A method for identifying smoke images in underground coal mines

Abstract

The present invention relates to the field of image recognition technology, and in particular to a method for identifying smoke images in coal mines. The method comprises: collecting coal mine images, constructing flow area feature values ​​of each pixel point, and calculating the air circulation of each pixel point; obtaining the dark channel gradient direction complexity of each pixel point according to the distribution of the dark channel gradient direction in the neighborhood window of each pixel point; obtaining the smoke area feature value of each pixel point according to the air circulation of each pixel point and the dark channel gradient direction complexity, and then obtaining the smoke pixel judgment value of each pixel point; obtaining the smoke gradient of each pixel point according to the air circulation of each pixel point, the smoke pixel judgment value and the dark channel gradient amplitude, and calculating the smoke influence degree of the smoke gradient image; constructing the smoke recognition factor of each pixel point, identifying the smoke pixel point in the coal mine image, and completing the recognition of smoke in the coal mine. The present invention can accurately identify smoke in coal mines based on image data.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to a method for recognizing smoke images in underground coal mines. Background Art

[0002] Coal occupies an important strategic position in the development of my country, so the safe mining of coal mines is of great significance to the social and economic development of my country. All equipment that produces high temperature or open flames can ignite flammable materials and cause fires if they are not used and managed properly. In medium and small coal mines, various open flames and blasting work are often the causes of external fires. With the improvement of the degree of mechanization, the proportion of electromechanical equipment fires has gradually increased. Once a fire occurs, it will endanger the lives of the entire mining area and the personnel in the mining area. Fires generally occur very suddenly, often unexpectedly, and develop rapidly and violently. If they are not discovered in time, they may often lead to major fire accidents, produce a large amount of toxic and harmful high-temperature smoke, and endanger the lives of the entire mine and the personnel in the mining area. Fire warning can reduce the losses caused by disasters. In most cases, the first thing to spread in a fire is smoke rather than open flames. Compared with open flames, smoke with rapid diffusion characteristics in space and time is more likely to appear within the range of video surveillance and is easier to be captured.

[0003] Timely detection of smoke at the early stage of a fire and accurate warning can give underground personnel sufficient time to take measures to avoid greater casualties and property losses. However, due to the high dust content, uneven lighting, and gradual diffusion of smoke in coal mines, it is often difficult to accurately detect it. Therefore, improving the accuracy of smoke image recognition in coal mines is of vital importance. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for identifying smoke images in underground coal mines to solve the existing problems.

[0005] A coal mine smoke image recognition method of the present invention adopts the following technical solution:

[0006] An embodiment of the present invention provides a method for identifying smoke images in a coal mine, the method comprising the following steps:

[0007] Collect underground coal mine images, combine dark channel algorithms to obtain coal mine dark channel images, and use edge detection to obtain the dark channel gradient amplitude and direction of each pixel in the coal mine dark channel images;

[0008] According to the distribution of the dark channel value and the R, G, and B channel values ​​in the neighborhood window of each pixel, the flow area characteristic value of each pixel is constructed; according to the flow area characteristic value and the dark channel gradient amplitude of each pixel, the air circulation of each pixel is constructed; according to the distribution of the dark channel gradient direction in the neighborhood window of each pixel, the dark channel gradient direction complexity of each pixel is obtained; according to the air circulation of each pixel and the dark channel gradient direction complexity of each pixel, the smoke area characteristic value of each pixel is obtained; according to the smoke area characteristic value of each pixel and the preset smoke pixel judgment threshold, the smoke pixel judgment value of each pixel is obtained; according to the air circulation of each pixel, the smoke pixel judgment value and the dark channel gradient amplitude, the smoke gradient of each pixel is obtained, and the smoke gradients of all pixels constitute a smoke gradient image; according to the distribution of the smoke gradient of each pixel in the smoke gradient image, a smoke area measurement matrix is ​​constructed; according to the distribution of each element in the smoke area measurement matrix, the smoke influence degree of the smoke gradient image is obtained;

[0009] The smoke recognition factor of each pixel is constructed according to the smoke influence degree of the smoke gradient image of the smoke area characteristic value of each pixel. The pixels whose smoke recognition factor is higher than the preset threshold are regarded as smoke pixels to complete the recognition of smoke in coal mines.

[0010] Furthermore, constructing the flow area characteristic value of each pixel point according to the distribution of the dark channel value and the R, G, and B channel values ​​in the neighborhood window of each pixel point includes:

[0011] A neighborhood window is constructed with each pixel as the center, the mean of the dark channel values ​​of all pixels in the neighborhood window of each pixel is calculated, the variance of the channel values ​​of all pixels in the neighborhood window of each pixel in the R, G, and B channels is calculated, the sum of the three variances is obtained, the ratio of the mean to the sum is calculated, and the ratio is used as the flow area characteristic value of each pixel.

[0012] Furthermore, the air flow degree of each pixel point includes: taking the product of the flow area characteristic value of each pixel point and the dark channel gradient amplitude as the air flow degree of the pixel point.

[0013] Furthermore, the dark channel gradient direction complexity of each pixel is obtained according to the distribution of the dark channel gradient direction in the neighborhood window of each pixel, including:

[0014] Divide 180° into directions, the dark channel gradient directions of the pixels in the neighborhood window of each pixel are counted to obtain the dark channel gradient direction histogram of each pixel, and the normalized dark channel gradient direction histogram is obtained after normalization. The complexity of the dark channel gradient direction of the pixel x is The expression is:

[0015]

[0016] In the formula, For the The corresponding value of each direction in the normalized dark channel gradient direction histogram of pixel point x.

[0017] Furthermore, the smoke region characteristic value of each pixel point is obtained according to the air circulation degree of each pixel point and the complexity of the dark channel gradient direction, including:

[0018] The ratio of the air circulation degree of each pixel to the complexity of the dark channel gradient direction is obtained, the opposite of the ratio is used as the exponent of an exponential function with a natural constant as the base, and the difference between the number 1 and the calculation result of the exponential function is used as the smoke area characteristic value of each pixel.

[0019] Furthermore, the smoke pixel judgment value of each pixel point is obtained according to the smoke region characteristic value of each pixel point and a preset smoke pixel judgment threshold value, including:

[0020] When the smoke region characteristic value of a pixel point is less than a preset smoke pixel judgment threshold, the smoke pixel judgment value of the pixel point is 0; otherwise, the smoke pixel judgment value of the pixel point is 1.

[0021] Furthermore, the smoke gradient of each pixel is obtained according to the air circulation degree of each pixel, the smoke pixel judgment value and the dark channel gradient amplitude, including:

[0022] The product of the air circulation degree of each pixel, the smoke pixel judgment value and the dark channel gradient amplitude is taken as the smoke gradient of each pixel.

[0023] Furthermore, the smoke region measurement matrix is ​​constructed according to the smoke gradient distribution of each pixel in the smoke gradient image, including:

[0024] The smoke gradients of all pixels in the smoke gradient image are evenly divided into M smoke gradient orders, the area of ​​each connected domain in the smoke gradient image is obtained, and the frequency of the area of ​​each connected domain under different smoke gradient orders in the smoke gradient image is counted. The elements in the smoke area measurement matrix It represents the frequency of occurrence of a connected domain with a smoke gradient order of i and an area of ​​j in the smoke gradient image. The dimension of the smoke region measurement matrix is ​​M×N, where N is the maximum area of ​​all connected domains in the smoke gradient image.

[0025] Furthermore, the smoke influence degree of the smoke gradient image is obtained according to the distribution of each element in the smoke area measurement matrix, and the expression is:

[0026]

[0027] In the formula, Represents the frequency of appearance of a connected domain with smoke gradient order i and area j in the smoke gradient image. The degree of smoke influence on the smoke gradient image.

[0028] Furthermore, the smoke recognition factor of each pixel is constructed according to the smoke influence degree of the smoke gradient image of the smoke region characteristic value of each pixel, including:

[0029] The smoke impact degree is taken as the exponent of an exponential function with a natural constant as the base, the sum of the exponential function and the number 1 is obtained, and the product of the smoke area characteristic value of each pixel point and the sum value is taken as the smoke identification factor of each pixel point.

[0030] The present invention has at least the following beneficial effects:

[0031] According to the characteristic that the gray value of the smoke area is small, the present invention calculates the dark channel value of each pixel in the coal mine underground image based on the dark channel prior theory, obtains the coal mine dark channel image, and reduces the influence of other background areas. The diffusion of smoke, dust, etc. is usually closely related to the circulation of air. The air circulation degree is obtained according to the size of the dark channel value in the pixel neighborhood window and the pixel value of the three channels R, G, and B. Since the coal mine underground usually contains a lot of dust and the illumination is uneven, mining lamps are required for illumination, and the dust area and the mining lamp area are very similar to the characteristics of smoke. In order to detect smoke more accurately, according to the characteristics that the dust area and the light area both diffuse from the center to the surroundings, while the smoke diffuses in a single direction, the dark channel gradient direction complexity is calculated to distinguish the smoke area from dust and illumination. The air circulation degree and the dark channel gradient direction complexity are combined to obtain the smoke area characteristic value, and to judge whether the pixel point is in the smoke area. Not only the characteristic that the dark channel gradient direction inside the smoke is basically consistent is considered, but also the flow characteristics of the smoke in the air are taken into account, so as to more accurately judge whether the pixel point is in the smoke area. The smoke gradient image is obtained according to the smoke pixel judgment value and the dark channel gradient amplitude, and then the smoke influence degree of the smoke gradient image is calculated, the smoke recognition factor is constructed, and the smoke area in the coal mine underground image is extracted. The present invention can realize accurate detection and recognition of coal mine underground smoke and improve the recognition accuracy of coal mine underground smoke. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 A flow chart of the steps of a coal mine underground smoke image recognition method provided by the present invention. DETAILED DESCRIPTION

[0034] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the coal mine underground smoke image recognition method proposed by the present invention, its specific implementation method, structure, characteristics and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0035] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0036] The specific scheme of the coal mine underground smoke image recognition method provided by the present invention is described in detail below with reference to the accompanying drawings.

[0037] An embodiment of the present invention provides a method for identifying smoke images in coal mines. Specifically, the following method for identifying smoke images in coal mines is provided. Figure 1 , the method comprises the following steps:

[0038] Step S001, using an explosion-proof camera to collect images of underground coal mines.

[0039] This embodiment aims to accurately identify smoke based on underground coal mine images, so it is necessary to first use an explosion-proof camera to collect images. Cameras are needed for daily inspections, documentary filming, geological surveys, etc. in coal mines, but there are flammable and explosive gases and substances such as gas and coal dust in coal mines. When the underground equipment is operating normally or has a fault, it will produce electric sparks, arcs, hot surfaces and burning particles, etc. They all have thermal energy and may become factors that ignite the ignition source and heat source of mine gas and coal dust. Ordinary cameras have not been improved with explosion-proof technology and may produce sparks with the flammable gas underground, thereby causing the risk of explosion. Explosion-proof cameras can be used in dangerous places in coal mines, mainly for coal mine disaster accident investigation and evidence collection and daily geological cataloging, recording underground production safety, electromechanical equipment operation status, roof support conditions and geological characteristics, etc., so explosion-proof cameras can be used to collect underground coal mine images.

[0040] So far, the coal mine underground images are collected according to the above method of this embodiment to obtain basic image data for coal mine underground smoke recognition.

[0041] Step S002, calculate the dark channel value of each pixel point to obtain the coal mine dark channel image, then use the Sobel operator to convolve with the coal mine dark channel image to obtain the dark channel gradient amplitude and direction of each pixel point, obtain the smoke area characteristic value in the neighborhood window of the pixel point according to its air circulation and the complexity of the dark channel gradient direction, and judge whether the pixel point is in the smoke area, and then obtain the smoke gradient image according to the smoke pixel judgment value and the dark channel gradient amplitude, and then calculate the smoke impact degree of the smoke gradient image.

[0042] Based on the dark channel prior theory, the size of the dark channel value reflects the degree to which the pixel is affected by dust, smoke, etc. The dark channel image of the coal mine can be obtained by calculating the dark channel value of each pixel in the coal mine underground image. However, due to the high dust content in the coal mine, the dark channel value cannot be directly used to approximate the degree to which the pixel is affected by smoke. It is necessary to distinguish the smoke area from the dust area and the light area in order to accurately extract the smoke area.

[0043] Compared with other areas, due to the diffusion of smoke, the pixel values ​​inside the smoke area show a certain regular change. In order to distinguish the smoke area from other areas, the Sobel operator is used to convolve with the coal mine dark channel image to obtain the dark channel gradient amplitude of each pixel point. and direction .

[0044] The diffusion of smoke is usually closely related to the circulation of air. The horizontal diffusion of smoke mainly depends on wind speed and air circulation. Generally speaking, the concentration of smoke in coal mines is inversely proportional to the wind speed and air circulation speed underground. The wind direction and air circulation direction affect the diffusion direction of smoke. Usually, the diffusion direction of smoke is in the direction of air circulation or downwind.

[0045] Since air is colorless, it cannot be seen by the naked eye and cannot be captured by the camera, but dust, smoke, etc. will be affected by air circulation. Therefore, the concentration changes of dust and smoke in the image correspond to the air circulation. The air circulation in the coal mine can be estimated based on the gradient changes of the dust and smoke pixels in the image. Compared with the interfering mine equipment, the pixel values ​​of the R, G, and B channels of dust and smoke are roughly equal, and the dark channel value is larger. With each pixel as the center, a neighborhood window of each pixel is constructed. The size of the neighborhood window is a×a, and the value in this embodiment is 13×13. The mean of the dark channel value in the neighborhood window of pixel point x is recorded as , the variances of the channel values ​​of R, G, and B channels in the neighborhood window of pixel point x are , and , then the flow area characteristic value at pixel point x It can be expressed as follows:

[0046]

[0047] In the formula, , , are the variances of the channel values ​​of the three channels R, G, and B of all pixels in the neighborhood window of pixel point x, which represents the consistency of the channel values ​​of the three channels R, G, and B in the neighborhood window of pixel point x. The larger the variance, the more inconsistent the channel values ​​of the three channels R, G, and B in the neighborhood window, and the less likely the pixel is to be in a flow area such as dust or smoke. The characteristic value of the flow area The smaller; is the mean of the dark channel values ​​in the neighborhood window of pixel point x, representing the degree to which the pixel point is affected by dust, smoke, etc. The larger the dark channel value, the more likely the pixel point is in a flow area such as dust and smoke. The bigger.

[0048] In the dust, smoke and other flowing areas, the size of the dark channel value reflects the degree of influence of dust, smoke, etc. on the pixel point, and the gradient amplitude of the dark channel value represents the flow speed of dust, smoke, etc., which indirectly reflects the air circulation at the pixel point. Combining the characteristic value of the flow area at the pixel point x and the gradient amplitude of the dark channel, the air circulation at the pixel point x is It can be expressed as follows:

[0049]

[0050] in, is the characteristic value of the flow area at the pixel point x, which represents the possibility that the pixel point is in the flow area such as dust and smoke. The larger the characteristic value, the more the area can reflect the air circulation, and the greater the weight is given; is the dark channel gradient amplitude at pixel point x, representing the degree of change of the dark channel value in the area. The greater the degree of change, the faster the flow speed of dust, smoke, etc., and the air circulation The higher.

[0051] The smoke area is distinguished from dust and light according to the characteristics that the smoke concentration inside the smoke area gradually decreases in one direction, while dust and light diffuse from the center to the surroundings. In the neighborhood window of each pixel, the directional complexity of the dark channel gradient is calculated.

[0052] In order to facilitate calculation and reduce the amount of calculation, 0-180 degrees is evenly divided into directions, In this embodiment, the value is 9, that is, the directions are Then, the dark channel gradient direction of each pixel in the neighborhood window is counted to obtain the dark channel gradient direction histogram of each pixel , and normalized to , for the data point in the normalized dark channel gradient direction histogram , representing the The frequency of the occurrence of a direction in all dark channel gradient directions in the neighborhood window of pixel point x. The histogram statistics acquisition process is a prior art and will not be described in detail in this embodiment. It can be expressed as follows:

[0053]

[0054] in, Indicates The corresponding value of each direction in the normalized dark channel gradient direction histogram of the pixel point x. When the number of dark channel gradient directions of each pixel in the neighborhood window of the central pixel point is more, the complexity of the dark channel gradient direction of the pixel point is greater; when the dark channel gradient direction of each pixel in the neighborhood window of the central pixel point is more single, the complexity of the dark channel gradient direction of the pixel point is smaller.

[0055] Smoke will flow with the circulation of air, which can indirectly reflect the air circulation. Therefore, the air circulation is larger in the smoke area, while the air circulation is smaller in other static areas. Due to the continuity of smoke flow, the smoke concentration gradually decreases from the smoke source to the boundary of the smoke diffusion direction. The gradient distribution of each pixel is often uniform, and the smoke is generated by the fire source. The gradient direction is often consistent with the smoke diffusion direction. If the pixel x is in the smoke area, the dark channel gradient direction in its neighborhood window is mostly consistent with the smoke diffusion direction. The comprehensive air circulation and the complexity of the dark channel gradient direction can be used to judge whether the pixel is in the smoke area. The smoke area feature value of pixel x is It can be expressed as follows:

[0056]

[0057] in, is the air circulation degree at pixel point x, representing the flow speed of dust and smoke. The greater the air circulation degree, the faster the flow speed of dust and smoke. The bigger; is the complexity of the dark channel gradient direction of pixel x, which represents the dispersion degree of the dark channel gradient direction of each pixel in the pixel neighborhood window. The lower the complexity of the dark channel gradient direction, the more single the dark channel gradient direction in the pixel neighborhood window is, and the more likely it is that the smoke is flowing in a single direction. The characteristic value of the smoke area The bigger.

[0058] Taking into account the internal characteristics of the smoke itself and its flow characteristics in the air, the threshold is set to determine whether the pixel is in the smoke area according to the characteristic value of the smoke area. , when the smoke area feature value of a pixel is greater than the threshold, the pixel is judged to be in the smoke area. The value of this embodiment is 0.7. Smoke pixel judgment value It can be expressed as follows:

[0059]

[0060] In the formula, The threshold for smoke pixels. is the smoke area eigenvalue of pixel x, representing the possibility that the pixel is in the smoke area. The larger the eigenvalue, the smaller the complexity of the dark channel gradient direction. Since smoke generally diffuses in one direction, the more complex the gradient direction of the dark channel value of the pixel is, the smaller the complexity of the dark channel gradient direction is, the more likely the pixel is to be in the smoke area. The smoke pixel judgment value Indicates that the pixel is in the smoke area.

[0061] Since the dark channel value reflects the degree to which the pixel is affected by dust, smoke, etc., for the smoke area, the magnitude of the dark channel gradient reflects the degree of smoke diffusion. The smoke gradient of pixel x is obtained by multiplying the dark channel gradient amplitude, air circulation, and smoke pixel judgment value, that is, , calculate the smoke gradient for each pixel in the coal mine underground image and obtain the smoke gradient image In the smoke gradient image, the pixel value of the non-smoke area is 0, and the pixel value of the smoke area is the size of its dark channel gradient, which reflects the degree of diffusion of the smoke at the pixel point.

[0062] The smoke region measurement matrix is ​​constructed based on the smoke gradient image to measure the smoke impact in the smoke gradient image. First, this embodiment divides the smoke gradient into M smoke gradient orders on average. M is 10 in this embodiment. The smoke region measurement matrix is ​​M×N, where M is the number of smoke gradient orders and N is the maximum area of ​​all connected domains in the smoke gradient image. The matrix elements are The number of occurrences of the connected domain with the order i of smoke gradient and the area j in the smoke gradient image is normalized to obtain the smoke area measurement matrix in the final frequency form. Then the smoke influence degree of the smoke gradient image is It can be expressed as follows:

[0063]

[0064] In the above formula represents the product, Represents the frequency of appearance of a connected domain with smoke gradient order i and area j in the smoke gradient image. The larger the value, the more times the connected area appears, indicating that the wider the influence of the smoke is; i represents the order of smoke gradient, the larger the value, the higher the smoke diffusion degree of the connected area, and the greater the influence of the smoke; j represents the size of the connected area, the larger the value, the larger the coverage of the connected area, and the greater the influence of the smoke. For the connected areas with larger smoke gradient order, larger area, and more occurrences in the smoke gradient image, they are more affected by the smoke.

[0065] Step S003, performing smoke recognition on the coal mine underground image according to the smoke area characteristic value of each pixel point and the smoke influence degree of the smoke gradient image.

[0066] Finally, this embodiment identifies the smoke situation in the coal mine underground image according to the smoke area feature value and the smoke influence degree of the smoke gradient image, and constructs a smoke recognition factor:

[0067]

[0068] In the above formula, is the smoke recognition factor of pixel x, is the smoke impact degree of the smoke gradient image, which represents the impact degree of smoke in coal mines. The greater the impact degree, the higher the weight given to the pixel points in the image when identifying smoke. is the smoke area eigenvalue of pixel x. The larger the smoke area eigenvalue is, the higher the possibility that the pixel is in the smoke area.

[0069] The larger the smoke recognition factor, the higher the possibility that smoke exists at the pixel point at the corresponding position. The smoke recognition factor of the pixel point is normalized, and a threshold is preset. When the normalized smoke recognition factor of the pixel point is greater than the threshold, the pixel point is a smoke pixel point. Among them, the threshold value in this embodiment is 0.6. The smoke recognition factor is calculated for all pixels in the coal mine underground image to extract all smoke areas in the coal mine underground image and realize accurate recognition of coal mine underground smoke.

[0070] In summary, according to the above process of this embodiment, accurate identification of smoke in coal mines can be achieved based on the image data processing and identification process, thereby ensuring the accuracy of smoke identification in coal mines.

[0071] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0072] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0073] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Modifications to the technical solutions recorded in the aforementioned embodiments, or equivalent replacement of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for recognizing smoke images in coal mines, characterized in that: The method comprises the following steps: Collect underground coal mine images, combine dark channel algorithms to obtain coal mine dark channel images, and use edge detection to obtain the dark channel gradient amplitude and direction of each pixel in the coal mine dark channel images; According to the distribution of the dark channel value and the R, G, and B channel values ​​in the neighborhood window of each pixel, the flow area characteristic value of each pixel is constructed; according to the flow area characteristic value and the dark channel gradient amplitude of each pixel, the air circulation of each pixel is constructed; according to the distribution of the dark channel gradient direction in the neighborhood window of each pixel, the dark channel gradient direction complexity of each pixel is obtained; according to the air circulation of each pixel and the dark channel gradient direction complexity of each pixel, the smoke area characteristic value of each pixel is obtained; according to the smoke area characteristic value of each pixel and the preset smoke pixel judgment threshold, the smoke pixel judgment value of each pixel is obtained; according to the air circulation of each pixel, the smoke pixel judgment value and the dark channel gradient amplitude, the smoke gradient of each pixel is obtained, and the smoke gradients of all pixels constitute a smoke gradient image; according to the distribution of the smoke gradient of each pixel in the smoke gradient image, a smoke area measurement matrix is ​​constructed; according to the distribution of each element in the smoke area measurement matrix, the smoke influence degree of the smoke gradient image is obtained; The smoke recognition factor of each pixel is constructed according to the smoke area characteristic value of each pixel and the smoke influence degree of the smoke gradient image. The pixel whose smoke recognition factor is higher than the preset threshold is regarded as the smoke pixel to complete the smoke recognition in the coal mine. The flow area characteristic value of each pixel is constructed according to the distribution of the dark channel value and the R, G, and B channel values ​​in the neighborhood window of each pixel, including: A neighborhood window is constructed with each pixel as the center, the mean of the dark channel values ​​of all pixels in the neighborhood window of each pixel is calculated, the variance of the channel values ​​of all pixels in the neighborhood window of each pixel in the R, G, and B channels is calculated, the sum of the three variances is obtained, the ratio of the mean to the sum is calculated, and the ratio is used as the flow area characteristic value of each pixel; The smoke area characteristic value of each pixel point is obtained according to the air circulation degree of each pixel point and the complexity of the dark channel gradient direction, including: Obtain the ratio of the air circulation degree of each pixel to the complexity of the gradient direction of the dark channel, use the inverse of the ratio as the exponent of an exponential function with a natural constant as the base, and use the difference between the number 1 and the calculation result of the exponential function as the characteristic value of the smoke area of ​​each pixel; The method of constructing a smoke identification factor for each pixel point according to the smoke region characteristic value of each pixel point and the smoke influence degree of the smoke gradient image includes: taking the smoke influence degree as an exponent of an exponential function with a natural constant as a base, obtaining a sum of the exponential function and the number 1, and taking the product of the smoke region characteristic value of each pixel point and the sum as the smoke identification factor for each pixel point; The dark channel gradient direction complexity of each pixel point is obtained according to the distribution of the dark channel gradient direction in the neighborhood window of each pixel point, including: Divide 180° into directions, the dark channel gradient directions of the pixels in the neighborhood window of each pixel are counted to obtain the dark channel gradient direction histogram of each pixel, and the normalized dark channel gradient direction histogram is obtained after normalization. The complexity of the dark channel gradient direction of the pixel x is The expression is: In the formula, For the The corresponding value of each direction in the normalized dark channel gradient direction histogram of pixel point x; The smoke gradient of each pixel is obtained according to the air circulation degree of each pixel, the smoke pixel judgment value and the dark channel gradient amplitude, including: The product of the air circulation degree of each pixel, the smoke pixel judgment value and the dark channel gradient amplitude is taken as the smoke gradient of each pixel; The smoke region measurement matrix is ​​constructed according to the smoke gradient distribution of each pixel in the smoke gradient image, including: The smoke gradients of all pixels in the smoke gradient image are evenly divided into M smoke gradient orders, the area of ​​each connected domain in the smoke gradient image is obtained, and the frequency of the area of ​​each connected domain under different smoke gradient orders in the smoke gradient image is counted. The elements in the smoke area measurement matrix Represents the frequency of occurrence of a connected domain with a smoke gradient order of i and an area of ​​j in the smoke gradient image. The dimension of the smoke region measurement matrix is ​​M×N, where N is the maximum area of ​​all connected domains in the smoke gradient image. The smoke influence degree of the smoke gradient image is obtained according to the distribution of each element in the smoke area measurement matrix, and the expression is: In the formula, Represents the frequency of appearance of a connected domain with smoke gradient order i and area j in the smoke gradient image. The degree of smoke influence on the smoke gradient image.

2. A coal mine smoke image recognition method as claimed in claim 1, characterized in that: The air flow degree of each pixel point includes: taking the product of the flow area characteristic value of each pixel point and the dark channel gradient amplitude as the air flow degree of the pixel point.

3. A coal mine smoke image recognition method as claimed in claim 1, characterized in that: The method of obtaining the smoke pixel judgment value of each pixel point according to the smoke region characteristic value of each pixel point and a preset smoke pixel judgment threshold value includes: When the smoke region characteristic value of a pixel point is less than a preset smoke pixel judgment threshold, the smoke pixel judgment value of the pixel point is 0; otherwise, the smoke pixel judgment value of the pixel point is 1.

Images