Drill pipe counting method and device applied to low-illumination in coal mine underground

Abstract

The application relates to the technical field of image pattern recognition, and provides a drill pipe counting method and device applied to low-illumination underground coal mines. Based on preset space factors and gray scale factors, original images obtained by a camera device are subjected to projection processing to remove noise to obtain first images; the first images are processed through a network structure of channel splicing, the mutual relationship between first components and second components in the first images is extracted on different levels, and the first components corresponding to the reflection side and the second components corresponding to the illumination side are extracted from the first images; the first components and the second components are fused to generate an enhanced image corresponding to the original image; features of the enhanced image are extracted, and the number of drill pipes in the original image is determined according to the extracted features. Through multi-scale feature extraction and detection, real-time counting of drill pipes with high precision and low error under low illumination is realized, and a complete closed loop from environment adaptation to target detection is formed.

Description

Technical Field

[0001] This application relates to the field of image pattern recognition technology, and more specifically, to a method and apparatus for counting drill rods in low-light conditions in underground coal mines. Background Technology

[0002] In recent years, the coal mining industry has accelerated its intelligent transformation driven by technological iteration. Currently, video monitoring systems are widely deployed in underground operations, which lays the data foundation for the application of image processing technology in the field of gas drainage borehole depth measurement.

[0003] However, existing monitoring systems mostly focus on video data acquisition and storage, with a low level of intelligence. Currently, drilling depth is typically calculated by combining real-time recordings by underground miners during drilling operations with verification of the drilling operation videos by surface personnel. This method is labor-intensive and inefficient, and prolonged continuous operation inevitably introduces errors, increasing the risk of accidents such as gas explosions and coal and gas outbursts. Therefore, in the low-light drilling rod inspection process in coal mines, there are problems with both low accuracy and efficiency. Summary of the Invention

[0004] This application provides a method and apparatus for counting drill pipes in low-light conditions in underground coal mines, which can at least partially solve the problem of low accuracy and efficiency in the detection of drill pipes in low-light conditions in underground coal mines.

[0005] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0006] According to one aspect of this application, a method for counting drill rods in low-light conditions in underground coal mines is provided, comprising: acquiring an original image of the underground coal mine using a camera device; performing projection processing on the original image based on preset spatial factors and grayscale factors to remove noise from the original image and generate a first image; processing the first image through a channel-stitched network structure to extract the relationship between a first component and a second component in the first image at different levels, and extracting the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image; fusing the first component and the second component to generate an enhanced image corresponding to the original image; performing feature extraction on the enhanced image, and determining the number of drill rods in the original image based on the extracted features.

[0007] In this application, based on the aforementioned scheme, the step of performing projection processing on the original image based on preset spatial factors and grayscale factors to remove noise from the original image and generate a first image includes: performing projection processing on the original image to generate a projected image; decomposing the projected image into image components of a preset number of levels; generating projection factors corresponding to each position in the image components based on preset spatial factors and grayscale factors; filtering the high-frequency components in the image components based on the projection factors to generate denoised components; and performing inverse reconstruction on the denoised components to generate the first image.

[0008] In this application, based on the aforementioned scheme, the processing of the first image through a channel-stitched network structure to extract the relationship between the first component and the second component in the first image at different levels, and to extract the first component corresponding to the reflection side and the second component corresponding to the illuminance side from the first image, includes: constructing a channel-stitched network structure based on a preset downsampling module and an upsampling module; processing the first image through the upsampling module in the network structure to generate the first component corresponding to the reflection side; and generating the second component corresponding to the illuminance side through the downsampling module and the upsampling module of the first component branch in the network structure.

[0009] In this application, based on the aforementioned scheme, fusing the first component and the second component to generate the enhanced image corresponding to the original image includes: performing progressive illumination updates on the second component to generate a third component; and fusing the first component and the third component to generate the enhanced image corresponding to the original image.

[0010] In this application, based on the aforementioned scheme, the step of progressively updating the illumination of the second component to generate the third component includes: generating a self-updating module formula based on the output component output at the current stage; and processing the second component based on the self-updating module formula to generate the third component.

[0011] In this application, based on the aforementioned scheme, the step of generating the self-updating module formula based on the output components of the current stage includes: generating the self-updating module formula G(X) based on the output components of the current stage. t )for:

[0012]

[0013] Where t≥1, Z t It is the low-light observation image y and the output component X of the current stage. t The calculated fusion result; S t It is the result of self-calibration mapping, derived from the parameterized operator K. θ Generate, Vt It is the input used for the next stage after calibration.

[0014] In this application, based on the aforementioned scheme, the step of extracting features from the enhanced image and determining the number of drill rods in the original image based on the extracted features includes: extracting features from the enhanced image through convolution with a preset number of layers and outputting feature maps of different scales; upsampling and downsampling the feature maps and fusing the sampling results to generate fused features; generating candidate detection boxes on the fused features and predicting the number of drill rods.

[0015] According to one aspect of this application, a drill pipe counting device for use in low-light conditions in underground coal mines is provided, comprising:

[0016] The acquisition unit is used to acquire raw images of underground coal mines via a camera device;

[0017] A noise reduction unit is used to perform projection processing on the original image based on a preset spatial factor and grayscale factor to remove noise from the original image and generate a first image.

[0018] The component unit is used to process the first image through a network structure of channel splicing, extract the relationship between the first component and the second component in the first image at different levels, and extract the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image.

[0019] An enhancement unit is used to fuse the first component and the second component to generate an enhanced image corresponding to the original image;

[0020] A counting unit is used to extract features from the enhanced image and determine the number of drill rods in the original image based on the extracted features.

[0021] According to one aspect of this application, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the drill pipe counting method for low-light conditions in coal mines as described in the above embodiments.

[0022] According to one aspect of this application, an electronic device is provided, comprising: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the drill pipe counting method for low-light conditions in coal mines as described in the above embodiments.

[0023] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the drill pipe counting method for low-light conditions in coal mines provided in the various alternative implementations described above.

[0024] In this technical solution, an explosion-proof and dustproof camera device is used to adapt to the harsh underground environment. Combined with low-light optimization, the original image containing drill rod information is acquired stably. Based on deep learning, spatial and grayscale factors are used to suppress noise while preserving edges, generating a clean first image. The channel stitching network automatically separates the reflection and illumination components, avoiding manual parameter adjustment. After component fusion, a high-contrast enhanced image is generated to highlight the drill rod edges. Finally, through multi-scale feature extraction and detection, high-precision and low-error real-time counting of drill rods under low light conditions is achieved, improving the efficiency and accuracy of drill rod identification and counting under low light conditions in coal mines, forming a complete closed loop from environmental adaptation to target detection.

[0025] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0026] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0027] Figure 1 The flowchart illustrating a drill pipe counting method applied in low-light conditions in underground coal mines is shown in one embodiment of this application.

[0028] Figure 2 The flowchart illustrating the generation of a first image is shown in one embodiment of this application.

[0029] Figure 3 The illustration shows a schematic diagram of a drill pipe counting device applied in low-light conditions in a coal mine, according to one embodiment of this application.

[0030] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0031] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.

[0032] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0033] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0034] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0035] The implementation details of the technical solution of this application are described below:

[0036] Figure 1 A flowchart illustrating a drill pipe counting method applied in low-light conditions in underground coal mines according to an embodiment of this application is shown. (Refer to...) Figure 1 As shown, the drill pipe counting method applied to low-light conditions in coal mines includes at least steps S110 to S150, which are described in detail below:

[0037] S110 acquires raw images of underground coal mines using a camera device.

[0038] In one embodiment of this application, a camera device adapted to the extreme underground environment is first deployed in a coal mine. The camera device employs an explosion-proof housing and a sealed structure to prevent the risk of methane explosions and coal dust intrusion, ensuring long-term stable operation. The optical system utilizes a large-aperture lens and infrared illumination technology to increase light intake under low-light conditions, solving the problem of blurred images caused by insufficient light underground. The dustproof design combines physical barriers and air curtain isolation to effectively reduce coal dust contamination of the lens and sensor, maintaining the clarity of image acquisition.

[0039] In this embodiment, the camera device converts downhole optical signals into electrical signals using a photoelectric sensor. The photosensitive unit generates an electric charge upon receiving photons, the amount of which is proportional to the light intensity. This charge is read out by the circuit and quantized into a digital signal, forming the original image. To address downhole scattering noise and Gaussian noise, hardware optimization and basic noise reduction algorithms are used for preprocessing. For example, scattering noise caused by coal dust and sensor thermal noise are filtered using a low-noise sensor, providing usable images for subsequent steps.

[0040] To cope with the complex underground environment, the camera device integrates automatic gain control and noise suppression functions. Under low light conditions, the system dynamically adjusts the signal amplification factor to improve image brightness while suppressing noise amplification. The dustproof design, combined with hardened optical windows and positive pressure protection, reduces the interference of coal dust on imaging. Although the final output raw image has noise and resolution limitations, through environmental adaptability optimization, it meets the basic quality requirements for subsequent denoising and enhancement processes, forming a reliable capture link from environment to image.

[0041] In this embodiment, the explosion-proof and dust-proof design of the camera device ensures stable operation in downhole environments with excessive methane concentrations, preventing monitoring interruptions due to equipment failure. Low-light adaptability is achieved through a combination of a large-aperture lens and infrared illumination, enabling the capture of sufficient light signals even under dust obstruction, forming a raw image containing the drill pipe's outline and texture, providing basic data for subsequent processing.

[0042] S120, based on preset spatial factors and grayscale factors, the original image is subjected to projection processing to remove noise from the original image and generate a first image.

[0043] When projecting the original image, a multi-level decomposition method (such as wavelet transform) is used to divide the image into different frequency components. High-frequency components contain noise and edge details, while low-frequency components retain the overall structure. Based on preset spatial factors (reflecting pixel position correlation) and grayscale factors (reflecting pixel grayscale differences), projection factors are generated. The spatial factor controls the neighborhood range to avoid over-smoothing edges; the grayscale factor adjusts the grayscale similarity weights to preserve detail contrast. The filtered high-frequency components and the unprocessed low-frequency components are then inversely reconstructed to generate the first image. This process removes Gaussian and scattering noise while preserving the sharpness of the drill rod edges, providing a low-noise, high-contrast intermediate image for subsequent component decomposition.

[0044] like Figure 2 As shown, in one embodiment of this application, based on a preset spatial factor and grayscale factor, the original image is subjected to projection processing to remove noise from the original image and generate a first image, including:

[0045] S210, the original image is subjected to projection processing to generate a projected image, and the projected image is decomposed into image components of a preset number of levels;

[0046] S220, Based on the preset spatial factor and grayscale factor, generate the projection factor corresponding to each position in the image component;

[0047] S230, based on the projection factor, the high-frequency components in the image components are filtered to generate denoised components;

[0048] S240, the denoised components are inversely reconstructed to generate the first image.

[0049] In underground coal mine applications, raw images acquired by cameras suffer from severe interference due to the underground environment, such as dust, low light, and sensor noise. Noise reduction processing is necessary to improve image quality and provide a clear foundation for subsequent analysis. Specifically, the noise in the raw images mainly comes from three sources: sensor thermal noise, camera dead pixels, and dust scattering. Sensor thermal noise is randomly distributed, similar to Gaussian noise resembling static on a television screen; camera dead pixels appear as randomly occurring black or white dots, i.e., salt-and-pepper noise; and dust scattering is caused by suspended particles in the air scattering light, creating a fog-like scattering noise.

[0050] Optionally, the wavelet type can be dynamically selected based on the image texture complexity, such as the sharpness of the drill rod edge, to balance the need to preserve details and suppress noise. Simultaneously, the image is decomposed into image components of different frequencies through multi-level decomposition, i.e., the image is split into contour layers and detail layers, with a focus on processing the noisy high-frequency detail layers.

[0051] In one embodiment of this application, the original image is subjected to projection processing to generate a projected image, and the projected image is decomposed into image components I of a preset number of levels. raw for:

[0052]

[0053] in, Let m ∈ {h,v,d} represent the high-frequency components of the l-th order direction, where h,v,d represent the diagonal view, the horizontal view, and the vertical view, respectively.

[0054] In this embodiment, similar texture regions, such as the repetitive pattern of drill rod edges, are found in the high-frequency detail layer. Noise is suppressed by calculating the projection factor. The calculation of the projection factor considers both the spatial factor (nearest neighbor priority in spatial distance) and the gray-level factor (pixels with similar gray levels are more important). In one embodiment of this application, based on the preset spatial factor and gray-level factor, the projection factor ω(x,y;i,j) corresponding to the position in the image component weight is generated as follows:

[0055]

[0056] Where x and y represent the total length and total width of the image component, respectively, i and j represent the coordinates in the image component, I represents the pixel value, and h s h r These represent the spatial factor and the gray factor, respectively.

[0057] In one embodiment of this application, based on the projection factor, the high-frequency components in the image components are filtered to generate corresponding denoised components. for:

[0058]

[0059] Here, N(x,y) represents a 7×7 neighborhood centered at (x,y).

[0060] In one embodiment of this application, the denoised components are inversely reconstructed, that is, the denoised components are combined to generate the first image. The processed denoised components of each frequency are then recombined to generate the denoised first image. At this point, the noise in the image has been significantly reduced, and the drill pipe edges and textures are clearer.

[0061] The above process divides the image into different frequency components through multi-level decomposition. For the noisy high-frequency components, adaptive filtering is used to suppress scattering noise and Gaussian noise while preserving edge details. The noise in the generated first image is significantly reduced, and the edge sharpness of the drill rod is improved, providing a clean base for component decomposition and avoiding noise interference with the accurate extraction of reflection and illumination components.

[0062] S130, the first image is processed through a network structure of channel splicing, and the relationship between the first component and the second component in the first image is extracted at different levels, and the first component corresponding to the reflection side and the second component corresponding to the illumination side are extracted from the first image.

[0063] When processing the first image using a channel-concatenated network structure, an encoding / decoding architecture is employed. Downsampling modules (such as convolutional layers) are used to progressively extract high-level semantic features, followed by upsampling modules to restore spatial details. At different levels (such as shallow edges and deep semantics), skip connections are used to concatenate downsampled and upsampled feature maps, fusing multi-scale information to simultaneously capture details of the reflection component (such as drill rod texture) and the illumination distribution of the illuminance component. The reflection component is directly generated through the upsampling path, preserving local contrast; the illuminance component is separated by using a downsampling path combined with global average pooling to isolate illumination variations.

[0064] In one embodiment of this application, the first image is processed using a channel-stitched network structure to extract the relationship between the first component and the second component in the first image at different levels, and to extract the first component corresponding to the reflection side and the second component corresponding to the illuminance side from the first image, including:

[0065] A network structure for channel splicing is constructed based on preset downsampling and upsampling modules;

[0066] The first image is processed by the upsampling module in the network structure to generate the first component corresponding to the reflection side;

[0067] The second component corresponding to the illuminance side is generated through the downsampling module and upsampling module of the first component branch in the network structure.

[0068] In one embodiment of this application, the network structure employs a dual-channel collaborative encoder and decoder architecture, used to extract the first component corresponding to the reflection side and the second component corresponding to the illumination side in the low-light image, respectively, to achieve high-quality decomposition. The first component branch is constructed using a U-Net structure, achieving feature extraction and reconstruction through symmetrical downsampling and upsampling processes.

[0069] In one embodiment of this application, a channel-stitched network structure is constructed based on a preset downsampling module and an upsampling module. In the downsampling stage, each module consists of two 3×3 convolutional layers and an activation function, and then the spatial dimension is compressed by 2×2 max pooling to gradually capture multi-scale contextual features.

[0070] During the upsampling stage, a skip connection mechanism is introduced to perform channel-by-channel concatenation and fusion of edge and texture details in the shallow feature map with the global semantic information in the deep feature map. This effectively alleviates the detail loss problem caused by the increase in network depth and ensures that the first component can retain the inherent properties and high-frequency details of the object. The first image is then processed by the upsampling module in the network structure to generate the first component corresponding to the reflection side.

[0071] In one embodiment of this application, a second component corresponding to the illumination side is generated through the downsampling module and upsampling module of the first component branch in the network structure. The second component branch reuses the downsampling and upsampling paths of the first component branch, and receives feature maps from the reflection branch as prior guidance through convolution and activation layers during the decoding stage. Combining the low-rank smoothness characteristics of the second component itself, feature aggregation is performed using a 1×1 convolutional layer, and finally the output is constrained to the range of [0,1] through an activation function to ensure the physical rationality of the second component.

[0072] In the above process, the loss function L of the decomposition module d Designed as follows:

[0073] L d =L res +λ r L ref +λ f L fill +λ m L mc

[0074] Where, λ r , λ i and λ m L represents the weighting coefficients used to balance the impact of each loss term on network training. res L represents the reconstruction loss. ref L represents consistent loss. ill L represents the smoothness loss on the illuminance side. mc This indicates a loss of mutual consistency.

[0075] Among them, reconstruction loss L res The constraint used to ensure the consistency between the product of the first and second components after decomposition and the original input image can be expressed as:

[0076] L res =||R l I l -L l ||1+||R n I n -L n ||1

[0077] Among them, I n Il These are the second and first components of the input image, respectively, R. n R l These are the second and first components of the output image, respectively, where |||1 is the L1 norm. If L res The smaller value indicates that the product of the first and second components closely approximates the original image, meaning that the decomposition process preserves the key information of the image, making subsequent component-based processing more meaningful.

[0078] The core of the first component lies in ensuring the consistency of reflectance patterns within the same scene under both low and normal illumination conditions. The first component represents the inherent properties of objects, such as texture and color, which theoretically should not be affected by changes in lighting. This is achieved through L... ref Under the constraints of illumination, the network can learn more stable, illumination-independent reflection features, avoiding unreasonable fluctuations in the first component caused by illumination differences, thus providing a more reliable representation of the object's inherent properties for subsequent processing. The consistency loss L of the first component... ref It can be represented as:

[0079] L ref =||R l -R n ||1

[0080] The smoothness loss L of the second component ill By constraining the gradient change of the second component in space, we ensure that the second component is as smooth as possible in terms of texture details, while preserving overall structural information. The formula is:

[0081]

[0082] in, It is a first-order gradient operator, including horizontal and vertical directions, where ε is a very small positive constant used to avoid division by zero error.

[0083] In this embodiment, a mutual consistency loss L is added. mc To ensure consistency of the second component in both smooth and edge regions, the formula is as follows:

[0084] L mc =||M·exp(-c·M)||1

[0085] Here, M represents the preset loss gradient, and c represents the speed of light. The loss term is constructed by exponentially weighting M. In regions with smooth illumination, the weight exp(-c·M) is relatively large to emphasize the consistency of illumination in these regions. In edge regions, the weight decays to avoid over-constraining illumination changes at the edges.

[0086] The above process extracts the relationship between the reflection component (details) and the illuminance component (brightness) at different levels through downsampling and upsampling modules. The reflection component preserves the texture and edge information of the drill rod, while the illuminance component reflects the illumination distribution and global brightness. The network structure overcomes the limitation of traditional deep learning and machine learning models requiring manual parameter adjustment through multi-scale feature fusion, achieving automatic component separation. This provides an independent control dimension for subsequent fusion, ensuring that the enhanced image possesses both high contrast and detail preservation capabilities.

[0087] S140, the first component and the second component are fused to generate an enhanced image corresponding to the original image.

[0088] When fusing the reflection component (first component) and the illuminance component (second component), a progressive illumination update strategy is employed. A self-updating module dynamically adjusts the brightness distribution of the illuminance component to avoid local overexposure or underexposure, while preserving the detailed texture of the reflection component. The fusion process incorporates multi-scale feature weighting to balance global illumination and local contrast, generating an enhanced image. This makes drill pipe edges more prominent and the background more uniform, improving the accuracy of subsequent feature extraction and ensuring the reliability of drill pipe counting in low-light environments.

[0089] In one embodiment of this application, the first component and the second component are fused to generate an enhanced image corresponding to the original image, including:

[0090] The second component is progressively updated with illumination to generate the third component;

[0091] The first component and the third component are fused to generate an enhanced image corresponding to the original image.

[0092] In one embodiment of this application, a third component is generated by progressively updating the illumination of the second component, including:

[0093] Generate a self-updating module formula based on the output components of the current stage;

[0094] The second component is processed based on the self-updating module formula to generate the third component.

[0095] In one embodiment of this application, the second component is progressively updated in terms of illumination to generate a third component. During this process, the self-updating module formula G(X) for generating the second component is... t )for:

[0096]

[0097] Where t≥1, Z t It is the low-light observation image y and the output component X of the current stage. tThe calculated fusion result; S t It is the result of self-calibration mapping, derived from the parameterized operator K. θ Generate, V t It is the input used for the next stage after calibration.

[0098] The second component is updated using a self-updating module formula to generate the third component. During the learning process, a cascaded illumination learning and self-calibration module is used to improve the image quality and sharpness of low-light images in underground coal mines.

[0099] After generating the third component, the first and third components are fused to generate the enhanced image corresponding to the original image. By establishing a connection between the input at each stage and the initial low-light observation, the input is corrected by adding it to the low-light observation values. Through automatic adjustment of network parameters, in-depth analysis of the difference between the output and the target, and intelligent optimization of network weights, the illumination estimation results at each stage converge to the same state. This significantly reduces the computational burden while maintaining performance, making the output closer to the target and improving the accuracy of illumination estimation. This not only improves the performance, accuracy, and stability of the system but also ensures that the output at each stage remains consistent, accelerates the algorithm's convergence speed, improves the overall training effect, and makes the entire system more robust and reliable.

[0100] S150, perform feature extraction on the enhanced image, and determine the number of drill rods in the original image based on the extracted features.

[0101] When extracting features from the enhanced image, a multi-layer convolutional network is used to capture visual features at different scales, such as drill pipe edges, textures, and shapes. Shallow details and deep semantic information are fused through upsampling, downsampling, and skip connections to generate high-order fused features. Candidate detection boxes are generated on the feature map based on the anchor box mechanism. Redundant boxes are filtered using non-maximum suppression. A classifier then determines whether the target within the box is a drill pipe, and a regressor adjusts the position and size of the box. Finally, the number of valid detection boxes is counted. This step utilizes the high contrast and clear edges of the enhanced image to address occlusion and scale variations in the downhole environment, achieving high-precision and low-error real-time drill pipe counting.

[0102] In one embodiment of this application, feature extraction is performed on the enhanced image, and the number of drill rods in the original image is determined based on the extracted features, including:

[0103] Features are extracted from the enhanced image through convolution with a preset number of layers, and feature maps of different scales are output.

[0104] The feature map is upsampled and subsampled, and the sampling results are fused to generate a fused feature;

[0105] Candidate detection boxes are generated on the fused features, and the number of drill pipes is predicted.

[0106] In one embodiment of this application, a pattern recognition-based target detection framework is employed to construct a feature extraction network. The network structure consists of three parts: a backbone network extracts multi-scale features, such as the edges, textures, and shapes of the drill rod, through a preset number of convolutional layers, outputting three feature maps (large, medium, and small scales); a feature fusion layer upsamples and fuses feature maps of different scales to enhance adaptability to changes in the drill rod's scale, such as the representation of a distant drill rod in a small feature map. The detection head generates candidate detection boxes on each fused feature map and predicts the positional offset of the drill rod.

[0107] Multiple predicted bounding boxes for the same drill pipe are filtered, retaining the box with the highest confidence and suppressing other boxes with an overlap exceeding a threshold (e.g., 0.5). The final number of retained predicted bounding boxes is the drill pipe count. The drill pipe count is directly used for downhole equipment monitoring or production statistics, forming a complete technical closed loop. This automates the process from image enhancement to quantity detection, providing reliable technical support for safe coal mine production.

[0108] In this application's technical solution, an original image of a coal mine is acquired using a camera device; based on preset spatial and grayscale factors, the original image is processed by projection to remove noise and generate a first image; the first image is processed using a channel-stitched network structure to extract the relationship between a first component and a second component at different levels, extracting the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image; the first component and the second component are fused to generate an enhanced image corresponding to the original image; feature extraction is performed on the enhanced image, and the number of drill rods in the original image is determined based on the extracted features. By adapting to the harsh underground environment with explosion-proof and dustproof camera devices, and combining low-light optimization to stably acquire original images containing drill pipe information, a clean first image is generated by suppressing noise and preserving edges based on spatial and grayscale factors. The channel stitching network automatically separates the reflection and illumination components, avoiding manual parameter adjustment. After component fusion, a high-contrast enhanced image is generated to highlight the drill pipe edges. Finally, through multi-scale feature extraction and detection, high-precision and low-error real-time counting of drill pipes under low light conditions is achieved, forming a complete closed loop from environmental adaptation to target detection.

[0109] The following describes embodiments of the drill pipe counting device for low-light conditions in coal mines, which can be used to execute the drill pipe counting method for low-light conditions in coal mines described in the above embodiments of this application. It is understood that the drill pipe counting device for low-light conditions in coal mines can be a computer program (including program code) running on a computer device; for example, the drill pipe counting device for low-light conditions in coal mines is an application software. The drill pipe counting device for low-light conditions in coal mines can be used to execute the corresponding steps in the method provided in the embodiments of this application. For details not disclosed in the embodiments of the drill pipe counting device for low-light conditions in coal mines of this application, please refer to the embodiments of the drill pipe counting method for low-light conditions in coal mines described above.

[0110] Figure 3 A block diagram of a drill pipe counting device for use in low-light conditions in underground coal mines, according to an embodiment of this application, is shown.

[0111] Reference Figure 3 As shown, a drill pipe counting device for use in low-light conditions in coal mines according to an embodiment of this application includes:

[0112] Acquisition unit 310 is used to acquire raw images of underground coal mines via a camera device;

[0113] The noise reduction unit 320 is used to perform projection processing on the original image based on a preset spatial factor and grayscale factor to remove noise from the original image and generate a first image.

[0114] The component unit 330 is used to process the first image through a network structure of channel splicing, extract the relationship between the first component and the second component in the first image at different levels, and extract the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image.

[0115] Enhancement unit 340 is used to fuse the first component and the second component to generate an enhanced image corresponding to the original image;

[0116] The counting unit 350 is used to extract features from the enhanced image and determine the number of drill rods in the original image based on the extracted features.

[0117] In this application, based on the aforementioned scheme, the step of performing projection processing on the original image based on preset spatial factors and grayscale factors to remove noise from the original image and generate a first image includes: performing projection processing on the original image to generate a projected image; decomposing the projected image into image components of a preset number of levels; generating projection factors corresponding to each position in the image components based on preset spatial factors and grayscale factors; filtering the high-frequency components in the image components based on the projection factors to generate denoised components; and performing inverse reconstruction on the denoised components to generate the first image.

[0118] In this application, based on the aforementioned scheme, the processing of the first image through a channel-stitched network structure to extract the relationship between the first component and the second component in the first image at different levels, and to extract the first component corresponding to the reflection side and the second component corresponding to the illuminance side from the first image, includes: constructing a channel-stitched network structure based on a preset downsampling module and an upsampling module; processing the first image through the upsampling module in the network structure to generate the first component corresponding to the reflection side; and generating the second component corresponding to the illuminance side through the downsampling module and the upsampling module of the first component branch in the network structure.

[0119] In this application, based on the aforementioned scheme, fusing the first component and the second component to generate the enhanced image corresponding to the original image includes: performing progressive illumination updates on the second component to generate a third component; and fusing the first component and the third component to generate the enhanced image corresponding to the original image.

[0120] In this application, based on the aforementioned scheme, the step of progressively updating the illumination of the second component to generate the third component includes: generating a self-updating module formula based on the output component output at the current stage; and processing the second component based on the self-updating module formula to generate the third component.

[0121] In this application, based on the aforementioned scheme, the step of generating the self-updating module formula based on the output components of the current stage includes: generating the self-updating module formula G(X) based on the output components of the current stage. t )for:

[0122]

[0123] Where t≥1, Z t It is the low-light observation image y and the output component X of the current stage. t The calculated fusion result; S t It is the result of self-calibration mapping, derived from the parameterized operator K. θ Generate, Vt It is the input used for the next stage after calibration.

[0124] In this application, based on the aforementioned scheme, the step of extracting features from the enhanced image and determining the number of drill rods in the original image based on the extracted features includes: extracting features from the enhanced image through convolution with a preset number of layers and outputting feature maps of different scales; upsampling and downsampling the feature maps and fusing the sampling results to generate fused features; generating candidate detection boxes on the fused features and predicting the number of drill rods.

[0125] In this application's technical solution, an original image of a coal mine is acquired using a camera device; based on preset spatial and grayscale factors, the original image is processed by projection to remove noise and generate a first image; the first image is processed using a channel-stitched network structure to extract the relationship between a first component and a second component at different levels, extracting the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image; the first component and the second component are fused to generate an enhanced image corresponding to the original image; feature extraction is performed on the enhanced image, and the number of drill rods in the original image is determined based on the extracted features. By adapting to the harsh underground environment with explosion-proof and dustproof camera devices, and combining low-light optimization to stably acquire original images containing drill pipe information, a clean first image is generated by suppressing noise and preserving edges based on spatial and grayscale factors. The channel stitching network automatically separates the reflection and illumination components, avoiding manual parameter adjustment. After component fusion, a high-contrast enhanced image is generated to highlight the drill pipe edges. Finally, through multi-scale feature extraction and detection, high-precision and low-error real-time counting of drill pipes under low light conditions is achieved, forming a complete closed loop from environmental adaptation to target detection.

[0126] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown.

[0127] It should be noted that the computer system of the electronic device in this embodiment is only an example and should not impose any limitations on the function and scope of use of the embodiments of this application.

[0128] In this embodiment, the computer system includes a central processing unit 401, which can perform various appropriate actions and processes based on a program stored in a read-only memory 402 or a program loaded from a storage section 408 into a random access memory 403, such as executing the drill pipe counting method for low-light conditions in coal mines described in the above embodiment. The random access memory 403 also stores various programs and data required for system operation. The central processing unit 401, the read-only memory 402, and the random access memory 403 are interconnected via a bus 404. An input / output interface 405 is also connected to the bus 404.

[0129] The following components are connected to the input / output interface 405: an input section 406 including a keyboard, mouse, etc.; an output section 407 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output interface 405 as needed. A removable medium 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 410 as needed so that computer programs read from it can be installed into the storage section 408 as needed.

[0130] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit 401, it performs various functions defined in the system of this application.

[0131] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. The transmitted data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0132] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0133] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0134] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various alternative implementations described above.

[0135] In another aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to implement the drill pipe counting method for low-light conditions in coal mines described in the above embodiments.

[0136] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0137] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0138] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0139] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for counting drill pipes in low-light conditions in underground coal mines, characterized in that, include: Raw images of underground coal mines are obtained using camera devices; Based on preset spatial factors and grayscale factors, the original image is subjected to projection processing to remove noise from the original image and generate a first image. The first image is processed by a network structure with channel splicing, and the relationship between the first component and the second component in the first image is extracted at different levels. The first component corresponding to the reflection side and the second component corresponding to the illumination side are extracted from the first image. The first component and the second component are fused to generate an enhanced image corresponding to the original image; Feature extraction is performed on the enhanced image, and the number of drill rods in the original image is determined based on the extracted features; The process of fusing the first component and the second component to generate the enhanced image corresponding to the original image includes: The second component is progressively updated with illumination to generate the third component; The first component and the third component are fused to generate an enhanced image corresponding to the original image; The process of progressively updating the illumination of the second component to generate the third component includes: Generate a self-updating module formula based on the output components of the current stage; The second component is processed based on the self-updating module formula to generate the third component; Among them, the self-updating module formula is generated based on the output components of the current stage, including: Based on the output components of the current stage Generate self-updating module formula for: Where t≥1, Low-light observation image y and the output components of the current stage The fusion results of the calculation; It is a self-calibration mapping result, generated by the parameterized operator. generate, It is the input used for the next stage after calibration.

2. The drill pipe counting method for low-light conditions in coal mines according to claim 1, characterized in that, Based on preset spatial and grayscale factors, the original image is subjected to projection processing to remove noise and generate a first image, including: The original image is subjected to projection processing to generate a projected image, and the projected image is decomposed into image components of a preset number of levels. Based on preset spatial factors and grayscale factors, the projection factors corresponding to each position in the image component are generated. Based on the projection factor, the high-frequency components in the image components are filtered to generate denoised components. The denoised components are then inversely reconstructed to generate the first image.

3. The drill pipe counting method for low-light conditions in coal mines according to claim 1, characterized in that, The first image is processed using a channel-stitched network structure to extract the relationship between the first component and the second component in the first image at different levels. This process extracts the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image, including: A network structure for channel splicing is constructed based on preset downsampling and upsampling modules; The first image is processed by the upsampling module in the network structure to generate the first component corresponding to the reflection side; The second component corresponding to the illuminance side is generated through the downsampling module and upsampling module of the first component branch in the network structure.

4. The drill pipe counting method for low-light conditions in coal mines according to claim 1, characterized in that, Feature extraction is performed on the enhanced image, and the number of drill rods in the original image is determined based on the extracted features, including: Features are extracted from the enhanced image through convolution with a preset number of layers, and feature maps of different scales are output. The feature map is upsampled and subsampled, and the sampling results are fused to generate a fused feature; Candidate detection boxes are generated on the fused features, and the number of drill pipes is predicted.

5. A drill pipe counting device for use in low-light conditions in underground coal mines, characterized in that, include: The acquisition unit is used to acquire raw images of underground coal mines via a camera device; A noise reduction unit is used to perform projection processing on the original image based on a preset spatial factor and grayscale factor to remove noise from the original image and generate a first image. The component unit is used to process the first image through a network structure of channel splicing, extract the relationship between the first component and the second component in the first image at different levels, and extract the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image. An enhancement unit is used to fuse the first component and the second component to generate an enhanced image corresponding to the original image; A counting unit is used to extract features from the enhanced image and determine the number of drill rods in the original image based on the extracted features. The process of fusing the first component and the second component to generate the enhanced image corresponding to the original image includes: The second component is progressively updated with illumination to generate the third component; The first component and the third component are fused to generate an enhanced image corresponding to the original image; The process of progressively updating the illumination of the second component to generate the third component includes: Generate a self-updating module formula based on the output components of the current stage; The second component is processed based on the self-updating module formula to generate the third component; Among them, the self-updating module formula is generated based on the output components of the current stage, including: Based on the output components of the current stage Generate self-updating module formula for: Where t≥1, Low-light observation image y and the output components of the current stage The fusion results of the calculation; It is a self-calibration mapping result, generated by the parameterized operator. generate, It is the input used for the next stage after calibration.

6. The drill pipe counting device for low-light conditions in coal mines according to claim 5, characterized in that, Based on preset spatial and grayscale factors, the original image is subjected to projection processing to remove noise and generate a first image, including: The original image is subjected to projection processing to generate a projected image, and the projected image is decomposed into image components of a preset number of levels. Based on preset spatial factors and grayscale factors, the projection factors corresponding to each position in the image component are generated. Based on the projection factor, the high-frequency components in the image components are filtered to generate denoised components. The denoised components are then inversely reconstructed to generate the first image.

7. The drill pipe counting device for low-light conditions in underground coal mines according to claim 5, characterized in that, The first image is processed using a channel-stitched network structure to extract the relationship between the first component and the second component in the first image at different levels. This process extracts the first component corresponding to the reflection side and the second component corresponding to the illumination side from the first image, including: A network structure for channel splicing is constructed based on preset downsampling and upsampling modules; The first image is processed by the upsampling module in the network structure to generate the first component corresponding to the reflection side; The second component corresponding to the illuminance side is generated through the downsampling module and upsampling module of the first component branch in the network structure.

Images