Three-dimensional feature mining method based on single image of blast furnace charge surface and related equipment

Abstract

The application discloses a kind of based on blast furnace charge surface monocular image three-dimensional feature mining method and related equipment, the method includes: after obtaining the two-dimensional video image of full charge surface coverage in high-temperature industrial endoscope, the two-dimensional video image sent by the high-temperature industrial endoscope is received;The two-dimensional video image is preprocessed to obtain a gray image, three-dimensional structured point cloud data is obtained based on the gray image, and a three-dimensional model is constructed based on the three-dimensional structured point cloud data;Three-dimensional plane features of the three-dimensional model are extracted, and different regions are obtained by region segmentation on the three-dimensional plane features;The parameters of different regions are compared horizontally to analyze the flow trend and the depression degree of different regions of charge surface.The application obtains two-dimensional image containing rich texture information, extracts three-dimensional regional features of blast furnace charge surface based on two-dimensional image, to analyze the difference between different regions of blast furnace charge surface, and then guide accurate material distribution to improve smelting efficiency.

Description

Technical Field

[0001] This invention relates to the field of control science and engineering, and in particular to a method, system, terminal and storage medium for three-dimensional feature mining based on monocular images of blast furnace burden surface. Background Technology

[0002] A blast furnace is a large, countercurrent reactor characterized by high temperature, high pressure, intense dust, and a closed, dark environment. Numerous intense physical and chemical reactions occur constantly within its interior. The charge level undergoes complex changes during top charging and the settling of reactants within the furnace. It is a critical piece of equipment in the steelmaking process, the core unit for ferrous material flow conversion, and the most energy-intensive and costly stage in steelmaking, accounting for approximately 60%-70% of total energy consumption and manufacturing costs. The essence of a blast furnace lies in controlling the distribution and development of the gas flow by adjusting the charge level, thereby reducing malfunctions and maintaining long-term, stable operation. Real-time, clear three-dimensional morphology of the blast furnace charge level is the most intuitive, reliable, and visual representation of the smelting state within the furnace. It also reflects dust movement, gas flow distribution, and temperature field changes, making it crucial for improving molten iron quality and achieving energy conservation and emission reduction. Further analysis of the blast furnace charge level's three-dimensional morphology provides further guidance for on-site operators. By observing the morphology and typical characteristics of the blast furnace charge level, operators can analyze the furnace's operating status and subsequently control the charge distribution. Therefore, the three-dimensional morphology and characteristics of the blast furnace charge surface are important means to achieve precise charge distribution and improve smelting efficiency.

[0003] Meanwhile, the gas flow above the blast furnace burden surface also affects the position of the descent of the burden, causing the burden surface to deviate from its expected shape. Currently, given the difficulty in obtaining real-time and accurate three-dimensional burden surface distribution through detection technology, many scholars have conducted research on the morphological characteristics of the blast furnace burden surface, mainly focusing on the definition and modeling of the radial burden line. However, research on the analysis of the entire burden surface coverage area is relatively weak. Due to the lack of advanced detection methods, this approach, while meeting industrial needs to some extent, is too simplistic and limited in its representation of the blast furnace burden surface morphology. In the fitting and reconstruction of the radial burden line in the blast furnace, the shape of the burden line is often fitted using a "curve" or "straight line" based on the actual burden distribution pattern and detection data, which cannot accurately express the morphological characteristics of the burden surface. This method is particularly problematic when a localized burden line descent fault occurs. Therefore, the methods for mining and visualizing the feature information of the blast furnace burden surface morphology still suffer from a simplistic and one-sided representation due to limitations in detection technology. Currently, there is an urgent need for an information mining and visualization method targeting the three-dimensional morphological characteristics of the entire blast furnace burden surface.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a three-dimensional feature mining method and related equipment based on monocular images of blast furnace burden surfaces, aiming to solve the problems in the prior art of being unable to obtain video images with rich texture information inside the blast furnace and being unable to perform precise burden control to improve smelting efficiency.

[0006] To achieve the above objectives, the present invention provides a method for three-dimensional feature mining based on monocular images of blast furnace burden surfaces. The method includes the following steps:

[0007] After acquiring a two-dimensional video image covering the entire material surface using a high-temperature industrial endoscope, the system receives the two-dimensional video image sent by the high-temperature industrial endoscope.

[0008] The two-dimensional video image is preprocessed to obtain a grayscale image, three-dimensional structured point cloud data is obtained based on the grayscale image, and a three-dimensional model is constructed based on the three-dimensional structured point cloud data.

[0009] Extract the three-dimensional planar features of the three-dimensional model, and perform region segmentation on the three-dimensional planar features to obtain different regions;

[0010] By comparing parameters across different regions, we can analyze the flow trends and subsidence levels in different areas of the material surface.

[0011] Optionally, in the three-dimensional feature mining method based on monocular images of blast furnace burden, the high-temperature industrial endoscope includes an imaging component, a cooling protection component, a backlight component, and a power supply component. The imaging component is composed of an imaging tube; the cooling protection component consists of a front housing, a fixing ring, and a rear housing; the backlight component consists of a backlight tube; and the power supply component consists of a power supply, a power connector, an imaging drive circuit, and a video signal cable interface.

[0012] Optionally, the three-dimensional feature mining method based on monocular images of blast furnace burden surface, wherein the step of receiving the two-dimensional video image sent by the high-temperature industrial endoscope after acquiring a two-dimensional video image covering the entire burden surface using a high-temperature industrial endoscope specifically includes:

[0013] When the high-temperature industrial endoscope enters the top of the blast furnace, the cooling protection component provides cooling.

[0014] After the cooling protection component has completed cooling, the backlight component introduces the external backlight source into the furnace to provide a light source for the imaging component;

[0015] The imaging tube of the imaging component exports the acquired blast furnace material surface image to the high-temperature industrial endoscope, and controls the photosensitive chip and imaging drive circuit of the imaging tube to perform digital imaging to obtain a two-dimensional video image.

[0016] The two-dimensional video image is output based on the video signal line interface of the power supply component.

[0017] Optionally, the method for three-dimensional feature mining based on monocular images of blast furnace burden surface, wherein processing the two-dimensional video image to obtain a grayscale image, obtaining three-dimensional structured point cloud data based on the grayscale image, and constructing a three-dimensional model based on the three-dimensional structured point cloud data specifically includes:

[0018] The two-dimensional video image is converted to grayscale to obtain a grayscale image, and the grayscale image is then subjected to Gaussian convolution kernel denoising operation to obtain the target grayscale image;

[0019] Obtain the depth value of the pixel in the target grayscale image, use the pixel and the depth value as planar coordinates and height respectively, and construct a three-dimensional model based on the planar coordinates and the height.

[0020] Optionally, the three-dimensional feature mining method based on monocular images of blast furnace burden surface, wherein the construction of the three-dimensional model based on the planar coordinates and the height further includes, before:

[0021] The horizontal rightward direction in the planar coordinate system is taken as the positive x-axis, the vertical upward direction is taken as the positive y-axis, and the lower left corner of the target grayscale image is taken as the origin coordinate.

[0022] Optionally, the three-dimensional feature mining method based on monocular images of blast furnace burden surface, wherein extracting the three-dimensional planar features of the three-dimensional model and performing region segmentation on the three-dimensional planar features to obtain different regions specifically includes:

[0023] The blast furnace charging system based on bell-less charging defines the three-dimensional material surface features, wherein the blast furnace charging system includes fixed-point charging, fan-shaped charging, single-ring charging, and multi-ring charging.

[0024] Once the three-dimensional material surface features are defined, the three-dimensional planar features of the three-dimensional model are extracted, and different planar diagrams are generated based on the three-dimensional planar features.

[0025] The boundary points of the plan are marked to determine different regional ranges, and the regions are divided based on the regional ranges to obtain different regions.

[0026] Optionally, the three-dimensional feature mining method based on monocular images of blast furnace burden surface, wherein the lateral comparison of parameters in different regions to analyze the flow trend and depression degree in different regions of the burden surface specifically includes:

[0027] The comparison results are obtained by comparing the characteristic parameters of different regions and the runoff profile characteristics of the material surface;

[0028] The comparison results are evaluated according to the evaluation indicators to obtain the evaluation results, and the flow trend and subsidence degree of different areas of the material surface are analyzed based on the evaluation results.

[0029] Optionally, in the three-dimensional feature mining method based on monocular images of blast furnace burden surface, the three-dimensional planar features include elevation distribution, gradient, regional stress, and orientation distribution, wherein the gradient, regional stress, and orientation distribution are calculated using gradient formula, regional stress formula, and orientation distribution formula, respectively.

[0030] Optionally, in the three-dimensional feature mining method based on monocular images of blast furnace burden surface, the gradient formula is: G(x,y)=max(G1(x,y):G8(x,y));

[0031] Where G(x,y) is the downward gradient value of a point in a certain direction, G1(x,y) is the gradient value obtained after applying the convolution kernel in the first direction, and G8(x,y) is the gradient value obtained after applying the convolution kernel in the eighth direction.

[0032] The formula for the regional stress is: LR = (H max -H min ) / R;

[0033] Where LR is the local stress value at a point, and H max H is the maximum height value. min R is the minimum height value, and R is the set specific radius;

[0034] The distribution orientation formula is:

[0035] Where α is the orientation value of a pixel, N is the projection vector of a pixel, and n is the direction vector of a pixel.

[0036] Optionally, in the three-dimensional feature mining method based on monocular images of blast furnace burden surface, the evaluation indicators include the area, average elevation, and preferred parameters of each region.

[0037] The formula for the area of ​​each region is: A = n * cell.size 2 ;

[0038] Where A is the area of ​​each region, cell.size is the area of ​​each pixel, and n is the direction vector of a pixel;

[0039] The formula for the average elevation is:

[0040] Among them, S ele The height standard deviation for a specific region. H represents the average height, where K is a natural number. i Let be the height of the i-th height, where i ranges from (1, K);

[0041] The formula for the preferred parameters is:

[0042] Where λ is the evaluation index for the highest priority fabric placement, SL represents the local slope value, and LR is the calculated local stress value.

[0043] Furthermore, to achieve the above objectives, the present invention also provides a three-dimensional feature mining system based on a monocular image of blast furnace burden surface, wherein the three-dimensional feature mining system based on the monocular image of blast furnace burden surface includes:

[0044] The image acquisition module is used to receive the two-dimensional video image sent by the high-temperature industrial endoscope after the high-temperature industrial endoscope acquires a two-dimensional video image covering the entire material surface.

[0045] The model building module is used to preprocess the two-dimensional video image to obtain a grayscale image, obtain three-dimensional structured point cloud data based on the grayscale image, and build a three-dimensional model based on the three-dimensional structured point cloud data.

[0046] The region segmentation module is used to extract the three-dimensional planar features of the three-dimensional model and perform region segmentation on the three-dimensional planar features to obtain different regions;

[0047] The results analysis module is used to make horizontal comparisons of parameters in different regions in order to analyze the flow trend and the degree of subsidence in different areas of the material surface.

[0048] Furthermore, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a three-dimensional feature mining program based on a monocular image of blast furnace burden stored in the memory and executable on the processor. When the three-dimensional feature mining program based on the monocular image of blast furnace burden is executed by the processor, it implements the steps of the three-dimensional feature mining method based on the monocular image of blast furnace burden as described above.

[0049] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a three-dimensional feature mining program based on a monocular image of blast furnace burden surface, and when the three-dimensional feature mining program based on the monocular image of blast furnace burden surface is executed by a processor, it implements the steps of the three-dimensional feature mining method based on the monocular image of blast furnace burden surface as described above.

[0050] In this invention, after acquiring a two-dimensional video image covering the entire blast furnace burden surface using a high-temperature industrial endoscope, the system receives the two-dimensional video image transmitted by the endoscope. The two-dimensional video image is preprocessed to obtain a grayscale image. Based on the grayscale image, three-dimensional structured point cloud data is acquired, and a three-dimensional model is constructed based on the three-dimensional structured point cloud data. Three-dimensional planar features of the three-dimensional model are extracted, and these features are segmented into different regions. Parameters of different regions are compared laterally to analyze the flow trend and depression degree of different regions of the burden surface. This invention acquires two-dimensional images containing rich texture information, extracts three-dimensional regional features of the blast furnace burden surface based on the two-dimensional images, analyzes the differences between different regions of the blast furnace burden surface, and thus guides precise burden distribution to improve smelting efficiency. Attached Figure Description

[0051] Figure 1 This is a flowchart of a preferred embodiment of the three-dimensional feature mining method based on monocular images of blast furnace burden surface in this invention;

[0052] Figure 2 This is a schematic diagram of the installation of the blast furnace model and data acquisition device in this invention;

[0053] Figure 3 This is a schematic diagram illustrating the definition of the three-dimensional features of the material surface in this invention;

[0054] Figure 4 This is a schematic diagram of the planar features of the material surface in this invention;

[0055] Figure 5 This is a schematic diagram of the material surface area segmentation and flow trend in this invention;

[0056] Figure 6 This is a schematic diagram of the region profile comparison of the three-dimensional feature mining method based on monocular images of blast furnace burden surface in this invention;

[0057] Figure 7 This is a schematic diagram of the simulation results of the three-dimensional feature mining method based on monocular images of blast furnace burden surface in this invention;

[0058] Figure 8 This is a schematic diagram illustrating the principle of a preferred embodiment of the three-dimensional feature mining system based on a monocular image of blast furnace burden surface in this invention.

[0059] Figure 9 This is a schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0061] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0062] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0063] The preferred embodiment of the present invention describes a three-dimensional feature mining method based on a monocular image of blast furnace burden surface, such as... Figure 1 As shown, the three-dimensional feature mining method based on monocular images of blast furnace burden surface includes the following steps:

[0064] Step S10: After acquiring a two-dimensional video image covering the entire material surface using a high-temperature industrial endoscope, receive the two-dimensional video image sent by the high-temperature industrial endoscope.

[0065] Step S10 includes:

[0066] Step S11: When the high-temperature industrial endoscope enters the top of the blast furnace, the cooling protection component is cooled.

[0067] Step S12: After the cooling protection component has completed cooling, the backlight component introduces the external backlight source into the furnace to provide a light source for the imaging component.

[0068] Step S13: The imaging tube of the imaging component exports the acquired blast furnace material surface image to the high-temperature industrial endoscope, and controls the photosensitive chip and imaging drive circuit of the imaging tube to perform digital imaging to obtain a two-dimensional video image.

[0069] Step S14: Output the two-dimensional video image based on the video signal line interface of the power supply component.

[0070] Specifically, the high-temperature industrial endoscope mainly comprises four parts: an imaging component, a cooling and protection component, a backlight component, and a power supply component. The imaging component is mainly composed of an imaging tube; the cooling and protection component consists of a front housing, a fixing ring, and a rear housing; the backlight component's function is realized by the backlight tube; and the power supply component mainly provides power to the device and transmits video signals through a power supply, a power connector, an imaging drive circuit, and a video signal cable interface. First, the cooling protection component ensures the normal operation of the high-temperature industrial endoscope in the harsh high-temperature environment of the blast furnace top. Second, the backlight component inside the high-temperature industrial endoscope guides the light emitted from the external backlight source into the furnace to provide sufficient light for the front-end imaging component. Third, the imaging tube of the imaging component exports the acquired image of the blast furnace material surface to the low-temperature region at the rear end of the high-temperature industrial endoscope, where the image is then digitally imaged by the image sensor chip and the imaging drive circuit to obtain a two-dimensional video image. Finally, the video signal cable interface of the power supply component transmits the acquired two-dimensional video image of the blast furnace material surface, completing the entire video image acquisition process. A clear two-dimensional video image of the material surface can be captured, such as... Figure 2 As shown, Figure 2 The invention includes the shape of the blast furnace, the chute (used to slide the placed material down to the bottom of the blast furnace and rotate around the center line of the blast furnace at an angular velocity ω, the curve formed during the descent is the zero-point material line), the shape of the material accumulation, the shape of the material surface, and the central heat flow, so that operators can intuitively observe the furnace condition information and adjust the material distribution method in a timely manner. The invention is based on the two-dimensional material surface image obtained by the high-temperature industrial endoscope and performs three-dimensional feature extraction. Its data source covers a more comprehensive material surface area and overcomes the significant impact of high temperature, high pressure, and high dust inside the blast furnace, obtaining video images containing richer texture information.

[0071] Step S20: Preprocess the two-dimensional video image to obtain a grayscale image, acquire three-dimensional structured point cloud data based on the grayscale image, and construct a three-dimensional model based on the three-dimensional structured point cloud data.

[0072] Step S20 includes:

[0073] Step S21: Convert the two-dimensional video image to grayscale to obtain a grayscale image, and perform Gaussian convolution kernel denoising operation on the grayscale image to obtain the target grayscale image;

[0074] Step S22: Obtain the depth value of the pixel in the target grayscale image, use the pixel and the depth value as planar coordinates and height respectively, and construct a three-dimensional model based on the planar coordinates and the height.

[0075] Specifically, the method for constructing a three-dimensional model based on a two-dimensional image uses methods from the field of remote sensing geomorphological analysis to construct a three-dimensional model of the surface, namely a Digital Elevation Model (DEM). The DEM uses pixels as planar coordinates and the depth value of the corresponding pixel as the height value. First, the two-dimensional video image is converted to grayscale to obtain a grayscale image, removing the influence of color on the information of each pixel value. Second, a Gaussian convolution kernel is performed on the grayscale image to obtain a target grayscale image, in order to reduce the influence of noise on the target grayscale image. Then, the depth values ​​of the pixels in the target grayscale image are obtained, and the pixel value and the depth value are used as planar coordinates and height, respectively. Finally, the DEM model can be constructed using the target grayscale image, with the depth value of each pixel value as the z-axis height, the horizontal rightward direction as the positive x-axis, the vertical upward direction as the positive y-axis, and the lower left corner of the target grayscale image as the origin coordinate. Thus, the desired three-dimensional model, i.e., the digital elevation model, is obtained.

[0076] Step S30: Extract the three-dimensional planar features of the three-dimensional model, and perform region segmentation on the three-dimensional planar features to obtain different regions.

[0077] Step S30 includes:

[0078] Step S31: Define the three-dimensional material surface features based on the blast furnace charging without a charging bell, wherein the blast furnace charging includes fixed-point charging, fan-shaped charging, single-ring charging and multi-ring charging;

[0079] Step S32: When the three-dimensional material surface features are defined, extract the three-dimensional planar features of the three-dimensional model, and generate different planar maps based on the three-dimensional planar features;

[0080] Step S33: Mark the boundary points of the plan view, determine different regional ranges, and complete the regional segmentation based on the regional ranges to obtain different regions.

[0081] Specifically, the region-based three-dimensional material surface feature definition method is based on the bell-less blast furnace charging method, which is mainly divided into four types: fixed-point charging, fan-shaped charging, single-ring charging, and multi-ring charging. These four methods are as follows: Figure 3 As shown, Figure 3In this invention, (a) represents single-ring fabric, which involves fabricating in different directions within a single ring; (b) represents multi-ring fabric, which involves fabricating within a region divided into multiple different rings; (c) represents fan-shaped fabric, which involves fabricating within a divided fan-shaped region; and (d) represents fixed-point fabric, which involves fabricating within a specified target area. Inspired by fan-shaped and multi-ring fabrics, this invention further proposes regional features, i.e., regions that can be composed of multiple fan-shaped features. This definition of three-dimensional features can be associated with a specific common region, allowing for the calculation of a fabric matrix for precise fabric control in a preferred region. Methods for extracting three-dimensional planar features include, but are not limited to, elevation distribution, gradient, regional stress, and orientation distribution. These features can generate different planar maps, preparing for subsequent region segmentation and comparison; the gradient refers to the result of gradients in eight directions, the regional stress refers to the rate of change of elevation within a certain radius, and the orientation refers to the distribution of the tilt angle at each point. The gradient, regional stress, and orientation distribution are calculated using gradient formulas, regional stress formulas, and orientation distribution formulas, respectively. The gradient formula is: G(x,y)=max(G1(x,y):G8(x,y)); where G(x,y) is the downward gradient value of a point in a certain direction, G1(x,y) is the gradient value obtained after applying a convolution kernel to the first direction, and G8(x,y) is the gradient value obtained after applying a convolution kernel to the eighth direction; the regional stress formula is: LR=(H max -H min ) / R; where LR is the local stress value at a point, H max H is the maximum height value. min Where R is the minimum height value and R is the set specific radius; the distribution orientation formula is: Where α is the orientation value of a pixel, N is the projection vector of a pixel, and n is the direction vector of a pixel.

[0082] Furthermore, such as Figure 4 As shown, Figure 4In the diagram, 'a' represents the shaded area of ​​the material surface, 'b' represents the slope of the material surface, 'c' represents the stress of the material surface, and 'd' represents the orientation of the material surface. The slope and local stress can be extracted from these data. Planar features such as relief and aspect are used to analyze the overall trend of the material surface, the distribution of depressions, and the rate of change of slope. The planar distribution map can provide a preliminary analysis of the distribution trend of material surface flow, reflecting the distribution area of ​​the central hot airflow. A region segmentation method based on confluence depression features is employed, including but not limited to the calculation of various topographic factors based on different principles, such as flow direction (distance transformation algorithm, D8 scan algorithm, weighted maze algorithm, etc.), runoff volume (multi-flow direction algorithm, balanced binary search tree, etc.), and key steps such as catchment areas. Simultaneously, during the calculation of depression areas, the endpoints of the longest runoff profiles within different regions need to be interactively provided to constrain the boundary range of region segmentation. Based on these runoff profiles, boundary points are further calibrated to determine the different regional ranges, i.e., their boundaries. In other words, by extracting key topographic factors such as material surface flow direction, runoff volume, and catchment area, different catchment areas are ultimately determined, i.e., region segmentation. The final result is as follows: Figure 5 As shown, Figure 5 In the diagram, 'a' represents the runoff profile and region segmentation of the material surface, and 'b' represents the flow trend of the material surface.

[0083] Step S40: Compare the parameters of different regions horizontally to analyze the flow trend and subsidence degree of different regions of the material surface.

[0084] Step S40 includes:

[0085] Step S41: Compare the characteristic parameters of different regions and the runoff profile characteristics of the material surface to obtain comparison results;

[0086] Step S42: Evaluate the comparison results according to the evaluation indicators to obtain the evaluation results, and analyze the flow trend and subsidence degree of different areas of the material surface based on the evaluation results.

[0087] Specifically, by comparing the characteristic parameters and runoff profiles of different regions based on the similar flowability and regional depression characteristics of the analyzed catchment areas, a more accurate comparison of the characteristics of different regions can be achieved. For profile analysis within different regions, including but not limited to the elevation, slope, stress, and other evaluation indicators of the longest characteristic profile within the region, the basic characteristics of the profile can be obtained from planar features and directly compared. The evaluation indicators can be comprehensively considered based on established composite planar features. The overall evaluation indicators for the region include the area, average elevation, and optimal parameters of each region; wherein the area, average elevation, and optimal parameters are calculated using the area formula, average elevation formula, and optimal parameter formula, respectively; the area formula is: A = n * cell.size 2Where A is the area of ​​each region, cell.size is the area of ​​each pixel, and n is the direction vector of a pixel; the average elevation formula is: Among them, S ele The height standard deviation for a specific region. H represents the average height, where K is a natural number. i Let be the height of the i-th height, where i ranges from (1, K); the formula for the preferred parameters is: Where λ is the evaluation index for the highest priority fabric placement, SL represents the local slope value, and LR is the calculated local stress value.

[0088] Furthermore, such as Figure 6 As shown, Figure 6 In the diagram, 'a' is a region depth comparison map with distance (pixels) as the x-axis and depth as the y-axis; 'b' is a region slope comparison map with distance (pixels) as the x-axis and slope as the y-axis; 'c' is a region stress comparison map with distance (pixels) as the x-axis and stress as the y-axis; and 'd' is an area and index distribution map with distance (pixels) as the x-axis and area and index as the y-axis. By extracting basic indicators such as elevation, slope, stress, and area of ​​representative runoff profiles from all segmented regions, as well as overall regional composite indicators, the flow trends and depression degrees of different regions are analyzed.

[0089] Furthermore, in embodiments of the present invention, representative images were selected from a large number of video images acquired by high-temperature industrial endoscopes, such as... Figure 7 As shown, four sets of images are presented: the original image (set a), the planar feature orientation map (set b), the result image based on runoff extraction and region segmentation (set c), and the parameter comparison map based on the region (set d). From the four planar feature orientation maps in set b, it can be seen that the distribution and orientation of the central hot airflow are closely related. From the four runoff extraction and region segmentation result images in set c, it can be seen that runoff can largely reflect the flow trend of materials on the material surface, and the catchment area based on runoff features naturally connects closely connected pixels into a region. The results of the horizontal comparative analysis based on the runoff extraction and region segmentation results are shown in the four region-based parameter comparison images in set d. The region segmentation results form a ring around the central hot airflow, which can further extract various parameter features of different runoff and regions. Based on the comparative analysis of these parameter features, the areas that urgently need material placement can be identified, thereby guiding the on-site material placement operation.

[0090] Furthermore, such as Figure 8As shown, based on the above-mentioned three-dimensional feature mining method based on monocular images of blast furnace burden surface, the present invention also provides a three-dimensional feature mining system based on monocular images of blast furnace burden surface, wherein the three-dimensional feature mining system based on monocular images of blast furnace burden surface includes:

[0091] The image acquisition module 51 is used to receive the two-dimensional video image sent by the high-temperature industrial endoscope after the high-temperature industrial endoscope acquires a two-dimensional video image covering the entire material surface.

[0092] The model building module 52 is used to preprocess the two-dimensional video image to obtain a grayscale image, obtain three-dimensional structured point cloud data based on the grayscale image, and build a three-dimensional model based on the three-dimensional structured point cloud data.

[0093] The region segmentation module 53 is used to extract the three-dimensional planar features of the three-dimensional model and perform region segmentation on the three-dimensional planar features to obtain different regions;

[0094] The results analysis module 54 is used to make horizontal comparisons of parameters in different regions in order to analyze the flow trend and subsidence degree in different regions of the material surface.

[0095] Furthermore, this invention extracts circumferential region comparison results based on region segmentation by mining the three-dimensional features of video images from high-temperature industrial endoscopes installed inside blast furnaces. Combined with the principles of multi-ring and fan-shaped fabric application, it can better determine the specific area range of the required fabric application, thereby improving the operability of precise fabric application.

[0096] Furthermore, such as Figure 9 As shown, based on the above-mentioned three-dimensional feature mining method and system based on monocular images of blast furnace burden, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 9 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0097] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a 3D feature mining program 40 based on a monocular image of blast furnace burden surface. This 3D feature mining program 40 based on a monocular image of blast furnace burden surface can be executed by the processor 10, thereby implementing the 3D feature mining method based on a monocular image of blast furnace burden surface in this application.

[0098] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the three-dimensional feature mining method based on a monocular image of blast furnace burden surface.

[0099] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0100] In one embodiment, when the processor 10 executes the three-dimensional feature mining program 40 based on the monocular image of the blast furnace burden surface in the memory 20, the following steps are performed:

[0101] After acquiring a two-dimensional video image covering the entire material surface using a high-temperature industrial endoscope, the system receives the two-dimensional video image sent by the high-temperature industrial endoscope.

[0102] The two-dimensional video image is preprocessed to obtain a grayscale image, three-dimensional structured point cloud data is obtained based on the grayscale image, and a three-dimensional model is constructed based on the three-dimensional structured point cloud data.

[0103] Extract the three-dimensional planar features of the three-dimensional model, and perform region segmentation on the three-dimensional planar features to obtain different regions;

[0104] By comparing parameters across different regions, we can analyze the flow trends and subsidence levels in different areas of the material surface.

[0105] The high-temperature industrial endoscope includes an imaging component, a cooling protection component, a backlight component, and a power supply component. The imaging component is composed of an imaging tube; the cooling protection component is composed of a front housing, a fixing ring, and a rear housing; the backlight component is composed of a backlight tube; and the power supply component is composed of a power supply, a power connector, an imaging drive circuit, and a video signal cable interface.

[0106] The step of receiving the two-dimensional video image transmitted by the high-temperature industrial endoscope after acquiring a two-dimensional video image covering the entire material surface specifically includes:

[0107] When the high-temperature industrial endoscope enters the top of the blast furnace, the cooling protection component provides cooling.

[0108] After the cooling protection component has completed cooling, the backlight component introduces the external backlight source into the furnace to provide a light source for the imaging component;

[0109] The imaging tube of the imaging component exports the acquired blast furnace material surface image to the high-temperature industrial endoscope, and controls the photosensitive chip and imaging drive circuit of the imaging tube to perform digital imaging to obtain a two-dimensional video image.

[0110] The two-dimensional video image is output based on the video signal line interface of the power supply component.

[0111] Specifically, the process of processing the two-dimensional video image to obtain a grayscale image, obtaining three-dimensional structured point cloud data based on the grayscale image, and constructing a three-dimensional model based on the three-dimensional structured point cloud data includes:

[0112] The two-dimensional video image is converted to grayscale to obtain a grayscale image, and the grayscale image is then subjected to Gaussian convolution kernel denoising operation to obtain the target grayscale image;

[0113] Obtain the depth value of the pixel in the target grayscale image, use the pixel and the depth value as planar coordinates and height respectively, and construct a three-dimensional model based on the planar coordinates and the height.

[0114] The construction of the 3D model based on the planar coordinates and the height also includes, prior to:

[0115] The horizontal rightward direction in the planar coordinate system is taken as the positive x-axis, the vertical upward direction is taken as the positive y-axis, and the lower left corner of the target grayscale image is taken as the origin coordinate.

[0116] Specifically, the step of extracting the three-dimensional planar features of the three-dimensional model and performing region segmentation on the three-dimensional planar features to obtain different regions includes:

[0117] The blast furnace charging system based on bell-less charging defines the three-dimensional material surface features, wherein the blast furnace charging system includes fixed-point charging, fan-shaped charging, single-ring charging, and multi-ring charging.

[0118] Once the three-dimensional material surface features are defined, the three-dimensional planar features of the three-dimensional model are extracted, and different planar diagrams are generated based on the three-dimensional planar features.

[0119] The boundary points of the plan are marked to determine different regional ranges, and the regions are divided based on the regional ranges to obtain different regions.

[0120] The method of comparing parameters across different regions to analyze the flow trend and subsidence degree of different areas of the material surface specifically includes:

[0121] The comparison results are obtained by comparing the characteristic parameters of different regions and the runoff profile characteristics of the material surface;

[0122] The comparison results are evaluated according to the evaluation indicators to obtain the evaluation results, and the flow trend and subsidence degree of different areas of the material surface are analyzed based on the evaluation results.

[0123] The three-dimensional planar features include elevation distribution, gradient, regional stress, and orientation distribution. The gradient, regional stress, and orientation distribution are calculated using gradient formula, regional stress formula, and orientation distribution formula, respectively.

[0124] The gradient formula is: G(x,y)=max(G1(x,y):G8(x,y));

[0125] Where G(x,y) is the downward gradient value of a point in a certain direction, G1(x,y) is the gradient value obtained after applying the convolution kernel in the first direction, and G8(x,y) is the gradient value obtained after applying the convolution kernel in the eighth direction.

[0126] The formula for the regional stress is: LR = (H max -H min ) / R;

[0127] Where LR is the local stress value at a point, and H max H is the maximum height value. min R is the minimum height value, and R is the set specific radius;

[0128] The distribution orientation formula is:

[0129] Where α is the orientation value of a pixel, N is the projection vector of a pixel, and n is the direction vector of a pixel.

[0130] The evaluation indicators include the area, average elevation, and preferred parameters of each region;

[0131] The formula for the area of ​​each region is: A = n * cell.size 2 ;

[0132] Where A is the area of ​​each region, cell.size is the area of ​​each pixel, and n is the direction vector of a pixel;

[0133] The formula for the average elevation is:

[0134] Among them, S ele The height standard deviation for a specific region. H represents the average height, where K is a natural number. i Let be the height of the i-th height, where i ranges from (1, K);

[0135] The formula for the preferred parameters is:

[0136] Where λ is the evaluation index for the highest priority fabric placement, SL represents the local slope value, and LR is the calculated local stress value.

[0137] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a three-dimensional feature mining program based on a monocular image of blast furnace burden surface, and the three-dimensional feature mining program based on the monocular image of blast furnace burden surface implements the steps of the three-dimensional feature mining method based on the monocular image of blast furnace burden surface as described above when executed by a processor.

[0138] In summary, this invention provides a method and related equipment for three-dimensional feature mining based on monocular images of blast furnace burden surfaces. The method includes: after acquiring a two-dimensional video image covering the entire burden surface using a high-temperature industrial endoscope, receiving the two-dimensional video image sent by the high-temperature industrial endoscope; preprocessing the two-dimensional video image to obtain a grayscale image; acquiring three-dimensional structured point cloud data based on the grayscale image; constructing a three-dimensional model based on the three-dimensional structured point cloud data; extracting three-dimensional planar features from the three-dimensional model; segmenting the three-dimensional planar features to obtain different regions; and comparing the parameters of different regions laterally to analyze the flow trend and depression degree of different regions of the burden surface. This invention acquires two-dimensional images containing rich texture information, extracts three-dimensional regional features of the blast furnace burden surface based on the two-dimensional images, analyzes the differences between different regions of the blast furnace burden surface, and thus guides precise burden distribution to improve smelting efficiency.

[0139] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0140] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0141] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A three-dimensional feature mining method based on monocular images of blast furnace burden surface, characterized in that, The three-dimensional feature mining method based on monocular images of blast furnace burden surface includes: After acquiring a two-dimensional video image covering the entire material surface using a high-temperature industrial endoscope, the system receives the two-dimensional video image sent by the high-temperature industrial endoscope. The two-dimensional video image is preprocessed to obtain a grayscale image, three-dimensional structured point cloud data is obtained based on the grayscale image, and a three-dimensional model is constructed based on the three-dimensional structured point cloud data. Extract the three-dimensional planar features of the three-dimensional model, and perform region segmentation on the three-dimensional planar features to obtain different regions; Horizontal comparisons of parameters in different regions were conducted to analyze the flow trends and subsidence degrees in different areas of the material surface. The horizontal comparison of parameters in different regions to analyze the flow trend and subsidence degree in different areas of the material surface specifically includes: The comparison results are obtained by comparing the characteristic parameters of different regions and the runoff profile characteristics of the material surface; The comparison results are evaluated according to the evaluation indicators to obtain the evaluation results, and the flow trend and subsidence degree of different areas of the material surface are analyzed based on the evaluation results.

2. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 1, characterized in that, The high-temperature industrial endoscope includes an imaging component, a cooling protection component, a backlight component, and a power supply component. The imaging component is composed of an imaging tube; the cooling protection component is composed of a front housing, a fixing ring, and a rear housing; the backlight component is composed of a backlight tube; and the power supply component is composed of a power supply, a power connector, an imaging drive circuit, and a video signal cable interface.

3. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 2, characterized in that, After acquiring a two-dimensional video image covering the entire material surface using a high-temperature industrial endoscope, receiving the two-dimensional video image transmitted by the high-temperature industrial endoscope specifically includes: When the high-temperature industrial endoscope enters the top of the blast furnace, the cooling protection component provides cooling. After the cooling protection component has completed cooling, the backlight component introduces the external backlight source into the furnace to provide a light source for the imaging component; The imaging tube of the imaging component exports the acquired blast furnace material surface image to the high-temperature industrial endoscope, and controls the photosensitive chip and imaging drive circuit of the imaging tube to perform digital imaging to obtain a two-dimensional video image. The two-dimensional video image is output based on the video signal line interface of the power supply component.

4. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 1, characterized in that, The process of preprocessing the two-dimensional video image to obtain a grayscale image, obtaining three-dimensional structured point cloud data based on the grayscale image, and constructing a three-dimensional model based on the three-dimensional structured point cloud data specifically includes: The two-dimensional video image is converted to grayscale to obtain a grayscale image, and the grayscale image is then subjected to Gaussian convolution kernel denoising operation to obtain the target grayscale image; Obtain the depth value of the pixel in the target grayscale image, use the pixel and the depth value as planar coordinates and height respectively, and construct a three-dimensional model based on the planar coordinates and the height.

5. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 4, characterized in that, The process of constructing a 3D model based on the planar coordinates and the height also includes, prior to: The horizontal rightward direction in the planar coordinate system is taken as the positive x-axis, the vertical upward direction is taken as the positive y-axis, and the lower left corner of the target grayscale image is taken as the origin coordinate.

6. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 1, characterized in that, The step of extracting the three-dimensional planar features of the three-dimensional model and performing region segmentation on the three-dimensional planar features to obtain different regions specifically includes: The blast furnace charging system based on bell-less charging defines the three-dimensional material surface features, wherein the blast furnace charging system includes fixed-point charging, fan-shaped charging, single-ring charging, and multi-ring charging. Once the three-dimensional material surface features are defined, the three-dimensional planar features of the three-dimensional model are extracted, and different planar diagrams are generated based on the three-dimensional planar features. The boundary points of the plan are marked to determine different regional ranges, and the regions are divided based on the regional ranges to obtain different regions.

7. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 1, characterized in that, The three-dimensional planar features include elevation distribution, gradient, regional stress, and orientation distribution, wherein the gradient, regional stress, and orientation distribution are calculated using gradient formula, regional stress formula, and orientation distribution formula, respectively.

8. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 7, characterized in that, The gradient formula is: ; in, Let be the downward gradient value in a certain direction at a point. This represents the gradient value obtained after applying the convolution kernel in the first direction. The gradient value obtained after applying the convolution kernel to the eighth direction; The formula for the regional stress is: ; in, The local stress value at a point. This is the maximum height value. This is the minimum height value. For a specific radius; The orientation distribution formula is: ; in, The orientation value of a pixel. This is the projection vector of a single pixel. This is the direction vector of a single pixel.

9. The three-dimensional feature mining method based on monocular images of blast furnace burden surface according to claim 1, characterized in that, The evaluation indicators include the area, average elevation, and regional parameters of each region, wherein the area, average elevation, and regional parameters are calculated using the area formula, average elevation formula, and regional parameter formula, respectively. The area formula is: ; in, The area of ​​each region, The area of ​​each pixel, The direction vector of a single pixel; The formula for the average elevation is: ; in, The standard deviation of the height of this region. Average height For natural numbers, For the first One height, among which The range is ; The formula for the region parameter is: ; in, The evaluation criterion for the highest priority fabric is... Represents the local slope value. This represents the local stress value at a single point.

10. A three-dimensional feature mining system based on a monocular image of blast furnace burden surface, characterized in that, The three-dimensional feature mining system based on monocular images of blast furnace burden surface is used to implement the three-dimensional feature mining method based on monocular images of blast furnace burden surface as described in any one of claims 1-9. The three-dimensional feature mining system based on monocular images of blast furnace burden surface includes: The image acquisition module is used to receive the two-dimensional video image sent by the high-temperature industrial endoscope after the high-temperature industrial endoscope acquires a two-dimensional video image covering the entire material surface. The model building module is used to preprocess the two-dimensional video image to obtain a grayscale image, obtain three-dimensional structured point cloud data based on the grayscale image, and build a three-dimensional model based on the three-dimensional structured point cloud data. The region segmentation module is used to extract the three-dimensional planar features of the three-dimensional model and perform region segmentation on the three-dimensional planar features to obtain different regions; The results analysis module is used to make horizontal comparisons of parameters in different regions in order to analyze the flow trend and the degree of subsidence in different areas of the material surface.

11. A terminal, characterized in that, The terminal includes a memory, a processor, and a three-dimensional feature mining program based on a monocular image of blast furnace burden stored in the memory and executable on the processor. When the three-dimensional feature mining program based on the monocular image of blast furnace burden is executed by the processor, it implements the steps of the three-dimensional feature mining method based on a monocular image of blast furnace burden as described in claim 1.

12. A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of the three-dimensional feature mining method based on a monocular image of blast furnace burden as described in claim 1.

Images