A deep learning-based coal slime foam image enhancement method and system

Abstract

The application discloses a coal slime foam image enhancement method and system based on deep learning. First, a coal slime foam image is collected; then, a deblurring proportion a and a brightness enhancement proportion b are calculated; then, a fusion parameter c is calculated through a formula c=a*X+b*Y; finally, the fusion parameter c is loaded into a preset image processing model, and an enhanced foam image is output. The application enhances the coal slime foam image based on deep learning, simplifies the foam image processing program, is simple and convenient to operate, saves the time and manpower spent on screening and processing of the foam image dataset, improves the utilization rate of the foam image dataset taken from the production site, and has important significance in improving resource utilization and economic benefits of coal preparation plants.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of froth flotation, and particularly relates to a coal slime froth image enhancement method and system based on deep learning. BACKGROUND

[0002] Froth flotation is a method of sorting minerals according to the difference in wettability of the mineral surface, and is one of the most effective methods for sorting clean coal in a coal preparation plant. The surface visual features such as the fineness, shape and size of the froth texture are closely related to the froth working condition. Since there are many factors affecting the flotation production, the production parameters are mainly adjusted by manually observing the visual features of the froth surface with the naked eye. However, the production mode of adjusting the production parameters according to the naked eye observation is easily affected by subjective factors, which may cause fluctuations in the flotation production indexes and large consumption of reagents.

[0003] Therefore, it is of great significance to use machine vision based on deep learning to monitor the froth in the flotation production process, so as to improve resource utilization and economic benefits of the coal preparation plant. SUMMARY

[0004] The present application provides a coal slime froth image enhancement method and system based on deep learning, which can enhance the coal slime froth image based on deep learning.

[0005] The present application provides a coal slime froth image enhancement method based on deep learning, comprising:

[0006] Collecting a coal slime froth image;

[0007] calculating a deblurring proportion a by the formula a=X / Xmax; wherein X is a variance of the coal slime froth image, and Xmax is a variance threshold; d D calculating a brightness enhancement proportion b by the formula b=Y / Ymax; wherein Y is a mean pixel value percentage of the coal slime froth image, and Ymax is a mean pixel percentage threshold; h H

[0008] calculating a fusion parameter c by the formula c=a*X+b*Y; wherein X is a deblurring parameter, and Y is a brightness enhancement parameter; loading the fusion parameter c into a preset image processing model to output an enhanced froth image.

[0009]

[0010] Specifically, the deblurring proportion a is calculated by the formula a=X / Xmax, comprising:

[0011] calculating a deblurring proportion a by the formula a=X / Xmax; wherein X is a variance of the coal slime froth image, and Xmax is a variance threshold; ​​​​​​

[0012] The variance x of the slime foam image is compared with the variance threshold value x d . D .

[0013] If x d is less than x D , the deblurring proportion a is calculated by the formula .

[0014] Specifically, the variance threshold value x D is calculated by the formula ; wherein n is the total number of the slime foam images, xi is the variance of the i-th image.

[0015] Specifically, the brightness enhancement proportion b is calculated by the formula , which comprises:

[0016] The average pixel value percentage x of the slime foam image is compared with the average pixel percentage threshold value x h . H

[0017] If x h is less than x H , the brightness enhancement proportion b is calculated by the formula .

[0018] Specifically, the average pixel percentage threshold value x H is calculated by the formula ; wherein is the average pixel percentage of the i-th image.

[0019] The application further provides a slime foam image enhancement system based on deep learning, comprising:

[0020] A slime foam image acquisition module is configured to acquire a slime foam image.

[0021] A deblurring proportion calculation module is configured to calculate a deblurring proportion a by the formula ; wherein x is the variance of the slime foam image, x is a variance threshold value. d D

[0022] A brightness enhancement proportion calculation module is configured to calculate a brightness enhancement proportion b by the formula ; wherein x is the average pixel value percentage of the slime foam image, x is an average pixel percentage threshold value. h H

[0023] ​​​​​​The fusion parameter calculation module is used to calculate the fusion parameter c using the formula c = a*X + b*Y; where X is the deblurring parameter and Y is the brightness enhancement parameter.

[0024] The image processing module is used to load the fusion parameter c into a preset image processing model and output an enhanced foam image.

[0025] Specifically, the deblurring weight calculation module includes:

[0026] The variance determination unit is used to determine the variance x of the coal slime foam image. d With the variance threshold x D Compare;

[0027] Defuzzy weight calculation unit, used if x d Less than x D Through formula The deblurring ratio 'a' is calculated.

[0028] Specifically, the deblurring weight calculation module further includes:

[0029] The variance threshold calculation unit is used to calculate the variance threshold using the formula. The variance threshold x is calculated. D Where n is the total number of the coal slime foam images, Let be the variance of the i-th image.

[0030] Specifically, the brightness enhancement ratio calculation module includes:

[0031] The average pixel percentage determination unit is used to calculate the average pixel value percentage of the coal slime foam image by multiplying it by the average pixel value percentage. h With the average pixel percentage threshold x H Compare;

[0032] Brightness enhancement weight calculation unit, used if x h Less than x H Through formula The brightness enhancement ratio b was calculated.

[0033] Specifically, the brightness enhancement ratio calculation module further includes:

[0034] The average pixel percentage threshold calculation unit is used to calculate the average pixel percentage threshold using the formula. The average pixel percentage threshold x is calculated. H ;in, is the average pixel percentage of the i-th image.

[0035] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0036] First, the coal slime foam image is collected; then, the deblurring ratio a and the brightness enhancement ratio b are calculated; then, the fusion parameter c is calculated through the formula c=a*X+b* Y; finally, the fusion parameter c is loaded into the preset image processing model, and the enhanced foam image is output. The coal slime foam image is enhanced based on deep learning, which simplifies the foam image processing program, is simple and convenient to operate, saves the time and manpower spent on screening and processing of the foam image dataset, improves the utilization rate of the foam image dataset taken from the production site, and has important significance for improving resource utilization and improving the economic benefits of the coal preparation plant. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The flowchart of the coal slime foam image enhancement method based on deep learning provided by the embodiment of the present application is shown in the figure.

[0038] Figure 2 The principle diagram of the coal slime foam image enhancement method based on deep learning provided by the embodiment of the present application is shown in the figure.

[0039] Figure 3 The structural diagram of the deblurring network, the brightness enhancement network and the image processing model in the embodiment of the present application is shown in the figure.

[0040] Figure 4 The module diagram of the coal slime foam image enhancement system based on deep learning provided by the embodiment of the present application is shown in the figure.

[0041] Figure 5 The structural diagram of the foam image data acquisition system is shown in the figure.

[0042] Figure 6 The contrast diagram of the visual effect of the enhanced image by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0043] The embodiment of the present application provides a coal slime foam image enhancement method and system based on deep learning, which can enhance the coal slime foam image based on deep learning.

[0044] The technical scheme in the embodiment of the present application is as follows to achieve the above technical effects:

[0045] The variance of the foam image dataset is calculated, and the variance value is set as the variance threshold. The average pixel value of the foam image dataset is calculated, and the percentage of the value in all pixels is set as the average pixel percentage threshold. The variance of a single foam image is compared with the variance threshold. If the variance value of the image is lower than the variance threshold, it is considered that the image is blurred and needs to be deblurred, and the proportion a of the image that needs to be deblurred is calculated. If the variance value of the image is greater than or equal to the variance threshold, it is considered that the image does not need to be deblurred. The average pixel of a single foam image is calculated, and the percentage of the pixel value in the image is calculated, which is recorded as the average pixel percentage of the image, and compared with the average pixel percentage threshold. If the average pixel percentage of the image is lower than the average pixel percentage threshold, it is considered that the image brightness is low and needs to be enhanced, and the proportion b of the image that needs to be enhanced is calculated. If the average pixel percentage of the image is greater than or equal to the average pixel percentage threshold, it is considered that the image does not need to be enhanced. The fusion parameter c is calculated through the proportions a and b, and the image is enhanced through the image processing network model.

[0046] In order to better understand the above technical solutions, the above technical solutions will be described in detail below in combination with the drawings in the specification and specific embodiments.

[0047] Referring to Figure 1 and Figure 2 , the coal slime foam image enhancement method based on deep learning provided by the embodiment of the present application comprises:

[0048] Collecting a coal slime foam image;

[0049] The deblurring proportion a is calculated through the formula ; wherein, x d is the variance of the coal slime foam image, x D is the variance threshold;

[0050] Specifically, the deblurring proportion a is calculated through the formula , which comprises:

[0051] The variance x d of the coal slime foam image is compared with the variance threshold x D .

[0052] If x d is less than x D , it is indicated that the image needs to be deblurred, and the deblurring proportion a is calculated through the formula .

[0053] Specifically, the variance threshold x D is calculated through the formula is calculated, wherein n is the total number of coal slime foam images, x di is the variance of the i-th image. The variance x d is calculated by the formula is calculated, wherein x jk is the pixel value at the (j+1)th row and (k+1)th column of the coal slime foam image, x q is the average pixel value of each image, which is calculated by the formula , wherein j*k is the image resolution.

[0054] The brightness enhancement proportion b is calculated by the formula , wherein x h is the average pixel value percentage of the coal slime foam image, x H is the average pixel percentage threshold.

[0055] The step is specifically described, the brightness enhancement proportion b is calculated by the formula , including:

[0056] The average pixel value percentage x h of the coal slime foam image is compared with the average pixel percentage threshold x H .

[0057] If x h is less than x H , it indicates that the image needs to be processed by brightness enhancement, and the brightness enhancement proportion b is calculated by the formula .

[0058] Specifically, the average pixel percentage threshold x H is calculated by the formula , wherein is the average pixel percentage of the i-th image. The average pixel value percentage x h of the coal slime foam image is calculated by the formula , wherein the number of image set average pixel values in each image The image set average pixel value is the average pixel of the i-th image.

[0059] The fusion parameter c is calculated by the formula c=a*X+b*Y, wherein X is the deblurring parameter, and Y is the brightness enhancement parameter.

[0060] Specifically, the deblurring network is trained by using the clear image in the GOPRO. When deblurring the blurred image, the network reads the parameter file learned by the training model, estimates the possible value of each pixel in the current image and assigns the value, to obtain the deblurring parameter X. Specifically, the deblurring parameter X includes weights, biases and the like. The brightness enhancement network is trained by using the low-light normal-light image in the LOL. When enhancing the brightness of the low-light image, the network reads the parameter file learned by the training model, estimates the possible value of each pixel in the current image and assigns the value, to obtain the brightness enhancement parameter Y. Specifically, the brightness enhancement parameter Y also includes weights, biases and the like.

[0061] The fusion parameter c is loaded into the preset image processing model, and an enhanced foam image is output.

[0062] Specifically, the fusion parameter c is stored in the model parameter file in the form of a dictionary. The model parameter file includes the weights, biases and the like of the model, and each parameter value in the model parameter file is obtained by the formula c=a*X+b*Y. The image processing model loads the fusion parameter c, extracts the value corresponding to the same key in the model parameter file through the key in the model, assigns values to the weights, biases and the like required by each network layer in the model during operation, and thus performs deblurring and brightness enhancement processing on the image. Each layer of network assigns values to the blurred pixel values and low-light pixel values in the image, and adjusts them in the next layer of network, selects the most possible value layer by layer until the model operation ends, and outputs the processed image.

[0063] Referring to Figure 3 The deblurring network, the brightness enhancement network and the image processing network in the embodiment of the application adopt the same network structure. The network design adopts the architecture of a multi-scale network. In order to make full use of the detail information of the image, the resolution of the image is reduced with the increase of the network fineness through the processing of the upsampling method. The original image is subjected to a convolution network with a convolution kernel size of 3x3 and a convolution kernel number of 64 in each scale, and an activation function is only set in the convolution network layer of the last layer. The resolution of the image is unchanged in each scale, and the resolution in the next scale becomes 1 / 2 of the resolution in the previous scale.

[0064] It should be noted that the processing effect of the foam image can be adjusted by adjusting the deblurring proportion a and the brightness enhancement proportion b.

[0065] Referring to Figure 4 The coal slime foam image enhancement system based on deep learning provided by the embodiment of the application comprises:

[0066] The coal slime foam image acquisition module 100 is configured to acquire a coal slime foam image.

[0067] The deblurring specific gravity calculation module 200 is configured to calculate a deblurring specific gravity a by a formula wherein x d is a variance of the coal slime foam image, and x D is a variance threshold.

[0068] Specifically, the deblurring specific gravity calculation module 200 comprises:

[0069] The variance judgment unit is configured to compare the variance x d of the coal slime foam image with the variance threshold x D .

[0070] The deblurring specific gravity calculation unit is configured to, if x d is less than x D , indicate that the image needs to be deblurred, and calculate the deblurring specific gravity a by a formula .

[0071] The variance threshold calculation unit is configured to calculate the variance threshold x D by a formula ; wherein n is a total number of the coal slime foam images, and x is a variance of the i-th image.

[0072] The image variance calculation unit is configured to calculate the variance x d of the coal slime foam image by a formula ; wherein x jk is a pixel value at a (j+1)-th row and a (k+1)-th column of the coal slime foam image, x q is an average pixel value of each image, and the average pixel value is calculated by a formula , wherein j*k is an image resolution.

[0073] The brightness enhancement specific gravity calculation module 300 is configured to calculate a brightness enhancement specific gravity b by a formula ; wherein x h is a percentage of the average pixel value of the coal slime foam image, and x H is a percentage threshold of the average pixel.

[0074] Specifically, the brightness enhancement specific gravity calculation module 300 comprises:

[0075] The average pixel percentage judgment unit is configured to compare the percentage x h of the average pixel value of the coal slime foam image with the percentage threshold x H of the average pixel.

[0076] a luminance enhancement proportion calculation unit configured to calculate a luminance enhancement proportion b according to the formula if x h is less than x H , it indicates that the image needs to be processed by the luminance enhancement, and the luminance enhancement proportion b is calculated according to the formula .

[0077] An average pixel percentage threshold calculation unit is configured to calculate an average pixel percentage threshold x according to the formula H ; wherein, is the average pixel percentage of the i-th image.

[0078] An image average pixel value percentage calculation unit is configured to calculate an average pixel value percentage x h of the coal slime foam image according to the formula ; wherein, the average pixel value number of the image set in each image is The average pixel value of the image set is the average pixel of the i-th image.

[0079] The fusion parameter calculation module 400 is configured to calculate a fusion parameter c according to the formula c = a * X + b * Y; wherein, X is a deblurring parameter, and Y is a luminance enhancement parameter.

[0080] Specifically, the embodiment of the present application includes a deblurring network and a luminance enhancement network. The deblurring network is trained by using the blur-clear image in GOPRO. When deblurring the blur image, the network reads the parameter file learned by the training model, estimates the possible value of each pixel in the current image and assigns the value, to obtain the deblurring parameter X. Specifically, the deblurring parameter X includes weights, biases, etc. The luminance enhancement network is trained by using the low-light normal-light image in LOL. When enhancing the luminance of the low-light image, the network reads the parameter file learned by the training model, estimates the possible value of each pixel in the current image and assigns the value, to obtain the luminance enhancement parameter Y. Specifically, the luminance enhancement parameter Y also includes weights, biases, etc.

[0081] The image processing module 500 is configured to load the fusion parameter c into a preset image processing model, and output an enhanced foam image.

[0082] Specifically, the fusion parameter c is stored in the model parameter file in the form of a dictionary. The model parameter file includes the weight, bias and other parameters of the model, and each parameter value in the model parameter file is obtained by the formula c=a*X+b*Y. The image processing model loads the fusion parameter c, extracts the value corresponding to the same "key" in the model parameter file through the "key" in the model, assigns values to the weight, bias and other parameters required by each network layer in the model during operation, and thus performs deblurring and brightness enhancement processing on the image. Each layer of network assigns values to the blurred pixel values and low-light pixel values in the image, and adjusts them in the next layer of network, selects the most likely value layer by layer until the model runs out, and outputs the processed image.

[0083] In this embodiment, the deblurring network, the brightness enhancement network and the image processing network in the embodiment of the application adopt the same network structure. The network design adopts the architecture of a multi-scale network. In order to make full use of the detailed information of the image, the resolution of the image is reduced with the increase of the network fineness through the processing of the upsampling method. The original image is processed by a convolutional network with a convolution kernel size of 3x3 and a convolution kernel number of 64 in each scale, and an activation function is only set in the convolutional network layer of the last layer. The resolution of the image is unchanged in each scale, and the resolution becomes 1 / 2 of the previous scale in the next scale.

[0084] It should be noted here that the processing effect of the foam image can be adjusted by adjusting the deblurring proportion a and the brightness enhancement proportion b.

[0085] The deep learning-based coal slime foam image enhancement method and system provided by the embodiment of the application will be described in detail below through specific embodiments:

[0086] The flotation site foam image data acquisition system is shown in Figure 5 The devices used include an industrial camera, a device box, a computer and a lighting device. Foam images of No. 1 and No. 2 flotation tanks in a coal washing company in Inner Mongolia are taken every 20 minutes, a total of 8 days, and the shooting time is 11 hours per day.

[0087] The enhancement process of the foam image is as follows:

[0088] (1) Input the image set.

[0089] (2) Network state judgment is performed, if network training is performed and the input data set is a Gopro image set, the output is a parameter file of the deblurring network. If the input data set is a LoL image set, the output is a parameter file of the brightness enhancement network. Otherwise, the network operation is stopped. In this embodiment, 2103 pairs of clear and blurred images in the public data set GOPRO are used to train the deblurring network. 485 pairs of low-light normal-light images in the public data set LOL are used to train the image brightness enhancement network, and the parameter files of the deblurring network and the brightness enhancement network are obtained respectively. 1000 randomly selected coal slime foam images are used for testing.

[0090] (3) If network testing is performed, the parameter files obtained by network training are fused to obtain a parameter file, which is used as the training parameter of the image processing network model to process the image. Whether the image deblurring processing is effective is determined by comparing the variance of the processed image with the data set variance threshold. If not, the deblurring is performed again, and the proportion a of images that need to be deblurred is recalculated. At the same time, whether the image brightness enhancement processing is effective is determined by comparing the average pixel value percentage of the processed image with the data set average pixel threshold. If not, the brightness enhancement is performed again, and the proportion b of images that need to be brightness enhanced is recalculated. At the same time, the parameters are fused to obtain c, and the image is processed again.

[0091] (4) The output image.

[0092] A no-reference image quality evaluation index is selected to evaluate the enhancement effect of the foam image, so as to prove the effectiveness of the embodiment of the present application. In the embodiment of the present application, entropy and standard deviation are selected as the evaluation indexes, and the larger the values are, the better the image quality effect is. The quality evaluation indexes of the original image and the image after enhancement by the enhancement method provided in the embodiment of the present application are shown in Table 1.

[0093] Table 1 Comparison of evaluation indexes of original image and image after enhancement by the enhancement method provided in the embodiment of the present application

[0094]

[0095]

[0096] As can be seen from Table 1, the coal slime foam image after enhancement by the enhancement method provided in the embodiment of the present application has better values in the quality evaluation indexes, which indicates that the image quality is better.

[0097] The visual effect comparison of the original image and the image after enhancement by the enhancement method provided in the embodiment of the present application is shown in Figure 6 Figure 6 ​It can be known that the image enhanced by the enhancement method provided by the embodiment has better effects in brightness, contrast and image definition, higher image quality and more detailed information.

[0098] The coal slime foam image enhancement method and system based on deep learning provided by the embodiment can enhance the coal slime foam image based on deep learning, simplifies the foam image processing program, is simple and convenient to operate, saves the time and manpower spent on screening and processing the foam image dataset, improves the utilization rate of the foam image dataset taken from the production site, and has important significance for improving resource utilization and improving the economic benefits of the coal preparation plant. The foam image processed by the embodiment can be directly used for foam image recognition, classification and segmentation.

[0099] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0100] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device that implements the flowcharts and / or block diagrams. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks.

[0101] These computer program instructions can also be stored in a computer readable storage medium that can guide the computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction devices that implement the flowcharts and / or block diagrams. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the function specified in one flow or multiple flows and / or blocks.

[0102] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are generated to realize the computer-implemented processes in the computer or other programmable devices, and the instructions executed in the computer or other programmable devices provide operational steps for implementing the functions specified in the flowchart Figure 1 one flow or multiple flows and / or the functions specified in one block or multiple blocks. Figure 1 one flow or multiple flows and / or the functions specified in one block or multiple blocks.

[0103] The embodiments of the present application not detailed are the known technology of the skilled in the art. Finally, it is explained that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit, although the present application is described in detail with reference to the preferred embodiments, the ordinary skilled in the art should understand that the technical solutions of the present application can be modified or replaced by the equivalent, without departing from the purpose and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A deep learning-based coal slime foam image enhancement method, characterized in that, The method comprises the steps of: collecting a slime foam image; The deblurring ratio a is calculated by the formula wherein x d is the variance of the slime foam image, x D is the variance threshold; The brightness enhancement ratio b is calculated by the formula wherein x h is the average pixel value percentage of the slime foam image, and x H is the average pixel percentage threshold value. calculating a fusion parameter c through a formula c=a*X+b*Y; wherein X is a deblurring parameter and Y is a brightness enhancement parameter; loading the fusion parameter c into a preset image processing model to output an enhanced foam image.

2. The deep learning-based coal slime froth image enhancement method of claim 1, wherein, The deblurring ratio a is calculated by the formula comprising: comparing the variance x of the coal slurry foam image to a variance threshold x d D ​​ If x d is less than x D , the deblurring ratio a is calculated by the formula ​ 3. The deep learning-based coal slime froth image enhancement method according to claim 1 or 2, characterized in that, The variance threshold x D By formula Calculated; wherein n is the total number of the slime foam images, The variance of the i-th image. 4.The deep learning-based coal slime froth image enhancement method of claim 1, wherein, The luminance enhancement ratio b is calculated by the formula comprising: comparing the average pixel value percentage x of the coal slime froth image to the average pixel percentage threshold x h H to the average pixel percentage threshold x​ If x h Less than x H The brightness enhancement ratio b is calculated by the formula ​ 5. The deep learning-based slime froth image enhancement method of claim 1 or 4, wherein, the average pixel percentage threshold x H by the formula calculated; wherein, is the average pixel percentage of the ith image.

6. A deep learning-based coal slime foam image enhancement system, characterized in that, The method comprises the steps of: a slime foam image collecting module for collecting a slime foam image; A deblurring specific gravity calculation module is configured to calculate a deblurring specific gravity a by a formula wherein x d is a variance of the coal slime foam image, x D is a variance threshold value; The brightness enhancement ratio calculation module is configured to calculate the brightness enhancement ratio b by the formula wherein x h is the average pixel value percentage of the slime foam image, x H is the average pixel percentage threshold value. a fusion parameter calculating module for calculating a fusion parameter c through a formula c=a*X+b*Y; wherein X is a deblurring parameter and Y is a brightness enhancement parameter; an image processing module for loading the fusion parameter c into a preset image processing model to output an enhanced foam image.

7. The deep learning based slurry froth image enhancement system as claimed in claim 6, wherein, The deblurring proportion calculating module comprises: a variance determination unit configured to compare the variance x of the coal slime foam image with a variance threshold value x d and the variance threshold value x D ​ a deblurring specific gravity calculating unit for calculating a deblurring specific gravity a if x d is smaller than x D , by the formula a = x + (x - x) / (x - x) 8. The deep learning-based slurry froth image enhancement system of claim 6 or 7, wherein, The deblurring proportion calculating module further comprises: a variance threshold calculation unit configured to calculate the variance threshold x by a formula D ; wherein n is the total number of the slime foam images, is the variance of the i-th image.​ 9. The deep learning based slurry froth image enhancement system as claimed in claim 6, wherein, The brightness enhancement proportion calculating module comprises: an average pixel percentage judging unit, configured to judge the average pixel value percentage x of the slime foam image h comparing with the average pixel percentage threshold value x H ; a luminance enhancement ratio calculating unit for calculating a luminance enhancement ratio b by a formula h if x H is smaller than x​ 10. The deep learning-based slurry froth image enhancement system of claim 6 or 9, wherein, The brightness enhancement proportion calculating module further comprises: an average pixel percentage threshold calculation unit configured to calculate the average pixel percentage threshold x by the formula x = 1 / n H ; wherein, is the average pixel percentage of the i-th image.

Images