System and method for comprehensively detecting state of curved screen

Abstract

The method for comprehensively detecting the state of the curved screen is characterized by comprising the following steps: acquiring images of curved screens in different states; preprocessing the image to obtain preprocessed data; marking the image, and performing feature extraction on the preprocessed data; inputting the obtained feature data into a model for training to obtain a trained verification model; and identifying the curved screen image by using the verification model to obtain the state of the curved screen. By integrating various machine vision technologies and algorithms, various state problems, including dark spots, bright spots, damaged points, color distortion and the like, of the curved screen can be accurately detected, and the comprehensiveness and accuracy of state detection are improved. The verification model can complete a state detection task in a shorter time, and the detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of screen detection technology, and in particular to a system and method for comprehensively detecting the state of a curved screen. Background Art

[0002] With the rapid development of curved screen technology, its smooth, immersive display effects and beautiful design have made it popular. However, with the rapid popularization of curved screen technology, it is crucial to accurately monitor its status and handle it in a timely manner.

[0003] Current curved screen status detection mainly relies on traditional methods, including manual inspection and simple sensor monitoring. However, these methods have obvious limitations:

[0004] Manual inspection is not accurate enough: Methods that rely on manual inspection are subject to subjective factors and cannot guarantee accurate assessment of the curved screen status. Limited sensor monitoring: Sensor monitoring is usually limited to basic states, such as on or off, and cannot provide more comprehensive monitoring of curved screens. Existing technologies may have insufficient accuracy when detecting the status of curved screens. Especially in complex situations, such as poor lighting conditions or when the screen is slightly damaged, existing technologies may not provide accurate detection results.

[0005] In summary, the existing technologies for curved screen status monitoring have defects such as insufficient accuracy, limited monitoring range, and slow response speed. These problems seriously affect the stability and user experience of curved screen devices, and a more advanced and efficient curved screen status detection system is urgently needed to solve these challenges. Summary of the invention

[0006] In view of the above-mentioned defects, the purpose of the present invention is to provide a system and method for comprehensively detecting the state of a curved screen, so as to solve the problems of low detection accuracy and limited detection range of the existing curved screen.

[0007] To achieve this purpose, the present invention adopts the following technical solution: A method for comprehensively detecting the state of a curved screen, comprising the following steps:

[0008] Step S1: collecting images of the curved screen in different states;

[0009] Step S2: preprocessing the image to obtain preprocessed data;

[0010] Step S3: label the image and perform feature extraction on the preprocessed data;

[0011] Step S4: input the acquired feature data into the model for training to obtain a trained verification model;

[0012] Step S5: using the verification model to identify the curved screen image and obtain the state of the curved screen.

[0013] Preferably, the preprocessing in step S2 includes a combination of one or more of image denoising, image enhancement, and illumination correction.

[0014] Preferably, the step of image denoising is: denoising is achieved by convolving each pixel of the image with a Gaussian kernel; wherein the formula for convolution denoising is as follows:

[0015]

[0016] in I(x, y) represents the image, x, y are the pixel coordinates of the image, σ is the standard deviation of the Gaussian distribution, and n is the size of the Gaussian kernel;

[0017] The steps of image enhancement are: by adjusting the grayscale distribution of the image, the grayscale histogram of the image is made uniform, thereby enhancing the contrast of the image;

[0018] The formula for enhancing the image is as follows:

[0019] S k =(L-1)CDF(r k );

[0020] (L-1) is the grayscale range of the image, Where P r (r i ) is the probability of the i-th gray level appearing;

[0021] The steps for illumination correction are as follows:

[0022] Processing the grayscale value of the image by using the reflection component and the illumination component of the reflected light to obtain a first image;

[0023] Taking the logarithm of the first image to obtain a first parameter;

[0024] Performing Fourier transform on the first parameter to obtain a second parameter;

[0025] Design a filter H(u,v) to filter the second parameter, separate and adjust the reflection component and the illumination component: H(u,v) = H L (u,v)+H H (u,v), where H L (u,v) is the reflection component, H H (u,v) is the illumination component;

[0026] The filtered image is then inverse Fourier transformed:

[0027] i f (x,y)=F -1 [H(u,v)F[lnI(x,y)]], F[] is the Fourier transform;

[0028] Finally, the corrected image is obtained through exponential operation:

[0029]

[0030] Preferably, the steps of feature extraction in step S3 are as follows:

[0031] Step S31: Take one of the pixels as the center pixel g c , construct a 3*3 neighborhood and compare the grayscale values ​​of the 8 pixels in the neighborhood with the central pixel g c The size between them is used to obtain its LBP value;

[0032] Step S32: Replace the next pixel as the new center pixel and re-execute step S31 until the LBP values ​​of all pixels of the image are extracted;

[0033] Step S33: Count the histogram of LBP values ​​of all pixels to obtain texture features;

[0034] Step S34: convert the image from the RGB color space to the HSV color space, and extract the histograms of the three channels H, S, and V respectively;

[0035] Combine the histograms of the three channels to obtain color features;

[0036] Step S35: using the Canny algorithm to extract edge features in the image;

[0037] Step S36: vectorize the texture features, color features and edge features, and combine them to obtain a comprehensive feature vector.

[0038] Preferably, the steps of step S4 are as follows: all comprehensive feature vectors and corresponding category labels are divided into a training set and a test set according to a preset ratio, wherein the training set is used to train the SVM model, and the test set is used to evaluate the performance of the model.

[0039] A system for comprehensively detecting the state of a curved screen, characterized in that the method for comprehensively detecting the state of a curved screen is used, comprising a data acquisition module, a preprocessing module, a feature extraction module, a training module and a detection module;

[0040] The data acquisition module is used to collect images of the curved screen in different states;

[0041] The preprocessing module is used to preprocess the image to obtain preprocessing data;

[0042] The feature extraction module is used to mark the image and extract features from the preprocessed data;

[0043] The training module is used to input the acquired feature data into the model for training to obtain a trained verification model;

[0044] The detection module is used to identify the curved screen image using the verification model to obtain the state of the curved screen.

[0045] Preferably, the preprocessing module includes a combination of one or more of image denoising, image enhancement, and illumination correction.

[0046] Preferably, the preprocessing module includes a denoising submodule, an enhancement submodule and a correction submodule;

[0047] The denoising submodule is used to achieve denoising by convolving each pixel of the image with a Gaussian kernel; the formula for convolution denoising is as follows:

[0048]

[0049] in I(x, y) represents the image, x, y are the pixel coordinates of the image, σ is the standard deviation of the Gaussian distribution, and n is the size of the Gaussian kernel;

[0050] The enhancement submodule is used to adjust the grayscale distribution of the image to make the grayscale histogram of the image uniform, thereby enhancing the contrast of the image;

[0051] The formula for enhancing the image is as follows:

[0052] S k =(L-1)CDF(r k );

[0053] (L-1) is the grayscale range of the image, Where P r (r i ) is the probability of the i-th gray level appearing;

[0054] The correction submodule is used to process the grayscale value of the image through the reflection component and the illumination component of the reflected light to obtain a first image;

[0055] Taking the logarithm of the first image to obtain a first parameter;

[0056] Performing Fourier transform on the first parameter to obtain a second parameter;

[0057] Design a filter H(u,v) to filter the second parameter, separate and adjust the reflection component and the illumination component: H(u,v) = H L (u,v)+H H (u,v);

[0058] The filtered image is then inverse Fourier transformed:

[0059] i f (x,y)=F -1 [H(u,v)F[lnI(x,y)]];

[0060] Finally, the corrected image is obtained through exponential operation:

[0061]

[0062] Preferably, the feature extraction module includes a first submodule, a second submodule, a texture extraction submodule, an edge feature extraction submodule, a color feature extraction submodule and a combination submodule;

[0063] The first submodule is used to take one of the pixels as the central pixel g c , construct a 3*3 neighborhood and compare the grayscale values ​​of the 8 pixels in the neighborhood with the central pixel g c The size between them is used to obtain its LBP value;

[0064] The second submodule is used to replace the next pixel as the new center pixel and re-call the first submodule until the LBP values ​​of all pixels of the image are extracted;

[0065] The texture extraction submodule is used to count the histogram of LBP values ​​of all pixels to obtain texture features;

[0066] The color feature extraction submodule is used to convert the image from the RGB color space to the HSV color space, and extract the histograms of the three channels of H, S, and V respectively;

[0067] Combine the histograms of the three channels to obtain color features;

[0068] The edge feature extraction submodule is used to extract edge features in the image using the Canny algorithm;

[0069] The combination submodule is used to vectorize texture features, color features and edge features, and combine them to obtain a comprehensive feature vector.

[0070] One of the above technical solutions has the following advantages or beneficial effects: the invention can accurately detect various status problems of the curved screen, including dark spots, bright spots, damaged spots, color distortion, etc., by integrating multiple machine vision technologies and algorithms, thereby improving the comprehensiveness and accuracy of status detection. The verification model can complete the status detection task in a shorter time, thereby improving the detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 is a flow chart of an embodiment of the method of the present invention.

[0072] Figure 2 It is a structural schematic diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0073] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0074] In the description of the embodiments of the present invention, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0075] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, the meaning of "plurality" is two or more. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0076] like Figures 1-2 As shown, a method for comprehensively detecting the state of a curved screen includes the following steps:

[0077] Step S1: collecting images of the curved screen in different states;

[0078] Collect a large number of images of curved screens in different states, including normal curved screen images and images with damaged areas, bright spots, dark spots, and other abnormal conditions. These images need to be representative, covering different surface shapes, resolutions, lighting conditions, etc.

[0079] Step S2: preprocessing the image to obtain preprocessed data;

[0080] By preprocessing the image, the edge features in the image are well preserved, the adjustability is strong, the calculation efficiency is high, and the ringing phenomenon is avoided, which facilitates the subsequent feature extraction of the preprocessed data.

[0081] Step S3: label the image and perform feature extraction on the preprocessed data;

[0082] The state of the curved screen in each image is clarified, such as normal, damaged, bright spot, dark spot, etc., and marked with corresponding category labels (for example, normal is 1, damaged is -1, bright spot is 2, dark spot is 3, etc.). Then, feature extraction is performed on the preprocessed data to obtain features corresponding to different states. These features can effectively feedback the characteristics of the image in different states. In the subsequent model, these features can be used to quickly identify the state.

[0083] Step S4: input the acquired feature data into the model for training to obtain a trained verification model; in the present invention, the SVM model is used as the training model, and the SVM model is used as a commonly used model for classification. The detection of the curved screen includes a variety of features, and the SVM model can process multidimensional data and nonlinear problems, and can handle complex classification tasks.

[0084] Step S5: using the verification model to identify the curved screen image and obtain the state of the curved screen.

[0085] The invention integrates a variety of machine vision technologies and algorithms to accurately detect various status problems of curved screens, including dark spots, bright spots, damaged spots, color distortion, etc., improving the comprehensiveness and accuracy of status detection. The verification model can complete status detection tasks in a shorter time, improving detection efficiency.

[0086] Preferably, the preprocessing in step S2 includes a combination of one or more of image denoising, image enhancement, and illumination correction.

[0087] Preferably, the step of image denoising is: denoising is achieved by convolving each pixel of the image with a Gaussian kernel; wherein the formula for convolution denoising is as follows:

[0088]

[0089] in I(x, y) represents the image, x, y are the pixel coordinates of the image, σ is the standard deviation of the Gaussian distribution, and n is the size of the Gaussian kernel;

[0090] The template coefficient of the Gaussian filter decreases as the distance from the template center increases, so it has a smaller blurring effect on the image and can better preserve the image details. Compared with other filtering and denoising methods, Gaussian filtering can better preserve the main features of the image, such as contours and edges, while smoothing the image. The edges are extracted later.

[0091] The steps of image enhancement are: by adjusting the grayscale distribution of the image, the grayscale histogram of the image is made uniform, thereby enhancing the contrast of the image;

[0092] The formula for enhancing the image is as follows:

[0093] S k =(L-1)CDF(r k );

[0094] (L-1) is the grayscale range of the image, Where P r (r i ) is the probability of the i-th gray level appearing;

[0095] In some images, due to light and shadow problems, some parts of the image are blurred, and some details are difficult to extract. For this reason, the present invention uses histogram equalization to process the image, so that its grayscale distribution is more uniform and the contrast is enhanced. After processing, the clarity of some detailed textures can be further improved, so as to facilitate the subsequent extraction of texture features in the image.

[0096] The steps for lighting correction are as follows:

[0097] Processing the grayscale value of the image by using the reflection component and the illumination component of the reflected light to obtain a first image;

[0098] Taking the logarithm of the first image to obtain a first parameter;

[0099] Performing Fourier transform on the first parameter to obtain a second parameter;

[0100] Design a filter H(u,v) to filter the second parameter, separate and adjust the reflection component and the illumination component: H(u,v) = H L (u,v)+H H (u,v), where H L (u,v) is the reflection component, H H (u,v) is the illumination component;

[0101] The filtered image is then inverse Fourier transformed:

[0102] i f (x,y)=F -1[H(u,v)F[lnI(x,y)]], F[] is the Fourier transform;

[0103] Finally, the corrected image is obtained through exponential operation:

[0104]

[0105] By separating and adjusting the reflection component and illumination component of the image, this method can significantly enhance the contrast of the image, thereby improving the overall color of the image and facilitating the subsequent extraction of color features.

[0106] Preferably, the steps of feature extraction in step S3 are as follows:

[0107] Step S31: Take one of the pixels as the center pixel g c , construct a 3*3 neighborhood and compare the grayscale values ​​of the 8 pixels in the neighborhood with the central pixel g c The size between them is used to obtain its LBP value;

[0108] Step S32: Replace the next pixel as the new center pixel and re-execute step S31 until the LBP values ​​of all pixels of the image are extracted;

[0109] Step S33: Count the histogram of LBP values ​​of all pixels to obtain texture features;

[0110] Texture extraction is performed through LBP values, and the extracted local texture information is rich in texture. Moreover, due to its simple calculation and rotation invariance, images of curved screens taken from different directions can be detected.

[0111] Step S34: convert the image from the RGB color space to the HSV color space, and extract the histograms of the three channels H, S, and V respectively;

[0112] Combine the histograms of the three channels to obtain color features;

[0113] In the detection of curved screens, since the curved screen has a certain curvature, the brightness and luminance of the curved part will be reduced. If a defect appears in this place, it will be difficult to identify. For this reason, in order to improve the feature extraction of the curved part in the present invention, the image is converted from the RGB color space to the HSV color space. Because some characteristics of the color may not be intuitively reflected in the RGB color space, such as the characteristics of the color purity and brightness (if a dark spot appears, it cannot be identified). In the HSV color space, by separating the three channels of hue, saturation and brightness, the color characteristics can be extracted more accurately.

[0114] Step S35: Use the Canny algorithm to extract edge features in the image; the Canny algorithm is insensitive to changes in the grayscale value of the image and can process any grayscale image. In addition, it can also adapt to edge detection of different scales and directions and has strong robustness.

[0115] Step S36: vectorize the texture features, color features and edge features, and combine them to obtain a comprehensive feature vector.

[0116] The present invention can form a more comprehensive and rich feature vector by combining texture features, color features and edge features. These features capture different aspects of the image, such as the subtle structure of the texture, the global distribution of the color and the geometric shape of the edge, thereby enhancing the expressive power of the feature vector and meeting the needs of defect detection for curved screens.

[0117] Preferably, the steps of step S4 are as follows: all comprehensive feature vectors and corresponding category labels are divided into a training set and a test set according to a preset ratio (such as 70% training set, 30% test set), wherein the training set is used to train the SVM model, and the test set is used to evaluate the performance of the model.

[0118] A system for comprehensively detecting the state of a curved screen, characterized in that the method for comprehensively detecting the state of a curved screen is used, comprising a data acquisition module, a preprocessing module, a feature extraction module, a training module and a detection module;

[0119] The data acquisition module is used to collect images of the curved screen in different states;

[0120] The preprocessing module is used to preprocess the image to obtain preprocessing data;

[0121] The feature extraction module is used to mark the image and extract features from the preprocessed data;

[0122] The training module is used to input the acquired feature data into the model for training to obtain a trained verification model;

[0123] The detection module is used to identify the curved screen image using the verification model to obtain the state of the curved screen.

[0124] Preferably, the preprocessing module includes a combination of one or more of image denoising, image enhancement, and illumination correction.

[0125] Preferably, the preprocessing module includes a denoising submodule, an enhancement submodule and a correction submodule;

[0126] The denoising submodule is used to achieve denoising by convolving each pixel of the image with a Gaussian kernel; the formula for convolution denoising is as follows:

[0127]

[0128] in I(x, y) represents the image, x, y are the pixel coordinates of the image, σ is the standard deviation of the Gaussian distribution, and n is the size of the Gaussian kernel;

[0129] The enhancement submodule is used to adjust the grayscale distribution of the image to make the grayscale histogram of the image uniform, thereby enhancing the contrast of the image;

[0130] The formula for enhancing the image is as follows:

[0131] S k =(L-1)CDF(r k );

[0132] (L-1) is the grayscale range of the image, Where P r (r i ) is the probability of the i-th gray level appearing;

[0133] The correction submodule is used to process the grayscale value of the image through the reflection component and the illumination component of the reflected light to obtain a first image;

[0134] Taking the logarithm of the first image to obtain a first parameter;

[0135] Performing Fourier transform on the first parameter to obtain a second parameter;

[0136] Design a filter H(u,v) to filter the second parameter, separate and adjust the reflection component and the illumination component: H(u,v) = H L (u,v)+H H (u,v);

[0137] The filtered image is then inverse Fourier transformed:

[0138] i f (x,y)=F -1 [H(u,v)F[lnI(x,y)]];

[0139] Finally, the corrected image is obtained through exponential operation:

[0140]

[0141] Preferably, the feature extraction module includes a first submodule, a second submodule, a texture extraction submodule, an edge feature extraction submodule, a color feature extraction submodule and a combination submodule;

[0142] The first submodule is used to take one of the pixels as the central pixel g c , construct a 3*3 neighborhood and compare the grayscale values ​​of the 8 pixels in the neighborhood with the central pixel g c The size between them is used to obtain its LBP value;

[0143] The second submodule is used to replace the next pixel as the new center pixel and re-call the first submodule until the LBP values ​​of all pixels of the image are extracted;

[0144] The texture extraction submodule is used to count the histogram of LBP values ​​of all pixels to obtain texture features;

[0145] The color feature extraction submodule is used to convert the image from the RGB color space to the HSV color space, and extract the histograms of the three channels of H, S, and V respectively;

[0146] Combine the histograms of the three channels to obtain color features;

[0147] The edge feature extraction submodule is used to extract edge features in the image using the Canny algorithm;

[0148] The combination submodule is used to vectorize texture features, color features and edge features, and combine them to obtain a comprehensive feature vector.

[0149] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0150] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A method for comprehensively detecting the state of a curved screen, characterized in that: The steps include: Step S1: collecting images of the curved screen in different states; Step S2: preprocessing the image to obtain preprocessed data; Step S3: label the image and perform feature extraction on the preprocessed data; Step S4: input the acquired feature data into the model for training to obtain a trained verification model; Step S5: using the verification model to identify the curved screen image and obtain the state of the curved screen.

2. The method for comprehensively detecting the state of a curved screen according to claim 1, characterized in that: The preprocessing in step S2 includes one or more combinations of image denoising, image enhancement, and illumination correction.

3. The method for comprehensively detecting the state of a curved screen according to claim 2, characterized in that: The image denoising step is: denoising is achieved by convolving each pixel of the image with a Gaussian kernel; wherein the formula for convolution denoising is as follows: in I(x, y) represents the image, x, y are the pixel coordinates of the image, σ is the standard deviation of the Gaussian distribution, and n is the size of the Gaussian kernel; The steps of image enhancement are: by adjusting the grayscale distribution of the image, the grayscale histogram of the image is made uniform, thereby enhancing the contrast of the image; The formula for enhancing the image is as follows: S k =(L-1)CDF(r k ); (L-1) is the grayscale range of the image, Where P r (r i ) is the probability of the i-th gray level appearing; The steps for lighting correction are as follows: Processing the grayscale value of the image by using the reflection component and the illumination component of the reflected light to obtain a first image; Taking the logarithm of the first image to obtain a first parameter; Performing Fourier transform on the first parameter to obtain a second parameter; Design a filter H(u,v) to filter the second parameter, separate and adjust the reflection component and the illumination component: H(u,v) = H L (u,v)+H H (u,v), where H L (u,v) is the reflection component, H H (u,v) is the illumination component; The filtered image is then inverse Fourier transformed: i f (x,y)=F -1 [H(u,v)F[lnI(x,y)]], F[] is the Fourier transform; Finally, the corrected image is obtained through exponential operation:

4. The method for comprehensively detecting the state of a curved screen according to claim 1, characterized in that: The steps of feature extraction in step S3 are as follows: Step S31: Take one of the pixels as the center pixel g c , construct a 3*3 neighborhood and compare the grayscale values ​​of the 8 pixels in the neighborhood with the central pixel g c The size between them is used to obtain its LBP value; Step S32: Replace the next pixel as the new center pixel and re-execute step S31 until the LBP values ​​of all pixels of the image are extracted; Step S33: Count the histogram of LBP values ​​of all pixels to obtain texture features; Step S34: convert the image from the RGB color space to the HSV color space, and extract the histograms of the three channels H, S, and V respectively; Combine the histograms of the three channels to obtain color features; Step S35: using the Canny algorithm to extract edge features in the image; Step S36: vectorize the texture features, color features and edge features, and combine them to obtain a comprehensive feature vector.

5. The method for comprehensively detecting the state of a curved screen according to claim 4, characterized in that: The steps of step S4 are as follows: all comprehensive feature vectors and corresponding category labels are divided into a training set and a test set according to a preset ratio, wherein the training set is used to train the SVM model, and the test set is used to evaluate the performance of the model.

6. A system for comprehensively detecting the state of a curved screen, characterized in that: A method for comprehensively detecting the state of a curved screen using any one of claims 1 to 5, comprising a data acquisition module, a preprocessing module, a feature extraction module, a training module and a detection module; The data acquisition module is used to collect images of the curved screen in different states; The preprocessing module is used to preprocess the image to obtain preprocessing data; The feature extraction module is used to mark the image and extract features from the preprocessed data; The training module is used to input the acquired feature data into the model for training to obtain a trained verification model; The detection module is used to identify the curved screen image using the verification model to obtain the state of the curved screen.

7. A system for comprehensively detecting the state of a curved screen according to claim 6, characterized in that: The pre-processing module includes one or more combinations of image denoising, image enhancement, and illumination correction.

8. The system for comprehensively detecting the state of a curved screen according to claim 7, characterized in that: The preprocessing module includes a denoising submodule, an enhancement submodule and a correction submodule; The denoising submodule is used to achieve denoising by convolving each pixel of the image with a Gaussian kernel; the formula for convolution denoising is as follows: in I(x, y) represents the image, x, y are the pixel coordinates of the image, σ is the standard deviation of the Gaussian distribution, and n is the size of the Gaussian kernel; The enhancement submodule is used to adjust the grayscale distribution of the image to make the grayscale histogram of the image uniform, thereby enhancing the contrast of the image; The formula for enhancing the image is as follows: S k =(L-1)CDF(r k ); (L-1) is the grayscale range of the image, Where P r (r i ) is the probability of the i-th gray level appearing; The correction submodule is used to process the grayscale value of the image through the reflection component and the illumination component of the reflected light to obtain a first image; Taking the logarithm of the first image to obtain a first parameter; Performing Fourier transform on the first parameter to obtain a second parameter; Design a filter H(u,v) to filter the second parameter, separate and adjust the reflection component and the illumination component: H(u,v) = H L (u,v)+H H (u,v); The filtered image is then inverse Fourier transformed: lnI f (x,y)=F -1 [H(u,v)F[lnI(x,y)]]; Finally, the corrected image is obtained through exponential operation:

9. The system for comprehensively detecting the state of a curved screen according to claim 6, characterized in that: The feature extraction module includes a first submodule, a second submodule, a texture extraction submodule, an edge feature extraction submodule, a color feature extraction submodule and a combination submodule; The first submodule is used to take one of the pixels as the central pixel g c , construct a 3*3 neighborhood and compare the grayscale values ​​of the 8 pixels in the neighborhood with the central pixel g c The size between them is used to obtain its LBP value; The second submodule is used to replace the next pixel as the new center pixel and re-call the first submodule until the LBP values ​​of all pixels of the image are extracted; The texture extraction submodule is used to count the histogram of LBP values ​​of all pixels to obtain texture features; The color feature extraction submodule is used to convert the image from the RGB color space to the HSV color space, and extract the histograms of the three channels of H, S, and V respectively; Combine the histograms of the three channels to obtain color features; The edge feature extraction submodule is used to extract edge features in the image using the Canny algorithm; The combination submodule is used to vectorize texture features, color features and edge features, and combine them to obtain a comprehensive feature vector.

Images