Intelligent online detection method and system for moisture content of sintering mixture

Abstract

The application discloses a kind of sintering mixing water content intelligent online detection method and system, by obtaining sintering mixing image, extract the color feature of sintering mixing image under HSV space, using gray level co-occurrence matrix to extract the texture feature of sintering mixing image, extract the morphological feature of sintering mixing image, according to the color feature, texture feature and morphological feature of sintering mixing image, establish sintering mixing water content detection model and based on sintering mixing water content detection model, sintering mixing water content is detected, the technical problem that the existing sintering mixing water content detection precision is low is solved, intelligent method is introduced into sintering mixing water content intelligent detection for the first time, greatly improve the stability and detection precision of sintering mixing water content detection.

Description

Technical Field

[0001] This invention mainly relates to the field of steel industry technology, specifically to an intelligent online detection method and system for moisture content in sintering mixtures. Background Technology

[0002] The steel industry is a crucial component of my country's economic development and infrastructure construction. The sintering process is a key step in the steel smelting process, primarily providing qualified raw materials for blast furnace ironmaking. The moisture content of the sinter mix is ​​a critical parameter affecting downstream process control and bed permeability. Online detection of the mix moisture is essential for improving sinter yield and quality and reducing harmful gas emissions. However, the complex environment surrounding the sinter mixing conveyor belt makes rapid and effective moisture detection difficult.

[0003] Currently, moisture detection at sintering sites combines offline drying with online detection. Online detection primarily involves installing moisture analyzers between the primary and secondary mixing zones to measure the material's moisture content. However, due to the characteristics of sintering mixes—frequent changes in material composition, irregular material stacking, high surface water vapor concentration, and large temperature fluctuations—accurate online moisture detection is difficult. Existing methods for detecting mix moisture can be divided into two main categories: direct and indirect detection. Direct detection includes thermal drying and the Karl Fischer method, but the requirement for offline sampling and analysis results in significant detection lag, making it difficult to meet the real-time requirements of moisture detection at the sintering site. Indirect measurement is divided into contact and non-contact methods. Contact measurement methods are rarely used at the sintering site due to the rough surface, large undulations, and high belt speeds of the mixes. Non-contact methods include near-infrared moisture analyzers and microwave moisture analyzers. However, due to the high surface water vapor concentration, large variations in material ratios, and large temperature fluctuations of the mixes, high-precision moisture detection using near-infrared and microwave moisture analyzers is difficult to achieve. Currently, the most effective method for moisture detection at the sintering site still relies on workers who rely on their long-accumulated prior knowledge and experience to judge the moisture content based on the color, texture, and shape of the sintered mixture surface.

[0004] Because non-contact moisture detectors inevitably encounter numerous external interference factors during the moisture detection process of sintering mixtures, and these interference factors are difficult to eliminate through hardware improvements to the instrument itself, existing instruments exhibit significant fluctuations in their readings. Currently, the most accurate moisture detection method still relies on the experience of operators to determine the moisture content of the sintering mixture. Therefore, this embodiment proposes an intelligent online moisture detection method for sintering mixtures based on the experience of sintering moisture inspectors. This method achieves continuous and accurate measurement of the moisture content of the sintering mixture, overcoming the challenges of non-contact detection methods being greatly affected by environmental factors and having significant time delays in direct measurement.

[0005] The invention patent with publication number CN113447392A proposes a calibration and verification method for a sintering mixture moisture measurement device. The patent uses the drying and weighing method to obtain the relative true value of the material moisture content at the sampling point at the sampling time, providing a reference value for the measurement benchmark for device verification. Then, it combines the expert experience method to express the current moisture percentage of the mixture using linguistic variables or fuzzy mathematics, which is used to verify the consistency of the device's dynamic measurement of large / small or slightly large / slightly small values. Finally, based on the drying and weighing method and the worker / expert experience method, the continuous measurement value of the device is calibrated and verified using the big data analysis relative correction method.

[0006] The big data correction part of this invention has a small data volume and does not take into account various external factors, making it difficult to cope with changes in working conditions.

[0007] The invention patent with publication number CN108133106B proposes a method for estimating the moisture content of sintering machine mixtures based on the amount of water added. This method allows operators to estimate the moisture content of the mixture in the sintering machine based on the amount of water added during the sintering mixing process. This enables operators to predict the moisture content of the mixture in the later stages of sintering based on the water added in the preceding mixing stages, facilitating adjustments to the water addition amount based on the process conditions of the sintering mixture's moisture content, thus optimizing the water addition during the sintering mixing process. While this invention establishes a formula for calculating moisture content, accurate calculation is difficult due to fluctuations in material proportions and numerous external interference conditions at the sintering site. Summary of the Invention

[0008] The intelligent online detection method and system for moisture content in sintering mixtures provided by this invention solves the technical problem of low accuracy in existing moisture content detection methods for sintering mixtures.

[0009] To solve the above-mentioned technical problems, the present invention proposes an intelligent online detection method for moisture content in sintering mixes, comprising:

[0010] Acquire images of the sintering mixture;

[0011] Extract color features of sintered mixture images in HSV space;

[0012] Texture features of sintered mixture images were extracted using the gray-level co-occurrence matrix.

[0013] Extract morphological features from sintering mixture images;

[0014] Based on the color, texture, and morphological features of the sintering mixture image, a moisture detection model for the sintering mixture is established.

[0015] Based on the sintering mixture moisture detection model, the moisture content of the sintering mixture is detected.

[0016] Furthermore, after acquiring the sintering mixture image, before extracting the color features of the sintering mixture image in the HSV space, the sintering mixture image also undergoes preprocessing.

[0017] Furthermore, the color features extracted from the sintering mixture image in HSV space include:

[0018] Convert the sintering and mixing image from RGB space to HSV space;

[0019] The mean, standard deviation, and slop of the sintered mixture image are calculated in HSV space to obtain the color characteristics of the sintered mixture image.

[0020] Furthermore, the extraction of texture features from the sintering mix image using the gray-level co-occurrence matrix includes:

[0021] The sintering mixture image is converted into a grayscale image, and the grayscale co-occurrence matrix of the grayscale image is calculated.

[0022] Five statistical measures were calculated based on the gray-level co-occurrence matrix: inverse difference, energy, entropy, moment of inertia, and correlation.

[0023] Texture features of sintered mixture images are obtained based on five statistical measures: contrast, energy, entropy, moment of inertia, and correlation.

[0024] Furthermore, the morphological features extracted from the sintering mixture image include:

[0025] Based on the characteristics that sintered mixtures with high moisture content have more and larger agglomerates on their surface, while sintered mixtures with low moisture content have smaller and smoother agglomerates on their surface, the shaded area formed by agglomerates is defined as indirectly representing the degree of agglomeration on the surface of the sintered mixture.

[0026] Extract the effective shadow blocks from the sintering mixture image;

[0027] Based on the effective shadow blocks in the sintering mixture image, the maximum shadow area, average shadow area, and number of shadows are obtained as morphological features of the sintering mixture image.

[0028] Furthermore, the effective shadow blocks extracted from the sintering mixture image include:

[0029] The sintering mixture image is converted to grayscale.

[0030] The grayscale image of the sintered mixture is binarized.

[0031] Invert the binarized image of the sintered mixture.

[0032] The 8-neighborhood traversal marking method is used to mark the connected components in the sintered mixture image after the inversion operation, thereby obtaining the effective shaded blocks of the sintered mixture image.

[0033] Furthermore, the 8-neighborhood traversal marking method is used to mark the connected components in the sintered mixture image after the inversion operation. The calculation formula for the effective shaded blocks of the sintered mixture image is as follows:

[0034]

[0035] Where f(x,y) is the pixel with coordinates x,y on a visible light image of size M×N, and S valid This represents the effective shaded area.

[0036] Furthermore, based on the effective shadow blocks in the sintering mixture image, the formulas for calculating the maximum shadow area and the average shadow area are as follows:

[0037] S max =max(S1,S2,…,S) N ),

[0038]

[0039] Where S1, S2, ... S N Let N be the area of ​​all valid shaded blocks, and N be the number of valid shaded blocks.

[0040] Furthermore, based on the color, texture, and morphological features of the sintering mixture image, a moisture detection model for the sintering mixture is established, including:

[0041] Calculate the maximum mutual information coefficients of color features, texture features, and morphological features of the sintered mixture image;

[0042] Feature quantities with the largest mutual information coefficient greater than a preset threshold are selected, and XGBoost is used to construct a moisture detection model for sintering mixtures.

[0043] The intelligent online moisture detection system for sintering mixes provided by this invention includes:

[0044] The present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the intelligent online detection method for moisture content of sintering mixtures provided by the present invention.

[0045] Compared with the prior art, the advantages of the present invention are as follows:

[0046] The present invention provides an intelligent online detection method and system for sintered material moisture. By acquiring images of sintered materials, extracting color features of the images in HSV space, extracting texture features of the images using a gray-level co-occurrence matrix, and extracting morphological features of the images, a moisture detection model for sintered materials is established based on the color, texture, and morphological features of the images. Based on this model, the moisture content of the sintered materials is detected. This invention solves the technical problem of low accuracy in existing sintered material moisture detection methods and introduces intelligent methods into intelligent moisture detection of sintered materials for the first time, greatly improving the stability and accuracy of moisture detection.

[0047] Key technical points of this invention:

[0048] (1) In view of the environment between the first and second mixing in the sintering site, a sintering mixing moisture detection system based on machine vision was built to transmit the required data to a remote computing device in real time and detect the results through the model proposed in this invention.

[0049] (2) This invention introduces intelligent methods into the moisture detection of sintering mixtures for the first time. By cooperating with workers on the sintering site to dry the sintering mixtures in batches from the hot drying oven, the true value of the moisture content is obtained. Simultaneously, image data of the sintering mixtures is acquired, and the required dataset is constructed. Based on the features of multiple heterogeneous images, an intelligent detection model for moisture of sintering mixtures based on the independently constructed dataset is established to detect the moisture of the sintering mixtures.

[0050] (3) Based on the prior knowledge of water monitoring workers, when the moisture content of the sintering mixture is high, the top of the mixture pile forms a sharp angle, the material agglomeration rate is high, the agglomerates are large, and the overall surface texture is messy, rough, and dark in color. The image color space is converted to the HSV space, which is more in line with human vision, and color features are extracted. The gray-level co-occurrence matrix of the image is calculated, and texture features are extracted based on this.

[0051] (4) Since the clumping features of sintered mixture images are difficult to extract using existing morphological feature extraction methods, this invention analyzes the clumping situation and characteristics of sintered mixtures on images with different moisture contents, defines the maximum area, average area and number of shadow blocks between the wrinkles on the surface of the mixture as clumping features, and proposes a corresponding clumping feature extraction method, providing key information for the subsequent establishment of an image-based moisture detection model for sintered mixtures.

[0052] (5) The sintering mixture moisture detection method of the present invention has for the first time realized the continuous and accurate measurement of the moisture content of the sintering mixture. Attached Figure Description

[0053] Figure 1 This is a diagram of the sintering mixture moisture detection system according to Embodiment 2 of the present invention;

[0054] Figure 2 The flowchart of the intelligent detection model for moisture in sintered mixtures based on multiple types of heterogeneous image features in Embodiment 2 of the present invention is shown below.

[0055] Figure 3 This is a structural block diagram of the intelligent online moisture detection system for sintering mixtures according to an embodiment of the present invention.

[0056] Figure label:

[0057] 1. Belt support; 2. Feeding belt; 3. Sintering and mixing; 4. Visible light camera; 5. Camera bracket; 6. LCD display; 7. Fiber optic cable; 8. Main unit; 10. Memory; 20. Processor. Detailed Implementation

[0058] To facilitate understanding of the present invention, the present invention will be described more fully and in detail below with reference to the accompanying drawings and preferred embodiments, but the scope of protection of the present invention is not limited to the following specific embodiments.

[0059] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

[0060] Example 1

[0061] The intelligent online detection method for moisture content in sintering mixes provided in Embodiment 1 of the present invention includes:

[0062] Step S101: Obtain the sintering mixture image;

[0063] Step S102: Extract the color features of the sintering mixture image in HSV space;

[0064] Step S103: Extract texture features from the sintering mixture image using the gray-level co-occurrence matrix;

[0065] Step S104: Extract the morphological features of the sintering mixture image;

[0066] Step S105: Based on the color features, texture features, and morphological features of the sintering mixture image, establish a moisture detection model for the sintering mixture.

[0067] Step S106: Detect the moisture content of the sintering mixture based on the sintering mixture moisture detection model.

[0068] The intelligent online detection method for moisture content in sintered materials provided in this invention acquires an image of the sintered material, extracts its color features in the HSV space, extracts its texture features using a gray-level co-occurrence matrix, and extracts its morphological features. Based on the color, texture, and morphological features of the sintered material image, a moisture detection model for the sintered material is established, and the moisture content of the sintered material is detected based on this model. This method solves the technical problem of low accuracy in existing sintered material moisture detection methods and is the first to introduce intelligent methods into the intelligent detection of moisture content in sintered materials, greatly improving the stability and accuracy of moisture detection.

[0069] Example 2

[0070] This invention proposes an online method for detecting moisture content in sintering mixes based on multi-source information fusion. Figure 1 This is a schematic diagram of an online moisture detection system for sintering mixture installed at the mixing belt after sintering. It mainly consists of a belt support 1, a feeding belt 2, sintering mixture 3, a visible light camera 4, a camera bracket 5, an LCD display 6, an optical fiber 7, and a host 8. Figure 2 This is a flowchart illustrating the implementation steps of the continuous and accurate moisture detection method proposed in this invention, including the following steps:

[0071] 1. After sintering and mixing, a visible light camera is installed to acquire visible light image data of the sintered mixture. The visible light image is cropped and rotated to remove as much irrelevant area outside the mixture as possible to obtain the region of interest (ROI). Bilateral filtering is then applied to the ROI to remove noise. The sintered mixture at the corresponding time point is sampled and dried as the true value for subsequent processing.

[0072] 2. Sintered mixtures with different moisture contents have obvious color differences. Using the HSV color model, which is more in line with the visual characteristics of the human eye, the mean, standard deviation, and slope of the mixture image in the HSV color space are extracted as features.

[0073] 3. To address the different roughness and irregularity exhibited by the mixed texture under different moisture contents, the gray-level co-occurrence matrix was selected to calculate its texture features. First, the image was converted into a grayscale image, and its gray-level co-occurrence matrix was calculated. Based on the gray-level co-occurrence matrix, five statistical quantities, namely, inverse difference, energy, entropy, moment of inertia, and correlation, were calculated as subsequent modeling variables.

[0074] 4. Moisture content is closely related to the degree of agglomeration on the surface of the mixture. Surfaces with higher moisture content produce more and larger agglomerates, while surfaces with lower moisture content produce smaller and smoother agglomerates. Since the boundary between agglomerates and their surroundings is not clear, directly extracting agglomeration characteristics is difficult. Considering that wrinkles form around agglomerated materials, and these wrinkles create coarse shadows that effectively distinguish them from the surrounding environment, the shadow areas formed by agglomerates are used to indirectly represent the degree of agglomeration on the surface of the mixture. The maximum shadow area, average shadow area, and number of shadows are defined and extracted as its morphological characteristics.

[0075] 5. Calculate the maximum information coefficients of image color, texture, and morphology features, then select the features with strong correlation and input them into XGBoost for training to obtain a sintering mixture moisture detection model based on image features.

[0076] The specific operational plan is as follows:

[0077] (1) Obtain visible light images and process variable data:

[0078] The visible light camera is installed on the suspended beam of the chamber away from the mixing outlet to ensure that the visible light camera can obtain a complete image of the mixing surface, while avoiding unstable video streams caused by belt vibration and contamination of the visible light camera lens by a large amount of dust and fog on site.

[0079] A visible light camera is used to capture images of the material between the first and second mixing stages. The video stream is then transmitted via a dedicated fiber optic network to a computer in the monitoring room for storage. Image data is captured at a sampling frequency of one hour, and the desired images are obtained remotely via a connection to the monitoring room computer. This data is input into the sintering mixture moisture detection model constructed in this embodiment, thereby enabling real-time detection of the moisture content of the sintering mixture. Sintering mixtures from the sintering conveyor belt after the first mixing stage are sampled at the same sampling frequency. The mixture is sealed, and ten samples are taken as a batch and sent to a hot drying oven for drying to obtain the true moisture content of the sintering mixture at the corresponding time.

[0080] Because the acquired visible light image contains too many irrelevant regions and noise, image preprocessing is necessary. First, the image is rotated counterclockwise by α degrees. Then, a rectangle formed by (x1, y1) and (x2, y2) is selected to crop the image, removing as many irrelevant regions as possible to obtain the region of interest. A bilateral filter, which better preserves edge characteristics, is then used to remove image noise interference. The specific preprocessing steps are as follows:

[0081] Step 1: The size of the acquired visible light image is 1920×1080. Since the shooting angle is from the side towards the mixed material, the image needs to be rotated 28 degrees counterclockwise first. Due to the interference of debris at the edge of the image, two diagonal points (40, 350) and (1600, 650) are selected in sequence to form a rectangular area to crop the image. The obtained rectangular area is taken as the region of interest of the image.

[0082] Step 2: Due to the harsh environment, the images captured by the camera are subject to noise interference during transmission, necessitating image denoising. While traditional denoising methods such as Gaussian filtering, mean filtering, and median filtering can effectively remove high-frequency noise, they also remove some surface texture of the sintered material because edges are also high-frequency signals. Since subsequent texture feature extraction using the gray-level co-occurrence matrix is ​​required, this patent utilizes bilateral filtering, which better preserves edge characteristics, to remove image noise interference. Bilateral filters smooth images and reduce noise while preserving edges; they are nonlinear bilateral filters. The kernel function of a bilateral filter is a combination of the spatial domain kernel and the pixel range domain kernel. In flat regions, the difference in pixel values ​​between neighboring pixels is small, corresponding to a domain weight close to 1. In this case, the spatial domain weight plays a major role, equivalent to directly applying Gaussian blur to this region. Therefore, flat regions are equivalent to Gaussian blur. In edge regions, the difference in pixel values ​​between neighboring pixels is large, corresponding to a domain weight close to 0, causing the kernel function to decrease. This reduces the impact on the current pixel, thus preserving the edge details of the original image.

[0083] (2) Extracting the color features of the sintered mixture image in HSV space:

[0084] Since sintered mixtures with different moisture contents exhibit significant color differences, the HSV color model, which better aligns with human visual perception, is used. It intuitively expresses the hue, vibrancy, and brightness of colors, facilitating color comparison. Here, H represents the hue of the image, indicating the basic attributes of different colors; S refers to the saturation of the color, representing its purity; higher saturation results in a more vibrant color; and V represents the brightness of the color, determining the overall brightness of the image. The mean, standard deviation, and slope of the mixture image in the HSV color space are extracted as features. First, the image is converted from RGB space to HSV space, and then its mean, standard deviation, and slope are calculated as color features of the mixture image. The calculation formulas are as follows:

[0085] 1) Mean:

[0086]

[0087] 2) Standard deviation:

[0088]

[0089] 3) Slope:

[0090]

[0091] Where P ij N represents the different pixels in the image after conversion to grayscale, where N is the number of pixels in the image.

[0092] (3) Extract texture features from sintered mixture images using the gray-level co-occurrence matrix:

[0093] The moisture content of the sintering mix varies, resulting in different textures, shapes, and colors on its surface. When the moisture content is high, the top of the mix pile forms sharp angles, indicating high agglomeration rates, large agglomerates, and a generally messy and rough surface texture. Conversely, when the moisture content is low, the top of the mix pile forms rounded arcs, indicating low agglomeration rates, small agglomerates, and a generally regular and smooth surface texture, indicating lower moisture content. On-site inspectors assess the moisture content of the sintering mix by observing its color, texture, and shape. This patent utilizes a grayscale co-occurrence matrix to extract image texture features based on the experience of on-site inspectors. The specific steps are as follows:

[0094] To address the varying roughness and irregularity of the mixture under different moisture contents, a gray-level co-occurrence matrix (GLCM) was selected to calculate its texture features. First, the image was converted to grayscale, and its GLCM was calculated. Based on the GLCM, five statistical measures—inverse difference, energy, entropy, moment of inertia, and correlation—were calculated as subsequent modeling variables. The calculation formulas are as follows:

[0095] 1) Energy:

[0096]

[0097] Energy characterizes the uniformity of gray levels in an image. If the gray values ​​of individual pixels in an image do not vary significantly, the elements on the main diagonal of the spatial gray-level co-occurrence matrix will have larger values ​​and higher energy.

[0098] When the surface of the mixture has a high moisture content, the texture becomes rougher and more uneven. Energy represents the uniformity of image grayscale. A high energy value indicates that the current texture is a relatively stable texture with regular variations.

[0099] 2) Entropy:

[0100]

[0101] Entropy characterizes the complexity of texture in an image and is a measure of the uniformity of element distribution in a spatial gray-level matrix. The smaller the entropy value, the more uniform the image.

[0102] The more complex the texture of an image, the higher its entropy. Conversely, if the gray levels are uniform but the elements in the spatial gray-level co-occurrence matrix vary significantly, the entropy will be lower. When the moisture content of an image changes, the mixed particles will form clumps of different sizes due to the effect of moisture; therefore, the overall irregularity of the texture changes continuously with the change in moisture content.

[0103] 3) Moment of inertia:

[0104]

[0105] Moment of inertia describes the total amount of gray-level variation in an image. It indicates that the more frequently pixel pairs with unequal gray-level values ​​appear, the more uneven the image and the finer the texture.

[0106] 4) Correlation:

[0107]

[0108] This value measures the similarity of gray levels in an image along the row or column direction. Therefore, the magnitude of the value reflects local gray-level correlation; a larger value indicates a stronger correlation. When the water content is low, the similarity of gray levels in the row or column direction increases.

[0109] 5) Contrast:

[0110]

[0111] This reflects the clarity and regularity of the texture. Clear, regular, and easily described textures have higher values; chaotic and difficult-to-describe textures have lower values. When the moisture content increases, the originally regular image becomes chaotic.

[0112] (4) Extraction of morphological features from sintering mixing images

[0113] Based on the experience of workers monitoring water levels, the degree of agglomeration on the surface of the mixture is closely related to moisture content. Surfaces with higher moisture content tend to have more and larger agglomerates, while surfaces with lower moisture content have smaller and smoother agglomerates. Since the boundary between agglomerates and their surroundings is not clearly defined, directly extracting agglomeration characteristics is difficult. Considering that wrinkles form around agglomerated materials, and these wrinkles create coarser shadows that effectively distinguish the agglomerate from its surroundings, this embodiment uses the shadowed areas formed by agglomerates to indirectly represent the degree of agglomeration on the surface of the mixture, and defines the maximum shadow area (S). max ), average shaded area (S) average The number of shadows (N) is used as its morphological feature, and the steps are as follows:

[0114] Step 1: Since shadows generally exist as black blocks in an image, extracting the shadow area first requires separating the black block regions. However, RGB images have colors distributed across three channels, making it difficult to extract black blocks. Therefore, the RGB image needs to be converted to a binary image before shadow extraction. First, the image is converted to grayscale. In this embodiment, a traditional grayscale conversion method is used:

[0115] Gray=0.299×R+0.587×G+0.114×B (9)

[0116] Step 2: To separate the shadow areas, further binarization processing of the image is required. Since the colors of the sintered mixture image are relatively fixed, a fixed threshold is used for binarization, and the binarized image is then inverted.

[0117] Step 3: The separated binarized image clearly shows the shadow areas. To extract the shadow area, average shadow area, and number of shadows, connected component analysis is needed to calculate the area and number of each shadow patch. Connected component analysis is the process of detecting connected regions in an image using a specific scanning method and assigning different labels to different connected regions. This study uses the 8-neighborhood traversal labeling method to label the connected regions in the image, and then statistically analyzes parameters such as the area and number of shadow patches. This study considers the interference of a large number of small shadow patches caused by the irregularity of the mixed surface itself, and sets the effective color patch area S. valid The value is 100. Shadow blocks smaller than this value are not included in the statistics. First, all valid shadow blocks in a single image are calculated. The formula for calculating a single shadow block is as follows:

[0118]

[0119] The extracted effective shadow blocks from a single image are: [S1, S2, ..., S N ].

[0120] The number of valid shadow blocks is denoted as the number of shadows (N). The maximum value among the valid shadow blocks is calculated and denoted as the maximum shadow area (S). max ):

[0121] S max =max(S1,S2,…,S) N (11)

[0122] Calculate the average area of ​​all valid shadow blocks in a single image, and denot it as the average shadow area (S). average ):

[0123]

[0124] Where S1, S2, ... SN Let N be the area of ​​all valid shaded blocks, and N be the number of valid shaded blocks.

[0125] (5) Construction of a moisture detection model for sintered mixtures based on features of multiple heterogeneous images:

[0126] The selected process variables are based solely on experience, suggesting a strong correlation with the moisture content of the sintering mix, which is highly subjective. Considering the difficulty in obtaining labels and the limited data volume, it is necessary to quantitatively analyze the correlation coefficients between process variables and the moisture content of the sintering mix, thereby identifying process variables with strong correlations. Because the relationship between process variables and the moisture content of the sintering mix is ​​highly non-linear, conventional correlation coefficient calculation methods are insufficient to characterize this relationship. Therefore, this patent uses the maximum information coefficient to calculate the correlation between variables. The maximum information coefficient can capture the complex relationships between variables and has the advantages of universality, fairness, and symmetry. Its calculation formula is as follows:

[0127]

[0128]

[0129] In the formula, I(x,y) is the mutual information between the process variable x and the moisture content y of the sintered mixture, a and b are the number of grids in the x and y directions, and B is a fixed value, usually set to about 0.6 times the amount of data. In formula (14), p(x,y) is the joint probability distribution of x and y, p(x) is the marginal probability distribution of x, and p(y) is the marginal probability distribution of y.

[0130] Features with large mutual information coefficients were selected through correlation analysis, and a model was built using XGBoost. XGBoost is an improvement on the gradient boosting algorithm. It features high accuracy, low overfitting risk, and strong scalability. Based on the decision tree algorithm, XGBoost adds a regularization term including the number of leaf nodes and the sum of squares of the leaf nodes. Therefore, its objective function consists of two parts: a loss function and a regularization term. The loss function is the difference between the predicted and actual values; reducing the loss optimizes the objective function. The regularization term controls the model's complexity and prevents extreme cases. The optimal objective function of the XGBoost algorithm is:

[0131]

[0132] In the formula, λ is a fixed coefficient, γ is a complexity coefficient, T is the number of leaf nodes in the tree, and G is a variable. j H is the sum of the first-order partial derivatives of the samples contained in leaf node j. j It is the sum of the second-order partial derivatives of the samples contained in leaf node j.

[0133] The test set was used as input to the trained model to obtain the moisture content of the sintered mixture. The RMSE, MSE, MAE, MAPE, and hit rate J of the fusion model and the near-infrared moisture detector were calculated respectively. The four evaluation indicators of the fusion model and the near-infrared moisture detector were compared and analyzed. The calculation formulas of RMSE, MSE, MAE, MAPE, and hit rate J are as follows:

[0134] Mean square error:

[0135]

[0136] Root mean square error:

[0137]

[0138] Mean absolute error:

[0139]

[0140] Where y (i) For the true value of the sample, These are predicted values.

[0141] Based on the experience of on-site experts, an absolute error between the predicted and actual test values ​​within 0.1% is considered acceptable. To more intuitively demonstrate the model's predictive performance, the model's hit rate J is considered, defined as:

[0142]

[0143] Where l represents the length of the predicted sample. H(·) represents the predicted value of the sample, and H(·) is the Heaviside function, defined as:

[0144]

[0145] This invention focuses on the mixing of materials on the conveyor belt between the primary and secondary mixing sections of a sintering plant, proposing an intelligent online detection method and system for sintering materials moisture. Based on the characteristics of the sintering environment, this invention establishes a machine vision-based detection system capable of extracting image data in real time and transmitting it to a terminal for processing. Targeting the color and texture characteristics of the sintering material surface, color features of the mixing image in HSV space are extracted, and texture features are extracted using a gray-level co-occurrence matrix. Based on the agglomeration morphology of the sintering material, the maximum shadow area, shadow quantity, and average shadow area are defined and extracted as morphological features. Finally, considering the strong fitting ability of XGBoost with small sample sizes, this invention inputs multiple heterogeneous features into XGBoost to establish an intelligent model. This invention is the first to introduce intelligent methods into the intelligent detection of moisture in sintering materials, improving the current problems of poor stability and accuracy in moisture detection.

[0146] Furthermore, based on the long-term experience of workers observing moisture levels at the sintering site, the degree of agglomeration on the surface of the mixture is closely related to moisture content. Surfaces with higher moisture content tend to have more and larger agglomerates, while surfaces with lower moisture content have smaller and smoother agglomerates. Because the boundary between agglomerates and the surrounding environment is not clear, and the complex surface texture of the mixture image significantly interferes with the extraction of agglomeration characteristics, directly extracting agglomeration features is quite difficult. Considering that when the sintering mixture contains a high amount of moisture, the particles will form agglomerates of a certain size due to the binding effect of moisture. These agglomerates will form wrinkles, and the wrinkled parts will have relatively coarse shadows. These shadows can effectively distinguish the mixture from the surrounding environment, offering advantages such as easy extraction, immunity to environmental interference, and strong correlation with moisture content. Therefore, this embodiment of the invention uses shadows to indirectly represent the agglomeration characteristics of the mixture.

[0147] Because the surface texture of the sintered mixture is complex, the shadows extracted by existing shadow extraction methods contain a lot of noise. The calculated average shadow area and shadow quantity characteristics are affected by a large number of small shadow blocks and have large errors, with weak correlation. The shadow extraction method proposed in this embodiment can effectively extract effective shadows, and the proposed maximum shadow area, average shadow area, and shadow quantity have been experimentally proven to reflect the moisture content of the sintered mixture well.

[0148] Example 3

[0149] This embodiment uses a 360m³ sintering plant as an example. 2 The sintering plant served as an experimental platform, where the invented moisture detection device was applied to the moisture detection of the mixed materials after the first mixing stage. The moisture detection device was installed on a suspended beam at the discharge port of the first mixing stage, 4 meters away from the discharge port, and its construction is shown in the attached figure. Figure 1 The sintering mixture moisture detection system shown.

[0150] To verify the effectiveness of the moisture detection device and method, the effectiveness of the moisture detection results will be indirectly explained from two perspectives:

[0151] ① Take samples and dry them to obtain the actual moisture value of the sintered mixture.

[0152] Currently, the most accurate method for moisture detection is the sampling and drying method. The drying method at the sintering site involves taking 500 grams or more of the mixture and drying it in a drying oven at a temperature greater than 120 degrees Celsius for 2 hours to obtain the true moisture value. The moisture content of the sintering mixture measured by the drying method is used as the true value. The effectiveness of the invention is verified by calculating the mean square error, root mean square error, mean absolute error, and hit rate of the detection results of the invention. The accuracy of the invention is verified by comparing it with the detection accuracy of existing near-infrared moisture detectors at the sintering site.

[0153] ②The experience of sintering workers in monitoring water flow

[0154] Through multiple surveys at the sintering site, the most effective method for judging the moisture content of sintering mixes is currently to have on-site inspectors observe the agglomeration rate, agglomerate size, material color, and surface texture of the mix to comprehensively determine the moisture content. After applying this method at the sintering site, the moisture content of the sintering mixes generally matches the moisture content judged by on-site inspectors. Furthermore, when the mix surface exhibits a high agglomeration rate, large agglomerates, and a pointed material accumulation shape, the moisture content measured by this method is mostly higher than 7.1%. Conversely, when the mix surface has a low agglomeration rate, small agglomerates, material accumulation in an arc shape, and a relatively smooth surface, the moisture content measured by this method is mostly lower than 6.9%. The moisture detection results of this invention are consistent with the experience of on-site inspectors.

[0155] Reference Figure 3 The intelligent online moisture detection system for sintering mixtures proposed in this embodiment of the invention includes:

[0156] The system includes a memory 10, a processor 20, and a computer program stored on the memory 10 and executable on the processor 20. When the processor 20 executes the computer program, it implements the steps of the intelligent online detection method for moisture content of sintering mixtures proposed in this embodiment.

[0157] The specific working process and working principle of the intelligent online detection system for moisture content in sintering mixtures in this embodiment can be referred to the working process and working principle of the intelligent online detection method for moisture content in sintering mixtures in this embodiment.

[0158] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent online detection of moisture content in sintering mixtures, characterized in that, The method includes: Acquire images of the sintering mixture; Extract color features of sintered mixture images in HSV space; Texture features of sintered mixture images were extracted using the gray-level co-occurrence matrix. Morphological features of the sintering mixture image are extracted, including: Based on the characteristics that sintered mixtures with high moisture content have more and larger agglomerates on their surface, while sintered mixtures with low moisture content have smaller and smoother agglomerates on their surface, the shaded area formed by agglomerates is defined as indirectly representing the degree of agglomeration on the surface of the sintered mixture. Extract the effective shadow blocks from the sintering mixture image; Based on the effective shadow blocks of the sintering mixture image, the maximum shadow area, average shadow area, and number of shadows are obtained as morphological features of the sintering mixture image. Based on the color, texture, and morphological features of the sintered mixture images, a moisture detection model for the sintered mixture is established. This model includes: Calculate the maximum mutual information coefficients of color features, texture features, and morphological features of the sintered mixture image; Select the feature quantity whose maximum mutual information coefficient is greater than the preset threshold, and use XGBoost to build a moisture detection model for sintering mixture; Based on the sintering mixture moisture detection model, the moisture content of the sintering mixture is detected.

2. The intelligent online detection method for moisture content of sintering mixtures according to claim 1, characterized in that, After acquiring the sintering mixture image, the process of preprocessing the sintering mixture image before extracting its color features in the HSV space also includes obtaining the sintering mixture image.

3. The intelligent online detection method for moisture content of sintering mixtures according to claim 1, characterized in that, The color features extracted from the sintering mixture image in HSV space include: Convert the sintering and mixing image from RGB space to HSV space; The mean, standard deviation, and slop of the sintered mixture image are calculated in HSV space to obtain the color characteristics of the sintered mixture image.

4. The intelligent online detection method for moisture content of sintering mixtures according to claim 3, characterized in that, Texture features extracted from sintered mixture images using the gray-level co-occurrence matrix include: The sintering mixture image is converted into a grayscale image, and the grayscale co-occurrence matrix of the grayscale image is calculated. Five statistical measures were calculated based on the gray-level co-occurrence matrix: inverse difference, energy, entropy, moment of inertia, and correlation. Texture features of sintered mixture images are obtained based on five statistical measures: contrast, energy, entropy, moment of inertia, and correlation.

5. The intelligent online detection method for moisture content of sintering mixtures according to claim 4, characterized in that, Effective shadow blocks for extracting sintered mixture images include: The sintering mixture image is converted to grayscale. The grayscale image of the sintered mixture is binarized. Invert the binarized image of the sintered mixture. The 8-neighborhood traversal marking method is used to mark the connected components in the sintered mixture image after the inversion operation, thereby obtaining the effective shaded blocks of the sintered mixture image.

6. The intelligent online detection method for moisture content of sintering mixtures according to claim 5, characterized in that, The connected components in the sintered mixture image after the inversion operation are marked using the 8-neighborhood traversal marking method. The formula for calculating the effective shaded blocks in the sintered mixture image is as follows: ， in, The corresponding coordinates on the visible light image are pixels, Set the effective shadow area value.

7. The intelligent online detection method for moisture content of sintering mixtures according to claim 6, characterized in that, Based on the effective shadow blocks in the sintering mixture image, the formulas for calculating the maximum shadow area and the average shadow area are as follows: ， in, To calculate the area of ​​all valid shaded blocks, The number of valid shaded blocks.

8. A smart online moisture detection system for sintering mixtures, the system comprising: The memory (10), the processor (20), and the computer program stored in the memory (10) and executable on the processor (20) are characterized in that the processor (20) implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.

Images