Monitoring system convenient for demisting in coal mine field

Abstract

The invention discloses a coal mine field monitoring system convenient for demisting, particularly relates to the field of monitoring and flow prediction in a humidity fluctuation environment, is used for solving the problems of picture blurring and flow data distortion caused by high humidity sudden change, and is characterized in that distributed sensors are deployed in a coal mine tunnel environment and a multi-dimensional environment matrix is constructed; after a humidity abrupt change interval is identified by using a differential algorithm, attenuation characteristics of energy efficiency of the demisting device are captured through a multi-scale wavelet and morphological filtering means, and a limit intervention critical value is determined, and then the critical value and a visual ambiguity index of a monitoring picture are subjected to comparison and grading matching; the definition change of pictures before and after defogging intervention is dynamically quantified, finally, a self-adaptive clustering regression mechanism is combined to carry out multi-dimensional deviation correction on hierarchical definition indexes and actual measurement data of a flow sensor, a regression optimization curve is generated, the intervention opportunity of a key time period is calibrated, the defogging efficiency and the flow prediction precision can be comprehensively improved, and the defogging efficiency and the flow prediction accuracy are improved. And the reliability of overall monitoring and scheduling is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of monitoring and flow prediction in a humidity fluctuation environment, and more specifically, to a monitoring system for coal mines that is convenient for demisting. Background Art

[0002] In the coal mine tunnel environment, the humidity fluctuates greatly and the airflow changes are complex. Especially in the rainy season or high humidity environment, a large amount of condensed fog often appears in the tunnel. After the dust particles are superimposed, it will seriously interfere with the monitoring screen, affecting the image clarity and the accuracy of sensor data. Common demisting devices rely on technical means such as thermal dehumidification, chemical adsorption or electrical demisting to maintain the clarity of the monitoring screen, but under drastic humidity changes or continuous high humidity conditions, the efficiency of the device is prone to rapid decay, making it difficult to meet the requirements of real-time monitoring and data accuracy. At the same time, flow prediction is highly dependent on environmental data and image quality. Blurred images and distorted sensor data in high humidity environments will cause a significant increase in the error of the prediction model, affecting the reliability of coal mine scheduling and safety warnings.

[0003] In the existing technology, no effective solution has been proposed for the combined impact of humidity mutation on the defogging device efficiency attenuation and its impact on tunnel flow prediction. Humidity mutation not only directly interferes with the performance of the defogging device, but also affects the blurriness of the monitoring image and the accuracy of the flow sensor data through a chain reaction, resulting in significant deviations in the flow prediction results. How to establish a technical means to dynamically capture the defogging device efficiency attenuation characteristics, synchronously correct the image clarity and flow data errors, and optimize the flow prediction results in real time under humidity mutation environment has become a core issue that needs to be solved urgently.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a coal mine monitoring system that is convenient for demisting. By deploying distributed sensors in a coal mine tunnel environment with frequent humidity fluctuations and constructing a multidimensional environmental matrix, a differential algorithm is used to identify the humidity mutation interval, and then the attenuation characteristics of the energy efficiency of the demisting device are captured by multi-scale wavelet and morphological filtering means to determine the extreme intervention critical value. The critical value is then compared and graded with the visual blur index of the monitoring picture to dynamically quantify the clarity change of the picture before and after the demisting intervention. Finally, an adaptive clustering regression mechanism is combined to perform multi-dimensional deviation correction on the graded clarity index and the measured data of the flow sensor, generate a regression optimization curve and calibrate the intervention timing of the key period, which can comprehensively improve the demisting efficiency and flow prediction accuracy, effectively deal with the problems of severe blurring of the monitoring picture and distortion of the sensor data in a high humidity mutation environment, reduce the flow prediction deviation and enhance the reliability of the overall monitoring and scheduling, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A coal mine monitoring system for demisting, comprising: a matrix analysis module, an attenuation capture module, a fogging segmentation module and a regression optimization module;

[0008] Matrix analysis module: Based on the multi-dimensional environmental data collected by distributed sensors, the humidity mutation sensitivity function is calculated and the humidity mutation boundary matrix is ​​established. The mutation interval is located and dynamically marked, and the multi-dimensional environmental data and the humidity mutation boundary matrix are transferred to the attenuation capture module;

[0009] Attenuation capture module: Combining the multi-dimensional environmental matrix with the humidity mutation boundary matrix, using multi-scale wavelet decomposition and morphological filtering to extract the performance attenuation characteristics of the defogging device at the humidity inflection point and train the attenuation model, output the extreme intervention critical value, and pass the extreme intervention critical value to the fogging segmentation module;

[0010] Fog classification module: Based on the relationship between the limit intervention critical value and the visual blur index, the degree of fogging of the picture is graded with the help of the virtual optical matching algorithm, and the clarity change amplitude is fed back to the attenuation model correction parameter set to form a closed-loop judgment, and the clarity index after graded processing is passed to the regression optimization module;

[0011] Regression optimization module: Based on the multidimensional regression model of support vector regression, the comprehensive evaluation index and clarity index are jointly analyzed and nonlinear deviation correction is performed to generate the airflow optimization curve and calibrate the defog intervention timing.

[0012] In a preferred embodiment, the matrix analysis module includes the following contents:

[0013] Use distributed sensors to collect environmental data; output all sensors with node number N i Construct a multidimensional environment matrix M for the index t =[T i ,H i ,F i ,P i ], where T i , H i 、F i , P i Represents node N respectively i Temperature, humidity, air flow and dust concentration at time t; through humidity data H i The difference operation ΔH i =H i (t)-H i (t-1), extract humidity change rate R H, and combined with the linkage trend of air flow data and dust concentration data, the humidity mutation sensitivity function is calculated: Where, ∈ is a small value to avoid the denominator being zero, P max is the maximum dust concentration reference value, which is used for normalization processing; when the sensitivity function value exceeds the preset threshold, the mutation area mark will be triggered and the time t of the mutation point will be recorded. k and position N k , and stored as the humidity mutation boundary matrix

[0014] In a preferred embodiment, the attenuation capture module includes the following:

[0015] First, in the multidimensional environment matrix M t In the humidity dimension, select the humidity mutation boundary matrix Recorded in [N k , t k ] area adjacent to the time segment, for each position N k The humidity series H k (t) performs multi-scale wavelet transform, denoted as in Represents the position N at different scales α k Time-frequency decomposition of humidity series, ψ represents the mother wavelet function, τ is the integral variable, and * represents the complex conjugate operation;

[0016] Next, morphological filtering is applied to the wavelet decomposition results at each position to remove random pulse interference and false peaks, and the filtering results are recorded as Then, define the performance decay function

[0017] Then, each position before and after the inflection point A k (t,α) The peak value, valley value and occurrence time are included in the attenuation model training set D a , according to the attenuation section corresponding to each position before and after the humidity mutation, the curve fitting is performed to extract the energy efficiency drop trigger point and attenuation slope, and the fitting results are stored in the behavior curve matrix B e =[n k ,α,A max ,A min ,onset,duration], where n k Indicates the position number, A max , A min Indicates the maximum and minimum performance attenuation values ​​of the corresponding period, onset is the start time of the sudden drop, and duration is the duration of the attenuation;

[0018] Finally, according to the attenuation characteristics of each position in the behavior curve matrix under different high humidity environments, the limit intervention critical value E is extracted. lim , that is, the maximum attenuation amplitude and the shortest attenuation duration that the demisting device can withstand in an environment with sudden humidity changes.

[0019] In a preferred embodiment, the attenuation capture module further includes the following contents:

[0020] According to the behavior curve matrix, extract the limit intervention critical value E lim The process is based on the joint analysis of the performance decay law of multiple nodes and environmental characteristics. The specific method is as follows: First, the performance decay amplitude A of each position in the behavior curve matrix is ​​calculated. max -A min The attenuation duration is classified into high attenuation area and low attenuation area according to the time of the mutation point. The high attenuation area meets A max >γ1 and duration>γ2, where γ1 and γ2 are empirical thresholds, and vice versa is a low attenuation area; in the high attenuation area, select the key position N with the largest attenuation amplitude max , extract the start time of the sudden drop and the duration of the attenuation from the behavior curve at the corresponding position, and calculate the time offset gradient of the performance sudden drop To evaluate the drop speed, normalize the frequency components and the maximum performance attenuation of each location in the same area and define the frequency response adaptability It is used to quantify the performance adaptability within the area; then, the maximum values ​​of the time offset gradient and the frequency response adaptability are calculated one by one for the behavior curves of all positions in the high attenuation area, and the weighted sum of these two maximum values ​​is taken to obtain the extreme intervention critical value.

[0021] In a preferred embodiment, the atomization and segmentation module includes the following contents:

[0022] S3.1, first, extract the real-time visual fuzziness index V(t) from the monitoring picture. The visual fuzziness index is used to quantify the degree of fuzziness of the monitoring picture. The calculation process is as follows:

[0023] First, the real-time monitoring image I(t) is converted into a grayscale image I gray (t), then, applying the Laplacian operator Perform edge detection on the grayscale image to generate an edge image E(t); next, perform binarization on the edge image, set a threshold T to mark pixels with edge strength higher than the threshold T as edge points, and generate a binary edge image E binary (t); Then, calculate the total number of edge pixels N in the binary edge image edges(t); Finally, the total number of edge pixels is normalized to the visual blur index V(t): Where W and H are the width and height of the image respectively.

[0024] In a preferred embodiment, S3.2, the limit intervention threshold E output by the attenuation model is lim Compare with the visual blur index V(t) in real time. When V(t)≥E lim When the virtual optical matching algorithm O is triggered M The virtual optical matching algorithm processes the monitoring images according to the degree of fogging and generates a graded fog image I through a preset optical model. graded (t), the specific implementation process is as follows:

[0025] A, first, define the optical model to simulate the interaction between light and fog in the tunnel environment; second, pre-process the real-time monitoring image and apply a high-pass filter H f Remove low-frequency background noise and enhance image edges and details: I pre (x,y,t)=H f *t(x,y,t); where I pre (x, y, t) is the preprocessed image, * indicates the convolution operation, H f is the high-pass filter kernel; then, the optical model is used to perform pixel-by-pixel optical matching on the preprocessed image to calculate the fogging degree score S(x,y,t) of each pixel point (x,y), and the formula is as follows: Among them, σ is the scattering cross section, ρ is the density of fog particles, and d is the optical path length; then, multiple fog level thresholds are set The fogging degree score is divided into different fogging levels L according to the fogging level threshold range. k ; Then, generate the graded fog image I graded (t).

[0026] In a preferred embodiment, B, by comparing the image clarity before and after the intervention, the clarity change amplitude ΔC(t) is defined as: ΔC(t)=C post (t)-C pre (t); where C post (t) and C pre (t) respectively represent the image clarity index before and after the intervention of the defogging device, and the calculation method adopts the gradient mean algorithm: C Where W and H are the width and height of the image, respectively, and I(x,y) is the image pixel value. and is the gradient of the image in the x and y directions;

[0027] C, the calculated clarity change amplitude is used as the correction parameter to update the parameter set Θ of the attenuation model. The specific update rule is: Θ new =Θ old +ηΔC(t); where η is the learning rate;

[0028] D. By feeding back the clarity change amplitude to the attenuation model, the operation strategy of the defogger device is adjusted in real time to ensure the continuous optimization of the defogger efficiency in the case of sudden humidity changes.

[0029] In a preferred embodiment, the regression optimization module includes the following contents:

[0030] First, obtain the clarity index matrix C after hierarchical processing graded (t) and actual flow data measured by flow sensor Take the two as input data sets; then calculate the airflow fluctuation index It analyzes the instantaneous rate of change of actual flow data It turns out that: Among them, ∈1 is a small constant to avoid the denominator being zero; then, the visual fuzziness index V(t) is fused with the airflow fluctuation index to generate a comprehensive evaluation index I combined (t): Next, an adaptive clustering algorithm was used to perform cluster analysis on the comprehensive evaluation index and clarity index to identify data subsets D with similar characteristics at different humidity mutation stages. v = {I combined (t),C graded (t)}, where v represents the cluster number.

[0031] In a preferred embodiment, the regression optimization module further includes the following contents:

[0032] Then, for each cluster subset, a multidimensional regression model was constructed The nonlinear regression method is used to fit the relationship between the comprehensive evaluation index and the clarity index. The specific form is: in, is the corrected traffic prediction value; this regression model achieves the initial traffic prediction value by learning the feature relationship in different cluster subsets Multi-dimensional deviation correction to generate optimized prediction curve The calculation formula is as follows: in, Represents the flow deviation correction amount predicted by the multidimensional regression model; then, the matching degree between the optimized prediction curve and the actual flow data is analyzed, and the error index is defined When the error index exceeds the corresponding preset threshold, the intervention operation of the defogger device is triggered.

[0033] The technical effects and advantages of the coal mine monitoring system convenient for demisting of the present invention are as follows:

[0034] By deploying distributed sensors in a coal mine tunnel environment with frequent humidity fluctuations and constructing a multidimensional environmental matrix, the differential algorithm is used to identify the humidity mutation interval, and then the attenuation characteristics of the energy efficiency of the defogging device are captured through multi-scale wavelet and morphological filtering methods to determine the extreme intervention critical value. The critical value is then compared and graded with the visual blur index of the monitoring image to dynamically quantify the clarity change of the image before and after the defogging intervention. Finally, the adaptive clustering regression mechanism is combined to perform multidimensional deviation correction on the graded clarity index and the measured data of the flow sensor, generate a regression optimization curve and calibrate the intervention timing of the key period, which can comprehensively improve the defogging efficiency and flow prediction accuracy, effectively deal with the serious blurring of the monitoring image and the distortion of the sensor data in the high humidity mutation environment, reduce the flow prediction deviation and enhance the reliability of the overall monitoring and scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The present invention is a schematic structural diagram of a coal mine monitoring system that is convenient for demisting. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Embodiment 1: Figure 1 The present invention provides a coal mine monitoring system that is convenient for demisting, including: a matrix analysis module, an attenuation capture module, a fogging segmentation module and a regression optimization module;

[0038] Matrix analysis module: Based on the multi-dimensional environmental data collected by distributed sensors, the humidity mutation sensitivity function is calculated and the humidity mutation boundary matrix is ​​established. The mutation interval is located and dynamically marked, and the multi-dimensional environmental data and the humidity mutation boundary matrix are transferred to the attenuation capture module;

[0039] Attenuation capture module: Combining the multi-dimensional environmental matrix with the humidity mutation boundary matrix, using multi-scale wavelet decomposition and morphological filtering to extract the performance attenuation characteristics of the defogging device at the humidity inflection point and train the attenuation model, output the extreme intervention critical value, and pass the extreme intervention critical value to the fogging segmentation module;

[0040] Fog classification module: Based on the relationship between the limit intervention critical value and the visual blur index, the degree of fogging of the picture is graded with the help of the virtual optical matching algorithm, and the clarity change amplitude is fed back to the attenuation model correction parameter set to form a closed-loop judgment, and the clarity index after graded processing is passed to the regression optimization module;

[0041] Regression optimization module: Based on the multidimensional regression model of support vector regression, the comprehensive evaluation index and clarity index are jointly analyzed and nonlinear deviation correction is performed to generate the airflow optimization curve and calibrate the defog intervention timing.

[0042] Humidity fluctuations in coal mine tunnels are frequent and large, and environmental factors are complex and changeable. Especially in the case of sudden humidity changes, conventional monitoring systems are difficult to accurately capture the specific location and time of the sudden change area, resulting in the inability to effectively evaluate and correct the efficiency attenuation of the demisting device. At the same time, humidity changes will have a nonlinear effect on dust diffusion and air flow, further exacerbating the error in flow prediction. To address these issues, it is necessary to build a basic data structure that can dynamically calibrate the boundaries of humidity mutations through precise data monitoring and mutation point identification technology, laying the foundation for subsequent analysis of demisting performance attenuation and flow prediction optimization.

[0043] The mutation caused by humidity fluctuation has a direct impact on the operating efficiency of monitoring equipment. The purpose of this module is to accurately identify the initial demarcation point of the sudden change of humidity in the tunnel environment by establishing a humidity mutation demarcation matrix, so as to provide a clear positioning basis for the subsequent demisting efficiency analysis and optimization.

[0044] The matrix analysis module includes the following:

[0045] In the tunnel environment, distributed sensors are used to collect environmental data such as temperature, humidity, air flow, and dust content. The output of each sensor is stored in the form of a time series. All sensor outputs are numbered as nodes N. i Construct a multidimensional environment matrix M for the index t =[T i ,H i ,F i ,P i ], where T i , H i 、F i , P i Represents node N respectively i Temperature, humidity, air flow and dust concentration at time t. i The difference operation ΔH i =H i (t)-H i (t-1), extract humidity change rate R H, and combined with the linkage trend of air flow data and dust concentration data, the humidity mutation sensitivity function is calculated: Where, ∈ is a small value to avoid the denominator being zero, P max is the maximum dust concentration reference value, which is used for normalization processing. When the sensitivity function value exceeds the preset threshold, the mutation area mark will be triggered and the time t of the mutation point will be recorded. k and position N k , and stored as the humidity mutation boundary matrix matrix It will serve as the core input data for subsequent analysis of humidity mutation characteristics.

[0046] Humidity mutation boundary matrix The construction realizes the precise calibration and spatiotemporal positioning of the mutation interval, ensuring that the subsequent modules can directly perform energy efficiency attenuation analysis and behavior curve training on sensitive areas, avoiding data redundancy and positioning deviation in the processing process.

[0047] In the humidity mutation boundary matrix generated in the previous step, the spatiotemporal coordinates of the rapid humidity change [N k , t k ]. The goal of the current module is to retrieve the attenuation law of the energy efficiency of the defogger device as the humidity inflection point appears around the above coordinates based on the multi-dimensional environmental matrix, laying a data support for the subsequent clarity grading processing and dynamic closed-loop judgment.

[0048] The decay capture module includes the following:

[0049] First, in the multidimensional environment matrix M t In the humidity dimension, select the humidity mutation boundary matrix Recorded in [N k , t k ] area adjacent to the time segment, for each position N k The humidity series H k (t) performs multi-scale wavelet transform, denoted as in Represents the position N at different scales α k Time-frequency decomposition of humidity series, ψ represents the mother wavelet function, τ is the integral variable, and * represents the complex conjugate operation.

[0050] Next, morphological filtering is applied to the wavelet decomposition results at each position to remove random pulse interference and false peaks, and the filtering results are recorded as Then, define the performance decay function It is used to measure the fluctuation range of energy efficiency of the defogger during periods of high humidity changes. The larger the value, the more obvious the attenuation.

[0051] Then, each position before and after the inflection point A k (t,α) The peak value, valley value and occurrence time are included in the attenuation model training set D a , according to the attenuation section corresponding to each position before and after the humidity mutation, the curve fitting is performed to extract the energy efficiency drop trigger point and attenuation slope, and the fitting results are stored in the behavior curve matrix B e =[n k ,α,A max ,A min ,onset,duration], where n k Indicates the position number, A max , A min Indicates the maximum and minimum performance attenuation values ​​in the corresponding period, onset is the start time of the sudden drop, and duration is the duration of the attenuation.

[0052] Finally, according to the behavior curve matrix B e The attenuation characteristics of each position in different high humidity environments are used to extract the limit intervention critical value E lim , that is, the maximum attenuation amplitude and the shortest attenuation duration that the demisting device can withstand in an environment with sudden humidity changes. This critical value will be passed to the atomization segmentation module for real-time monitoring and graded matching of the demisting efficiency.

[0053] According to the behavior curve matrix, extract the limit intervention critical value E lim The process is based on the joint analysis of the performance decay law of multiple nodes and environmental characteristics. The specific method is as follows: First, the performance decay amplitude A of each position in the behavior curve matrix is ​​calculated. max -A min The attenuation duration is classified and the grouping is carried out according to the time of the mutation point as the center, and the high attenuation area (satisfying A max >γ1 and duration>γ2, where γ1 and γ2 are empirical thresholds) and low attenuation areas (areas that do not meet the above conditions); in the high attenuation area, select the key position N with the largest attenuation amplitude max , extract the start time of the sudden drop and the duration of the attenuation from the behavior curve at the corresponding position, and calculate the time offset gradient of the performance sudden drop To evaluate the sudden drop speed, the frequency component α and the maximum performance attenuation A of each location in the same area are calculated. max Perform normalization analysis and define frequency response fitness It is used to quantify the performance adaptability in the area; then, the maximum values ​​of the time offset gradient and frequency response adaptability are calculated one by one for the behavior curves of all positions in the high attenuation area, which are respectively denoted as G t,max and F resp,max , take the weighted sum of these two maximum values ​​and obtain the limit intervention critical value: Elim =β1G t,max +β2F resp,max ; Among them, β1 and β2F resp,max The importance factor is dynamically adjusted according to the actual scenario, reflecting the influence weight of time response and frequency response in high humidity environment. Through this method, the dynamic characteristics in the behavior curve can be quantified into key performance limit values, and high priority locations and times that require real-time intervention can be marked, providing a direct basis for dynamic monitoring and hierarchical processing of subsequent modules.

[0054] Through the time-frequency analysis method combining multi-scale wavelet decomposition and morphological filtering, the attenuation behavior of the energy efficiency of the defogger near the humidity sudden change point is accurately characterized, and the attenuation section characteristics are recorded in the behavior curve matrix according to the curve fitting results, providing a quantifiable reference for subsequent clarity grading enhancement and dynamic closed-loop judgment. The performance sudden drop threshold triggered by high humidity is confirmed in this module, laying a key data foundation for further reducing the flow prediction deviation caused by insufficient defogger efficiency.

[0055] In the previous step, the performance attenuation characteristics of the defogging device at different humidity inflection points were obtained through multi-scale wavelet decomposition and morphological filtering analysis, and the critical value of extreme intervention was established. The goal of this module is to compare this critical value with the visual blur index in the real-time monitoring picture, and to grade the degree of picture fogging through the virtual optical matching algorithm, so as to quantify the change in clarity before and after the intervention of the defogging device. This process requires not only real-time monitoring and evaluation of picture quality, but also feedback of clarity changes back to the attenuation model to achieve dynamic closed-loop judgment of defogging efficiency.

[0056] The atomization subdivision module includes the following:

[0057] S3.1, first, extract the real-time visual fuzziness index V(t) from the monitoring picture. The visual fuzziness index is used to quantify the degree of fuzziness of the monitoring picture. The calculation process is as follows:

[0058] First, the real-time monitoring image I(t) is converted into a grayscale image I gray (t) to simplify the subsequent module processing. Then, the Laplacian operator is applied Perform edge detection on the grayscale image to generate the edge image E(t): The Laplacian operator highlights the edges and details in the image by calculating the second-order derivative of the image grayscale value.

[0059] Next, the edge image is binarized, and a threshold T is set to mark the pixels with edge strength higher than the threshold T as edge points, generating a binary edge image E. binary (t):

[0060] Where (x, y) represents the pixel coordinates of the image, and |E(t)(x, y)| is the absolute edge strength output by the Laplace operator.

[0061] Then, the total number of edge pixels N in the binary edge image is calculated edges (t): Where H and H are the width and height of the image respectively.

[0062] Finally, the total number of edge pixels is normalized to the visual blur index V(W):

[0063] The value range of the visual blur index is between 0 and 1. The higher the value, the more edge information there is in the image and the clearer the picture is. Conversely, the lower the value, the more blurred the image is.

[0064] S3.2, the extreme intervention threshold E output by the attenuation model lim Compare with the visual blur index V(t) in real time to determine whether the current working state of the defogger has reached the critical point requiring intervention. lim When the virtual optical matching algorithm O is triggered M The virtual optical matching algorithm processes the monitoring images according to the degree of fogging and generates a graded fog image I through a preset optical model. graded (t). The specific implementation process is as follows:

[0065] A,first, an optical model is defined to simulate the interaction between light and fog in the tunnel environment.,The model is based on the radiation transfer theory and considers the scattering and,absorption effects of light, which are specifically manifested as light intensity attenuation and,direction change.

[0066] Secondly, the real-time monitoring image is pre-processed and a high-pass filter H is applied. f Remove low-frequency background noise and enhance image edges and details: I pre (x,y,t)=H f *I(x,y,t); where, I pre (x, y, t) is the preprocessed image, * indicates the convolution operation, H f is the high pass filter kernel.

[0067] Next, the optical model is used to perform pixel-by-pixel optical matching on the preprocessed image to calculate the fogging degree score S(x,y,t) of each pixel point (x,y). The formula is as follows: Among them, σ is the scattering cross section, ρ is the density of fog particles, and d is the optical path length. The larger the fogging degree score, the higher the degree of fogging in the area.

[0068] Then, set multiple atomization level thresholds The fogging degree score is divided into different fogging levels L according to the fogging level threshold range. k :

[0069]

[0070] Each level L k Corresponding to different degrees of fogging severity, such as light fog, medium fog, heavy fog, etc.

[0071] Then, the graded fog image I is generated graded (t), where the value of each pixel is determined by its corresponding fog level. Specifically, different levels are assigned different colors or grayscale values ​​to intuitively display the degree of fog in each area:

[0072]

[0073] Among them, Color k Predefined color codes corresponding to the level of atomization.

[0074] B. By comparing the image clarity before and after the intervention, the clarity change amplitude ΔC(t) is defined as: ΔC(t) = C post (t)-C pre (t); where C post (t) and C pre (t) represents the image clarity index before and after the intervention of the defogging device, and the calculation method adopts the gradient mean algorithm: Where W and H are the width and height of the image, respectively, and I(x,y) is the image pixel value. and is the gradient of the image in the x and y directions.

[0075] C, the calculated clarity change amplitude is used as the correction parameter to update the parameter set Θ of the attenuation model. The specific update rule is: Θ new =Θ old +ηΔC(t); where η is the learning rate, which is used to control the pace of parameter updating to ensure the stability and adaptability of the model.

[0076] D. By feeding back the clarity change amplitude to the attenuation model, the operation strategy of the defogger is adjusted in real time to ensure the continuous optimization of the defogger efficiency under the condition of sudden humidity changes. The dynamic closed-loop process includes real-time evaluation of the clarity of the monitoring screen, adjustment of the defogger status, and dynamic update of the attenuation model parameters, forming an adaptive defogger efficiency determination mechanism.

[0077] By comparing the extreme intervention threshold with the real-time visual blur index, the grading of the degree of fogging and the quantification of the clarity change of the picture were successfully achieved. The application of the virtual optical matching algorithm ensures the accurate grading of the degree of fogging of the picture, and the feedback mechanism of the clarity change amplitude enables the attenuation model to dynamically correct the parameters according to the actual intervention effect, thus forming a closed-loop judgment process for the defogging efficiency. This process not only improves the clarity of the monitoring picture, but also improves the responsiveness of the defogging device and the reliability of the overall monitoring system through the dynamic feedback mechanism.

[0078] After comparing the extreme intervention critical value with the real-time visual blur index through the virtual optical matching algorithm in the previous step, and realizing the graded processing of the image fogging degree and the quantification of the clarity change, this module aims to further optimize the accuracy of airflow flow prediction. By introducing an adaptive clustering regression mechanism, combining the clarity index after graded processing with the measured data of the flow sensor, the airflow flow prediction value is corrected in multiple dimensions, and a regression optimization curve is generated at different humidity mutation stages, and the defogging intervention timing in key periods is calibrated. This process ensures that in a high humidity mutation environment, the monitoring system can accurately identify and respond to deviations in flow prediction, thereby improving the reliability of overall monitoring and scheduling.

[0079] The regression optimization module includes the following:

[0080] First, obtain the clarity index matrix C after hierarchical processing graded (t) and actual flow data measured by flow sensor Take the two as input data sets. Next, calculate the airflow fluctuation index It analyzes the instantaneous rate of change of actual flow data It turns out that: Among them, ∈1 is a small constant to avoid the denominator being zero. The airflow fluctuation index reflects the fluctuation amplitude of the actual airflow flow. The larger the value, the more severe the flow fluctuation.

[0081] Then, the visual fuzziness index V(t) is fused with the airflow fluctuation index to generate a comprehensive evaluation index I combined (t):

[0082] Next, an adaptive clustering algorithm is applied to the comprehensive evaluation index I combined (t) and clarity index C graded (t) Perform cluster analysis to identify data subsets D with similar characteristics at different humidity mutation stages v = {I combined (t),C graded(t)}, where v represents the cluster number. By dynamically adjusting the clustering parameters, it is ensured that each cluster subset can accurately reflect the clarity and flow fluctuation characteristics under a specific humidity mutation stage.

[0083] Then, for each cluster subset, a multidimensional regression model was constructed The nonlinear regression method is used to fit the relationship between the comprehensive evaluation index and the clarity index. The specific form is: in, is the corrected traffic prediction value. This regression model achieves the initial traffic prediction value by learning the feature relationship in different cluster subsets. Multi-dimensional deviation correction to generate optimized prediction curve The calculation formula is as follows: in, Represents the flow deviation correction amount predicted by the multidimensional regression model.

[0084] Then, analyze the matching degree between the optimized prediction curve and the actual flow data and define the error index When the error index exceeds the corresponding preset threshold, the intervention operation of the defogger device is triggered.

[0085] Finally, the optimization results of different cluster subsets are integrated to form a global regression optimization strategy to ensure accurate flow prediction correction and timely defogging intervention in various humidity mutation scenarios.

[0086] Among them, the multidimensional regression model The construction of a nonlinear regression model based on support vector regression is specifically used to fit the relationship between the comprehensive evaluation index and the graded clarity index on the corrected traffic prediction value. Specifically, for each cluster subset D v =(I combined (t),C graded (t)), the model first transforms the input feature (I combined (t),C graded (t)) is mapped to a high-dimensional feature space to capture the complex nonlinear relationships between input variables. v The actual flow correction value in is used as the target variable, and the parameters of the support vector regression model, including the penalty parameter and the kernel parameter, are optimized by minimizing the objective function of the prediction error and the model complexity. After the training is completed, the multidimensional regression model can accurately predict the flow correction amount according to the new input, thereby generating an optimized flow prediction value. The multidimensional regression model achieves high-precision correction of airflow flow prediction by finely fitting the features under different humidity mutation stages, ensuring the reliability and response speed of the monitoring system under various complex environmental conditions.

[0087] Through the adaptive clustering regression mechanism, combined with the clarity index and airflow fluctuation index after hierarchical processing, the multi-dimensional deviation correction of the airflow flow prediction value was successfully achieved, and regression optimization curves for different humidity mutation stages were generated. This process not only significantly improved the accuracy of flow prediction, but also ensured the efficient response capability of the monitoring system in a high humidity mutation environment by calibrating the timing of defog intervention during key periods, further enhancing the reliability and real-time performance of the overall monitoring and scheduling.

[0088] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0089] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0090] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0091] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A coal mine monitoring system that is convenient for demisting, characterized in that: include: Matrix analysis module, attenuation capture module, atomization segmentation module and regression optimization module; Matrix analysis module: Based on the multi-dimensional environmental data collected by distributed sensors, the humidity mutation sensitivity function is calculated and the humidity mutation boundary matrix is ​​established. The mutation interval is located and dynamically marked, and the multi-dimensional environmental data and the humidity mutation boundary matrix are transferred to the attenuation capture module; Attenuation capture module: Combining the multi-dimensional environmental matrix with the humidity mutation boundary matrix, using multi-scale wavelet decomposition and morphological filtering to extract the performance attenuation characteristics of the defogging device at the humidity inflection point and train the attenuation model, output the extreme intervention critical value, and pass the extreme intervention critical value to the fogging segmentation module; Fog classification module: Based on the relationship between the limit intervention critical value and the visual blur index, the degree of fogging of the picture is graded with the help of the virtual optical matching algorithm, and the clarity change amplitude is fed back to the attenuation model correction parameter set to form a closed-loop judgment, and the clarity index after graded processing is passed to the regression optimization module; Regression optimization module: Based on the multidimensional regression model of support vector regression, the comprehensive evaluation index and clarity index are jointly analyzed and nonlinear deviation correction is performed to generate the airflow optimization curve and calibrate the defog intervention timing.

2. A coal mine monitoring system for demisting according to claim 1, characterized in that: The matrix analysis module includes the following: Use distributed sensors to collect environmental data; output all sensors with node number N i Construct a multidimensional environment matrix M for the index t =[T i ,H i ,F i ,P i ], where T i , H i 、F i , P i Represents node N respectively i Temperature, humidity, air flow and dust concentration at time t; through humidity data H i The difference operation ΔH i =H i (t)-H i (t-1), extract humidity change rate R H , and combined with the linkage trend of air flow data and dust concentration data, the humidity mutation sensitivity function is calculated: Among them, ∈ is to avoid small values ​​of the denominator being zero, P max It is the reference value of the maximum dust concentration, used for normalization; When the sensitivity function value exceeds the preset threshold, the mutation area mark will be triggered and the time t of the mutation point will be recorded. k and position N k , and stored as the humidity mutation boundary matrix 3. A coal mine monitoring system for demisting according to claim 2, characterized in that: The decay capture module includes the following: First, in the multidimensional environment matrix M t In the humidity dimension, select the humidity mutation boundary matrix Recorded in [N k , t k ] area adjacent to the time segment, for each position N k The humidity series H k (t) performs multi-scale wavelet transform, denoted as in Represents the position N at different scales α k Time-frequency decomposition of humidity series, ψ represents the mother wavelet function, τ is the integral variable, and * represents the complex conjugate operation; Next, morphological filtering is applied to the wavelet decomposition results at each position to remove random pulse interference and false peaks, and the filtering results are recorded as Then, define the performance decay function Then, each position before and after the inflection point A k (t,α) The peak value, valley value and occurrence time are included in the attenuation model training set D a , according to the attenuation section corresponding to each position before and after the humidity mutation, the curve fitting is performed to extract the energy efficiency drop trigger point and attenuation slope, and the fitting results are stored in the behavior curve matrix B e =[n k ,α,A max ,A min ,onset,duration], where n k Indicates the position number, A max , A min Indicates the maximum and minimum performance attenuation values ​​of the corresponding period, onset is the start time of the sudden drop, and duration is the duration of the attenuation; Finally, according to the attenuation characteristics of each position in the behavior curve matrix under different high humidity environments, the limit intervention critical value E is extracted. lim , that is, the maximum attenuation amplitude and the shortest attenuation duration that the demisting device can withstand in an environment with sudden humidity changes.

4. A coal mine monitoring system for demisting according to claim 3, characterized in that: The decay capture module also Includes the following: According to the behavior curve matrix, extract the limit intervention critical value E lim The process is based on the joint analysis of the performance decay law of multiple nodes and environmental characteristics. The specific method is as follows: First, the performance decay amplitude A of each position in the behavior curve matrix is ​​calculated. max -A min The attenuation duration is classified into high attenuation area and low attenuation area according to the time of the mutation point. The high attenuation area meets A max >γ1 and duration>γ2, where γ1 and γ2 are empirical thresholds, and vice versa is a low attenuation area; in the high attenuation area, select the key position N with the largest attenuation amplitude max , extract the start time of the sudden drop and the duration of the attenuation from the behavior curve at the corresponding position, and calculate the time offset gradient of the performance sudden drop To evaluate the drop speed, normalize the frequency components and the maximum performance attenuation of each location in the same area and define the frequency response adaptability Used to quantify performance adaptability within a region; Then, the maximum values ​​of the time offset gradient and the frequency response fitness are calculated one by one for the behavior curves of all positions in the high attenuation area, and the weighted sum of these two maximum values ​​is taken to obtain the extreme intervention critical value.

5. A coal mine monitoring system for demisting according to claim 3, characterized in that: The atomization subdivision module includes the following: S3.1, first, extract the real-time visual fuzziness index V(t) from the monitoring picture. The visual fuzziness index is used to quantify the degree of fuzziness of the monitoring picture. The calculation process is as follows: First, the real-time monitoring image I(t) is converted into a grayscale image I gray (t), then, applying the Laplacian operator Perform edge detection on the grayscale image to generate an edge image E(t); next, perform binarization on the edge image, set a threshold T to mark pixels with edge strength higher than the threshold T as edge points, and generate a binary edge image E binary (t); Then, calculate the total number of edge pixels N in the binary edge image edges (t); Finally, the total number of edge pixels is normalized to the visual blur index V(t): Where W and H are the width and height of the image respectively.

6. A coal mine monitoring system for demisting according to claim 5, characterized in that: S3.2, the extreme intervention threshold E output by the attenuation model lim Compare with the visual blur index V(t) in real time. When V(t)≥E lim When the virtual optical matching algorithm O is triggered M The virtual optical matching algorithm processes the monitoring images according to the degree of fogging and generates a graded fog image I through a preset optical model. graded (t), the specific implementation process is as follows: A, first, define the optical model to simulate the interaction between light and fog in the tunnel environment; second, pre-process the real-time monitoring image and apply a high-pass filter H f Remove low-frequency background noise and enhance image edges and detail features: I pre (x,y,t)=H f *I(x,y,t); where, I pre (x, y, t) is the preprocessed image, * indicates the convolution operation, H f is the high-pass filter kernel; then, the optical model is used to perform pixel-by-pixel optical matching on the preprocessed image to calculate the fogging degree score S(x,y,t) of each pixel point (x,y), and the formula is as follows: Among them, σ is the scattering cross section, ρ is the density of fog particles, and d is the optical path length; then, multiple fog level thresholds {T1, T2, …, T n }, the atomization degree score is divided into different atomization levels L according to the atomization level threshold range k ; Then, generate the graded fog image I graded (t).

7. A coal mine monitoring system for demisting according to claim 6, characterized in that: B. By comparing the image clarity before and after the intervention, the clarity change amplitude ΔC(t) is defined as: ΔC(t) = C post (t)-C pre (t); where C post (t) and C pre (t) represents the image clarity index before and after the intervention of the defogging device, and the calculation method adopts the gradient mean algorithm: Where W and H are the width and height of the image, respectively, and I(x,y) is the image pixel value. and is the gradient of the image in the x and y directions; C, the calculated clarity change amplitude is used as the correction parameter to update the parameter set Θ of the attenuation model. The specific update rule is: Θ new =Θ old +ηΔC(t); where η is the learning rate; D. By feeding back the clarity change amplitude to the attenuation model, the operation strategy of the defogger device is adjusted in real time to ensure the continuous optimization of the defogger efficiency in the case of sudden humidity changes.

8. A coal mine monitoring system for demisting according to claim 7, characterized in that: The regression optimization module includes the following: First, obtain the clarity index matrix C after hierarchical processing graded (t) and actual flow data measured by flow sensor Take the two as input data sets; then calculate the airflow fluctuation index It analyzes the instantaneous rate of change of actual flow data It turns out that: Among them, ∈1 is a small constant to avoid the denominator being zero; then, the visual fuzziness index V(t) is fused with the airflow fluctuation index to generate a comprehensive evaluation index I combined (t): Next, an adaptive clustering algorithm was used to perform cluster analysis on the comprehensive evaluation index and clarity index to identify data subsets D with similar characteristics at different humidity mutation stages. v = {I combined (t),C graded (t)}, where v represents the cluster number.

9. A coal mine monitoring system for demisting according to claim 8, characterized in that: The regression optimization module also includes the following: Then, for each cluster subset, a multidimensional regression model was constructed The nonlinear regression method is used to fit the relationship between the comprehensive evaluation index and the clarity index. The specific form is: C graded (t)); where is the corrected traffic prediction value; this regression model achieves the initial traffic prediction value by learning the feature relationship in different cluster subsets Multi-dimensional deviation correction to generate optimized prediction curve The calculation formula is as follows: in, Represents the flow deviation correction amount predicted by the multidimensional regression model; then, the matching degree between the optimized prediction curve and the actual flow data is analyzed, and the error index is defined When the error index exceeds the corresponding preset threshold, the intervention operation of the defogger device is triggered.

Images