of the enhanced image. If the evaluation result meets the preset threshold, output the enhanced image; otherwise, iteratively optimize the enhancement parameters until the threshold is met.

[0007] Furthermore, in step S1, the raw image data of the intelligent connected vehicle test scenario is acquired, and the raw image data is preprocessed, specifically including: The image acquisition equipment on the intelligent connected vehicle is used to collect multiple frames of raw image data in the test scenario to form an image sequence. The original image data is preprocessed, which includes at least noise removal and format normalization. Noise removal uses a filtering algorithm to reduce noise in each frame of the original image, and format normalization normalizes the pixel values ​​of the denoised image to a preset range, resulting in a preprocessed image sequence.

[0008] Furthermore, the filtering algorithm is a Gaussian filtering algorithm, and the size of the filtering kernel is adaptively set according to the image noise intensity; The preset numerical range is [0,1]. Pixel value normalization is achieved by calculating the difference between the pixel value of the image after noise reduction and the minimum pixel value of the image, and then dividing it by the difference between the maximum and minimum pixel values ​​of the image.

[0009] Furthermore, in step S2, a scene feature recognition model is constructed to extract dynamic scene features from the preprocessed image data and determine the environmental category parameters of the current test scene, specifically including: A scene feature recognition model based on spatial feature extraction network and temporal feature extraction network is constructed. The input of the model is a preprocessed image sequence, and the output is environmental category parameters. The environmental category parameters include at least one or more of the following: weather condition parameters, road condition parameters, traffic environment parameters, and light intensity parameters. Spatial features of each preprocessed image frame are extracted using a spatial feature extraction network to obtain a spatial feature sequence; Spatial feature sequences are input into a temporal feature extraction network to capture the dynamic temporal features of the scene; The dynamic temporal features are classified using a classification network, and environmental category parameters are output.

[0010] Furthermore, the spatial feature extraction network is a convolutional neural network, the temporal feature extraction network is a long short-term memory network, and the classification network is a fully connected layer.

[0011] Furthermore, in step S3, based on the environment category parameters, the corresponding scene adaptive enhancement strategy library is invoked, and the image enhancement parameters are adjusted in real time through a dynamic image enhancement algorithm, specifically including: A scene-adaptive enhancement strategy library is constructed. The strategy library contains enhancement strategies corresponding to different environment category parameters. Each strategy has a preset initial set of enhancement parameters. The enhancement parameters include at least one or more of the following: contrast adjustment coefficient, brightness adjustment coefficient, and detail enhancement intensity coefficient. Based on the environment category parameter, match the corresponding initial enhancement strategy from the strategy library and obtain the initial enhancement parameters; A dynamic image enhancement algorithm is constructed, and the initial enhancement parameters are optimized in real time based on a deep learning model to obtain the final adjusted enhancement parameters.

[0012] Furthermore, the deep learning model is a generative adversarial network (GAN). The generator of the GAN performs preliminary enhancement on the preprocessed image and the initial enhancement parameters to obtain a preliminary enhanced image. The quality of the pre-enhanced image is evaluated using a discriminator, and a quality score is output. Based on the difference between the quality score and the preset target score, the enhancement parameters are iteratively optimized using the gradient descent algorithm until the quality score meets the preset error threshold, thus obtaining the final adjusted enhancement parameters.

[0013] Furthermore, in step S4, the preprocessed image data is enhanced using the adjusted enhancement parameters to generate an enhanced image, specifically including: Based on the adjusted enhancement parameters, the preprocessed image is enhanced in steps, and the enhancement steps include at least one or more of contrast adjustment, brightness adjustment, and detail enhancement. The enhanced image pixel values ​​are mapped back to a preset pixel value range to obtain the final enhanced image.

[0014] Furthermore, step S5 involves a quality assessment of the enhanced image, specifically including: Construct an image quality assessment index system, wherein the index system includes at least one or more of the following: peak signal-to-noise ratio, structural similarity index, and information entropy; Calculate the values ​​of various quality assessment metrics for the enhanced image and the original sharp reference image; Set preset thresholds for each quality assessment indicator. If the enhanced image meets all the threshold conditions at the same time, then output the enhanced image. If any threshold is not met, return to step S3, adjust the relevant parameters, and iterate to optimize the enhancement parameters again until the enhanced image meets all threshold conditions.

[0015] An image enhancement system for intelligent connected vehicle testing scenarios, applied to perform the aforementioned image enhancement method for intelligent connected vehicle testing scenarios, the system comprising: The image acquisition and preprocessing module is used to acquire raw image data in the intelligent connected vehicle test scenario and preprocess the raw image data. The scene feature recognition module is used to build a scene feature recognition model, extract dynamic scene features from preprocessed image data, and determine the environmental category parameters of the current test scene. The enhancement parameter adjustment module is used to call the corresponding scene adaptive enhancement strategy library based on the environment category parameters, and adjust the image enhancement parameters in real time through dynamic image enhancement algorithms; The image enhancement module is used to enhance the preprocessed image data using adjusted enhancement parameters to generate an enhanced image. The quality assessment and iterative optimization module is used to assess the quality of the enhanced image. If the assessment result meets the preset threshold, the enhanced image is output; otherwise, the enhancement parameter adjustment module is controlled to iteratively optimize the enhancement parameters until the threshold is met.

[0016] The beneficial effects of this invention are as follows: 1. By constructing a scene recognition model that integrates spatial and temporal features, the environmental category parameters of the current test scene can be identified in real time, enabling accurate classification of various complex environments and providing a basis for subsequent enhancements.

[0017] 2. A dynamic image enhancement algorithm based on generative adversarial networks is introduced. Initial parameters are automatically matched from the enhancement strategy library according to the scene category, and the algorithm is optimized through quality feedback loop to achieve intelligent adjustment of multiple dimensions such as contrast, brightness, and detail, resulting in a more natural and smooth enhancement effect.

[0018] 3. This invention establishes a multi-index image quality assessment system, including peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and information entropy. After each enhancement, a quantitative evaluation is performed, and the enhancement parameters are automatically adjusted based on this feedback signal, forming a closed-loop iterative optimization process to ensure that the output image meets the quality threshold requirements.

[0019] 4. By combining a pre-defined strategy library with a deep model, the system can maintain temporal consistency of enhancement parameters in consecutive video frames, avoid image flickering, and significantly improve the continuity and real-time processing capabilities of video-level enhancement, making it suitable for dynamic driving scenarios. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the overall method of the present invention; Figure 2 This is a flowchart of the preprocessing module of the present invention; Figure 3 This is a flowchart illustrating the enhanced parameter optimization and quality assessment process of the present invention. Figure 4 This is a system block diagram of the present invention. Detailed Implementation

[0021] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0022] Example 1: Please see Figure 1-3 This invention provides a technical solution: an image enhancement method for intelligent connected vehicle testing scenarios, the method comprising: S1. Acquire raw image data in the intelligent connected vehicle test scenario and preprocess the raw image data; Among them, the intelligent connected vehicle test scenario refers to a set of specific environments and situations set up to test the various performance and functions of intelligent connected vehicles. It covers a variety of complex factors such as road conditions, traffic flow, and weather conditions, and is used to simulate various situations that intelligent connected vehicles may encounter in actual operation. The raw image data is image information directly obtained from the image acquisition device without any processing or modification. This data contains the original visual content under the intelligent connected vehicle test scenario. Preprocessing is a series of preliminary processing operations performed on the raw image data. The purpose is to improve image quality, remove noise, adjust image size or format, etc., so as to better extract and analyze features in the future. S2. Construct a scene feature recognition model, extract dynamic scene features from the preprocessed image data, and determine the environmental category parameters of the current test scene; The scene feature recognition model is a specially designed machine learning or deep learning model that analyzes and processes input image data to identify and extract information that represents the characteristics of different test scenes. Through training, the model can learn unique patterns and features in various scenes, thereby accurately determining the scene type to which the image belongs. Scene dynamic feature extraction uses the scene feature recognition model to mine dynamic change features related to the current test scene from preprocessed image data. These features may include the motion state of objects in the scene, changes in lighting, weather-related features, etc., which can reflect the dynamic characteristics of the test scene in real time. The environment category parameter determines the environment category to which the current test scene belongs based on the results of scene dynamic feature extraction and represents it with corresponding parameters. S3. Based on the environment category parameters, call the corresponding scene adaptive enhancement strategy library and adjust the image enhancement parameters in real time through dynamic image enhancement algorithms; The scene-adaptive enhancement strategy library is a pre-built collection containing image enhancement strategies designed for different environment categories. These strategies, based on the characteristics and requirements of different environments, define corresponding image enhancement methods and parameter adjustment rules to achieve optimal image enhancement results in various test scenarios. The dynamic image enhancement algorithm is an algorithm that can automatically adjust image enhancement parameters based on real-time input environment category parameters. It can flexibly change the image enhancement method according to different scene requirements, such as adjusting parameters like contrast, brightness, and sharpness, to adapt to image quality requirements in different environments. Real-time adjustment of image enhancement parameters, during intelligent connected vehicle testing, uses the dynamic image enhancement algorithm to immediately modify and optimize the parameters required for image enhancement based on the current test scenario's environment category parameters, ensuring that the image enhancement effect can match the changes in the current scene in real time. S4. Use the adjusted enhancement parameters to enhance the preprocessed image data and generate the enhanced image. Among them, enhancement processing involves performing a series of operations on the preprocessed image data using adjusted enhancement parameters to improve the visual quality of the image, highlight important information in the image, and improve the image's clarity, contrast, and other indicators, making the image more suitable for subsequent analysis or application needs. The enhanced image is the image obtained after enhancement processing, which has a significant improvement in visual quality compared to the original image or the preprocessed image, and can better display the details and information in the intelligent connected vehicle test scenario. S5. Evaluate the quality of the enhanced image. If the evaluation result meets the preset threshold, output the enhanced image; otherwise, iteratively optimize the enhancement parameters until the threshold is met. The quality assessment process involves comprehensively evaluating the enhanced image using specific evaluation metrics and methods to determine whether its quality meets the expected requirements. Evaluation metrics may include image sharpness, contrast, signal-to-noise ratio, structural similarity, etc. By comprehensively considering these metrics, it is determined whether the enhanced image meets the needs of the actual application. The preset threshold is a pre-set quality standard value or range used as the basis for judging whether the enhanced image is qualified. Only when the quality assessment result of the enhanced image reaches or exceeds this preset threshold is the image considered to meet the requirements. The iterative optimization of enhancement parameters involves repeatedly adjusting and modifying the image enhancement parameters based on the evaluation feedback information, and then performing enhancement processing and quality assessment again. This process is repeated until the quality of the enhanced image meets the preset threshold. This process aims to find the most suitable image enhancement scheme for the current test scenario by continuously optimizing the parameters.

[0023] It should be noted that, during use, the original image data is preprocessed to remove noise and adjust the format, providing a good foundation for subsequent processing, improving overall processing efficiency and accuracy. A scene feature recognition model is built to extract dynamic features and determine environmental category parameters, accurately grasping the characteristics of different test scenarios, making image enhancement more targeted. Based on the environmental category parameters, an adaptive enhancement strategy library is called and parameters are adjusted in real time to adapt to complex and ever-changing test environments, ensuring that the enhancement effect is always optimal. The enhanced image is quality evaluated and iteratively optimized to ensure that the output image quality meets the preset standards, providing high-quality image data for the test analysis of intelligent connected vehicles, which helps to more accurately evaluate vehicle performance and improve the reliability and effectiveness of testing.

[0024] In one embodiment, raw image data from a smart connected vehicle test scenario is acquired, and the raw image data is preprocessed, including: The image acquisition equipment on the intelligent connected vehicle collects multiple frames of raw image data in the test scenario to form an image sequence. ,in Indicates the number of image frames, each image , , These represent the height and width of the image, respectively, and 3 represents the RGB three channels; Preprocessing of the raw image data includes noise removal and format normalization: Noise Removal: A Gaussian filtering algorithm is used to reduce noise in each frame of the original image. The filter kernel size is adaptively set according to the noise intensity of the image, and the formula is as follows: ,in The standard deviation is expressed as Gaussian kernel, This represents the convolution operation; Format standardization: Normalizes the pixel values ​​of the denoised image to... The standardized formula for the interval is: Obtain the preprocessed image sequence .

[0025] This design uses image acquisition devices mounted on intelligent connected vehicles to acquire multiple frames of original images to form a sequence. Gaussian filtering is used for adaptive noise reduction, and pixel values ​​are normalized to the [0,1] range. The multi-frame image sequence can provide richer scene information. The Gaussian filtering adaptively sets the filter kernel size to effectively reduce noise according to different image noise intensities, avoiding over-smoothing or insufficient noise reduction. Format standardization unifies pixel values ​​to a fixed range, eliminating differences in numerical ranges between different images, making subsequent processing more stable and consistent. This lays a good data foundation for subsequent scene feature extraction and image enhancement operations, improving the overall processing effect and efficiency.

[0026] In one embodiment, a scene feature recognition model is constructed to extract dynamic scene features from preprocessed image data and determine the environmental category parameters of the current test scene, including: A scene feature recognition model based on convolutional neural networks (CNN) and long short-term memory networks (LSTM) is constructed. The input to the model is a preprocessed image sequence. The output is the environment category parameter. ,in Indicates weather condition parameters, Indicates road condition parameters, Indicates traffic environment parameters, Indicates the light intensity parameter; The spatial features of each preprocessed image frame are extracted using a CNN, with the following formula: in, ; The spatial feature sequence is obtained: The spatial feature sequence is input into the LSTM network to capture the dynamic temporal features of the scene, as shown in the formula: Dynamic temporal features are classified using a fully connected layer, and environmental category parameters are output. The formula is ,in This indicates a fully connected layer.

[0027] This design constructs a model based on CNN and LSTM. CNN extracts spatial features of each frame of the image, while LSTM captures dynamic temporal features. The environment category parameters are output through a fully connected layer. CNN is good at processing spatial information of images and can accurately extract the spatial features of each frame of the image. LSTM can effectively process sequential data and capture the dynamic temporal features of the scene changing over time. The combination of the two can comprehensively and accurately extract the dynamic features of the scene and determine the environment category parameters, enabling the model to have a more accurate recognition ability for different test scenarios. This provides a reliable basis for subsequently calling appropriate image enhancement strategies and improves the targeting of image enhancement.

[0028] In one embodiment, based on the environment category parameter, the corresponding scene adaptive enhancement strategy library is invoked, and the image enhancement parameters are adjusted in real time through a dynamic image enhancement algorithm, including: A scene-adaptive enhancement strategy library is constructed, containing enhancement strategies corresponding to different environment category parameters. Each strategy has a preset initial set of enhancement parameters, including contrast adjustment coefficients. Brightness adjustment coefficient Detail enhancement strength coefficient ; Based on environmental category parameters Match the corresponding initial enhancement strategy from the strategy library and obtain the initial enhancement parameters. ; A dynamic image enhancement algorithm is constructed, and the initial enhancement parameters are optimized in real time based on a deep learning model: With the preprocessed image Using the initial enhancement parameters as input, the image is initially enhanced by the generator of a Generative Adversarial Network (GAN), resulting in a preliminary enhanced image: The discriminator is used to evaluate the quality of the initially enhanced image and output a quality score. Based on quality rating Compared with the preset target score The difference is used to iteratively optimize the enhancement parameters using the gradient descent algorithm. The optimization formula is: in Indicates the number of iterations. Indicates the learning rate. Indicates the parameter The gradient is applied until the quality score is satisfied. , To obtain the final adjusted enhancement parameters, a preset error threshold is set. .

[0029] This design involves building a strategy library to match initial enhancement strategies, using the GAN generator to initially enhance the image, having the discriminator evaluate the quality, and then iteratively optimizing the enhancement parameters through gradient descent. The strategy library contains enhancement strategies corresponding to various environments, quickly matching initial parameters and providing a foundation for image enhancement. The GAN generator can initially enhance the image, the discriminator evaluates the quality and provides feedback, and the parameters can be iteratively optimized through gradient descent. This allows for real-time adjustments based on the actual image conditions, making the enhancement parameters more suitable for the current image requirements, achieving dynamic and accurate image enhancement, improving enhancement effects, and adapting to complex and ever-changing test scenarios.

[0030] In one embodiment, enhancing the preprocessed image data using adjusted enhancement parameters to generate an enhanced image includes: Based on the adjusted enhancement parameters The preprocessed image is processed in steps. Enhancement: Contrast adjustment: Image contrast is adjusted through linear transformation, using the following formula: in This represents the average pixel value of the preprocessed image; Brightness Adjustment: Adjust the brightness based on the image after contrast adjustment. The formula is: in The function is used to limit pixel values ​​to a certain range. interval; Detail Enhancement: A guided filtering algorithm is used to enhance the details of the brightness-adjusted image. The formula is as follows: in This represents the window radius of the guided filter. This represents the regularization parameter for the guided filter; Map the enhanced image pixel values ​​back The interval, the formula is: The final enhanced image is obtained. .

[0031] This design, based on adjusted parameters, processes the image in three steps: contrast, brightness, and detail enhancement. Finally, the pixel values ​​are mapped back to the [0, 255] range. The step-by-step enhancement logic is clear. Contrast adjustment uses linear transformation to highlight the differences between light and dark areas in the image. Brightness adjustment reasonably changes the brightness based on contrast to avoid being too bright or too dark. Detail enhancement uses a guided filtering algorithm, which can effectively improve the clarity of image details. Finally, the pixel values ​​are mapped back to the commonly used range for convenient subsequent display and storage. This design can systematically and comprehensively improve image quality, making the enhanced image clearer and more realistic in visual effect, meeting the image quality requirements of intelligent connected vehicle testing.

[0032] In one embodiment, the enhanced image is quality evaluated. If the evaluation result meets a preset threshold, the enhanced image is output; otherwise, the enhancement parameters are iteratively optimized until the threshold is met, including: An image quality assessment index system was constructed, which includes peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and information entropy (Entropy). Computational image enhancement Distortion-free images acquired under standard test conditions, compared to the original sharp reference image. The formula for PSNR is: in Indicates mean square error; The SSIM of the enhanced image and the original sharp reference image is calculated using the following formula: in , These represent the pixel mean values ​​of the enhanced image and the reference image, respectively. , These represent the pixel standard deviations of the two, respectively. Indicates the covariance of the two. , This is a preset constant; The formula for calculating the information entropy of an enhanced image is: in Indicates the pixel value in the enhanced image The probability of; Set quality assessment thresholds: , , , in, , , They are respectively , The preset threshold for information entropy; If the enhanced image meets all three threshold conditions, then the enhanced image is output. If any threshold is not met, return to step S3 and adjust the learning rate. The enhancement parameters are then iterated and optimized again until the enhanced image meets all threshold conditions.

[0033] This design constructs an evaluation index system that includes PSNR, SSIM, and information entropy. It calculates various indicators for the enhanced image and the reference image, sets thresholds to determine whether to output, and adjusts the learning rate to re-optimize the parameters if the thresholds are not met. The multi-index evaluation system can comprehensively measure image quality from different perspectives. PSNR reflects the degree of image distortion, SSIM measures structural similarity, and information entropy reflects the richness of image information. Setting thresholds can strictly control the quality of the output image, ensuring that only high-quality images are output. If the thresholds are not met, adjusting the learning rate to re-optimize the parameters can avoid getting stuck in local optima, continuously approach the best enhancement effect, and ensure that the quality of the final output image is reliable, meeting the stringent requirements of intelligent connected vehicle testing.

[0034] Example 2, please refer to Figure 4 An image enhancement system for intelligent connected vehicle testing scenarios, applied to the image enhancement method described above for intelligent connected vehicle testing scenarios, the system comprising: The image acquisition and preprocessing module is used to acquire raw image data in intelligent connected vehicle test scenarios and preprocess the raw image data. The scene feature recognition module is used to build a scene feature recognition model, extract dynamic scene features from preprocessed image data, and determine the environmental category parameters of the current test scene. The enhancement parameter adjustment module is used to call the corresponding scene adaptive enhancement strategy library based on the environment category parameters, and adjust the image enhancement parameters in real time through dynamic image enhancement algorithms; The image enhancement module is used to enhance the preprocessed image data using adjusted enhancement parameters to generate an enhanced image. The quality assessment and iterative optimization module is used to assess the quality of the enhanced image. If the assessment result meets the preset threshold, the enhanced image is output; otherwise, the enhancement parameter adjustment module is controlled to iteratively optimize the enhancement parameters until the threshold is met.

[0035] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can 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.

[0036] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An image enhancement method of an intelligent networked vehicle test scene, characterized in that, The method includes the following steps: S1. Obtain raw image data in the intelligent connected vehicle test scenario and preprocess the raw image data; S2. Construct a scene feature recognition model, extract dynamic scene features from the preprocessed image data, and determine the environmental category parameters of the current test scene; S3. Based on the environment category parameters, call the corresponding scene adaptive enhancement strategy library and adjust the image enhancement parameters in real time through dynamic image enhancement algorithm; S4. Use the adjusted enhancement parameters to enhance the preprocessed image data and generate the enhanced image. S5. Evaluate the quality of the enhanced image. If the evaluation result meets the preset threshold, output the enhanced image; otherwise, iteratively optimize the enhancement parameters until the threshold is met.

2. The image enhancement method for intelligent connected vehicle test scenarios according to claim 1, characterized in that, Step S1 involves acquiring raw image data from a smart connected vehicle test scenario and preprocessing the raw image data, specifically including: The image acquisition equipment on the intelligent connected vehicle is used to collect multiple frames of raw image data in the test scenario to form an image sequence. The original image data is preprocessed, which includes at least noise removal and format normalization. Noise removal uses a filtering algorithm to reduce noise in each frame of the original image, and format normalization normalizes the pixel values ​​of the denoised image to a preset range, resulting in a preprocessed image sequence.

3. The image enhancement method for intelligent connected vehicle test scenarios according to claim 2, characterized in that: The filtering algorithm is a Gaussian filtering algorithm, and the size of the filtering kernel is adaptively set according to the image noise intensity. The preset numerical range is [0,1]. Pixel value normalization is achieved by calculating the difference between the pixel value of the image after noise reduction and the minimum pixel value of the image, and then dividing it by the difference between the maximum and minimum pixel values ​​of the image.

4. The image enhancement method for intelligent connected vehicle test scenarios according to claim 1, characterized in that, Step S2 involves constructing a scene feature recognition model, extracting dynamic scene features from the preprocessed image data, and determining the environmental category parameters of the current test scene. Specifically, this includes: A scene feature recognition model based on spatial feature extraction network and temporal feature extraction network is constructed. The input of the model is a preprocessed image sequence, and the output is environmental category parameters. The environmental category parameters include at least one or more of the following: weather condition parameters, road condition parameters, traffic environment parameters, and light intensity parameters. Spatial features of each preprocessed image frame are extracted using a spatial feature extraction network to obtain a spatial feature sequence; Spatial feature sequences are input into a temporal feature extraction network to capture the dynamic temporal features of the scene; The dynamic temporal features are classified using a classification network, and environmental category parameters are output.

5. The image enhancement method for intelligent connected vehicle test scenarios according to claim 4, characterized in that, The spatial feature extraction network is a convolutional neural network, the temporal feature extraction network is a long short-term memory network, and the classification network is a fully connected layer.

6. The image enhancement method for intelligent connected vehicle test scenarios according to claim 1, characterized in that, In step S3, based on the environment category parameters, the corresponding scene adaptive enhancement strategy library is invoked, and the image enhancement parameters are adjusted in real time through a dynamic image enhancement algorithm. Specifically, this includes: Construct a scene-adaptive enhancement strategy library. The strategy library contains enhancement strategies corresponding to different environment category parameters. Each strategy has a preset initial set of enhancement parameters. The enhancement parameters include at least one or more of the following: contrast adjustment coefficient, brightness adjustment coefficient, and detail enhancement intensity coefficient. Based on the environment category parameter, match the corresponding initial enhancement strategy from the strategy library and obtain the initial enhancement parameters; A dynamic image enhancement algorithm is constructed, and the initial enhancement parameters are optimized in real time based on a deep learning model to obtain the final adjusted enhancement parameters.

7. The image enhancement method for intelligent connected vehicle test scenarios according to claim 6, characterized in that: The deep learning model is a generative adversarial network (GAN). The generator of the GAN performs preliminary enhancement on the preprocessed image and the initial enhancement parameters to obtain a preliminary enhanced image. The quality of the pre-enhanced image is evaluated using a discriminator, and a quality score is output. Based on the difference between the quality score and the preset target score, the enhancement parameters are iteratively optimized using the gradient descent algorithm until the quality score meets the preset error threshold, thus obtaining the final adjusted enhancement parameters.

8. The image enhancement method for intelligent connected vehicle test scenarios according to claim 1, characterized in that, Step S4 involves using the adjusted enhancement parameters to enhance the preprocessed image data, generating an enhanced image. Specifically, this includes: Based on the adjusted enhancement parameters, the preprocessed image is enhanced in steps, and the enhancement steps include at least one or more of contrast adjustment, brightness adjustment, and detail enhancement. The enhanced image pixel values ​​are mapped back to a preset pixel value range to obtain the final enhanced image.

9. The image enhancement method for intelligent connected vehicle test scenarios according to claim 1, characterized in that, Step S5 involves a quality assessment of the enhanced image, specifically including: Construct an image quality assessment index system, wherein the index system includes at least one or more of the following: peak signal-to-noise ratio, structural similarity index, and information entropy; Calculate the values ​​of various quality assessment metrics for the enhanced image and the original sharp reference image; Set preset thresholds for each quality assessment indicator. If the enhanced image meets all the threshold conditions at the same time, then output the enhanced image. If any threshold is not met, return to step S3, adjust the relevant parameters, and iterate to optimize the enhancement parameters again until the enhanced image meets all threshold conditions.

10. An image enhancement system for intelligent connected vehicle testing scenarios, characterized in that, An image enhancement method applied to a test scenario of intelligent connected vehicles according to any one of claims 1-9, the system comprising: The image acquisition and preprocessing module is used to acquire raw image data in the intelligent connected vehicle test scenario and preprocess the raw image data. The scene feature recognition module is used to build a scene feature recognition model, extract dynamic scene features from preprocessed image data, and determine the environmental category parameters of the current test scene. The enhancement parameter adjustment module is used to call the corresponding scene adaptive enhancement strategy library based on the environment category parameters, and adjust the image enhancement parameters in real time through dynamic image enhancement algorithms; The image enhancement module is used to enhance the preprocessed image data using adjusted enhancement parameters to generate an enhanced image. The quality assessment and iterative optimization module is used to assess the quality of the enhanced image. If the assessment result meets the preset threshold, the enhanced image is output; otherwise, the enhancement parameter adjustment module is controlled to iteratively optimize the enhancement parameters until the threshold is met.

Images