Coal mine underground image enhancement method and system, computer equipment and storage medium

Abstract

The invention belongs to the technical field of underground coal mine image enhancement, and particularly discloses an underground coal mine image enhancement method and system, computer equipment and a storage medium. The method comprises the following steps: firstly, establishing a glow imaging degradation model to obtain a glow image, and removing glow by adopting a layer separation method to obtain a glow-removed image; secondly, on the basis of a low-illumination enhancement SRLLIE method, an objective function of the low-illumination enhancement SRLLIE method is improved according to the glow-removed image, iterative solution is carried out on the improved objective function, and an illumination image which suppresses overexposure and retains structural details and a denoised reflection image are obtained; adjusting the brightness of the illuminance diagram by using an S-type gamma correction function to obtain an optimized illuminance diagram; and finally, according to a Retinex theory, carrying out point multiplication on the optimized illumination diagram and the reflection diagram to obtain an enhanced image. According to the method, the brightness of the image can be effectively improved, excessive enhancement is inhibited, the detail information of the image can be recovered, and the visual effect of the image is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of coal mine underground image enhancement, and in particular relates to a coal mine underground image enhancement method, system, computer equipment and storage medium. Background Art

[0002] As an important source of underground information, underground video images, with a large number of images and video technologies deeply integrated into the entire process of coal mining, can provide a solid data basis for intelligent monitoring of underground coal mines. However, due to the absorption and multiple scattering of light from artificial light sources by floating coal dust particles in underground coal mines, a glow phenomenon appears in the image, resulting in uneven illumination of underground images. In addition, due to insufficient light and poor imaging environment in coal mines, video images show degradation phenomena such as low illumination and high noise, which hinders the identification of key information, thereby hindering the real-time detection of mine workers, the monitoring of equipment operation status, and the precise decision-making of intelligent systems. These have limited the application of digital video technology in coal mine production.

[0003] Ying Dongjie and others used wavelet transform technology to decompose and reconstruct underground coal mine images to enhance the image details. Fan Lingyun and others performed histogram equalization and wavelet transform on underground images, and then performed matching processing and inverse wavelet transform again to enhance the image. However, this type of image enhancement method dominated by wavelet transform does not fully address the enhancement of low-frequency information, making the image information unable to be effectively displayed, and fails to significantly improve the overall image brightness.

[0004] Zhi et al. used the concept of dual gamma function and fine-tuned it according to the distribution characteristics of the illumination map to improve the grayscale level of the low-light area and weaken the grayscale influence of the local strong light area. Yu et al. used the inverse tangent hyperbola and optimized hyperbolic tangent profile to map the hue components and enhance the low-light image. In addition, Park et al. proposed a two-layer autoencoder network model based on Retinex theory (an image processing theory), which combines stacked autoencoders and convolutional autoencoders to achieve low-light enhancement and noise reduction. Kimmel et al. used a priori assumptions to propose a Retinex image enhancement method based on a variational framework, which transforms the illumination estimation problem into an optimal quadratic optimization problem. Although the complexity is high, the effect is significant. Elad et al. proposed a non-iterative Retinex image enhancement method to effectively control the edges in the image and suppress the noise in the dark area.

[0005] Although the above methods have achieved certain results in image enhancement, due to the problems of low and uneven illumination and high noise in coal mine underground images, the processing effect of the existing image enhancement methods is not ideal. Summary of the invention

[0006] In view of the problems of low and uneven illumination and high noise in underground coal mine images, the present invention proposes a method for underground coal mine image enhancement. The method improves the objective function of the low illumination enhancement SRLLIE method by utilizing the de-glow image, and can effectively enhance and denoise the low illumination image while suppressing overexposure, so that the obtained enhanced image can effectively improve the brightness of the image, suppress over-enhancement, and restore the image detail information, thereby improving the image visual effect.

[0007] In order to achieve the above object, the present invention adopts the following technical scheme:

[0008] Step 1. Establish a glow imaging degradation model to obtain a glow image, use a layer separation method to remove the glow, suppress overexposure, and obtain a de-glow image;

[0009] Step 2. Based on the low-illumination enhancement SRLLIE method, the objective function of the low-illumination enhancement SRLLIE method is improved according to the de-glow image, and the improved objective function is iteratively solved to obtain the illumination map and denoised reflectance map that suppress overexposure and retain structural details;

[0010] Step 3. Use the S-type gamma correction function to adjust the brightness of the illumination map to obtain an optimized illumination map;

[0011] Step 4. According to the Retinex theory, multiply the optimized illumination map with the reflectance map to obtain the enhanced image.

[0012] In addition, based on the coal mine underground image enhancement method, the present invention also proposes a coal mine underground image enhancement system adapted thereto, and its technical solution is as follows:

[0013] An underground coal mine image enhancement system, comprising:

[0014] Image de-glow module, used to establish a glow imaging degradation model to obtain a glow image, and use a layer separation method to remove the glow to avoid overexposure problems and obtain a de-glow image;

[0015] The improved SRLLIE module is used to improve the objective function of the low-illumination enhancement SRLLIE method based on the de-glow image, iteratively solve the improved objective function, and obtain the illumination map and denoised reflectance map that suppress overexposure and retain structural details;

[0016] An illumination map optimization module is used to adjust the brightness of the illumination map using an S-type gamma correction function to obtain an optimized illumination map;

[0017] The image enhancement module is used to obtain an enhanced image by multiplying the optimized illumination map with the reflectance map according to the Retinex theory.

[0018] In addition, based on the above-mentioned coal mine underground image enhancement method, the present invention also proposes a computer device, which includes a memory and one or more processors;

[0019] The memory stores executable codes, and when the processor executes the executable codes, the executable codes are used to implement the steps of the above-mentioned coal mine underground image enhancement method.

[0020] In addition, based on the above-mentioned coal mine underground image enhancement method, the present invention also proposes a computer-readable storage medium on which a program is stored; when the program is executed by a processor, it is used to implement the steps of the above-mentioned coal mine underground image enhancement method.

[0021] The present invention has the following advantages:

[0022] As described above, the present invention relates to a method for enhancing images underground in coal mines. The method first establishes a glow imaging degradation model to obtain a glow image, and adopts a layer separation method to subtract the input image from the glow image to remove the glow. The obtained de-glow image suppresses overexposure and improves the clarity and recognition of the image; then, on the basis of the original low-illumination enhancement SRLLIE method, its objective function is improved according to the de-glow image, and then the alternating direction multiplier method is used to iteratively solve the improved objective function to obtain an illumination map and a denoised reflection map that suppresses overexposure and retains structural details, thereby avoiding overexposure, improving the problem of uneven illumination of images underground in coal mines, and reducing noise interference; then, the brightness of the illumination map is adjusted using an S-type gamma correction function, and the obtained optimized illumination map can not only improve the brightness of the dark area, but also avoid excessive enhancement of the over-bright area; finally, according to the Retinex theory, the optimized illumination map is dot-multiplied with the reflection map to obtain an enhanced image. The method of the present invention can not only effectively improve the brightness of the image and suppress over-enhancement, but also restore the image detail information, improve the image visual effect, and effectively solve the problems of low and uneven illumination and high noise in coal mine underground images. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Schematic diagram of the process of the coal mine underground image enhancement method in an embodiment of the present invention;

[0024] Figure 2 It is a flowchart of the coal mine underground image enhancement method in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0026] Example 1

[0027] In view of the problems of low and uneven illumination and high noise in underground coal mine images, this embodiment describes a method for enhancing underground coal mine images. The general inventive concept of the method is as follows: first, the glow is removed through a light source (glow) model to avoid overexposure and obtain a de-glow image; then, on the basis of the low illumination enhancement SRLLIE algorithm, its objective function is improved according to the de-glow image, and then the alternating direction multiplier method is used for iterative solution to obtain an illumination map and a denoised reflection map that suppresses overexposure and retains structural details; then, the S-type gamma correction function is used to adjust the brightness of the illumination map to obtain an optimized illumination map; finally, according to the Retinex theory, the optimized illumination map is dot-multiplied with the reflection map to obtain an enhanced image.

[0028] like Figure 1 to Figure 2 As shown, the coal mine underground image enhancement method in this embodiment includes the following steps:

[0029] Step 1. Establish a glow imaging degradation model to obtain a glow image, use a layer separation method to remove the glow, suppress overexposure, and obtain a de-glow image.

[0030] The light from the light source is scattered multiple times by fine particles, thus forming a glow, which will cause overexposure near the light source in the image. This glow is modeled as an atmospheric point spread function. Based on this, the glow imaging degradation model is established as:

[0031] I=J+L a *APSF (1)

[0032] Where I is the input image, J is the de-glow image; L a *APSF is for glow images G, L a is the light source, * represents the convolution operation, and APSF represents the atmospheric point spread function.

[0033] A glow image is obtained based on the glow imaging degradation model of formula (1), which is used to suppress overexposure and improve the clarity and recognition of the image.

[0034] Due to the multiple scattering of light by fine particles, the brightness of the glow gradually decreases smoothly, and the gradient histogram of the glow shows the characteristics of a "short-tail" distribution. This property is used to remove the glow using the layer separation method. The objective function of the layer separation is:

[0035]

[0036] Among them, E(J) represents the objective function for layer separation, f 1,2is the first-order derivative filter in the horizontal and vertical directions, f3 is the second-order Laplace filter; ρ(s) is a robust function that can remove the glow image and retain the information with larger gradient in the input image, ρ(s) = min(s 2 ,τ) where s is J(x,y)*f 1,2 , τ is a fixed parameter; λ is the smoothness weight parameter, J r represents the r channel of the de-glow image, J g represents the g channel of the de-glow image, J b Represents the b channel of the deflluminated image.

[0037] The inequality constraint in formula (2) ensures that the solution J of the objective function is within an appropriate range (ie, ensures that J is greater than 0 and less than 1), and the equality constraint in formula (2) can solve the color shift problem well.

[0038] Solve the objective function of the above formula (2) to obtain the de-glow image.

[0039] Since there are a lot of coal dust particles floating in the air in coal mines, and the underground lighting relies only on artificial light sources, these suspended particles scatter the light emitted by the light source multiple times, causing a glow phenomenon near the image light source, which in turn causes the image to be overexposed. Therefore, the present invention uses the glow model to remove the glow, suppress overexposure, and improve the clarity and recognition of the image.

[0040] Step 2. Based on the low-illumination enhancement SRLLIE method, the objective function of the low-illumination enhancement SRLLIE method is improved according to the de-glow image obtained in step 1, and the improved objective function is iteratively solved to obtain the illumination map and denoised reflection map that suppress overexposure and retain structural details, thereby improving the uneven illumination problem of underground coal mine images, avoiding overexposure, and reducing noise interference.

[0041] SRLLIE is a low-light image enhancement method with good performance. However, when processing images with uneven illumination in coal mines, it will cause overexposure near the near light point. In view of this, the present invention proposes an underground coal mine image enhancement algorithm based on the glow model to improve SRLLIE on the basis of the original SRLLIE algorithm. That is, the input image in the objective function of the original SRLLIE algorithm is replaced with the de-glow image obtained in step 1. The improved objective function formula is as follows:

[0042]

[0043] Where I represents the input image, R represents the reflectance map, L represents the illumination map, G represents the glow image, and N represents the noise map; β, ω, and δ are coefficients; ||·|| Fand ||·||1 represent the F norm and L1 norm respectively. is a first-order differential operator and P is the adjusted gradient of the input image.

[0044] According to the improved objective function of the low-light enhancement SRLLIE method, the iterative solution is performed through the alternating direction multiplier method. The specific solution process is as follows:

[0045] First, we introduce the auxiliary variable T to replace the second term of the objective function in formula (3). The objective function is transformed into the following form:

[0046]

[0047] By introducing the Lagrangian multiplier Z to eliminate the equality constraint, the augmented Lagrangian function is obtained:

[0048]

[0049] in, A is Z, B is <*,*> represents matrix inner product, and μ represents penalty scalar.

[0050] The first term on the right side of formula (5) is the fidelity term; the second term on the right side of the equation in formula (5) ||T||1 is the smoothness term, which takes into account the smoothness of the illumination map L; the third term on the right side of the equation in formula (5) Constrain the fidelity between the gradient of the reflection map and the gradient of the input image, and strengthen the structural information of the reflection map; the fourth term on the right side of the formula (5) Limit the overall intensity of the noise, The fidelity term of the noise map is used to ensure the accuracy of the model. It is expected that the estimated illumination map, reflectance map, and noise map can accurately reconstruct the input image.

[0051] The objective function is solved by step-by-step iteration. The following is the calculation process of the k-th iteration solution of each sub-problem:

[0052] R subproblem: Ignore the terms in formula (5) that are not related to R and obtain the following objective function:

[0053]

[0054] Among them, L (k) represents the kth iteration result of L, N (k) represents the k-th iteration result of N; by reformulating the objective function in formula (6), it becomes a classic least squares problem:

[0055]

[0056] Among them, l, r, n, i, g, and p are the vectorization of L, R, N, I, G, and P respectively. represents a diagonal matrix with l as its element, express The k-th iteration result, n (k) Represents the k-th iteration result of n.

[0057] By differentiating r in formula (7), considering the variables other than r in formula (7) as constants, and setting the derivative with respect to r to 0, the following formula is obtained:

[0058]

[0059] Where D is the discrete gradient operator, f(x) = x T x, r (k+1) Represents the k+1th iteration result of r.

[0060] L subproblem: Ignore the terms in formula (5) that are not related to L and obtain the following objective function:

[0061]

[0062] Among them, R (k+1) represents the k+1th iteration result of R, Z (k) represents the kth iteration result of Z, T (k) represents the k-th iteration result of T; the solution of the k+1-th iteration of L is as follows:

[0063]

[0064] Among them, t is the vectorization of T, z is the vectorization of Z, express The k+1th iteration result of represents a diagonal matrix with l as its element, t (k) represents the kth iteration result of t, z (k) Represents the k-th iteration result of z.

[0065] N subproblems: Ignore the terms in formula (5) that are not related to N and obtain the following objective function:

[0066]

[0067] The solution for the k+1th iteration of N is:

[0068]

[0069] T subproblem: Ignore the terms in formula (5) that are not related to T and obtain the following objective function:

[0070]

[0071] Among them, Z (k) represents the k-th iteration result of Z. The solution of the k+1-th iteration of T is:

[0072]

[0073] Among them, S ε (x)=sign(x)max(|x|-ε,0), ε is The calculation is performed element by element; μ (k) Represents the kth iteration result of μ.

[0074] The auxiliary matrix Z and the penalty scalar μ are updated as follows:

[0075]

[0076] Where h is a constant and h>1. Only when R (k) and R (k+1) The difference between (k) and L (k+1) The entire iteration stops when the difference between them is less than a preset critical value or the maximum number of iterations is reached.

[0077] In this embodiment, h is set to 1.5, for example, and the threshold value is preset to 10 -3 , the maximum number of iterations is set to 30.

[0078] After the iteration, an illumination map and a denoised reflectance map that suppress overexposure and retain structural details are obtained.

[0079] The present invention improves the objective function of the original SRLLIE algorithm according to the de-glow image on the basis of the original SRLLIE algorithm, obtains the illumination map and denoised reflection map with the functions of suppressing overexposure and retaining structural details, avoids overexposure, improves the problem of uneven illumination of images in coal mines, and also reduces noise interference.

[0080] Step 3. Use the S-type gamma correction function to adjust the brightness of the illumination map to obtain the optimized illumination map.

[0081] The illuminance map obtained by iterative solution is still low in brightness, so the brightness of the illuminance map is adjusted through the S-type gamma correction function:

[0082]

[0083] Among them, L' is the optimized illumination map, and γ is the brightness adjustment factor. The optimized illumination map obtained by using the S-type gamma correction function can not only improve the brightness of the dark area, but also avoid excessive enhancement of the overly bright area.

[0084] Step 4. According to the Retinex theory, multiply the optimized illumination map with the reflectance map to obtain the enhanced image.

[0085] According to the Retinex theory, the optimized illumination map is multiplied by the reflectance map to obtain the enhanced image:

[0086] I′=R·L′ (17)

[0087] Where I′ represents the enhanced image.

[0088] In addition, in order to verify the effectiveness of the image enhancement method proposed in the present invention, the following specific experiments are also given:

[0089] Since there is glow near the light source in the illumination map obtained by the SRLLIE image enhancement method, this will eventually lead to overexposure in the enhanced image, reducing the visibility of the area. The image enhancement method proposed in the present invention improves the original objective function of SRLLIE based on the de-glow image, and the illumination map obtained reduces the glow near the light source, thereby suppressing the overexposure problem.

[0090] The experiment selected three coal mine underground images Ⅰ, Ⅱ, and Ⅲ in the data set, and conducted experimental comparative analysis on the image enhancement method proposed in the present invention, the LR3M image enhancement method, the BioEnh image enhancement method, and the SRLLIE image enhancement method.

[0091] Among them, the LR3M image enhancement method is an image enhancement method based on the low-rank regularized Retinex model to estimate the illumination map and reflectance map, and perform low-illumination image enhancement. The BioEnh image enhancement method proposes a dual-channel model, namely the structure channel and the detail channel, decomposes it based on the total variation, improves the brightness and suppresses the noise of the structure channel and the detail channel respectively, and finally integrates the two channels to obtain an enhanced image. The SRLLIE image enhancement method is an image enhancement method based on the robust Retinex model to estimate the illumination map and reflectance map, and then obtain an enhanced image.

[0092] In the experimental results, the image enhancement method proposed by the present invention and the three compared image enhancement methods can achieve the enhancement of coal mine underground images to varying degrees. Although the LR3M image enhancement method can improve the brightness of dark areas, it does not eliminate or suppress the glow around the light source, resulting in overexposure problems in areas near the light source. The BioEnh image enhancement method enhances the overall brightness of the image and restores the detail information of the low-illuminance area, but it also has overexposure problems near the light source, resulting in serious loss of image texture details. The overexposure problem of the BioEnh image enhancement method is more obvious than that of other image enhancement methods. The SRLLIE image enhancement method has the function of significantly enhancing the brightness of low-illuminance images and eliminating the influence of noise, but it will also cause overexposure at the near-light point of the image, seriously reducing the visibility of the scene near the light source.

[0093] Compared with the other three traditional image enhancement methods, the image enhancement method of the present invention can effectively improve the brightness of dark areas and restore image details and edge information. At the same time, it can effectively weaken the influence of glow and avoid overexposure problems in high-brightness areas, thereby enhancing image contrast and improving the overall visual effect.

[0094] In addition, in order to further verify the effectiveness of the image enhancement method proposed in the present invention, three non-reference indicators, namely information entropy, natural image quality evaluation NIQE and average gradient (AG), were used to experimentally evaluate the image enhancement method proposed in the present invention and the LR3M image enhancement method, BioEnh image enhancement method and SRLLIE image enhancement method.

[0095] Among them, information entropy is used to evaluate the information content of an image. When the information entropy value is large, it means that the image contains richer visual information; natural image quality assessment evaluates the quality of an image by analyzing the naturalness and distortion of the image. The lower the natural image quality assessment value, the better the visual effect of the image; the average gradient is a key indicator for measuring the difference in local details and texture change characteristics of an image. It can effectively reflect the overall quality of the image and, to a certain extent, the clarity of the image. The higher the average gradient value, the more image details and the higher the clarity.

[0096] The objective quality evaluation results are shown in Table 1. The ones with the highest information entropy and average gradient values ​​in the evaluation indicators are displayed in bold, and the ones with the second highest values ​​are displayed in an inclined manner; the ones with the lowest natural image quality evaluation values ​​are displayed in bold, and the ones with the second highest values ​​are displayed in an inclined manner.

[0097] Table 1

[0098]

[0099]

[0100] As shown in Table 1, for low-illuminance images in coal mines, the four image enhancement methods can effectively enhance the overall illumination of the image. Among them, the LR3M image enhancement method has higher Entropy and AG scores, which means that the image enhancement method can effectively enhance the brightness of low-illuminance areas and restore detail information, but the image enhancement method also has a high NIQE score, which means that the image enhancement method has a certain degree of distortion. The BioEnh image enhancement method is the only image enhancement method that has the highest Entropy and NIQE scores among the four image enhancement methods. This method can enhance the brightness of the entire image and restore the texture edge information of the low-illuminance area, but the image will be overexposed and distorted in some areas. The SRLLIE image enhancement method, with its superior Entropy and AG scores, fully demonstrates the role of this image enhancement method in improving the brightness of the image and restoring the image detail information, making the image present a higher clarity, but it will also cause overexposure at the near light point of the image. The image enhancement method proposed in the present invention has high entropy, the highest AG score and the lowest NIQE score, which means that the image enhancement method proposed in the present invention can not only effectively improve the brightness of the image and suppress over-enhancement, but also restore the image detail information and improve the image visual effect. The above experiments verify the effectiveness of the method proposed in the present invention.

[0101] Example 2

[0102] This embodiment 2 describes an underground coal mine image enhancement system, which is based on the same inventive concept as the underground coal mine image enhancement method in embodiment 1.

[0103] Specifically, the coal mine underground image enhancement system includes the following modules:

[0104] Image de-glow module, used to establish a glow imaging degradation model to obtain a glow image, and use a layer separation method to remove the glow to avoid overexposure problems and obtain a de-glow image;

[0105] The improved SRLLIE module is used to improve the objective function of the low-illumination enhancement SRLLIE method based on the de-glow image on the basis of the low-illumination enhancement SRLLIE image enhancement method, iteratively solve the improved objective function, and obtain the illumination map and denoised reflectance map that suppress overexposure and retain structural details;

[0106] An illumination map optimization module is used to adjust the brightness of the illumination map using an S-type gamma correction function to obtain an optimized illumination map;

[0107] The image enhancement module is used to obtain an enhanced image by multiplying the optimized illumination map with the reflectance map according to the Retinex theory.

[0108] It should be noted that, in the coal mine underground image enhancement system, the implementation process of the functions and effects of each functional module is specifically described in the implementation process of the corresponding steps in the method in Example 1, and will not be repeated here.

[0109] Example 3

[0110] This embodiment 3 describes a computer device, which includes a memory and one or more processors.

[0111] An executable code is stored in the memory, and when the processor executes the executable code, it is used to implement the steps of the coal mine underground image enhancement method in the above-mentioned embodiment 1.

[0112] In this embodiment, the computer device is any device or apparatus with data processing capability, which will not be described in detail here.

[0113] Example 4

[0114] This embodiment 4 describes a computer-readable storage medium on which a program is stored. When the program is executed by a processor, it is used to implement the steps of the coal mine underground image enhancement method.

[0115] The computer-readable storage medium can be an internal storage unit of any device or apparatus with data processing capabilities, such as a hard disk or memory, or an external storage device of any device with data processing capabilities, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), an SD card, a flash card, etc. equipped on the device.

[0116] Of course, the above description is only a preferred embodiment of the present invention, and the present invention is not limited to the above embodiments. It should be noted that all equivalent substitutions and obvious deformation forms made by any technician familiar with the field under the guidance of this specification fall within the essential scope of this specification and should be protected by the present invention.

Claims

1. A method for enhancing an image in an underground coal mine, characterized in that: The steps include: Step 1. Establish a glow imaging degradation model to obtain a glow image, use a layer separation method to remove the glow, suppress overexposure, and obtain a de-glow image; Step 2. Based on the low-illumination enhancement SRLLIE method, the objective function of the low-illumination enhancement SRLLIE method is improved according to the de-glow image, and the improved objective function is iteratively solved to obtain the illumination map and denoised reflectance map that suppress overexposure and retain structural details; Step 3. Use the S-type gamma correction function to adjust the brightness of the illumination map to obtain an optimized illumination map; Step 4. According to the Retinex theory, multiply the optimized illumination map with the reflectance map to obtain the enhanced image.

2. The coal mine underground image enhancement method according to claim 1, characterized in that: The step 1 is specifically as follows: A glow imaging degradation model is established and a glow image is obtained based on the glow imaging degradation model. The glow imaging degradation model is: I=J+L a *APSF (1) Where I is the input image, J is the de-glow image; L a *APSF represents the glow image G, L a is the light source, * indicates the convolution operation, and APSF indicates the atmospheric point spread function; The glow is removed by layer separation method, and the objective function of layer separation is: Among them, E(J) represents the objective function for layer separation, f 1,2 is the first-order derivative filter in the horizontal and vertical directions, f3 is the second-order Laplace filter; ρ(s) is the robust function ρ(s)=min(s 2 ,τ), where s is J*f 1,2 , τ is a fixed parameter; λ is the smoothness weight parameter, J r represents the r channel of the de-glow image, J g represents the g channel of the de-glow image, J b represents the b channel of the de-glow image; Solve the objective function in the above formula (2) to obtain the de-glow image.

3. The coal mine underground image enhancement method according to claim 2, characterized in that: The step 2 is specifically as follows: The improved objective function formula of the low-light enhancement SRLLIE method is as follows: Where R represents the reflectance map, L represents the illumination map, and N represents the noise map; β, ω, and δ are coefficients; ||·|| F and ||·||1 represent the F norm and L1 norm respectively. is a first-order differential operator and P is the adjusted gradient of the input image.

4. The coal mine underground image enhancement method according to claim 3, characterized in that: In the step 2, according to the improved objective function of the low illumination enhancement SRLLIE method, an iterative solution is performed by an alternating direction multiplier method; Introduce the auxiliary variable T to replace the second term of the objective function Then the objective function in formula (3) is transformed into the following form: By introducing the Lagrangian multiplier Z to eliminate the equality constraint, the augmented Lagrangian function is obtained: in, A is Z, B is <*,*> represents the inner product of the matrix, μ represents the penalty scalar; the objective function is solved by step-by-step iteration. The following is the calculation process of the k-th iteration solution of each subproblem: R subproblem: Ignore the terms in formula (5) that are not related to R and obtain the following objective function: Among them, L (k) represents the kth iteration result of L, N (k) represents the k-th iteration result of N; by reformulating the objective function in formula (6), it becomes a classic least squares problem: Among them, l, r, n, i, g, and p are the vectorization of L, R, N, I, G, and P respectively. represents a diagonal matrix with l as its element, express The kth iteration result, n (k) represents the k-th iteration result of n; by differentiating r in formula (7), considering the variables other than r in formula (7) as constants, and setting the derivative with respect to r to 0, the following formula is obtained: Where D is the discrete gradient operator, f(x) = x T x, r (k+1) represents the k+1th iteration result of r; L subproblem: Ignore the terms in formula (5) that are not related to L and obtain the following objective function: Among them, R (k+1) represents the k+1th iteration result of R, Z (k) represents the kth iteration result of Z, T (k) represents the k-th iteration result of T; the solution of the k+1-th iteration of L is as follows: Among them, t is the vectorization of T, z is the vectorization of Z, express The k+1th iteration result of express The k-th iteration result, t (k) represents the kth iteration result of t, z (k) represents the kth iteration result of z; N subproblems: Ignore the terms in formula (5) that are not related to N and obtain the following objective function: The solution for the k+1th iteration of N is: N (k+1) =(I-G-R (k+1) ·L (k+1) ) / (1+δ) (12) T subproblem: Ignore the terms in formula (5) that are not related to T and obtain the following objective function: The solution of the k+1th iteration of T is: Among them, S ε (x)=sign(x)max(|x|-ε,0), ε is The calculation is performed element by element; μ (k) represents the k-th iteration result of μ; The auxiliary matrix Z and the penalty scalar μ are updated as follows: Among them, h is a constant and h>1; only when R (k) and R (k+1) The difference between (k) and L (k+1) The entire iteration stops when the difference between them is less than a preset critical value or the maximum number of iterations is reached; After the iteration, an illumination map and a denoised reflectance map that suppress overexposure and retain structural details are obtained.

5. The coal mine underground image enhancement method according to claim 4, characterized in that: In step 3, the brightness of the illumination image is adjusted by an S-type gamma correction function: Among them, L′ is the optimized illumination map, and γ is the brightness adjustment factor.

6. The coal mine underground image enhancement method according to claim 5, characterized in that: The step 4 is specifically as follows: According to the Retinex theory, the optimized illumination map is multiplied by the reflectance map to obtain the enhanced image: I′=R·L′ (17) Where I′ represents the enhanced image.

7. An underground coal mine image enhancement system, characterized in that: include: Image de-glow module, used to establish a glow imaging degradation model to obtain a glow image, and use a layer separation method to remove the glow to avoid overexposure problems and obtain a de-glow image; The improved SRLLIE module is used to improve the objective function of the low-illumination enhancement SRLLIE method based on the de-glow image, iteratively solve the improved objective function, and obtain the illumination map and denoised reflectance map that suppress overexposure and retain structural details; An illumination map optimization module is used to adjust the brightness of the illumination map using an S-type gamma correction function to obtain an optimized illumination map; The image enhancement module is used to obtain an enhanced image by multiplying the optimized illumination map with the reflectance map according to the Retinex theory.

8. A computer device comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that: When the processor executes the executable code, the steps of the coal mine underground image enhancement method as described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the coal mine underground image enhancement method as described in any one of claims 1 to 6 are implemented.

Images