Visual-based liquid level state recognition method and system for barium salt feeding process

Abstract

The present application relates to the technical field of image analysis, and particularly relates to a vision-based liquid level state recognition method and system for a barium salt feeding process. The method comprises: acquiring a liquid level image; extracting a foam breakage area from the liquid level image; extracting a to-be-recognized area, obtaining spatial distribution dispersion, and a fitting deviation between an edge of each to-be-recognized area and a circle; calculating a confidence degree of the to-be-recognized area belonging to a precipitation area; screening out the precipitation area from the to-be-recognized area according to a preset confidence threshold; calculating an equivalent geometric size of the precipitation area when the precipitation area is located on a liquid level surface layer; calculating a precipitation quantitative value; and generating a liquid level state category based on the deviation. The present application can improve the precision and production efficiency of barium sulfate product granularity control.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, specifically to a vision-based method and system for recognizing the liquid level state during barium salt feeding. Background Technology

[0002] In the continuous production of precipitated barium sulfate, a barium salt-sulfuric acid synthesis process based on the "reverse addition" principle is employed. A high-precision machine vision monitoring system is constructed by installing a high-definition industrial camera and auxiliary light source on the inner wall of the reactor. This system can capture real-time images of the dynamic evolution of barium sulfate precipitate at the liquid surface and perform online analysis of crystal particle size distribution, sedimentation rate, and morphological characteristics. Based on the real-time crystal particle size data fed back by the vision system, the control system performs closed-loop adjustment of the injection pressure of the barium chloride atomizing nozzle, thereby precisely controlling the contact mode and mixing intensity of the reactants. This visual feedback-based dynamic pressure control method can effectively intervene in the nucleation and growth stages of barium sulfate particles, optimize the crystal development environment, and ultimately achieve controllable preparation of product particle size.

[0003] However, existing technologies have the following technical problems during implementation: In the reaction process of precipitating barium sulfate, when using the reverse addition process, dilute sulfuric acid is pre-added to the reactor as a base liquid, and agitation is used to maintain a uniform flow field. Barium chloride solution is sprayed into the reactor at high speed through an atomizing nozzle, where it fully contacts the dilute sulfuric acid and reacts. Due to the high-speed jet impact of the barium chloride solution and the continuous disturbance of the liquid surface by the agitator, a large number of bubbles are generated on the reaction surface, forming a stable foam layer. This foam layer has similar texture, grayscale, and dynamic characteristics to barium sulfate precipitate, making it easily misidentified as precipitate by computer vision monitoring systems. This misidentification interferes with the system's accurate perception of the true precipitation state, leading to incorrect adjustment of the barium chloride spray pressure, affecting the nucleation and growth process of barium sulfate crystals, and ultimately reducing the accuracy of product particle size control and production efficiency. Summary of the Invention

[0004] This invention provides a vision-based method and system for recognizing the liquid level state during barium salt feeding, in order to solve existing problems.

[0005] The vision-based method for recognizing the liquid level state during barium salt feeding in this invention adopts the following technical solution: One embodiment of the present invention provides a vision-based method for recognizing the liquid level state during a barium salt feeding process, the method comprising the following steps: Acquire images of the liquid level inside the reaction vessel; Based on the projection position of the jet source on the liquid surface and the direction of liquid flow, the foam bursting area is extracted from the liquid surface image; The connected regions with gray values ​​higher than a preset threshold in the foam rupture area are extracted as the regions to be identified. The spatial distribution dispersion of gray-scale maxima in each region to be identified and the fitting deviation between the edge of each region to be identified and the circle are obtained. The spatial distribution dispersion is used to characterize the degree of aggregation of each gray-scale maxima relative to the center of the region to be identified, and the fitting deviation is used to characterize the degree of conformity between the edge of the region to be identified and the standard circle. The confidence level that the region to be identified belongs to the sedimentation region is calculated based on the product of the spatial distribution dispersion and the fitting bias. Based on a preset confidence threshold, sedimentation areas are filtered out from the areas to be identified; The average gray value and measured geometric dimensions of each sedimentation region are obtained. Based on the pre-established mapping relationship between gray value, geometric dimensions and depth, the actual depth of the sedimentation region below the liquid surface is determined. Based on the actual depth, the measured geometric dimensions are converted into the equivalent geometric dimensions of the sedimentation region when it is located on the surface of the liquid. The sedimentation quantification value is calculated based on the deviation between the equivalent geometric size of each sedimentation region after conversion and the standard particle size. Based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size, a liquid surface state category determined by the deviation is generated.

[0006] Furthermore, based on the projection position of the jet source on the liquid surface and the direction of liquid flow, the foam bursting area is extracted from the liquid surface image, specifically including: Obtain the location of the jet source from the liquid surface image; A conical atomizing cone is constructed with the position of the jet source as the cone apex, the distance from the jet source to the liquid surface as the cone height, and the atomization angle as the cone apex angle. The projection position of the bottom surface of the atomizing cone onto the liquid surface is determined as the reaction contact surface. The area in front of the reaction contact surface along the direction of liquid flow is defined as the foam-dense area; By removing the reaction contact surface and the area of ​​dense foam from the liquid surface image, the area of ​​foam rupture is obtained.

[0007] Furthermore, the spatial distribution dispersion of gray-level maxima in each region to be identified, and the fitting deviation between the edge of each region to be identified and the circle are obtained, specifically including: Obtain the center of each region to be identified and the pixel position corresponding to the maximum grayscale value; A position coordinate system is constructed with the center of each region to be identified as the origin, and the coordinates of the pixel corresponding to each maximum gray value are obtained in the coordinate system. The local outlier factor algorithm is used to obtain the local outlier factor of each pixel in the coordinate system, and the average value of all local outliers is calculated as the spatial distribution dispersion of the region to be identified. The radius is determined by the area of ​​each region to be identified, and a circle is constructed based on the radius. The Euclidean distance between each edge pixel of the region to be identified and the nearest point on the circle is calculated, and the average value of each Euclidean distance is determined as the fitting bias.

[0008] Furthermore, based on the product of spatial distribution dispersion and fitting bias, the confidence level that the region to be identified belongs to the sedimentation region is calculated, specifically including: The product of spatial distribution dispersion and fitting deviation is calculated, and the negative exponential function value with the natural constant as the base and the product as the exponent is determined as the confidence level that the region to be identified belongs to the sedimentation region.

[0009] Furthermore, based on the pre-established mapping relationship between grayscale values, geometric dimensions, and depth, the actual depth of the sedimentation region below the liquid surface is determined, and the measured geometric dimensions are converted into the equivalent geometric dimensions of the sedimentation region when it is located at the liquid surface, specifically including: Based on the mapping relationship, determine the actual depth of the sedimentation region below the liquid surface, which corresponds to the average gray value and measured geometric dimensions of the sedimentation region. Based on the geometric dimension variation of standard precipitates at different depths, the conversion factor between depth and geometric dimension is determined; Using conversion factors and actual depth, the measured geometric dimensions are converted into the equivalent geometric dimensions when the sedimentation area is located at the surface of the liquid.

[0010] Furthermore, based on the deviation between the converted equivalent geometric dimensions of each sedimentation region and the standard particle size, the sedimentation quantification value is calculated, specifically including: Obtain the maximum equivalent geometric dimension among the converted equivalent geometric dimensions of each sedimentation region; Calculate the deviation between the maximum equivalent geometric size of each sedimentation region and the standard particle size, and average the deviations of all sedimentation regions to obtain the sedimentation quantification value.

[0011] Furthermore, based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size, a liquid surface state category determined by the deviation is generated, specifically including: Calculate the difference between the quantification value of the sedimentation and the target quantification value, and use it as the state deviation value; The corresponding liquid surface state category is determined based on the preset range in which the state deviation value is located.

[0012] Furthermore, the method also includes: The ratio of the precipitated quantification value to the target quantification value is determined as the pressure adjustment coefficient; Adjust the injection pressure according to the pressure adjustment coefficient.

[0013] Furthermore, the methods for establishing the mapping relationship between grayscale values, geometric dimensions, and depth specifically include: Standard precipitates were placed at different depths in the medium, and images were acquired at each depth. Extract the grayscale values ​​and measured geometric dimensions of standard precipitates at each depth from images at each depth; A three-dimensional space is constructed using depth, grayscale value, and the measured geometric dimensions of the standard precipitate as coordinate dimensions. Data points in the three-dimensional space are fitted to obtain the mapping relationship between grayscale value, geometric dimensions, and depth.

[0014] This invention proposes a vision-based liquid level status recognition system for barium salt feeding process, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the vision-based barium salt feeding process liquid level status recognition method.

[0015] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, by extracting the foam bursting area from the liquid surface image, the spatial distribution dispersion of the gray-scale maxima and the fitting deviation between the edge and the circle are used to accurately distinguish the foam area from the sedimentation area, effectively eliminating the interference of foam on visual monitoring. Furthermore, by establishing a mapping relationship between gray-scale values, geometric dimensions, and depth, the sedimentation geometric dimensions at different depths are uniformly converted into equivalent geometric dimensions below the liquid surface, eliminating the influence of solution refraction on sedimentation size measurement. Finally, the sedimentation quantification value is calculated based on the deviation between the equivalent geometric dimensions and the standard particle size, and a liquid surface state category is generated, realizing accurate identification of sedimentation agglomeration state, providing a reliable basis for closed-loop control of injection pressure, thereby improving the accuracy of barium sulfate product particle size control and production efficiency. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a vision-based method for recognizing the liquid level state during barium salt feeding, as provided in an embodiment of the present invention. Figure 2 This is a structural diagram of a vision-based barium salt feeding process liquid level state recognition system provided in one embodiment of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the vision-based barium salt feeding process liquid level state recognition method proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of the vision-based barium salt feeding process liquid level state recognition method provided by the present invention.

[0021] This invention provides a vision-based method and system for recognizing the liquid level state during barium salt feeding. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates a flowchart of a vision-based method for recognizing the liquid level state during barium salt feeding, according to an embodiment of the present invention. The method includes the following steps: S101. Obtain an image of the liquid level inside the reaction vessel.

[0022] In this embodiment, the barium salt-sulfuric acid synthesis process based on the "reverse addition" principle is adopted in the continuous production of precipitated barium sulfate, as detailed below: High-purity barium chloride and sulfuric acid solutions are used as raw materials. In a reactor equipped with a high-efficiency stirrer, a measured amount of sulfuric acid or dilute sulfuric acid is first added as a base liquid. Under precisely controlled speed, temperature, and stirring intensity, the barium chloride solution is added to the reactor via a metering pump in a spray manner, ensuring full contact and reaction with the dilute sulfuric acid to form barium sulfate precipitate. After the reaction is complete, multiple washings are performed to thoroughly remove reaction byproducts (such as HCl) and soluble impurities, ensuring high purity and application stability of the product. The filter cake is dried to remove moisture, yielding a primary powder. Since the drying process often leads to slight particle agglomeration, it needs to be deagglomerated using equipment such as an air jet mill and then precisely classified to obtain the required particle size distribution.

[0023] To achieve real-time monitoring of the liquid level, a high-definition industrial camera and auxiliary light source are installed at the observation position on the inner wall of the reactor. Liquid level images are acquired at fixed intervals; in this embodiment, the fixed interval is 10Hz, meaning 10 frames per second. The industrial camera and auxiliary light source are installed in close proximity to ensure uniform illumination. The reactor cross-section is a circle, and the industrial camera, auxiliary light source, and nozzle are located at the same radius of this circle, arranged in the following order along the stirring direction: industrial camera, auxiliary light source, nozzle. The camera's aspect ratio is 16:9, and its field of view covers the contact surface between barium chloride and dilute sulfuric acid, as well as a portion of the upstream area of ​​the rotating contact surface.

[0024] With the above settings, the industrial camera can capture real-time images of the dynamic evolution of barium sulfate precipitate at the liquid surface, providing raw image data for subsequent liquid surface status identification.

[0025] S102. Based on the projection position of the jet source on the liquid surface and the direction of liquid flow, extract the foam rupture area from the liquid surface image.

[0026] In this embodiment, based on the projection position of the jet source on the liquid surface and the direction of liquid flow, the foam bursting area is extracted from the liquid surface image, specifically including: Obtain the location of the jet source from the liquid surface image; A conical atomizing cone is constructed with the position of the jet source as the cone apex, the distance from the jet source to the liquid surface as the cone height, and the atomization angle as the cone apex angle. The projection position of the bottom surface of the atomizing cone onto the liquid surface is determined as the reaction contact surface. The area in front of the reaction contact surface along the direction of liquid flow is defined as the foam-dense area; By removing the reaction contact surface and the area of ​​dense foam from the liquid surface image, the area of ​​foam rupture is obtained.

[0027] For example, a stirring device is installed in the reaction vessel to ensure uniform mixing of the dilute sulfuric acid solution, thereby maintaining a stable concentration of the dilute sulfuric acid solution that comes into contact with the barium chloride solution sprayed into the reaction vessel. When the barium chloride solution enters the reaction vessel through the spray system, the atomization angle of the nozzle and the height of the nozzle above the liquid surface together define the reaction contact area between the barium chloride solution and the dilute sulfuric acid solution. Because there is a certain distance between the spray system and the dilute sulfuric acid liquid surface, and under agitation, the sprayed barium chloride droplets collide with the flowing dilute sulfuric acid, generating a large amount of foam on the liquid surface through impact and exothermic reaction.

[0028] After foam is formed, it rotates with the liquid surface. After one rotation cycle, the foam temperature drops, and the surface tension can no longer maintain the foam shape, causing it to naturally burst and dissipate. Therefore, in the acquired liquid surface images, there are areas where the foam has dissipated, which are suitable for identifying the sedimentation state.

[0029] Based on the above principle, this embodiment extracts the foam rupture region (i.e., the region where the surface tension can no longer maintain the foam shape, thus causing it to naturally rupture and dissipate) from the liquid surface image through the following steps: Obtain the location of the nozzle from the liquid surface image. Construct an atomizing cone with the nozzle location as the cone apex, the distance from the nozzle to the liquid surface as the cone height, and the atomization angle as the cone apex angle. Record the bottom surface of the atomizing cone as the reaction contact surface. This reaction contact surface is the area where barium chloride solution and dilute sulfuric acid solution react violently, producing a large amount of foam; it needs to be excluded from the analysis area.

[0030] The liquid surface image is cropped along the y-axis past the nozzle position. The area in the direction of liquid flow that is in front of the reaction contact face (i.e., the part where the y-axis is in the same direction as the rotation) is identified as the foam-dense area. The foam in this area has not yet burst and also needs to be excluded.

[0031] By subtracting the aforementioned reaction contact surface and dense foam area from the liquid surface image, the remaining portion is designated as the foam breakage area. In this area, the foam has naturally broken and dissipated, exposing the precipitate to the liquid surface, making it suitable for identifying and analyzing the precipitation state.

[0032] S103. Extract connected regions with gray values ​​higher than a preset threshold in the foam rupture area as regions to be identified. Obtain the spatial distribution dispersion of gray-scale maxima in each region to be identified, as well as the fitting deviation between the edge of each region to be identified and the circle. The spatial distribution dispersion is used to characterize the degree of aggregation of each gray-scale maxima relative to the center of the region to be identified, and the fitting deviation is used to characterize the degree of conformity between the edge of the region to be identified and the standard circle.

[0033] In this embodiment, the spatial distribution dispersion of gray-level maxima in each region to be identified, and the fitting deviation between the edge of each region to be identified and the circle are obtained, specifically including: Obtain the center of each region to be identified and the pixel position corresponding to the maximum grayscale value; A position coordinate system is constructed with the center of each region to be identified as the origin, and the coordinates of the pixel corresponding to each maximum gray value are obtained in the coordinate system. The local outlier factor algorithm is used to obtain the local outlier factor of each pixel in the coordinate system, and the average value of all local outliers is calculated as the spatial distribution dispersion of the region to be identified. The radius is determined by the area of ​​each region to be identified, and a circle is constructed based on the radius. The Euclidean distance between each edge pixel of the region to be identified and the nearest point on the circle is calculated, and the average value of each Euclidean distance is determined as the fitting bias.

[0034] For example, in areas where foam has broken, there may still be a small amount of foam that has not completely broken, and its visual characteristics are similar to barium sulfate precipitate, requiring further differentiation.

[0035] Barium sulfate precipitate is a white powdery crystal, and the precipitate formed in solution is white granular. When the injection pressure is too high, a large amount of barium chloride comes into contact with dilute sulfuric acid to form barium sulfate precipitate. If the agitation speed of the dilute sulfuric acid solution is insufficient to remove the precipitate, the barium sulfate precipitate will agglomerate, forming large particles. By visually monitoring the precipitation in the foam rupture area, the precipitation state of barium sulfate can be analyzed, and the injection pressure can be adjusted accordingly.

[0036] Because foam has a curved and smooth surface, under the illumination of an auxiliary light source, the relative positions of the high-brightness areas on the foam surface are fixed, and the edges of the foam area are smooth. In contrast, the high-brightness areas in the sedimentation area are randomly positioned, and the edges are irregular. Based on this characteristic, this embodiment distinguishes between the foam area and the sedimentation area through the following steps: Threshold segmentation identifies regions with grayscale values ​​higher than a preset threshold as areas to be identified. Preferably, this preset threshold is an adaptive threshold, dynamically determined based on the grayscale distribution of the current liquid surface image.

[0037] Obtain the center and the location of the maximum grayscale value for each region to be identified. Construct a coordinate system with the center of each region as the origin, and project the location of the maximum grayscale value onto this coordinate system. Use the Local Outlier Factor (LOF) algorithm to obtain the local outlier factor for each data point in the coordinate system. The larger the outlier factor value, the more dispersed the surrounding data points are, and the more likely it is a random bright spot generated by flocculent sedimentation; conversely, the smaller the outlier factor, the denser the surrounding data points are, and the more likely it is a bright spot of foam. Calculate the average of all local outliers as the spatial distribution dispersion of the region to be identified.

[0038] Since the region to be identified is an irregular connected region, not a standard circle, its center cannot be directly obtained. In this embodiment, the centroid of the region to be identified is used as its center. The centroid is calculated by taking a weighted average of the coordinates of all pixels within the region to be identified, using the grayscale value of each pixel as the weight, to calculate the centroid coordinates.

[0039] In another embodiment, the geometric center of the region to be identified is used as its center. The geometric center is the center point of the bounding rectangle of the region to be identified, or the center of the smallest circumcircle of the region's boundary. Specifically, the minimum and maximum coordinate values ​​of the region to be identified in the horizontal and vertical directions are obtained, and the midpoint coordinates are calculated as the geometric center.

[0040] Based on the area of ​​each region to be identified, the radius is inversely calculated and a standard circle is constructed. The Euclidean distance between each edge pixel of the region to be identified and the nearest point on the standard circle is calculated. The average value of all Euclidean distances is determined as the fitting deviation of the region to be identified. The smaller the fitting deviation, the closer the edge of the region to be identified is to a circle, and the more likely it is to be a foam region; the larger the fitting deviation, the more irregular the edge is, and the more likely it is to be a sedimentation region.

[0041] Therefore, based on the above analysis, the first Fitting deviation of each region to be identified The calculation formula can be: ; in, For the first The number of edge pixels in the region to be identified Indicates by the first After constructing a circle from the radius of the area of ​​the region to be identified, the first region is identified. The Euclidean distance between the k-th edge pixel of the region to be identified and the nearest point on the circle; the smaller the Euclidean distance, the more significant the difference between the k-th edge pixel and the nearest point on the circle. The edges of the region to be identified tend to be closer to a circle.

[0042] S104. Calculate the confidence level that the region to be identified belongs to the sedimentation region based on the product of the spatial distribution dispersion and the fitting deviation.

[0043] In this embodiment, the confidence level that the region to be identified belongs to the sedimentation region is calculated based on the product of the spatial distribution dispersion and the fitting deviation, specifically including: The product of spatial distribution dispersion and fitting deviation is calculated, and the negative exponential function value with the natural constant as the base and the product as the exponent is determined as the confidence level that the region to be identified belongs to the sedimentation region.

[0044] For example, the first Confidence that the region to be identified belongs to the sedimentation region The calculation formula can be: ; in, Indicates the first Spatial distribution dispersion of the regions to be identified.

[0045] S105. Select sedimentation areas from the areas to be identified based on the preset confidence threshold.

[0046] In this embodiment, the preset confidence threshold can be determined by the following method: Multiple sets of sample images containing known foam and sediment regions were collected, and the confidence score of each region to be identified was calculated using the aforementioned method. The confidence score distributions of foam and sediment regions were statistically analyzed, and the confidence score value that best distinguishes the two types of regions was selected as the preset confidence threshold.

[0047] Alternatively, calculate the mean and standard deviation of the confidence scores for all regions to be identified in the current liquid surface image, set the confidence threshold to the mean plus one standard deviation, or adjust it dynamically according to actual needs.

[0048] Alternatively, a large number of sample images labeled with foam and sediment regions can be collected, and feature vectors (including spatial distribution dispersion, fitting bias, etc.) of the regions to be identified in each sample image can be extracted. A classifier (such as support vector machine, decision tree, or neural network) can be used for training, and the classifier can automatically output the classification boundary. The confidence value corresponding to the boundary is the preset confidence threshold.

[0049] Experiments have shown that a good differentiation effect can be obtained when the confidence threshold is set between 0.5 and 0.7. In this embodiment, the preset confidence threshold is preferably set to 0.6. That is, when the confidence of the area to be identified is greater than 0.6, it is determined to be a sediment area; when the confidence of the area to be identified is less than or equal to 0.6, it is determined to be a foam area.

[0050] In actual production, any of the above methods can be selected to determine the preset confidence threshold; no specific restrictions are imposed here.

[0051] S106. Obtain the average gray value and measured geometric dimensions of each sedimentation area. Determine the actual depth of the sedimentation area below the liquid surface based on the pre-established mapping relationship between gray value, geometric dimensions and depth. Convert the measured geometric dimensions of the sedimentation area to the equivalent geometric dimensions when the sedimentation area is located on the surface of the liquid based on the actual depth.

[0052] In this embodiment, the actual depth of the sedimentation region below the liquid surface is determined based on a pre-established mapping relationship between grayscale values, geometric dimensions, and depth. Then, the measured geometric dimensions are converted to the equivalent geometric dimensions of the sedimentation region when it is located at the liquid surface, specifically including: Based on the mapping relationship, determine the actual depth of the sedimentation region below the liquid surface, which corresponds to the average gray value and measured geometric dimensions of the sedimentation region. Based on the geometric dimension variation of standard precipitates at different depths, the conversion factor between depth and geometric dimension is determined; Using conversion factors and actual depth, the measured geometric dimensions are converted into the equivalent geometric dimensions when the sedimentation area is located at the surface of the liquid.

[0053] The methods for establishing the mapping relationship between grayscale values, geometric dimensions, and depth specifically include: Standard precipitates were placed at different depths in the medium, and images were acquired at each depth. Extract the grayscale values ​​and measured geometric dimensions of standard precipitates at each depth from images at each depth; A three-dimensional space is constructed using depth, grayscale value, and the measured geometric dimensions of the standard precipitate as coordinate dimensions. Data points in the three-dimensional space are fitted to obtain the mapping relationship between grayscale value, geometric dimensions, and depth.

[0054] For example, since the liquid surface image and the obtained sedimentation area acquired in this application are both obtained after the foam generated at the reaction contact surface has rotated once and naturally broken, the images are then captured. At this time, the generated sediment gradually sinks under the influence of gravity. The light generated by the auxiliary light source is reflected by the sediment and enters the camera. During its propagation, the light needs to pass through the dilute sulfuric acid solution and undergoes refraction. This refraction effect amplifies the optical characteristic of "nearer objects appearing larger and farther objects appearing smaller," meaning that sediments of different sizes are located at different depths in the solution. This results in the apparent size of the area captured when larger sediment particles are located at a deeper position being similar to, or even smaller than, the area of ​​smaller sediment particles on the surface.

[0055] In addition to large flocculent or lumpy precipitates, the solution also contains a large number of unaggregated fine precipitates. During stirring, the structurally unstable flocculent or lumpy precipitates may break down and redisperse into fine precipitates. These precipitates, due to their density being similar to that of the solution, are distributed in a suspended state within the solution. Simultaneously, Rayleigh scattering occurs when light strikes these suspended particles; that is, when the size of the suspended particles is close to or smaller than the wavelength of light, light is scattered. Therefore, the deeper the flocculent precipitate is located, the weaker the light it receives and reflects. Based on this characteristic, even if the measured geometric dimensions of the precipitate at depth are small, its depth can be reflected by its grayscale value (brightness / darkness), thus enabling the reconstruction of the geometric dimensions of large particles of precipitate at depth.

[0056] For a standard precipitate, its apparent geometric dimensions and corresponding grayscale values ​​were measured at different depths, resulting in multiple sets of depth-grayscale-measured geometric dimension data points. Using depth and grayscale values ​​as inputs and measured geometric dimensions as outputs, a nonlinear least squares method was employed to fit the data points to a surface, establishing a mapping model. For the actual detected precipitation region, its average grayscale value and measured geometric dimensions were substituted into the inverse function of the mapping model to obtain the actual depth of the precipitation region below the liquid surface. Subsequently, based on the geometric dimension variation pattern of the standard precipitate at different depths, a conversion factor between depth and geometric dimensions was determined. This conversion factor was then used to convert the measured geometric dimensions into equivalent geometric dimensions below the liquid surface.

[0057] Because light from the auxiliary light source refracts when passing through the dilute sulfuric acid solution, and the path length of the light varies depending on the depth of the precipitate, the apparent geometric dimensions of the precipitate area captured by the camera change with depth. Specifically, for precipitates located at deeper depths, the light travels a longer path through the solution, resulting in a more pronounced refraction effect and a compressed apparent geometric dimension; for precipitates located at shallower depths, the apparent geometric dimensions are closer to the true dimensions. Through the aforementioned mapping relationship and conversion factors, the measured geometric dimensions at different depths can be uniformly restored to the equivalent geometric dimensions below the liquid surface, thus enabling a fair comparison between different precipitates.

[0058] S107. Calculate the precipitation quantification value based on the deviation between the equivalent geometric size of each precipitation zone and the standard particle size.

[0059] In this embodiment, the precipitation quantification value is calculated based on the deviation between the equivalent geometric dimensions of each precipitation region after conversion and the standard particle size. Specifically, this includes: Obtain the maximum equivalent geometric dimension among the converted equivalent geometric dimensions of each sedimentation region; Calculate the deviation between the maximum equivalent geometric size of each sedimentation region and the standard particle size, and average the deviations of all sedimentation regions to obtain the sedimentation quantification value.

[0060] For example, for each sedimentation region, the maximum equivalent geometric dimension among its converted equivalent geometric dimensions is obtained. The equivalent geometric dimension can be a geometric measure such as equivalent diameter, equivalent area, or equivalent perimeter. In this embodiment, the equivalent diameter is preferably used as the geometric dimension measurement standard.

[0061] Calculate the deviation between the maximum equivalent geometric size of each sedimentation region and the standard particle size required for the finished product. Sum the deviations of all sedimentation regions and divide by the total number of sedimentation regions to obtain the sedimentation quantification value for the m-th liquid surface image. The calculation formula can be: ; in, This represents the maximum diameter of the nth sedimentation region in the mth liquid surface image. This represents the total number of sedimentation regions in the m-th liquid surface image. This indicates the standard particle size required for the finished product.

[0062] The precipitation quantification value is used to characterize the overall degree of precipitation agglomeration below the current liquid surface. When the precipitation quantification value is greater than 1, it indicates that the current precipitation is too large compared to the standard particle size, and there is a tendency for excessive agglomeration; when the precipitation quantification value is less than 1, it indicates that the current precipitation is too small compared to the standard particle size, and there is a tendency for insufficient agglomeration; when the precipitation quantification value is close to 1, it indicates that the current precipitation state is relatively close to the standard particle size, and the degree of precipitation agglomeration is moderate.

[0063] S108. Generate a liquid surface state category based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size.

[0064] In this embodiment, based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size, a liquid surface state category determined by the deviation is generated, specifically including: Calculate the difference between the quantification value of the sedimentation and the target quantification value, and use it as the state deviation value; The corresponding liquid surface state category is determined based on the preset range in which the state deviation value is located.

[0065] The method also includes: The ratio of the precipitated quantification value to the target quantification value is determined as the pressure adjustment coefficient; Adjust the injection pressure according to the pressure adjustment coefficient.

[0066] For example, in this embodiment, based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size, a liquid surface state category determined by the deviation is generated, and the injection pressure is adjusted accordingly, specifically including: The difference between the precipitation quantification value and the target quantification value is calculated as the state deviation value. The target quantification value is obtained based on the standard particle size using the same calculation method as the precipitation quantification value. Specifically, the required standard particle size A of the finished product is substituted into the calculation formula for the precipitation quantification value to obtain the target quantification value.

[0067] Determine the liquid level state category. Based on the preset range in which the state deviation value falls, determine the corresponding liquid level state category. For example: When the precipitation quantification value is greater than the target quantification value, that is, when the state deviation value is positive, it indicates that the current precipitation agglomeration degree is higher than the standard, and it is judged as the excessive precipitation agglomeration category. When the precipitation quantification value is less than the target quantification value, that is, when the state deviation value is negative, it indicates that the current precipitation agglomeration degree is lower than the standard and is judged as insufficient precipitation agglomeration. When the precipitation quantification value is equal to the target quantification value or the deviation is within the preset tolerance range, it is judged as a normal precipitation category.

[0068] The method also includes the step of adjusting the injection pressure based on the liquid level identification result, as follows: The ratio of the precipitated quantification value to the target quantification value is determined as the pressure adjustment coefficient. The injection pressure is adjusted according to the pressure adjustment coefficient. Specifically: When the pressure adjustment coefficient is greater than 1, it indicates that the precipitation and agglomeration are excessive, and the spray pressure needs to be reduced to reduce the spray volume of barium chloride solution, thereby reducing the formation rate of barium sulfate and inhibiting the agglomeration phenomenon. When the pressure adjustment coefficient is less than 1, it indicates that the precipitation is insufficient and the spray pressure needs to be increased to increase the spray volume of barium chloride solution and promote the nucleation and growth of barium sulfate crystals. When the pressure adjustment coefficient is equal to 1 or within the preset tolerance range, it indicates that the current sedimentation state meets the requirements, and the current injection pressure is maintained unchanged.

[0069] The system adjusts the nozzle injection pressure in real time based on the calculated required pressure, achieving closed-loop control of barium sulfate precipitate particle size. Through this dynamic pressure control based on visual feedback, the nucleation and growth stages of barium sulfate particles can be effectively intervened, the crystal development environment optimized, and ultimately, controllable particle size of the product achieved.

[0070] In summary, the embodiments of the present invention can accurately distinguish between foam areas and sediment areas in liquid surface images, eliminating the interference of foam on visual monitoring; by correcting the sediment size through a depth-grayscale-size mapping model, the influence of solution refraction on measurement accuracy is eliminated; finally, based on the corrected sediment quantification value, a liquid surface state category is generated and the injection pressure is adjusted, thereby achieving accurate identification and closed-loop control of sediment agglomeration state, thus improving the accuracy of barium sulfate product particle size control and production efficiency.

[0071] This invention also proposes a vision-based liquid level recognition system for the barium salt feeding process; please refer to [link / reference]. Figure 2 The diagram shows a structure of a vision-based barium salt feeding process liquid level status recognition system provided in an embodiment of the present invention. The system includes: a data acquisition module 101, a data processing module 102, and a status recognition module 103.

[0072] Data acquisition module 101 is used to acquire images of the liquid level inside the reaction vessel; Data processing module 102 is used to extract the foam bursting area from the liquid surface image based on the projection position of the jet source on the liquid surface and the direction of liquid surface flow. The connected regions with gray values ​​higher than a preset threshold in the foam rupture area are extracted as the regions to be identified. The spatial distribution dispersion of gray-scale maxima in each region to be identified and the fitting deviation between the edge of each region to be identified and the circle are obtained. The spatial distribution dispersion is used to characterize the degree of aggregation of each gray-scale maxima relative to the center of the region to be identified, and the fitting deviation is used to characterize the degree of conformity between the edge of the region to be identified and the standard circle. The confidence level that the region to be identified belongs to the sedimentation region is calculated based on the product of the spatial distribution dispersion and the fitting bias. Based on a preset confidence threshold, sedimentation areas are filtered out from the areas to be identified; The average gray value and measured geometric dimensions of each sedimentation region are obtained. Based on the pre-established mapping relationship between gray value, geometric dimensions and depth, the actual depth of the sedimentation region below the liquid surface is determined. Based on the actual depth, the measured geometric dimensions are converted into the equivalent geometric dimensions of the sedimentation region when it is located on the surface of the liquid. The sedimentation quantification value is calculated based on the deviation between the equivalent geometric size of each sedimentation region after conversion and the standard particle size. The state recognition module 103 is used to generate a liquid surface state category determined based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size.

[0073] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the vision-based barium salt feeding process liquid level state recognition system and the vision-based barium salt feeding process liquid level state recognition method embodiments provided in the above embodiments belong to the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0074] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0075] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A vision-based method for recognizing the liquid level state during barium salt feeding, characterized in that, include: Acquire images of the liquid level inside the reaction vessel; Based on the projection position of the jet source on the liquid surface and the direction of liquid flow, the foam bursting area is extracted from the liquid surface image; The connected regions with gray values ​​higher than a preset threshold in the foam rupture area are extracted as the regions to be identified. The spatial distribution dispersion of gray-scale maxima in each region to be identified and the fitting deviation between the edge of each region to be identified and the circle are obtained. The spatial distribution dispersion is used to characterize the degree of aggregation of each gray-scale maxima relative to the center of the region to be identified, and the fitting deviation is used to characterize the degree of conformity between the edge of the region to be identified and the standard circle. The confidence level that the region to be identified belongs to the sedimentation region is calculated based on the product of the spatial distribution dispersion and the fitting bias. Based on a preset confidence threshold, sedimentation areas are filtered out from the areas to be identified; The average gray value and measured geometric dimensions of each sedimentation region are obtained. Based on the pre-established mapping relationship between gray value, geometric dimensions and depth, the actual depth of the sedimentation region below the liquid surface is determined. Based on the actual depth, the measured geometric dimensions are converted into the equivalent geometric dimensions of the sedimentation region when it is located on the surface of the liquid. The sedimentation quantification value is calculated based on the deviation between the equivalent geometric size of each sedimentation region after conversion and the standard particle size. Based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size, a liquid surface state category determined by the deviation is generated.

2. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The step of extracting the foam bursting area from the liquid surface image based on the projection position of the jet source on the liquid surface and the direction of liquid flow specifically includes: Obtain the location of the jet source from the liquid surface image; A conical atomizing cone is constructed with the position of the jet source as the cone apex, the distance from the jet source to the liquid surface as the cone height, and the atomization angle as the cone apex angle. The projection position of the bottom surface of the atomizing cone onto the liquid surface is determined as the reaction contact surface. The area in front of the reaction contact surface along the direction of liquid flow is defined as the foam-dense area; By removing the reaction contact surface and the area of ​​dense foam from the liquid surface image, the area of ​​foam rupture is obtained.

3. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The process of obtaining the spatial distribution dispersion of gray-level maxima in each region to be identified, and the fitting deviation between the edge of each region to be identified and the circle, specifically includes: Obtain the center of each region to be identified and the pixel position corresponding to the maximum grayscale value; A position coordinate system is constructed with the center of each region to be identified as the origin, and the coordinates of the pixel corresponding to each maximum gray value are obtained in the coordinate system. The local outlier factor algorithm is used to obtain the local outlier factor of each pixel in the coordinate system, and the average value of all local outliers is calculated as the spatial distribution dispersion of the region to be identified. The radius is determined by the area of ​​each region to be identified, and a circle is constructed based on the radius. The Euclidean distance between each edge pixel of the region to be identified and the nearest point on the circle is calculated, and the average value of each Euclidean distance is determined as the fitting bias.

4. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The step of calculating the confidence level that the region to be identified belongs to the sedimentation region based on the product of the spatial distribution dispersion and the fitting deviation specifically includes: The product of spatial distribution dispersion and fitting deviation is calculated, and the negative exponential function value with the natural constant as the base and the product as the exponent is determined as the confidence level that the region to be identified belongs to the sedimentation region.

5. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The process of determining the actual depth of the sedimentation region below the liquid surface based on a pre-established mapping relationship between grayscale values, geometric dimensions, and depth, and then converting the measured geometric dimensions of the sedimentation region to the equivalent geometric dimensions when it is located at the liquid surface based on this actual depth, specifically includes: Based on the mapping relationship, determine the actual depth of the sedimentation region below the liquid surface, which corresponds to the average gray value and measured geometric dimensions of the sedimentation region. Based on the geometric dimension variation of standard precipitates at different depths, the conversion factor between depth and geometric dimension is determined; Using conversion factors and actual depth, the measured geometric dimensions are converted into the equivalent geometric dimensions when the sedimentation area is located at the surface of the liquid.

6. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The step of calculating the sedimentation quantification value based on the deviation between the equivalent geometric dimensions of each sedimentation region and the standard particle size specifically includes: Obtain the maximum equivalent geometric dimension among the converted equivalent geometric dimensions of each sedimentation region; Calculate the deviation between the maximum equivalent geometric size of each sedimentation region and the standard particle size, and average the deviations of all sedimentation regions to obtain the sedimentation quantification value.

7. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The step of generating a liquid surface state category based on the deviation between the sedimentation quantification value and the target quantification value calculated based on the standard particle size specifically includes: Calculate the difference between the quantification value of the sedimentation and the target quantification value, and use it as the state deviation value; The corresponding liquid surface state category is determined based on the preset range in which the state deviation value is located.

8. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The method further includes: The ratio of the precipitated quantification value to the target quantification value is determined as the pressure adjustment coefficient; Adjust the injection pressure according to the pressure adjustment coefficient.

9. The vision-based method for recognizing the liquid level state during barium salt feeding as described in claim 1, characterized in that, The method for establishing the mapping relationship between grayscale values, geometric dimensions, and depth specifically includes: Standard precipitates were placed at different depths in the medium, and images were acquired at each depth. Extract the grayscale values ​​and measured geometric dimensions of standard precipitates at each depth from images at each depth; A three-dimensional space is constructed using depth, gray value, and the measured geometric dimensions of the standard precipitate as coordinate dimensions. The data points in the three-dimensional space are fitted using the nonlinear least squares method to obtain the mapping relationship between gray value, geometric dimensions, and depth.

10. A vision-based barium salt feeding process liquid level status recognition system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of the vision-based barium salt feeding process liquid level state recognition method as described in any one of claims 1-9.

Images