Mine video enhancement method and system based on super-resolution

Abstract

The invention provides a mine video enhancement method and system based on super-resolution, and the method comprises the steps: collecting a mine video image, and carrying out the preprocessing of the collected image; constructing a degradation model by using a super-resolution technology, and training a super-resolution network to obtain a newly generated high-resolution image; the generated high-resolution image is calculated through pixel-level artifact mapping and local variance; and carrying out image segmentation and identification on the generated high-resolution image. The system comprises an image acquisition module which is used for acquiring an original image of a mine video; the degradation module is used for carrying out degradation processing on the original high-resolution image; the super-resolution processing module is used for processing the degraded low-resolution image to generate a high-resolution image; and the image segmentation and recognition module is used for generating a high-resolution image for image segmentation and feature recognition.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent management of coal mines, and in particular relates to a method and system for enhancing mine videos based on super-resolution. Background Art

[0002] With the continuous development of the coal mining industry and the popularization of intelligent management systems, the application of sensors in coal mine production has become more and more extensive. Sensors can monitor a variety of parameters in coal mines in real time, such as gas concentration, temperature, humidity, etc., which provides important support for coal mine production. However, due to the complexity and harsh conditions of the coal mine working environment, sensor images are often interfered by factors such as light and dust, resulting in poor image quality. This poses a huge challenge to the accurate recognition of sensor data. How to effectively improve the quality and accuracy of sensor images has become a problem that coal mine managers and technicians need to solve urgently. In recent years, the development of super-resolution technology and deep learning technology has provided new solutions to this problem.

[0003] The prior art has the following technical problems:

[0004] (1) Extreme lighting conditions: There is a serious lack of light in coal mines, and the sensors usually rely on weak light sources or artificial lighting, which leads to uneven lighting, low contrast, and easy local overexposure or dark areas. Such extreme lighting conditions significantly affect the clarity of the image, making it difficult to identify important details in the image.

[0005] (2) Sample scarcity and angle deviation problems: In coal mining environments, it is very challenging to obtain high-quality sensor images due to harsh working conditions and complex equipment layout. The quality and quantity of sensor images are crucial for training deep learning models, but traditional data augmentation and image recognition techniques have obvious limitations when dealing with sample scarcity and angle deviation problems. As a result, the generalization ability and recognition accuracy of the model are limited.

[0006] (3) Sensor image quality issues: Factors such as dim light and turbid air in the underground environment of coal mines often lead to poor sensor image quality, blur and distortion. These image quality issues directly affect the accurate recognition of sensor data, and thus affect subsequent data analysis and decision-making. Therefore, improving the clarity and stability of sensor images is the key to realizing intelligent management of coal mines.

[0007] (4) High dust and gas disturbance: High concentrations of dust and gas in coal mine environments scatter light, causing blurred images and reduced clarity. Dust also easily adheres to the sensor lens, exacerbating image noise issues. These unique interference factors are particularly significant in coal mine environments, further degrading image data quality. Summary of the invention

[0008] To address the deficiencies in the existing technology, the present invention provides a method and system for enhancing mine videos based on super-resolution, applying super-resolution technology and deep learning technology to the intelligent management system of the coal mining industry. With the development of the coal mining industry and the advancement of the intelligentization process, effectively monitoring and managing underground sensors has become a key task for coal mine managers. By combining super-resolution technology with deep learning algorithms for sensor identification, the present invention aims to improve the accuracy and real-time response ability of sensor data, thereby providing more reliable technical support for coal mine safety production.

[0009] The present invention adopts the following technical solutions.

[0010] The present invention provides a method for enhancing mine videos based on super-resolution, including:

[0011] Collecting mine video images and preprocessing the collected images;

[0012] Using super-resolution technology to construct a degradation model, training the super-resolution network, and obtaining a newly generated high-resolution image;

[0013] Calculating the generated high-resolution image through pixel-level artifact mapping and local variance;

[0014] Performing image segmentation and recognition on the generated high-resolution image.

[0015] Preferably, the preprocessing of the collected images includes:

[0016] Performing grayscale processing on the image,

[0017] Gray = 0.299×R + 0.587×G + 0.114×B

[0018] Gray is the grayscale value, R is the red channel value, G is the green channel value, and B is the blue channel value;

[0019] Performing binary processing on the grayscale processed image;

[0020]

[0021] where B(x,y) represents the pixel value of the binary image, G(x,y) represents the pixel value of the grayscale image, and T is a preset threshold.

[0022] Preferably, the construction of the degradation model includes:

[0023] I LR = D(I HR ) + N

[0024] where ILR is a low-resolution image, I HR is the original high-resolution image; D is the degradation operation, including downsampling and blurring; N is the noise added during the degradation process.

[0025] Preferably, the training of the super-resolution network includes:

[0026] Using the SRGAN model to train the super-resolution network. The SRGAN model consists of a generator and a discriminator. The generator is used to convert the low-resolution image into a high-resolution image, and the discriminator is used to determine whether the generated high-resolution image is consistent with the real image. The loss function of SRGAN includes content loss and adversarial loss.

[0027] L G = L content + λL adv

[0028] where L content is the content loss, which is used to ensure that the generated image is consistent with the real image in content. λ is a weight parameter used to balance the contributions between the content loss and the adversarial loss. L adv is the adversarial loss.

[0029] Preferably, the calculation of the content loss includes:

[0030]

[0031] W and H are the width and height of the image respectively. φ represents the feature extraction function of the VGG network. I HR is the original high-resolution image, and G(I LR ) is the image generated by the generator.

[0032] Preferably, the calculation of the adversarial loss includes:

[0033] L adv = -log(A(G(I LR )))

[0034] where A is the discriminator, and G(I LR ) is the high-resolution image generated by the generator according to the input low-resolution image I LR .

[0035] Preferably, the image segmentation of the generated high-resolution image includes:

[0036]

[0037] σ 2 is the local variance;

[0038] S is the number of pixels within the window;

[0039] x i is the grayscale value of the i-th pixel within the window;

[0040] μ is the average of the grayscale values of all pixels within the window.

[0041] The present invention also provides a mine video enhancement system based on super-resolution. The system is the system used in the aforementioned mine video enhancement method based on super-resolution, and includes:

[0042] An image acquisition module, configured to acquire the original mine video image;

[0043] A degradation module, configured to perform degradation processing on the original high-resolution image;

[0044] A super-resolution processing module, configured to process the degraded low-resolution image to generate a high-resolution image;

[0045] An image segmentation and recognition module, configured to perform image segmentation and feature recognition on the generated high-resolution image.

[0046] The present invention also provides a terminal, including a processor and a storage medium;

[0047] The storage medium is used to store instructions;

[0048] The processor is used to operate according to the instructions to execute the steps of the aforementioned method.

[0049] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps of the aforementioned method are implemented.

[0050] The beneficial effects of the present invention are as follows. Compared with the prior art, the present invention significantly improves the recognition accuracy and real-time performance of underground sensor images by comprehensively applying various technical means such as super-resolution technology, grayscale processing, binarization processing, local discriminant learning, image segmentation and threading method recognition, real-time processing optimization, and early warning system integration. The specific advantages are as follows:

[0051] 1. Super-resolution technology: By converting a low-resolution image into a high-resolution image, the clarity and details of the image are significantly improved, enabling the 7-segment code to be clearly displayed even in harsh environments;

[0052] 2. Grayscale and binarization processing: Simplify the image data, highlight the 7-segment code features, reduce background noise, and improve the recognition efficiency;

[0053] 3. Local discriminant learning: By pixel-level artifact mapping and local variance calculation, enhance the important features in the image and further improve the recognition accuracy;

[0054] 4. Image segmentation and threading method recognition: Effectively segment the 7-segment code area and quickly recognize specific numbers through the threading method, suitable for real-time processing;

[0055] 5. Real-time processing optimization: Significantly improve the processing efficiency through technologies such as parallel computing, multi-threading processing, and hardware acceleration to ensure real-time performance;

[0056] 6. Early warning system integration: Integrate the recognition results with the early warning system to achieve full-range monitoring and management of the coal mine production process and improve safety. Brief Description of the Drawings

[0057] Figure 1 It is a schematic diagram of the system framework of the present invention;

[0058] Figure 2 It is a schematic diagram of the binaryzation processing flow in the present invention;

[0059] Figure 3 It is a schematic diagram of the image super-resolution reconstruction process of SRGAN in the present invention. Detailed Embodiments

[0060] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. The embodiments described in this application are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the spirit of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0061] Embodiment 1 of the present invention provides a mine video enhancement method based on super-resolution technology, covering a series of technical steps for image processing and recognition in harsh environments. First, the resolution of the sensor image is significantly improved through super-resolution technology, and combined with grayscale processing and binaryzation processing, the image quality is effectively improved. Secondly, through technologies such as local discriminant learning, pixel-level artifact mapping, and local variance calculation, the recognition accuracy of the image is further improved. Especially in the recognition of the sensor 7-segment code, a local discriminant learning method is adopted, combined with pixel-level artifact mapping and local variance calculation, significantly enhancing the recognition effect. Finally, these processing steps are integrated into a real-time processing and early warning system to achieve real-time monitoring and abnormal early warning functions. The system flow chart is as Figure 1 shown:

[0062] Step 1: Collect mine video images and preprocess the collected images. The main purpose of image preprocessing is to remove noise and interference, improve image quality, and provide a good foundation for subsequent processing and recognition. The preprocessing process includes two main steps: grayscale processing and binarization processing.

[0063] Step 1.1: Perform grayscale processing on the image;

[0064] Grayscale processing is the process of converting the original color sensor image into a grayscale image. By performing grayscale processing, the computational complexity is reduced and the contour features of the 7-segment code are highlighted. A color image usually contains three channels: red, green, and blue, while a grayscale image only contains one channel, thus greatly reducing the amount of data and computational requirements. Grayscale processing is achieved by calculating the brightness value of each pixel, usually using the weighted average method for calculation, that is, the grayscale value of each pixel is the weighted sum of its red, green, and blue channel values. The specific formula is as follows:

[0065] Gray = 0.299×R + 0.587×G + 0.114×B

[0066] Gray represents the grayscale value, R represents the red channel value, G represents the green channel value, and B represents the blue channel value. Grayscale processing not only simplifies the image data but also effectively highlights the contrast of the 7-segment code, making it easier to identify in subsequent processing steps.

[0067] Step 1.2: Perform binarization processing on the grayscale processed image;

[0068] Binarization processing converts the grayscale image into a binary image. By setting a threshold, using an image processing library to read the image, obtaining the pixel values, and classifying the pixel values into two categories, namely foreground and background. Binarization processing further removes background noise and highlights the feature area of the 7-segment code. Commonly used binarization methods include the global threshold method and the adaptive threshold method. The global threshold method uses a fixed threshold to divide the image into foreground and background, and the formula is as follows:

[0069]

[0070] As Figure 2 shown, where B(x, y) represents the pixel value of the binarized image, G(x, y) represents the pixel value of the grayscale image, and T is the preset threshold. The adaptive threshold method dynamically adjusts the threshold according to the grayscale values of local regions and is more suitable for images with uneven illumination. Binarization processing significantly improves the contrast of the image, making the features of the 7-segment code more obvious and providing a good foundation for the subsequent recognition process.

[0071] Step 2: Use super-resolution technology to construct a degradation model and train the super-resolution network;

[0072] Step 2.1, Degradation model construction;

[0073] The degradation model is used to simulate the generation process of low-resolution images and usually includes operations such as downsampling, blurring, and adding noise. These operations aim to simulate various factors that cause image quality degradation during the actual shooting process. By constructing a reasonable degradation model, the super-resolution network can be effectively trained to better restore high-resolution images in practical applications. When constructing the degradation model, various factors in practical applications need to be considered, such as the resolution of the sensor, the degree of blurring of the lens, and environmental noise. The formula of the degradation model is as follows:

[0074] I LR = D(I HR ) + N

[0075] where, I LR represents the low-resolution image, and I HR represents the original high-resolution image, that is, the originally captured mine video image. D represents the degradation operation, which includes processing steps such as downsampling and blurring that reduce image quality during the construction of the degradation model. Downsampling simulates the decrease in image quality by reducing the resolution of the image, usually involving reducing the image size, which may introduce pixelation effects. Blurring simulates defocus or motion blur of the image by applying a convolution kernel (such as Gaussian blur), affecting the sharpness and details of the image. These processing steps can effectively simulate the performance of actual images under different quality conditions. N represents the noise added during the degradation process, which can simulate noise interference in the actual image capture process such as sensor noise and compression noise.

[0076] Step 2.2, Super-resolution network training;

[0077] In the present invention, we adopt the SRGAN (Super-Resolution Generative Adversarial Network) model to train the super-resolution network. The SRGAN model consists of two parts: a generator and a discriminator. The generator is responsible for converting the low-resolution image into a high-resolution image, while the discriminator is used to judge whether the generated high-resolution image is consistent with the real image. Through adversarial training, the generator and the discriminator continuously improve the quality of the generated image. The VGG network (VGGNet) is a deep convolutional neural network, usually VGG16 or VGG19, referring to a network with 16 layers or 19 layers. In the SRGAN model, the VGG network is commonly used as a feature extractor. Specifically, the pre-trained model of the VGG network is used to calculate the content loss. The content loss measures the difference between the generated image and the high-resolution real image in the feature space, and the VGG network can provide rich intermediate layer features, which are particularly effective for measuring image quality.

[0078] The loss function of SRGAN consists of two parts: content loss and adversarial loss, and its formula is as follows:

[0079] L G = L content + λL adv

[0080] Among them, L content is the content loss, which is used to ensure that the generated image is consistent with the real image in content. λ is a weight parameter used to balance the contributions between the content loss and the adversarial loss. L adv is the adversarial loss.

[0081] L content The content loss usually uses the perceptual loss, and the formula is:

[0082]

[0083] Here, W and H are the width and height of the image respectively, φ represents the feature extraction function of the VGG network, I HR is the original high-resolution image, and GI LR is the image generated by the generator.

[0084] Step 2.3, calculate the adversarial loss;

[0085] The adversarial loss is used to enhance the realism of the image generated by the generator, making it difficult for the discriminator to distinguish between the generated image and the original image. The formula of the adversarial loss is as follows:

[0086] L adv = -log(A(G(I LR )))

[0087] Among them, A is the discriminator, and G(I LR ) is the high-resolution image generated by the generator according to the input low-resolution image I LR .

[0088] As Figure 3 shown, by alternately training the generator and the discriminator, the SRGAN model can generate high-quality super-resolution images, significantly improving the clarity and details of the images.

[0089] Step 3, further improve the image quality and recognition accuracy through pixel-level artifact mapping and local variance calculation.

[0090] The pixel-level artifact mapping technique analyzes the detailed features in an image by generating an artifact image. Artifacts refer to artificially created interference patterns introduced in the image, and these interference patterns can highlight certain features in the image, making them more obvious. In the present invention, through pixel-level artifact mapping, the edges and details of the 7-segment code can be highlighted, making them more obvious and easier to identify in subsequent processing. Artifact mapping is usually achieved by performing edge detection and feature extraction on the image, and common methods include Canny edge detection and Sobel operator, etc.

[0091] Local variance calculation is a method for analyzing the changes in local regions of an image. By calculating the variance value of the local region of the image, significant change regions can be detected. The local region of the image refers to a specific region in the image, usually defined by a sliding window or a region of a fixed size. Regions with high local variance usually contain more detailed and edge information, while regions with low local variance are relatively smooth. By performing local variance calculation on the image, significant features in the 7-segment code region can be effectively identified, and the interference of background noise can be reduced. In a specific implementation, by calculating the variance of pixel values within a sliding window, the degree of change in this region can be quantified. Regions with a relatively large local variance can highlight important details in the image, such as the edges of the 7-segment code, while regions with a relatively small local variance help remove background noise, thereby improving the recognition accuracy of the 7-segment code region.

[0092] The connection between the two steps of pixel-level artifact mapping and local variance calculation lies in their joint action on the feature enhancement and noise suppression of the image:

[0093] Feature enhancement: Pixel-level artifact mapping highlights important features in the image, such as edges and details, by introducing artifacts, making the target region (such as the 7-segment code) in the image more prominent. This enhancement helps subsequent steps more easily identify the key information in the image.

[0094] Noise suppression: Local variance calculation helps distinguish between detailed regions and background noise regions by analyzing the changes in local regions of the image. Regions with high variance usually correspond to more details and edges, while regions with low variance may be smooth backgrounds. By calculating the variance, the influence of background noise can be reduced, further improving the recognition accuracy of the feature region.

[0095] Combining these two methods, artifact mapping first enhances the target features in the image, while local variance calculation helps suppress noise and improves the recognition accuracy of these features.

[0096] Step 4, segment and recognize the image;

[0097] Image segmentation and recognition are the key steps for extracting and recognizing the 7-segment code features from the preprocessed and enhanced image, including image segmentation and recognition by the threading method.

[0098] Step 4.1, Image segmentation

[0099] Image segmentation is to divide an image into multiple regions, and each region corresponds to different objects or features. In the present invention, the 7-segment code region of the digital display in the sensor is extracted through image segmentation to separate it from the background. Common image segmentation methods include threshold-based segmentation, edge-based segmentation, and region-based segmentation, etc. The threshold-based segmentation method is simple and effective and is suitable for images with obvious gray-scale differences; the edge-based segmentation method is suitable for images with clear edges by detecting the edges in the image; the region-based segmentation method is suitable for the case where the same object in the image has similar features by analyzing the connected regions in the image. The specific formula is as follows:

[0100]

[0101] σ 2 represents the local variance, which is an index to measure the dispersion degree of pixel values within the window.

[0102] S is the number of pixels within the considered window, which determines the scale of local variance calculation.

[0103] x i is the gray value of the i-th pixel within the window.

[0104] μ is the average value of all pixel gray values within the window and is used to measure the average level of the overall brightness of this window.

[0105] Step 4.2, Threading method recognition

[0106] The threading method recognition is a classic 7-segment code recognition method. By performing threading processing on the segmented 7-segment code region, the specific digits of the 7-segment code are recognized. The threading method draws horizontal and vertical lines within the 7-segment code region and detects whether there are bright spots on each line to determine the state of each segment code. According to the state of each segment code, the specific digit can be recognized. The threading method recognition has the advantages of simple calculation and high efficiency and is suitable for scenarios with high requirements for real-time processing.

[0107] Embodiment 2 of the present invention provides a mine video enhancement system based on super-resolution. The system is the system used for the foregoing mine video enhancement method based on super-resolution, and includes:

[0108] An image acquisition module, which is used to acquire the original mine video image;

[0109] A degradation module, which is used to perform degradation processing on the original high-resolution image;

[0110] A super-resolution processing module for processing the degraded low-resolution image to generate a high-resolution image;

[0111] An image segmentation and recognition module for performing image segmentation and feature recognition on the high-resolution image.

[0112] The beneficial effects of the present invention are as follows. Compared with the prior art, the present invention significantly improves the recognition accuracy and real-time performance of the downhole sensor image by comprehensively applying various technical means such as super-resolution technology, grayscale processing, binarization processing, local discriminant learning, image segmentation and threading method recognition, real-time processing optimization, and early warning system integration. The specific advantages are as follows:

[0113] 1. Super-resolution technology: By converting the low-resolution image into a high-resolution image, the clarity and details of the image are significantly improved, enabling the 7-segment code to be clearly displayed even in harsh environments;

[0114] 2. Grayscale and binarization processing: Simplify the image data, highlight the 7-segment code features, reduce background noise, and improve the recognition efficiency;

[0115] 3. Local discriminant learning: By pixel-level artifact mapping and local variance calculation, enhance the important features in the image and further improve the recognition accuracy;

[0116] 4. Image segmentation and threading method recognition: Effectively segment the 7-segment code area and quickly identify the specific numbers through the threading method, which is suitable for real-time processing;

[0117] 5. Real-time processing optimization: By technologies such as parallel computing, multi-threading processing, and hardware acceleration, significantly improve the processing efficiency and ensure real-time performance;

[0118] 6. Early warning system integration: Integrate the recognition results with the early warning system to achieve comprehensive monitoring and management of the coal mine production process and improve safety.

[0119] The present disclosure can be a system, a method, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0120] A computer-readable storage medium can be a tangible device that can hold and store instructions for use by an instruction execution device. A computer-readable storage medium may be, for example—but not limited to—an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, a mechanically encoded device such as a punched card or raised structures in grooves having instructions stored thereon, and any suitable combination of the foregoing. The computer-readable storage medium as used herein is not construed as being a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable), or an electrical signal transmitted through a wire.

[0121] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include a copper transmission cable, an optical fiber transmission, a wireless transmission, a router, a firewall, a switch, a gateway computer, and / or an edge server. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0122] Computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state-setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider). In some embodiments, by using the state information of the computer-readable program instructions to customize an electronic circuit, such as a programmable logic circuit, a field-programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit can execute the computer-readable program instructions to implement various aspects of the present disclosure.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent substitutions can still be made to the specific embodiments of the present invention, and any modification or equivalent substitution that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A mine video enhancement method based on super-resolution, characterized in that: include: Collect mine video images and pre-process the collected images; Use super-resolution technology to build a degradation model, train the super-resolution network, and obtain a newly generated high-resolution image; The resulting high-resolution image is computed through pixel-level artifact mapping and local variance; Perform image segmentation and recognition on the generated high-resolution images.

2. The method for enhancing mine video based on super-resolution according to claim 1, characterized in that: The preprocessing of the collected images comprises: Grayscale the image. Gray=0.299×R+0.587×G+0.114×B Gray is the gray value, R is the red channel value, G is the green channel value, and B is the blue channel value; Binarize the grayscale image. Among them, B(x, y) represents the pixel value of the binarized image, G(x, y) represents the pixel value of the grayscale image, and T is the preset threshold.

3. The method for enhancing mine video based on super-resolution according to claim 2, characterized in that: The constructing of the degradation model comprises: I LR =D(I HR )+N Among them, I LR is a low-resolution image, I HR is the original high-resolution image; D is the degradation operation, including downsampling and blurring; N is the noise added during the degradation process.

4. The method for enhancing mine video based on super-resolution according to claim 3, characterized in that: The training of the super-resolution network comprises: The SRGAN model is used to train the super-resolution network. The SRGAN model consists of two parts: a generator and a discriminator. The generator is used to convert low-resolution images into high-resolution images, and the discriminator is used to determine whether the generated high-resolution images are consistent with the real images. The loss function of SRGAN includes content loss and adversarial loss. THE G =L content +λL adv Among them, L content is the content loss, which is used to ensure that the generated image is consistent with the real image in terms of content. λ is a weight parameter used to balance the contribution between content loss and adversarial loss. adv To combat losses.

5. The method for enhancing mine video based on super-resolution according to claim 4, characterized in that: The calculation of the content loss includes: W and H are the width and height of the image, respectively. φ represents the feature extraction function of the VGG network. I HR is the original high-resolution image, G(I LR ) is the image generated by the generator.

6. The method for enhancing mining video based on super-resolution according to claim 5, characterized in that: The calculation of the adversarial loss includes: L adv =-log(A(G(I LR ))) Among them, A is the discriminator, G(I LR ) is the generator according to the input low-resolution image I LR Generated high-resolution image.

7. The method for enhancing mine video based on super-resolution according to claim 6, characterized in that: The image segmentation of the generated high-resolution image comprises: σ 2 is the local variance; S is the number of pixels in the window; x i is the gray value of the i-th pixel in the window; μ is the average grayscale value of all pixels in the window.

8. A mine video enhancement system based on super-resolution, the system being a system used by a mine video enhancement method based on super-resolution as claimed in any one of claims 1 to 7, characterized in that: include: Image acquisition module, used to collect original images of mine videos; A degradation module, used to perform degradation processing on the original high-resolution image; A super-resolution processing module is used to process the low-resolution image after degradation to generate a high-resolution image; Image segmentation and recognition module, used to generate high-resolution images for image segmentation and feature recognition.

9. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images