AI Analysis and Detection Methods and Devices for Coal Mine Power Distribution Rooms

Abstract

This invention discloses an AI analysis and detection method and device for coal mine power distribution rooms, belonging to the field of AI intelligent analysis and detection. The method includes: determining whether the current video image from the AI ​​camera in the coal mine power distribution room needs image enhancement; if so, inputting an infrared image captured by a thermal imaging camera at the same location as the current video image, after angle transformation, into a pre-trained growth adversarial network model, to generate a new visible light video image for AI analysis and detection using the learned mapping relationship between the thermal radiation information of the infrared image and the texture information of the visible light image; based on the AI ​​analysis and detection results and the temperature monitoring of the thermal imaging camera, providing early warnings for corresponding detection events inside and outside the coal mine power distribution room. This solution can inject detailed texture information from the infrared image into the current video image, thereby effectively enhancing the visible light image under dim coal mine conditions and improving the accuracy of AI analysis, detection, and early warning.

Description

Technical Field

[0001] This invention relates to the field of AI intelligent analysis and detection technology, and in particular to an AI analysis and detection method and device for a coal mine power distribution room. Background Technology

[0002] With the rapid development of my country's coal mining industry, AI intelligent analysis technology can be used not only to monitor the operating status of coal mine machinery and equipment, but also to detect and supervise the safety behavior of coal mine personnel.

[0003] Coal mine power distribution rooms not only require scene detection but also temperature alarms. With the development of AI technology, high-definition dome cameras can be installed in coal mine power distribution rooms to utilize AI intelligent analysis technology for detection, while mine thermal imaging cameras can be used for temperature monitoring. However, coal mines are complex environments with dim lighting, which can easily lead to problems such as low brightness, severe noise, and loss of detail in the images captured by the cameras. Traditional image enhancement methods mainly increase image contrast through histogram equalization and Laplacian transform, but these methods do not take into account the contextual information in the image, failing to obtain truly clear images. This significantly affects the accuracy of AI analysis and detection in the hoisting inclined roadway of the coal mine's auxiliary shaft, resulting in ineffective warnings.

[0004] Therefore, there is an urgent need to provide a new AI analysis and detection method for coal mine power distribution rooms. Summary of the Invention

[0005] To address the problem that existing image enhancement methods cannot restore the details of blurry images captured by AI cameras in underground coal mines with high precision, thus affecting the effectiveness of AI monitoring and early warning in the hoisting inclined roadway of the coal mine auxiliary shaft, this invention provides a comprehensive monitoring and early warning method and device for the hoisting inclined roadway of the coal mine auxiliary shaft.

[0006] On the one hand, an AI analysis and detection method for the power distribution room of a coal mine is provided, the method comprising:

[0007] Determine whether the current video image from the AI ​​camera in the coal mine's power distribution room needs image enhancement;

[0008] If so, the infrared image captured by the mining thermal imaging camera at the same position as the current video image is angle-transformed and input into the pre-trained growth adversarial network model. This model utilizes the pre-learned mapping relationship between the thermal radiation information of the infrared image and the texture information of the visible light image to generate a new visible light video image. The new visible light video image is then fused with the current video image for AI analysis and detection. If not, AI analysis and detection are performed directly based on the current video image.

[0009] Based on the AI ​​analysis and detection results of the AI ​​camera and the temperature monitoring of the mining thermal imaging camera, early warnings are issued for corresponding detection events inside and outside the coal mine power distribution room.

[0010] On the other hand, an AI analysis and detection device for a coal mine power distribution room is provided, used to implement the steps described in any method embodiment of the specification, the device comprising:

[0011] The judgment unit is used to determine whether the current video image of the AI ​​camera in the coal mine power distribution room needs image enhancement;

[0012] If the analysis unit is used, it is used to transform the angle of the infrared image captured by the mining thermal imaging camera at the same position as the current video image and input it into the pre-trained growth adversarial network model. It uses the pre-learned mapping relationship between the thermal radiation information of the infrared image and the texture information of the visible light image to generate a new visible light video image. The new visible light video image and the current video image are then fused together for AI analysis and detection. If not, AI analysis and detection are performed directly based on the current video image.

[0013] The early warning unit is used to provide early warnings for corresponding detection events inside and outside the coal mine power distribution room based on the AI ​​analysis and detection results of the AI ​​camera and the temperature monitoring of the mine thermal imaging camera.

[0014] On the other hand, a computer device is provided, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for executing the computer program stored in the memory to implement the steps of the method described above.

[0015] On the other hand, a computer-readable storage medium is provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, it implements the steps of the method described above.

[0016] On the other hand, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described above.

[0017] The technical solution provided by this invention can bring at least the following beneficial effects:

[0018] By utilizing a growth adversarial network model, infrared images captured by thermal imaging cameras are mapped and converted into visible light video images, which are then fused with the current video image. This allows for the injection of detailed texture information from the infrared images into the current video image, thereby effectively enhancing visible light images in dimly lit coal mines and improving the accuracy of AI analysis, detection, and early warning in coal mine power distribution rooms. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of an AI analysis and detection method for a coal mine power distribution room according to an embodiment of the present invention;

[0021] Figure 2 This is a structural diagram of an AI analysis and detection device for a coal mine power distribution room according to an embodiment of the present invention;

[0022] Figure 3 This is a hardware architecture diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0024] The specific implementation of the above concept is described below.

[0025] Please refer to Figure 1 This invention provides an AI analysis and detection method for a coal mine power distribution room, the method comprising:

[0026] Step 100: Determine whether the current video image from the AI ​​camera in the coal mine's power distribution room needs image enhancement;

[0027] Step 102: If yes, then after angular transformation, input the infrared image taken by the mining thermal imaging camera at the same position as the current video image into the pre-trained growth adversarial network model, so as to generate a new visible light video image by utilizing the pre-learned mapping relationship between the thermal radiation information of the infrared image and the texture information of the visible light image, and then fuse the new visible light video image with the current video image for AI analysis and detection; if no, then directly perform AI analysis and detection based on the current video image.

[0028] Step 104: Based on the AI ​​analysis and detection results of the AI ​​camera and the temperature monitoring of the mine thermal imaging camera, provide early warnings for each corresponding detection event in the coal mine power distribution room, both inside and outside the mine.

[0029] In this embodiment of the invention, by utilizing a growth adversarial network model, infrared images captured by thermal imaging cameras are mapped and converted into visible light video images, which are then fused with the current video image. This allows for the injection of detailed texture information from the infrared images into the current video image, thereby effectively enhancing visible light images in dimly lit coal mines and improving the accuracy of AI analysis, detection, and early warning in coal mine power distribution rooms.

[0030] The following description Figure 1 The execution method for each step is shown.

[0031] For step 100:

[0032] The power distribution room can be equipped with high-definition dome cameras and KBA12R mining thermal imaging cameras. The equipment features dual channels: a visible light channel for AI scene detection and a thermal imaging channel for equipment temperature measurement and alarm. By installing dual-channel imaging equipment in the power distribution room, real-time monitoring of the hoist's details and the temperature of the power distribution equipment can be achieved, enabling real-time monitoring and alarm of the status of the coal mine's auxiliary shaft hoist equipment, ensuring equipment performance and safety.

[0033] It is understandable that AI video analytics includes comprehensive monitoring and full-process recording of worker behavior and machinery status, forming a visualized, traceable, and accountable video safety supervision and management system.

[0034] In some implementations, step 100 may include:

[0035] Based on the confidence level and image frame rate of AI analysis and detection within a set historical time period, the operating status of the AI ​​camera is determined;

[0036] Evaluate the quality of the current video image based on its contrast and blur.

[0037] Based on the confidence level, image frame rate, contrast, and blur of the current video image detected by AI within a set historical time period, it is determined whether to perform image enhancement on the current video image.

[0038] In this embodiment, for each AI camera, the operating status of the AI ​​camera and the quality of the video image can be combined to determine whether the current video image needs image enhancement. If it does, enhancement and fusion are performed before AI analysis and detection. Otherwise, AI analysis and detection are performed directly to avoid wasting the computing resources of the backend intelligent analysis platform.

[0039] The operational status of an AI camera is evaluated by using the confidence level and image frame rate of AI analysis and detection within a historical set time period. Abnormal confidence levels may lead to chaotic and unreliable AI analysis and detection. Low or unstable frame rates may cause discontinuous motion trajectories of targets between consecutive frames, which may not be directly reflected in the video image quality. However, due to the discontinuity of video frames, noise and changes in the image may be incorrectly identified as targets, thus increasing the probability of false alarms. Therefore, monitoring the confidence level and image frame rate of AI analysis and detection within a historical set time period allows for the notification of staff for fault maintenance and the simultaneous image enhancement and fusion of the video image to improve the accuracy of AI analysis and detection when the confidence level or video frame rate of the AI ​​intelligent analysis is abnormal. This approach not only determines the need for image enhancement based on the quality of the video image but also considers the operational status of the AI ​​camera, comprehensively integrating adverse factors affecting the accuracy of AI analysis to ensure the accuracy and effectiveness of the judgment. Consequently, it can improve the accuracy of AI analysis and early warning while saving computing resources.

[0040] In some implementations, "based on the confidence level of AI analysis and detection within a historical set time period, the image frame rate, and the contrast and blur of the current video image, it is determined whether to perform image enhancement on the current video image," including:

[0041] The composite index z is calculated based on the following formula:

[0042]

[0043] Among them, Q p Q represents the average confidence level of AI analysis and detection over a set historical period. p′ F is the confidence level standard value. ps F is the image frame rate. ps ′ represents the standard value of the image frame rate, D v x is the standard value of the current video image. i This represents the grayscale value of each pixel in the current video image. Let be the average grayscale value of the current video image, and N be the number of pixels in the current video image. i Where S is the area of ​​the blurred block, n is the effective area of ​​the current video image, g1, g2, g3 and g4 are weighting factors;

[0044] If the overall index of the current video image is less than a preset threshold, it is determined that the current video image needs image enhancement; otherwise, image enhancement is not required.

[0045] In this embodiment, the operating status of the AI ​​camera is an auxiliary factor compared to the quality of the video image. Therefore, using logarithmic operations can reduce the impact of the AI ​​camera's operating status on the comprehensive index, and can more accurately quantify and evaluate the comprehensive index. This allows for a comprehensive consideration of adverse factors affecting the accuracy of AI analysis, ensuring the accuracy and effectiveness of the judgment on whether image enhancement is needed. In this way, the accuracy of AI analysis and early warning can be improved while saving computing resources.

[0046] Regarding step 102:

[0047] In some implementations, the growth adversarial network model is trained in the following manner:

[0048] The generator of the growth adversarial network model contains several learning modules with progressively increasing convolutional channels, until the last learning module has convolutional channels the same size as the video image. The convolutional channels of the learning modules are progressively increased to map the input infrared image from a low-resolution image to a high-resolution image layer by layer to generate a visible light image.

[0049] For each learning module of the generator, a discriminant module corresponding to that learning module is designed to form a discriminator; wherein, the generator is used to generate visible light images based on infrared images, and the discriminator is used to distinguish between the visible light images generated by the generator and real visible light image samples;

[0050] Initialize the weights using a standard normal distribution;

[0051] The growth adversarial network model is trained using the training set to generate visible light images from infrared image samples in the training set. The difference between the generated visible light images and real visible light image samples in the training set is calculated. The network parameters are optimized using stochastic gradient descent to update the parameters of the generator and discriminator networks.

[0052] In this embodiment, the growth adversarial network includes a generator and a discriminator. The generator contains several learning modules with progressively increasing convolutional channels, until the last learning module has convolutional channels the same size as the video image. The discriminator contains several discrimination modules, each corresponding to a learning module. Additionally, a layer is added at the end of the discriminator to progressively downsample and learn a large tensor. This tensor projects the input onto a set of statistical information. The loss function of the generative adversarial network is constructed with the optimization objective of minimizing generation error and feature matching error.

[0053] The growth adversarial network model in this embodiment includes four learning modules. The convolutional channels of the generator's learning modules are progressively increased until the output image of the last learning module is the same size as the video image output by the AI ​​camera. This smooth growth adversarial network can smoothly advance the mapping relationship between the thermal radiation information of infrared images and the texture information of visible light images from low-resolution images to high-resolution images. Compared with traditional generative adversarial networks, which mostly train directly on the original video images, this progressive learning scheme can effectively learn the mapping relationship of rich details and improve the quality of the generated new visible light video images.

[0054] In some implementations, each learning module includes a low-resolution convolutional block and a high-resolution convolutional block, wherein the convolutional channels of the high-resolution convolutional block in the learning module are greater than those of the low-resolution convolutional block, and the convolutional channels of the low-resolution convolutional block in the next learning module are equal to those of the high-resolution convolutional block in the previous learning module, so that the convolutional channels of the learning modules in the generator gradually increase, so as to progressively map the input infrared image from the low-resolution image to the high-resolution image to generate a visible light image.

[0055] In this embodiment, the low-resolution convolutional block in the first learning module is 16×16, and the high-resolution convolutional block is 32×32; the low-resolution convolutional block in the second learning module is 32×32, and the high-resolution convolutional block is 64×64, and so on. Assuming the video image size is 256×256, there is also a learning module and a discrimination module in between, with a low-resolution convolutional block of 64×64 and a high-resolution convolutional block of 128×128. The last learning module has a low-resolution convolutional block of 128×128 and a high-resolution convolutional block of 256×256.

[0056] In this embodiment, the learning module is trained using a standard deviation structure, which includes a low-resolution convolutional block, an upsampling layer, two branches, and a weighted summation layer. One branch generates a first feature image from the upsampled image, while the other branch contains a high-resolution convolutional block to generate a second feature image. The first and second feature images are then weighted and summed in a 1-t:t ratio to output the final feature image. This image enhancement method using the standard deviation structure can effectively learn rich detail mapping relationships, improving the quality of newly generated visible light video images.

[0057] In this embodiment, the discrimination module is also trained using the standard deviation structure. The discrimination module includes two branches, a weighted summation layer, several convolutional blocks with gradually decreasing kernel size, and a downsampling layer between every two convolutional blocks from top to bottom. The third branch downsamples the feature image to obtain a third feature image. The fourth branch contains a convolutional layer and a downsampling layer of the same size as the high-resolution convolutional block of the corresponding learning module to obtain a fourth feature image. The third feature image and the fourth feature image are weighted and summed in a ratio of 1-t:t until they are downsampled to the same size as the low-resolution convolutional block of the first learning module, thereby improving the training accuracy of the corresponding learning module.

[0058] In some implementations, the step "training the growth adversarial network model using the training set to generate visible light images from infrared image samples in the training set, calculating the difference between the generated visible light images and real visible light image samples in the training set, and optimizing the network parameters using stochastic gradient descent to update the generator and discriminator network parameters" may include:

[0059] Infrared image samples from the training set are input into the first learning module. Based on the standard deviation structure of the current scaling factor, the network parameters of the high-resolution convolutional block of the first learning module are trained, and the feature image is output to the corresponding first discriminant module.

[0060] The first discrimination module calculates the loss function based on the feature image and the corresponding real visible light image sample, and feeds it back to the first learning module.

[0061] The scaling factor of the first learning module is increased multiple times, and after each increase, the execution jumps to the standard deviation structure based on the current scaling factor to train the network parameters of the high-resolution convolutional block of the first learning module until the scaling factor is 1. This completes the multiple training of the low-resolution and high-resolution convolutional blocks of the first learning module. The scaling factor of the learning module is selected by comparing the loss function value under each scaling factor.

[0062] The network parameters of the high-resolution convolutional block in the first learning module are inherited by the low-resolution convolutional block in the second learning module.

[0063] The feature image output by the first learning module at the selected scaling factor is input into the second learning module. Based on the standard deviation structure of the current scaling factor, the network parameters of the high-resolution convolutional block of the second learning module are trained, and the feature image is output to the corresponding second discriminant module.

[0064] The second discrimination module calculates the loss function based on the feature image and the corresponding real visible light image sample, and feeds it back to the second learning module;

[0065] The scaling factor of the first learning module is increased multiple times, and after each increase, the execution jumps to the standard deviation structure based on the current scaling factor to train the network parameters of the high-resolution convolutional block of the second learning module until the scaling factor is 1. This completes the multiple training of the low-resolution and high-resolution convolutional blocks of the second learning module. The scaling factor of the learning module is selected by comparing the loss function value under each scaling factor.

[0066] This process is repeated for each of the remaining learning modules until the loss function of the last discriminant module is less than a set threshold, at which point the growth adversarial network model is obtained.

[0067] In this embodiment, each learning module undergoes multiple training iterations. The first iteration uses a scaling factor t of 0, which is incremented by a step size, for example, 0.25. The second iteration uses a scaling factor t of 0.25, and so on, until the fifth iteration uses a scaling factor t of 1. The first learning module and the first discriminator module are executed five times to obtain the final network parameters of the high-resolution convolutional block of the first learning module. These final network parameters are then copied to the low-resolution convolutional block of the second learning module. In essence, each learning module is executed five times to complete its training, achieving a smooth learning progression from low-resolution to high-resolution images.

[0068] Therefore, this invention improves upon traditional adversarial networks by constructing a smooth growth adversarial network. This trained model smoothly progresses from low-resolution to high-resolution images, learning the mapping relationship between thermal radiation information in infrared images and texture information in visible light images at multiple levels. This generates new visible light video images. Compared to traditional image enhancement methods, progressive enhancement effectively learns richer detail mapping relationships, improving the quality of the generated new visible light video images and thus enhancing the accuracy of AI analysis and detection in coal mine power distribution rooms. Furthermore, the decision to enhance images is not limited to quality; the operating status of the AI ​​camera is also considered, comprehensively addressing factors affecting AI analysis accuracy and ensuring the accuracy and effectiveness of the judgment. This improves the accuracy of AI analysis and early warning while conserving computing resources. Therefore, this solution not only effectively improves the quality of generated new visible light video images and enhances the accuracy of AI analysis, detection, and early warning in coal mine auxiliary shaft hoisting inclined roadways, but also avoids wasting computing resources.

[0069] In some implementations, "fusing new visible light video images with current video images and then performing AI analysis and detection" includes:

[0070] Based on the pixel brightness of the new visible light video image, calculate the fusion weight for each pixel:

[0071]

[0072] In the formula, I(x) is the brightness of the x-th pixel in the new visible light video image, and T(x) is the pixel value of the x-th pixel in the new visible light video image. This represents the pixel value of the x-th pixel in the current video image. For convolution operation, K is the convolution kernel, ρ is the control parameter, k is the brightness threshold, and α is the adjustment system, which takes a value less than e;

[0073] The pixel value at each position in the new visible light video image is multiplied by its corresponding fusion weight, and then added to the corresponding pixel in the current video image to obtain the fused video image, which is then used for AI analysis and detection.

[0074] In this embodiment, the fusion weight of each pixel is calculated based on the pixel brightness of the new visible light video image. It can be understood that the higher the brightness, the higher the fusion weight in the formula. By fusing the new visible light video image with the current video image based on the fusion weight, the spatial texture details of the current video image can be effectively enhanced, thereby achieving effective enhancement of visible light images in dim coal mines and improving the accuracy of AI analysis, detection and early warning in coal mine power distribution rooms.

[0075] Please refer to Figure 2 This invention provides an AI analysis and detection device for a coal mine power distribution room, used to implement the steps of any method embodiment in the specification. The device includes:

[0076] The judgment unit 201 is used to determine whether the current video image of the AI ​​camera in the coal mine power distribution room needs to be enhanced.

[0077] If the analysis unit 202 is selected, it is used to transform the angle of the infrared image captured by the mining thermal imaging camera at the same position as the current video image and input it into the pre-trained growth adversarial network model. This model utilizes the pre-learned mapping relationship between the thermal radiation information of the infrared image and the texture information of the visible light image to generate a new visible light video image. The new visible light video image is then fused with the current video image for AI analysis and detection. If not, AI analysis and detection are performed directly based on the current video image.

[0078] The early warning unit 203 is used to provide early warnings for corresponding detection events inside and outside the coal mine power distribution room based on the AI ​​analysis and detection results of the AI ​​camera and the temperature monitoring of the mine thermal imaging camera.

[0079] In one embodiment of the present invention, the determination unit 201 is used to perform:

[0080] Based on the confidence level and image frame rate of AI analysis and detection within a set historical time period, the operating status of the AI ​​camera is determined;

[0081] Evaluate the quality of the current video image based on its contrast and blur.

[0082] Based on the confidence level, image frame rate, contrast, and blur of the current video image detected by AI within a set historical time period, it is determined whether to perform image enhancement on the current video image.

[0083] In one embodiment of the present invention, when the judging unit 201 performs an analysis based on the confidence level, image frame rate, contrast, and blur of the current video image within a historical set time period to determine whether to perform image enhancement on the current video image, it is configured to:

[0084] The composite index z is calculated based on the following formula:

[0085]

[0086] Among them, Q p Q represents the average confidence level of AI analysis and detection over a set historical period. p′ F is the confidence level standard value. ps F is the image frame rate. ps ′ represents the standard value of the image frame rate, D v x is the standard value of the current video image. i This represents the grayscale value of each pixel in the current video image. Let be the average grayscale value of the current video image, and N be the number of pixels in the current video image. i Where S is the area of ​​the blurred block, n is the effective area of ​​the current video image, g1, g2, g3 and g4 are weighting factors;

[0087] If the overall index of the current video image is less than a preset threshold, it is determined that the current video image needs image enhancement; otherwise, image enhancement is not required.

[0088] In one embodiment of the present invention, the growth adversarial network model in the analysis unit 202 is trained in the following manner:

[0089] The generator of the growth adversarial network model contains several learning modules with progressively increasing convolutional channels, until the last learning module has convolutional channels the same size as the video image. The convolutional channels of the learning modules are progressively increased to map the input infrared image from a low-resolution image to a high-resolution image layer by layer to generate a visible light image.

[0090] For each learning module of the generator, a discriminant module corresponding to that learning module is designed to form a discriminator; wherein, the generator is used to generate visible light images based on infrared images, and the discriminator is used to distinguish between the visible light images generated by the generator and real visible light image samples;

[0091] Initialize the weights using a standard normal distribution;

[0092] The growth adversarial network model is trained using the training set to generate visible light images from infrared image samples in the training set. The difference between the generated visible light images and real visible light image samples in the training set is calculated. The network parameters are optimized using stochastic gradient descent to update the parameters of the generator and discriminator networks.

[0093] In one embodiment of the present invention, each learning module in the analysis unit 202 includes a low-resolution convolutional block and a high-resolution convolutional block. The convolutional channels of the high-resolution convolutional block in the learning module are greater than those of the low-resolution convolutional block, and the convolutional channels of the low-resolution convolutional block in the next learning module are equal to those of the high-resolution convolutional block in the previous learning module. This allows the convolutional channels of the learning modules in the generator to gradually increase, so as to progressively map the input infrared image from the low-resolution image to the high-resolution image to generate a visible light image.

[0094] In one embodiment of the present invention, when the analysis unit 202 performs AI analysis and detection after fusing the new visible light video image and the current video image, it is used to:

[0095] Based on the pixel brightness of the new visible light video image, calculate the fusion weight for each pixel:

[0096]

[0097] In the formula, I(x) is the brightness of the x-th pixel in the new visible light video image, and T(x) is the pixel value of the x-th pixel in the new visible light video image. This represents the pixel value of the x-th pixel in the current video image. For convolution operation, K is the convolution kernel, ρ is the control parameter, k is the brightness threshold, and α is the adjustment system, which takes a value less than e;

[0098] The pixel value at each position in the new visible light video image is multiplied by its corresponding fusion weight, and then added to the corresponding pixel in the current video image to obtain the fused video image, which is then used for AI analysis and detection.

[0099] It should be noted that the AI ​​analysis and detection device for the coal mine power distribution room provided in the above embodiments is only an example of the division of the above functional units. In practical applications, the above functions can be assigned to different functional units as needed, that is, the internal structure of the device can be divided into different functional units to complete all or part of the functions described above. In addition, the above device embodiments and method embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0100] Embodiments of this application also provide a computer device, please refer to... Figure 3 The computer device includes a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, at least one program, code set or instruction set being loaded and executed by the processor to implement the AI ​​analysis and detection method for coal mine power distribution room provided in the above-described method embodiments.

[0101] The embodiments of this application also provide a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the AI ​​analysis and detection method for coal mine power distribution rooms provided in the above-described method embodiments.

[0102] The embodiments of this application also provide a computer program product, which includes a computer program. The processor of a computer device reads the computer program from a computer-readable storage medium and executes the computer program, causing the computer device to perform any of the AI ​​analysis and detection methods for coal mine power distribution rooms described in the above embodiments.

[0103] For ease of description, the above systems or devices are described separately as various modules or units based on their functions. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware components.

[0104] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of the embodiments of this application.

[0105] Finally, it should be noted that in this document, relational terms such as first, second, third, and fourth are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0106] The above are merely preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. An AI analysis and detection method for a coal mine power distribution room, characterized in that, The method includes: Determine whether the current video image from the AI ​​camera in the coal mine's power distribution room needs image enhancement; If so, the infrared image captured by the mining thermal imaging camera at the same position as the current video image is angle-transformed and input into the pre-trained growth adversarial network model. This utilizes the pre-learned mapping relationship between the thermal radiation information of the infrared image and the texture information of the visible light image to output a new visible light video image. The new visible light video image is then fused with the current video image for AI analysis and detection. If not, AI analysis and detection are performed directly based on the current video image. Each learning module of the growth adversarial network model includes a low-resolution convolutional block and a high-resolution convolutional block. In this learning module, the convolutional channels of the high-resolution convolutional block are greater than those of the low-resolution convolutional block, and the convolutional channels of the low-resolution convolutional block in the next learning module are equal to those of the high-resolution convolutional block in the previous learning module. This allows the convolutional channels of the learning modules in the generator to gradually increase, progressively mapping the input infrared image from a low-resolution image to a high-resolution image to generate a visible light image. Based on the AI ​​analysis and detection results of the AI ​​camera and the temperature monitoring of the mining thermal imaging camera, early warnings are issued for corresponding detection events in the coal mine power distribution room, both inside and outside the mine. The determination of whether the current video image of the AI ​​camera in the coal mine's power distribution room needs image enhancement includes: Based on the confidence level and image frame rate of AI analysis and detection within a set historical time period, the operating status of the AI ​​camera is determined; Evaluate the quality of the current video image based on its contrast and blur. Based on the confidence level, image frame rate, and contrast and blur of the current video image detected by AI within a set historical time period, a comprehensive index z is calculated according to the following formula: in, The average confidence level of AI analysis and detection over a set historical period. The confidence level standard value, For image frame rate, This is the standard value for image frame rate. The standard value for the current video image. This represents the grayscale value of each pixel in the current video image. The mean grayscale value of the current video image is given, and N is the number of pixels in the current video image. For the area of ​​the fuzzy block, Let n be the effective area of ​​the current video image, and n be the number of blocks. , , and As a weighting factor; When the comprehensive index of the current video image is less than the preset threshold, it is determined that the current video image needs to be enhanced; otherwise, it does not need to be enhanced. The process of fusing the new visible light video image with the current video image and then performing AI analysis and detection includes: Based on the pixel brightness of the new visible light video image, calculate the fusion weight for each pixel: In the formula, Let x be the brightness of the x-th pixel in the new visible light video image. Let x be the pixel value of the x-th pixel in the new visible light video image. This represents the pixel value of the x-th pixel in the current video image. For convolution operations, K is the convolution kernel. For control parameters, The brightness threshold. To adjust the system, the value should be less than e; The pixel value at each position of the new visible light video image is multiplied by its corresponding fusion weight, and then added to the corresponding pixel of the current video image to obtain a fused video image, which is then used for AI analysis and detection.

2. The method as described in claim 1, characterized in that, The growth adversarial network model is trained in the following manner: The generator of the growth adversarial network model contains several learning modules with progressively increasing convolutional channels, until the convolutional channels of the last learning module are the same size as the video image. The convolutional channels of the learning modules are progressively increased to map the input infrared image from a low-resolution image to a high-resolution image layer by layer to generate a visible light image. For each learning module of the generator, a discriminant module corresponding to that learning module is designed to form a discriminator; wherein, the generator is used to generate a visible light image based on an infrared image, and the discriminator is used to distinguish the visible light image generated by the generator from real visible light image samples; Initialize the weights using a standard normal distribution; The growth adversarial network model is trained using the training set to generate visible light images from infrared image samples in the training set. The difference between the generated visible light images and real visible light image samples in the training set is calculated. The network parameters are optimized using stochastic gradient descent to update the parameters of the generator and discriminator networks.

3. An AI analysis and detection device for a coal mine power distribution room, used to implement the steps of the method described in any one of claims 1-2, characterized in that, The device includes: The judgment unit is used to determine whether the current video image of the AI ​​camera in the coal mine power distribution room needs image enhancement; If so, the analysis unit is used to transform the angle of the infrared image captured by the mining thermal imaging camera at the same position as the current video image and input it into the pre-trained growth adversarial network model. It uses the pre-learned mapping relationship between the thermal radiation information of the infrared image and the texture information of the visible light image to output a new visible light video image. The new visible light video image and the current video image are then fused together for AI analysis and detection. If not, AI analysis and detection are performed directly based on the current video image. The early warning unit is used to provide early warnings for corresponding detection events inside and outside the coal mine power distribution room based on the AI ​​analysis and detection results of the AI ​​camera and the temperature monitoring of the mine thermal imaging camera.

4. A computer device, characterized in that, The computer device includes a memory and a processor. The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the steps of the method according to any one of claims 1-2.

5. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the method described in any one of claims 1-2.

6. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1-2.

Images