An image enhancement-based license plate intelligent recognition method and system

Abstract

The application provides a license plate intelligent recognition method and system based on image enhancement, and belongs to the technical field of image processing. The method performs Fourier transform on license plate image data to determine the spectral distribution characteristics; then, a gray level histogram is constructed based on the characteristics, and when the gray level variance exceeds the preset variance threshold, it is determined that the license plate is affected by light changes and the dynamic characteristic index of the gray level distribution is determined; then, the index is analyzed using a convolutional neural network to generate modulation parameters of the contrast enhancement factor and the sharpening intensity; when the deviation between the spectral peak position and the preset peak value exceeds the preset deviation threshold, it is determined that the license plate is affected by motion blur, and an enhanced spectral representation is constructed; finally, the enhanced license plate image data is generated through inverse Fourier transform. The quality of the license plate image in a complex environment is significantly improved, the characters are clear and identifiable, and the license plate recognition accuracy and system robustness are effectively improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, specifically to a method and system for intelligent license plate recognition based on image enhancement. Background Technology

[0002] Currently, license plate recognition technology, as a core component of intelligent transportation systems, plays an irreplaceable role in key application scenarios such as urban traffic management, security monitoring, and automated toll collection. With the rapid increase in traffic flow and the increasing complexity of monitoring needs, accurate and rapid license plate recognition has become a crucial technological support for the construction of modern smart cities. However, current license plate recognition methods generally suffer from insufficient adaptability when handling complex environmental conditions. Traditional recognition algorithms often employ fixed image processing parameters and uniform enhancement strategies, making it difficult to cope with the diverse needs of real-world applications, such as changes in lighting conditions and differences in vehicle motion states. This lack of environmental adaptability leads to unstable performance of the recognition system under different shooting conditions, making it impossible to maintain consistently reliable recognition accuracy.

[0003] The information provided in the background section of this application is only for enhancing the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] In view of this, this application provides a license plate intelligent recognition method based on image enhancement, which can improve the recognition accuracy of license plate images.

[0005] In a first aspect, embodiments of this application provide a license plate intelligent recognition method based on image enhancement. The method includes: acquiring license plate image data and performing Fourier transform processing on the license plate image data to determine spectral distribution characteristics; constructing a grayscale histogram based on the spectral distribution characteristics; determining whether the grayscale variance of the grayscale histogram exceeds a preset variance threshold; if it exceeds the preset variance threshold, determining that the license plate image data is affected by illumination changes, and determining dynamic characteristic indicators of the grayscale distribution based on the current grayscale histogram; and using a convolutional neural network to determine the license plate image based on the dynamic characteristic indicators. The modulation parameters of the license plate image data include a contrast enhancement factor and a sharpening intensity. The system determines whether the deviation between the peak position of the spectral distribution feature and a preset peak position exceeds a preset deviation threshold. If the deviation exceeds the preset deviation threshold, it determines that the license plate image data has motion blur and determines an enhanced spectral representation based on the license plate image data. An inverse Fourier transform is used to determine enhanced license plate image data based on the enhanced spectral representation. The system then determines whether the grayscale variance of the enhanced license plate image data exceeds a preset variance threshold. If the variance exceeds the preset variance threshold, the modulation parameters are updated iteratively.

[0006] Secondly, embodiments of this application provide a license plate intelligent recognition system based on image enhancement. This system includes: an acquisition module, a construction module, a first judgment module, a first determination module, a second judgment module, a second determination module, and a third judgment module. The acquisition module acquires license plate image data and performs Fourier transform processing on the license plate image data to determine spectral distribution characteristics. The construction module constructs a grayscale histogram based on the spectral distribution characteristics. The first judgment module determines whether the grayscale variance of the grayscale histogram exceeds a preset variance threshold. If the determination exceeds the preset variance threshold, it determines that the license plate image data is affected by illumination changes and determines dynamic characteristic indicators of the grayscale distribution based on the current grayscale histogram. The first determination module uses a convolutional neural network to determine the modulation parameters of the license plate image data based on the dynamic characteristic indicators. The modulation parameters include a contrast enhancement factor and a sharpening intensity. The second judgment module determines whether the deviation between the peak position of the spectral distribution characteristics and a preset peak position exceeds a preset deviation threshold. If the determination exceeds the preset deviation threshold, it determines that the license plate image data is affected by motion blur and determines an enhanced spectral representation based on the license plate image data. The second determining module is used to determine the enhanced license plate image data based on the enhanced spectrum representation using inverse Fourier transform; the third judging module is used to judge whether the gray-level variance of the enhanced license plate image data exceeds a preset variance threshold. If the judgment exceeds the preset variance threshold, the iterative modulation parameters are updated.

[0007] This application provides a method and system for intelligent license plate recognition based on image enhancement. By performing Fourier transform processing on the acquired license plate image data, the spectral distribution characteristics are clarified, providing a precise basis for subsequent image enhancement modulation. A grayscale histogram is constructed based on these spectral distribution characteristics, which can intuitively reflect the grayscale distribution state of the image. Then, by determining whether the grayscale variance of the grayscale histogram exceeds a preset variance threshold, if it does, it is determined that the license plate image data is affected by changes in illumination. Simultaneously, dynamic characteristic indicators of the grayscale distribution are determined based on the current grayscale histogram. A convolutional neural network is used to perform deep analysis and feature extraction on the aforementioned dynamic characteristic indicators, enabling accurate calculation of modulation parameters adapted to the current environment. These parameters include a contrast enhancement factor for optimizing image brightness differences and a sharpening intensity for improving character edge clarity, thus achieving… The system employs intelligent adaptive determination of modulation parameters. It determines whether the deviation between the peak position of the spectral distribution characteristics and a preset peak position exceeds a preset deviation threshold. If the deviation exceeds the threshold, motion blur is detected in the license plate image data. This allows for targeted determination of the enhanced spectral representation, laying the foundation for eliminating motion blur and restoring image details. An inverse Fourier transform is used to convert the enhanced spectral representation into enhanced license plate image data, accurately mapping the enhancement effect in the frequency domain to the spatial domain, effectively restoring character edge information and improving the overall quality of the license plate image. By determining whether the grayscale variance of the enhanced license plate image data exceeds a preset variance threshold, the modulation parameters are updated and iteratively adjusted to ensure the grayscale distribution of the license plate image remains optimal, continuously optimizing the enhancement effect. This significantly improves the image quality of license plate images in complex environments, ensuring clear and legible character information, increasing the accuracy of license plate recognition and the robustness of the system. It can stably adapt to diverse practical application scenarios, including varying lighting conditions and vehicle motion states. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating an exemplary embodiment of the intelligent license plate recognition method based on image enhancement provided in this application.

[0010] Figure 2 This is a flowchart illustrating another exemplary embodiment of the intelligent license plate recognition method based on image enhancement provided in this application.

[0011] Figure 3This is a flowchart illustrating an image-enhanced intelligent license plate recognition method provided in another exemplary embodiment of this application.

[0012] Figure 4 This is a flowchart illustrating an image-enhanced intelligent license plate recognition method provided in another exemplary embodiment of this application.

[0013] Figure 5 This is a flowchart illustrating an image-enhanced intelligent license plate recognition method provided in another exemplary embodiment of this application.

[0014] Figure 6 This is a flowchart illustrating an image-enhanced intelligent license plate recognition method provided in another exemplary embodiment of this application.

[0015] Figure 7 This is a flowchart illustrating an image-enhanced intelligent license plate recognition method provided in another exemplary embodiment of this application. Detailed Implementation

[0016] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this application will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this application.

[0017] The terms “a,” “one,” and “the” are used to indicate the existence of one or more elements / components / etc.; the terms “including” and “having” are used to indicate an open-ended inclusion and to mean that there may be other elements / components / etc. in addition to the listed elements / components / etc. The terms “first” and “second” are used only as markers and are not a limitation on the number of objects.

[0018] Currently, license plate recognition technology, as a core component of intelligent transportation systems, plays an irreplaceable role in key application scenarios such as urban traffic management, security monitoring, and automated toll collection. With the rapid increase in traffic flow and the increasing complexity of monitoring needs, accurate and rapid license plate recognition has become a crucial technological support for the construction of modern smart cities. However, current license plate recognition methods generally suffer from insufficient adaptability when handling complex environmental conditions. Traditional recognition algorithms often employ fixed image processing parameters and uniform enhancement strategies, making it difficult to cope with the diverse needs of real-world applications, such as changes in lighting conditions and differences in vehicle motion states. This lack of environmental adaptability leads to unstable performance of the recognition system under different shooting conditions, making it impossible to maintain consistently reliable recognition accuracy.

[0019] For example, license plate characters exhibit specific spectral distribution patterns in the frequency domain, but existing methods lack specialized modulation processing mechanisms for these spectral features, failing to fully utilize frequency domain information to guide the image enhancement process. Furthermore, the grayscale distribution characteristics of the license plate region exhibit significant dynamism with changes in the shooting environment, requiring modulation parameters to have real-time adaptive adjustment capabilities. In tunnel entrance / exit scenes, license plate images are simultaneously affected by abrupt changes in illumination and vehicle motion blur, leading to severe degradation of character edge information. Traditional static parameter modulation methods are completely incapable of handling this complex image degradation pattern.

[0020] Therefore, how to construct an image enhancement mechanism that can perform specialized modulation processing based on the spectral characteristics of license plate images and dynamically and adaptively adjust parameters based on grayscale distribution characteristics has become a technical problem that needs to be solved to improve the environmental adaptability of intelligent license plate recognition systems.

[0021] This application provides a method for intelligent license plate recognition based on image enhancement, such as... Figure 1 The illustrated method is an image enhancement-based intelligent license plate recognition method. This method may include the following steps: Step S110: Acquire license plate image data and perform Fourier transform processing on the license plate image data to determine the spectral distribution characteristics; Step S120: Construct a grayscale histogram based on the spectral distribution characteristics; Step S130: Determine whether the grayscale variance of the grayscale histogram exceeds the preset variance threshold. If it is determined that the grayscale variance exceeds the preset variance threshold, it is determined that the license plate image data is affected by illumination changes, and the dynamic characteristic index of grayscale distribution is determined based on the current grayscale histogram. Step S140: Based on dynamic feature indicators, a convolutional neural network is used to determine the modulation parameters of the license plate image data. The modulation parameters include contrast enhancement factor and sharpening intensity. Step S150: Determine whether the deviation between the peak position of the spectral distribution feature and the preset peak position exceeds the preset deviation threshold. If it is determined that the deviation exceeds the preset deviation threshold, then it is determined that the license plate image data has motion blur effects, and the enhanced spectral representation is determined based on the license plate image data. Step S160: Use inverse Fourier transform to determine the enhanced license plate image data based on the enhanced spectrum representation; Step S170: Determine whether the grayscale variance of the enhanced license plate image data exceeds a preset variance threshold. If it exceeds the preset variance threshold, update the iterative modulation parameters.

[0022] According to the image enhancement-based intelligent license plate recognition method provided in this application, this method can clarify the spectral distribution characteristics by performing Fourier transform processing on the acquired license plate image data, providing a precise basis for subsequent image enhancement modulation; constructing a gray-level histogram based on the spectral distribution characteristics can intuitively reflect the gray-level distribution state of the image; then, by judging whether the gray-level variance of the gray-level histogram exceeds a preset variance threshold, if it exceeds the preset variance threshold, it is determined that the license plate image data is affected by changes in illumination. At the same time, the dynamic characteristic index of gray-level distribution is determined based on the current gray-level histogram; using a convolutional neural network to perform deep analysis and feature extraction on the above dynamic characteristic index, it can accurately calculate the modulation parameters adapted to the current environment, including a contrast enhancement factor for optimizing the difference between image brightness and darkness and a factor for improving character edge clarity. The system intelligently and adaptively determines modulation parameters by adjusting the sharpening intensity. It checks if the deviation between the peak position of the spectral distribution characteristics and a preset peak position exceeds a preset deviation threshold. If the deviation exceeds the threshold, motion blur is detected in the license plate image data, leading to a targeted enhancement of the spectral representation. This lays the foundation for eliminating motion blur and restoring image details. An inverse Fourier transform is used to convert the enhanced spectral representation into enhanced license plate image data, accurately mapping the enhancement effect in the frequency domain to the spatial domain. This effectively restores character edge information, optimizes image contrast, and improves the overall quality of the license plate image. By checking if the grayscale variance of the enhanced license plate image data exceeds a preset variance threshold, the modulation parameters are updated and iteratively adjusted to ensure the grayscale distribution of the license plate image remains optimal, continuously optimizing the enhancement effect. This significantly improves the image quality of license plate images in complex environments, ensuring clear and legible character information, increasing the accuracy of license plate recognition and the robustness of the system. It can stably adapt to diverse practical application scenarios, including varying lighting conditions and vehicle motion states.

[0023] The steps of the image enhancement-based intelligent license plate recognition method provided in this application are described in detail below: In one embodiment of this application, step S110 involves acquiring license plate image data and performing Fourier transform processing on the license plate image data to determine its spectral distribution characteristics. Specifically, original license plate image data containing the license plate area can be captured using image acquisition terminals such as traffic monitoring cameras or toll station cameras. The image format can support common RGB or grayscale image formats, and the pixel size is determined according to the acquisition device parameters, such as the common 256×128 pixel or 512×512 pixel specifications. Performing Fourier transform processing on the acquired license plate image data essentially transforms the image from the spatial domain (i.e., the spatial distribution of pixels) to the frequency domain (i.e., the frequency distribution of the image signal), thereby extracting the frequency characteristics of the image. For example, if the acquired license plate image is a 256×128 pixel grayscale image, processing the image using a two-dimensional fast Fourier transform algorithm yields a spectral distribution matrix. After dividing the frequency range, it was found that the peak amplitude of the low-frequency part (0-0.2) was 1500, concentrated in a 3×3 neighborhood of the center of the spectrum; the peak amplitude of the high-frequency part (0.6-1.0) was 450, distributed at 8 coordinate points at the four corners of the spectrum. The mean of the spectrum distribution was calculated to be 820 and the standard deviation to be 400, determining the effective frequency range to be 15-1250Hz. The peak position coordinates of the high-frequency part were recorded as (200,100), (200,28), (56,100), (56,28), etc., with the corresponding amplitude values ​​all being 450. These data together constitute the spectrum distribution characteristics of the license plate image.

[0024] In one embodiment of this application, step S120, constructing a grayscale histogram based on spectral distribution characteristics, further includes the following steps: Figure 2 As shown, the specific content is as follows: Step S210: Determine the spectral amplitude spectrum based on the spectral distribution characteristics; Step S220: Construct a grayscale histogram based on the spectral amplitude spectrum.

[0025] Specifically, the spectral distribution feature is a set of frequency domain data obtained by performing a Fourier transform on the license plate image data. This includes a complex spectral distribution matrix corresponding to the frequency domain coordinates. This matrix reflects both the distribution location of each frequency component and the amplitude and phase information of the frequency signal. The spectral amplitude spectrum is the core quantitative representation of the spectral distribution feature, used to extract the intensity information of each frequency component individually (eliminating phase influence), providing standardized data for the subsequent construction of the grayscale histogram. The grayscale histogram is a statistical chart used to visually present the distribution pattern of image grayscale values. Its horizontal axis represents the grayscale level (usually set to 0-255 levels, covering all brightness ranges of image pixels), and the vertical axis represents the percentage or absolute number of pixels at the corresponding grayscale level. The core logic of constructing a grayscale histogram based on the spectral amplitude spectrum is to normalize all amplitude data in the spectrum, mapping them to a grayscale level range of 0-255, and then count the number of amplitude data (i.e., the number of pixels) corresponding to each grayscale level, ultimately forming a complete grayscale distribution statistical chart.

[0026] For example, the spectral distribution matrix F(u,v) of a 512×512 pixel license plate image can be obtained through Fourier transform, where the frequency domain coordinates (32,32) correspond to F(32,32)=100+60i (i is the imaginary unit), and its conjugate complex number F (32,32)=100-60i. According to the formula for calculating the amplitude spectrum, first calculate the product of the two, and then take the square root of the result to get H(32,32)=116.62. Perform this calculation on all frequency domain coordinates in the spectrum distribution matrix in turn to obtain the amplitude spectrum containing 512×512 amplitude data. When constructing a grayscale histogram based on the amplitude spectrum, first normalize the amplitude to gray levels 0-255 through the linear mapping formula. For example, after mapping the amplitude H=900, we get 128. Then, we traverse all amplitude data to count the number of pixels corresponding to each gray level. Among them, the number of pixels corresponding to gray level 128 is 31457, accounting for 0.12. The total number of pixels corresponding to gray levels 0-50 is 78643, accounting for about 0.3. Finally, we draw a grayscale histogram with gray levels 0-255 as the horizontal axis and the percentage as the vertical axis, which presents a single-peak feature and the peak value is located at gray level 128.

[0027] In the above method, by determining the spectral amplitude spectrum, the phase information in the spectral distribution matrix after Fourier transform is removed, retaining only the amplitude data reflecting the intensity of frequency components. This quantifies the energy distribution of each frequency component and eliminates the impact of phase interference on subsequent analysis, ensuring data standardization and reliability. A grayscale histogram is constructed based on the spectral amplitude spectrum, transforming the dispersed frequency domain amplitude data into an intuitive chart statistically analyzed by grayscale level. This clearly presents the distribution pattern, central tendency, and dispersion of grayscale values ​​in the license plate image, transforming previously difficult-to-analyze frequency domain features into quantifiable grayscale distribution indicators (such as peak position and variance). This not only bridges the gap between frequency domain features and grayscale distribution analysis but also provides a direct analytical framework for subsequent steps to determine illumination changes and identify dynamic characteristics of grayscale distribution through grayscale variance, ensuring the accuracy of illumination impact identification and the objectivity of dynamic characteristic indicators.

[0028] In one embodiment of this application, step S130 involves determining whether the grayscale variance of the grayscale histogram exceeds a preset variance threshold. If the threshold is exceeded, it is determined that the license plate image data is affected by illumination changes, and the dynamic characteristic index of the grayscale distribution is determined based on the current grayscale histogram. The method further includes the following steps: Figure 3 As shown, the specific content is as follows: Step S310: Perform Gaussian filtering on the current grayscale histogram to determine a smoothed grayscale histogram; Step S320: Determine the mean gray level based on the smoothed gray level histogram; Step S330: Determine the gray-level variance of the smoothed gray-level histogram based on the gray-level mean; Step S340: Calculate the ratio between the gray-level variance of the smoothed gray-level histogram and the gray-level variance under standard illumination, and use the ratio result as a dynamic characteristic index of gray-level distribution.

[0029] Specifically, the gray-level variance of the gray-level histogram is a core indicator for quantifying the dispersion of gray-level values ​​in a license plate image. Its value directly reflects the degree of difference between bright and dark areas in the image. A larger variance indicates a more dispersed gray-level distribution, stronger image contrast, and a greater likelihood of being affected by changes in lighting (such as direct sunlight, shadow occlusion, or sudden changes in lighting at tunnel entrances / exits). Conversely, a smaller variance indicates a more concentrated gray-level distribution, smoother image contrast, and more stable lighting conditions. The preset variance threshold is a benchmark value derived from statistical analysis of a large number of license plate image samples under standard lighting conditions. This benchmark value is determined by calculating the gray-level variance of the gray-level histogram of license plate images acquired under conditions of no lighting interference and uniform, stable lighting, and taking the statistical mean. The purpose is to provide an objective comparative standard for judging whether the current image is affected by changes in lighting, whether the image contrast is abnormal, and to ensure the consistency and reliability of the judgment results.

[0030] For example, suppose we acquire a 512×512 pixel license plate image. The total number of pixels in this license plate image is N = 512 × 512 = 262144. We can obtain the pixel proportion h(r) of each gray level using a gray-level histogram, where the peak position r... peak =128, corresponding to h(128)=0.12, grayscale mean r mean =130; Next, according to the variance formula, the square of the difference between each gray level r and the gray mean 130 is multiplied by the corresponding pixel proportion h(r), and then all results are summed to finally obtain the gray variance σ²=1800 of the image gray histogram. The preset variance threshold is determined based on the statistics of 10,000 license plate images under standard lighting, and is set to 1500. Comparing the current gray variance 1800 with the preset variance threshold 1500, since 1800>1500, it indicates that the gray distribution dispersion of the license plate image is significantly higher than the normal level, and the proportion of dark area pixels (such as the proportion of gray levels 0-50 rising to 0.3) has increased significantly. Therefore, it is determined that the license plate image data is affected by changes in lighting.

[0031] For example, a 512×512 pixel license plate image is acquired, with a total pixel count N = 512×512 = 262144. A grayscale histogram has been constructed based on its spectral amplitude. This grayscale histogram is processed using a 5×1 one-dimensional Gaussian filter kernel with a standard deviation σ = 1.2. A weighted average is then applied to each gray level and its four neighboring gray levels using convolution operations. For example, the original proportion h of gray level 128... original (128) = 0.11, the original proportions of neighboring gray levels 127 and 129 are 0.09 and 0.10 respectively. After Gaussian filtering and weighting, the smoothed proportion h smooth (128) = 0.115, resulting in a smooth grayscale histogram with no "spiking" and continuous distribution. Multiplying each grayscale level by its corresponding proportion and summing the results yields the grayscale variance σ. smooth 2 =1750. Gray-scale variance under preset standard illumination. =1200. The ratio of the smoothed grayscale variance to the standard grayscale variance is calculated to obtain Δ=1750 / 1200≈1.458. This ratio of 1.458 is the dynamic characteristic index of the grayscale distribution of the license plate image, indicating that the grayscale dispersion of the current license plate image is 1.458 times that under standard illumination, and there is a significant influence of illumination changes.

[0032] In one embodiment of this application, step S140 involves determining the modulation parameters of the license plate image data using a convolutional neural network based on dynamic feature indicators. The modulation parameters include a contrast enhancement factor and sharpening intensity. The process also includes the following steps: Figure 4As shown, the specific content is as follows: Step S410: Determine the grayscale features based on the dynamic feature indicators; Step S420: Input the grayscale features into the fully connected layer of the convolutional neural network to output the contrast enhancement factor and sharpening intensity; Step S430: Determine the modulation parameters of the license plate image data based on the contrast enhancement factor and sharpening intensity.

[0033] Specifically, using dynamic feature indicators as the core, key statistical features of the grayscale distribution of license plate images are extracted, including the mean μ, standard deviation σ, skewness g1, and kurtosis g2 of the grayscale histogram. These features are all calculated based on a smoothed grayscale histogram. The mean reflects the overall average brightness of the image; the standard deviation σ quantifies the dispersion of grayscale values; skewness describes the asymmetry of the grayscale distribution (positive skewness indicates concentrated dark pixels, negative skewness indicates concentrated bright pixels); and kurtosis g2 characterizes the steepness of the grayscale distribution (kurtosis greater than 0 indicates a steeper distribution with more concentrated details; less than 0 indicates a smoother distribution). Combining the dynamic feature indicators with the above four statistical features forms a fixed-dimensional vector (e.g., a 1×5 vector), which serves as the grayscale feature for subsequent network input.

[0034] Next, the grayscale features are input into a convolutional neural network and extracted through three convolutional layers (kernel size 3×1, stride 1, and filter numbers 32, 64, and 128 respectively). After each convolutional layer, an activation function is used to introduce non-linearity. Then, a dimensionality reduction operation is performed to transform the high-dimensional feature map into a low-dimensional feature vector, which is then input into the first fully connected layer (256 neurons) for feature fusion and then passed to the second fully connected layer (128 neurons) for precise mapping. Finally, the second fully connected layer outputs two values, corresponding to the contrast enhancement factor and sharpening intensity, respectively.

[0035] For example, after processing a 512×512 pixel license plate image, the dynamic feature index Δ=1.458 is obtained. Further, multi-dimensional grayscale features are extracted from the smoothed grayscale histogram: mean grayscale μ=72.5, standard deviation σ=18.3, skewness g1=0.8, and kurtosis g2=3.2. These features are combined with the dynamic feature index to form a grayscale feature vector [72.5, 18.3, 0.8, 3.2, 1.458]. This grayscale feature vector is input into a trained convolutional neural network. After three convolutional layers, higher-order features are extracted, and then mapped and calculated through two fully connected layers, finally outputting a contrast enhancement factor α=1.42 and a sharpening intensity β=0.67. Based on the output, the contrast enhancement factor α=1.42 and the sharpening intensity β=0.67 are integrated to determine the modulation parameters of the license plate image data as (1.42, 0.67).

[0036] In the above method, by supplementing multi-dimensional gray-scale statistical features based on dynamic feature indicators, a comprehensive and concrete gray-scale feature vector is constructed. This not only retains the quantitative information of illumination changes but also covers the morphological features of gray-scale distribution, providing sufficient recognition basis for convolutional neural networks and avoiding parameter calculation deviations caused by single indicators. By utilizing a well-trained convolutional neural network and mining the correlation between gray-scale features and modulation parameters through deep learning algorithms, the limitations of traditional fixed parameters or manual experience adjustments are overcome. The output contrast enhancement factor and sharpening intensity can accurately match the current environment (such as illumination changes and the degree of detail degradation), ensuring the adaptive characteristics of the parameters. This significantly improves the adaptability and reliability of the modulation parameters, laying a key foundation for subsequent frequency domain enhancement processing, and thus supporting the entire license plate recognition system to maintain high recognition accuracy and strong robustness in complex environments.

[0037] In one embodiment of this application, after determining the modulation parameters of the license plate image data using a convolutional neural network based on dynamic feature indicators in step S140, the following steps are further included: Figure 5 As shown, the specific content is as follows: Step S510: Determine whether the matching degree between the modulation parameters and the standard modulation parameters under the current environment does not exceed the preset matching degree threshold; Step S520: If it is determined that the preset matching degree threshold is not exceeded, the contrast enhancement factor is adjusted.

[0038] Specifically, modulation parameters refer to the contrast enhancement factors and sharpening intensities output by a convolutional neural network, used to adapt to the current environment. Their core function is to specifically optimize the image contrast and character edge clarity of license plate images. Standard modulation parameters for the current environment refer to the optimal set of modulation parameters determined through extensive sample training and expert annotation in scenarios consistent with the current environment type (e.g., dim high contrast, direct strong light, weak light with motion). This parameter set serves as the benchmark for achieving the best image enhancement effect in the current environment. It is obtained by categorizing images from a dataset of 10,000 images across multiple environments by type (e.g., intensity of light changes, degree of motion blur), taking the statistical average of the optimal modulation parameters for each environment type, and forming a standard modulation parameter library corresponding to different environments. The system can then match the corresponding standard parameters based on the characteristics of the current environment (e.g., dynamic feature indicators, spectral peak shift). Matching degree is an indicator that quantifies the degree of fit between the current modulation parameters and the standard modulation parameters. A higher matching degree indicates that the current modulation parameters are more suitable for the current environment, and the image enhancement effect is closer to the optimal level; conversely, a lower matching degree indicates that the current modulation parameters have an adaptation deviation and need adjustment. The matching degree is calculated using the mean squared error (MSE) inverse mapping method. First, the difference between the two is calculated using the mean squared error, and then the difference value is converted into a matching degree in the 0-1 interval.

[0039] For example, when the matching degree does not exceed the preset matching degree threshold, the contrast enhancement factor is adjusted first. The core reason is that the current modulation parameter adaptation deviation mainly stems from the imbalance of gray-level distribution caused by changes in illumination (the dynamic feature index has quantified the influence of illumination). The contrast enhancement factor is directly responsible for stretching the gray-level dynamic range, optimizing image contrast, and balancing the difference between brightness and darkness, making it a key parameter for adapting to changes in illumination. Sharpening intensity mainly targets character edge details, and its deviation has a relatively minor impact on the overall enhancement effect. Therefore, adjusting the contrast enhancement factor is prioritized to quickly improve the matching degree of the modulation parameters. The adjustment method employs a deviation ratio adaptive adjustment strategy, that is, based on the deviation ratio between the current contrast enhancement factor and the standard value, combined with the degree of inadequacy in the matching degree, the adjustment amount is determined to ensure that the image contrast accurately adapts to the current environmental requirements.

[0040] In the above method, using the standard modulation parameters of the current environment as a benchmark, the inadequacy of the initial modulation parameters is accurately identified through matching degree and threshold judgment, avoiding poor license plate image enhancement results caused by blindly applying modulation parameters. Based on the deviation ratio, the contrast enhancement factor is adaptively adjusted, ensuring the accuracy of the adjustment direction (towards the standard parameters) and avoiding over-adjustment or under-adjustment by adjusting the gain and value range constraints, thus rapidly improving the parameter matching degree. In this way, the core adaptation problem caused by illumination changes can be focused on, prioritizing the optimization of key parameters. Without increasing the computational load excessively, the environmental adaptability of the modulation parameters is further enhanced, ensuring that subsequent frequency domain enhancement processing can more accurately balance image contrast and offset the impact of illumination changes. This lays a more reliable parameter foundation for generating high-quality enhanced license plate images, thereby improving the stability and recognition accuracy of the entire intelligent license plate recognition system in complex environments.

[0041] In one embodiment of this application, step S150 involves determining whether the deviation between the peak position of the spectral distribution feature and a preset peak position exceeds a preset deviation threshold. If the deviation exceeds the preset deviation threshold, it is determined that the license plate image data is affected by motion blur, and an enhanced spectral representation is determined based on the license plate image data. The step also includes the following steps: Figure 6 As shown, the specific content is as follows: Step S610: Determine the first time-domain image signal based on the license plate image data; Step S620: Based on the first time-domain image signal, use Fourier transform to determine the frequency-domain image signal; Step S630: Determine the initial amplitude frequency based on the frequency domain image signal; Step S640: Determine the degree of motion blur based on the deviation between the peak position of the spectral distribution characteristics and the preset peak position; Step S650: Determine the amplitude gain of the initial amplitude frequency amplitude based on the degree of motion blur; Step S660: Determine the enhanced spectrum representation based on the amplitude gain.

[0042] Specifically, the peak position of the spectral distribution features refers to the frequency coordinates corresponding to the maximum amplitude point of the high-frequency part in the frequency domain after the license plate image undergoes Fourier transform. This high-frequency part corresponds to key details such as the edges and contours of the license plate characters, and the stability of its peak position directly reflects the clarity of the character edges. When there is no motion blur, the peak position is fixed. When motion blur exists, the degradation of character edge information leads to the dispersion of high-frequency components, causing the peak position to shift. For example, from the spectral distribution features of the current license plate image, the peak position of the high-frequency part is extracted (determined by locating the maximum amplitude point); secondly, the absolute deviation between this peak position and the preset peak position (150Hz) is calculated; finally, the calculated deviation is compared with the preset deviation threshold (2Hz). If the deviation exceeds the preset deviation threshold, it indicates that the high-frequency components of the character edges in the current license plate image are significantly dispersed due to motion blur, and the peak position shift exceeds the normal range, thus confirming that the license plate image data is affected by motion blur; if the deviation does not exceed the preset deviation threshold, it indicates that there is no significant motion blur, and the character edge information is basically intact.

[0043] For example, given a 256×128 pixel license plate image, the grayscale value corresponding to each spatial coordinate is extracted to form a 256×128-dimensional first time-domain image signal. The grayscale value at coordinate (100, 50) is 120, and the grayscale value at coordinate (150, 80) is 85. A two-dimensional Fast Fourier Transform (FFT) algorithm is used to transform the first time-domain image signal, obtaining a complex frequency-domain image signal. The frequency-domain coordinate (30, 20) corresponds to F(30, 20) = 100 + 60i, and the frequency-domain coordinate (40, 30) corresponds to F(40, 30) = 80 + 40i. The initial amplitude-frequency magnitude is calculated for the complex elements of each frequency-domain coordinate; similarly, the initial amplitude-frequency magnitude for all frequency-domain coordinates is calculated. Given that the peak position of the current spectral distribution is 147.2Hz, the preset peak position is 150Hz, and the deviation is |147.2-150|=2.8Hz, the motion blur level is determined to be 1.87%. The amplitude gain of the initial amplitude is then determined based on the motion blur level. Amplitude gain is an adjustment coefficient used to compensate for the amplitude attenuation of high-frequency components caused by motion blur. Its core function is to restore the intensity of character edge details by amplifying the amplitude of high-frequency components, thus counteracting the negative impact of motion blur. For example, if the amplitude gain is determined to be G=1.019, the initial amplitude is multiplied by the amplitude gain 1.019 sequentially. For instance, if the initial amplitude is 116.62, the enhanced amplitude is 116.62×1.019≈118.84. Simultaneously, the original phase information is preserved. The enhanced amplitude is combined with the original phase information to form a new complex spectral distribution matrix, which is the enhanced spectral representation.

[0044] In the above method, the abstract effects of motion blur are transformed into quantifiable frequency domain features by converting the time-domain signal to the frequency-domain signal. By extracting the initial amplitude and frequency, phase interference is removed, and the focus is on the intensity analysis of frequency components, ensuring the targeted nature of subsequent gain adjustments. The degree of motion blur is quantified by the deviation ratio, transforming the blurring effect from an abstract phenomenon into a calculable numerical indicator, providing an objective basis for gain determination. Furthermore, the amplitude gain is dynamically calculated based on the degree of blur, achieving precise matching between gain and blur level, effectively compensating for high-frequency component attenuation while avoiding distortion caused by over-enhancement. Thus, quantitative analysis and precise compensation of motion blur effects are achieved, ensuring that the subsequent inverse Fourier transform can generate enhanced images with clear character edges and complete details. This provides high-quality frequency domain data support for improving license plate recognition accuracy and effectively enhances the robustness and adaptability of the entire intelligent license plate recognition system in vehicle motion scenarios.

[0045] In one embodiment of this application, in step S160, an inverse Fourier transform is used to determine the enhanced license plate image data based on the enhanced spectral representation. Specifically, the enhanced spectral representation is a complex-form spectral distribution matrix obtained after the aforementioned steps. Its core feature is that, through amplitude gain compensation and modulation parameter optimization, it specifically enhances the high-frequency components corresponding to the edges of the license plate characters, while retaining the stable information of the low-frequency background. Each element in the matrix contains the enhanced amplitude and the original phase information (the phase information ensures that the positional correspondence of the frequency components is not distorted). The inverse Fourier transform is used to determine the enhanced license plate image data based on the enhanced spectral representation, which maps the enhancement processing effect completed in the frequency domain back to the spatial domain, restoring it to visualized license plate image data, thus realizing a closed loop of frequency domain enhancement → spatial domain image conversion.

[0046] In one embodiment of this application, after determining the enhanced license plate image data based on the enhanced spectral representation using inverse Fourier transform in step S160, the following steps are further included: Figure 7 As shown, the specific content is as follows: Step S710: Determine the character edge positions based on the enhanced license plate graphic data; Step S720: Determine the edge restoration map based on the character edge position; Step S730: Use the Laplacian operator to determine whether the sharpness score of the edge restoration map exceeds the preset sharpness score threshold; Step S740: If it is determined that the preset clarity score threshold is not exceeded, the amplitude gain is adjusted to optimize and enhance the license plate image data.

[0047] Specifically, for example, when enhancing a license plate image data of 512×512 pixel grayscale, the Sobel gradient operator is used for edge detection. The horizontal and vertical gradient magnitudes are calculated using horizontal and vertical convolution kernels respectively, resulting in a comprehensive gradient magnitude. A preset gradient magnitude threshold of 50 is used. After traversing all coordinates, a set of character edge positions is determined, including continuous coordinates such as (120,150), (121,150), and (120,151). However, the position (130,160) is not included due to its gradient magnitude of 45, indicating an edge break. A 512×512 pixel blank binary image is constructed, setting the pixel values ​​of the coordinates in the edge position set to 255, while keeping the remaining coordinates 0. This results in an edge reconstruction map where the character edges are presented as white lines, with a clear break near (130,160). The edge reconstruction map was convolved using a Laplacian operator convolution kernel. The sum of the absolute values ​​of the Laplacian responses of all pixels was calculated, resulting in a sharpness score of 3200. After comparing this score with the preset sharpness score threshold of 3500, it was determined that the score did not exceed the threshold. Therefore, the original amplitude gain of 1.019 was adjusted to 1.089, which did not exceed the upper limit of gain of 1.5. The adjusted amplitude gain of 1.089 was then reapplied to the enhanced spectral representation construction. After inverse Fourier transform, the optimized enhanced license plate image data was generated. The sharpness score of the edge reconstruction map was improved to 3580, exceeding the preset threshold. The broken edges of the characters were repaired, and the sharpness met the recognition requirements.

[0048] In the above method, the gradient operator is used to accurately extract the character edge positions, avoiding interference from background information in the sharpness judgment. By constructing an edge reconstruction map, abstract edge features are transformed into intuitive binary images, making the integrity and blurriness of the edges directly observable and quantifiable. The Laplacian operator is used to calculate the sharpness score, combined with a preset threshold, to achieve an objective and quantitative judgment of edge sharpness, providing a clear trigger for parameter adjustment and avoiding the bias of subjective judgment. Finally, the amplitude gain is dynamically adjusted to ensure that the gain optimization is precisely matched with the edge blurriness, which not only further amplifies the high-frequency components corresponding to the character edges and repairs edge breakage and blurring problems, but also avoids image distortion through the upper limit constraint of the gain. In this way, the key detail quality of the enhanced license plate image data is significantly improved, effectively strengthening the robustness and recognition accuracy of the entire license plate intelligent recognition system in complex environments.

[0049] This application also provides a license plate intelligent recognition system based on image enhancement. The system may include an acquisition module, a construction module, a first judgment module, a first determination module, a second judgment module, a second determination module, and a third judgment module. The acquisition module acquires license plate image data and performs Fourier transform processing on the license plate image data to determine spectral distribution characteristics. The construction module constructs a grayscale histogram based on the spectral distribution characteristics. The first judgment module determines whether the grayscale variance of the grayscale histogram exceeds a preset variance threshold. If it exceeds the preset variance threshold, it determines that the license plate image data is affected by illumination changes and determines the dynamic characteristic index of the grayscale distribution based on the current grayscale histogram. The first determination module uses a convolutional neural network to determine the modulation parameters of the license plate image data based on the dynamic characteristic index. The modulation parameters include a contrast enhancement factor and a sharpening intensity. The second judgment module determines whether the deviation between the peak position of the spectral distribution characteristics and a preset peak position exceeds a preset deviation threshold. If it exceeds the preset deviation threshold, it determines that the license plate image data is affected by motion blur and determines the enhanced spectral representation based on the license plate image data. The second determining module is used to determine the enhanced license plate image data based on the enhanced spectrum representation using inverse Fourier transform; the third judging module is used to judge whether the gray-level variance of the enhanced license plate image data exceeds a preset variance threshold. If the judgment exceeds the preset variance threshold, the iterative modulation parameters are updated.

[0050] It should be noted that the embodiments of the image-enhanced license plate intelligent recognition system provided in this application can be used to execute the processing flow of the embodiments of the image-enhanced license plate intelligent recognition method in the above embodiments. Its functions will not be repeated here, but can be referred to the detailed description of the above method embodiments.

[0051] This application also provides an electronic device, which includes one or more processors and memory resources represented by memory for storing instructions executable by the processor, such as application programs. The application programs stored in the memory may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor is configured to execute instructions to perform the aforementioned graphics-enhanced intelligent license plate recognition method.

[0052] The electronic device may also include a power supply component configured to perform power management of the electronic device, a wired or wireless network interface configured to connect the electronic device to a network, and an input / output (I / O) interface. The electronic device can be operated based on operating devices stored in memory, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0053] In one embodiment, a computer device, which may be a server, is also provided. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device stores data. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a graphics-enhanced intelligent license plate recognition method.

[0054] In one embodiment, a computer device is provided, which may be a terminal. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a graphics-enhanced intelligent license plate recognition method. The display unit of the computer device is used to form a visually visible image and may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0055] This application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of the electronic device, enables the electronic device to execute a graphics-enhanced intelligent license plate recognition system method.

[0056] This application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0057] It should be noted that although the steps of the image-enhanced license plate intelligent recognition method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps, such as omitting certain steps, combining multiple steps into one step, and / or decomposing one step into multiple steps, should all be considered part of this application.

[0058] It should be understood that this application is not limited to the detailed structure and arrangement of the modules in the graphics-enhanced intelligent license plate recognition system proposed in this specification. This application can have other implementations and can be implemented and executed in various ways. The foregoing variations and modifications fall within the scope of this application. It should be understood that the application and its definition in this specification extend to all alternative combinations of two or more individual features mentioned or apparent in the text and / or drawings. All these different combinations constitute multiple alternative aspects of this application. The embodiments described in this specification illustrate the best known mode for implementing this application and will enable those skilled in the art to utilize this application.

Claims

1. A method for intelligent license plate recognition based on image enhancement, characterized in that, include: Acquire license plate image data and perform Fourier transform processing on the license plate image data to determine the spectral distribution characteristics; Construct a grayscale histogram based on the aforementioned spectral distribution characteristics; Determine whether the grayscale variance of the grayscale histogram exceeds a preset variance threshold. If it is determined that the grayscale variance exceeds the preset variance threshold, then it is determined that the license plate image data is affected by illumination changes, and the dynamic characteristic index of grayscale distribution is determined based on the current grayscale histogram. Based on the dynamic feature index, a convolutional neural network is used to determine the modulation parameters of the license plate image data, the modulation parameters including contrast enhancement factor and sharpening intensity; Determine whether the deviation between the peak position of the spectral distribution feature and the preset peak position exceeds a preset deviation threshold. If it is determined that the deviation exceeds the preset deviation threshold, then it is determined that the license plate image data has motion blur effects, and an enhanced spectral representation is determined based on the license plate image data. The enhanced license plate image data is determined by using inverse Fourier transform based on the enhanced spectrum representation; Determine whether the grayscale variance of the enhanced license plate image data exceeds the preset variance threshold. If it is determined that it exceeds the preset variance threshold, then update the modulation parameters iteratively.

2. The intelligent license plate recognition method based on image enhancement according to claim 1, characterized in that, Determining the grayscale histogram based on the spectral distribution characteristics includes: Determine the spectral amplitude spectrum based on the aforementioned spectral distribution characteristics; A grayscale histogram is constructed based on the amplitude spectrum.

3. The image enhancement-based intelligent license plate recognition method according to claim 1, characterized in that, The step of determining the dynamic characteristic index of grayscale distribution based on the current grayscale histogram includes: The current grayscale histogram is subjected to Gaussian filtering to determine a smooth grayscale histogram; The mean gray level is determined based on the smoothed gray level histogram. The gray-level variance of the smoothed gray-level histogram is determined based on the gray-level mean. The gray-level variance of the smoothed gray-level histogram is compared with the gray-level variance under standard illumination, and the result of the ratio calculation is used as the dynamic characteristic index of the gray-level distribution.

4. The image enhancement-based intelligent license plate recognition method according to claim 1, characterized in that, The step of determining the modulation parameters of the license plate image data using a convolutional neural network based on the dynamic feature index includes: The grayscale features are determined based on the dynamic feature indicators; The grayscale features are input into the fully connected layer of the convolutional neural network to output a contrast enhancement factor and sharpening intensity. The modulation parameters of the license plate image data are determined based on the contrast enhancement factor and the sharpening intensity.

5. The image enhancement-based intelligent license plate recognition method according to claim 4, characterized in that, After determining the modulation parameters of the license plate image data based on the contrast enhancement factor and the sharpening intensity, the method further includes: Determine whether the matching degree between the modulation parameters and the standard modulation parameters in the current environment does not exceed a preset matching degree threshold; If the preset matching threshold is not exceeded, the contrast enhancement factor is adjusted.

6. The intelligent license plate recognition method based on image enhancement according to claim 1, characterized in that, The step of determining the enhanced spectral representation based on the license plate image data includes: A first time-domain image signal is determined based on the license plate image data; Based on the first time-domain image signal, the frequency-domain image signal is determined by Fourier transform; The initial amplitude frequency is determined based on the frequency domain image signal; The degree of motion blur is determined based on the deviation between the peak position of the spectral distribution characteristics and the preset peak position; The amplitude gain of the initial amplitude frequency amplitude is determined based on the degree of motion ambiguity; The enhanced spectrum representation is determined based on the amplitude gain.

7. The intelligent license plate recognition method based on image enhancement according to claim 6, characterized in that, After determining the enhanced license plate image data based on the enhanced spectral representation using the inverse Fourier transform, the method further includes: Determine the character edge positions based on enhanced license plate graphic data; Determine the edge restoration map based on the character edge position; The Laplacian operator is used to determine whether the sharpness score of the edge reconstruction map exceeds a preset sharpness score threshold. If it is determined that the preset clarity score threshold is not exceeded, the amplitude gain is adjusted to optimize the enhanced license plate image data.

8. A license plate intelligent recognition system based on image enhancement, characterized in that, include: The acquisition module is used to acquire license plate image data and perform Fourier transform processing on the license plate image data to determine the spectral distribution characteristics; The construction module is used to construct a grayscale histogram based on the spectral distribution characteristics; The first judgment module is used to determine whether the gray-level variance of the gray-level histogram exceeds a preset variance threshold. If it is determined that the gray-level variance exceeds the preset variance threshold, it is determined that the license plate image data is affected by illumination changes, and the dynamic characteristic index of gray-level distribution is determined based on the current gray-level histogram. The first determining module is used to determine the modulation parameters of the license plate image data based on the dynamic feature index using a convolutional neural network. The modulation parameters include a contrast enhancement factor and a sharpening intensity. The second judgment module is used to determine whether the deviation between the peak position of the spectral distribution feature and the preset peak position exceeds a preset deviation threshold. If the deviation exceeds the preset deviation threshold, it is determined that the license plate image data has motion blur effects, and the enhanced spectral representation is determined based on the license plate image data. The second determining module is used to determine the enhanced license plate image data based on the enhanced spectrum representation using inverse Fourier transform; The third judgment module is used to determine whether the grayscale variance of the enhanced license plate image data exceeds the preset variance threshold. If it is determined that the grayscale variance exceeds the preset variance threshold, the modulation parameters are updated iteratively.

Images