Belt material conveying fire early warning method based on image processing

Abstract

The invention relates to the technical field of image data processing, in particular to a belt material conveying fire early warning method based on image processing. The method comprises the following steps: firstly, acquiring a reference RGB image when a conveyor belt is normally not on fire, and extracting an ROI from the reference RGB image; the method comprises the following steps: firstly, acquiring an actual scene image at the same shooting angle position, performing smoke analysis on an ROI region of the actual scene image to obtain smoke confidence of each ROI pixel point in the ROI region, and performing fire early warning identification on the actual scene image based on the smoke confidence of each ROI pixel point so as to detect whether a conveyor belt has a fire or not. And early fire smoke generated by comparing flames is detected, so that fire signs can be found in advance, and measures can be taken earlier.

Description

Technical Field

[0001] The invention relates to the technical field of image data processing, and in particular to a belt conveyor fire early warning method based on image processing. Background Art

[0002] In mines, coal is usually transported by belt conveyors. When electrical failures, roller failures, or spontaneous combustion of coal occur in belt conveyors, the conveyor belts may catch fire. The space in mines is small and there are many flammable gases. Once a belt catches fire, it may directly ignite the coal, and the fire spreads to various nearby equipment and buildings, causing equipment damage, building burns, and even the loss of life of mine workers. In order to deal with the fire, production must be suspended and repair work must be carried out, which will delay production plans and directly cause economic losses to the company.

[0003] The existing fire warning system mainly focuses on flame recognition. However, when the flame appears, the fire has already occurred. Therefore, this flame detection fire warning method cannot provide early warning and avoid fire. Summary of the invention

[0004] The technical problem to be solved by the present invention is that the existing fire warning system cannot provide early warning and avoid fire. In view of the defects in the prior art, a belt conveyor fire warning method based on image processing is provided.

[0005] In order to solve the above technical problems, the present invention provides a belt conveyor fire warning method based on image processing, comprising: Capturing an image of a normal conveyor belt that is not on fire to obtain a reference RGB image, and extracting a ROI region from the reference RGB image; Acquire an actual scene image of the conveyor belt at the same shooting angle position as the reference RGB image; Performing smoke analysis on the ROI region of the actual scene image to obtain smoke confidence of each ROI pixel in the ROI region; Fire warning recognition is performed on the actual scene image based on the smoke confidence of each ROI pixel point.

[0006] Optionally, extracting a ROI region from the reference RGB image includes: Converting the reference RGB image into a reference grayscale image; Performing Sobel operator edge detection on the reference grayscale image to obtain edge contours in the reference grayscale image; Performing Hough line transform on the edge contour to obtain the coordinates of each straight line segment in the reference grayscale image and each pixel point on the straight line segment; Counting the number of pixels on each straight line segment as the length of the straight line segment, and selecting two straight line segments with the largest lengths to be determined as edge lines; Obtain edge endpoints of the two edge lines; Connect the edge endpoints located at the same image boundary and extend them to both sides to obtain boundary lines; draw endpoint vertical lines passing through the edge endpoints and perpendicular to the boundary lines respectively; The largest area enclosed by the boundary line and the vertical line of the endpoints is taken as the ROI area.

[0007] Optionally, capturing an actual scene image of a conveyor belt at the same shooting angle as the reference RGB image comprises: Using a CMOS camera to periodically capture an actual RGB image of the conveyor belt at the same shooting angle as the RGB image; Preprocessing the actual RGB image using median filtering to obtain an actual scene image; The actual scene image is converted into an actual scene grayscale image.

[0008] Optionally, performing smoke analysis on the ROI region of the actual scene image to obtain smoke confidence of each ROI pixel in the ROI region includes: Extracting the ROI region of the actual scene grayscale image; Constructing a neighborhood sliding window for ROI pixel points in the ROI area; The smoke feature of the ROI area is extracted by using the neighborhood sliding window to obtain the smoke probability factor of the ROI pixel point; Calculate the clarity of each ROI pixel based on the neighborhood sliding window; Calculate the smoke clarity diffusion degree of each ROI pixel point based on the clarity of each ROI pixel point; The smoke confidence of each ROI pixel is calculated based on the smoke probability factor and the smoke clarity diffusion.

[0009] Optionally, extracting smoke features of the ROI region using the neighborhood sliding window to obtain a smoke probability factor of the ROI pixel point includes: Using the neighborhood sliding window to obtain the gray level co-occurrence matrix of each ROI pixel in the ROI area; Calculating the contrast and entropy of the gray level co-occurrence matrix as the neighborhood contrast and neighborhood entropy of the ROI pixel point; Calculating the texture richness of the ROI pixel point according to the neighborhood contrast and the neighborhood entropy; For the ROI pixel point whose texture richness is less than or equal to the texture threshold, calculating the neighborhood grayscale variance of the neighborhood sliding window corresponding to the ROI pixel point; The smoke probability factor of the ROI pixel is calculated based on the texture richness and the neighborhood grayscale variance.

[0010] Optionally, the texture richness is calculated according to the following formula: In the formula, is the texture richness; is the neighborhood contrast; is the neighborhood entropy.

[0011] Optionally, the smoke probability factor is calculated according to the following formula: In the formula, is the smoke probability factor; is the normalized texture richness, is the neighborhood grayscale variance; is the texture threshold.

[0012] Optionally, calculating the clarity of each ROI pixel based on the neighborhood sliding window includes: Calculate the gradient value of each ROI pixel point; Performing Harris corner point detection on the ROI area, and counting the number of corner points in the neighborhood sliding window corresponding to each ROI pixel point; Calculate the Euclidean distance between each non-central ROI pixel point and the central ROI pixel point in the neighborhood sliding window corresponding to each ROI pixel point; The clarity of the ROI pixel point as the center is calculated based on the gradient value of each ROI pixel point in the neighborhood sliding window, the number of corner points, and the Euclidean distance from the non-central ROI pixel point to the central ROI pixel point.

[0013] Optionally, the clarity is calculated according to the following formula: In the formula, is the first The gradient value of the non-central ROI pixel point; is the gradient value of the ROI pixel at the center of the neighborhood sliding window; is the first The Euclidean distance from the non-central ROI pixel point to the central ROI pixel point; is the sum of all Euclidean distances within the neighborhood sliding window; is the number of corner points in the neighborhood sliding window.

[0014] Optionally, calculating the smoke clarity diffusion degree of each ROI pixel point based on the clarity of each ROI pixel point includes: The clarity of each of the ROI pixel points in the neighborhood sliding window is constructed into a clarity matrix according to the arrangement of the ROI pixel points; Calculating the clarity mean of each row in the clarity matrix and the Euclidean distance between different rows; The smoke clarity diffusion of the ROI pixel point as the center is calculated based on the clarity mean value of each row and the Euclidean distance between different rows.

[0015] Optionally, the smoke clarity diffusion is calculated according to the following formula: In the formula, It is the clear diffusion of smoke; and They are the first and The mean clarity of the rows, and ; is the clarity matrix and Euclidean distance between rows.

[0016] Optionally, calculating the smoke confidence of each ROI pixel point based on the smoke probability factor and the smoke clarity diffusion degree includes: Classify the ROI pixel points whose smoke probability factor is greater than or equal to the smoke probability threshold as high-probability smoke points; classify the ROI pixel points whose smoke probability factor is less than the smoke probability threshold as non-high-probability smoke points; The smoke confidence of each of the high-probability smoke points is calculated based on the smoke probability factor and the smoke clarity diffusion degree, and the smoke confidence of each of the non-high-probability smoke points is calculated based on the smoke probability factor.

[0017] Optionally, the smoke confidence is calculated according to the following formula: In the formula, is the smoke confidence; is the normalized smoke probability factor; is the normalized smoke clarity diffusion; is the smoke probability threshold.

[0018] Optionally, performing fire warning recognition on the actual scene image based on the smoke confidence of each ROI pixel point includes: Performing threshold segmentation on each of the ROI pixel points of the actual scene image according to the smoke confidence, and classifying the ROI pixel points whose smoke confidence is greater than or equal to the confidence threshold as smoke points; The number of the smoke points in the actual scene image is counted, and when the number of the smoke points exceeds a smoke threshold, it is determined that a fire occurs on the conveyor belt.

[0019] Optionally, perform threshold segmentation according to the following formula: In the formula, is the coordinate point The corresponding binary value of the ROI pixel point, a binary value of 0 indicates that the ROI pixel point is the smoke point, and a binary value of 1 indicates that the ROI pixel point is a non-smoke point; is the coordinate point The normalized smoke confidence of the corresponding ROI pixel point; is the confidence threshold.

[0020] The implementation of the belt conveyor fire warning method based on image processing of the present invention has the following beneficial effects: Different from the traditional flame detection and fire warning method, the present invention detects fire smoke that is generated earlier than the flame during the fire, so that signs of fire can be discovered earlier and measures can be taken earlier. The present invention improves the image threshold segmentation and no longer uses grayscale threshold for image segmentation. Instead, it analyzes the smoke characteristics such as texture richness and smoke probability factor classification of the ROI area in combination with the specific scene, and then constructs the smoke clarity diffusion degree in combination with the characteristics that the smoke caused by the conveyor belt fire will spread, so as to segment the actual scene image and improve the time and accuracy of fire warning. In addition, in combination with the actual conveyor belt fire scene, the ROI area is extracted based on the characteristics of the conveyor belt in the actual scene image, and only the ROI area is analyzed, which improves the efficiency of detection and warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 1 is a schematic diagram of the main steps of a belt conveyor fire warning method based on image processing according to an embodiment of the present invention; Figure 2It is a schematic diagram of the overall logic of a belt conveyor fire warning method based on image processing according to an embodiment of the present invention; Figure 3 It is a flowchart of image preprocessing of a belt conveyor fire warning method based on image processing, which can be used as a reference embodiment of the present invention; Figure 4 is a schematic diagram of the ROI area of ​​an embodiment of the present invention; Figure 5 It is a flow chart of constructing smoke clarity diffusion of a belt conveyor fire warning method based on image processing, which can be used as a reference embodiment of the present invention; Figure 6 It is a flowchart of threshold segmentation of a belt conveyor fire warning method based on image processing, which can be used as a reference embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, a belt conveyor fire warning method based on image processing provided by an embodiment of the present invention mainly includes the following steps: Step 101: Capture an image of a normal conveyor belt that is not on fire to obtain a reference RGB image, and extract a ROI area from the reference RGB image.

[0024] Different from the traditional flame detection fire warning method, the embodiment of the present invention adopts the fire smoke detection fire warning method. During a fire, smoke is generated earlier than flames. Detecting fire smoke can detect signs of fire earlier and take measures to deal with it earlier. The embodiment of the present invention uses machine vision technology to monitor the conveyor belt in real time and detect signs of fire in time. In order to monitor the conveyor belt, it is necessary to obtain visual data of the normal conveyor belt that has not caught fire in advance, that is, a reference RGB image. It should be noted that the reference RGB image can be obtained by a CMOS camera.

[0025] Among them, a CMOS camera is a camera that uses a complementary metal oxide semiconductor (CMOS sensor) to convert optical images into digital signals for transmission. An RGB image is an image based on a red, green, and blue color model in the industry. Various colors are obtained by changing the three color channels of red (R), green (G), and blue (B) and superimposing them on each other. RGB represents the colors of the three channels of red, green, and blue. This color model includes almost all colors that can be perceived by human vision and is one of the most widely used color systems. In addition, bilateral filtering is a nonlinear filtering method, and its specific implementation process can refer to existing technical solutions or be implemented using existing computer programs, and the embodiments of the present invention will not be repeated.

[0026] In an embodiment of the present invention, the step of extracting the ROI area from the reference RGB image can be implemented in the following manner: converting the reference RGB image into a reference grayscale image; performing Sobel operator edge detection on the reference grayscale image to obtain the edge contour in the reference grayscale image; performing Hough line transform on the edge contour to obtain the coordinates of each straight line segment in the reference grayscale image and each pixel point on the straight line segment; counting the number of pixels on each straight line segment as the length of the straight line segment, and selecting two straight line segments with the largest lengths to be determined as edge lines; obtaining edge endpoints of the two edge lines; connecting the edge endpoints located at the same image boundary and extending them to both sides to obtain a boundary line; drawing endpoint perpendicular lines passing through the edge endpoints and perpendicular to the boundary line respectively; and taking the largest area enclosed by the boundary line and the endpoint perpendicular line as the ROI area.

[0027] Among them, the Sobel operator is an important processing method in the field of computer vision. It is to detect the edge by taking the weighted difference of the grayscale values ​​of the four areas of up, down, left, and right of each pixel in the image and reaching the extreme value at the edge. The Sobel operator is mainly used to obtain the first-order gradient of a digital image. Its common application and physical meaning is edge detection. The specific implementation process of the Sobel operator can refer to the existing technical solutions or be implemented using existing computer programs, and the embodiments of the present invention will not be repeated.

[0028] In the actual scene image of the conveyor belt, the target is the belt. Under normal circumstances, the belt is relatively straight, the edge is obvious, and the edge of the belt is quite different from the surrounding environment. The Sobel operator edge detection is performed on the grayscale image of the reference RGB image to obtain the edge contour in the reference grayscale image. Then, the Hough line transform is used to obtain the straight line segments in the reference grayscale image. At the same time, the coordinates of each pixel on each straight line segment are obtained. The number of pixels on each straight line segment is counted, which is the length of the straight line segment. The edge straight line of the belt is the longest on the reference grayscale image. The lengths of all straight line segments are arranged in descending order. The two longest straight line segments are the two edge straight lines of the belt. At this time, by comparing the horizontal and vertical coordinates of all points in the edge segment, the two endpoints of each edge segment are found, and a total of four endpoints are obtained. The edge endpoints located at the same image boundary are connected and extended to both sides to obtain the boundary line. Then, the endpoint perpendicular lines passing through the endpoints and perpendicular to the boundary line are drawn for the four endpoints, and a total of four endpoint perpendicular lines are obtained. The area enclosed by the two outermost endpoint perpendicular lines and the boundary line can be used as the ROI (region of interest) area at most for subsequent analysis, and the remaining two endpoint perpendicular lines are included in the ROI area. The ROI area is the area that actually needs to be analyzed, and the pixels within it are called ROI pixels.

[0029] Step 102: Capture an actual scene image of the conveyor belt at the same shooting angle as the reference RGB image.

[0030] The embodiment of the present invention uses a fixed camera position to capture images, so that the ROI area in each frame of image or each image is unchanged, that is, the ROI area of ​​the reference RGB image and the actual scene image is the same, so as to extract the ROI area in each frame of image for analysis. It should be noted that the actual scene image can be shot at a fixed frequency, or it can be a single frame image extracted from a real-time monitoring image at a fixed frequency.

[0031] In an embodiment of the present invention, step 102 can be implemented in the following manner: using a CMOS camera to periodically capture an actual RGB image of the conveyor belt at the same shooting angle as the RGB image; using a median filter to preprocess the actual RGB image to obtain an actual scene image; and converting the actual scene image into an actual scene grayscale image.

[0032] The actual RGB image can also be collected by a CMOS camera, and the frequency of collection can be adjusted as needed, which is not limited in the embodiment of the present invention. All actual RGB images are preprocessed to eliminate the interference caused by their own equipment and external environment, thereby improving the accuracy of subsequent analysis and operation. As a preferred implementation, median filtering can be used to preprocess the actual RGB image, and then all actual RGB images can be converted into grayscale images. In addition, other denoising methods can also be used, without being limited to the median filtering method.

[0033] Step 103: Perform smoke analysis on the ROI region of the actual scene image to obtain the smoke confidence of each ROI pixel in the ROI region.

[0034] Since the conveyor belt is a chemical material, an organic mixture, burning will produce black smoke, so the smoke that appears in the early stage of the belt fire is a mixture of gas, solid or particles produced by the belt combustion. At the same time, the combustion in the early stage of the fire will generate a lot of heat and smoke, causing the air to expand and contract greatly. A lot of hot air rises from the fire source. These hot gases gather together to form a relatively stable air column, and the surrounding air is also pulled to flow to the middle of the air column, thus forming a rising column of smoke in the early stage of the fire. Based on this, this step analyzes the smoke in the ROI area to obtain the smoke confidence of each ROI pixel.

[0035] In the embodiment of the present invention, step 103 can be implemented in the following manner: construct a neighborhood sliding window for the ROI pixel points in the ROI area; use the neighborhood sliding window to extract the smoke characteristics of the ROI area to obtain the smoke probability factor of the ROI pixel points; calculate the clarity of each ROI pixel point based on the neighborhood sliding window; calculate the smoke clarity diffusion of each ROI pixel point based on the clarity of each ROI pixel point; calculate the smoke confidence of each ROI pixel point based on the smoke probability factor and the smoke clarity diffusion.

[0036] In the actual scene grayscale image, the pixel points in the ROI area (i.e., ROI pixel points) are the objects of analysis, and the embodiment of the present invention uses a neighborhood sliding window to obtain the smoke features of the ROI pixel points. As a preferred implementation, the size of the neighborhood sliding window is a 5*5 square with a step size of 1, that is, each ROI pixel point corresponds to a 5*5 neighborhood sliding window with it as the center.

[0037] In an embodiment of the present invention, a neighborhood sliding window is used to extract smoke features in an ROI area to obtain a smoke probability factor of an ROI pixel, which can be achieved in the following manner: a grayscale co-occurrence matrix of each ROI pixel in the ROI area is obtained using a neighborhood sliding window; the contrast and entropy of the grayscale co-occurrence matrix are calculated as the neighborhood contrast and neighborhood entropy of the ROI pixel; the texture richness of the ROI pixel is calculated according to the neighborhood contrast and neighborhood entropy; for ROI pixels whose texture richness is less than or equal to a texture threshold, the neighborhood grayscale variance of the neighborhood sliding window corresponding to the ROI pixel is calculated; and the smoke probability factor of the ROI pixel is calculated based on the texture richness and the neighborhood grayscale variance.

[0038] In the mine environment, there are four kinds of materials in the ROI area of ​​the grayscale image of the actual scene. The first kind is the black coal transported by the belt. After mining and crushing, the coal is of different sizes, uneven, and has rich textures. The second kind is the belt body. The belt surface is extremely smooth and has almost no texture. The third kind is the black smoke produced after the belt is on fire. When the belt is on fire, a large number of particles will be generated, so that the smoke contains some particles and has less texture, but it is far less rich than the texture of coal. The fourth kind is other materials on the belt at the conveyor, including the rollers and railings of the conveyor. This part is messy and has rich textures. According to the richness of the material texture, the four materials can be divided into two categories. The first category is coal and rollers and railings with high texture richness, and the second category is belts and smoke with low texture richness.

[0039] As a preferred implementation, the texture richness of the ROI pixel is constructed using the neighborhood contrast and neighborhood entropy corresponding to the ROI pixel, and the texture richness can be calculated according to the following formula: In the formula, is the texture richness, is the neighborhood contrast, is the neighborhood entropy. Among them, the neighborhood contrast It is used to measure the contrast or difference of the grayscale values ​​in the neighborhood sliding window, which reflects the depth of the grooves in the image texture. The shallower the texture grooves, the greater the contrast, indicating a richer texture. It reflects the uniformity and randomness of the grayscale distribution of the neighborhood pixels. The larger the difference in grayscale values ​​of the ROI pixels, the more chaotic and random the distribution. The larger the value, the richer the texture. Normalized, the value range is [0,1], denoted as . Use Texture Richness A distinction can be made between coal, rollers or railing sections with high texture richness and belts or smoke sections with low texture richness. The first category is coal, rollers or railing sections with high texture richness that meet , while the texture-poor belt or smoke part of the second type of material satisfies . is the texture threshold, the value of the texture threshold can be determined as needed, and is preferably 0.3.

[0040] However, the texture richness of the belt and the smoke is low. If only the texture threshold is used, it cannot be distinguished and the two will be mixed together. Through research and investigation of specific scenarios, it is found that the surface of a normal, unburned belt is smooth, the grayscale values ​​between the ROI pixels in the neighborhood sliding window change evenly, and the grayscale values ​​of each ROI pixel are not much different. When the belt is on fire, it will release various gases, particles and combustion products, including fine soot particles. The soot particles are unevenly distributed in the smoke. The greater the concentration in a certain area, the darker the area. Different regional concentrations will cause the grayscale value of the smoke to change greatly. The grayscale value changes of each ROI pixel in the smoke are more obvious than those of the belt. Therefore, according to the different uniformity of the grayscale value changes of the belt and the smoke, the neighborhood grayscale variance in each neighborhood sliding window is calculated and recorded as , using neighborhood grayscale variance Distinguish between belts and black smoke. When it is small, it means that the grayscale value in the neighborhood sliding window changes evenly, that is, it may be a belt. A larger value indicates that the grayscale value changes in the sliding window of the neighborhood are chaotic, which may be a smoke point.

[0041] Using normalized texture richness It can distinguish between coal, rollers or railing parts with high texture richness and belts or smoke with low texture richness, and then use the neighborhood grayscale variance Belts and smoke can be distinguished. As a preferred implementation, texture richness TR and neighborhood grayscale variance are used. Construct the smoke probability factor, which can be calculated according to the following formula: In the formula, is the smoke probability factor, is the normalized texture richness, is the neighborhood grayscale variance; is the texture threshold. When , it means that the texture of the ROI pixel is richer. The higher the texture richness, the higher the smoke probability factor of the ROI pixel. The smaller. When the TR of the ROI pixel is small, the neighborhood grayscale variance of smoke is larger than that of the belt, and the neighborhood grayscale variance of the ROI pixel in the neighborhood sliding window is The larger the value, the smoke probability factor of the ROI pixel It should be noted that the same smoke probability factor Normalized, the value range is [0,1], denoted as .

[0042] In an embodiment of the present invention, the clarity of each ROI pixel is calculated based on the neighborhood sliding window, which can be implemented in the following manner: calculating the gradient value of each ROI pixel; performing Harris corner point detection on the ROI area, and counting the number of corner points in the neighborhood sliding window corresponding to each ROI pixel; calculating the Euclidean distance from each non-central ROI pixel to the central ROI pixel in the neighborhood sliding window corresponding to each ROI pixel; and calculating the clarity of the ROI pixel as the center based on the gradient value of each ROI pixel in the neighborhood sliding window, the number of corner points, and the Euclidean distance from the non-central ROI pixel to the central ROI pixel.

[0043] When the belt is on fire, it will release various gases, particles and combustion products, including fine carbon smoke particles, producing black smoke. These particles will rise and spread in the early stage of the fire. There are many and concentrated particles below the smoke, the smoke texture is richer, there are more corners, and it is clearer in the image, but there are fewer and scattered particles above, which are suspended in the air and interact with the light, scattering and absorbing the light, causing the light in the smoke to scatter, and the image is blurred, the smoke texture is scarce, and there are fewer corners. Therefore, the clarity of the ROI pixel point is constructed and calculated based on the neighborhood sliding window.

[0044] In a clear image, the obvious boundaries and textures of objects will cause drastic changes in pixel values, thereby generating larger gradient values. In a blurred smoke image, due to the blurring of boundary and texture information, the pixel values ​​change relatively slowly and the gradient values ​​are small. Therefore, the gradient value represents the rate of change of the grayscale value of the pixel in the image. By calculating the gradient values ​​of each point in the image, the degree of change of each pixel in the image can be obtained, thereby reflecting the clarity of the image. A larger gradient value indicates that the edges and textures are clearer and the image is relatively clearer; while a smaller gradient value indicates that the edges and textures are more blurred and the image is relatively blurred. Based on this, the gradient values ​​of all ROI pixels in the ROI area are calculated, and Harris corner detection is performed at the same time. The core of Harris corner detection is to use a local window to move on the image to determine whether the grayscale value has changed significantly. If the grayscale value in the window (on the gradient map) has changed significantly, then there is a corner point in the area where the window is located. The specific implementation process of Harris corner detection can refer to the existing technical solution or be implemented using an existing computer program, and the embodiments of the present invention will not be repeated.

[0045] As a preferred implementation, the clarity can be calculated according to the following formula: In the formula, It’s clarity; is the first The gradient value of the non-center ROI pixel; It is the gradient value of the ROI pixel at the center of the neighborhood sliding window; is the first The Euclidean distance from the non-central ROI pixel to the central ROI pixel; is the sum of all Euclidean distances within the neighborhood sliding window; is the number of corner points in the neighborhood sliding window. The smaller the value, the Euclidean distance from the non-central ROI pixel to the central ROI pixel. The smaller it is, the greater the influence of the non-central ROI pixel on the central ROI pixel. The larger the gradient value, the more non-central the ROI pixel point is. The larger the difference between the ROI pixel and the center, the more obvious the gradient value change in the neighborhood sliding window is, and the clarity of the ROI pixel at the center is. The higher, is the number of corner points in the neighborhood sliding window. The larger the number of corner points, the more obvious the change in grayscale value in the neighborhood sliding window is, and the clarity of the ROI pixel in the center is The higher.

[0046] Finally, the clarity of all ROI pixels in the ROI area of ​​the actual scene grayscale image is obtained. The detection target is the smoke generated by the belt fire. The smoke has a unique characteristic of spreading upward. Compared with the smoke above, the texture of the smoke below is richer and the clarity is better. For any ROI pixel point in the ROI area, the neighborhood sliding window corresponding to the ROI pixel point has the clarity of each ROI pixel point in the neighborhood sliding window. Based on this, the smoke clarity diffusion of the ROI pixel point is further calculated.

[0047] In an embodiment of the present invention, the smoke clarity diffusion of each ROI pixel is calculated based on the clarity of each ROI pixel, which can be achieved in the following manner: the clarity of each ROI pixel in the neighborhood sliding window is constructed into a clarity matrix according to the ROI pixel arrangement; the clarity mean of each row in the clarity matrix and the Euclidean distance between different rows are calculated; and the smoke clarity diffusion of the ROI pixel as the center is calculated based on the clarity mean of each row and the Euclidean distance between different rows.

[0048] The clarity can be constructed as a clarity matrix according to the ROI pixel arrangement , the clarity of different lines There is a difference, and the difference increases with distance. It can be expressed as: Calculate the mean clarity of each row and record it as , , , , , the clarity means of different rows are compared in pairs, so that different rows can be compared in pairs without duplication. There are 10 cases in total, and the distance between different rows is recorded as , the smoke clarity diffusion is calculated according to the following formula: In the formula, is the smoke clarity diffusion, and They are the first and The mean clarity of the rows (where and ); is the clarity matrix and The Euclidean distance between rows. The larger the value, the greater the difference in the mean clarity between different rows, the more intense the smoke diffusion, and the larger the smoke clarity diffusion b. It is a distance influencing factor. The farther the rows are from each other, the more obvious the image clarity changes, the more violent the smoke spreads, and the larger the smoke clarity diffusion b is. It should be noted that the smoke clarity diffusion b is also normalized to a value range of [0,1], denoted as B.

[0049] In an embodiment of the present invention, the smoke confidence of each ROI pixel is calculated based on the smoke probability factor and the smoke clarity diffusion degree, which can be implemented in the following manner: the ROI pixel points whose smoke probability factor is greater than or equal to the smoke probability threshold are classified as high-probability smoke points; the ROI pixel points whose smoke probability factor is less than the smoke probability threshold are classified as non-high-probability smoke points; the smoke confidence of each high-probability smoke point is calculated based on the smoke probability factor and the smoke clarity diffusion degree, and the smoke confidence of each non-high-probability smoke point is calculated based on the smoke probability factor.

[0050] Since smoke clarity is a unique feature of smoke, other substances do not have this feature. Therefore, for high probability smoke points ( ), whether the high-probability smoke point in the actual scene grayscale image is a smoke point caused by the belt fire depends on the smoke probability factor SP and smoke clarity diffusion B of the corresponding ROI pixel point; for non-high-probability smoke points ( ), in the actual scene grayscale image, the non-high probability smoke point is more likely to be a non-smoke point. Whether the non-high probability smoke point is a smoke point caused by belt ignition depends only on the smoke probability factor SP of the corresponding ROI pixel point. Among them, the smoke probability threshold The value of can be determined as needed, and is preferably 0.8.

[0051] As a preferred implementation, the smoke confidence can be expressed as: In the formula, is the smoke confidence; is the normalized smoke probability factor; is the normalized smoke clarity diffusion. Whether a ROI pixel belongs to a smoke point depends only on the smoke probability factor SP of the ROI pixel. The larger the smoke probability factor, the higher the smoke confidence of the ROI pixel. The bigger; for Whether a ROI pixel belongs to a smoke point depends not only on the smoke probability factor SP of the ROI pixel, but also on the smoke clarity diffusion B, a unique feature of smoke. The larger the smoke clarity diffusion B of the ROI pixel, the higher the smoke confidence of the ROI pixel. It should be noted that the same smoke confidence Normalized, the value range is [0,1], denoted as , It is the coordinates corresponding to the ROI pixel point.

[0052] It should be noted that the texture richness, smoke probability factor, smoke clarity diffusion and smoke confidence calculated by the embodiment of the present invention need to be normalized before being used in subsequent calculations. The specific implementation process can refer to the existing technical solutions or be implemented using existing computer programs, and the embodiments of the present invention will not be repeated here.

[0053] Step 104: Perform fire warning recognition on the actual scene image based on the smoke confidence of each ROI pixel point.

[0054] The smoke confidence of each ROI pixel can be used to analyze whether there are signs of fire in the ROI area of ​​the entire actual scene image, thereby realizing fire warning identification of the conveyor belt, so as to provide early warning before the fire occurs.

[0055] In an embodiment of the present invention, step 104 can be implemented in the following manner: threshold segmentation is performed on each ROI pixel point of the actual scene image according to the smoke confidence level, and the ROI pixel points whose smoke confidence level is greater than or equal to the confidence threshold level are classified as smoke points; the number of smoke points in the actual scene image is counted, and when the number of smoke points exceeds the smoke threshold level, it is determined that a fire has occurred on the conveyor belt.

[0056] This step uses threshold segmentation, takes the smoke confidence of all ROI pixels in the actual scene image as input, performs binarization on the smoke confidence of each ROI pixel, and outputs a binary image. Segmentation is achieved by comparing with the confidence threshold during the binarization process, wherein when the smoke confidence is greater than or equal to the confidence threshold, the binarization value of the ROI pixel is 0; when the smoke confidence is less than the confidence threshold, the binarization value of the ROI pixel is 1. The 0-value point in the binary image is a smoke point, and the 1-value point is a non-smoke point. The value of the confidence threshold can be determined as needed, preferably 0.9.

[0057] When the number of 0 value points (i.e. smoke points) in the binary image Exceeding the smoke threshold 4 o'clock ( ), indicating that the number of smoke points in the ROI pixel points in the actual scene image exceeds the smoke threshold. At this time, the conveyor belt has shown signs of fire, and there is a high probability that a fire will occur. A fire warning is required, such as issuing a fire warning signal through an intelligent system or triggering an alarm device to remind the staff to take appropriate measures. Smoke Threshold The value of 4 can be determined as needed, and is preferably 16.

[0058] As a preferred implementation, threshold segmentation can be performed according to the following formula: In the formula, is the coordinate point The corresponding ROI pixel point is binarized. A binarized value of 0 indicates that the ROI pixel point is a smoke point, and a binarized value of 1 indicates that the ROI pixel point is a non-smoke point. is the coordinate point The normalized smoke confidence of the corresponding ROI pixel; is the confidence threshold.

[0059] A belt conveyor fire warning method based on image processing in an embodiment of the present invention can monitor the conveyor belt in real time and detect signs of fire in time in a mine through machine vision technology, thereby providing early warning and minimizing potential risks and hazards. Moreover, the machine vision system can continuously monitor the conveyor belt for signs of fire during transportation without the need for continuous manual observation, thereby improving work efficiency and reducing the burden on staff. Through machine vision technology, the signs of fire detected by the system can be displayed in the form of images or videos, so that operators can intuitively understand the fire situation, make decisions quickly, and take measures. Finally, the machine vision system can record and store fire monitoring data for subsequent analysis and evaluation. These data can be used for investigation and analysis of fire incidents to improve safety precautions.

[0060] like Figure 2 As shown, the overall logic of a belt conveyor fire warning method based on image processing in an embodiment of the present invention includes the following three steps: 1. Obtain the conveyor belt image through the CMOS camera and pre-process the image.

[0061] First, an image of a normal, unfired actual scene is captured to obtain a reference RGB image I1. Then, the actual scene is continuously captured at the same shooting position to obtain actual RGB images I2, I3, ..., and all actual RGB images are preprocessed to eliminate interference caused by the own equipment and external environment, thereby improving the accuracy of subsequent analysis and operations. The embodiment of the present invention uses median filtering to preprocess the actual RGB image to obtain an actual scene image, and finally converts all actual scene images into actual scene grayscale images.

[0062] Second, according to the characteristics of the conveyor belt, the ROI area is extracted from the reference RGB image, and then the smoke and non-smoke substances in the grayscale image of the actual scene are analyzed to construct the smoke clarity diffusion.

[0063] The Sobel operator edge detection is performed on the grayscale image of the reference RGB image I1 to obtain the edge contour in the reference grayscale image. Then, the Hough line transform is used to obtain the straight line segments in the reference grayscale image. At the same time, the coordinates of each pixel on each straight line segment are obtained. The number of pixels on each straight line segment is counted, which is the length of the straight line segment. The edge straight line of the belt is the longest on the reference grayscale image. The lengths of all straight line segments are arranged in descending order and recorded as a1, a2, ..., an. The two longest straight line segments in the image (a1 and a2) are the two edge straight lines of the belt. At this time, by comparing the horizontal and vertical coordinates of all points in the edge segment, the coordinates of each endpoint of a1 and a2 are found, which are recorded as a1(x1, y1), a1(x2, y2), a2(x3, y3), and a2(x4, y4). The ROI area can be obtained by combining the image boundary and the four endpoints.

[0064] For the ROI pixel points in the ROI area of ​​the actual scene grayscale image, a 5*5 neighborhood sliding window is constructed with a step size of 1. The grayscale co-occurrence matrix corresponding to each ROI pixel point is obtained, and the contrast and entropy of the grayscale co-occurrence matrix are calculated as the neighborhood contrast con and neighborhood entropy ent of the ROI pixel point. The texture richness tr is constructed using the con and ent corresponding to the point. The texture richness is normalized to a value range of [0,1], denoted as TR. For ROI pixel points whose texture richness is less than or equal to the texture threshold, the grayscale variance of each neighborhood is calculated and recorded as Using texture richness TR and neighborhood grayscale variance Construct the smoke probability factor sp. Normalize the smoke probability factor sp to a value range of [0,1], denoted as SP.

[0065] At this point, all ROI pixels in the actual scene grayscale image have a smoke probability factor SP. According to the size of the SP value, all ROI pixels can be divided into two categories of pixels. The first category is high probability smoke points ( ,), the second category is non-high probability smoke points ( ).

[0066] The gradient values ​​of all ROI pixels are calculated for the ROI area, and Harris corner detection is performed at the same time. In the 5*5 neighborhood sliding window, the gradient value of the ROI pixel as the center is , No. The gradient value of the non-center ROI pixel is , No. The Euclidean distance from the non-central ROI pixel to the central ROI pixel is , the sum of all Euclidean distances is , and the number of corner points in the neighborhood sliding window is counted as r. Thus, based on the gradient value of each ROI pixel point in the neighborhood sliding window, the number of corner points, and the Euclidean distance from the non-central ROI pixel point to the central ROI pixel point, the clarity of the ROI pixel point as the center is calculated. .

[0067] Finally, the clarity of all ROI pixels in the ROI area is obtained. For any 5*5 neighborhood sliding window in the ROI area, the clarity of each ROI pixel and the clarity of different rows are obtained. There is a difference between them, and the farther the distance, the greater the difference. The mean clarity of each row is calculated and recorded as , , , , , the clarity means of different rows in the neighborhood are compared pairwise, there are 10 cases in total, and the distance between different rows is recorded as The smoke clarity diffusion b of the ROI pixel as the center is calculated based on the clarity mean of each row and the Euclidean distance between different rows. The smoke clarity diffusion b is normalized to a value range of [0,1], denoted as B.

[0068] for ROI pixel, the ROI pixel belongs to a non-high probability smoke point. Whether the ROI pixel is smoke depends only on the smoke probability factor SP of the ROI pixel. The larger the smoke probability factor, the greater the smoke confidence s of the ROI pixel. ROI pixel, this ROI pixel belongs to a high probability smoke point. Whether this ROI pixel is smoke depends not only on the smoke probability factor SP of the ROI pixel, but also on the unique feature of smoke, that is, the smoke clarity diffusion B. The larger the smoke clarity diffusion B of the ROI pixel, the larger the smoke confidence s of the ROI pixel. Based on this, the smoke confidence s of each ROI pixel is calculated and normalized to a value range of [0,1], denoted as .

[0069] Third, according to the smoke confidence of each ROI pixel in the grayscale image of the actual scene, the actual scene grayscale image is threshold segmented to detect whether a fire occurs on the conveyor belt.

[0070] Image threshold segmentation is used to input the smoke confidence of all ROI pixels in the actual scene image. , the smoke confidence of all ROI pixels in the actual scene image Perform binarization processing and output a binarized image. When the smoke confidence is greater than or equal to the confidence threshold, the binarization value of the ROI pixel is 0, that is, the ROI pixel is a smoke point; when the smoke confidence is less than the confidence threshold, the binarization value of the ROI pixel is 1, that is, the ROI pixel is a non-smoke point.

[0071] The number of 0 value points when the binary image appears Exceeding the smoke threshold 4 o'clock ( ), if there are signs of fire on the conveyor belt, the intelligent system can issue a fire warning signal or trigger the alarm device to remind the staff to take appropriate measures.

[0072] like Figure 3 As shown, when applying a belt conveyor fire warning method based on image processing according to an embodiment of the present invention, the image preprocessing can refer to the following process: 301. Capture an image of a normal conveyor belt that is not on fire to obtain a reference RGB image, and convert the reference RGB image into a reference grayscale image; 302. Perform Sobel operator edge detection on the reference grayscale image to obtain an edge contour in the reference grayscale image; 303. Perform Hough line transform on the edge contour to obtain the coordinates of each straight line segment and each pixel point on the straight line segment in the reference grayscale image; 304. Count the number of pixels on each straight line segment as the length of the straight line segment, and select two straight line segments with the largest lengths to determine as edge lines; 305. Obtain edge endpoints of two edge lines; 306. Connect edge endpoints located at the same image boundary and extend to both sides to obtain a boundary line; 307. Draw endpoint perpendicular lines passing through edge endpoints and perpendicular to the boundary line respectively; 308. The largest area enclosed by the boundary line and the vertical line of the endpoint is taken as the ROI area; 309. Using a CMOS camera to periodically capture an actual RGB image of the conveyor belt at the same shooting angle as the RGB image; 310. Use median filtering to preprocess the actual RGB image to obtain an actual scene image.

[0073] like Figure 4As shown, after the grayscale image of the reference RGB image is processed by Sobel operator edge detection and Hough line transform according to step 101, two edge lines of the belt (i.e., belt edge line segment a1 and belt edge line segment a2) are obtained, and then the coordinates of the endpoints of the belt edge line segment a1 and the belt edge line segment a2 are found by comparing the horizontal and vertical coordinates of all points in the edge lines, which are recorded as a1(x1,y1), a1(x2,y2), a2(x3,y3), and a2(x4,y4) respectively. The edge endpoints located at the same image boundary (i.e., a1(x1,y1) and a2(x3,y3), a1(x2,y2) and a2(x4,y4)) are connected and extended to both sides to obtain the boundary line (the horizontal dotted line and the image boundary in the figure), and then the endpoint perpendicular lines passing through the four endpoints and perpendicular to the boundary line are drawn respectively, and a total of four endpoint perpendicular lines are obtained. The area enclosed by the two outermost endpoint perpendicular lines (the vertical dotted lines in the figure) and the boundary line can be used as the ROI area at most.

[0074] like Figure 5 As shown, when applying a belt conveyor fire warning method based on image processing according to an embodiment of the present invention, the smoke confidence level can be calculated by referring to the following process: 501. Extracting the ROI region of the actual scene grayscale image; 502. Construct a 5*5 neighborhood sliding window for the ROI pixel points in the ROI area; 503. Obtain the gray level co-occurrence matrix of each ROI pixel using a neighborhood sliding window; 504. Calculate the texture richness of the ROI pixel: The texture richness is calculated according to the following formula ; In the formula, is the neighborhood contrast, obtained by calculating the contrast of the gray-level co-occurrence matrix; is the neighborhood entropy, which is obtained by calculating the entropy of the gray-level co-occurrence matrix. Normalized, the value range is [0,1], denoted as .

[0075] For ROI pixel points with texture richness greater than 0.3, execute step 505; for ROI pixel points with texture richness less than or equal to 0.3, execute step 506; 505. Calculate the smoke probability factor of the ROI pixel point according to the texture richness; 506. Calculate the neighborhood grayscale variance of the neighborhood sliding window corresponding to the ROI pixel point, and calculate the smoke probability factor of the ROI pixel point according to the texture richness and the neighborhood grayscale variance; 507. Calculate the clarity of each ROI pixel: The clarity is calculated according to the following formula ; In the formula, is the first The gradient value of the non-center ROI pixel point, is the gradient value of the ROI pixel at the center of the neighborhood sliding window, is the first The Euclidean distance from the non-central ROI pixel to the central ROI pixel, is the sum of all Euclidean distances in the neighborhood sliding window, is the number of corner points in the neighborhood sliding window.

[0076] 508. The clarity of each ROI pixel point in the neighborhood sliding window is constructed into a clarity matrix according to the ROI pixel point arrangement: Calculate the smoke clarity diffusion of the ROI pixel point as the center of the neighborhood sliding window ; In the formula, is the smoke clarity diffusion, and They are the first and The mean clarity of the rows ( and ), is the clarity matrix and The Euclidean distance between rows. The smoke clarity diffusion b is normalized to a value range of [0,1], denoted as B.

[0077] 509. Classify the ROI pixel points with smoke probability factors greater than or equal to 0.8 as high-probability smoke points; 510. Calculate the smoke confidence of each high-probability smoke point based on the smoke probability factor and the smoke clarity diffusion; 511. Classify the ROI pixel points with a smoke probability factor less than 0.8 as non-high probability smoke points; 512. Calculate the smoke confidence of each non-high probability smoke point based on the smoke probability factor.

[0078] In the embodiment of the present invention, the texture threshold is set to 0.3 and the smoke probability threshold is set to 0.8.

[0079] like Figure 6 As shown, when applying a belt conveyor fire warning method based on image processing according to an embodiment of the present invention, the following process can be referred to: 601. Perform threshold segmentation on each ROI pixel point of the actual scene image; The ROI pixels with smoke confidence less than 0.9 are classified as non-smoke points; The ROI pixels with smoke confidence greater than or equal to 0.9 are classified as smoke points; 602. Count the number of smoke points in the actual scene image; If the number of smoke points is greater than or equal to 16, it is determined that a fire has occurred on the conveyor belt.

[0080] If the number of smoke points is less than 16, the conveyor belt is judged to be normal.

[0081] In the embodiment of the present invention, the confidence threshold is set to 0.9 and the smoke threshold is set to 16.

[0082] In summary, the belt conveyor fire warning method based on image processing in the embodiment of the present invention is different from the traditional flame detection fire warning method. The present invention detects fire smoke that is generated earlier than the flame during the fire, so that signs of fire can be discovered earlier and measures can be taken earlier. The present invention improves the image threshold segmentation. Instead of using grayscale threshold for image segmentation, the ROI area is analyzed for smoke characteristics such as texture richness and smoke probability factor classification in combination with specific scenes. In combination with the characteristics that smoke caused by the conveyor belt fire will spread, the smoke clarity diffusion is constructed to segment the actual scene image, thereby improving the time and accuracy of fire warning. In addition, in combination with the actual conveyor belt fire scene, the ROI area is extracted based on the characteristics of the conveyor belt in the actual scene image, and only the ROI area is analyzed, which improves the efficiency of detection and warning.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A belt conveyor fire warning method based on image processing, characterized in that: include: Capturing an image of a normal conveyor belt that is not on fire to obtain a reference RGB image, and extracting a ROI region from the reference RGB image; Acquire an actual scene image of the conveyor belt at the same shooting angle position as the reference RGB image; Performing smoke analysis on the ROI region of the actual scene image to obtain smoke confidence of each ROI pixel in the ROI region; Fire warning recognition is performed on the actual scene image based on the smoke confidence of each ROI pixel point.

2. The belt conveyor fire warning method based on image processing according to claim 1 is characterized in that: Extracting the ROI region from the reference RGB image includes: Converting the reference RGB image into a reference grayscale image; Performing Sobel operator edge detection on the reference grayscale image to obtain edge contours in the reference grayscale image; Performing Hough line transform on the edge contour to obtain the coordinates of each straight line segment in the reference grayscale image and each pixel point on the straight line segment; Counting the number of pixels on each straight line segment as the length of the straight line segment, and selecting two straight line segments with the largest lengths to be determined as edge lines; Obtain edge endpoints of the two edge lines; Connect the edge endpoints located at the same image boundary and extend them to both sides to obtain boundary lines; draw endpoint vertical lines passing through the edge endpoints and perpendicular to the boundary lines respectively; The largest area enclosed by the boundary line and the vertical line of the endpoints is taken as the ROI area.

3. The belt conveyor fire warning method based on image processing according to claim 1 is characterized in that: The actual scene image of the conveyor belt is collected at the same shooting angle position as the reference RGB image, including: Using a CMOS camera to periodically capture an actual RGB image of the conveyor belt at the same shooting angle as the RGB image; Preprocessing the actual RGB image using median filtering to obtain an actual scene image; The actual scene image is converted into an actual scene grayscale image.

4. The belt conveyor fire warning method based on image processing according to claim 3 is characterized in that: Performing smoke analysis on the ROI region of the actual scene image to obtain smoke confidence of each ROI pixel in the ROI region includes: Extracting the ROI region of the actual scene grayscale image; Constructing a neighborhood sliding window for ROI pixel points in the ROI area; The smoke feature of the ROI area is extracted by using the neighborhood sliding window to obtain the smoke probability factor of the ROI pixel point; Calculate the clarity of each ROI pixel based on the neighborhood sliding window; Calculate the smoke clarity diffusion degree of each ROI pixel point based on the clarity of each ROI pixel point; The smoke confidence of each ROI pixel is calculated based on the smoke probability factor and the smoke clarity diffusion.

5. The belt conveyor fire warning method based on image processing according to claim 4 is characterized in that: The smoke features of the ROI area are extracted using the neighborhood sliding window to obtain the smoke probability factor of the ROI pixel point, including: Using the neighborhood sliding window to obtain the gray level co-occurrence matrix of each ROI pixel in the ROI area; Calculating the contrast and entropy of the gray level co-occurrence matrix as the neighborhood contrast and neighborhood entropy of the ROI pixel point; Calculating the texture richness of the ROI pixel point according to the neighborhood contrast and the neighborhood entropy; For the ROI pixel point whose texture richness is less than or equal to the texture threshold, calculating the neighborhood grayscale variance of the neighborhood sliding window corresponding to the ROI pixel point; Calculate the smoke probability factor of the ROI pixel based on the texture richness and the neighborhood grayscale variance; in: The texture richness is calculated according to the following formula: In the formula, is the texture richness; is the neighborhood contrast; is the neighborhood entropy; The smoke probability factor is calculated according to the following formula: In the formula, is the smoke probability factor; is the normalized texture richness, is the neighborhood grayscale variance; is the texture threshold.

6. The belt conveyor fire warning method based on image processing according to claim 4 is characterized in that: Calculating the clarity of each ROI pixel based on the neighborhood sliding window includes: Calculate the gradient value of each ROI pixel point; Performing Harris corner point detection on the ROI area, and counting the number of corner points in the neighborhood sliding window corresponding to each ROI pixel point; Calculate the Euclidean distance between each non-central ROI pixel point and the central ROI pixel point in the neighborhood sliding window corresponding to each ROI pixel point; Calculate the clarity of the ROI pixel point as the center based on the gradient value of each ROI pixel point in the neighborhood sliding window, the number of corner points, and the Euclidean distance from the non-central ROI pixel point to the central ROI pixel point; The clarity is calculated according to the following formula: In the formula, is the first The gradient value of the non-central ROI pixel point; is the gradient value of the ROI pixel at the center of the neighborhood sliding window; is the first The Euclidean distance from the non-central ROI pixel point to the central ROI pixel point; is the sum of all Euclidean distances within the neighborhood sliding window; is the number of corner points in the neighborhood sliding window.

7. The belt conveyor fire warning method based on image processing according to claim 4 is characterized in that: Calculating the smoke clarity diffusion degree of each ROI pixel point based on the clarity of each ROI pixel point includes: The clarity of each of the ROI pixel points in the neighborhood sliding window is constructed into a clarity matrix according to the arrangement of the ROI pixel points; Calculating the clarity mean of each row in the clarity matrix and the Euclidean distance between different rows; Calculate the smoke clarity diffusion of the ROI pixel point as the center based on the clarity mean of each row and the Euclidean distance between different rows; The smoke clarity diffusion is calculated according to the following formula: In the formula, It is the clear diffusion of smoke; and They are the first and The mean clarity of the rows, and ; is the clarity matrix and Euclidean distance between rows.

8. The belt conveyor fire warning method based on image processing according to claim 4 is characterized in that: Calculating the smoke confidence of each ROI pixel point based on the smoke probability factor and the smoke clarity diffusion degree includes: Classify the ROI pixel points whose smoke probability factor is greater than or equal to the smoke probability threshold as high-probability smoke points; classify the ROI pixel points whose smoke probability factor is less than the smoke probability threshold as non-high-probability smoke points; Calculating the smoke confidence of each of the high-probability smoke points based on the smoke probability factor and the smoke clarity diffusion, and calculating the smoke confidence of each of the non-high-probability smoke points based on the smoke probability factor; The smoke confidence is calculated according to the following formula: In the formula, is the smoke confidence; is the normalized smoke probability factor; is the normalized smoke clarity diffusion; is the smoke probability threshold.

9. The belt conveyor fire warning method based on image processing according to claim 1 is characterized in that: Performing fire warning recognition on the actual scene image based on the smoke confidence of each ROI pixel point includes: Performing threshold segmentation on each of the ROI pixel points of the actual scene image according to the smoke confidence, and classifying the ROI pixel points whose smoke confidence is greater than or equal to the confidence threshold as smoke points; The number of the smoke points in the actual scene image is counted, and when the number of the smoke points exceeds a smoke threshold, it is determined that a fire occurs on the conveyor belt.

10. The belt conveyor fire warning method based on image processing according to claim 9 is characterized in that: Threshold segmentation is performed according to the following formula: In the formula, is the coordinate point The corresponding binary value of the ROI pixel point, a binary value of 0 indicates that the ROI pixel point is the smoke point, and a binary value of 1 indicates that the ROI pixel point is a non-smoke point; is the coordinate point The normalized smoke confidence of the corresponding ROI pixel point; is the confidence threshold.

Images