Blast furnace burden surface segmentation method and device based on dynamic radius and multi-modal features

Abstract

This invention discloses a method and apparatus for blast furnace burden surface segmentation based on dynamic radius and multimodal features, comprising: initializing the center point coordinates and radius of the infrared image of the blast furnace burden surface; dividing the blast furnace burden surface into a core high-temperature zone, a relatively high-temperature zone, and a low-temperature zone based on dual temperature thresholds; establishing dual radius constraints and optimizing the radius under the condition of satisfying the constraints; fusing temperature features and texture gradient features, calculating the multimodal feature weights, and updating the centroid coordinates based on a weighted average; executing a closed-loop feedback process of radius-centroid alternating optimization, iterating until the center point and radius converge; segmenting the image into a high-temperature central region and a low-temperature peripheral region based on the converged optimal parameters; employing differentiated processing algorithms for different regions, and using Gaussian weighted fusion to obtain a globally consistent final result. This invention achieves robust and adaptive region segmentation of the blast furnace burden surface, providing a foundation for subsequent regional 3D reconstruction, and has certain applicability and reliability.

Description

Technical Field

[0001] This invention relates to the field of industrial image processing technology, and in particular to a method and apparatus for blast furnace charge surface segmentation based on dynamic radius and multimodal features. Background Technology

[0002] Blast furnace ironmaking is a core process in the modern steel industry, and its operational stability and efficiency directly affect the quality and cost of steel products. The distribution of the burden at the blast furnace throat is a decisive factor affecting the distribution of gas flow, heat transfer efficiency, and the uniformity of chemical reactions within the furnace. The surface of the burden accumulation, i.e., its three-dimensional shape, is an important characteristic reflecting the burden distribution. Therefore, real-time and accurate monitoring of the blast furnace burden surface shape is crucial for optimizing the burden distribution system and stabilizing blast furnace operation.

[0003] Among numerous monitoring technologies, infrared imaging technology has become a key means of obtaining information on the temperature distribution and morphology of blast furnace burden due to its advantages such as non-contact operation and strong resistance to dust interference. Infrared images of the burden surface acquired by infrared thermal imagers can not only reflect the temperature field but also be used for subsequent three-dimensional reconstruction to restore the morphology of the raw material surface.

[0004] However, infrared images of blast furnace burden surfaces exhibit a very unique regional characteristic difference. The central region of the image usually corresponds to the area with the highest temperature inside the furnace. Due to the extremely high temperature, this area appears as saturated or even overexposed in the infrared image, resulting in severe loss of texture information and insufficient detail. Conversely, the surrounding areas of the image (i.e., the low-temperature periphery) have lower temperatures and relatively darker brightness, but their texture clarity is higher, preserving rich details of the burden surface.

[0005] The significant differences in characteristics between the central and surrounding areas pose a major challenge to traditional image processing methods. If a uniform parameter setting and algorithm are used to process the entire image, the results cannot adequately represent the characteristics of different regions, leading to low reconstruction accuracy.

[0006] In the process of realizing this invention, the inventors discovered at least the following problems in the prior art:

[0007] Segmentation methods based on geometric assumptions are too rigid. Some existing material surface analysis methods tend to spatially divide the material surface into a central coke zone, a funnel zone, and a plateau zone, and then perform three-dimensional morphology fitting. These methods largely rely on fixed geometric model assumptions rather than adaptive segmentation based on real-time infrared image features. When changes in furnace conditions cause the heat distribution to deviate from strict geometric centrosymmetry, the accuracy of this method drops significantly.

[0008] Texture-based analysis methods fail in the central region. Other studies have focused on characterizing the roughness and grain size of the material surface using fractal analysis and other techniques. However, these methods heavily rely on the texture information of the image. Due to severe overexposure, a significant amount of texture information is lost in the central high-temperature region of infrared images of blast furnace burden. Therefore, any texture-dependent analysis method will fail in this critical area, failing to obtain accurate analytical results.

[0009] Traditional centroid positioning methods fail in overexposed areas. When segmenting areas, the center of the material surface needs to be located first. Traditional centroid methods typically rely solely on weighted averaging of grayscale information. However, in high-temperature overexposed areas with uniform grayscale at the center of the material surface, the grayscale information is almost identical. In such cases, traditional centroid methods are highly susceptible to local disturbances, leading to inaccurate or drifting center point positioning.

[0010] In summary, existing technologies lack a robust and adaptive region segmentation method capable of addressing the extreme differences in characteristics of infrared images of blast furnace burden surfaces. Therefore, a new technical solution is urgently needed that can intelligently segment the image into a high-temperature central region and a low-temperature peripheral region based on the image's actual thermodynamic and textural features, laying the foundation for subsequent targeted and differentiated image reconstruction and analysis. Summary of the Invention

[0011] In view of this, this application provides a blast furnace burden surface segmentation method and apparatus based on dynamic radius and multimodal features to solve the technical problem in related technologies that the large differences in characteristics between the center and the surrounding area of ​​the infrared image of the blast furnace burden surface make it difficult for a unified processing algorithm to take into account the characteristics of different areas and the low reconstruction accuracy.

[0012] To achieve the above objectives, the present invention adopts the following technical solution:

[0013] According to a first aspect of the present invention, a method for blast furnace burden segmentation based on dynamic radius and multimodal characteristics is provided, comprising:

[0014] Initialize the center point coordinates of the infrared image of the blast furnace burden surface, and set a circular region with a variable radius centered on the center point coordinates as a candidate radius set;

[0015] Statistical analysis of the temperature of infrared images of blast furnace burden surface was performed, and a dual temperature threshold mechanism was introduced to divide the image into three types of regions based on the temperature range.

[0016] Based on the three types of regions, a radius double constraint condition is established. Through the radius double constraint condition, the radius parameter that satisfies the temperature distribution law is initially determined in the candidate radius set, so as to realize the adaptive adjustment of the central region range.

[0017] Within the central region determined by the radius parameter, the multimodal centroid coordinates are calculated by combining temperature features and texture gradient features;

[0018] The center point coordinates are updated using the multimodal centroid coordinates. The radius parameter and center position are updated alternately through iterative loops until convergence, thus obtaining the optimal center point coordinates and optimal radius. In this way, the infrared image of the blast furnace material surface is segmented into a high-temperature central region and a low-temperature peripheral region.

[0019] Different parameters are set for the high-temperature central region and the low-temperature peripheral region for reconstruction, and the reconstruction results of the two regions are Gaussian weighted fusion to generate a globally consistent final result.

[0020] According to a second aspect of the present invention, a blast furnace burden surface segmentation device based on dynamic radius and multimodal characteristics is provided, comprising:

[0021] An initialization module is used to initialize the center point coordinates of the infrared image of the blast furnace burden surface, and to set a circular region with a variable radius centered on the center point coordinates as a candidate radius set.

[0022] The region division module is used for statistical analysis of the temperature of infrared images of blast furnace burden surface. It introduces a dual temperature threshold mechanism to divide the image into three types of regions based on the temperature range.

[0023] The radius constraint optimization module is used to establish radius double constraint conditions based on the three types of regions. Through the radius double constraint conditions, the radius parameters that meet the temperature distribution law are initially determined in the candidate radius set, so as to realize the adaptive adjustment of the central region range.

[0024] The centroid calculation module calculates multimodal centroid coordinates within the central region determined by the radius parameter, combining temperature features and texture gradient features.

[0025] The iterative optimization module updates the center point coordinates using the multimodal centroid coordinates, and alternately updates the radius parameter and center position through iterative loops until convergence, thereby obtaining the optimal center point coordinates and optimal radius, and then segments the infrared image of the blast furnace material surface into a high-temperature central region and a low-temperature peripheral region.

[0026] The partitioning module sets different parameters for the high-temperature central region and the low-temperature peripheral region to reconstruct them, and performs Gaussian weighted fusion on the reconstruction results of the two regions to generate a globally consistent final result.

[0027] According to a third aspect of the present invention, an electronic device is provided, comprising:

[0028] One or more processors;

[0029] Memory, used to store one or more programs;

[0030] When the one or more programs are executed by the one or more processors, the one or more processors perform the method as described in the first aspect.

[0031] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having computer instructions stored thereon that, when executed by a processor, implement the steps of the method as described in the first aspect.

[0032] The technical solutions provided by the embodiments of this application may include the following beneficial effects:

[0033] By employing dynamic radius optimization based on dual temperature thresholds and dual constraints, this invention adaptively determines the range of the high-temperature center zone according to real-time furnace conditions (temperature field distribution), overcoming the rigidity of existing technologies that rely on fixed geometric assumptions. Addressing the challenges of overexposure and texture loss in the center zone, this invention utilizes a multimodal centroid calculation method that fuses temperature and texture gradients. This effectively leverages residual edge response information, significantly improving the accuracy and robustness of centroid localization. It overcomes the problem of traditional grayscale centroid methods being susceptible to overexposure disturbances, ultimately achieving effective partitioning of the blast furnace burden infrared image. This allows for subsequent application of optimal differentiated processing algorithms to the unique characteristics of the center and surrounding areas, significantly enhancing the overall effect and accuracy of blast furnace burden 3D reconstruction. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating a blast furnace burden surface segmentation method based on dynamic radius and multimodal characteristics, according to an exemplary embodiment.

[0035] Figure 2 This is a clearly imaged infrared schematic diagram of the blast furnace charge surface according to an exemplary embodiment.

[0036] Figure 3 This is a schematic diagram illustrating the blast furnace burden surface segmentation result according to an example embodiment.

[0037] Figure 4 This is a block diagram illustrating a blast furnace burden surface segmentation device based on dynamic radius and multimodal characteristics, according to an exemplary embodiment.

[0038] Figure 5 This is a schematic diagram of the structure of an electronic device according to an exemplary embodiment. Detailed Implementation

[0039] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0040] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0041] Figure 1 This is a flowchart illustrating a blast furnace burden surface segmentation method based on dynamic radius and multimodal characteristics, according to an exemplary embodiment. Figure 1 As shown, this method, when applied to a terminal, may include the following steps:

[0042] In step S1, the center point coordinates of the infrared image of the blast furnace charge surface are initialized, and a circular region with a variable radius is set as a candidate radius with the center point coordinates as the center.

[0043] Specifically, based on prior knowledge of infrared images of the blast furnace burden surface, namely that the central region is usually located in the center of the image, the coordinates of the center point of the burden surface are initialized. :

[0044]

[0045]

[0046] in, and These represent the width and height of the image, respectively. A variable radius is defined with this point as the center. and its range of variation, that is, setting a set of candidate radii. .

[0047] In step S2, the temperature of the infrared image of the blast furnace burden surface is statistically analyzed. A dual temperature threshold mechanism is introduced to divide the image into three regions based on the temperature range:

[0048] S21: Perform statistical analysis on the temperature of the infrared image of the blast furnace burden surface, and set a first temperature threshold and a second temperature threshold, wherein the first temperature threshold is greater than the second temperature threshold.

[0049] Specifically, Figure 2 This is a clear infrared image of the blast furnace burden surface. It can be observed that the brightness of the infrared image of the blast furnace burden surface exhibits a radial decrease from the center outwards. The central region is brighter than the surrounding regions, representing a higher temperature; the surrounding regions are darker, representing a lower temperature. Simultaneously, there is a transition area between the central and surrounding regions. Therefore, the blast furnace burden surface image is divided into a core high-temperature zone, a relatively high-temperature zone, and a low-temperature zone. Through statistical analysis of the brightness of the blast furnace burden surface infrared image, the first temperature threshold... It can be the top 10% quantile of the temperature statistical distribution, the second temperature threshold. It can be the top 30 percentile of the temperature statistical distribution.

[0050] S22: Based on the first temperature threshold and the second temperature threshold, the infrared image of the blast furnace burden surface is divided into regions. The core high-temperature zone is defined as the region with a temperature greater than the first temperature threshold, representing the part of the burden surface where combustion is most intense. The relatively high-temperature zone is defined as the region with a temperature between the second temperature threshold and the first temperature threshold, reflecting the heat diffusion process. The remaining part is defined as the low-temperature zone.

[0051] Specifically, temperatures greater than The area is defined as the core high-temperature zone, representing the part of the material surface where combustion is most intense, and the temperature is between [temperature range missing]. and The area between the central high-temperature region and the peripheral low-temperature region is defined as the relatively high-temperature region, which is usually located between the central high-temperature region and the peripheral low-temperature region. It also contains some characteristic information of the central region to some extent, reflecting the heat diffusion process. Other areas are defined as low-temperature regions.

[0052] In step S3, based on the three types of regions, a radius double constraint condition is established. Through the radius double constraint condition, the radius parameter that satisfies the temperature distribution law is initially determined from the candidate radius set. This enables adaptive adjustment of the central area; this step includes the following sub-steps:

[0053] S31: Based on the temperature distribution characteristics of infrared images of blast furnace burden, design a double constraint condition for the radius;

[0054] Specifically, the radius double constraint condition is as follows:

[0055] (1) First constraint condition: at a radius of R Within the central region, the proportion of pixels in the Relatively High Temperature Zone (RHTZ) to the total number of pixels in the entire image should exceed a threshold. :

[0056]

[0057] in, For the current radius R The number of pixels in the relatively high-temperature zone inside. This represents the total number of pixels in the relatively high-temperature areas of the entire image. These are pre-defined hyperparameters. This constraint, by requiring the segmented region to cover a certain proportion of the relatively high-temperature area of ​​the entire image, solves the problem of traditional methods ignoring temperature change characteristics. It can effectively filter out the interference of low-temperature areas at the edges on the characteristics of the central region, while avoiding the central region being limited to the absolutely high-temperature area due to an excessively small radius.

[0058] (2) Second constraint condition: at a radius of R Within the central region, the proportion of pixels in the core high-temperature zone (CHTZ) to the total number of pixels in the central region should exceed a given threshold. :

[0059]

[0060] in, For the current radius R The number of pixels in the inner core high-temperature zone (CHTZ). This represents the total number of pixels in the central region. These are pre-defined hyperparameters. This constraint ensures that the central region contains sufficient absolute high-temperature information, preventing the importance of the core high-temperature region from being overlooked;

[0061] S32: Within the range of the candidate radius set, dynamically adjust the radius. R To ensure that it satisfies the double radius constraint condition, the radius parameter that satisfies the temperature distribution law is initially determined. .

[0062] In step S4, within the central region determined by the radius parameter, the multimodal centroid coordinates are calculated by combining temperature features and texture gradient features; this step includes the following sub-steps:

[0063] S41: In the case of the radius parameter Within a defined central region, the edge response is extracted using the Sobel operator. As a texture feature, it is linearly fused with temperature features (pixel grayscale values) to calculate the feature weight of each pixel. And normalize it:

[0064]

[0065]

[0066] in The set of pixels in the current central region. , This is an adjustable fusion coefficient.

[0067] S42: Based on the normalized feature weights, the position coordinates of each pixel in the central region are weighted and averaged to calculate the new multimodal centroid coordinates. :

[0068]

[0069]

[0070] This multimodal fusion method enables the calculation of a more robust centroid, even within the high-temperature core region where temperatures are saturated, by utilizing the weak gradient information at the boundary between its edges and the relatively high-temperature region.

[0071] In step S5, the center point coordinates are updated using the multimodal centroid coordinates. The radius parameter and center position are updated alternately through iterative loops until convergence, thus obtaining the optimal center point coordinates and optimal radius. In this way, the infrared image of the blast furnace material surface is segmented into a high-temperature central region and a low-temperature peripheral region.

[0072] Specifically, the new centroid coordinates calculated in step S42 are used. Update the center point coordinates. Then determine if the center point coordinates and radius have converged, i.e., the change is less than a certain small threshold. If not, return to step S3 and continue the next round of radius optimization and centroid update using the new center point. If converged, exit the iteration loop, and obtain the center point coordinates. and That is, the optimal center point and the optimal radius.

[0073] In step S6, different parameters are set for the high-temperature central region and the low-temperature peripheral region for reconstruction, and the reconstruction results of the two regions are Gaussian weighted fusion to generate a globally consistent final result.

[0074] S61: Analyze the characteristic differences between the high temperature center region and the low temperature periphery region, set differentiated parameters for different regions, and construct a larger support window for the high temperature center region compared with the low temperature periphery region when aggregating costs, that is, set a larger arm length limit when constructing the cross arm, and increase the gradient constraint strength.

[0075] Specifically, based on the optimal center point and optimal radius obtained after convergence in step S5, the infrared image of the blast furnace burden surface is formally segmented into a high-temperature central region and a low-temperature peripheral region. Different parameters and processing algorithms are set for the distinct characteristics of these two regions. The high-temperature central region is characterized by high brightness and weak texture, requiring suppression of its overexposure. The low-temperature peripheral region is characterized by low brightness and rich texture, requiring preservation of its rich details. The window construction rules supported during cost aggregation are as follows:

[0076] (1)

[0077] (2)

[0078] (3) if

[0079] (4)

[0080] (5) if

[0081] in This represents the color difference (grayscale difference) between pixel p and pixel q. It is a pre-set threshold. , Represents the distance between pixels p and q, through Control the arm length. When the arm length exceeds... When setting a smaller threshold This ensures that the support arm only propagates between pixels that are very similar in color. This represents the gradient difference between pixel p and pixel q. , This represents the gradient along the x-axis. This represents the gradient along the y-axis. For the high-temperature central region, the support window should be enlarged compared to the low-temperature peripheral region, i.e., the area of ​​the support window should be increased. At the same time, increase the gradient constraint strength, that is, reduce and .

[0082] This improves the ability of the central area to capture subtle texture and brightness variations, and suppresses the negative impact of overexposure on the reconstruction results.

[0083] S62: Perform stereo matching on different regions using pre-set differential parameters to obtain two disparity maps. Then, perform Gaussian weighted fusion on the disparity maps to obtain the final result.

[0084] Specifically, after differentiating the two regions separately, a Gaussian weighted fusion strategy is adopted to ensure the global consistency of the final output. By introducing a smooth Gaussian weight distribution, a seamless transition is achieved at the boundary between the high-temperature central region and the low-temperature peripheral region. This fusion method not only preserves the characteristic expression of each region but also eliminates the potential boundary artifacts caused by hard segmentation, improving the visual consistency and reconstruction accuracy of the overall result.

[0085] refer to Figure 3The image shows the regional segmentation results of the infrared image of the blast furnace burden surface. The part circled in the middle is the high-temperature central area, and the remaining part is the low-temperature peripheral area.

[0086] The accuracy of the reconstruction results was measured by comparing the results of 3D reconstruction using differentiated parameters after material surface region segmentation with the results of 3D reconstruction without region segmentation. The specific calculation formulas are as follows:

[0087]

[0088]

[0089]

[0090]

[0091] in, These are actual measured values. It is the depth of the reconstruction result. The number of data points.

[0092] Table 1 shows the comparison results of the evaluation indicators:

[0093]

[0094] As can be seen from Table 1, the differential reconstruction after segmentation of the blast furnace surface region proposed in this invention has better accuracy than the three-dimensional reconstruction of the entire blast furnace surface image directly. This indicates that the blast furnace surface segmentation method based on dynamic radius and multimodal features proposed in this invention can reasonably divide the blast furnace surface into regions and improve the accuracy of three-dimensional reconstruction, proving the effectiveness of the method proposed in this invention.

[0095] As demonstrated by the above embodiments, the three-dimensional morphology and temperature distribution information of the blast furnace burden surface are crucial for understanding the burden reaction status, optimizing the charging system, and ensuring stable blast furnace operation. However, infrared images of the blast furnace burden surface exhibit extreme differences in characteristics, with the central region showing high temperature overexposure and texture loss, while the surrounding region shows low temperature and rich texture. This invention provides a blast furnace burden surface segmentation method and apparatus based on dynamic radius and multimodal features. By constructing a closed-loop feedback system that alternately optimizes the radius and centroid, the aforementioned technical challenges are effectively solved. On one hand, this method, based on dual temperature thresholds and applying dual constraints, achieves dynamic adaptive determination of the central region radius, overcoming the rigidity of traditional methods that rely on fixed geometric assumptions. On the other hand, this method innovatively integrates temperature features and texture gradient features to calculate the multimodal centroid, effectively solving the problem of inaccurate positioning or drift caused by the traditional grayscale centroid method due to high temperature overexposure and texture loss in the central region. Based on this, this invention applies differentiated processing algorithms to address the characteristic differences between the segmented high-temperature central region and the low-temperature surrounding region, and uses Gaussian weighted fusion to eliminate boundary artifacts. The technical solution provided by this invention can intelligently adapt to the characteristic differences of infrared images of blast furnace burden, significantly improving the accuracy and robustness of subsequent three-dimensional reconstruction of blast furnace burden, and providing a new solution for the accurate monitoring and analysis of blast furnace burden status.

[0096] Corresponding to the aforementioned embodiments of the blast furnace burden surface segmentation method based on dynamic radius and multimodal characteristics, this application also provides embodiments of the blast furnace burden surface segmentation device based on dynamic radius and multimodal characteristics.

[0097] Figure 4 This is a block diagram of an apparatus for a blast furnace burden surface segmentation method based on dynamic radius and multimodal characteristics, according to an exemplary embodiment. (Refer to...) Figure 4 The device includes:

[0098] Initialization module 1 is used to initialize the center point coordinates of the infrared image of the blast furnace burden surface, and set a circular region with a variable radius centered on the center point coordinates as a candidate radius set;

[0099] The region division module 2 is used to perform statistical analysis on the temperature of infrared images of blast furnace burden surface. It introduces a dual temperature threshold mechanism to divide the image into three types of regions based on the temperature range: core high temperature region, relatively high temperature region and low temperature region.

[0100] The radius constraint optimization module 3 is used to establish a radius double constraint condition based on the three types of regions. Through the radius double constraint condition, the radius parameter that satisfies the temperature distribution law is initially determined in the candidate radius set, so as to realize the adaptive adjustment of the central region range.

[0101] Centroid calculation module 4 calculates multimodal centroid coordinates within the central region determined by the radius parameter, combining temperature features and texture gradient features;

[0102] Iterative optimization module 5 updates the center point coordinates using the multimodal centroid coordinates, and alternately updates the radius parameter and center position through iterative loop until convergence, thereby obtaining the optimal center point coordinates and optimal radius, and then dividing the infrared image of the blast furnace material surface into a high-temperature center region and a low-temperature peripheral region.

[0103] The partitioning module 6 sets different parameters for the high-temperature central region and the low-temperature peripheral region to reconstruct them, and performs Gaussian weighted fusion on the reconstruction results of the two regions to generate a globally consistent final result.

[0104] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0105] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0106] Accordingly, this application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the above-described method for three-dimensional reconstruction of blast furnace burden surface based on binocular infrared images. Figure 5 The diagram shown is a hardware structure diagram of any device with data processing capabilities, including a 3D reconstruction device for blast furnace burden surface based on binocular infrared images provided in an embodiment of the present invention. (Except for...) Figure 5 In addition to the processor and memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0107] Accordingly, this application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the aforementioned method for three-dimensional reconstruction of blast furnace burden surface based on binocular infrared images. The computer-readable storage medium can be an internal storage unit of any data processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0108] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0109] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A blast furnace charge level segmentation method based on dynamic radius and multi-modal features, characterized in that, include: Initialize the center point coordinates of the infrared image of the blast furnace burden surface, and set a circular region with a variable radius centered on the center point coordinates as a candidate radius set; Statistical analysis of the temperature of infrared images of blast furnace burden surface was performed, and a dual temperature threshold mechanism was introduced to divide the image into three types of regions based on the temperature range. Based on the three types of regions, a radius double constraint condition is established. Through the radius double constraint condition, the optimal radius parameter that satisfies the temperature distribution law is initially determined in the candidate radius set, so as to realize the adaptive adjustment of the central region range. Within the central region determined by the optimal radius parameter, the multimodal centroid coordinates are calculated by combining temperature features and texture gradient features; The center point coordinates are updated using the multimodal centroid coordinates. The radius parameter and center position are updated alternately through iterative loops until convergence, thus obtaining the optimal center point coordinates and optimal radius. In this way, the infrared image of the blast furnace material surface is segmented into a high-temperature central region and a low-temperature peripheral region. Different parameters are set for the high-temperature central region and the low-temperature peripheral region for reconstruction, and the reconstruction results of the two regions are Gaussian weighted fusion to generate a globally consistent final result; Specifically, based on the temperature distribution characteristics, a double constraint condition for the radius is established. Using this double constraint condition, radius parameters that satisfy the temperature distribution law are initially determined from the candidate radius set, including: Based on the temperature distribution characteristics of the infrared image of the blast furnace burden surface, a double constraint condition for the radius is designed, wherein: The first constraint is: within a radius of... R Within the central region, the proportion of pixels in the relatively high-temperature area to the total number of pixels in the relatively high-temperature area of ​​the entire image should exceed the first threshold. The second constraint is: within a radius of... R Within the central region, the proportion of pixels in the core high-temperature zone to the total number of pixels in the central region should exceed a given second threshold. Within the set of candidate radii, the radius is dynamically adjusted to satisfy the double radius constraint condition, and the radius parameters that satisfy the temperature distribution law are initially determined. Specifically, different parameters are set for the high-temperature central region and the low-temperature peripheral region for reconstruction, and the reconstruction results of the two regions are Gaussian-weightedly fused to generate a globally consistent final result, including: The characteristics of the high-temperature center region and the low-temperature periphery region are analyzed, and differentiated parameters are set for different regions. For the high-temperature center region, compared with the low-temperature periphery region, a larger support window is constructed when the cost is aggregated. That is, a larger arm length limit is set when constructing the cross arm, and the gradient constraint strength is increased. Stereo matching is performed on different regions using pre-set differential parameters to obtain two disparity maps. The disparity maps are then fused using Gaussian weighting to obtain the final result. The construction rules for the support window are as follows: （1） ； （2） ； (3) ,if ; （4） ； (5) ,if ; in This represents the color difference between pixel p and pixel q. It is a pre-set threshold. Represents the distance between pixels p and q, through Control the arm length; For the threshold; This represents the gradient difference between pixel p and pixel q.

2. The method according to claim 1, characterized in that, Statistical analysis was performed on the temperature of infrared images of blast furnace burden surfaces. A dual temperature threshold mechanism was introduced, dividing the images into three regions based on temperature range: a core high-temperature region, a relatively high-temperature region, and a low-temperature region, including: Statistical analysis was performed on the temperature of infrared images of blast furnace burden surface, and a first temperature threshold and a second temperature threshold were set, with the first temperature threshold being greater than the second temperature threshold. Based on the first temperature threshold and the second temperature threshold, the infrared image of the blast furnace burden surface is divided into regions. The core high-temperature region is defined as the region with a temperature greater than the first temperature threshold, the relatively high-temperature region is defined as the region with a temperature between the second temperature threshold and the first temperature threshold, and the remaining part is defined as the low-temperature region.

3. The method according to claim 1, characterized in that, Within the central region determined by the radius parameter, multimodal centroid coordinates are calculated by combining temperature characteristics and texture gradient characteristics, including: Within the central region determined by the radius parameter, the edge response is extracted as a texture feature by the Sobel operator, linearly fused with the temperature feature, and the feature weight of each pixel is calculated and normalized. Based on the normalized feature weights, the position coordinates of each pixel in the central region are weighted and averaged to calculate the new multimodal centroid coordinates.

4. A blast furnace burden surface segmentation device based on dynamic radius and multimodal characteristics, characterized in that, include: An initialization module is used to initialize the center point coordinates of the infrared image of the blast furnace burden surface, and to set a circular region with a variable radius centered on the center point coordinates as a set of candidate radii. The region division module is used for statistical analysis of the temperature of infrared images of blast furnace burden surface. It introduces a dual temperature threshold mechanism to divide the image into three types of regions based on the temperature range. The radius constraint optimization module is used to establish radius double constraint conditions based on the three types of regions. Through the radius double constraint conditions, the radius parameters that meet the temperature distribution law are initially determined in the candidate radius set, so as to realize the adaptive adjustment of the central region range. The centroid calculation module calculates multimodal centroid coordinates within the central region determined by the radius parameter, combining temperature features and texture gradient features. The iterative optimization module updates the center point coordinates with the multimodal centroid coordinates, and alternately updates the radius parameter and center position through iterative loop until convergence, thereby obtaining the optimal center point coordinates and optimal radius, and then segments the infrared image of the blast furnace material surface into a high-temperature central region and a low-temperature peripheral region. The partitioning module reconstructs the high-temperature central region and the low-temperature peripheral region using different parameters, and performs Gaussian weighted fusion on the reconstruction results of the two regions to generate a globally consistent final result. Specifically, based on the temperature distribution characteristics, a radius double constraint condition is established. Through this radius double constraint condition, radius parameters that satisfy the temperature distribution law are initially determined from the candidate radius set, including: Based on the temperature distribution characteristics of the infrared image of the blast furnace burden surface, a double constraint condition for the radius is designed, wherein: The first constraint is: within a radius of... R Within the central region, the proportion of pixels in the relatively high-temperature area to the total number of pixels in the relatively high-temperature area of ​​the entire image should exceed the first threshold. The second constraint is: within a radius of... R Within the central region, the proportion of pixels in the core high-temperature zone to the total number of pixels in the central region should exceed a given second threshold. Within the set of candidate radii, the radius is dynamically adjusted to satisfy the double radius constraint condition, and the radius parameters that satisfy the temperature distribution law are initially determined. Specifically, different parameters are set for the high-temperature central region and the low-temperature peripheral region for reconstruction, and the reconstruction results of the two regions are Gaussian-weightedly fused to generate a globally consistent final result, including: The characteristics of the high-temperature center region and the low-temperature periphery region are analyzed, and differentiated parameters are set for different regions. For the high-temperature center region, compared with the low-temperature periphery region, a larger support window is constructed when the cost is aggregated. That is, a larger arm length limit is set when constructing the cross arm, and the gradient constraint strength is increased. Stereo matching is performed on different regions using pre-set differential parameters to obtain two disparity maps. The disparity maps are then fused using Gaussian weighting to obtain the final result. The construction rules for the support window are as follows: （1） ； （2） ； (3) ,if ; （4） ； (5) ,if ; in This represents the color difference between pixel p and pixel q. It is a pre-set threshold. Represents the distance between pixels p and q, through Control the arm length; For threshold; This represents the gradient difference between pixel p and pixel q.

5. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-3.

6. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-3.

Images