Enhanced recognition method and system for optical image of valve in severe weather based on multispectral fusion

Abstract

The invention provides a multispectral fusion-based valve optical image enhancement identification method and system in severe weather, and the method comprises the steps: obtaining multispectral optical image data and mechanical dynamic response characteristics of an industrial pipeline valve in severe weather; based on the mechanical dynamic response characteristics, separating the multispectral optical image data to extract multiband texture characteristics associated with the reflection characteristics of the valve material; double amplitude correction of a spatial domain and a frequency domain is carried out on the multiband texture features; distributing a spectrum fusion weight for the multispectral optical image data according to the corrected multiband texture features; and generating an enhanced optical image matched with the actual mechanical state of the valve based on the spectrum fusion weight and the corrected multiband texture features. According to the method, the recognition accuracy of defects such as valve surface cracks and deformation in severe weather is improved.

Description

Technical Field

[0001] The present application relates to the field of multispectral fusion technology, and particularly to a method and system for enhancing the optical image recognition of valves under adverse weather conditions based on multispectral fusion. Background Art

[0002] When industrial pipeline valves operate under adverse weather conditions such as strong winds, sandstorms, rain, and fog, their surfaces are easily interfered by the environment (such as sand adhesion, water vapor scattering, etc.), making it difficult for traditional optical detection methods to accurately identify defects such as valve cracks and corrosion. Existing technologies need to solve the collaborative optimization problem of noise suppression and effective feature enhancement in multispectral images, and at the same time, it is necessary to fuse valve mechanical state data (such as vibration, stress) to improve the physical relevance of defect detection.

[0003] Currently, the more advanced solutions adopt an image enhancement method based on the fusion of infrared and visible light dual bands. By establishing a statistical model of sand and dust scattering noise, background difference processing is performed on multispectral images, and the visible light texture details and infrared thermal radiation characteristics are fused with fixed weights, and finally an enhanced image is output.

[0004] This solution has the problems that it is difficult to distinguish between sand and dust adhesion and real defects; the fixed weight fusion strategy cannot adapt to the reliability changes of each band feature under different weather conditions; and the matching degree between the structural deformation features in the enhanced image and the real mechanical state is low. Summary of the Invention

[0005] The present application provides a method and system for enhancing the optical image recognition of valves under adverse weather conditions based on multispectral fusion, so as to solve the problem of low recognition accuracy of defects such as surface cracks and deformations of valves under adverse weather conditions in the prior art.

[0006] In a first aspect, the present application provides a method for enhancing the optical image recognition of valves under adverse weather conditions based on multispectral fusion, including: Under adverse weather conditions, obtaining multispectral optical image data and mechanical dynamic response characteristics of industrial pipeline valves; Based on the mechanical dynamic response characteristics, separating the multispectral optical image data to extract multi-band texture features associated with the reflection characteristics of the valve material; Performing double amplitude correction on the multi-band texture features in the spatial domain and the frequency domain; According to the corrected multi-band texture features, assigning spectral fusion weights to the multispectral optical image data; Based on the spectral fusion weights and the corrected multi-band texture features, generating an enhanced optical image that matches the actual mechanical state of the valve.

[0007] Optionally, based on the mechanical dynamic response characteristics, separating the multispectral optical image data to extract multi-band texture features associated with the reflection characteristics of the valve material, including: Constructing a dynamic noise separation model related to dust scattering interference, where the parameters of the dynamic noise separation model are jointly determined by the vibration frequency time-varying curve and the mechanical stress spatial gradient distribution in the mechanical dynamic response characteristics; According to the dynamic noise separation model, determining the time-varying interference intensity and the noise diffusion region of the dust scattering noise in the multispectral optical image data; Performing a coupling calculation on the time-varying interference intensity and the noise diffusion region to determine the noise separation boundary condition; According to the noise separation boundary condition, performing a cross-band correlation analysis on the visible light band reflectance and the infrared band radiation intensity at the same spatial coordinates in the multispectral optical image data to screen out target pixels with the visible light band reflectance lower than a first threshold and the infrared band radiation intensity higher than a second threshold, and taking the set composed of all target pixels as the dust scattering noise component; Extracting multi-band texture features from the multispectral optical image data after removing the dust scattering noise component.

[0008] Optionally, the step of performing a cross-band correlation analysis on the visible light band reflectance and the infrared band radiation intensity at the same spatial coordinates in the multispectral optical image data according to the noise separation boundary condition to screen out target pixels with the visible light band reflectance lower than a first threshold and the infrared band radiation intensity higher than a second threshold includes: Based on the noise separation boundary condition and the noise diffusion region, establishing an inverse change relationship mapping table between the visible light band reflectance and the infrared band radiation intensity at the same spatial coordinates; According to the inverse change relationship mapping table, constructing a joint determination rule for the attenuation amplitude of the visible light reflectance and the increase amplitude of the infrared radiation intensity, where the boundary conditions of the joint determination rule include a first threshold and a second threshold. The first threshold is dynamically calculated from the difference between the average reflectance of the visible light band without dust interference and the current time-varying interference intensity, and the second threshold is obtained by superposing the radiation increment of the noise diffusion region on the reference intensity within the normal radiation range of the valve material in the infrared band; Screening out target pixels with the visible light band reflectance lower than the first threshold and the infrared band radiation intensity higher than the second threshold.

[0009] Optionally, the step of establishing an inverse change relationship mapping table between the visible light band reflectance and the infrared band radiation intensity at the same spatial coordinates based on the noise separation boundary condition and the noise diffusion region includes: Convert the time-varying interference intensity in the noise separation boundary condition into a spatio-temporal matrix according to the distribution of amplitude mutation points of the vibration frequency time-varying curve; Calculate the diffusion area ratio parameter of each spatial coordinate position according to the spatial coverage shape of the noise diffusion region; Perform a pixel-by-pixel multiplication operation on the spatio-temporal matrix and the diffusion area ratio parameter to generate a noise interference superposition weight map; Based on the gradient distribution direction in the noise interference superposition weight map, identify low-reflectivity pixels with a gradient distribution direction angle greater than a set angle in the visible light band reflectivity, and at the same time identify high-radiation intensity pixels with a gradient distribution direction angle less than the set angle in the infrared band radiation intensity; Construct an inverse change relationship mapping table with the spatial overlap degree of the low-reflectivity pixels and the high-radiation intensity pixels as the index key value.

[0010] Optionally, the coupling calculation of the time-varying interference intensity and the noise diffusion region to determine the noise separation boundary condition includes: Calculate the instantaneous influence value of the time-varying interference intensity based on the vibration frequency time-varying curve; Divide the noise diffusion region into a core influence area and an edge transition area according to the mechanical stress spatial gradient distribution; In the core influence area, perform a convolution operation on the time-varying interference intensity and the spatial stress gradient to generate the interference intensity of the core influence area, and perform extended compensation in the edge transition area to obtain the interference intensity of the edge transition area; Adjust the numerical value of the parameter used to separate the core influence area and the edge transition area according to the interference intensity of the core influence area and the interference intensity of the edge transition area to obtain the noise separation boundary condition.

[0011] Optionally, the generation of an enhanced optical image matching the actual mechanical state of the valve based on the spectral fusion weight and the corrected multi-band texture feature includes: Dynamically adjust the spectral fusion weight according to the mechanical dynamic response characteristics; According to the corrected multi-band texture feature, perform hierarchical fusion on the multi-spectral optical image data according to the adjusted spectral fusion weight; Mark potential structural anomaly areas in the hierarchically fused multi-spectral optical image data based on the mechanical stress spatial gradient distribution; Generate an enhanced optical image matching the actual mechanical state of the valve based on the multi-spectral optical image data marked with the potential structural anomaly areas.

[0012] Optionally, the dual amplitude correction of the multi-band texture features in the spatial domain and the frequency domain includes: Based on the mechanical dynamic response features, extract the main peak value of the energy distribution of the multi-band texture features in the frequency domain, and use the matching degree between the amplitude of the frequency component corresponding to the main peak value of the energy distribution and the natural vibration frequency of the valve material as the frequency domain amplitude correction coefficient; According to the maximum gradient direction of the mechanical stress spatial gradient distribution, calculate the anisotropy intensity ratio of the multi-band texture features in the spatial domain; In the spatial domain, along the maximum gradient direction, perform enhancement correction on the amplitude of the multi-band texture features that is proportional to the anisotropy intensity ratio, and perform attenuation correction on the amplitude perpendicular to the maximum gradient direction that is inversely proportional to the anisotropy intensity ratio. In the frequency domain, apply amplitude suppression to the multi-band texture features that is inversely proportional to the frequency domain amplitude correction coefficient to obtain the modified multi-band texture features.

[0013] In a second aspect, the present application provides a valve optical image enhancement and recognition system based on multi-spectral fusion under harsh weather conditions, including: An acquisition module for acquiring multi-spectral optical image data and mechanical dynamic response features of industrial pipeline valves under harsh weather conditions; A separation module for separating the multi-spectral optical image data based on the mechanical dynamic response features to extract multi-band texture features associated with the reflection characteristics of the valve material; A correction module for performing dual amplitude correction on the multi-band texture features in the spatial domain and the frequency domain; An allocation module for allocating spectral fusion weights to the multi-spectral optical image data according to the corrected multi-band texture features; A generation module for generating an enhanced optical image that matches the actual mechanical state of the valve based on the spectral fusion weights and the corrected multi-band texture features.

[0014] In a third aspect, the present application provides a computing device including a processor and a memory, where a computer program is stored in the memory, and the processor is configured to run the computer program to execute any one of the methods for enhancing and recognizing valve optical images based on multi-spectral fusion under harsh weather conditions in the first aspect.

[0015] In a fourth aspect, the present application provides a computer storage medium with computer program instructions stored thereon, and when the computer program instructions are executed by a processor, they implement any one of the methods for enhancing and recognizing valve optical images based on multi-spectral fusion under harsh weather conditions in the first aspect.

[0016] In this application, an optical image enhancement and recognition method for valves under adverse weather conditions based on multispectral fusion is provided. The method includes: under adverse weather conditions, acquiring multispectral optical image data and mechanical dynamic response characteristics of industrial pipeline valves; based on the mechanical dynamic response characteristics, separating the multispectral optical image data to extract multi-band texture characteristics associated with the reflection characteristics of the valve material; performing double amplitude correction on the multi-band texture characteristics in the spatial domain and frequency domain; according to the corrected multi-band texture characteristics, assigning spectral fusion weights to the multispectral optical image data; and generating an enhanced optical image that matches the actual mechanical state of the valve based on the spectral fusion weights and the corrected multi-band texture characteristics.

[0017] The technical solution provided by this application has the following beneficial effects: This application realizes the synchronous acquisition of multi-modal data of the valve state, providing a time-space aligned input source for subsequent fusion processing. A physical noise model is established using vibration and stress data to improve the separation accuracy between sand-dust scattering noise and real valve characteristics. Through directional enhancement (spatial domain) and resonance frequency band screening (frequency domain), texture characteristics strongly related to the valve material / deformation are retained. The band weights are adaptively adjusted according to the reliability of the corrected characteristics to optimize the feature complementarity under different weather conditions. An image that simultaneously retains high-resolution texture details and mechanical state-related characteristics is output to improve the defect detection accuracy.

[0018] Furthermore, this application also constructs a dynamic noise separation model driven jointly by vibration frequency and mechanical stress to determine the coupling boundary conditions of the time-varying interference intensity and the noise diffusion region, and then screens noise pixels based on the inverse change characteristics of visible light reflectivity and infrared radiation intensity, and finally extracts effective multi-band texture characteristics from the denoised data.

[0019] Moreover, this solution breakthroughly converts the valve mechanical state data into physical constraint conditions for image noise separation, improving the recognition accuracy of sand-dust scattering noise and avoiding the mis-removal of real defect characteristics.

[0020] These aspects or other aspects of this application will be more clearly understood in the following description of the embodiments. Description of the Drawings

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1A flowchart of a method for enhancing the recognition of valve optical images under severe weather conditions based on multispectral fusion provided by an embodiment of the present application; Figure 2 A schematic structural diagram of a system for enhancing the recognition of valve optical images under severe weather conditions based on multispectral fusion provided by an embodiment of the present application; Figure 3 A schematic structural diagram of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0023] In order to enable those skilled in the art of the present technology to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application.

[0024] In some processes described in the specification, claims and the above-mentioned accompanying drawings of the present application, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0025] Researchers have found that traditional optical detection methods for industrial pipeline valves under severe weather conditions have problems such as serious interference from sand and dust scattering noise and inaccurate extraction of defect features, and the existing technologies lack dynamic association with the mechanical state of the valves. Based on this, an embodiment of the present application provides a method for enhancing the recognition of valve optical images under severe weather conditions based on multispectral fusion. This method synchronously collects multispectral images and mechanical dynamic response data, constructs a noise separation model driven by physical characteristics, and adopts spatial-frequency domain collaborative correction and dynamic weight fusion technologies to improve the recognition accuracy of valve surface defects in sand / dust / rain / fog environments. The technical solution of the present application is applicable to on-line detection and health status assessment of pipeline valves in industrial fields such as petroleum and chemical industries under severe working conditions such as strong wind and sand and high humidity.

[0026] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0027] Figure 1 The flowchart of a method for enhancing the recognition of valve optical images under harsh weather conditions based on multispectral fusion provided by an embodiment of the present application is as follows. Figure 1 As shown, the method includes: Step 101: Under harsh weather conditions, obtain multispectral optical image data and mechanical dynamic response characteristics of industrial pipeline valves.

[0028] In this step, harsh weather refers to a strong wind and sand environment with a sand dust concentration ≥ 200 μg / m³ and a visibility ≤ 500 m, or a rain and fog environment with a rainfall intensity ≥ 10 mm / h and a relative humidity ≥ 90%. The sand dust concentration is obtained in real time by a laser scattering particle monitor, the visibility is measured by a transmissive visibility meter, the rainfall intensity is collected by a tipping bucket rain gauge, and the relative humidity is detected by a capacitive humidity sensor. Multispectral optical image data: Optical image data containing visible light bands and at least two infrared bands. Mechanical dynamic response characteristics include vibration frequency signals (in the range of 10 - 2000 Hz, collected by an acceleration sensor), surface mechanical stress distribution (measured by a strain gauge array with a resolution ≥ 100 points / m²), and acoustic emission signals during the valve opening and closing process (in the frequency band of 20 - 100 kHz, obtained by a broadband acoustic emission sensor). The sampling rate of the vibration frequency signal and the acoustic emission signal is ≥ 10 kHz, and the time synchronization error between the mechanical stress distribution data and the multispectral optical image data is ≤ 1 ms.

[0029] In an embodiment of the present application, multispectral optical image data is obtained by an image acquisition device installed near the valve, and mechanical dynamic response characteristics are collected using vibration sensors and stress sensors. The image acquisition device uses a multispectral camera, which can still work stably in harsh environments; the vibration sensor uses a high-precision accelerometer to monitor the vibration state of the valve in real time; the stress sensor uses a distributed strain gauge array to accurately measure the surface stress distribution of the valve. All data acquisition devices ensure the time consistency of the data through a time synchronization module.

[0030] For example, during the monitoring of a gas pipeline valve in a chemical plant, when encountering strong sand dust weather, a multispectral camera is used to collect visible light and infrared images of the valve. At the same time, an acceleration sensor and a strain gauge array installed on the valve flange are used to obtain the vibration frequency signal and surface stress distribution data of the valve in the sand dust environment respectively. These data are transmitted to a central processor through an industrial Internet of Things gateway for subsequent processing.

[0031] Step 102: Based on the mechanical dynamic response characteristics, separate the multispectral optical image data to extract multi-band texture characteristics associated with the reflection characteristics of the valve material.

[0032] In this step, the reflection characteristics of the valve material refer to the reflection laws of different valve material surfaces for light in the multi-spectral band. The multi-band texture features refer to the combination of physically meaningful features extracted from the multi-spectral image.

[0033] In the embodiment of the present application, based on the collected vibration frequency signal, calculate its time variation curve, and analyze the correlation between the vibration characteristics and the noise; at the same time, calculate the spatial gradient distribution according to the stress distribution data to determine the stress concentration area. Input these two features into the dynamic noise separation model. First, calculate the time-varying interference intensity, then determine the spatial range of the noise diffusion area, and finally obtain the noise separation boundary conditions through coupled calculation. According to the boundary conditions, perform cross-band analysis on the visible light reflectivity and infrared radiation intensity of each pixel point in the image, screen out and remove the pixel set that conforms to the noise characteristics, and finally extract the effective multi-band texture features.

[0034] For example, continuing with the example of the chemical plant valve, the processor calculates the variation curve of the main vibration frequency over time based on the collected vibration signal and finds that the vibration frequency suddenly increases during a certain period; at the same time, the stress distribution shows that stress concentration appears at the bottom of the valve. The system determines that this period and area are high-noise interferences, establishes the corresponding noise separation boundary, and successfully separates the dust interference attached to the valve surface from the image, retaining the true valve texture features.

[0035] Step 103: Perform double amplitude correction on the multi-band texture features in the spatial domain and the frequency domain.

[0036] In this step, the spatial domain amplitude correction represents the adjustment of the feature intensity in the image spatial dimension. The frequency domain amplitude correction represents the adjustment of the feature intensity in the frequency dimension.

[0037] In the embodiment of the present application, first analyze the energy distribution of the texture features in the frequency domain, find the main peak frequency, and calculate its matching degree with the natural frequency of the valve material as the frequency domain correction coefficient. At the same time, in the spatial domain, calculate the anisotropy intensity ratio according to the maximum direction of the stress gradient distribution. Enhance the feature amplitude along the maximum gradient direction in the spatial domain and weaken the amplitude in the perpendicular direction; suppress the irrelevant frequency components in the frequency domain according to the correction coefficient. Finally, obtain the double correction result that retains the important features.

[0038] For example, for the texture features of the chemical plant valve, the system detects an obvious peak in a certain frequency band, which coincides with the natural frequency of the stainless steel valve; the spatial analysis finds that the texture features show obvious anisotropy along the main stress direction of the valve. Based on this, the system enhances the feature frequency band in the frequency domain and enhances the feature display along the main stress direction in the spatial domain, effectively highlighting the possible crack features of the valve.

[0039] Step 104: Assign spectral fusion weights to the multi-spectral optical image data according to the corrected multi-band texture features.

[0040] In this step, the spectral fusion weight represents a parameter that determines the proportion of images in different bands in the final fusion result. The corrected multi-band texture feature refers to an optimized feature set after double amplitude correction processing.

[0041] In the embodiment of the present application, according to the quality evaluation result of the corrected multi-band texture feature, the reliability of each band image is dynamically determined. When the dust is severe, the weight of the visible light band is reduced and the weight of the infrared band is increased; when the humidity is high, the weight ratio of different infrared bands is appropriately adjusted. The weight adjustment considers factors such as the clarity of the texture feature and the matching degree with the mechanical state, and realizes adaptive allocation through a preset weight calculation model.

[0042] For example, in the case of a chemical plant, the system automatically reduces the weight of the visible light band and increases the weight of the mid-wave infrared in case of dust weather; when it is detected that the stress in a certain area of the valve is abnormal, the weight ratio of this area in the long-wave infrared image is increased to ensure that the characteristics of the stress abnormal area are fully displayed.

[0043] Step 105: Generate an enhanced optical image that matches the actual mechanical state of the valve based on the spectral fusion weight and the corrected multi-band texture feature.

[0044] In this step, mechanical state matching means the consistency between the image features and the actual physical state of the valve. The enhanced optical image refers to a high-quality valve image after processing.

[0045] In the embodiment of the present application, the images in each band are pixel-level fused according to the allocated weights, and the feature enhancement processing is performed on the marked potential abnormal areas. During the fusion process, the balance between texture details and thermal radiation characteristics is maintained to ensure that the generated image can clearly display the surface topography and reflect the internal stress distribution. In the finally output image, the valve defect features are highly consistent with the mechanical state data.

[0046] For example, in the enhanced image of the chemical plant valve finally generated, a small crack at the bottom of the valve is clearly shown, and this position exactly corresponds to the stress concentration area, and the crack direction is consistent with the stress gradient direction, verifying the reliability of the detection result.

[0047] This method effectively overcomes the problem of noise interference under bad weather through the collaborative processing of multi-spectral images and mechanical dynamic responses, and realizes the accurate identification of valve surface defects. The generated enhanced image not only clearly shows the surface topography features, but also can reflect the changes in the internal mechanical state, providing a reliable basis for the health state assessment of the valve. The whole set of solutions shows good adaptability and accuracy in a complex industrial environment.

[0048] In order to solve the problem of insufficient detection accuracy of valve surface defects under bad weather conditions, in some embodiments, step 102: based on the mechanical dynamic response characteristics, separate the multi-spectral optical image data to extract multi-band texture characteristics associated with the reflection characteristics of the valve material, including: Step 201: Construct a dynamic noise separation model related to sand and dust scattering interference, where the parameters of the dynamic noise separation model are jointly determined by the vibration frequency time-varying curve and the mechanical stress spatial gradient distribution in the mechanical dynamic response characteristics.

[0049] In step 201, the dynamic noise separation model is a physical driving model for distinguishing real valve characteristics and environmental noise. The vibration frequency time-varying curve represents a characteristic curve reflecting the change of the main vibration frequency of the valve over time. The vibration frequency time-varying curve is converted into noise model parameters through the following steps: perform a short-time Fourier transform on the original vibration signal collected by the acceleration sensor, and extract the curve of the amplitude of the main frequency component in the frequency band of 1-100 Hz changing with time; calculate the standard deviation of this curve within a sliding time window (0.1 second) as the time-varying interference intensity reference value; combine the valve structure resonance frequency (pre-stored in the material database) to normalize the reference value, and finally generate the time-varying interference intensity parameter for the dynamic noise separation model. The mechanical stress spatial gradient distribution characterizes the spatial distribution of the stress change rate on the valve surface. The mechanical stress spatial gradient distribution is obtained through the following methods: use an embedded fiber Bragg grating sensor array (spacing ≤ 2 cm) to measure the surface strain distribution of the valve in real time; convert the strain data into stress values according to the stress-strain constitutive relationship of the valve material (pre-stored in the material database); finally, obtain the spatial gradient distribution through three-dimensional spline interpolation, and its resolution is kept in a 1:1 match with the spatial sampling rate of the multi-spectral optical image.

[0050] In the embodiments of the present application, the original vibration signal is obtained through a vibration sensor, and the main frequency component is extracted through time-frequency analysis to generate a time-varying curve; at the same time, the stress gradient distribution is calculated according to the strain gauge array data. The two are input into a neural network model based on physical constraints, and a dynamic noise separation model that can reflect the vibration-stress-noise correlation relationship is trained.

[0051] Step 202: According to the dynamic noise separation model, determine the time-varying interference intensity and noise diffusion region of the sand and dust scattering noise in the multi-spectral optical image data.

[0052] In step 202, the time-varying interference intensity represents a strength parameter that quantifies the fluctuation of the sand and dust scattering noise over time. The noise diffusion region represents the spatial range affected by sand and dust in the image.

[0053] In the embodiments of the present application, the trained model is used to analyze the vibration spectrum characteristics and stress gradient distribution at the current moment, and the time-varying interference intensity value is calculated; at the same time, the main noise diffusion path is determined according to the stress gradient direction, and the current noise area range is predicted by combining historical diffusion data.

[0054] Step 203: Perform a coupling calculation on the time-varying interference intensity and the noise diffusion area to determine the noise separation boundary condition.

[0055] In step 203, the coupling calculation represents a fusion operation of time-domain and spatial-domain features. The noise separation boundary condition represents the decision boundary for defining noise and effective features.

[0056] In the embodiments of the present application, the time-varying interference intensity is mapped onto the spatial grid of the noise diffusion area, and a two-dimensional boundary distribution map is generated through spatial convolution operations. Among them, linear superposition is used for the core influence area, and weighted fusion based on the stress gradient direction is used for the edge area, and finally a noise separation boundary with physical significance is generated.

[0057] Step 204: According to the noise separation boundary condition, perform a cross-band correlation analysis on the visible light band reflectance and the infrared band radiation intensity with the same spatial coordinates in the multi-spectral optical image data to screen out the target pixels whose visible light band reflectance is lower than the first threshold and whose infrared band radiation intensity is higher than the second threshold, and use the set composed of all target pixels as the dust scattering noise component.

[0058] In step 204, the visible light band reflectivity refers to the quantitative index of the reflection ability of the valve surface to visible light with a wavelength of 400 - 700 nm, which is obtained by collecting through the visible light channel of a multispectral camera. During specific calculation, taking the reflectivity of a standard whiteboard as a reference, the ratio of the reflected light intensity to the incident light intensity on the valve surface is normalized to a reflectivity value within the range of 0 - 1. This value is jointly affected by the valve material, surface condition, and environmental interference. In a sandy dust environment, the attachment of sandy dust will cause a local abnormal decrease in reflectivity. The infrared band radiation intensity refers to the thermal radiation energy intensity of the valve surface in a specific infrared band (such as 3 - 5 μm or 8 - 12 μm), which is measured by an infrared thermal imager or an infrared multispectral camera. Its value reflects the comprehensive characteristics of the valve surface temperature and emissivity. During calculation, based on the blackbody radiation law and combined with the sensor calibration parameters, the original radiation signal is converted into a standardized radiation intensity value. The coverage of sandy dust will change the local emissivity, resulting in an abnormal increase in radiation intensity. Cross - band correlation analysis represents the collaborative processing of multispectral data. The target pixels represent a set of pixels that meet the noise characteristics. The sandy dust scattering noise component refers to the set of optical interference characteristics caused by sandy dust particles in the multispectral image, manifested as pixel regions with an abnormal decrease in reflectivity in the visible light band (sandy dust occlusion) and an abnormal increase in radiation intensity in the infrared band (sandy dust thermal radiation). This component is separated from the original image through cross - band correlation analysis, and its spatial distribution matches the noise diffusion region determined by mechanical vibration and stress data. It is the main noise source that needs to be suppressed in subsequent image enhancement processing.

[0059] In the embodiment of the present application, for each spatial coordinate point, the compliance of the visible light reflectivity and the infrared radiation intensity with the boundary conditions is synchronously compared. When a certain point simultaneously meets the conditions that the visible light reflectivity is lower than the dynamic threshold (affected by the time - varying interference intensity) and the infrared radiation is higher than the reference value (affected by the characteristics of the noise region), it is determined as a noise pixel.

[0060] Step 205: Extract multi - band texture features from the multispectral optical image data after removing the sandy dust scattering noise component.

[0061] In step 205, the multi - band texture features represent a set of effective features after denoising.

[0062] In the embodiment of the present application, for the multispectral data after removing noise pixels, the texture features of each band (such as the gray - level co - occurrence matrix, etc.) are respectively extracted, and the correlation features between bands (such as the reflectivity ratio, gradient correlation, etc.) are calculated to form a complete feature expression.

[0063] The following is a specific example: In an actual monitoring case of a gas pipeline valve in a chemical plant, when encountering a strong sandstorm with a visibility of less than 300 meters, the average reflectance of the valve in the visible light band collected by the system decreased from the normal value of 0.65 to 0.32 (calculation formula: reflectance = measured reflected light intensity / reflected light intensity of the standard whiteboard × environmental attenuation coefficient of 0.9). At the same time, the long-wave infrared radiation intensity increased from the standard value of 0.45 to 0.83 (affected by the thermal radiation of sand particles). Through vibration signal analysis, it was found that there was abnormal vibration at 125 Hz at the flange connection of the valve (the normal value is 80 - 100 Hz), and the vibration amplitude reached 0.8 mm / s² (obtained by integrating the acceleration sensor); the strain gauge data showed that there was a stress concentration area with a maximum value of 25 MPa at the lower part of the flange (calculated according to the strain-stress conversion formula σ = Eε, where E is the elastic modulus of the valve material). After the system input these parameters into the dynamic noise model, the time-varying interference intensity was determined to be 0.72 (calculation formula: 0.5 × vibration anomaly coefficient + 0.5 × stress gradient coefficient), and a noise diffusion area with a radius of 15 cm centered on the flange was demarcated (determined according to the safety factor of 1.5 times the stress gradient extension length). By setting the reflectance threshold of 0.4 and the radiation intensity threshold of 0.7 (calculated based on the 3σ principle of historical data statistics), about 8% of the pixel points (a total of 2,356 pixels) at the flange connection were successfully separated as sandstorm noise components, and finally two radial cracks with lengths of 3.2 mm and 4.5 mm were clearly shown in the denoised image.

[0064] In the embodiment of the present application, this method effectively solves the problem of inaccurate extraction of valve surface features in harsh environments through noise separation driven by mechanical state data. The generated texture features can not only clearly reflect the surface morphology but also be highly consistent with the actual mechanical state of the valve, providing a reliable basis for defect diagnosis.

[0065] To solve the problem of serious noise interference in the identification of valve surface defects in a sandstorm environment, in some embodiments, step 203: performing cross-band correlation analysis on the reflectance of the visible light band and the radiation intensity of the infrared band with the same spatial coordinates in the multi-spectral optical image data according to the noise separation boundary conditions to screen out target pixels with the reflectance of the visible light band lower than the first threshold and the radiation intensity of the infrared band higher than the second threshold, including: Step 301: Based on the noise separation boundary conditions and the noise diffusion area, establish a mapping table of the reverse change relationship between the reflectance of the visible light band and the radiation intensity of the infrared band with the same spatial coordinates.

[0066] In step 301, the reverse change relationship mapping table refers to a query table that records the corresponding relationship between the decrease in visible light reflectance and the increase in infrared radiation.

[0067] In the embodiments of the present application, first, the image space grid is divided according to the noise separation boundary conditions, and the corresponding relationship between the decrease in visible light reflectivity and the increase in infrared radiation is statistically analyzed within each grid cell. A continuous change relationship surface is generated through a spatial interpolation algorithm, and a mapping relationship table is established with spatial coordinates as the index and the change rate of reflectivity-radiation intensity as the value. This table is updated in real time to dynamically reflect the noise characteristics of different regions.

[0068] Step 302: According to the reverse change relationship mapping table, construct a joint determination rule for the attenuation amplitude of visible light reflectivity and the increase amplitude of infrared radiation intensity. The boundary conditions of the joint determination rule include a first threshold and a second threshold. The first threshold is dynamically calculated from the difference between the average reflectivity of the visible light band without dust interference and the current time-varying interference intensity. The second threshold is calculated by superimposing the radiation increment in the noise diffusion region on the reference intensity within the normal radiation range of the valve material in the infrared band.

[0069] In step 302, the joint determination rule represents the noise recognition logic that synthesizes visible light and infrared features. The first threshold represents the dynamic segmentation threshold of visible light reflectivity. The second threshold represents the dynamic segmentation threshold of infrared radiation intensity. The visible light band refers to the spectral acquisition channel with a wavelength range of 400 - 700 nm. The visible light band reflectivity represents the measured value of the reflection ability of the valve surface to incident light within this band. The attenuation amplitude of visible light reflectivity refers to the decreased value of the current reflectivity compared to the non-interference state. The current time-varying interference intensity specifically refers to the instantaneous interference intensity value calculated based on the real-time vibration frequency mutation point at a specific moment or time period, which is the specific instantiation of "time-varying interference intensity" on the time axis. Time-varying interference intensity refers to the dynamic change intensity of dust scattering noise in the time dimension determined by the time-varying curve of vibration frequency. The relationship between the two is that "the current time-varying interference intensity" is the real-time value of "time-varying interference intensity" during the dynamic noise separation process, and the former inherits the definition of the latter and adds a time point constraint. The infrared band refers to the spectral acquisition channel with a wavelength range of 700 nm - 1 mm (including the at least two sub-bands). The infrared band radiation intensity represents the measured value of the thermal radiation energy of the valve surface within this band. The increase amplitude of infrared radiation intensity refers to the increased value of the current radiation intensity compared to the reference state.

[0070] In the embodiments of the present application, based on the mapping table, the statistical distribution characteristics of reflectivity and radiation intensity are analyzed, and a bivariate joint probability model is established. The first threshold is calculated by subtracting 0.8 times the time-varying interference intensity (output from step 202) from the reference reflectivity (obtained from the valve material parameter library); the second threshold is calculated by adding 1.2 times the radiation increment in the noise region (estimated through a heat diffusion model) to the reference radiation intensity (calculated based on the valve operating temperature). The two thresholds are automatically calibrated once per hour.

[0071] Step 303: Screen out target pixels whose reflectance in the visible light band is lower than the first threshold and whose radiation intensity in the infrared band is higher than the second threshold.

[0072] In the embodiment of the present application, a parallel processing architecture is adopted to perform triple determination on each pixel point: first, check whether it is within the noise diffusion area, then verify whether the visible light reflectance is lower than the first threshold, and finally confirm whether the infrared radiation is higher than the second threshold. Pixels that meet all three conditions are marked as target pixels, and the spatial distribution map of the noise component is composed of all target pixels.

[0073] The following is a specific example: In the actual monitoring of the gas pipeline valve in a chemical plant, based on the established noise separation boundary conditions (time-varying interference intensity 0.72, noise diffusion area radius 15 cm), the system first constructs a mapping table of the inverse change relationship between visible light reflectance and infrared radiation intensity: in the flange connection area, the measured visible light reflectance drops from the reference value of 0.65 to 0.28 (calculation process: 0.65×(1 - 0.72 time-varying interference coefficient)×0.95 environmental correction factor), while the infrared radiation rises from the reference value of 0.45 to 0.81 (calculation process: 0.45 + 0.3 radiation increment×1.2 stress concentration coefficient). According to this mapping relationship, the system dynamically sets the first threshold to 0.38 (calculation: 0.65 reference reflectance - 0.72 interference intensity×0.38 adjustment coefficient), and the second threshold to 0.75 (calculation: 0.45 reference radiation + 0.25 minimum radiation increment×1.2 safety factor). By scanning the image of 1280×1024 pixels in the flange area, 2150 target pixels (visible light reflectance < 0.38 and infrared radiation > 0.75) are screened out. After morphological closing operation, it is confirmed that the corresponding dust-covered area is 4.2 cm². After removing the noise, the image clearly shows a 3.8-mm-long crack on the flange sealing surface, whose position exactly corresponds to the 25-MPa stress concentration area. The crack direction forms an angle of 18° with the main stress direction, and the crack width is 0.15 mm (obtained by converting pixel size to actual size), providing an accurate defect location basis for the emergency repair of the valve.

[0074] In the embodiment of the present application, through the dual verification of the inverse change characteristics of multi-spectral features and mechanical state data, the method realizes the accurate identification and separation of dust noise, enables the display clarity of the true defects on the valve surface to meet the requirements of industrial inspection standards, and effectively supports the valve health status assessment work in harsh environments.

[0075] In order to solve the problem of insufficient accuracy in separating multi-spectral image noise in a sandy dust environment, in some embodiments, step 301: establishing a mapping table of the inverse variation relationship between the reflectance in the visible light band and the radiation intensity in the infrared band at the same spatial coordinates based on the noise separation boundary conditions and the noise diffusion region, including: Step 401: Convert the time-varying interference intensity in the noise separation boundary conditions into a spatio-temporal matrix according to the distribution of amplitude mutation points of the vibration frequency time-variation curve.

[0076] In step 401, the amplitude mutation point represents a fluctuation point in the vibration frequency curve that exceeds 30% of the average amplitude. The spatio-temporal matrix represents a two-dimensional matrix reflecting the spatial distribution of the time-varying interference intensity. Wherein the row vector of the matrix represents the change in interference intensity in the time dimension, and the column vector represents the distribution of interference intensity in the spatial dimension.

[0077] In the embodiments of the present application, first, identify the amplitude mutation points in the vibration frequency curve, construct an initial matrix with the duration of the mutation points as rows and the spatial positions as columns, perform spatial interpolation filling on the time-varying interference intensity according to the influence range of the mutation points, and finally generate a spatio-temporal matrix with the same resolution as the image. The value of each element of this matrix represents the quantization value of the time-varying interference corresponding to the pixel point.

[0078] Step 402: Calculate the diffusion area ratio parameter at each spatial coordinate position according to the spatial coverage shape of the noise diffusion region.

[0079] In step 402, the spatial coverage shape represents the noise diffusion form determined by the direction of the mechanical stress gradient. The diffusion area ratio parameter is a parameter characterizing the influence degree of noise in a local area.

[0080] In the embodiments of the present application, with each pixel point as the center, take a circular area with a radius 1.2 times the extension length of the stress gradient, and calculate the pixel ratio of the area within this region that belongs to the noise diffusion region as the diffusion area ratio parameter. The bilinear interpolation algorithm is used to ensure the spatial continuity of the parameter.

[0081] Step 403: Perform a pixel-by-pixel multiplication operation on the spatio-temporal matrix and the diffusion area ratio parameter to generate a noise interference superposition weight map.

[0082] In step 403, the noise interference superposition weight map represents a weight distribution map that comprehensively considers time-varying and space-varying interferences.

[0083] In the embodiments of the present application, perform a dot multiplication operation on the spatio-temporal matrix and the diffusion area ratio parameter, and perform normalization processing on the result to generate a weight map within the range of 0-1. Among them, the area where the weight value exceeds 0.7 is determined as a high interference area, and the area below 0.3 is a low interference area.

[0084] Step 404: Based on the gradient distribution direction in the noise interference superposition weight map, identify low-reflectivity pixels in the visible light band reflectivity with an included angle of the gradient distribution direction greater than a set angle, and simultaneously identify high-radiation-intensity pixels in the infrared band radiation intensity with an included angle of the gradient distribution direction less than the set angle.

[0085] In step 404, the set angle represents the allowable deviation angle determined according to the stress gradient direction. Low-reflectivity pixels refer to the image pixels with reflectivity lower than the normal value in the visible light band, and their reflectivity values are usually lower than the dynamic threshold determined according to the valve material characteristics. The gradient distribution direction represents the main diffusion direction of noise interference in space. High-radiation-intensity pixels refer to the image pixels with radiation intensity higher than the normal value in the infrared band, and their radiation values usually exceed the dynamic threshold calculated from the valve operating temperature and material emissivity.

[0086] In the embodiment of the present application, first calculate the gradient field of the weight map to determine the main gradient direction. In the visible light image, screen the pixels with reflectivity lower than the mean value and the included angle between the gradient direction and the main direction greater than 25 degrees; in the infrared image, screen the pixels with radiation intensity higher than the mean value and the included angle less than 15 degrees.

[0087] Step 405: Construct an inverse variation relationship mapping table with the spatial overlap degree between the low-reflectivity pixels and the high-radiation-intensity pixels as the index key value.

[0088] In step 405, the spatial overlap degree represents the degree of consistency of the two features in the spatial position. The specific generation process of the spatial overlap degree: Set of low-reflectivity pixels: Pixels with reflectivity values lower than the dynamically calculated threshold screened from the visible light band reflectivity. Set of high-radiation-intensity pixels: Pixels with radiation intensity values higher than the dynamically calculated threshold screened from the infrared band radiation intensity. Perform binary determination on each spatial coordinate point: If the current coordinate exists in both the set of low-reflectivity pixels and the set of high-radiation-intensity pixels, it is marked as an overlapping pixel, otherwise it is marked as a non-overlapping pixel. Calculate the proportion of the number of overlapping pixels to the total number of pixels in the local area (3×3 pixel range) at the current coordinate of the noise interference superposition weight map as the spatial overlap degree at the current coordinate. When the mechanical stress spatial gradient direction is consistent with the maximum gradient direction of the noise diffusion region: Apply a weighting coefficient proportional to the gradient intensity to the calculation result of the overlap degree. When the time change rate of the vibration frequency exceeds the set fluctuation threshold: Apply an attenuation compensation of the time-varying interference intensity to the calculation result of the overlap degree. Linearly map the spatial overlap degrees of all coordinates to the [0,1] interval, and use the normalized value as the index key value of the inverse variation relationship mapping table. The index key value represents the query basis of the mapping table.

[0089] In the embodiments of the present application, the coexistence situation of low-reflectivity pixels and high-radiation-intensity pixels at each spatial coordinate point is statistically analyzed, and the coexistence pixel ratio within a 3×3 neighborhood is calculated as the overlap degree. A mapping relationship table of reflectivity-radiation intensity change rate is constructed based on this key value.

[0090] The following is a specific example: In a monitoring case of a gas pipeline valve in a chemical plant, the system first converts the time-varying interference intensity of 0.72 into a spatio-temporal matrix with a resolution of 1280×1024 according to the vibration frequency mutation points (125 Hz, lasting for 8 minutes). The value in the flange area of the matrix is 0.65 (calculation: 0.72×0.9 position attenuation coefficient). According to the stress distribution, the noise diffusion area ratio parameter of each pixel point is calculated for the noise diffusion area (a sector area with a radius of 15 cm). The parameter near the flange bolt hole reaches 0.85 (calculation: actual diffusion area 12.8 cm² / theoretical maximum area 15 cm²×π). After multiplying the spatio-temporal matrix and the diffusion parameter point by point, a noise interference superposition weight map is generated. The weight value above the flange in the map reaches 0.74 (0.65×0.85×1.34 stress gradient coefficient). Based on the gradient analysis of the weight map, a pixel set with a reflectivity of 0.26 (reference 0.65×0.4) and a gradient direction angle of 32° (accounting for 7.2% of the total pixels) is identified in the visible light image; a pixel set with a radiation intensity of 0.83 (reference 0.45 + 0.38) and an angle of 12° (accounting for 8.5%) is identified in the infrared image. The spatial overlap degree of the two feature sets within a 3×3 neighborhood is calculated as 0.71 (number of overlapping pixels / total number of pixels). The constructed inverse change mapping table shows that when the reflectivity drops to 0.3, the corresponding radiation intensity rises to 0.8±0.05.

[0091] In the embodiments of the present application, through vibration-stress collaborative noise modeling and spatial correlation analysis of multi-spectral features, the method realizes the accurate quantification and positioning of dust interference, enables the detection rate of true valve defects to meet the requirements of industrial detection standards, and provides reliable technical support for equipment status assessment in harsh environments.

[0092] To solve the problem of insufficient accuracy in identifying the noise boundary in a dust environment, in some embodiments, step 203: The coupling calculation of the time-varying interference intensity and the noise diffusion area to determine the noise separation boundary condition includes: Step 501: Based on the time-varying curve of the vibration frequency, calculate the instantaneous influence value of the time-varying interference intensity.

[0093] In step 501, the instantaneous influence value represents the intensity of the action of the time-varying interference intensity on a spatial point at a specific moment.

[0094] In the embodiment of the present application, first, a first-order difference processing is performed on the vibration frequency-time change curve, and the time period with the amplitude change rate exceeding the set threshold is extracted as the effective interval. In each effective interval, the product of the square of the vibration amplitude and the duration is used as the basic influence value, and then multiplied by the valve structure transfer coefficient (obtained from the pre-stored data of the finite element model) to finally obtain the instantaneous influence value matrix of each spatial point.

[0095] Step 502: According to the spatial gradient distribution of the mechanical stress, divide the noise diffusion region into a core influence area and a marginal transition area.

[0096] In step 502, the core influence area represents the area with the largest mechanical stress gradient and the most serious noise interference. The range of the core influence area is determined by the product of the extension length and the elastic modulus of the valve surface material. The marginal transition area represents the surrounding area where the noise influence gradually weakens.

[0097] In the embodiment of the present application, according to the spatial gradient distribution of the mechanical stress, the area with the gradient value exceeding 1.5 times the average gradient is designated as the core influence area (accounting for about 30% of the area), and the rest is the marginal transition area. The boundary of the core influence area is processed by morphological dilation based on the gradient direction to ensure that the stress concentration area is completely included.

[0098] Step 503: In the core influence area, perform a convolution operation on the time-varying interference intensity and the spatial stress gradient to generate the interference intensity of the core influence area, and perform extended compensation in the marginal transition area to obtain the interference intensity of the marginal transition area.

[0099] In step 503, the interference intensity of the core influence area represents a quantitative index reflecting the comprehensive noise level of the core area. The extended compensation adopts a non-linear interpolation method based on the amplitude of the mechanical stress gradient: in the marginal transition area, based on the stress gradient value at the boundary of the core influence area, the compensation intensity decays according to the inverse square relationship of the distance from the boundary; at the same time, the first derivative of the vibration frequency-time change curve is superimposed as a dynamic adjustment factor, and when the derivative exceeds the threshold, a morphological dilation operation (3×3 circular structuring element) is started to ensure the spatial continuity of the noise diffusion area. The interference intensity of the marginal transition area refers to the quantitative characterization value during the process of the noise influence gradually decaying from the core area to the periphery.

[0100] In the embodiment of the present application, in the core area, a 3×3 convolution kernel is used to perform a spatial convolution operation on the time-varying interference intensity and the stress gradient, and the weight coefficients of the convolution kernel are dynamically adjusted according to the stress gradient direction. In the marginal transition area, an exponential decay compensation algorithm based on distance is adopted, and the compensation intensity is inversely proportional to the distance from the boundary of the core area, and the maximum compensation amount is 0.6 times the interference intensity of the core area.

[0101] Step 504: Adjust the numerical value of the parameter used to separate the core influence area and the edge transition area according to the interference intensity of the core influence area and the interference intensity of the edge transition area, so as to obtain the noise separation boundary condition.

[0102] In step 504, the noise separation boundary condition is implemented by a two-parameter adaptive threshold function: Let the interference intensity of the core area be Ic and the interference intensity of the edge area be Ie, then the boundary condition is α·Ic+(1-α)·Ie>T, where the weight coefficient α is determined by the consistency of the mechanical stress gradient direction (0.6 - 0.9), and the dynamic threshold T = κ·(1 + e^(-βΔt)), κ is a material-related constant, β is the attenuation rate of the time-varying interference intensity, and Δt is the time difference between the current moment and the vibration mutation point.

[0103] In the embodiment of the present application, the interference intensity of the core area and the interference intensity of the edge area are input into a double-threshold adaptive model. The core area adopts a rigid boundary (threshold coefficient 1.0), and the edge area adopts a flexible boundary (threshold coefficient 0.7). The boundary parameters are adjusted through an iterative optimization algorithm to make the final boundary continuous in space and geometrically consistent with the mechanical stress distribution.

[0104] The following is a specific example: In the actual monitoring of the gas pipeline valve in a chemical plant, the system calculates the instantaneous influence value of 0.78 based on the vibration frequency time-varying curve (abnormal vibration at 125Hz for 12 minutes, amplitude 0.85mm / s²) (calculation formula: 0.5×vibration amplitude coefficient + 0.5×duration coefficient, where the amplitude coefficient = measured amplitude / reference amplitude = 0.85 / 0.6 = 1.42, and the duration coefficient = actual duration / standard sampling duration = 12 / 10 = 1.2). According to the mechanical stress spatial gradient distribution (the maximum gradient at the lower part of the flange is 28MPa / m, and the extension length is 10cm), a circular area with a radius of 12cm is defined as the core influence area (area 113cm², calculation: π×12²), and the outer 3cm annular belt is the edge transition area. In the core area, a 3×3 convolution kernel (weight coefficient [0.2,0.3,0.2;0.3,0.5,0.3;0.2,0.3,0.2]) is used to perform convolution operations on the time-varying interference intensity and the stress gradient, and the interference intensity of the core area is obtained as 0.82 (calculation: 0.78×0.7 + 28 / 35×0.3). The edge transition area adopts distance attenuation compensation (compensation amount = core intensity×e^(-0.15×distance)), and the interference intensity measured 1cm away from the core area boundary is 0.58 (calculation: 0.82×e^(-0.15×1)). The finally generated noise separation boundary condition shows that the core area threshold is 0.75 (calculation: 0.82×0.9 safety factor), and the edge area threshold is 0.52 (calculation: 0.58×0.9).

[0105] In the embodiments of the present application, through the spatio-temporal coupling calculation of vibration and stress characteristics, the accurate division of the noise boundary is achieved, the recognition accuracy of the true defects of the valve reaches the requirements of industrial inspection standards, and the technical problem of equipment condition assessment in harsh environments is effectively solved.

[0106] To solve the problem of insufficient fusion of multi-spectral images and the mechanical state of the valve, in some embodiments, step 105: generating an enhanced optical image matching the actual mechanical state of the valve based on the spectral fusion weight and the corrected multi-band texture features, includes: Step 601: Dynamically adjust the spectral fusion weight according to the mechanical dynamic response characteristics.

[0107] In the embodiments of the present application, first, the main frequency change rate of the vibration signal and the gradient change amount of the stress distribution are analyzed, and the reliability coefficients of each band image are calculated. Visible light band weight = 0.6 × vibration stability coefficient + 0.4 × stress uniformity coefficient; Infrared band weight = 0.7 × thermal radiation correlation coefficient + 0.3 × vibration-thermal coupling coefficient. The weight value is dynamically updated once per hour to ensure adaptation to working condition changes.

[0108] Step 602: According to the corrected multi-band texture features, perform hierarchical fusion on the multi-spectral optical image data according to the adjusted spectral fusion weight.

[0109] In step 602, hierarchical fusion represents the integration of multi-spectral data classified by the importance of features.

[0110] In the embodiments of the present application, the multi-spectral data is processed in three layers: the basic layer (80% weight) fuses visible light texture and infrared radiation distribution; the detail layer (15% weight) strengthens anisotropic features; the abnormal layer (5% weight) highlights the features of stress concentration areas. Each layer uses a fusion algorithm based on wavelet transform, and the fusion parameters are dynamically adjusted according to the clarity of the texture features and the matching degree with the mechanical state.

[0111] Step 603: Based on the mechanical stress space gradient distribution, mark potential structural abnormal areas in the hierarchically fused multi-spectral optical image data.

[0112] In step 603, the potential structural abnormal area represents a high-risk area where defects may exist.

[0113] In the embodiments of the present application, first, identify the texture abnormal areas (such as fractures, mutations, etc.) in the fused image, and then perform spatial matching analysis with the stress gradient distribution. When the spatial coincidence degree between the texture abnormal area and the high stress gradient area (gradient value > 25 MPa / m) exceeds 70%, it is marked as a potential structural abnormal area. The marking intensity is divided into three levels, corresponding to different maintenance priorities.

[0114] Step 604: Generate an enhanced optical image that matches the actual mechanical state of the valve based on the multi-spectral optical image data marked with the potential structural anomaly area.

[0115] In step 604, the specific generation process of the enhanced optical image is as follows: perform the following processing on the multi-spectral optical image data marked with the potential structural anomaly area: First, perform anisotropic diffusion filtering based on texture sharpening on the visible light band (the diffusion coefficient is inversely proportional to the amplitude of the mechanical stress gradient); Second, perform radiation intensity normalization on the infrared band (based on the standard value in the radiation characteristic database of the valve material); Finally, perform pixel-level weighted superposition on the processed multi-band data according to the spectral fusion weight, where the fusion weight of the pixels in the anomaly area is additionally increased by a dynamic compensation coefficient proportional to the amplitude of the vibration frequency, and the final enhanced optical image is output.

[0116] In the embodiment of the present application, perform directional enhancement processing on the marked area: enhance the edge sharpness in the visible light channel (enhancement coefficient 1.2 - 1.5), and enhance the radiation contrast in the infrared channel (enhancement coefficient 1.3 - 1.8). At the same time, maintain the natural transition of the non-anomaly area, and finally generate an enhanced image that not only maintains optical authenticity but also highlights the mechanical state-related features.

[0117] The following is a specific example: In the monitoring of the gas transmission pipeline valve in a chemical plant, the system dynamically adjusts the spectral fusion weight according to the real-time collected vibration data (the main frequency suddenly increases from the normal 110Hz to 148Hz) and stress data (the stress concentration at the lower part of the flange reaches 28MPa): the visible light weight is set to 0.4 (calculation: 0.6 × vibration stability coefficient 0.5 + 0.4 × stress uniformity coefficient 0.25), and the mid-wave infrared weight is set to 0.6. In the hierarchical fusion process, the basic layer uses a 0.7 weight to fuse the visible light texture (reflectivity 0.35) and the mid-wave infrared feature (radiation intensity 0.78), the detail layer uses a 0.25 weight to enhance the anisotropic feature (anisotropy ratio 2.8) along the main stress direction (azimuth angle 135°), and the anomaly layer uses a 0.05 weight to highlight the feature of the stress concentration area (gradient value 26MPa / m). After fusion, 3 potential anomaly areas are identified on the flange sealing surface, and one of them coincides with the 32MPa stress point by 85% (calculation: the number of overlapping pixels / the total number of pixels in the stress area), which is marked as a first-level anomaly. Perform directional enhancement processing on this area: the visible light edge sharpening coefficient is 1.3 (benchmark value 1.0 + 0.3 × stress gradient coefficient), and the long-wave infrared contrast is increased by 1.4 times. The finally generated enhanced image clearly shows two cracks (lengths 4.2mm and 3.5mm, width 0.2mm) on the flange sealing surface, and the stress at the crack tip reaches 35MPa.

[0118] In the embodiments of the present application, the method realizes the accurate correlation expression of the surface topography features and the internal stress state of the valve through mechanically state-guided multi-spectral fusion and directional enhancement. The generated enhanced image meets the requirements of both visual interpretation and quantitative analysis, improving the reliability of valve defect detection in harsh environments.

[0119] To solve the problem of insufficient matching degree between multi-band texture features and the mechanical state of the valve, in some embodiments, step 103: The dual amplitude correction of the multi-band texture features in the spatial domain and the frequency domain includes: Step 701: Based on the mechanical dynamic response features, extract the main peak value of the energy distribution of the multi-band texture features in the frequency domain, and use the matching degree between the amplitude value of the frequency component corresponding to the main peak value of the energy distribution and the natural vibration frequency of the valve material as the frequency domain amplitude correction coefficient.

[0120] In step 701, the main peak value of the energy distribution represents the frequency component with the largest amplitude in the energy spectrum of the multi-band texture features in the frequency domain. The frequency domain amplitude correction coefficient is a parameter characterizing the matching degree between the characteristic frequency and the natural frequency of the material.

[0121] In the embodiments of the present application, first, perform a fast Fourier transform on the multi-band texture features to obtain the frequency domain energy spectrum, and identify the frequency component corresponding to the main peak value of the energy distribution through a peak detection algorithm. Match this frequency with the pre-stored database of the natural vibration frequencies of the valve material, and calculate the matching degree (calculation formula: matching degree = 1 - |measured frequency - natural frequency| / natural frequency), which is used as the frequency domain amplitude correction coefficient. This coefficient will be used for subsequent frequency domain amplitude suppression.

[0122] Step 702: Calculate the anisotropy intensity ratio of the multi-band texture features in the spatial domain according to the maximum gradient direction of the mechanical stress spatial gradient distribution.

[0123] In step 702, the maximum gradient direction represents the direction with the largest amplitude of the mechanical stress spatial gradient. The anisotropy intensity ratio refers to the ratio of the average intensity value of the multi-band texture features parallel to the maximum gradient direction of the mechanical stress to the average intensity value perpendicular to this direction, which is used to quantify the directional difference of the texture features.

[0124] In the embodiments of the present application, first calculate the gray level co-occurrence matrix of the texture features in the spatial domain, and extract the contrast features in four directions of 0°, 45°, 90°, and 135°. At the same time, extract the maximum gradient direction (such as 120°) from the mechanical stress spatial gradient distribution. Calculate the ratio of the average intensity of all directions with an angle less than 15° to the maximum gradient direction to the average intensity in the perpendicular direction as the anisotropy intensity ratio.

[0125] Step 703: In the spatial domain, enhance and correct the amplitude of the multi-band texture features along the maximum gradient direction in direct proportion to the anisotropy intensity ratio, and perform attenuation correction in the direction perpendicular to the maximum gradient direction in inverse proportion to the anisotropy intensity ratio. In the frequency domain, apply amplitude suppression to the multi-band texture features in inverse proportion to the frequency domain amplitude correction coefficient to obtain the modified multi-band texture features.

[0126] In step 703, the enhancement correction represents texture strengthening along the principal stress direction. The amplitude suppression represents the attenuation of non-material feature frequencies.

[0127] In the embodiment of the present application, in the spatial domain, a directional filter bank is used to perform band-pass filtering (passband width = anisotropy intensity ratio × 15°) along the maximum gradient direction to enhance the texture features in this direction; perform low-pass filtering in the vertical direction to suppress irrelevant details. In the frequency domain, weight the energy spectrum, and the weight coefficient = (1 - frequency domain amplitude correction coefficient) × 0.8 + 0.2, retain the frequency components of the material features, and suppress other frequency components. Finally, the modified multi-band texture features are obtained through inverse transformation.

[0128] The following is a specific example: In a monitoring case of the gas pipeline valve in a chemical plant, the system performs frequency domain analysis on the denoised multi-band texture features and finds that the main peak of energy is located at 126 Hz (the inherent frequency of the stainless steel valve is 128 Hz), and the matching degree reaches 0.98 (calculation: 1 - |126 - 128| / 128). The stress distribution shows that there is a stress concentration area with a maximum gradient direction of 50° (gradient value 24 MPa / m) around the flange bolt holes of the valve. The measured mean texture intensity along the 50° direction in the spatial domain is 82, and the vertical direction is 35, and the anisotropy intensity ratio is 2.34 (calculation: 82 / 35). During the correction process: in the frequency domain, retain the 120 - 132 Hz frequency band (weight coefficient 0.81, calculation: (1 - 0.98) × 0.8 + 0.2 = 0.81), and attenuate other frequency bands to 30% of the original amplitude; in the spatial domain, perform directional enhancement with a passband of 50° ± 17.55° (2.34 × 7.5°) along the 50° direction, and perform Gaussian smoothing with σ = 2.34 in the vertical direction. After processing, the image clearly shows two defects on the flange sealing surface: one is a radial crack with a length of 3.5 mm (deviating 3° from the principal stress direction), and the other is an erosion pit with a diameter of 2.1 mm (located at the center of the 26 MPa stress point).

[0129] In the embodiment of the present application, through the spatial-frequency domain collaborative correction guided by the mechanical state, the texture features not only retain the inherent frequency domain characteristics of the material but also strengthen the spatial features related to the stress distribution, improving the accuracy and reliability of valve defect detection.

[0130] Figure 2 The structural schematic diagram of a valve optical image enhancement and recognition system based on multi - spectral fusion provided by an embodiment of the present application is as follows Figure 2 shown. The system includes: An acquisition module 21, configured to acquire multi - spectral optical image data and mechanical dynamic response characteristics of an industrial pipeline valve under adverse weather conditions.

[0131] A separation module 22, configured to separate the multi - spectral optical image data based on the mechanical dynamic response characteristics to extract multi - band texture characteristics associated with the reflection characteristics of the valve material.

[0132] A correction module 23, configured to perform double - amplitude correction on the multi - band texture characteristics in the spatial domain and the frequency domain.

[0133] An allocation module 24, configured to allocate spectral fusion weights to the multi - spectral optical image data according to the corrected multi - band texture characteristics.

[0134] A generation module 25, configured to generate an enhanced optical image that matches the actual mechanical state of the valve based on the spectral fusion weights and the corrected multi - band texture characteristics.

[0135] Figure 2 The valve optical image enhancement and recognition system based on multi - spectral fusion can execute Figure 1 the valve optical image enhancement and recognition method based on multi - spectral fusion described in the embodiments shown. Its implementation principle and technical effects will not be elaborated further. For the valve optical image enhancement and recognition system based on multi - spectral fusion in the above - mentioned embodiments, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0136] In a possible design, Figure 2 the valve optical image enhancement and recognition system based on multi - spectral fusion described in the embodiments shown can be implemented as a computing device, as Figure 3 shown. The computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, and the one or more computer instructions are called and executed by the processing component 32.

[0137] The processing component 32 performs Figure 1 the valve optical image enhancement and recognition method based on multi - spectral fusion described in the above - mentioned embodiments.

[0138] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0139] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disc.

[0140] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0141] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0142] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0143] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0144] The embodiments of the present application also provide a computer storage medium storing a computer program, and when the computer program is executed by a computer, the above Figure 1 optical image enhancement and recognition method for valves in bad weather based on multispectral fusion shown in the embodiments can be implemented.

[0145] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0147] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. An optical image enhancement and recognition method for valves under adverse weather conditions based on multispectral fusion, characterized in that Including: Under harsh weather conditions, obtaining multi-spectral optical image data and mechanical dynamic response characteristics of industrial pipeline valves; Based on the mechanical dynamic response characteristics, separating the multi-spectral optical image data to extract multi-band texture features associated with the reflection characteristics of the valve material; Performing double amplitude correction in the spatial domain and frequency domain on the multi-band texture features; According to the corrected multi-band texture features, assigning spectral fusion weights to the multi-spectral optical image data; Based on the spectral fusion weights and the corrected multi-band texture features, generating an enhanced optical image that matches the actual mechanical state of the valve.

2. The method according to claim 1, wherein The separating the multi-spectral optical image data based on the mechanical dynamic response characteristics to extract multi-band texture features associated with the reflection characteristics of the valve material includes: Constructing a dynamic noise separation model related to dust scattering interference, where the parameters of the dynamic noise separation model are jointly determined by the vibration frequency time-varying curve and the mechanical stress spatial gradient distribution in the mechanical dynamic response characteristics; According to the dynamic noise separation model, determining the time-varying interference intensity and noise diffusion region of dust scattering noise in the multi-spectral optical image data; Performing coupling calculation on the time-varying interference intensity and the noise diffusion region to determine the noise separation boundary condition; According to the noise separation boundary condition, performing cross-band correlation analysis on the visible light band reflectance and the infrared band radiation intensity at the same spatial coordinates in the multi-spectral optical image data to screen out target pixels with the visible light band reflectance lower than the first threshold and the infrared band radiation intensity higher than the second threshold, and taking the set composed of all target pixels as the dust scattering noise component; Extracting multi-band texture features from the multi-spectral optical image data after removing the dust scattering noise component.

3. The method according to claim 2, characterized in that The performing cross-band correlation analysis on the visible light band reflectance and the infrared band radiation intensity at the same spatial coordinates in the multi-spectral optical image data according to the noise separation boundary condition to screen out target pixels with the visible light band reflectance lower than the first threshold and the infrared band radiation intensity higher than the second threshold includes: Based on the noise separation boundary condition and the noise diffusion region, establishing an inverse change relationship mapping table between the visible light band reflectance and the infrared band radiation intensity at the same spatial coordinates; According to the inverse change relationship mapping table, constructing a joint determination rule for the attenuation amplitude of the visible light reflectance and the increase amplitude of the infrared radiation intensity. The boundary conditions of the joint determination rule include the first threshold and the second threshold. The first threshold is dynamically calculated from the difference between the average reflectance of the visible light band without dust interference and the current time-varying interference intensity, and the second threshold is obtained by superimposing the radiation increment in the noise diffusion region on the reference intensity within the normal radiation range of the valve material in the infrared band; Screening out target pixels with the visible light band reflectance lower than the first threshold and the infrared band radiation intensity higher than the second threshold.

4. The method according to claim 3, wherein Based on the noise separation boundary condition and the noise diffusion region, establishing a mapping table of the reverse variation relationship between the visible light band reflectivity and the infrared band radiation intensity under the same spatial coordinates, including: Converting the time-varying interference intensity in the noise separation boundary condition into a spatio-temporal matrix according to the amplitude mutation point distribution of the vibration frequency time variation curve; Calculating the diffusion area ratio parameter of each spatial coordinate position according to the spatial coverage shape of the noise diffusion region; Performing a pixel-by-pixel multiplication operation on the spatio-temporal matrix and the diffusion area ratio parameter to generate a noise interference superposition weight map; Based on the gradient distribution direction in the noise interference superposition weight map, identifying low reflectivity pixels with a gradient distribution direction angle greater than a set angle in the visible light band reflectivity, and simultaneously identifying high radiation intensity pixels with a gradient distribution direction angle less than the set angle in the infrared band radiation intensity; Constructing a reverse variation relationship mapping table with the spatial overlap degree of the low reflectivity pixels and the high radiation intensity pixels as the index key value.

5. The method according to claim 2, characterized in that, Coupling and calculating the time-varying interference intensity and the noise diffusion region to determine the noise separation boundary condition, including: Calculating the instantaneous influence value of the time-varying interference intensity based on the vibration frequency time variation curve; Dividing the noise diffusion region into a core influence area and an edge transition area according to the mechanical stress spatial gradient distribution; In the core influence area, performing a convolution operation on the time-varying interference intensity and the spatial stress gradient to generate the interference intensity of the core influence area, and performing extended compensation in the edge transition area to obtain the interference intensity of the edge transition area; Adjusting the numerical value of the parameter used to separate the core influence area and the edge transition area according to the interference intensity of the core influence area and the interference intensity of the edge transition area to obtain the noise separation boundary condition.

6. The method according to claim 2, characterized in that, Generating an enhanced optical image matching the actual mechanical state of the valve based on the spectral fusion weight and the corrected multi-band texture features, including: Dynamically adjusting the spectral fusion weight according to the mechanical dynamic response characteristics; According to the corrected multi-band texture features, performing hierarchical fusion on the multi-spectral optical image data according to the adjusted spectral fusion weight; Based on the mechanical stress spatial gradient distribution, marking potential structural abnormal areas in the hierarchically fused multi-spectral optical image data; Generating an enhanced optical image matching the actual mechanical state of the valve based on the multi-spectral optical image data marked with the potential structural abnormal areas.

7. The method according to claim 1, wherein Performing double amplitude correction on the multi-band texture features in the spatial domain and the frequency domain, including: Based on the mechanical dynamic response characteristics, extracting the main peak value of the energy distribution of the multi-band texture features in the frequency domain, and using the matching degree between the amplitude of the frequency component corresponding to the main peak value of the energy distribution and the natural vibration frequency of the valve material as the frequency domain amplitude correction coefficient; Calculating the anisotropy intensity ratio of the multi-band texture features in the spatial domain according to the maximum gradient direction of the mechanical stress spatial gradient distribution; In the spatial domain, the amplitude of the multi-band texture feature is enhanced and corrected in direct proportion to the anisotropy intensity ratio along the maximum gradient direction, and is attenuated and corrected in inverse proportion to the anisotropy intensity ratio perpendicular to the maximum gradient direction. In the frequency domain, amplitude suppression inversely proportional to the frequency domain amplitude correction coefficient is applied to the multi-band texture feature to obtain the modified multi-band texture feature.

8. An optical image enhancement and recognition system for valves in bad weather conditions based on multispectral fusion, characterized in that, It includes: An acquisition module, configured to acquire multi-spectral optical image data and mechanical dynamic response characteristics of an industrial pipeline valve under bad weather conditions; A separation module, configured to separate the multi-spectral optical image data based on the mechanical dynamic response characteristics to extract multi-band texture features associated with the reflection characteristics of the valve material; A correction module, configured to perform double amplitude correction in the spatial domain and the frequency domain on the multi-band texture feature; An allocation module, configured to allocate spectral fusion weights to the multi-spectral optical image data according to the corrected multi-band texture feature; A generation module, configured to generate an enhanced optical image matching the actual mechanical state of the valve based on the spectral fusion weights and the corrected multi-band texture feature.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for enhancing and identifying valve optical images under bad weather conditions based on multi-spectral fusion as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by the computer, it implements a method for enhancing and identifying valve optical images under bad weather conditions based on multi-spectral fusion as described in any one of claims 1 to 7.

Images