Coal mine AI intelligent video recognition system and use method

Abstract

The present application relates to the technical field of coal mine video recognition, and particularly relates to a coal mine AI intelligent video recognition system and a use method. The coal mine AI intelligent video recognition system comprises a data acquisition module, which acquires video frame data of a coal mine tunnel, structural data of the coal mine tunnel and tunnel environment data; a coal dust distribution module, which processes the video frame data and the structural data of the coal mine tunnel, and constructs a coal dust concentration distribution model of a tunnel space; a contact angle calculation module, which processes the video frame data, and calculates a coal kind hydrophilic contact angle of coal dust in the coal mine tunnel; and a fog droplet particle size solving module, which solves an optimal dust-settling fog droplet particle size according to the coal kind hydrophilic contact angle of the coal dust and the coal dust concentration distribution model. The present application ensures that the coal dust concentration in the whole tunnel space is stably lower than a safety threshold value, and reduces the energy consumption and drainage burden of a spray dust-settling system.

Description

Technical Field

[0001] This invention relates to the field of coal mine video recognition technology, and in particular to a coal mine AI intelligent video recognition system and its usage method. Background Technology

[0002] During underground coal mine operations, large amounts of coal dust are inevitably generated during fully mechanized mining, tunneling, transportation, and crushing. This coal dust not only suspends in the tunnel space, causing a sharp decrease in visibility, but also spreads and migrates along the longitudinal direction of the tunnel with the ventilation airflow, forming a dynamically changing three-dimensional concentration distribution field. Different coal types exhibit vastly different wetting characteristics due to differences in coalification degree and surface functional groups. In addition, the real-time fluctuations in tunnel cross-sectional shape, equivalent diameter, airflow Reynolds number, and environmental temperature and humidity parameters make the generation, transport, and settling behavior of coal dust highly nonlinear and time-varying. Spray dust suppression, as a solution, is crucial for effective dust control. The most common and effective dust control method in coal mines relies on a strict physical matching relationship between droplet size, coal dust particle size, surface tension, and collision agglomeration efficiency. If the droplet size is too large, it is difficult to follow the turbulent pulsation of coal dust, resulting in low collection efficiency. If the particle size is too small, it is easy to evaporate and disperse, and its kinetic energy is insufficient, making it difficult to effectively wet coal dust particles. Therefore, there is an urgent need for a coal mine AI intelligent video recognition system and its usage method to ensure that the coal dust concentration in the entire roadway is stably below the safety threshold while reducing the energy consumption and drainage burden of the spray dust suppression system and improving the economic operation performance of the underground spray dust suppression system in coal mines. Summary of the Invention

[0003] To overcome the shortcomings of existing spray dust suppression systems, which are difficult to dynamically adjust precisely based on dust distribution, roadway structure, and coal properties, this invention provides a coal mine AI intelligent video recognition system and its usage method.

[0004] The technical implementation scheme of the present invention is as follows: a coal mine AI intelligent video recognition system, comprising the following steps: The data acquisition module collects video frame data, structural data, and environmental data of the coal mine roadways. The coal dust distribution module processes the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model in the roadway space. The contact angle calculation module processes the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway. The droplet size calculation module calculates the optimal dustfall droplet size based on the hydrophilic contact angle of the coal type and the coal dust concentration distribution model. The feedback adjustment module adjusts the dust suppression in the coal mine roadway based on the optimal dust droplet size.

[0005] Preferably, the acquisition of video frame data, structural data, and environmental data of the coal mine roadway includes: the video frame data includes visible light video frame data and multispectral video frame data; the structural data includes roadway cross-sectional height, roadway cross-sectional wet perimeter, and effective ventilation cross-sectional area; and the roadway environmental data includes real-time ventilation wind speed, roadway relative humidity, roadway temperature, and roadway air pressure.

[0006] Preferably, the process of processing the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model of the roadway space includes: calculating the equivalent diameter of the roadway based on the wetted perimeter of the roadway cross-section and the effective ventilation cross-sectional area of ​​the roadway; constructing a real-time point cloud model of the roadway based on the visible light video frame data and the structural data of the roadway; solving the optical path length along the longitudinal direction of the roadway using the Monte Carlo algorithm based on the point cloud model of the roadway; solving the roadway airflow Reynolds number based on the equivalent diameter of the roadway, the real-time ventilation wind speed, the roadway temperature, and the roadway air pressure; calculating the transmittance at each pixel of the visible light video frame data using dark channel priors; calculating the roughness factor using texture extraction; calculating the turbulent diffusion variance of coal dust in the transverse direction, the standard deviation of turbulent diffusion of coal dust in the vertical direction, and the turbulent diffusion variance of coal dust in the longitudinal direction of the roadway based on the real-time ventilation wind speed and the real-time point cloud model; constructing a coal dust concentration distribution model of the roadway space; and solving for the coal dust concentration distribution value at the roadway coordinates.

[0007] Preferably, the step of constructing a coal dust concentration distribution model for the roadway space and solving for the coal dust concentration distribution value at the roadway coordinates includes: The formula for the coal dust concentration distribution model is: ; In the formula, For the coordinates of the tunnel The coal dust concentration distribution value at the location, video frame in pixel coordinates Transmittance at that location Extinction coefficient, This represents the optical path length along the longitudinal direction of the tunnel. This is the shape correction factor. To correct the index, For the Reynolds number of the alleyway, The critical Reynolds number, The square of the transverse coordinate of the tunnel. Let Variance be the turbulent diffusion variance of coal dust in the transverse direction of the roadway. The vertical coordinates of the tunnel, The height of the tunnel cross section, The variance of turbulent diffusion of coal dust in the longitudinal direction of the roadway is denoted as .

[0008] Preferably, the step of processing the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway includes: performing image preprocessing and segmentation on the multispectral video frame data; selecting characteristic bands based on the spectral absorption characteristics of functional groups of coal and rock molecules; calculating the average reflectance of pixels in each characteristic band; arranging the characteristic bands in order of wavelength to establish a multispectral reflectance feature vector; selecting coal samples and obtaining the mean vector of standard coal rank spectra of different coal types, the covariance matrix of standard coal rank spectra of different coal types, and the spectral temperature compensation factor through standard experiments and Gaussian mixture distribution fitting; calculating the probability that coal dust in the roadway belongs to different coal types; the coal samples include anthracite, bituminous coal, and lignite; statistically analyzing the coal samples to obtain the total number of coal types and the Young's contact angle of each coal type; and combining the probability that coal dust belongs to different coal types, the roughness factor, and the relative humidity of the roadway to calculate the hydrophilic contact angle of coal dust.

[0009] Preferably, the calculation of the hydrophilic contact angle of coal dust based on the probability of coal dust belonging to different coal types, roughness factor, and relative humidity of the roadway includes: The membership probability formula is: ; In the formula, Belonging to the category of coal The probability, This is a multispectral reflectance feature vector. coal type The mean vector of the standard coal rank spectrum, coal type The inverse matrix of the covariance matrix of the standard coal rank spectrum. coal type The mean vector of the standard coal rank spectrum, The total number of coal types, for transpose, for transpose, The square of the spectral temperature compensation factor. The formula for hydrophilic contact angle is: ; In the formula, The hydrophilic contact angle of coal dust is the angle between the coal type and the coal. The total number of coal types, Belonging to the category of coal The probability, It is an inverse cosine function. Roughness factor coal type Young's contact angle, The attenuation coefficient is... The relative humidity of the tunnel.

[0010] Preferably, the step of solving for the optimal dustfall droplet size based on the hydrophilic contact angle of the coal type and the coal dust concentration distribution model includes: obtaining the roadway spatial volume and median diameter of the coal dust based on the real-time point cloud model of the roadway; calculating the partial pressure of water vapor in the roadway air based on the roadway relative humidity and roadway temperature; obtaining the capillary number and Stokes number of droplet movement through standard experiments; obtaining the saturated water vapor pressure; and calculating the optimal dustfall droplet size by combining the roadway cross-sectional height and the standard deviation of vertical turbulent diffusion of coal dust.

[0011] Preferably, the step of obtaining the saturated water vapor pressure and calculating the optimal dust droplet size by combining the roadway cross-sectional height and the standard deviation of vertical turbulent diffusion of coal dust includes: The optimal formula for dust droplet size is: ; In the formula, To achieve the optimal dust droplet size, The median diameter of coal dust. The number of capillaries responsible for droplet movement. For Stokes numbers, Let be the cosine of the hydrophilic contact angle of coal dust. The standard deviation of vertical turbulent diffusion of coal dust. The height of the tunnel cross section, The turbulence index, The saturated water vapor pressure It refers to the partial pressure of water vapor in the air of the tunnel.

[0012] Preferably, the step of adjusting the dust suppression feedback of the coal mine roadway based on the optimal dust suppression droplet size includes: obtaining the basic rated flow rate, standard particle size, real-time water supply pressure, and set water supply pressure of the spray system; calculating the water supply pressure difference based on the real-time water supply pressure and the set water supply pressure; setting a safe threshold for coal dust concentration; and solving for the target adjustment flow rate of the spray system by combining the optimal dust suppression droplet size, the coal dust concentration distribution value at the roadway coordinates, and the roadway space volume. The formula for the target regulating flow rate is: ; In the formula, for The target flow rate of the constant spray system is adjusted. Based on rated flow rate, for The optimal dust droplet size at any given time. Standard particle size, The coal dust response coefficient, It is the hyperbolic tangent function. For the coordinates of the tunnel The coal dust concentration distribution value at the location, The safe threshold for coal dust concentration For the volume of the alleyway space, For an integral infinitesimal element, This is the gain coefficient. Due to water supply pressure difference, To set the water supply pressure.

[0013] Preferably, a method of using a coal mine AI intelligent video recognition system further includes: S1: Collect video frame data, structural data, and environmental data of coal mine roadways; S2: Process the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model in the roadway space; S3: Process the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway; S4: Based on the hydrophilic contact angle of coal type and the coal dust concentration distribution model, solve for the optimal dust droplet size; S5: Adjust the dust suppression feedback in the coal mine roadway according to the optimal dust droplet size.

[0014] The beneficial effects of this invention are: This invention constructs a closed-loop control system encompassing perception, analysis, decision-making, and execution by setting up a data acquisition module, a coal dust distribution module, a contact angle calculation module, a droplet size calculation module, and a feedback adjustment module. Compared to traditional open-loop spraying methods, it can automatically reconstruct the coal dust concentration field and collaboratively solve for the optimal dust suppression parameters based on real-time visual images, structural parameters, and environmental variables in the coal mine. It also feeds back control commands to the spraying actuator in real time, achieving fully autonomous, high-precision, and highly adaptive control of the spraying dust suppression process, thereby improving the level of intelligence in underground coal mine dust control.

[0015] This invention integrates visible light video frame data, roadway cross-sectional wetted perimeter, effective ventilation cross-sectional area, and real-time point cloud model. It calculates transmittance a priori through dark channels and solves for optical path length using the Monte Carlo algorithm. Simultaneously, it introduces the roadway airflow Reynolds number and three-dimensional turbulent diffusion variance to construct a refined roadway spatial coal dust concentration distribution model. This model can accurately depict the non-uniform and unsteady migration law of coal dust under the action of complex turbulent fields, overcoming the shortcomings of traditional single-point concentration monitoring that cannot reflect the spatial distribution of the entire roadway. It provides a high-fidelity spatial concentration data foundation for subsequent precise dust suppression.

[0016] This invention combines multispectral video frame data with the spectral absorption characteristics of functional groups of coal and rock molecules, uses mixed Gaussian distribution fitting to construct a standard coal grade spectral feature library for different coal types, and introduces a spectral temperature compensation factor to eliminate environmental temperature interference; it determines the coal type to which suspended coal dust belongs in real time, and then dynamically calculates the hydrophilic contact angle of coal dust by combining roughness factor and roadway relative humidity, thus solving the technical problem of existing technologies that make it difficult to know the physicochemical properties of coal dust surface in real time. Attached Figure Description

[0017] Figure 1 This is a structural diagram of an AI-powered intelligent video recognition system for coal mines according to the present invention. Figure 2 This is a flowchart of an AI-powered intelligent video recognition method for coal mines according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: A coal mine AI intelligent video recognition system, such as Figure 1 As shown, it includes the following steps: The data acquisition module collects video frame data, structural data, and environmental data of the coal mine roadways. The video frame data includes visible light video frame data and multispectral video frame data. The structural data includes the tunnel cross-sectional height, the tunnel cross-sectional wet perimeter, and the effective ventilation cross-sectional area of ​​the tunnel. The tunnel environmental data includes real-time ventilation wind speed, tunnel relative humidity, tunnel temperature, and tunnel air pressure.

[0020] It should be further explained that the video frame data includes two types: visible light video frame data and multispectral video frame data. When collecting video frame data underground in coal mines, intrinsically safe mining cameras are deployed at key locations in the roadways. Visible light video frame data is collected by mining high-definition visible light cameras. Multispectral video frame data is collected by mining multispectral cameras. These cameras have built-in beam splitters or filter wheels and collect multiple narrow-band spectral images at preset wavelength intervals within the visible light to short-wave infrared range. The image data of each band is collected synchronously with the visible light video frame data in time, ensuring that the visible light image acquired at the same time and the spectral images of each band can be spatially registered at the pixel level. Structural data includes the tunnel cross-sectional height, wetted perimeter, and effective ventilation cross-sectional area. The tunnel cross-sectional height is the vertical distance from the tunnel floor to the roof, in meters. The wetted perimeter is the circumference of the tunnel's inner wall in contact with airflow, in meters. The effective ventilation cross-sectional area is the effective cross-sectional area within the tunnel for airflow passage, in square meters. The structural data is obtained using the following method: After tunnel excavation or support is completed, a mining laser profiler is used to select a measurement section at regular intervals along the longitudinal direction of the tunnel. The contour of each measurement section is scanned to obtain the coordinate point cloud data of the section contour. Based on the coordinate point cloud of the section contour, the actual height, actual perimeter, and actual area of ​​the section are calculated. For regular roadways with anchor spraying support or masonry arch support, the cross-sectional shape is close to a rectangle, a semi-circular arch, or a three-centered arch. The roadway cross-sectional height, wetted perimeter, and effective ventilation cross-sectional area are directly calculated based on the design width, height, and arch parameters of the roadway. For roadways with irregular cross-sections, the coordinate point cloud of the section contour is obtained through on-site measurement before calculation.

[0021] The tunnel environmental data includes real-time ventilation wind speed, tunnel relative humidity, tunnel temperature, and tunnel air pressure. Real-time ventilation wind speed is the average wind speed within the tunnel cross-section, expressed in meters per second; tunnel relative humidity is the ratio of the partial pressure of water vapor in the tunnel air to the saturated water vapor pressure at the same temperature, expressed as a percentage; tunnel temperature is the dry-bulb temperature of the tunnel, expressed in degrees Celsius; and tunnel air pressure is the absolute atmospheric pressure of the tunnel, expressed in kilopascals. The acquisition of tunnel environmental data relies on mining multi-parameter sensors deployed within the tunnel. These multi-parameter sensors integrate wind speed and direction measurement units, capacitive temperature and humidity measurement units, and air pressure measurement units.

[0022] The coal dust distribution module processes the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model in the roadway space. The equivalent diameter of the roadway is calculated based on the wetted perimeter of the roadway cross-section and the effective ventilation cross-sectional area of ​​the roadway. A real-time point cloud model of the roadway is constructed based on the visible light video frame data and the structural data of the roadway. The optical path length along the longitudinal direction of the roadway is solved using the Monte Carlo algorithm based on the point cloud model of the roadway. The Reynolds number of the roadway airflow is solved based on the equivalent diameter of the roadway, the real-time ventilation wind speed, the roadway temperature, and the roadway air pressure. The transmittance at each pixel of the visible light video frame is calculated using the dark channel prior. The roughness factor is calculated using texture extraction. The turbulent diffusion variance of coal dust in the transverse direction, the standard deviation of turbulent diffusion of coal dust in the vertical direction, and the turbulent diffusion variance of coal dust in the longitudinal direction of the roadway are calculated based on the real-time ventilation wind speed and the real-time point cloud model. A coal dust concentration distribution model in the roadway space is constructed, and the coal dust concentration distribution value at the roadway coordinates is solved.

[0023] The formula for the coal dust concentration distribution model is: ; In the formula, For the coordinates of the tunnel The coal dust concentration distribution value at the location, video frame in pixel coordinates Transmittance at that location Extinction coefficient, This represents the optical path length along the longitudinal direction of the tunnel. This is the shape correction factor. To correct the index, For the Reynolds number of the alleyway, The critical Reynolds number, The square of the transverse coordinate of the tunnel. Let Variance be the turbulent diffusion variance of coal dust in the transverse direction of the roadway. The vertical coordinates of the tunnel, The height of the tunnel cross section, The variance of turbulent diffusion of coal dust in the longitudinal direction of the roadway is denoted as .

[0024] It should be further explained that the equivalent diameter of the roadway is calculated based on the wetted perimeter and effective ventilation cross-sectional area of ​​the roadway obtained by the aforementioned data acquisition module. The calculation of the equivalent diameter adopts the conversion relationship between the wetted perimeter and the effective ventilation cross-sectional area of ​​the roadway, and the ventilation cross-section of the non-circular roadway is equivalent to the diameter of a circular cross-section with the same flow capacity. This is used as the characteristic scale for subsequent fluid dynamics calculations. The specific value of the equivalent diameter of the roadway is equal to four times the effective ventilation cross-sectional area of ​​the roadway divided by the wetted perimeter of the roadway cross-section. The unit of the calculation result is meters.

[0025] A real-time point cloud model of the tunnel is constructed by combining visible light video frame data and tunnel structural data. This is achieved through texture features extracted from visible light video frames and structured light visual measurement technology: feature points are extracted and matched from visible light video frames, and triangulation is performed in combination with the camera's internal and external parameters to obtain the three-dimensional spatial coordinates of the tunnel's inner wall and internal facilities. These coordinates are then fused and registered with the tunnel's cross-sectional height, wetted perimeter, and effective ventilation cross-sectional area structural data to generate a real-time point cloud model that reflects the current geometry of the tunnel. This point cloud model records the positional information of the tunnel walls, roof, and floor in three-dimensional space, and can accurately reflect the actual shape and size of the tunnel.

[0026] Based on a real-time point cloud model, the Monte Carlo algorithm is used to solve for the optical path length along the longitudinal direction of the tunnel. The basic idea of ​​the Monte Carlo algorithm is to simulate the process of a large number of photons starting from each pixel position of the video frame and propagating along the longitudinal direction of the tunnel. During the propagation process, the photons randomly collide and scatter with the coal dust particles suspended in the tunnel. By statistically analyzing the statistical path length traversed by a large number of photons from entering the tunnel space to being received by the camera, the optical path length at each position along the longitudinal direction of the tunnel is obtained. At the same time, the tunnel airflow Reynolds number is solved based on four parameters: the equivalent diameter of the tunnel, the real-time ventilation velocity, the tunnel temperature, and the tunnel air pressure. The calculation of the tunnel airflow Reynolds number requires obtaining the kinematic viscosity coefficient of the air in the tunnel. The kinematic viscosity coefficient of the air is obtained through a standard reference table of air kinematic viscosity coefficients. After obtaining the air kinematic viscosity coefficient, the product of the equivalent diameter of the tunnel and the real-time ventilation velocity is divided by... The air kinematic viscosity coefficient yields the Reynolds number of the tunnel airflow. For visible light video frame data, the dark channel prior algorithm is used to calculate the transmittance at each pixel of the video frame. The dark channel prior algorithm is based on statistical laws: in most outdoor images without smog or with low dust concentration, at least one color channel has an intensity value close to zero at some pixel locations. When applying this law to visible light video frames of coal mine tunnels, the minimum value of the red, green, and blue color channels is taken for each pixel and local minimum filtering is performed to obtain the dark channel value of that pixel. Combined with the estimated ambient light intensity in the tunnel, the transmittance at each pixel can be calculated through the inverse transformation of the atmospheric scattering model. The transmittance reflects the proportion of visible light signal propagating from a point in the tunnel depth direction to the camera that is not scattered and absorbed by coal dust. Its value is between zero and one, and the smaller the value, the higher the coal dust concentration at the corresponding pixel location.

[0027] The roughness factor is calculated using a texture extraction algorithm. The specific steps are as follows: After converting the visible light video frame to grayscale, the texture features of the image are extracted using the gray-level co-occurrence matrix method. The gray-level co-occurrence matrix statistically calculates the joint probability of grayscale values ​​appearing for pixels at specific directions and distances in the image. Based on the gray-level co-occurrence matrix, multiple texture parameters such as contrast, correlation, and entropy are calculated. After normalizing the calculated texture parameters, the roughness factor is obtained. This factor is used to characterize the influence of the geometric roughness of the coal dust particle surface on the wetting contact angle.

[0028] Based on real-time ventilation wind speed and a real-time point cloud model, turbulent diffusion parameters in three directions are calculated: the variance of turbulent diffusion of coal dust in the transverse direction of the roadway, the standard deviation of turbulent diffusion of coal dust in the vertical direction, and the variance of turbulent diffusion of coal dust in the longitudinal direction. The variance of turbulent diffusion of coal dust in the transverse direction characterizes the dispersion of coal dust particles in the horizontal direction across the roadway; the standard deviation of turbulent diffusion of coal dust in the vertical direction characterizes the amplitude of fluctuations in coal dust particles in the height direction of the roadway; and the variance of turbulent diffusion of coal dust in the longitudinal direction characterizes the dispersion of coal dust particles in the depth direction of the roadway. The calculation process for these three parameters is as follows: the turbulence intensity of the roadway is determined based on the real-time ventilation wind speed, which is positively correlated with the wind speed; then, combined with the geometric constraints of the roadway cross-section in the real-time point cloud model, the turbulent pulsation velocity components in the transverse, vertical, and longitudinal directions are extracted respectively; the diffusion variance or standard deviation in each direction is obtained by statistically averaging the turbulent pulsation velocity components. These three parameters together describe the turbulent diffusion capacity of coal dust in the three-dimensional space of the roadway.

[0029] After calculating all the above parameters, a coal dust concentration distribution model of the tunnel space is constructed, and the coal dust concentration distribution value at the tunnel coordinates is solved. The design logic of the coal dust concentration distribution model formula lies in combining the optical attenuation principle with the turbulent diffusion theory to characterize the three-dimensional distribution law of coal dust in the tunnel space. The basic starting point of the formula is: visible light will be scattered and absorbed when passing through dusty air, and the degree of light intensity attenuation is proportional to the total amount of coal dust on the path. The average coal dust concentration along the light propagation path is deduced from the transmittance at each pixel of the video frame. This is the logical basis of the first term of the formula. The first term can only give the average concentration along the light direction, which is difficult to describe the lateral and vertical changes in concentration in the tunnel cross section. The second term of the formula introduces a Gaussian distribution correction. The logic of this correction is: coal dust is carried by turbulent airflow The concentration of coal dust follows a theoretical distribution similar to turbulent diffusion in both the transverse and vertical dimensions of the roadway. Specifically, the concentration is highest in the central area of ​​the roadway and gradually decreases towards the wall. This distribution pattern is jointly controlled by the turbulent diffusion variance of coal dust in the transverse and longitudinal directions of the roadway. The airflow conditions in the roadway also affect the concentration distribution. When the Reynolds number exceeds the critical Reynolds number, the turbulence intensifies, making the coal dust diffusion more uniform. A Reynolds number ratio term is introduced into the formula to adjust the shape of the concentration distribution. The extinction coefficient in the formula is obtained through on-site calibration experiments. Specifically, a transmittance measuring device is placed at a location in the roadway with a known coal dust concentration. Visible light video frames at that location are simultaneously collected, and the transmittance is calculated. Multiple sets of concentration and transmittance data are fitted with linear regression, and the regression slope is the extinction coefficient. The shape correction coefficient is determined by least-squares fitting of the measured coal dust concentration distribution in the roadway with the theoretical distribution. The coefficient value that minimizes the deviation between the actual and theoretical distribution is the value of the shape correction coefficient. The correction index λ was obtained through nonlinear regression analysis of multiple sets of Reynolds number and concentration distribution data in the tunnel under different wind speed conditions. The best-fit index obtained from the regression is the value of λ. The critical Reynolds number adopts the critical value of flow in a circular pipe in fluid mechanics. This value is used to distinguish between laminar and turbulent flow states and does not require on-site calibration.

[0030] The contact angle calculation module processes the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway. The multispectral video frame data is preprocessed and segmented. Feature bands are selected based on the spectral absorption characteristics of functional groups of coal and rock molecules. The average reflectance of pixels in each feature band is calculated. Multispectral reflectance feature vectors are established by arranging the feature bands in order of wavelength. Coal samples are selected and the mean vector of standard coal rank spectra of different coal types, the covariance matrix of standard coal rank spectra of different coal types, and the spectral temperature compensation factor are obtained through standard experiments and Gaussian distribution fitting. The probability that coal dust in the roadway belongs to different coal types is calculated. The coal samples include anthracite, bituminous coal, and lignite. The total number of coal types and the Young's contact angle of each coal type are obtained by statistical analysis of the coal samples. The hydrophilic contact angle of coal dust is calculated by combining the probability that coal dust belongs to different coal types, the roughness factor, and the relative humidity of the roadway.

[0031] The membership probability formula is: ; In the formula, Belonging to the category of coal The probability, This is a multispectral reflectance feature vector. coal type The mean vector of the standard coal rank spectrum, coal type The inverse matrix of the covariance matrix of the standard coal rank spectrum. coal type The mean vector of the standard coal rank spectrum, The total number of coal types, for transpose, for transpose, The square of the spectral temperature compensation factor. The formula for hydrophilic contact angle is: ; In the formula, The hydrophilic contact angle of coal dust is the angle between the coal type and the coal. The total number of coal types, Belonging to the category of coal The probability, It is an inverse cosine function. Roughness factor coal type Young's contact angle, The attenuation coefficient is... The relative humidity of the tunnel.

[0032] It should be further explained that image preprocessing and image segmentation are performed on the multispectral video frame data. The image preprocessing stage includes radiometric correction, bad pixel replacement, noise filtering, and reflectance conversion. Radiometric correction aims to eliminate the response differences between detectors in different bands of the multispectral camera, ensuring that the grayscale values ​​of each band accurately reflect the radiation brightness reflected by the target. Bad pixel replacement corrects individual pixels with abnormal or malfunctioning responses in the detector array. The values ​​of these bad pixels are replaced by the interpolation results of the surrounding valid pixels. Noise filtering uses median filtering or Gaussian filtering to suppress random noise in the multispectral image caused by interference from underground electronic equipment and dark current from the detector. Reflectance conversion converts the grayscale values ​​of the radiometrically corrected image into reflectance values. A standard reference plate with known reflectance is placed in the camera's field of view, and the reflectance of each pixel in each band is calculated by comparing the grayscale response of the target pixel with that of the standard reference plate. The image segmentation stage distinguishes the coal dust particle area from the roadway wall, equipment surface, and background area based on the edge information provided by the visible light video frame and the interspectral differences of the multispectral image. This ensures that the subsequent average reflectance calculation of the characteristic bands is performed only for the coal dust particle area, eliminating interference from non-coal dust areas.

[0033] After image preprocessing and segmentation, characteristic bands are screened based on the spectral absorption characteristics of coal molecular functional groups. Different coal types differ in molecular structure, specifically in the types and quantities of oxygen-containing and aliphatic functional groups. Oxygen-containing functional groups include hydroxyl, carboxyl, and carbonyl groups, which have characteristic absorption peaks at specific wavelengths in the short-wave infrared band. Their absorption intensity and peak shape change regularly with the degree of coalification. Aliphatic functional groups such as methyl and methylene groups also have characteristic absorption peaks at specific wavelengths in the mid-infrared band, exhibiting combination and overtone absorption in the short-wave infrared range. Based on these known spectral absorption characteristics, characteristic bands sensitive to coal type variations are screened from the numerous bands covered by the multispectral camera. These characteristic bands are typically located near the center wavelength of the absorption peaks of coal functional groups.

[0034] After the characteristic bands are determined, the arithmetic mean of the reflectance of all pixels belonging to the coal dust particle region is calculated on the preprocessed and segmented multispectral video frame data for each characteristic band image. This yields the average reflectance of pixels in each characteristic band. The average reflectance of each band is arranged sequentially according to the center wavelength of each characteristic band from short to long wavelength, thus establishing a multispectral reflectance feature vector. This feature vector solidifies the reflectance characteristics of coal dust in the roadway under each multispectral characteristic band in numerical form, serving as a quantitative basis for subsequent coal type identification.

[0035] Before calibration, coal samples were selected for standard tests and Gaussian mixture distribution fitting. The coal samples included three categories: anthracite, bituminous coal, and lignite, covering the main coal types. The standard test procedure was as follows: representative coal block samples were collected from each mine, crushed, ground, and sieved to prepare coal powder samples. The multispectral reflectance spectral curves and Young's contact angles of each coal powder sample were measured under the same experimental conditions. The multispectral reflectance spectral curves were measured using a spectrometer with the same wavelength configuration as the underground multispectral camera under standard laboratory light source conditions. The Young's contact angle was measured using the seat-drop method; the coal powder was pressed into a sheet and placed on the sample stage of the contact angle measuring instrument, and distilled water was dropped onto the surface of the coal sheet. A high-precision camera system captures the droplet contour, and image analysis software is used to fit the droplet morphology for calculation. For each type of coal, multiple independent samples are measured. Regarding the multispectral reflectance spectral curve, the mean reflectance of these samples in each characteristic band is calculated to form the mean vector of the standard coal rank spectrum of that coal type. The covariance matrix of the reflectance of these samples in each characteristic band is calculated to describe the degree of spectral variation between different samples within the same coal type. The spectral temperature compensation factor is obtained by repeating the above measurement process under different ambient temperatures, comparing the offset of the spectral curve of the same coal type at different temperatures, and establishing the relationship between temperature and spectral offset through regression analysis to obtain the spectral temperature compensation factor.

[0036] The measured multispectral reflectance feature vectors were compared with the mean vectors and covariance matrices of the standard coal ranks for each coal type stored in advance. The probability of coal dust in the roadway belonging to different coal types was calculated using the membership probability formula. The logic of the membership probability formula is as follows: the multispectral reflectance features of the measured coal dust are compared with the pre-stored standard spectral features of each coal type. The closer the feature is to the standard feature of a certain coal type, the higher the similarity score is assigned to that coal type. Finally, the similarity scores of all coal types are normalized so that the sum of the membership probabilities of each coal type equals one. This is used to infer the probability of coal type composition of the coal dust in the roadway. The total number of coal type samples and the Young's contact angle of each coal type are counted in the same way. The total number of coal type samples includes anthracite, bituminous coal, and lignite. The Young's contact angle of each coal type has been determined in standard tests using the seated drop method. After obtaining the probability of coal dust belonging to each coal type, the hydrophilic contact angle of the coal dust is calculated by combining three factors: the probability of coal dust belonging to different coal types, the roughness factor, and the relative humidity of the roadway. The logic of the hydrophilic contact angle formula is as follows: first, the Young's contact angles of each coal type are weighted and averaged according to the measured membership probabilities; then, the roughness factor is introduced to correct the surface microstructure of the weighted result. The rough surface will amplify the original wetting characteristics; finally, the contact angle is attenuated and corrected according to the roadway environmental humidity. The higher the humidity, the more water molecules are adsorbed on the surface of the coal dust, and the smaller the effective contact angle. In the formula, the attenuation coefficient is obtained through wettability experiments. The hydrophilic contact angle of the same coal dust sample is measured under different relative humidity conditions. The experimental data of the contact angle changing with humidity are subjected to exponential fitting, and the humidity sensitivity coefficient obtained by fitting is the attenuation coefficient.

[0037] The droplet size calculation module calculates the optimal dustfall droplet size based on the hydrophilic contact angle of the coal type and the coal dust concentration distribution model. The roadway spatial volume and median diameter of coal dust are obtained based on the real-time point cloud model of the roadway. The partial pressure of water vapor in the roadway air is calculated based on the roadway relative humidity and roadway temperature. The capillary number and Stokes number of droplet movement are obtained through standard experiments. The saturated water vapor pressure is obtained. The optimal dust droplet size is calculated by combining the roadway cross-sectional height and the standard deviation of vertical turbulent diffusion of coal dust.

[0038] The optimal formula for dust droplet size is: ; In the formula, To achieve the optimal dust droplet size, The median diameter of coal dust. The number of capillaries responsible for droplet movement. For Stokes numbers, Let be the cosine of the hydrophilic contact angle of coal dust. The standard deviation of vertical turbulent diffusion of coal dust. The height of the tunnel cross section, The turbulence index, The saturated water vapor pressure It refers to the partial pressure of water vapor in the air of the tunnel.

[0039] It should be further explained that the roadway spatial volume and median diameter of coal dust are obtained based on the real-time point cloud model of the roadway. The roadway spatial volume is obtained as follows: the roadway space covered by the real-time point cloud model is processed into three-dimensional voxels, and the entire roadway is divided into several thin slices along the longitudinal direction. For each thin slice, the cross-sectional contour is extracted based on the point cloud data, and the cross-sectional area of ​​the thin slice is calculated. The cross-sectional area is multiplied by the thickness of the thin slice and then accumulated along the longitudinal direction to obtain the total volume of the roadway. The unit of this volume value is cubic meters. The median diameter of coal dust is obtained as follows: the contours of suspended coal dust particles in the roadway are extracted using the aforementioned visible light video frame data combined with image segmentation technology. The equivalent diameter of each particle is statistically analyzed, and a cumulative distribution curve of coal dust particle size is established. The particle size value corresponding to the cumulative distribution reaching 50% is taken as the median diameter of coal dust. This median diameter represents the central trend of coal dust particle size distribution in the roadway, and the unit is micrometers.

[0040] The partial pressure of water vapor in the roadway air is calculated based on the relative humidity and temperature of the roadway. The calculation of the partial pressure of water vapor in the roadway air requires first determining the saturated water vapor pressure at the roadway temperature, and then multiplying the saturated water vapor pressure by the relative humidity of the roadway to obtain the partial pressure of water vapor in the roadway air. The relative humidity and temperature of the roadway are measured in real time by multi-parameter sensors in the data acquisition module. The value of the water vapor partial pressure reflects the magnitude of the actual water vapor content in the roadway air, and the unit is kilopascal.

[0041] The capillary number and Stokes number of droplet motion are obtained through standard experiments. The standard experimental procedure for the capillary number is as follows: a droplet generation and observation platform is set up under laboratory conditions, and droplets are generated using nozzles of the same type as those used in downhole spray systems. The motion morphology of the droplets in the air is captured, and the droplet velocity and surface tension are measured simultaneously. The capillary number is equal to the ratio of the droplet's kinetic viscosity to its surface tension. This value reflects the ease with which the droplet deforms and breaks up, and is dimensionless. The standard experimental procedure for the Stokes number is as follows: the terminal velocity of the droplet in still air is measured, and the Stokes number is calculated using the Stokes formula, taking into account the droplet's characteristic size and the air viscosity coefficient. This value characterizes the droplet's ability to follow changes in airflow and is dimensionless. Simultaneously, the saturated water vapor pressure is obtained. The saturated water vapor pressure is obtained based on the tunnel temperature parameter. The water vapor saturation pressure calculation formula is used. After substituting the tunnel temperature into the formula, the saturated water vapor pressure value at that temperature is obtained. The difference between the saturated water vapor pressure and the aforementioned tunnel air water vapor partial pressure reflects the gap between the water vapor content in the air and the saturation state. This difference directly affects the evaporation rate of the droplets during flight.

[0042] After obtaining all the above parameters, the optimal dust droplet size is calculated by combining the tunnel cross-sectional height and the standard deviation of vertical turbulent diffusion of coal dust. The design logic of the formula integrates the microscopic physical processes of coal dust wetting and capture, the macroscopic transport processes of turbulent diffusion, and the environmental dissipation processes of droplet evaporation into a unified optimization framework. The first factor of the formula reflects the microscopic dynamic conditions for effective collision and wetting of droplets with coal dust. Its logic is as follows: the optimal particle size needs to maintain a reasonable scale match with the coal dust particle size. When the particle size is too large, the droplet number density decreases and it becomes difficult to follow the coal dust movement; when the particle size is too small, the inertial collision efficiency decreases. The capillary number and Stokes number respectively characterize the droplet's deformability and motion relaxation characteristics, which together determine the dimensionless scale corresponding to the peak capture efficiency when the droplet contacts the coal dust. The cosine value of the hydrophilic contact angle of the coal type determines the effective wetting after the collision. The first factor gives the basic optimal particle size. The second factor of the formula corrects for turbulent diffusion. Its logic is as follows: the larger the standard deviation of the vertical turbulent diffusion of coal dust in the roadway, the more intense the movement amplitude of the coal dust particles in the height direction, and the more vigorous the droplet needs to be. Sufficiently large particle size is needed to maintain its own momentum and prevent it from being carried away by turbulence and deviating from the collision trajectory with coal dust. Therefore, the particle size increases exponentially with the increase of the ratio of the standard deviation of turbulent diffusion to the height of the roadway cross section. The third factor in the formula is to correct for evaporation loss. The logic is that the difference between the saturated water vapor pressure and the partial pressure of water vapor in the roadway air reflects the degree of air unsaturation. The larger the difference, the faster the droplets evaporate during flight. Droplets with too small a particle size will evaporate completely before reaching the coal dust. Increasing the initial particle size ensures that the remaining particle size is large enough for effective collision. The logic of the entire formula is from the basic capture scale to turbulence following compensation and then to evaporation loss compensation. The product of the three factors makes the optimal particle size able to take into account collision efficiency, diffusion adaptability and evaporation tolerance in complex downhole environments. In the formula, the turbulence index is determined by comparing the deviation between the measured optimal particle size and the theoretical calculated value under different turbulence intensities. The deviation data is fitted with a power function to the turbulence intensity, and the power exponent obtained by the fitting is the turbulence index.

[0043] The feedback adjustment module adjusts the dust suppression in the coal mine roadway based on the optimal dust droplet size.

[0044] The basic rated flow rate, standard particle size, real-time water supply pressure and set water supply pressure of the spray system are obtained. The water supply pressure difference is calculated based on the real-time water supply pressure and set water supply pressure. The safe threshold of coal dust concentration is preset. The target regulating flow rate of the spray system is solved by combining the optimal dust droplet particle size, the coal dust concentration distribution value at the roadway coordinates and the roadway space volume. The formula for the target regulating flow rate is: ; In the formula, for The target flow rate of the constant spray system is adjusted. Based on rated flow rate, for The optimal dust droplet size at any given time. Standard particle size, The coal dust response coefficient, It is the hyperbolic tangent function. For the coordinates of the tunnel The coal dust concentration distribution value at the location, The safe threshold for coal dust concentration For the volume of the alleyway space, For an integral infinitesimal element, This is the gain coefficient. Due to water supply pressure difference, To set the water supply pressure.

[0045] It should be further explained that the four parameters of the spray system are: basic rated flow rate, standard particle size, real-time water supply pressure, and set water supply pressure. The basic rated flow rate refers to the spray water flow rate specified by the spray system under standard operating conditions. This value is provided by the spray system manufacturer or obtained through on-site calibration, and the unit is liters per minute. The standard particle size refers to the median particle size of the droplets produced by the spray system under the basic rated flow rate and set water supply pressure conditions. This value is determined by the nozzle's factory performance parameters, and the unit is micrometers. The real-time water supply pressure refers to the actual water pressure value of the spray system's water supply pipeline at the current moment. It is collected in real time by a pressure transmitter installed in the water supply pipeline, and the unit is megapascals. The set water supply pressure refers to the rated operating water pressure specified by the design of the spray system. This value is determined by the system parameter settings, and the unit is megapascals. After obtaining the real-time water supply pressure and the set water supply pressure, the water supply pressure difference is calculated. This water supply pressure difference is equal to the set water supply pressure minus the real-time water supply pressure. If the real-time water supply pressure is lower than the set water supply pressure, the water supply pressure difference is positive, indicating that the water supply pressure is insufficient. If the real-time water supply pressure is higher than the set water supply pressure, the water supply pressure difference is negative, indicating that the water supply pressure is too high. The calculation result of the water supply pressure difference provides a basis for subsequent flow correction.

[0046] A pre-set safety threshold for coal dust concentration is established. This threshold, determined according to national coal mine safety regulations and occupational health standards, represents the highest permissible coal dust concentration within the roadway, measured in milligrams per cubic meter (mg / m³), consistent with the coal dust concentration distribution value. Combining the optimal dust-falling droplet size, the coal dust concentration distribution value at the roadway coordinates, and the roadway volume, the target regulating flow rate of the spray system is calculated. The optimal dust-falling droplet size is calculated and provided in real-time by the droplet size calculation module; the coal dust concentration distribution value at the roadway coordinates is given by the coal dust distribution model, representing the concentration of each droplet within the roadway space. The concentration value at the coordinate location; the tunnel volume is calculated from the real-time point cloud model. The specific solution process uses the target regulating flow rate formula. The formula's design logic is to nonlinearly fuse the control requirements of three levels: the basic hydraulic characteristics of the spray system, the coal dust pollution load response, and the water supply compensation, to solve for the flow command. The first factor of the formula is based on the atomization mechanism for particle size correction. Its logic is: for pressure-type atomizing nozzles, under otherwise unchanged conditions, increasing the flow rate will increase the droplet size, and decreasing the flow rate will decrease the droplet size. The relationship between the two... The formula approximates a square relationship between the two. Therefore, when the optimal particle size deviates from the standard particle size, the actual atomized particle size is changed by adjusting the flow rate, so that the actual generated droplet particle size is as close as possible to the optimal value. The larger the particle size deviation, the larger the flow rate adjustment range. The second factor of the formula is based on the dynamic flow response of the pollution load. The logic is: the base value of the spray flow rate should be adjusted according to the actual degree of coal dust exceeding the limit in the roadway. When the total amount of coal dust is lower than the safety threshold, only the base flow rate needs to be maintained. When the total amount of coal dust is close to or exceeds the safety threshold, the spray intensity needs to be increased significantly. The introduction of the hyperbolic tangent function makes this response have a smooth transition characteristic near the threshold, reducing the system oscillation caused by the sudden flow jump. The coal dust response coefficient is used to adjust the final maximum flow rate increase. The third factor of the formula is based on the compensation correction of the water supply pressure deviation. The logic is: the deviation between the actual water supply pressure and the set value will directly affect the atomization effect and flow rate of the nozzle. When the pressure is insufficient, the atomized particle size is too large and the flow rate is too small. An additional set value is added at the upstream flow control end to compensate for the insufficient downstream pressure. When the pressure is too high, the flow rate setting is appropriately reduced to reduce the excessively fine droplets and energy waste. The logic of the entire formula is from adjusting the basic flow rate through particle size matching to adjusting the incremental flow rate through concentration response, and then to adjusting the final output through pressure compensation. In the formula, the coal dust response coefficient is obtained through on-site step response tests. A sudden increase in coal dust concentration is artificially created in the roadway, and the dynamic response curve of the spray system flow rate from the basic value to the stable adjustment value is recorded. The value of the coal dust response coefficient is determined based on the steady-state gain value. The gain coefficient is obtained through water supply pressure disturbance tests. During the operation of the spray system, the water supply pressure is artificially changed, and the deviation between the target adjustment flow rate and the actual flow rate is recorded. The corresponding data of multiple pressure deviations and flow deviations are linearly regressed, and the regression slope is the gain coefficient.

[0047] Example 2: Based on Example 1, a method for using a coal mine AI intelligent video recognition system, such as... Figure 2 As shown, it also includes: S1: Collect video frame data, structural data, and environmental data of coal mine roadways; S2: Process the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model in the roadway space; S3: Process the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway; S4: Based on the hydrophilic contact angle of coal type and the coal dust concentration distribution model, solve for the optimal dust droplet size; S5: Adjust the dust suppression feedback in the coal mine roadway according to the optimal dust droplet size.

[0048] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A coal mine AI intelligent video recognition system, characterized in that, Includes the following steps: The data acquisition module collects video frame data, structural data, and environmental data of the coal mine roadways. The coal dust distribution module processes the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model in the roadway space. The contact angle calculation module processes the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway. The droplet size calculation module calculates the optimal dustfall droplet size based on the hydrophilic contact angle of the coal type and the coal dust concentration distribution model. The feedback adjustment module adjusts the dust suppression in the coal mine roadway based on the optimal dust droplet size.

2. The coal mine AI intelligent video recognition system according to claim 1, characterized in that, The acquisition of video frame data, structural data, and environmental data of coal mine roadways includes: the video frame data includes visible light video frame data and multispectral video frame data; the structural data includes roadway cross-sectional height, roadway cross-sectional wet perimeter, and effective ventilation cross-sectional area; and the roadway environmental data includes real-time ventilation wind speed, roadway relative humidity, roadway temperature, and roadway air pressure.

3. The coal mine AI intelligent video recognition system according to claim 1, characterized in that, The process of processing the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model in the roadway space includes: calculating the equivalent diameter of the roadway based on the wetted perimeter of the roadway cross-section and the effective ventilation cross-sectional area of ​​the roadway; constructing a real-time point cloud model of the roadway based on the visible light video frame data and the structural data of the roadway; solving the optical path length along the longitudinal direction of the roadway using the Monte Carlo algorithm based on the point cloud model of the roadway; solving the roadway airflow Reynolds number based on the equivalent diameter of the roadway, the real-time ventilation wind speed, the roadway temperature, and the roadway air pressure; calculating the transmittance at each pixel of the visible light video frame data using dark channel priors; calculating the roughness factor using texture extraction; calculating the turbulent diffusion variance of coal dust in the transverse direction, the standard deviation of turbulent diffusion of coal dust in the vertical direction, and the turbulent diffusion variance of coal dust in the longitudinal direction of the roadway based on the real-time ventilation wind speed and the real-time point cloud model; constructing a coal dust concentration distribution model in the roadway space; and solving for the coal dust concentration distribution value at the roadway coordinates.

4. A coal mine AI intelligent video recognition system according to claim 3, characterized in that, The construction of the coal dust concentration distribution model in the roadway space, and the solution of the coal dust concentration distribution values ​​at the roadway coordinates, includes: The formula for the coal dust concentration distribution model is: ； In the formula, For the coordinates of the tunnel The coal dust concentration distribution value at the location, video frame in pixel coordinates Transmittance at that location Extinction coefficient, This represents the optical path length along the longitudinal direction of the tunnel. This is the shape correction factor. To correct the index, For the Reynolds number of the alleyway, The critical Reynolds number, The square of the transverse coordinate of the tunnel. Let Variance be the turbulent diffusion variance of coal dust in the transverse direction of the roadway. The vertical coordinates of the tunnel, The height of the tunnel cross section, The variance of turbulent diffusion of coal dust in the longitudinal direction of the roadway is denoted as .

5. A coal mine AI intelligent video recognition system according to claim 2, characterized in that, The process of processing the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway includes: performing image preprocessing and segmentation on the multispectral video frame data; selecting characteristic bands based on the spectral absorption characteristics of functional groups of coal and rock molecules; calculating the average reflectance of pixels in each characteristic band; arranging the characteristic bands in order of wavelength to establish a multispectral reflectance feature vector; selecting coal samples and obtaining the mean vector of standard coal rank spectra of different coal types, the covariance matrix of standard coal rank spectra of different coal types, and the spectral temperature compensation factor through standard experiments and Gaussian mixture distribution fitting; calculating the probability that coal dust in the roadway belongs to different coal types; the coal samples include anthracite, bituminous coal, and lignite; statistically analyzing the coal samples to obtain the total number of coal types and the Young's contact angle of each coal type; and combining the probability that coal dust belongs to different coal types, the roughness factor, and the relative humidity of the roadway to calculate the hydrophilic contact angle of coal dust.

6. A coal mine AI intelligent video recognition system according to claim 5, characterized in that, The calculation of the hydrophilic contact angle of coal dust based on the probability of coal dust belonging to different coal types, roughness factor, and relative humidity of the roadway includes: The membership probability formula is: ； In the formula, Belonging to the category of coal The probability, This is a multispectral reflectance feature vector. coal type The mean vector of the standard coal rank spectrum, coal type The inverse matrix of the covariance matrix of the standard coal rank spectrum. coal type The mean vector of the standard coal rank spectrum, The total number of coal types, for transpose, for transpose, The square of the spectral temperature compensation factor. The formula for hydrophilic contact angle is: ； In the formula, The hydrophilic contact angle of coal dust is the angle between the coal type and the coal. The total number of coal types, Belonging to the category of coal The probability, It is an inverse cosine function. Roughness factor coal type Young's contact angle, The attenuation coefficient is... The relative humidity of the tunnel.

7. A coal mine AI intelligent video recognition system according to claim 3, characterized in that, The step of determining the optimal dustfall droplet size based on the hydrophilic contact angle of the coal type and the coal dust concentration distribution model includes: obtaining the roadway spatial volume and median diameter of the coal dust based on the real-time point cloud model of the roadway; calculating the partial pressure of water vapor in the roadway air based on the roadway relative humidity and roadway temperature; obtaining the capillary number and Stokes number of droplet movement through standard experiments; obtaining the saturated water vapor pressure; and calculating the optimal dustfall droplet size by combining the roadway cross-sectional height and the standard deviation of vertical turbulent diffusion of coal dust.

8. A coal mine AI intelligent video recognition system according to claim 7, characterized in that, The process of obtaining the saturated water vapor pressure and calculating the optimal dust droplet size by combining the roadway cross-sectional height and the standard deviation of vertical turbulent diffusion of coal dust includes: The optimal formula for dust droplet size is: ； In the formula, To achieve the optimal dust droplet size, The median diameter of coal dust. The number of capillaries responsible for droplet movement. For Stokes numbers, Let be the cosine of the hydrophilic contact angle of coal dust. The standard deviation of vertical turbulent diffusion of coal dust. The height of the tunnel cross section, The turbulence index, The saturated water vapor pressure It refers to the partial pressure of water vapor in the air of the tunnel.

9. A coal mine AI intelligent video recognition system according to claim 3, characterized in that, The step of adjusting the dust suppression feedback in the coal mine roadway based on the optimal dust droplet size includes: obtaining the basic rated flow rate, standard particle size, real-time water supply pressure and set water supply pressure of the spray system; calculating the water supply pressure difference based on the real-time water supply pressure and set water supply pressure; setting a safe threshold for coal dust concentration; and solving the target adjustment flow rate of the spray system by combining the optimal dust droplet size, the coal dust concentration distribution value at the roadway coordinates and the roadway space volume. The formula for the target regulating flow rate is: ； In the formula, for The target flow rate of the constant spray system is adjusted. Based on rated flow rate, for The optimal dust droplet size at any given time. Standard particle size, The coal dust response coefficient, It is the hyperbolic tangent function. For the coordinates of the tunnel The coal dust concentration distribution value at the location, The safe threshold for coal dust concentration, For the volume of the alleyway space, For an integral infinitesimal element, This is the gain coefficient. Due to water supply pressure difference, To set the water supply pressure.

10. A method for using a coal mine AI intelligent video recognition system, characterized in that it also... include: S1: Collect video frame data, structural data, and environmental data of coal mine roadways; S2: Process the video frame data and the structural data of the coal mine roadway to construct a coal dust concentration distribution model in the roadway space; S3: Process the video frame data to calculate the hydrophilic contact angle of coal dust in the coal mine roadway; S4: Based on the hydrophilic contact angle of coal type and the coal dust concentration distribution model, solve for the optimal dust droplet size; S5: Adjust the dust suppression feedback in the coal mine roadway according to the optimal dust droplet size.

Images