Boiler scale detection method and device based on multi-modal image

Abstract

The invention relates to the technical field of image detection, in particular to a boiler scale detection method and device based on a multi-modal image, and the method comprises the steps: carrying out the panoramic scanning of the inner wall of a boiler through a dual-spectrum imaging system at a detection temperature, and synchronously collecting a visible light image and an infrared thermal imaging image; based on the scale chromaticity difference value in the visible light image, scale types are judged, and a scale classification graph is generated; based on a standard infrared thermal imaging image, obtained at the detection temperature, of the scale-free boiler, the infrared thermal imaging image is converted into a temperature anomaly distribution diagram with temperature anomaly blocks; inputting the temperature anomaly distribution diagram into a deep neural network to output a scale thickness distribution diagram, and aligning the scale classification diagram with the scale thickness distribution diagram through a spatial registration technology to ensure that pixels at the same spatial position contain scale types and thickness values at the same time; and generating a scale feature fusion graph of scale types presenting different color areas and corresponding thicknesses, thereby realizing accurate detection and quantification of the boiler scale.

Description

Technical Field

[0001] This application relates to the technical field of image detection, and particularly to a method and device for detecting boiler scale based on multimodal images. Background Art

[0002] In the process system of a natural gas station, as a key heat source device, the boiler continuously supplies high-temperature hot water to the heat exchanger. This hot water transfers heat energy to the natural gas flow through the heat exchanger wall to compensate for the temperature loss of the natural gas caused by subsequent sudden pressure drops. This heat transfer process is crucial for maintaining the natural gas transportation temperature and preventing hydrate blockages. However, scale accumulates on the heating surface of the boiler during long-term operation. The scale layer hinders the normal diffusion of heat to the water body and reduces the heat transfer efficiency.

[0003] In the existing boiler scale detection technology, the temperature distribution on the surface of the boiler during operation is captured by a single infrared thermal imaging method for scale discrimination. Its core principle is based on the physical characteristics of heat conduction: due to high thermal conductivity, the metal matrix of the boiler quickly transfers heat to the internal working medium, presenting a low-temperature dark tone in the thermal imaging; while the scale layer attached to the metal surface hinders heat transfer due to the thermal resistance effect, resulting in local heat accumulation to form a high-temperature area, which shows as a bright patch in the infrared thermal imaging image. In engineering practice, the empirical criterion of "the brighter, the thicker" is usually adopted, that is, the deposition degree of scale is qualitatively judged by the brightness value of the patch in the infrared thermal imaging image, and the central area of the bright spot is interpreted as the thickest part of the scale. However, this single infrared thermal imaging method can only roughly indicate the abnormal area and is difficult to support accurate boiler maintenance decisions. Summary of the Invention

[0004] In view of the above problems, this application provides a method and device for detecting boiler scale based on multimodal images, which realizes the accurate detection and quantification of boiler scale through the image features of visible light images and infrared thermal imaging and the heat conduction theory.

[0005] To achieve the purpose of this application, the following technical solutions are provided: In a first aspect, this application provides a method for detecting boiler scale based on multimodal images, including: At the detection temperature, a panoramic scan of the inner wall of the boiler is performed by a dual-spectrum imaging system, and visible light images and infrared thermal imaging images are synchronously collected; Perform polarization filtering on the visible light image to suppress the mirror reflection of the boiler metal substrate, retain the diffuse reflection area on the scale surface, then perform LAB color space conversion on the diffuse reflection area, and judge the scale type and generate a scale classification map based on the chromaticity differences in the AB chromaticity channels of different types of scale; Based on the standard infrared thermal imaging image of the scale-free boiler obtained at the detected temperature, convert the infrared thermal imaging image into a temperature anomaly distribution map presenting temperature anomaly blocks, where the temperature anomaly blocks characterize the location and scope of scale deposition; Input the temperature anomaly distribution map into a deep neural network, extract multi-scale features of the temperature anomaly distribution image through an encoder downsampling structure to generate the law of thermal resistance change, and gradually convert the law of thermal resistance change learned by the encoder into a spatially distributed thickness value through a decoder upsampling structure, and output a scale thickness distribution map, where each pixel in the scale thickness distribution map stores a floating-point thickness value; Align the scale classification map with the scale thickness distribution map through a spatial registration technique to ensure that pixels at the same spatial position simultaneously contain scale type and thickness values, and generate a scale feature fusion map presenting different color regions and their corresponding scale thicknesses.

[0006] In a possible implementation, perform polarization filtering on the visible light image to suppress the specular reflection of the boiler metal substrate and retain the diffuse reflection region on the scale surface, and then perform LAB color space conversion on the diffuse reflection region. Based on the chromaticity differences of different types of scales in the AB chromaticity channels, determine the scale type and generate a scale classification map, which further includes: By analyzing the intensity changes of the visible light image at different polarization angles, distinguish the specular reflection region generated by the boiler metal substrate and the diffuse reflection region generated on the scale surface, and suppress the specular reflection information and retain the diffuse reflection information; Enhance the difference between the specular reflection region and the diffuse reflection region by adjusting the local contrast and global gray distribution to obtain a visible light image retaining the diffuse reflection region; Perform color space conversion on the visible light image retaining the diffuse reflection region to generate an LAB image including the A chromaticity channel and the B chromaticity channel. Determine the scale type based on the aggregation regions of the scale in the A chromaticity channel and the B chromaticity channel, and generate the scale classification map, where each pixel in the scale classification map stores a scale type label value, and the scale types include rust scale, carbonate scale, and silicate scale with different aggregation regions.

[0007] In a possible implementation, based on the standard infrared thermal imaging image of the scale-free boiler, convert the infrared thermal imaging image into an infrared thermal imaging image presenting temperature anomaly blocks, where the temperature anomaly blocks characterize the location and scope of scale deposition, and further include; At the same detected temperature, obtain the standard infrared thermal imaging image of a scale-free boiler with the same model and specification, and align the standard infrared thermal imaging image with the infrared thermal imaging image; Traverse the aligned infrared thermal imaging image and the standard infrared thermal imaging image using a sliding window, and calculate the mean difference and fluctuation difference of the thermal radiation intensity within the window. The mean difference of the thermal radiation intensity is the difference between the window mean in the infrared thermal imaging image and the window mean in the standard infrared thermal imaging image, and the fluctuation difference of the thermal radiation is the absolute value of the difference between the window standard deviation in the infrared thermal imaging image and the window standard deviation in the standard infrared thermal imaging image; Let Δμ w > 2°C and the corresponding Δσ w < 0.5°C for the pixels as candidate seed points, cluster the continuously adjacent candidate seed points, and take the geometric center of each cluster as the growth starting point; If the neighborhood pixels of the growth starting point satisfy Δμ n > 1.5°C and |Δσ n - Δσ o | < 0.3°C, then incorporate the neighborhood pixels into the scale deposit area to obtain a temperature anomaly distribution map showing temperature anomaly blocks, where the neighborhood pixels include multiple pixel points around the growth starting point, Δμ represents the mean difference of the thermal radiation intensity, Δσ represents the fluctuation difference of the thermal radiation, Δσ o represents the temperature standard deviation of the area where the growth starting point is located, w represents the pixel points within the window, and n represents the pixel points within the neighborhood.

[0008] In a possible implementation, input the temperature anomaly distribution map into a deep neural network, extract multi-scale features of the temperature anomaly distribution map through an encoder downsampling structure to generate the law of thermal resistance change, and gradually transform the law of thermal resistance change learned by the encoder into a spatially distributed thickness value through a decoder upsampling structure to output a scale deposit thickness distribution map. Further including: Gradually extract the multi-scale heat conduction features of the temperature anomaly image through the four-layer downsampling structure of the encoder, and capture the law of thermal resistance change in different regions through convolution operations in each layer, and cooperate with the max-pooling operation to increase the number of feature channels while reducing the spatial resolution. The deformable convolution module is introduced in the third layer to fit the heat flow distortion features caused by the sudden change of the thermal conductivity coefficient by adaptively adjusting the sampling position; Gradually restore the spatial resolution through the symmetric four-layer upsampling structure of the decoder, and the feature map output by each layer is skip-connected with the same-scale features of the encoder to fuse the local morphology of the scale deposit and the global thermal resistance semantic features to generate fused features; Compress the fused features into a single channel through a 1×1 convolution of the network output layer, and linearly map them to a thickness value of 0.1 - 5.0 mm after normalization using the Sigmoid function to output a pixel-level scale deposit thickness distribution map.

[0009] In a possible implementation, the scale classification map is aligned with the scale thickness distribution map through spatial registration technology to ensure that pixels at the same spatial position contain both the scale type and the thickness value, generating a scale feature fusion map presenting different colored regions with the corresponding scale type and thickness, further including: Performing sub-pixel alignment on the scale classification map obtained by analyzing the visible light image and the thickness distribution map generated based on the infrared thermal imaging to form a composite data matrix, where in the matrix, each spatial position corresponds to a binary group including a type label and a thickness value, and at the same time, the coverage area percentage and average thickness of each type of scale are statistically analyzed.

[0010] In some embodiments, the above scale detection method is also applicable to devices that rely on heat transfer processes, such as distillation tanks, distillers for seawater desalination, condensers in power plant refrigeration equipment, etc., where scale causes observable heat conduction anomalies.

[0011] In a possible implementation, the method further includes: Dividing the scale in the scale thickness distribution map into thin-layer scale, medium-thick scale, and thick scale according to the thickness value, and establishing a piecewise linear mapping function between the scale thickness and the nozzle pressure; According to the piecewise linear mapping function, calculating the nozzle pressure corresponding to the thickest thickness in the scale area of the scale thickness distribution map; When the area of the scale area is less than the threshold area, setting the nozzle to be directed towards the thickest point of the scale area for spraying; When the area of the scale area is greater than the threshold area, taking the centroid of the scale area as the reference coordinate, calculating the offset vector in combination with the coordinate of the thickest point of the scale area, and setting the nozzle to perform a spiral motion with the thickest point as the center and the offset vector as the diameter to spray the scale area.

[0012] In a second aspect, the present application provides a boiler scale detection device based on multi-modal images, including: An image acquisition module, configured to panoramically scan the inner wall of the boiler through a dual-spectrum imaging system at the detection temperature, and synchronously acquire visible light images and infrared thermal imaging images; A first image processing module, configured to perform polarization filtering processing on the visible light image to suppress the mirror reflection of the boiler metal substrate, retain the diffuse reflection area on the scale surface, then perform LAB color space conversion on the diffuse reflection area, and judge the scale type and generate a scale classification map based on the chromaticity difference of different types of scales in the AB chromaticity channels; A second image processing module, configured to convert the standard infrared thermal imaging image of the scale-free boiler acquired at the detected temperature into a temperature anomaly distribution map presenting temperature anomaly blocks, where the temperature anomaly blocks characterize the position and range of scale deposition; An analysis module, configured to input the temperature anomaly distribution map into a deep neural network, extract multi-scale features of the temperature anomaly distribution map through an encoder downsampling structure, generate a law of thermal resistance change, and gradually convert the law of thermal resistance change learned by the encoder into a spatially distributed thickness value through a decoder upsampling structure, and output a scale thickness distribution map, where each pixel in the scale thickness distribution map stores a floating-point thickness value; A generation module, configured to align the scale classification map with the scale thickness distribution map through a spatial registration technique, ensure that pixels at the same spatial position simultaneously include the scale type and thickness value, and generate a scale feature fusion map presenting different color regions and their corresponding scale thicknesses.

[0013] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is run by a processor, it executes the steps of the method for detecting boiler scale based on multi-modal images.

[0014] The embodiments of the present application provide a method and device for detecting boiler scale based on multi-modal images. Through multi-modal image fusion technology, the combination of image feature analysis and heat conduction modeling is realized, and accurate detection and quantification of boiler scale are achieved. Specifically, visible light polarization imaging effectively strips the interference of the metal substrate, and the chromaticity feature of scale diffuse reflection is used to identify the scale type; at the same time, infrared thermal imaging is combined with a deep learning network to quantify the temperature anomaly situation into the scale thickness. The spatial registration and fusion of the two-modal image data can not only distinguish the chemical composition of the scale, but also quantify the deposition thickness, improve the recognition reliability of mixed scale, and provide a comprehensive basis for formulating an accurate cleaning strategy. Description of the Drawings

[0015] The drawings are used to provide a further understanding of the present application, and constitute a part of the specification. They are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application; Figure 1 It is a flowchart of the method for detecting boiler scale based on multi-modal images provided by the embodiments of the present application. Detailed Embodiments

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in this application. Obviously, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the protection scope of this application.

[0017] The terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of this application, unless otherwise stated, the meaning of "plurality" is two or more.

[0018] To solve the above technical problems, the following technical solutions and corresponding embodiments are proposed in this application.

[0019] Figure 1 The flowchart of the boiler scale detection method based on multimodal images provided for the embodiments of this application is as Figure 1 shown. The embodiments of this application provide a boiler scale detection method based on multimodal images, including the following steps: S100 At the detection temperature, panoramically scan the inner wall of the boiler through a dual-spectrum imaging system, and synchronously collect visible light images and infrared thermal imaging images.

[0020] Among them, the inner wall of the boiler includes the inner wall surface of the boiler drum and the inner wall surfaces of the upper and lower heads. The dual-spectrum imaging system includes a visible light imaging module and an infrared thermal imaging module coaxially installed. The embodiments of this application realize scale detection by synchronously collecting multimodal images of visible light images and infrared thermal images. Among them, the morphological and color features of the scale are obtained through the visible light images, and the scale position can be accurately identified and its thickness distribution can be evaluated through the temperature distribution features in the infrared thermal imaging images.

[0021] The visible light image has a high spatial resolution and can clearly show the microscopic morphology and color features of the scale. However, it is greatly interfered by surface reflection, such as the reflection interference of the boiler metal substrate. Therefore, it is necessary to use a polarization filter to collect visible light images of the inner wall of the boiler at different rotation angles.

[0022] The infrared thermal imaging module selects a thermal imager with a temperature resolution of not less than 0.1 °C. Its detection principle is based on the significant thermal conductivity difference between the water scale and the metal substrate: the thermal conductivity of the water scale is usually only 1 / 50 - 1 / 30 of that of the boiler metal substrate. For every 1 mm thick water scale formed in an industrial boiler, the thermal efficiency will be reduced by about 5%. This thermal conductivity difference causes detectable temperature differences between the metal substrate and the water scale, as well as between water scale regions with different thicknesses. To ensure effective detection of the temperature difference, the boiler needs to be in a heating state to generate sufficient heat conduction differences. Specifically, during implementation, the boiler is first heated to the operating temperature, then the hot water is quickly drained, and image acquisition is carried out when the internal temperature of the boiler is at the detection temperature of 80 - 120 °C. At this time, the temperature difference between the metal substrate and the water scale is the most significant, and the draining operation avoids the interference of liquid water on imaging. This method is applicable to industrial boilers with relatively low working pressures, taking into account engineering operability while ensuring detection accuracy. The location of the water scale can be accurately identified and its thickness distribution can be evaluated through the temperature distribution characteristics in the thermal imaging image.

[0023] It should also be noted that when collecting multi-modal images, a pan-tilt control dual-spectrum imaging system is used to scan along the axial and circumferential directions of the boiler. During scanning, ensure that the overlap rate of adjacent image frames is not less than 30%. The scanning path can be pre-planned according to the 3D model of the boiler, and the shooting distance is corrected in real time through a laser rangefinder during the scanning process.

[0024] S200 performs polarization filtering processing on the visible light image to suppress the specular reflection of the boiler metal substrate and retain the diffuse reflection area on the water scale surface. Then, color space conversion of the diffuse reflection area is carried out. Based on the chromaticity differences of different types of water scales in the AB chromaticity channels, the type of water scale is judged and a water scale classification map is generated, which further includes: By analyzing the intensity changes of the visible light image at different polarization angles, the specular reflection area generated by the boiler metal substrate and the diffuse reflection area generated on the water scale surface are distinguished, and the specular reflection information is suppressed while the diffuse reflection information is retained.

[0025] The differences between the specular reflection area and the diffuse reflection area are enhanced by adjusting the local contrast and the global gray level distribution to obtain a visible light image that retains the diffuse reflection area.

[0026] The visible light image that retains the diffuse reflection area is subjected to color space conversion to generate an LAB image including the A chromaticity channel and the B chromaticity channel. According to the aggregation areas of the water scale in the A chromaticity channel and the B chromaticity channel, the type of water scale is determined and a water scale classification map is generated. Among them, each pixel in the water scale classification map stores a type label value. For example, a label value of 0 represents rust-type water scale, a label value of 1 represents carbonate-type water scale, and a label value of 2 represents silicate-type water scale.

[0027] It should be noted that since the RGB channels of visible light images are affected by the illumination intensity, there are significant differences in the RGB values of the same water scale under strong light and weak light. Therefore, to avoid illumination interference and accurately obtain the chromaticity information of the water scale to avoid misjudgment of the water scale type, we convert the visible light image into an independent LAB image with a luminance channel and a chromaticity channel.

[0028] By analyzing and identifying a large number of water scale pictures, it is found that carbonate water scale is yellowish-white and granular on the cross-section; silicate water scale is grayish-white; iron rust water scale is blackish-brown or reddish-brown, and its density is more than twice that of general water scale. Further reflected in the color channel, the water scale types include iron rust water scale with an aggregation area of A > 25 and B < 12, carbonate water scale with an aggregation area of A ∈ [-12, 25] and B > 12, and silicate water scale with an aggregation area of A ∈ [-20, 0] and B ∈ [-12, 12].

[0029] S300 converts the infrared thermal imaging image into a temperature anomaly distribution map showing temperature anomaly blocks based on the standard infrared thermal imaging image of the scale-free boiler obtained at the detection temperature, where the temperature anomaly blocks characterize the location and scope of water scale deposition.

[0030] In some embodiments, since the thermal conductivity of the water scale is much lower than that of the metal substrate, obvious local high-temperature characteristics will be formed in the covered area. Therefore, by comparing the infrared thermal imaging of the boiler in the scale-free state with the currently collected infrared thermal imaging image, a temperature anomaly distribution map is generated. Specifically: First, at the same detection temperature, obtain the standard infrared thermal imaging image of the scale-free boiler of the same model and specification, and align the standard infrared thermal imaging image with the infrared thermal imaging image.

[0031] Then, use a sliding window to traverse the aligned infrared thermal imaging image and the standard infrared thermal imaging image, and calculate the mean difference and fluctuation difference of the thermal radiation intensity within the window. Among them, the mean difference of the thermal radiation intensity directly reflects the thermal resistance effect caused by the water scale. The mean difference of the thermal radiation intensity (Δμ w ) = the mean value of the window in the infrared thermal imaging image - the mean temperature value of the window in the standard infrared thermal imaging image. The fluctuation difference of the thermal radiation intensity reflects the uneven degree of the water scale distribution within the window. The fluctuation difference of the thermal radiation intensity (Δσ w ) = |the standard deviation of the window in the infrared thermal imaging image and the standard deviation of the window temperature in the standard infrared thermal imaging image|, and w represents the pixel points within the window.

[0032] Finally, the pixels with Δμ w > 2°C and the corresponding Δσ w < 0.5°C are used as candidate seed points, and the continuously adjacent candidate seed points are clustered, and the geometric center of each cluster is taken as the growth starting point. If the neighboring pixels of the growth starting point satisfy Δμ n>1.5°C and |Δσ n -Δσ o |<0.3°C, the neighboring pixels are included in the temperature anomaly block, and a temperature anomaly distribution map showing the temperature anomaly block is obtained, where the neighboring pixels include multiple pixel points around the growth starting point, and Δσ o represents the temperature standard deviation of the area where the growth starting point is located, and n represents the pixel points in the neighborhood. When the neighboring pixels are at the image boundary, the growth stops.

[0033] In this embodiment, the thermodynamic characteristics Δμ and Δσ are used for collaborative judgment. By using the characteristics that real scale has high Δμ and low Δσ, while noise / artifacts have high Δμ and high Δσ, real scale is effectively distinguished from noise / artifacts, and the geometric shape and area of the temperature anomaly block are accurately identified.

[0034] The above temperature anomaly distribution map identifies the area where there is heat conduction anomaly, but it cannot directly quantify the corresponding relationship between the anomaly degree and the scale thickness. To solve this problem, according to Fourier's law of heat conduction, scale deposition will form an additional thermal resistance layer, resulting in a local temperature anomaly increase. This temperature change shows a non-linear positive correlation with the scale thickness, that is, the greater the thickness, the more significant the temperature difference ΔT. In this embodiment, an encoder-decoder network architecture is used to achieve an accurate inversion from temperature anomaly to scale thickness.

[0035] S400 inputs the temperature anomaly distribution map into a deep neural network, extracts multi-scale features of the temperature anomaly distribution map through the encoder downsampling structure to generate the law of thermal resistance change, and gradually transforms the law of thermal resistance change learned by the encoder into a spatially distributed thickness value through the decoder upsampling structure, and outputs a scale thickness distribution map, where each pixel in the temperature anomaly distribution map stores a floating-point thickness value. Further including: The multi-scale heat conduction features of the temperature anomaly image are gradually extracted through the four-layer downsampling structure of the encoder, and each layer captures the law of thermal resistance change in different regions through convolution operations, and cooperates with the max-pooling operation to increase the number of feature channels while reducing the spatial resolution. Among them, a deformable convolution module is introduced in the third layer to fit the heat flow distortion features caused by sudden changes in thermal conductivity by adaptively adjusting the sampling position. Among them, the heat flow distortion features characterize the local heat conduction path variation caused by scale deposition. When the heat flow passes through the interface between the scale and the boiler metal substrate, due to the difference in thermal conductivity between the scale and the substrate, the heat flow lines will deflect and change in density, and this feature is captured through the learning of the offset of the deformable convolution, which is used to locate the thickness mutation area at the scale edge. The law of thermal resistance change refers to the mapping relationship between the scale spatial distribution and the thermal resistance, including the thermal resistance accumulation effect along the thickness direction of the boiler wall and the thermal conductivity difference of different components of scale (such as rust and carbonate).

[0036] The spatial resolution is gradually restored through a symmetric four-layer upsampling structure of the decoder. The feature maps output by each layer are skip-connected to the feature maps of the same scale of the encoder to fuse the local morphology and global thermal resistance semantic features of the water scale, generating fused features. Among them, the law of thermal resistance change includes the non-linear mapping relationship between temperature and thickness learned by the deep neural network, which conforms to Fourier's law of heat conduction. The global thermal resistance semantic feature is a high-order representation of the temperature-thickness correlation output by the deep decoder of the network. Essentially, it is the learned equivalent thermal resistance model R fouling : R fouling =d / λ scale =ΔT / q, λ scale represents the thermal conductivity. The global thermal resistance semantic feature establishes a physical constraint between the temperature anomaly (ΔT) and the thickness (d) through the heat flux density (q), ensuring that the prediction conforms to the laws of thermodynamics.

[0037] The fused features are compressed into a single channel through a 1×1 convolution of the network output layer, and after normalization using the Sigmoid function, they are linearly mapped to the actual thickness range of 0.1 - 5.0 mm to output a pixel-level water scale thickness distribution map.

[0038] In this embodiment, a technical closed-loop of "temperature anomaly input - thermal resistance feature extraction - thickness distribution output" is formed. Among them, the multi-scale thermal resistance features output by the encoder serve as the input basis for the decoder, and the thickness prediction generated by the decoder provides a spatial registration basis for subsequent water scale composition analysis.

[0039] S500 aligns the water scale classification map and the thickness distribution map through spatial registration technology, ensuring that the pixels at the same spatial position contain both the water scale type and the thickness value, generating a water scale feature fusion map presenting different color regions and their corresponding thicknesses of the water scale type. Further including: Among them, the scale classification map obtained based on visible light image analysis is sub-pixel aligned with the thickness distribution map generated based on infrared thermal imaging to form a composite data matrix. In this matrix, each spatial position (x, y) corresponds to a binary tuple <type label, thickness value>. For example, a pixel storing <0, 1.2> indicates that the deposit at this position is 1.2 mm thick rust. This data structure is visually presented through pseudo-color rendering technology: the rust area is represented by reddish-brown and overlaid with contour lines to show the 1.0 - 3.0 mm thickness gradient, the carbonate area is represented by grayish-white and coordinated with the 0.3 - 1.5 mm color scale, and the silicate area is represented by light gray and marked with the 0.1 - 0.8 mm thickness value. Through pixel-level type-thickness binding, a complete characterization of the scale characteristics is achieved. At the same time, rust scale is shown as reddish-brown overlaid with high-thickness contour lines, carbonate scale is shown as yellowish-white coordinated with medium-thickness color scale, silicate scale is shown as grayish-white coordinated with medium-thickness color scale, and the mixed deposition area is distinguished by a cross-hatch pattern. Meanwhile, the coverage area percentage and average thickness statistics of each type of scale are output, realizing the improvement of detection reliability by establishing the physical association rule between color features and thickness parameters and mutual verification based on multi-modal data. In some embodiments, the method of the embodiment of the present application further includes determining the nozzle pressure and spraying method in the high-pressure nozzle cleaning device based on the scale thickness in the scale thickness distribution map, and providing a scale cleaning solution. Further including: The scale in the scale thickness distribution map is divided into thin-layer scale, medium-thick scale, and thick scale according to the thickness value, and a piecewise linear mapping function between the scale thickness and the nozzle pressure is established; according to the piecewise linear mapping function, the nozzle pressure corresponding to the thickest thickness in the scale area in the scale thickness distribution map is calculated; when the area of the scale area is less than the threshold area, the nozzle is set to spray directly towards the thickest point of the scale area; when the area of the scale area is greater than the threshold area, with the centroid of the scale area as the reference coordinate, the offset vector is calculated in combination with the thickest point coordinate, and the nozzle is set to perform a spiral motion with the thickest point as the center and the offset vector as the diameter to spray the scale area.

[0040] In some embodiments, the softening situation of the soft water device at the front end of the boiler is inferred by the coverage area percentage and average thickness statistics of each type of scale. The distribution of the scale directly reflects the exceeding standard situation of specific metal ions in the boiler water. For example, when it is detected that the iron oxide content in the scale increases significantly, it indicates that the Fe 3+ concentration in the water has exceeded the safety threshold, indirectly proving that the ion exchange resin of the front-end soft water device has a decreased adsorption capacity for iron ions or the oxidation inhibition fails; similarly, the increase in carbonate scale and silicate scale indicates a reduction in the hardness filtration efficiency of the soft water device. Based on this, the front-end soft water device of the boiler can be adjusted accordingly.

[0041] In some embodiments, the present application provides a boiler scale detection device based on multi-modal images, including: An image acquisition module, which is used to panoramically scan the inner wall of the boiler through a dual-spectrum imaging system at the detected temperature, and synchronously acquire visible light images and infrared thermal imaging images.

[0042] A first image processing module, which is used to perform polarization filtering processing on the visible light image, suppress the specular reflection of the boiler metal substrate, retain the diffuse reflection area on the scale surface, then perform LAB color space conversion on the diffuse reflection area, and judge the scale type and generate a scale classification map based on the chromaticity difference of different types of scales in the AB chromaticity channels.

[0043] A second image processing module, which is used to convert the infrared thermal imaging image into a temperature anomaly distribution map presenting temperature anomaly blocks based on the standard infrared thermal imaging image of the scale-free boiler obtained at the detected temperature, where the temperature anomaly blocks characterize the position and range of scale deposition.

[0044] An analysis module, which is used to input the temperature anomaly distribution map into a deep neural network, extract multi-scale features of the temperature anomaly distribution image through an encoder downsampling structure to generate a thermal resistance change law, and gradually convert the thermal resistance change law learned by the encoder into a spatially distributed thickness value through a decoder upsampling structure, and output a scale thickness distribution map, where each pixel in the scale thickness distribution map stores a floating-point thickness value.

[0045] A generation module, which is used to align the scale classification map with the scale thickness distribution map through spatial registration technology to ensure that the pixels at the same spatial position simultaneously contain the scale type and thickness value, and generate a scale feature fusion map presenting different color regions with the corresponding scale type and thickness.

[0046] In some embodiments, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is run by a processor, it executes the steps of the method for detecting boiler scale based on multi-modal images.

[0047] The present application combines image feature analysis with heat conduction modeling through multi-modal image fusion technology, realizes accurate detection and quantitative evaluation of boiler scale. Specifically, visible light polarization imaging effectively strips the interference of the metal substrate, and the scale type is identified through the chromaticity characteristics of scale diffuse reflection; at the same time, through infrared thermal imaging combined with a deep learning network, the temperature anomaly situation is quantified as the scale thickness, and the spatial registration and fusion of the two-modal image data can not only distinguish the chemical composition of the scale, but also quantify the deposition thickness, improve the recognition reliability of mixed scale, and provide a comprehensive basis for formulating accurate cleaning strategies.

[0048] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, as well as combinations of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0049] It should be noted that although several modules or units of devices for action execution are mentioned in the above detailed description, such a division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more of the above-mentioned modules or units can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0050] From the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a portable hard drive, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present application.

[0051] In several embodiments provided in the present application, it should be understood that the disclosed systems, modules, and methods can be implemented in other ways. For example, the above-described module embodiments are merely illustrative. For example, the division of units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be indirect couplings or communication connections through some interfaces, modules, or units, and can be in electrical, mechanical, or other forms.

[0052] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting it. The present application is not limited to the exact structures described above and illustrated in the drawings. It cannot be determined that the specific implementation of the present application is only limited to these descriptions. For those of ordinary skill in the technical field to which the present application pertains, without departing from the concept of the present application, various changes and deformations made should be regarded as falling within the protection scope of the present application.

Claims

1. A boiler scale detection method based on multimodal images, characterized in that Including: At the detection temperature, the inner wall of the boiler is panoramically scanned by a dual-spectrum imaging system, and a visible light image and an infrared thermal imaging image are synchronously collected; The visible light image is subjected to polarization filtering to suppress the specular reflection of the boiler metal substrate and retain the diffuse reflection area on the scale surface. Then, the diffuse reflection area is subjected to LAB color space conversion. Based on the chromaticity difference of different types of scales in the AB chromaticity channels, the scale type is judged and a scale classification map is generated; Based on the standard infrared thermal imaging image of the scale-free boiler obtained at the detection temperature, the infrared thermal imaging image is converted into a temperature anomaly distribution map showing temperature anomaly blocks, where the temperature anomaly blocks characterize the position and range of scale deposition; The temperature anomaly distribution map is input into a deep neural network. The multi-scale features of the temperature anomaly distribution image are extracted through an encoder downsampling structure to generate a thermal resistance change law. Through a decoder upsampling structure, the thermal resistance change law learned by the encoder is gradually converted into a spatially distributed thickness value, and a scale thickness distribution map is output, where each pixel in the scale thickness distribution map stores a floating-point thickness value; The scale classification map and the scale thickness distribution map are aligned through a spatial registration technique to ensure that the pixels at the same spatial position simultaneously contain the scale type and the thickness value, and a scale feature fusion map presenting different color regions of the scale type and its corresponding thickness is generated.

2. The method for detecting boiler scale based on multimodal images according to claim 1, wherein, The visible light image is subjected to polarization filtering to suppress the specular reflection of the boiler metal substrate and retain the diffuse reflection area on the scale surface. Then, the diffuse reflection area is subjected to LAB color space conversion. Based on the chromaticity difference of different types of scales in the AB chromaticity channels, the scale type is judged and a scale classification map is generated, which further includes: By analyzing the intensity change of the visible light image at different polarization angles, the specular reflection area generated by the boiler metal substrate and the diffuse reflection area generated by the scale surface are distinguished, and the specular reflection information is suppressed and the diffuse reflection information is retained; The difference between the specular reflection area and the diffuse reflection area is enhanced by adjusting the local contrast and the global gray distribution to obtain a visible light image retaining the diffuse reflection area; The visible light image retaining the diffuse reflection area is subjected to color space conversion to generate an LAB image including the A chromaticity channel and the B chromaticity channel. According to the aggregation areas of the scale in the A chromaticity channel and the B chromaticity channel, the scale type is determined and the scale classification map is generated, where each pixel in the scale classification map stores a scale type label value, and the scale types include rust scale, carbonate scale, and silicate scale with different aggregation areas.

3. The method for detecting boiler scale based on multimodal images according to claim 1, wherein, Based on the standard infrared thermal imaging image of the scale-free boiler, the infrared thermal imaging image is converted into an infrared thermal imaging image showing temperature anomaly blocks, where the temperature anomaly blocks characterize the position and range of scale deposition, and further includes; At the same detection temperature, a standard infrared thermal imaging image of a scale-free boiler of the same model and specification is obtained, and the standard infrared thermal imaging image is aligned with the infrared thermal imaging image; The aligned infrared thermal imaging image and the standard infrared thermal imaging image are traversed using a sliding window, and the mean difference and the fluctuation difference of the thermal radiation intensity within the window are calculated, where the mean difference of the thermal radiation intensity is the difference between the window mean in the infrared thermal imaging image and the window mean in the standard infrared thermal imaging image, and the fluctuation difference of the thermal radiation is the absolute value of the difference between the window standard deviation in the infrared thermal imaging image and the window standard deviation in the standard infrared thermal imaging image; Take Δμ w > 2 °C and corresponding Δσ w < 0.5 °C pixels as candidate seed points, cluster the continuously adjacent candidate seed points, and take the geometric center of each cluster as the growth starting point; If the neighborhood pixels of the growth starting point satisfy Δμ n > 1.5 °C and |Δσ n - Δσ o | < 0.3 °C, then the neighborhood pixels are included in the scale region, and a temperature anomaly distribution map with temperature anomaly blocks is obtained, where the neighborhood pixels include multiple pixel points around the growth starting point, Δμ represents the mean difference in thermal radiation intensity, Δσ represents the difference in thermal radiation fluctuation, Δσ o represents the temperature standard deviation of the region where the growth starting point is located, w represents the pixel points in the window, and n represents the pixel points in the neighborhood.

4. The method for detecting boiler scale based on multimodal images according to claim 3, characterized in that The temperature anomaly distribution map is input into a deep neural network, and multi-scale features of the temperature anomaly distribution map are extracted through an encoder downsampling structure to generate the law of thermal resistance change. Through a decoder upsampling structure, the law of thermal resistance change learned by the encoder is gradually transformed into a spatially distributed thickness value, and a water scale thickness distribution map is output. Further included are: Multi-scale heat conduction features of the temperature anomaly image are gradually extracted through four layers of downsampling structures of the encoder, and the law of thermal resistance change in different regions is captured through convolution operations in each layer. At the same time, the maximum pooling operation is used to increase the number of feature channels while reducing the spatial resolution. A deformable convolution module is introduced in the third layer to fit the heat flow distortion features caused by sudden changes in the thermal conductivity by adaptively adjusting the sampling position; The spatial resolution is gradually restored through four symmetric upsampling structures of the decoder, and the feature map output by each layer is skip-connected with the same-scale features of the encoder to fuse the local morphology of the water scale and the global thermal resistance semantic features to generate fused features; The fused features are compressed into a single channel through a 1×1 convolution of the network output layer, and after being normalized by the Sigmoid function, they are linearly mapped to thickness values of 0.1 - 5.0 mm, and a pixel-level water scale thickness distribution map is output.

5. The method for detecting boiler scale based on multimodal images according to claim 4, wherein, The water scale classification map and the water scale thickness distribution map are aligned through a spatial registration technique to ensure that the pixels at the same spatial position simultaneously contain the water scale type and the thickness value, and a water scale feature fusion map presenting different color regions and their corresponding thicknesses is generated. Further included are: The water scale classification map obtained based on the visible light image analysis and the thickness distribution map generated based on the infrared thermal imaging are aligned at the sub-pixel level to form a composite data matrix. In the matrix, each spatial position corresponds to a binary group including a type label and a thickness value, and at the same time, the percentage of the coverage area and the average thickness of each type of water scale are statistically analyzed.

6. The method for detecting boiler scale based on multimodal images according to claim 4, wherein, The method further includes: According to the thickness value, the water scale in the water scale thickness distribution map is divided into thin-layer water scale, medium-thick water scale, and thick water scale, and a piecewise linear mapping function between the water scale thickness and the nozzle pressure is established; According to the piecewise linear mapping function, the nozzle pressure corresponding to the thickest thickness in the water scale area in the water scale thickness distribution map is calculated; When the area of the water scale area is less than the threshold area, the nozzle is set to be directed towards the thickest point of the water scale area for spraying; When the area of the water scale area is greater than the threshold area, with the centroid of the water scale area as the reference coordinate, an offset vector is calculated in combination with the coordinates of the thickest point of the water scale area, and the nozzle is set to perform a spiral motion with the thickest point as the center and the offset vector as the diameter to spray the water scale area.

7. A boiler scale detection device based on multimodal images, characterized in that, Including: An image acquisition module, configured to perform panoramic scanning on the inner wall of the boiler through a dual-spectrum imaging system at the detected temperature, and synchronously acquire a visible light image and an infrared thermal imaging image; A first image processing module, configured to perform polarization filtering on the visible light image to suppress the specular reflection of the boiler metal substrate, retain the diffuse reflection region on the scale surface, then perform LAB color space conversion on the diffuse reflection region, and judge the scale type and generate a scale classification map based on the chromaticity difference of different types of scales in the AB chromaticity channels; A second image processing module, configured to convert the infrared thermal imaging image into a temperature anomaly distribution map presenting temperature anomaly blocks based on the standard infrared thermal imaging image of the scale-free boiler obtained at the detected temperature, wherein the temperature anomaly blocks characterize the position and range of scale deposition; An analysis module, configured to input the temperature anomaly distribution map into a deep neural network, extract multi-scale features of the temperature anomaly distribution image through an encoder downsampling structure to generate a thermal resistance change rule, and gradually convert the thermal resistance change rule learned by the encoder into a spatially distributed thickness value through a decoder upsampling structure, and output a scale thickness distribution map, wherein each pixel in the scale thickness distribution map stores a floating-point thickness value; A generation module, configured to align the scale classification map with the scale thickness distribution map through a spatial registration technique to ensure that pixels at the same spatial position simultaneously contain the scale type and the thickness value, and generate a scale feature fusion map presenting different color regions with the corresponding scale type and thickness; 8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by a processor, it executes the steps of the multi-modal image-based boiler scale detection method according to any one of claims 1 to 6.

Images