Remote automatic acquisition system and method for pressure gauge data

Abstract

The invention discloses a remote automatic acquisition system and method for pressure gauge data, and the method comprises the steps: obtaining pointer type pressure gauge image data collected by a camera, optimizing the feature extraction of a pressure gauge image through the transformation of a polar coordinate system, and further enhancing the precision and robustness of pointer recognition through local-global feature phase alignment. Therefore, the reading accuracy of the reading value of the pressure gauge is improved. Furthermore, a dynamic analysis model is constructed based on the time sequence distribution of the reading values of the pressure gauge to distinguish the real equipment fault from the environment interference, and the false alarm rate is reduced. Thus, the system can significantly improve the accuracy of reading of the pressure gauge under complex working conditions, realizes reliable early warning of early hidden dangers through an anomaly detection mechanism combining a dynamic threshold value and a statistical rule, reduces manual inspection requirements, enhances the real-time performance and the intelligent level of equipment state monitoring, and improves the working efficiency. And efficient technical support is provided for safety management of an industrial site.

Description

Technical Field

[0001] The present application relates to the field of intelligent data collection, and more specifically, to a system and method for remote automatic data collection of pressure gauges. Background Art

[0002] As a key instrument for monitoring pressure status in industrial sites, pressure gauges are widely used in the fields of electricity, chemical industry, energy, etc. However, traditional pressure gauges can usually only provide local readings, lack real-time remote data transmission and abnormal warning capabilities, resulting in equipment status monitoring being highly dependent on manual inspections. Although there are remote meter reading systems in the prior art, such as monitoring solutions based on dual-channel communication, their functions are limited to data collection and transmission, and they cannot perform abnormal diagnosis for dynamic changes in pressure data, especially in complex working conditions, where it is difficult to distinguish between environmental interference and real equipment failures.

[0003] For pointer pressure gauges, existing image recognition technologies mostly rely on feature extraction in a plane coordinate system, which is easily affected by perspective deviation or dial deformation, resulting in insufficient reading accuracy. In addition, existing systems mostly use static thresholds when determining abnormalities, lack dynamic analysis based on statistical laws, and have a high false alarm rate. These problems not only increase the manpower cost of equipment management, but may also lead to hidden dangers not being dealt with in a timely manner, causing safety accidents.

[0004] Therefore, an optimized remote automated data collection method for pressure gauges is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a remote automatic acquisition system and method for pressure gauge data, which obtains the pointer pressure gauge image data collected by the camera, optimizes the pressure gauge image feature extraction through polar coordinate system transformation, and further utilizes the local-global feature phase alignment to enhance the accuracy and robustness of pointer recognition, thereby improving the reading accuracy of the pressure gauge reading value. Furthermore, a dynamic analysis model is constructed based on the time series distribution of the pressure gauge reading value to distinguish between real equipment failures and environmental interference and reduce the false alarm rate. In this way, the system can significantly improve the accuracy of pressure gauge readings under complex working conditions, and through the abnormality detection mechanism combining dynamic thresholds with statistical laws, a reliable early warning of hidden dangers can be achieved, while reducing the need for manual inspections, enhancing the real-time and intelligent level of equipment status monitoring, and providing efficient technical support for industrial site safety management.

[0006] According to one aspect of the present application, a method for remote automatic data collection of a pressure gauge is provided, comprising: Acquire the image data of the pointer pressure gauge collected by the camera; Performing coordinate system transformation on the pointer pressure gauge image data to transform the pointer pressure gauge image data from a plane coordinate system to a polar coordinate system to obtain pointer pressure gauge coordinate system transformed image data; Inputting the coordinate system transformed image data of the pointer pressure gauge into an image recognition network to obtain a pressure gauge reading value, including: determining the pressure gauge reading value based on the phase alignment of local-global features in the coordinate system transformed image data of the pointer pressure gauge; Based on the time series distribution of the pressure gauge readings, determine whether the pressure data is in an abnormal state.

[0007] According to another aspect of the present application, a pressure gauge data remote automatic acquisition system is provided, comprising: An image acquisition module, used to acquire the image data of the pointer pressure gauge collected by the camera; A coordinate system transformation module, used for performing coordinate system transformation on the pointer pressure gauge image data to transform the pointer pressure gauge image data from a plane coordinate system to a polar coordinate system to obtain the pointer pressure gauge coordinate system transformation image data; An image recognition module is used to input the coordinate system transformation image data of the pointer pressure gauge into the image recognition network to obtain the pressure gauge reading value; The abnormality judgment module is used to judge whether the pressure data is in an abnormal state based on the time series distribution of the pressure gauge reading value.

[0008] Compared with the prior art, the present application provides a remote automated data acquisition system and method for pressure gauges, which obtains the image data of the pointer-type pressure gauge collected by the camera, optimizes the pressure gauge image feature extraction through polar coordinate system transformation, and further utilizes the local-global feature phase alignment to enhance the accuracy and robustness of pointer recognition, thereby improving the reading accuracy of the pressure gauge reading value. Furthermore, a dynamic analysis model is constructed based on the time series distribution of the pressure gauge reading value to distinguish between real equipment failures and environmental interference and reduce the false alarm rate. In this way, the system can significantly improve the accuracy of the pressure gauge readings under complex working conditions, and through the abnormality detection mechanism that combines dynamic thresholds with statistical laws, it can achieve reliable early warning of hidden dangers, while reducing the need for manual inspections, enhancing the real-time and intelligent level of equipment status monitoring, and providing efficient technical support for industrial site safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 It is a flow chart of a method for remote automatic collection of pressure gauge data according to an embodiment of the present application; Figure 2 A data flow diagram of a method for remote automatic data collection of pressure gauges according to an embodiment of the present application; Figure 3 It is a flowchart of sub-step S3 of the method for remote automatic collection of pressure gauge data according to an embodiment of the present application; Figure 4 It is a flowchart of sub-step S32 of the method for remote automatic collection of pressure gauge data according to an embodiment of the present application; Figure 5 It is a block diagram of a remote automated data collection system for pressure gauges according to an embodiment of the present application. DETAILED DESCRIPTION

[0011] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0012] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0013] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0014] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0015] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0016] In the technical solution of the present application, a method for remote and automated collection of pressure gauge data is proposed. Figure 1 A flowchart of a method for remote automated data collection of pressure gauges according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the method for remote automatic data collection of pressure gauges according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the remote automatic acquisition method of pressure gauge data according to the embodiment of the present application includes the steps of: S1, acquiring the pointer pressure gauge image data acquired by the camera; S2, performing coordinate system transformation on the pointer pressure gauge image data to convert the pointer pressure gauge image data from the plane coordinate system to the polar coordinate system to obtain the pointer pressure gauge coordinate system transformed image data; S3, inputting the pointer pressure gauge coordinate system transformed image data into the image recognition network to obtain the pressure gauge reading value; S4, judging whether the pressure data is in an abnormal state based on the time series distribution of the pressure gauge reading value.

[0017] In particular, S1 obtains the image data of the pointer pressure gauge collected by the camera. The image data of the pointer pressure gauge carries key information such as dial scale, pointer angle, logo text and environmental background. Considering that traditional manual inspection relies on reading the pressure value with the naked eye, it is impossible to achieve real-time remote monitoring, and the image data, as a digital carrier, can completely preserve the instantaneous state of the pressure gauge and provide original input for subsequent automated processing.

[0018] In particular, the S2 performs a coordinate system transformation on the pointer pressure gauge image data to transform the pointer pressure gauge image data from a plane coordinate system to a polar coordinate system to obtain the pointer pressure gauge coordinate system transformation image data. It should be understood that in an industrial site, the camera may cause the pointer pressure gauge image to have problems such as tilt and perspective deformation due to the installation position or angle limitation, and it is difficult to effectively eliminate such geometric distortions in the image processing under the traditional plane coordinate system. Therefore, in the technical solution of the present application, the pointer pressure gauge image data is transformed by a coordinate system transformation to transform the pointer pressure gauge image data from a plane coordinate system to a polar coordinate system to obtain the pointer pressure gauge coordinate system transformation image data. It should be understood that the dial of the pointer pressure gauge is essentially a circular structure, and its scale distribution and pointer movement are both angularly expanded with the center of the circle as the center. The rectangular characteristics of the plane coordinate system cannot directly adapt to such annular features, which can easily cause the scale line to bend or the pointer angle calculation deviation. By converting the image from a plane coordinate system to a polar coordinate system, the dial can be expanded into a two-dimensional mapping of angle-radius with the center of the circle as the origin, so that the elliptical dial caused by the tilt of the viewing angle can be restored to a standard circular distribution, and the position information of the scale line and the pointer can be recalibrated to a radial linear relationship. The coordinate system transformation image data of the pointer pressure gauge after the coordinate system transformation can more clearly present the spatial correlation between local features (such as the position of the pointer tip) and global features (such as the complete scale ring), providing a geometric consistency basis for phase alignment, thereby improving the robustness of the image recognition network for the interpretation of the pointer angle, reducing the risk of misidentification caused by dial occlusion, light reflection or mechanical vibration, ensuring the stability and accuracy of pressure readings under complex working conditions, and providing reliable underlying data support for dynamic anomaly analysis.

[0019] In particular, in S3, the coordinate system transformed image data of the pointer pressure gauge is input into the image recognition network to obtain the pressure gauge reading value. That is, in the technical solution of the present application, the pressure gauge reading value is determined based on the phase alignment of the local-global features in the coordinate system transformed image data of the pointer pressure gauge. Specifically, in a specific example of the present application, Figure 3 As shown, the S3 includes: S31, respectively extracting the local features and the global features of the pointer pressure gauge coordinate system transformation image data to obtain the local coding features of the pointer pressure gauge image and the global coding features of the pointer pressure gauge image; S32, performing image local-global joint perceptual alignment coding on the local coding features of the pointer pressure gauge image and the global coding features of the pointer pressure gauge image to obtain the local-global joint significant perceptual coding features of the pointer pressure gauge image; S33, determining the pressure gauge reading value based on the local-global joint significant perceptual coding features of the pointer pressure gauge image.

[0020] Specifically, the S31 extracts the local features and global features of the pointer pressure gauge coordinate system transformed image data respectively to obtain the pointer pressure gauge image local coding features and the pointer pressure gauge image global coding features. In an embodiment of the present application, first, the pointer pressure gauge coordinate system transformed image data is passed through an image local feature extractor based on a hole convolutional neural network model to obtain a pointer pressure gauge image local feature coding map as the pointer pressure gauge image local coding feature; it should be understood that although the polar coordinate system image of the pointer pressure gauge corrects the geometric distortion of the dial through transformation, its local details (such as the tip of the pointer, the edge of the scale line, the stains on the dial or the reflective area) still have complex noise interference and multi-scale feature distribution. Traditional convolution kernels are difficult to balance fine-grained feature extraction and context information association under a fixed receptive field, while the dilated convolution expands the receptive field range by interval sampling, which can enable the network to penetrate the local noise layer while maintaining a high-resolution feature map, and capture the topological correlation between the pointer edge and the scale line edge in the radial-angular space, such as the tiny displacement of the pointer tip, the local break of the scale line and other key details, while correlating the redundant information in the dial background (such as the edge of the protective cover or the mounting bracket) to distinguish effective features. Therefore, in the technical solution of the present application, the coordinate system transformation image data of the pointer pressure gauge is passed through an image local feature extractor based on a dilated convolutional neural network model to mine the expression ability of the local features of the coordinate system transformation image of the pointer pressure gauge, and obtain a local feature encoding map of the pointer pressure gauge image. The generated local feature encoding map of the pointer pressure gauge image not only retains the geometric information of the pointer angle, but also suppresses the interference caused by illumination changes or local occlusion through the cross-layer receptive field of the dilated convolution, so that the image recognition network has stronger noise resistance and detail resolution under complex working conditions, thereby ensuring high-precision output of pressure readings and laying a reliable local feature foundation for dynamic anomaly analysis.

[0021] Furthermore, the image data of the pointer pressure gauge coordinate system transformation is passed through a converter-based image global feature extractor to obtain a pointer pressure gauge image global feature encoding map as the pointer pressure gauge image global encoding feature. It should be understood that although the pressure gauge image after polar coordinate system transformation has eliminated the perspective distortion, its global features (such as the overall scale distribution of the dial, the relative position of the pointer and the full scale, and the background noise of the dial) still need to be semantically understood through context association. The traditional convolution operation is limited by the local receptive field, and it is difficult to model the continuous distribution characteristics of the scale lines in the dial ring structure or the geometric relationship between the pointer angle and the overall scale. The converter can dynamically capture the long-range dependency between any two points in the image through the self-attention mechanism, such as identifying the angular correspondence between the pointer tip and the zero scale and the full scale, or the spatial isolation characteristics of the dial stain area and the effective scale. Therefore, in the technical solution of the present application, in order to establish an abstract representation of the global semantics of the dial, the image data of the pointer pressure gauge coordinate system transformation is passed through a converter-based image global feature extractor to obtain a pointer pressure gauge image global feature encoding map. The generated global feature encoding map of the pointer pressure gauge image can characterize the integrity of the dial structure, dynamically focus on key areas (such as the pointer swing range) through attention weights, suppress irrelevant background interference, and enable the image recognition network to accurately infer the global rationality of the pressure value under conditions of mechanical vibration, uneven lighting or local occlusion, thereby improving the ability to distinguish between environmental interference and real faults in abnormal state judgment, and providing a highly reliable global semantic basis for the dynamic analysis model.

[0022] Specifically, the S32 performs local-global joint perceptual alignment coding on the local coding features of the pointer pressure gauge image and the global coding features of the pointer pressure gauge image to obtain the local-global joint significant perceptual coding features of the pointer pressure gauge image. It should be understood that although the local feature coding map of the pointer pressure gauge image can capture the micro-movement of the pointer tip or the wear of the scale line, it lacks the understanding of the overall range distribution of the dial; and although the global feature coding map of the pointer pressure gauge image can perceive the annular structure and the logo position of the dial, it is easy to ignore the pixel-level changes in key areas. It is difficult to fully describe the state of the pressure gauge from a single feature perspective. For example, the pointer jitter caused by mechanical vibration needs to be analyzed in combination with the local displacement and the global swing range, and the dial stain interference needs to distinguish the semantic relevance between local noise and the overall background. Therefore, in order to construct a cross-level feature interaction network so that the high-resolution details of the local features of the pointer pressure gauge and the contextual semantics of the global features can complement each other, in the technical solution of the present application, the local coding features of the pointer pressure gauge image and the global coding features of the pointer pressure gauge image are subjected to image local-global joint perceptual alignment encoding to obtain the local-global joint significant perceptual coding features of the pointer pressure gauge image. By dynamically searching and aligning the phase relationship between local and global features, the system can break through the problem of feature space isolation in traditional methods, such as geometrically associating the pointer tip angle (local) with the full scale position (global) in the polar coordinate system, and at the same time decoupling the redundant coupling between the edge of the dial shield (global interference) and the effective scale line (local key information). The generated local-global joint salient perceptual coding features of the pointer pressure gauge image not only retain the geometric accuracy of the pointer position, but also incorporate the integrity constraints of the dial structure, so that the image recognition network can adaptively focus on key feature areas (such as the pointer swing trajectory) under interference such as uneven lighting, local occlusion or mechanical vibration, suppress the misleading influence of environmental noise, thereby improving the robustness of pressure readings under complex working conditions and the credibility of abnormality discrimination, and providing a high-quality feature expression with both detail sensitivity and global consistency for the dynamic analysis model. In a specific example of this application, Figure 4 As shown, the S32 includes: S321, feature decoupling the local feature coding map of the pointer pressure gauge image and the global feature coding map of the pointer pressure gauge image to obtain a set of local monomer feature coding matrices of the pointer pressure gauge image and a set of local correlation feature coding matrices of the pointer pressure gauge image; S322, performing adaptive aggregation analysis based on image local-global feature phase alignment on the set of local monomer feature coding matrices of the pointer pressure gauge image and the set of local correlation feature coding matrices of the pointer pressure gauge image to obtain a local-global joint significant perceptual coding map of the pointer pressure gauge image as a local-global joint significant perceptual coding feature of the pointer pressure gauge image.

[0023] More specifically, the S321 performs feature decoupling on the local feature coding map of the pointer pressure gauge image and the global feature coding map of the pointer pressure gauge image to obtain a set of local monomer feature coding matrices of the pointer pressure gauge image and a set of local correlation feature coding matrices of the pointer pressure gauge image. It should be understood that the processing of pointer pressure gauge image data faces multiple challenges such as perspective offset, dial deformation and light interference. When traditional image recognition methods directly operate on the global feature map, it is easy to cause semantic confusion due to excessive feature coupling. For example, the pointer contour feature interferes with the dial scale background, thereby affecting the reading accuracy. Therefore, in the technical solution of the present application, the local feature coding map of the pointer pressure gauge image and the global feature coding map of the pointer pressure gauge image are feature decoupled to obtain a set of local monomer feature coding matrices of the pointer pressure gauge image and a set of local correlation feature coding matrices of the pointer pressure gauge image. By deconstructing the spatial coupling relationship of global features, the high-dimensional local feature coding map of the pointer pressure gauge image and the global feature coding map of the pointer pressure gauge image are converted into a set of local feature matrices of the pointer pressure gauge image with clear semantic boundaries, thereby breaking through the limitations of traditional methods at the feature representation level. This decoupling does not simply cut the feature map into isolated fragments, but decomposes the feature map into a local monomer feature matrix of the pointer pressure gauge image that represents local structural information and an associated feature matrix that describes the association pattern between the features of the pointer pressure gauge image through topological analysis, forming a dynamically combinable feature primitive library. In a specific example of the present application, the local feature coding map of the pointer pressure gauge image and the global feature coding map of the pointer pressure gauge image are feature decoupled using the following decoupling formula to obtain a set of local monomer feature coding matrices of the pointer pressure gauge image and a set of local associated feature coding matrices of the pointer pressure gauge image; wherein the decoupling formula is:

[0024] in, is the local feature coding diagram of the pointer pressure gauge image, is the global feature coding map of the pointer pressure gauge image, To perform feature decoupling operations, and are respectively the first, second, and third in the set of local monomer feature encoding matrices of the pointer pressure gauge image. and The local monomer feature encoding matrix of the pointer pressure gauge image, and are respectively the first, second, and third in the set of local correlation feature encoding matrices of the pointer pressure gauge image. and The local correlation feature encoding matrix of the pointer pressure gauge image.

[0025] More specifically, the S322 performs an adaptive aggregation analysis based on the image local-global feature phase alignment on the set of local monomer feature coding matrices of the pointer pressure gauge image and the set of local correlation feature coding matrices of the pointer pressure gauge image to obtain the pointer pressure gauge image local-global joint significant perceptual coding map as the pointer pressure gauge image local-global joint significant perceptual coding feature. In an embodiment of the present application, first, based on the feature phase alignment between any two local monomer feature coding matrices of the pointer pressure gauge image and the set of local correlation feature coding matrices of the pointer pressure gauge image, the set of local monomer feature coding matrices of the pointer pressure gauge image and the set of local correlation feature coding matrices of the pointer pressure gauge image are subjected to image local-global feature phase dynamic search alignment to obtain a set of phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs. It should be understandable that when the pressure gauge dial is slightly deformed due to mechanical vibration, or the camera is not photographed at a normal viewing angle due to installation conditions, the traditional static feature matching method often leads to misjudgment of readings due to the failure of the fixed spatial position correspondence. For example, a certain scale area may cause geometric feature compression due to perspective distortion in a plane coordinate system, and although the radial deformation is alleviated after the polar coordinate system conversion, the topological relationship between different scale segments and the pointer may still be dynamically offset due to slight warping of the dial. At this time, if only relying on the preset spatial position correspondence rules, the system cannot adaptively capture the feature association under this non-rigid deformation. Therefore, in the technical solution of the present application, based on the feature phase alignment between any two local monomer feature coding matrices of the pointer pressure gauge image and the local association feature coding matrices of the pointer pressure gauge image, the local monomer feature coding matrices of the pointer pressure gauge image and the local association feature coding matrices of the pointer pressure gauge image are subjected to dynamic image local-global feature phase search alignment to obtain a set of phase-aligned {local monomer feature coding matrices of the pointer pressure gauge image, local association feature coding matrices of the pointer pressure gauge image} feature pairs. That is, by dynamically matching the decoupled local monomer features (such as the texture and edge of the independent scale segment) with the local association features (such as the angle relationship between the scale and the pointer), the system can break through the traditional method's reliance on rigid strong constraints on spatial position. For example, when the monomer features of a certain area are distorted due to contamination, the system can dynamically screen other associated features that have a strong logical association with the area (such as the relative angle change trend of adjacent scale segments) through the phase alignment analysis of the associated features, thereby reconstructing a credible pointer-scale mapping relationship.In this way, the spatial reasoning ability of the real position of the pointer is improved, thereby providing highly robust feature pair input for subsequent attention weight allocation, and ultimately supporting the stable recognition of pressure readings under complex working conditions and the reliable discrimination of abnormal states. In a specific example of the present application, the set of local monomer feature coding matrices of the pointer pressure gauge image and the set of local correlation feature coding matrices of the pointer pressure gauge image are subjected to image local-global feature phase dynamic search alignment using the following alignment formula to obtain a set of phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs; wherein the alignment formula is:.

[0026] in, is the transpose operation, To calculate the Frobenius norm, yes and The characteristic phase alignment between To return the maximum value value, is to find the position of the maximum approximate matching value in the set of local correlation feature encoding matrices of the pointer pressure gauge image, is the trace value of the matrix, is the phase-aligned {local monomer feature encoding matrix of the pointer pressure gauge image, local correlation feature encoding matrix of the pointer pressure gauge image} feature pair, yes and The pointer pressure gauge image traces the measurement value between.

[0027] Next, each phase-aligned {local monomer feature encoding matrix of pointer pressure gauge image, local association feature encoding matrix of pointer pressure gauge image} feature pair in the set of phase-aligned {local monomer feature encoding matrix of pointer pressure gauge image, local association feature encoding matrix of pointer pressure gauge image} feature pairs is input into the local-global feature joint perception attention network to obtain the set of local-global feature joint perception attention weights of pointer pressure gauge image. It should be understood that under complex working conditions in industrial sites, the semantic association between feature pairs often shifts dynamically due to mechanical vibration, installation tilt or environmental interference in pointer pressure gauge images. When the local monomer feature matrix (such as the tip of the pointer) and the association feature matrix (such as the distribution of scale lines) are phase-aligned, although a preliminary matching relationship is established, there are significant differences in the credibility of different feature pairs in complex deformation or occlusion scenes. For example, the radial stretching caused by the thermal expansion of the dial may weaken the feature association strength of some scale areas, while oil stain occlusion may completely destroy the structural consistency of local features. At this time, if all phase-aligned feature pairs are equally aggregated, the noise features will interfere with the accuracy of reading recognition. More importantly, local image blurring caused by common oil stains or condensed water mist at industrial sites may cause some associated features (such as the relationship between adjacent scale spacing) to form a pseudo-alignment relationship with monomer features (such as the edge gradient of the blurred area), directly threatening the reliability of reading recognition. Therefore, in the technical solution of the present application, through the deep analysis of phase-aligned feature pairs by the attention network, the system can identify the implicit hierarchical dependencies in the feature space. For example, when a monomer feature encoding matrix (such as the axis area of ​​the pointer) is phase-aligned with multiple associated feature encoding matrices (such as the distribution of scales at different radial angles), the attention mechanism can automatically determine the strength of association between the axis feature and the specific angle scale based on the structural rules in the historical data, and suppress the redundant association caused by dial contamination. This dynamic weight allocation essentially constructs a self-verification mechanism for feature contribution, so that when the physical state of the dial is abnormal, the system can still maintain the stability of reading reasoning through the probabilistic coupling relationship between features. In this way, when facing local deformation of the dial, the attention network can autonomously enhance the feature weights of the area not affected by the deformation by analyzing the trace measurement parameters of the feature pair. For example, if the edge of the dial warps due to mechanical stress, causing the phase of the associated features to shift, but the individual features of the central area (such as the color contrast of the pointer) and the associated features (such as the angle of the central symmetric scale) remain highly consistent, the system will use attention weights to strengthen the decision-making influence of the central area features.

[0028] Preferably, by introducing an attention mechanism based on the number of dual connections and manifold compactification, the topological correlation strength of feature pairs is deeply analyzed. Specifically, after calculating the L1 / L2 distance between feature matrices, the distribution stability of feature pairs is quantified by the dual connection numbers α and β, revealing the structural stability of phase-aligned feature pairs in the manifold space. For example, when a monomer feature encoding matrix (such as the pointer axis area) is phase-aligned with multiple associated feature encoding matrices (such as the distribution of scales at different radial angles), the manifold compactification operation embeds and optimizes the original matrix through an exponential function, effectively suppressing the divergence of the feature manifold caused by the warping of the dial. By embedding the trace measurement parameters after manifold optimization, the system can autonomously identify highly robust feature pairs: when the edge of the dial is asymmetrically deformed due to mechanical stress, the monomer features (such as the pointer chromatic aberration contrast) and the associated features (such as the angular relationship of the symmetrical scales) in the central area will present a higher number of dual connections, triggering the attention network to strengthen the decision weight of the feature pairs in this area. This attention mechanism based on manifold topology not only effectively filters out pseudo-feature associations caused by local deformation, but also ensures the logical consistency of reading reasoning under extreme working conditions by retaining the stability of long-range associations (such as the continuity of scale angles across deformation areas). It provides a highly reliable data foundation for subsequent dynamic anomaly detection based on time series distribution, and significantly improves the level of intelligent safety management in industrial sites. Specifically, the specific steps of inputting each phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pair in the set of phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs into the local-global feature joint perception attention network of the image are as follows: calculate the correlation matrix between each phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pair in the set of phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs to obtain a set of pointer pressure gauge image feature pair scale correlation matrices; perform trace measurement on each pointer pressure gauge image feature pair scale correlation matrix in the set of pointer pressure gauge image feature pair scale correlation matrices to obtain a set of pointer pressure gauge image trace measurement values; based on the set of pointer pressure gauge image trace measurement values, obtain a set of pointer pressure gauge image local-global feature joint perception attention weights.Among them, based on the set of pointer pressure gauge image trace measurement values, a set of local-global feature joint perceptual attention weights of the pointer pressure gauge image is obtained, including: based on the set of pointer pressure gauge image trace measurement values, each phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pair in the set of phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs is optimized based on the number of dual connections to obtain a set of phase-optimized aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs; based on the set of phase-optimized aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs, a set of local-global feature joint perceptual attention weights of the pointer pressure gauge image is obtained.

[0029] In particular, here, for each matrix trace metric operation and ,make and , by calculating the corresponding and Between distance and distance The number of eigenvalues ​​of the matrix and Number of dual connections between and , that is, the distribution map stationarity between the local monomer feature coding matrix of the pointer pressure gauge image and the local correlation feature coding matrix of the pointer pressure gauge image used to represent the phase alignment.

[0030] Then, in the discrete manifold representation of the local monomer feature coding matrix of the pointer pressure gauge image and the local correlation feature coding matrix of the pointer pressure gauge image, the trace metric operation of the local monomer feature coding matrix of the pointer pressure gauge image and the local correlation feature coding matrix of the pointer pressure gauge image is embedded and optimized by compactification based on the number of dual connections, so as to improve the calculation accuracy of the joint matrix trace metric on the basis of the robustness of long-range correlation connectivity:

[0031] in, yes The optimized phase-optimized alignment pointer pressure gauge image local monomer feature encoding matrix, yes The optimized local correlation feature encoding matrix of the phase-optimized aligned pointer pressure gauge image.

[0032] That is, based on the initial and Calculated , and In the case of , , ,initial and Optimize and then calculate Compute the optimized trace metric.

[0033] In a specific example of the present application, each phase-aligned {pointer pressure gauge image local monomer feature encoding matrix, pointer pressure gauge image local association feature encoding matrix} feature pair in the set of phase-aligned {pointer pressure gauge image local monomer feature encoding matrix, pointer pressure gauge image local association feature encoding matrix} feature pairs is input into the image local-global feature joint perception attention network to obtain a set of pointer pressure gauge image local-global feature joint perception attention weights; wherein, the joint perception attention formula is:

[0034] in, Calculate the feature joint perception attention weight, is the phase optimization alignment {pointer pressure gauge image local monomer feature encoding matrix, pointer pressure gauge image local correlation feature encoding matrix} feature pair, yes The optimized phase-optimized alignment pointer pressure gauge image local monomer feature encoding matrix, yes The optimized phase-optimized alignment pointer pressure gauge image local correlation feature encoding matrix, yes and The optimized pointer pressure gauge image trace measurement value between is the normalization function, for and The local-global features of the pointer pressure gauge image are jointly perceived by the attention weight.

[0035] Then, based on the set of local-global joint perceptual attention weights of the pointer pressure gauge image, the set of phase-aligned {local monomer feature encoding matrix of the pointer pressure gauge image, local correlation feature encoding matrix of the pointer pressure gauge image} feature pairs is subjected to attention-driven saliency aggregation to obtain the local-global joint saliency perceptual encoding map of the pointer pressure gauge image. It should be understood that the non-rigid deformation of the dial due to mechanical vibration may cause the association dislocation between local features and global features, while the perspective offset or uneven illumination will cause the feature response intensity of some key areas (such as the pointer tip and scale lines) to fluctuate significantly. Traditional feature aggregation methods use equal weights or fixed rules to fuse features, which cannot adaptively distinguish high-confidence features from noise interference. For example, when the dial is radially stretched due to thermal expansion, the phase alignment relationship between some scale line features and the pointer may be weakened due to deformation. If all feature pairs are still aggregated with fixed weights at this time, the low consistency features in the deformation area will dilute the contribution of effective information, resulting in reading deviation. Therefore, in the technical solution of the present application, based on the set of local-global joint perceptual attention weights of the pointer pressure gauge image, the set of phase-aligned {local monomer feature encoding matrix of the pointer pressure gauge image, local associated feature encoding matrix of the pointer pressure gauge image} is subjected to attention-driven significant aggregation to obtain the local-global joint significant perceptual coding map of the pointer pressure gauge image. That is, by introducing the attention-driven significant aggregation mechanism, the problem of unbalanced feature contribution under complex working conditions is solved through dynamic weight allocation, breaking through the technical bottleneck of traditional methods that are sensitive to noise. In this process, the system can quantify the importance of each phase-aligned feature pair in the global feature expression. For example, when the dial is blocked by oil stains, the phase alignment between the monomer feature matrix (such as blurred scale lines) and the associated feature matrix (such as the relative position of the pointer and the scale lines) in the contaminated area is low, and the attention network will automatically reduce the weight of such feature pairs and instead strengthen the precise matching features of the uncontaminated area. This dynamic weight allocation is not based on preset physical rules, but through the correlation matrix and trace value between feature pairs, it mines the intrinsic dependencies between features from a data-driven perspective, so that the local-global feature interaction in key areas (such as the precise correspondence between the pointer tip and the precise scale) can dominate the final aggregation process, thereby forming a high-fidelity joint feature expression. In this way, by focusing on high-weight feature pairs, the stability of the pointer pointing judgment can be maintained when there is local deformation or environmental noise on the dial. For example, in a scene where the dial is reflective due to sudden changes in illumination, the attention mechanism effectively avoids the interference of artifacts on reading recognition by suppressing the feature contribution of the reflective area. At the same time, the manifold compactification process based on the optimization of the number of dual connections further enhances the robustness of long-range feature associations, so that the aggregated feature map can accurately reflect the true state of the pressure gauge under dynamic conditions.This intelligent feature fusion strategy not only improves the accuracy of reading recognition, but also provides high-quality data input for abnormal diagnosis based on statistical laws.

[0036] In a specific example of the present application, the following aggregation formula is used to perform attention-driven saliency aggregation on the set of phase-aligned {local monomer feature encoding matrix of pointer pressure gauge image, local association feature encoding matrix of pointer pressure gauge image} feature pairs to obtain the local-global joint saliency perceptual encoding map of the pointer pressure gauge image; wherein the aggregation formula is:

[0037] in, and are the first joint perception weight matrix and the second joint perception weight matrix, It is subtracted by position point. It is added by location point. yes and The combined joint perception matrix of the pointer pressure gauge image features is the first one in the set of the joint perception matrix of the pointer pressure gauge image features. The joint perception matrix of the pointer pressure gauge image features, , and are the first, second, and third in the set of the joint perception matrix of the pointer pressure gauge image feature. The joint perception matrix of the pointer pressure gauge image features, It is the local-global joint salient perceptual coding map of the pointer pressure gauge image.

[0038] Specifically, the S33 determines the pressure gauge reading value based on the local-global joint significant perceptual coding features of the pointer pressure gauge image. That is, in the technical solution of the present application, the local-global joint significant perceptual coding map of the pointer pressure gauge image is input into the pressure gauge reading recognition engine based on the classifier to obtain the pressure gauge reading value. It should be understood that although the local-global joint significant perceptual coding map of the pointer pressure gauge image contains the deep features of the pointer tip position, the scale line distribution and the dynamic association between the two, these abstract representations still need to be converted into operable engineering values. The traditional method directly uses a regression model to map the relationship between the feature vector and the reading value, but under the conditions of viewing angle offset, dial deformation or uneven illumination, such methods are prone to reading jumps or deviations due to nonlinear distortion of the feature space. Therefore, in the technical solution of the present application, a reading recognition engine based on a classifier is introduced to convert the continuous pressure value range into a robust classification decision problem through discretized category space modeling, thereby effectively resisting the destruction of the continuity of the reading by the feature space disturbance. In this process, the classifier maps the multi-dimensional features in the joint significant perception coding map to the most matching category label through the predefined discrete interval of pressure values ​​(such as 0.1MPa as the minimum classification unit). In this way, the false alarm rate caused by environmental factors is reduced, allowing the system to maintain high performance under various conditions, improving the real-time and intelligent level of equipment status monitoring, and providing strong technical support for the safety management of industrial sites.

[0039] In particular, the S4 determines whether the pressure data is in an abnormal state based on the time series distribution of the pressure gauge reading value. In one example, the general behavior of the pressure gauge data can be understood by calculating basic statistics such as the mean and variance; the autoregressive integrated moving average model (ARIMA), seasonal decomposition time series analysis (STL) or time series prediction model in machine learning can also be applied to capture the long-term trend and periodic changes in the data. In addition, a control chart or a dynamic threshold can be introduced to set reasonable upper and lower limits based on historical data. In response to the current data point exceeding the limit, it is considered that there may be an abnormal situation. It is worth mentioning that considering that some types of faults may not immediately cause the readings to deviate significantly from the normal value, but instead appear as a slow change process, therefore, when judging whether there is an abnormal state, it is necessary not only to consider whether the readings at a single time point exceed the normal range, but also to pay attention to the relationship between multiple continuous time points. Through a detailed analysis of the time series distribution of the pressure gauge reading value, not only many problems existing in traditional monitoring methods, such as high false alarm rate and low efficiency, are solved, but also strong support is provided for safety management and equipment maintenance of industrial sites.

[0040] In summary, the remote automated data collection method for pressure gauges according to the embodiment of the present application is explained, which obtains the pointer pressure gauge image data collected by the camera, optimizes the pressure gauge image feature extraction through polar coordinate system transformation, and further utilizes the local-global feature phase alignment to enhance the accuracy and robustness of pointer recognition, thereby improving the reading accuracy of the pressure gauge reading value. Furthermore, a dynamic analysis model is constructed based on the time series distribution of the pressure gauge reading value to distinguish between real equipment failures and environmental interference and reduce the false alarm rate. In this way, the system can significantly improve the accuracy of pressure gauge readings under complex working conditions, and through the abnormality detection mechanism that combines dynamic thresholds with statistical laws, it can achieve reliable early warning of hidden dangers, while reducing the need for manual inspections, enhancing the real-time and intelligent level of equipment status monitoring, and providing efficient technical support for industrial site safety management.

[0041] Furthermore, a remote automatic data collection system for pressure gauges is also provided.

[0042] Figure 5 FIG. 1 is a block diagram of a remote automated data acquisition system for pressure gauges according to an embodiment of the present application. Figure 5 As shown, according to the pressure gauge data remote automatic acquisition system 300 of the embodiment of the present application, it includes: an image acquisition module 310, which is used to acquire the pointer pressure gauge image data acquired by the camera; a coordinate system transformation module 320, which is used to perform a coordinate system transformation on the pointer pressure gauge image data to convert the pointer pressure gauge image data from the plane coordinate system to the polar coordinate system to obtain the pointer pressure gauge coordinate system transformed image data; an image recognition module 330, which is used to input the pointer pressure gauge coordinate system transformed image data into an image recognition network to obtain the pressure gauge reading value; an abnormality judgment module 340, which is used to judge whether the pressure data has an abnormal state based on the time series distribution of the pressure gauge reading value.

[0043] As described above, the pressure gauge data remote automatic acquisition system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a pressure gauge data remote automatic acquisition algorithm. In a possible implementation, the pressure gauge data remote automatic acquisition system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the pressure gauge data remote automatic acquisition system 300 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the pressure gauge data remote automatic acquisition system 300 can also be one of the many hardware modules of the wireless terminal.

[0044] Alternatively, in another example, the pressure gauge data remote automated collection system 300 and the wireless terminal may be separate devices, and the pressure gauge data remote automated collection system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0045] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A remote automatic data collection method for a pressure gauge, characterized in that: include: Acquire the image data of the pointer pressure gauge collected by the camera; Performing coordinate system transformation on the pointer pressure gauge image data to transform the pointer pressure gauge image data from a plane coordinate system to a polar coordinate system to obtain pointer pressure gauge coordinate system transformed image data; Inputting the coordinate system transformed image data of the pointer pressure gauge into an image recognition network to obtain a pressure gauge reading value, including: determining the pressure gauge reading value based on the phase alignment of local-global features in the coordinate system transformed image data of the pointer pressure gauge; Based on the time series distribution of the pressure gauge readings, determine whether the pressure data is in an abnormal state.

2. The method for remote automatic data collection of pressure gauges according to claim 1, characterized in that: Determining a pressure gauge reading value based on a phase alignment of local-global features in the coordinate system transformation image data of the pointer pressure gauge includes: Extracting local features and global features of the pointer pressure gauge coordinate system transformation image data respectively to obtain local coding features of the pointer pressure gauge image and global coding features of the pointer pressure gauge image; Performing image local-global joint perceptual alignment coding on the local coding features and the global coding features of the pointer pressure gauge image to obtain the local-global joint salient perceptual coding features of the pointer pressure gauge image; The pressure gauge reading value is determined based on the local-global joint salient perceptual coding features of the pointer pressure gauge image.

3. The method for remote automatic data collection of pressure gauges according to claim 2, characterized in that: The local features and global features of the pointer pressure gauge coordinate system transformation image data are extracted respectively to obtain the local coding features of the pointer pressure gauge image and the global coding features of the pointer pressure gauge image, including: The coordinate system transformation image data of the pointer pressure gauge is passed through an image local feature extractor based on a hole convolutional neural network model to obtain a pointer pressure gauge image local feature coding map as a pointer pressure gauge image local coding feature; The pointer pressure gauge coordinate system transformed image data is passed through a converter-based image global feature extractor to obtain a pointer pressure gauge image global feature coding map as a pointer pressure gauge image global coding feature.

4. The method for remote automatic data collection of pressure gauges according to claim 2, characterized in that: The local coding features and the global coding features of the pointer pressure gauge image are subjected to image local-global joint perceptual alignment coding to obtain the local-global joint significant perceptual coding features of the pointer pressure gauge image, including: Performing feature decoupling on the local feature coding map of the pointer pressure gauge image and the global feature coding map of the pointer pressure gauge image to obtain a set of local monomer feature coding matrices of the pointer pressure gauge image and a set of local correlation feature coding matrices of the pointer pressure gauge image; The set of local monomer feature coding matrices of pointer pressure gauge images and the set of local correlation feature coding matrices of pointer pressure gauge images are subjected to adaptive aggregation analysis based on the local-global feature phase alignment of the images to obtain the local-global joint salient perceptual coding map of the pointer pressure gauge images as the local-global joint salient perceptual coding features of the pointer pressure gauge images.

5. The method for remote automatic data collection of pressure gauges according to claim 4, characterized in that: The set of local monomer feature coding matrices of the pointer pressure gauge image and the set of local correlation feature coding matrices of the pointer pressure gauge image are subjected to adaptive aggregation analysis based on the local-global feature phase alignment of the image to obtain the local-global joint salient perceptual coding map of the pointer pressure gauge image, including: Based on the feature phase alignment between any two local monomer feature coding matrices of pointer pressure gauge images and the local correlation feature coding matrices of pointer pressure gauge images in the set of local monomer feature coding matrices of pointer pressure gauge images and the set of local correlation feature coding matrices of pointer pressure gauge images, the set of local monomer feature coding matrices of pointer pressure gauge images and the set of local correlation feature coding matrices of pointer pressure gauge images are subjected to dynamic search and alignment of image local-global feature phases to obtain a set of phase-aligned {local monomer feature coding matrices of pointer pressure gauge images, local correlation feature coding matrices of pointer pressure gauge images} feature pairs; Input each phase-aligned {local monomer feature encoding matrix of pointer pressure gauge image, local association feature encoding matrix of pointer pressure gauge image} feature pair in the set of phase-aligned {local monomer feature encoding matrix of pointer pressure gauge image, local association feature encoding matrix of pointer pressure gauge image} feature pairs into the image local-global feature joint perception attention network to obtain a set of pointer pressure gauge image local-global feature joint perception attention weights; Based on the set of local-global joint perceptual attention weights of the pointer pressure gauge image, the set of phase-aligned {pointer pressure gauge image local monomer feature encoding matrix, pointer pressure gauge image local correlation feature encoding matrix} feature pairs is subjected to attention-driven salient aggregation to obtain the local-global joint salient perceptual coding map of the pointer pressure gauge image.

6. The method for remote automatic data collection of pressure gauges according to claim 5, characterized in that: Each phase-aligned {pointer pressure gauge image local monomer feature encoding matrix, pointer pressure gauge image local association feature encoding matrix} feature pair in the set of phase-aligned {pointer pressure gauge image local monomer feature encoding matrix, pointer pressure gauge image local association feature encoding matrix} feature pairs is input into the image local-global feature joint perception attention network to obtain a set of pointer pressure gauge image local-global feature joint perception attention weights, including: Calculate the correlation matrix between each phase-aligned {local monomer feature coding matrix of pointer-type pressure gauge image, local correlation feature coding matrix of pointer-type pressure gauge image} feature pair in the set of phase-aligned {local monomer feature coding matrix of pointer-type pressure gauge image, local correlation feature coding matrix of pointer-type pressure gauge image} feature pairs to obtain a set of scale correlation matrices of pointer-type pressure gauge image feature pairs; Performing trace measurement on each pointer type pressure gauge image feature pair scale association matrix in the set of pointer type pressure gauge image feature pair scale association matrices to obtain a set of pointer type pressure gauge image trace measurement values; Based on the set of pointer pressure gauge image trace measurement values, a set of joint perceptual attention weights of local and global features of the pointer pressure gauge image is obtained.

7. The method for remote automatic data collection of pressure gauges according to claim 2, characterized in that: Based on the set of pointer pressure gauge image trace measurement values, a set of local-global feature joint perception attention weights of the pointer pressure gauge image is obtained, including: Based on the set of pointer pressure gauge image trace measurement values, each phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pair in the set of phase-aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs is optimized based on the number of dual connections to obtain a set of phase-optimized aligned {pointer pressure gauge image local monomer feature coding matrix, pointer pressure gauge image local correlation feature coding matrix} feature pairs; Based on the phase optimization alignment of the set of feature pairs {local monomer feature encoding matrix of the pointer pressure gauge image, local correlation feature encoding matrix of the pointer pressure gauge image}, a set of joint perception attention weights of the local-global features of the pointer pressure gauge image is obtained.

8. The method for remote automatic data collection of pressure gauges according to claim 2, characterized in that: Based on the local-global joint salient perceptual coding features of the pointer pressure gauge image, the pressure gauge reading value is determined, including: The local-global joint salient perceptual coding map of the pointer pressure gauge image is input into the classifier-based pressure gauge reading recognition engine to obtain the pressure gauge reading value.

9. A remote automatic data collection system for pressure gauges, characterized in that: include: An image acquisition module, used to acquire the image data of the pointer pressure gauge collected by the camera; A coordinate system transformation module, used for performing coordinate system transformation on the pointer pressure gauge image data to transform the pointer pressure gauge image data from a plane coordinate system to a polar coordinate system to obtain the pointer pressure gauge coordinate system transformation image data; An image recognition module is used to input the coordinate system transformation image data of the pointer pressure gauge into the image recognition network to obtain the pressure gauge reading value; The abnormality judgment module is used to judge whether the pressure data is in an abnormal state based on the time series distribution of the pressure gauge reading value.

Images