CN-115619720-B - Three-dimensional feature mining method based on blast furnace burden surface monocular image and related equipment

CN115619720BCN 115619720 BCN115619720 BCN 115619720BCN-115619720-B

Abstract

The invention discloses a three-dimensional feature mining method and related equipment based on a blast furnace burden surface monocular image, wherein the method comprises the steps of receiving a two-dimensional video image sent by a high-temperature industrial endoscope after the high-temperature industrial endoscope obtains the two-dimensional video image covered by a full burden surface; the method comprises the steps of preprocessing a two-dimensional video image to obtain a gray level image, obtaining three-dimensional structured point cloud data based on the gray level image, constructing a three-dimensional model based on the three-dimensional structured point cloud data, extracting three-dimensional plane characteristics of the three-dimensional model, carrying out region segmentation on the three-dimensional plane characteristics to obtain different regions, and carrying out transverse comparison on parameters of the different regions to analyze the flow trend and depression degree of different regions of a material surface. According to the invention, the two-dimensional image containing abundant texture information is obtained, and the three-dimensional region characteristics of the blast furnace burden surface are extracted based on the two-dimensional image, so that the difference between different regions of the blast furnace burden surface is analyzed, and the accurate burden distribution is guided to improve the smelting efficiency.

Inventors

CAO TING
JIANG CHAOHUI
CHEN ZHIWEN
GUI WEIHUA
REN HAO
LUO WEICHAO
ZHANG CHAOBO

Assignees

鹏城实验室

Dates

Publication Date: 20260512
Application Date: 20220926

Claims (12)

1. The three-dimensional feature mining method based on the blast furnace burden surface monocular image is characterized by comprising the following steps of: After a high-temperature industrial endoscope acquires a two-dimensional video image covered by a full material surface, receiving the two-dimensional video image sent by the high-temperature industrial endoscope; Preprocessing the two-dimensional video image to obtain a gray level image, acquiring three-dimensional structured point cloud data based on the gray level image, and constructing a three-dimensional model based on the three-dimensional structured point cloud data; extracting three-dimensional plane characteristics of the three-dimensional model, and carrying out region segmentation on the three-dimensional plane characteristics to obtain different regions; Transversely comparing parameters of different areas to analyze the flowing trend and depression degree of different areas of the material surface The transverse comparison of the parameters of different areas is performed to analyze the flow trend and the depression degree of different areas of the material surface, and specifically comprises the following steps: comparing the characteristic parameters of different areas with the runoff profile characteristics of the material surface to obtain a comparison result; And evaluating the comparison result according to the evaluation index to obtain an evaluation result, and analyzing the flow trend and the depression degree of different areas of the material surface based on the evaluation result.
2. The blast furnace burden surface monocular image-based three-dimensional feature mining method according to claim 1, wherein the high-temperature industrial endoscope comprises an imaging assembly, a cooling protection assembly, a backlight assembly and a power supply assembly, wherein the imaging assembly is composed of an imaging tube, the cooling protection assembly is composed of a front sleeve shell, a fixing collar and a rear sleeve shell, the backlight assembly is composed of a backlight tube, and the power supply assembly is composed of a power supply, a power supply wiring port, an imaging driving circuit and a video signal wire interface.
3. The blast furnace burden surface monocular image-based three-dimensional feature mining method according to claim 2, wherein the receiving the two-dimensional video image transmitted by the high-temperature industrial endoscope after the high-temperature industrial endoscope acquires the two-dimensional video image covered by the full burden surface comprises: When the high-temperature industrial endoscope enters the blast furnace top, the cooling protection component cools; after the cooling protection assembly finishes cooling, the backlight assembly guides an external backlight light source into the furnace to provide a light source for the imaging assembly; The imaging tube of the imaging assembly leads the acquired blast furnace burden surface image to the high-temperature industrial endoscope, and controls the photosensitive chip and the imaging driving circuit of the imaging tube to carry out digital imaging to obtain a two-dimensional video image; And outputting the two-dimensional video image based on a video signal line interface of the power supply component.
4. The blast furnace burden surface monocular image-based three-dimensional feature mining method according to claim 1, wherein the preprocessing the two-dimensional video image to obtain a gray scale image, obtaining three-dimensional structured point cloud data based on the gray scale image, and constructing a three-dimensional model based on the three-dimensional structured point cloud data, specifically comprises: graying the two-dimensional video image to obtain a gray image, and performing Gaussian convolution kernel denoising operation on the gray image to obtain a target gray image; And obtaining a depth value of a pixel point in the target gray level image, taking the pixel point and the depth value as a plane coordinate and a height respectively, and constructing a three-dimensional model based on the plane coordinate and the height.
5. The blast furnace burden surface monocular image-based three-dimensional feature mining method according to claim 4, wherein the constructing a three-dimensional model based on the planar coordinates and the height further comprises: the horizontal rightward direction in the plane coordinates is taken as an x-axis positive direction, the vertical upward direction is taken as a y-axis positive direction, and the lower left corner of the target gray scale image is taken as an origin coordinate.
6. The method for mining three-dimensional characteristics based on monocular images of blast furnace burden surface according to claim 1, wherein the extracting three-dimensional planar characteristics of the three-dimensional model, and performing region segmentation on the three-dimensional planar characteristics to obtain different regions, comprises the following steps: Defining three-dimensional burden surface characteristics based on bell-less blast furnace burden distribution, wherein the blast furnace burden distribution comprises fixed-point burden distribution, fan-shaped burden distribution, single-ring burden distribution and multi-ring burden distribution; when the three-dimensional material surface features are defined, extracting three-dimensional plane features of the three-dimensional model, and generating different plane diagrams based on the three-dimensional plane features; And calibrating boundary points of the plane graph, determining different area ranges, and completing area segmentation based on the area ranges to obtain different areas.
7. The method for mining three-dimensional features based on monocular images of blast furnace burden surface according to claim 1, wherein the three-dimensional planar features include an elevation distribution, a gradient, a regional stress, and an orientation distribution, wherein the gradient, the regional stress, and the orientation distribution are calculated by a gradient formula, a regional stress formula, and an orientation distribution formula, respectively.
8. The blast furnace burden surface monocular image-based three-dimensional feature mining method of claim 7, wherein the gradient formula is: ; Wherein, the Is the downward gradient value of a certain direction of a point, For the gradient values obtained after the use of the convolution kernel for the first direction, Using a convolution kernel for the eighth direction to obtain a gradient value; The regional stress formula is: ; Wherein, the Is the local stress value of a point, For the maximum height value of the material, At the level of the minimum height value of the device, Is a set specific radius; The orientation distribution formula is as follows: ; Wherein, the Is the orientation value of one pixel point, Is a projection vector of one pixel point, Is the direction vector of one pixel point.
9. The blast furnace burden surface monocular image-based three-dimensional feature mining method according to claim 1, wherein the evaluation index includes an area, an average elevation, and a region parameter of each region, wherein the area, the average elevation, and the region parameter are calculated by an area formula, an average elevation formula, and a region parameter formula, respectively; The area formula is: ; Wherein, the For the area of each region, For the area of each pixel point, Is the direction vector of a pixel point; the average elevation formula is as follows: ; Wherein, the As the standard deviation of the height of the area, In order for the height to be an average height, Is a natural number of the Chinese characters, Is the first And a height, wherein, Is in the range of ; The regional parameter formula is as follows: ; Wherein, the Is the evaluation index of the most preferable cloth, Representing the value of the local gradient, Is the local stress value of a point.
10. A blast furnace burden surface monocular image-based three-dimensional feature mining system for implementing the blast furnace burden surface monocular image-based three-dimensional feature mining method according to any one of claims 1 to 9, the blast furnace burden surface monocular image-based three-dimensional feature mining system comprising: The image acquisition module is used for receiving the two-dimensional video image sent by the high-temperature industrial endoscope after the high-temperature industrial endoscope acquires the two-dimensional video image covered by the full material surface; the model construction module is used for preprocessing the two-dimensional video image to obtain a gray image, obtaining three-dimensional structured point cloud data based on the gray image, and constructing a three-dimensional model based on the three-dimensional structured point cloud data; The region segmentation module is used for extracting three-dimensional plane characteristics of the three-dimensional model and carrying out region segmentation on the three-dimensional plane characteristics to obtain different regions; and the result analysis module is used for transversely comparing the parameters of the different areas so as to analyze the flow trend and the depression degree of the different areas of the material surface.
11. A terminal comprising a memory, a processor, and a blast furnace burden level monocular image-based three-dimensional feature mining program stored on the memory and executable on the processor, the blast furnace burden level monocular image-based three-dimensional feature mining program when executed by the processor implementing the blast furnace burden level monocular image-based three-dimensional feature mining method steps of claim 1.
12. A computer-readable storage medium having stored thereon a computer program for execution by a processor to perform the steps of implementing the blast furnace burden level monocular image-based three-dimensional feature mining method of claim 1.

Description

Three-dimensional feature mining method based on blast furnace burden surface monocular image and related equipment Technical Field The invention relates to the field of control science and engineering, in particular to a three-dimensional feature mining method, a system, a terminal and a storage medium based on a blast furnace burden surface monocular image. Background The blast furnace is a large-sized countercurrent reactor with high temperature, high pressure, strong dust and sealing without light, a great amount of intense physical reaction and chemical reaction occur at any moment in the internal environment, the material surface is complicated and changed along with the processes of top material distribution, reaction sinking in the furnace and the like, the blast furnace is also key equipment for carrying out the steel manufacturing process, is a core unit for converting ferrite material flow, is the link with the largest energy consumption and the highest production cost in the steel manufacturing process, and accounts for about 60-70% of the total cost of the steel comprehensive energy consumption and the steel manufacturing. The blast furnace has the essence that the distribution and development of the gas flow are controlled by reasonably adjusting the distribution of the charge level, the occurrence of faults is reduced, the long-term and stable operation of the blast furnace is kept, the real-time and clear three-dimensional appearance of the charge level of the blast furnace is the most visual, reliable and visual expression form of the smelting state in the blast furnace, and meanwhile, the dust movement, the gas flow distribution and the temperature field change in the furnace can be reflected, so that the blast furnace has important research significance for improving the quality of molten iron and saving energy and reducing emission. And the further analysis of the three-dimensional shape of the blast furnace burden surface plays a further guiding role for on-site operators. The operators analyze the running state of the blast furnace by observing the shape and the typical characteristics of the surface of the blast furnace, and further control the distribution of the materials. Therefore, the three-dimensional shape and the characteristics of the blast furnace burden surface are important means for realizing accurate distribution and improving smelting efficiency. In the prior art, on the premise that real-time accurate three-dimensional burden surface distribution is difficult to obtain through detection technology, a plurality of students develop researches on the shape and characteristics of blast furnace burden surface, mainly concentrate on the definition and modeling aspects of radial stocklines of the blast furnace, but are weaker in the analysis aspect of areas covered by the whole burden surface, and due to the lack of advanced detection means, the method meets the industrial requirements to a certain extent, but the shape and the appearance of the blast furnace burden surface are too single and limited. In the fitting and reconstruction of the radial material line of the blast furnace, the shape of the material line is often fitted by adopting a curve or a straight line based on a real distribution rule and detection data, so that the shape characteristic information of the material surface cannot be accurately expressed, and the defects of the method are obvious especially when the material surface has a partial material line descending fault. Therefore, the characteristic information mining and visualization method of the blast furnace burden surface morphology still has the problem of single expression mode due to the restriction of detection technology, and the information mining and visualization method aiming at the blast furnace burden surface three-dimensional morphology characteristics is needed at present. Accordingly, the prior art is still in need of improvement and development. Disclosure of Invention The invention mainly aims to provide a three-dimensional feature mining method based on a blast furnace burden surface monocular image and related equipment, and aims to solve the problems that video images of rich texture information in a blast furnace cannot be obtained and precise distribution control cannot be performed so as to improve smelting efficiency in the prior art. In order to achieve the above object, the present invention provides a three-dimensional feature mining method based on a blast furnace burden surface monocular image, the three-dimensional feature mining method based on a blast furnace burden surface monocular image comprising the steps of: After a high-temperature industrial endoscope acquires a two-dimensional video image covered by a full material surface, receiving the two-dimensional video image sent by the high-temperature industrial endoscope; Preprocessing the two-dimensional video image to obtain a gray level image, acquiring three-di