Submarine pipeline multi-parameter monitoring data fusion visualization method and system based on Internet of Things

Abstract

The invention discloses a subsea pipeline multi-parameter monitoring data fusion visualization method and system based on the Internet of Things, and relates to the technical field of subsea pipeline monitoring data processing and visualization. Generating a reference deviation graph, a sensing response uniformity correction graph, a first monitoring parameter channel gain and a second monitoring parameter channel gain; and S2, acquiring four original monitoring data graphs under the conditions of the first monitoring dimension and the second monitoring dimension at the first Internet of Things acquisition node and the second Internet of Things acquisition node, and performing correction processing on the four original monitoring data graphs. According to the invention, a monitoring calibration base line is established through the reference deviation graph and the sensing response uniformity correction graph.

Description

Technical Field

[0001] This invention relates to the field of subsea pipeline monitoring data processing and visualization technology, and in particular to a method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things. Background Technology

[0002] In the online monitoring and data processing of subsea pipelines under the Internet of Things (IoT) architecture, the marine environment in which the subsea pipelines are located is complex, with various interference factors such as changes in seawater flow velocity, temperature gradient, salinity fluctuations, and seabed topographic disturbances. The IoT acquisition nodes are distributed in different sections of the pipeline, and there are problems such as installation posture deviation, signal transmission delay, and sensor unit aging. In addition, the monitored multi-dimensional parameters such as pressure, temperature, strain, and corrosion have different dimensions and different data fluctuation characteristics.

[0003] Monitoring systems often employ multi-node IoT acquisition and multi-dimensional parameter monitoring configurations to obtain multi-source monitoring data sensitive to the operational status of subsea pipelines. However, marine sites also present factors such as sensor unit reference drift, uneven response between nodes, signal transmission noise, random interference from the marine environment, and scale mismatch of data from different monitoring dimensions. These factors work together across multiple acquisition nodes and multiple monitoring dimensions, causing four monitoring data streams at the same monitoring location to exhibit similar overall fluctuations over time and with the marine environment. This poses challenges to the stable extraction of effective data related to pipeline operational status and the comparability of monitoring data across nodes and batches.

[0004] Existing technologies often employ single-node, single-dimensional data calibration or simple multi-data mean fusion, or data normalization only in a single monitoring dimension. Some solutions also use global numerical scaling to reduce environmental interference. While these approaches are effective in compensating for reference drift in a single sensor unit, they often struggle to distinguish between changes in pipeline operating status and those caused by the environment or equipment when faced with additional signal interference from the marine environment, overall gain drift of the sensor unit, and fluctuations in monitoring data due to changes in the attitude of the acquisition nodes. This can easily lead to incomparability between batches of monitoring data and unstable anomaly detection thresholds. Furthermore, there is a lack of a processing chain to unify complementary information from multiple acquisition nodes and various monitoring dimensions at the same monitoring location to a consistent scale. There is also a lack of saturated data removal and local noise assessment, and data-level confidence and sample-level statistics are typically not output. This results in monitoring results that are sensitive to noise and saturated data points, highly dependent on set parameters, and lack traceability, making it difficult to meet the requirements of subsea pipeline safety monitoring for data stability, consistency, and intuitive visualization. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies, such as difficulty in obtaining stable and comparable monitoring data-level indicators and the inability to achieve efficient fusion visualization under factors such as marine environmental interference, changes in the attitude of acquisition nodes, and gain drift of sensing units. Therefore, this invention proposes a method and system for fusion visualization of multi-parameter monitoring data of submarine pipelines based on the Internet of Things.

[0006] To address the problems existing in the prior art, the present invention adopts the following technical solution: A data fusion and visualization method for multi-parameter monitoring of subsea pipelines based on the Internet of Things includes: S1. Based on the reference zero value map and the standard calibration map, generate the reference deviation map, the sensor response uniformity correction map, the first monitoring parameter channel gain and the second monitoring parameter channel gain; S2. Obtain four original monitoring data images from the first IoT acquisition node and the second IoT acquisition node, respectively, under the conditions of the first monitoring dimension and the second monitoring dimension, and perform correction processing on the four original monitoring data images; S3. Perform uniformization processing on the four original monitoring data images after correction to obtain four uniform monitoring images, and calculate the common mode term based on the pixel intensity of the four uniform monitoring images; S4. Calculate the difference term based on the pixel intensity of the first and second monitoring dimensions of the four consistent monitoring maps, and normalize the difference term based on the common mode term to obtain the multi-node multi-dimensional monitoring difference index map. S5. Generate a saturated data mask based on four consistent monitoring maps, and perform confidence calculation and robust smoothing on the multi-node multi-dimensional monitoring difference index map based on the saturated data mask to obtain a confidence map and a robust multi-node multi-dimensional monitoring difference index map. S6. Threshold filtering and morphological purification are performed on the confidence graph to obtain an effective data mask. Based on the effective data mask, statistical calculations are performed on the robust multi-node multi-dimensional monitoring difference index graph to obtain a set of statistics. S7. Package the robust multi-node multi-dimensional monitoring differential index map, confidence map, effective data mask and statistical set to obtain the monitoring sample data package, and complete the fusion visualization of multi-parameter monitoring data of the subsea pipeline based on the monitoring sample data package.

[0007] Preferably, based on the reference zero-value map and the standard calibration map, a reference deviation map, a sensor response uniformity correction map, and the gains of the first monitoring parameter channel and the second monitoring parameter channel are generated, including: Multi-frame averaging noise reduction is performed on the baseline zero-value map to obtain the baseline deviation map; Based on the effective acquisition range defined by the internal parameters of the IoT monitoring terminal and the numerical threshold of the standard calibration map, the effective area of ​​standard calibration is obtained; the reference deviation map is subtracted from the standard calibration map, the average pixel intensity of the subtracted standard calibration map within the effective area of ​​standard calibration is calculated and normalized to obtain the sensor response uniformity correction map. At the first IoT acquisition node, the standard calibration maps of the first monitoring dimension and the second monitoring dimension are respectively subjected to benchmark deviation subtraction and sensing response uniformity correction to obtain the corrected standard calibration map. The average value of the corrected standard calibration chart within the effective area of ​​the standard calibration is calculated to obtain the gain of the first monitoring parameter channel and the gain of the second monitoring parameter channel.

[0008] Preferably, four raw monitoring data images are acquired at the first IoT acquisition node and the second IoT acquisition node, respectively, under the conditions of the first monitoring dimension and the second monitoring dimension, and the four raw monitoring data images are corrected, including: The parameters of the combination of the first monitoring dimension, the second monitoring dimension, the first IoT acquisition node, and the second IoT acquisition node are fixed, and the acquisition time and signal gain of the monitoring dimension are recorded to obtain a metadata group containing the monitoring dimension, acquisition node, acquisition time, and signal gain. Under the same pipeline monitoring attitude triggering condition, based on the metadata group, four original monitoring data maps are obtained corresponding to the first monitoring dimension and the first IoT acquisition node, the first monitoring dimension and the second IoT acquisition node, the second monitoring dimension and the first IoT acquisition node, and the second monitoring dimension and the second IoT acquisition node. The four original monitoring data images were subjected to benchmark deviation subtraction and sensor response uniformity correction, and the channel gain was made consistent using the channel gain of the first monitoring parameter and the channel gain of the second monitoring parameter, respectively, to obtain the corrected four-channel monitoring data stack.

[0009] Preferably, the four corrected original monitoring data images are subjected to a standardization process to obtain four standardized monitoring images, and the common mode term is calculated based on the pixel intensity of the four standardized monitoring images, including: Divide the corrected four-channel monitoring data stack by the acquisition time and signal gain respectively to obtain four uniform monitoring images; add the pixel intensity of the same pixel position in the four uniform monitoring images and multiply by one-half to obtain the common mode term.

[0010] Preferably, a difference term is calculated based on the pixel intensity of the first and second monitoring dimensions of the four consistent monitoring images, and the difference term is normalized based on the common mode term to obtain a multi-node, multi-dimensional monitoring difference index map, including: The pixel intensity of the first monitoring dimension consistent monitoring map and the second monitoring dimension consistent monitoring map of the first IoT acquisition node are subtracted to obtain the multi-dimensional difference term of the first acquisition node. The pixel intensity of the first and second monitoring dimension consistent monitoring maps of the second IoT acquisition node is subtracted to obtain the multi-dimensional difference terms of the second acquisition node. The difference term is obtained by subtracting the multi-dimensional difference term of the second acquisition node from the multi-dimensional difference term of the first acquisition node; the difference term is then divided by the sum of the common modulus term and the small positive number to obtain the multi-node multi-dimensional monitoring difference index map.

[0011] Preferably, a saturated data mask is generated based on four consistent monitoring maps, and confidence level calculation and robust smoothing are performed on the multi-node, multi-dimensional monitoring difference index map based on the saturated data mask to obtain a confidence map and a robust multi-node, multi-dimensional monitoring difference index map, including: The pixel intensity at the same pixel location in the four consistent monitoring images is compared. The maximum pixel intensity is compared with the saturation threshold. If the maximum value is greater than or equal to the saturation threshold, the pixel location is marked as a saturated data point; otherwise, it is marked as a non-saturated data point to generate a saturated data mask. The median absolute deviation of the multi-node, multi-dimensional monitoring differential index map is calculated within a preset size neighborhood of unsaturated data points, and the coefficients are converted to obtain a local noise estimate. The confidence plot is obtained by dividing the absolute value of the difference term by the sum of the local noise estimate and the small positive number. Median filtering is performed on the multi-node, multi-dimensional monitoring difference index map under the constraint of saturated data mask to obtain a robust multi-node, multi-dimensional monitoring difference index map.

[0012] Preferably, an effective data mask is obtained by thresholding and morphological cleansing of the confidence graph. Based on the effective data mask, a set of statistics is obtained by statistically calculating the robust multi-node, multi-dimensional monitoring difference index graph, including: A coarse mask is obtained by thresholding the confidence map, and an effective data mask is obtained by opening and closing operations on the coarse mask. The mean, standard deviation and quantile of the preset percentile are calculated for the robust multi-node multi-dimensional monitoring difference index map within the pixel set of the effective data mask. The obtained mean, standard deviation and quantile of each preset percentile are combined to form a statistical set.

[0013] Based on monitoring sample data packages, multi-parameter monitoring data of subsea pipelines are fused and visualized, including: The robust multi-node multi-dimensional monitoring differential index map, confidence map, and effective data mask in the monitoring sample data package are fused to generate a multi-dimensional fused monitoring base map. The numerical indicators in the statistical set are visualized and labeled, and then overlaid with the multi-dimensional integrated monitoring base map; Based on the geographic information data of the Internet of Things submarine pipeline, the superimposed fused monitoring map is registered with the spatial location of the submarine pipeline to generate a three-dimensional / two-dimensional visualization map of multi-parameter monitoring data of the submarine pipeline. The visualization map is dynamically updated and interactively designed, supporting multi-view viewing of monitoring data, highlighting of abnormal data, and retrospective comparison of historical data.

[0014] The IoT-based multi-parameter monitoring data fusion and visualization system for subsea pipelines includes: a baseline calibration module, which generates a baseline deviation map, a sensor response uniformity correction map, and the gain of the first monitoring parameter channel and the gain of the second monitoring parameter channel based on the baseline zero value map and the standard calibration map; The node acquisition and correction module is used to acquire four original monitoring data images from the first IoT acquisition node and the second IoT acquisition node, under the conditions of the first monitoring dimension and the second monitoring dimension, and to perform correction processing on the four original monitoring data images. The unification module is used to unify the four original monitoring data images after correction to obtain four unified monitoring images, and to calculate the common mode term based on the pixel intensity of the four unified monitoring images; The differential normalization module is used to calculate the differential term based on the pixel intensity of the first and second monitoring dimensions of the four consistent monitoring maps, and to normalize the differential term based on the common mode term to obtain a multi-node multi-dimensional monitoring differential index map. The confidence robustness module is used to generate a saturated data mask based on four consistent monitoring maps, and to perform confidence calculation and robust smoothing on the multi-node multi-dimensional monitoring difference index map based on the saturated data mask, so as to obtain a confidence map and a robust multi-node multi-dimensional monitoring difference index map. The effective domain statistics module is used to perform threshold filtering and morphological purification on the confidence graph to obtain an effective data mask. Based on the effective data mask, statistical calculations are performed on the robust multi-node multi-dimensional monitoring difference index graph to obtain a set of statistics. The data packaging module is used to package the robust multi-node multi-dimensional monitoring difference index map, confidence map, effective data mask and statistical set to obtain the monitoring sample data package; The fusion visualization module is used to complete the fusion visualization of multi-parameter monitoring data of subsea pipelines based on monitoring sample data packets.

[0015] The fusion visualization module includes: The layer fusion unit is used to fuse the robust multi-node multi-dimensional monitoring difference index map, confidence map, and effective data mask in the monitoring sample data package to generate a multi-dimensional fused monitoring base map; the indicator annotation unit is used to visualize and annotate the numerical indicators in the statistical set and overlay them with the multi-dimensional fused monitoring base map. The spatial registration unit is used to register the superimposed fused monitoring map with the spatial location of the subsea pipeline based on the geographic information data of the Internet of Things subsea pipeline, and generate a three-dimensional / two-dimensional visualization map of multi-parameter monitoring data of the subsea pipeline. The interactive update unit is used to dynamically update and interactively design the visualization map, supporting multi-view viewing of monitoring data, highlighting of abnormal data, and retrospective comparison of historical data.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention establishes a monitoring calibration baseline by using a benchmark deviation map and a sensor response uniformity correction map, and obtains channel gain in two monitoring dimensions. After performing benchmark deviation subtraction and sensor response correction on the four-channel monitoring data, the data is then uniformized according to their respective acquisition time and signal gain to construct a common mode term for the same monitoring location. This reduces the overall fluctuations caused by marine environmental interference, sensor dark noise, electronic gain, and slight changes in the attitude of the acquisition nodes from the source, enabling the entire subsequent process to operate on a unified scale and have comparability of monitoring data across nodes and shifts.

[0017] 2. This invention performs multi-dimensional differencing within different IoT acquisition nodes, then uses the difference between the two as the numerator and the sum of the common mode term and a small positive number as the denominator to achieve normalization, forming a dimensionless index map that is more sensitive to the operating status of subsea pipelines and insensitive to overall environmental drift. At the same time, a saturated data mask is generated based on four uniform monitoring maps, and local noise is estimated by converting the median absolute deviation. A confidence map is output, and median filtering is applied to the index map, thereby suppressing saturation distortion and isolated anomalies, and significantly improving the stability and repeatability of monitoring data.

[0018] 3. This invention obtains an effective data mask by threshold screening and morphological purification of the confidence graph. The mean, standard deviation and several quantiles are calculated within the mask to form a set of statistics for safety assessment and operation and maintenance control of subsea pipelines. The robust index graph, confidence graph, effective data mask and set of statistics are packaged and output by sample to achieve traceable data delivery from the data level to the sample level.

[0019] 4. This invention realizes the fusion and visualization of multi-parameter monitoring data of subsea pipelines based on monitoring sample data packages. Through layer fusion, index labeling, spatial registration and interactive updates, the abstract monitoring data is transformed into intuitive two-dimensional / three-dimensional visualization maps, realizing the accurate correspondence between monitoring data and pipeline physical location. It supports multi-view viewing, anomaly highlighting and historical backtracking, providing intuitive and efficient visualization support for the safety monitoring, fault early warning and operation and maintenance decision-making of subsea pipelines, and solving the problems of poor visualization and weak interactivity in traditional monitoring data processing. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things, according to an embodiment of the present invention. Figure 2 This is a functional block diagram of an IoT-based multi-parameter monitoring data fusion and visualization system for submarine pipelines, provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] A method for fusion and visualization of multi-parameter monitoring data for subsea pipelines based on the Internet of Things (IoT) includes: S1, generating a benchmark deviation map, a sensor response uniformity correction map, and the gains of the first and second monitoring parameter channels based on a benchmark zero-value map and a standard calibration map; S2, acquiring four original monitoring data maps at the first and second IoT acquisition nodes, respectively, under the conditions of the first and second monitoring dimensions, and performing correction processing on the four original monitoring data maps; S3, performing uniformization processing on the four corrected original monitoring data maps to obtain four uniform monitoring maps, and calculating the common mode term based on the pixel intensity of the four uniform monitoring maps; S4, calculating the difference term based on the pixel intensity of the first and second monitoring dimensions of the four uniform monitoring maps, and based on the common mode term... S5. Normalize the difference terms to obtain a multi-node, multi-dimensional monitoring difference index map; S6. Generate a saturated data mask based on four uniform monitoring maps, and perform confidence calculation and robust smoothing on the multi-node, multi-dimensional monitoring difference index map based on the saturated data mask to obtain a confidence map and a robust multi-node, multi-dimensional monitoring difference index map; S7. Perform threshold screening and morphological purification on the confidence map to obtain an effective data mask, and perform statistical calculations on the robust multi-node, multi-dimensional monitoring difference index map based on the effective data mask to obtain a set of statistics; S8. Package the robust multi-node, multi-dimensional monitoring difference index map, confidence map, effective data mask, and set of statistics to obtain a monitoring sample data package, and complete the fusion and visualization of multi-parameter monitoring data of the subsea pipeline based on the monitoring sample data package.

[0023] Preferably, based on the baseline zero-value map and the standard calibration map, a baseline deviation map, a sensor response uniformity correction map, a first monitoring parameter channel gain, and a second monitoring parameter channel gain are generated, including: performing multi-frame averaging noise reduction on the baseline zero-value map to obtain the baseline deviation map; obtaining the standard calibration effective area according to the effective acquisition range defined by the internal parameters of the IoT monitoring terminal and the numerical threshold of the standard calibration map; subtracting the baseline deviation map from the standard calibration map, calculating the average pixel intensity of the subtracted standard calibration map within the standard calibration effective area, and performing normalization processing to obtain the sensor response uniformity correction map; performing baseline deviation subtraction and sensor response uniformity correction on the standard calibration maps of the first monitoring dimension and the second monitoring dimension respectively under the first IoT acquisition node to obtain the corrected standard calibration map; averaging the corrected standard calibration map within the standard calibration effective area to obtain the first monitoring parameter channel gain and the second monitoring parameter channel gain.

[0024] Specifically, the baseline zero-value map is a raw monitoring data map obtained from the submarine pipeline when it is not in operation, maintaining the same monitoring dimensions, acquisition nodes, acquisition time, and signal gain as the actual monitoring. It is used to characterize the baseline deviation, dark noise, and fixed pattern noise baseline of the IoT sensing unit, facilitating subsequent baseline deviation subtraction on the four raw monitoring data maps. The standard calibration map is a raw monitoring data map obtained under the same monitoring environment and parameters, using a standard calibration piece as the monitoring target. It is used to characterize the uniformity of the sensing unit response, node transmission characteristics, and environmental baseline. After subtracting the baseline deviation and normalizing the standard calibration map within the effective area of ​​the standard calibration, a sensing response uniformity correction map can be obtained. Based on this, the gain of the first monitoring parameter channel and the gain of the second monitoring parameter channel are obtained, serving as the baseline for the corrected four-channel monitoring data stack and the uniformization processing.

[0025] Specifically, the baseline zero-value map obtained through pipelineless operation only includes the reference deviation of the sensing unit, dark noise, and fixed pattern noise. Pixel-to-pipe averaging of multiple frames of the baseline zero-value map reduces the variance of independent random noise by the inverse of the frame number and the standard deviation by the inverse square root of the frame number, thus obtaining a stable and reliable reference deviation map. The standard calibration map acquired using standard calibration components can be considered an ideal standard monitoring response, whose ideal value should be a constant in space. Actual deviations mainly originate from uneven sensing unit responses, node transmission losses, and local interference from the marine environment. Therefore, the effective acquisition range is limited based on the intrinsic parameters of the IoT monitoring terminal, and threshold values ​​from the standard calibration map are used for filtering. By removing invalid monitoring areas and abnormal data areas, the most representative standard calibration effective area for the true response of standard calibration can be obtained. Subtracting the reference deviation map from the standard calibration map by pixel can remove additive bias, leaving only multiplicative terms related to sensing response and transmission characteristics. Averaging the standard calibration effective area and using this average value to normalize the full-frame standard calibration map can form a sensing response uniformity correction map. This allows any subsequent actual monitoring data to first deduct the reference deviation and then divide by this correction map to compensate for sensing response non-uniformity and transmission loss in a multiplicative manner, achieving flat field correction and providing a scaled and traceable baseline for the calculation of the differential index of subsequent multi-node multi-dimensional monitoring.

[0026] Specifically, a baseline zero-value map is acquired under multi-frame pipelineless operation conditions. This baseline zero-value map reflects the baseline drift characteristics of the IoT sensing unit. Then, for each pixel location, the average value of all baseline zero-value maps at that pixel location is calculated. This averaging of multiple frames suppresses random noise in the baseline drift, ultimately yielding a baseline deviation map characterizing the baseline deviation of each pixel. Based on pre-acquired IoT monitoring terminal intrinsic parameters, which include the acquisition geometry parameters of the sensing unit and acquisition nodes, the effective acquisition range of monitoring data is determined to exclude invalid data areas where normal acquisition is impossible at the edges of the sensing unit. Simultaneously, a standard calibration map of a standard calibration component is acquired. Numerical analysis is performed on this standard calibration map, and reasonable numerical thresholds are set to filter out pixel areas with uniform and standard values. These areas represent the true monitoring response of the standard calibration component, excluding invalid areas such as interference or non-calibrated areas on the surface of the calibration component. The effective acquisition range determined by the monitoring terminal intrinsic parameters and the pixel areas filtered by the numerical thresholds are intersected to obtain the effective standard calibration area used for subsequent correction parameter calculations. The acquired standard calibration map undergoes a reference deviation subtraction process. Specifically, for each pixel position in the standard calibration map, the value of the corresponding pixel position in the reference deviation map is subtracted from the value of that pixel position to eliminate the influence of reference drift on the standard calibration map values. Within the defined effective area of ​​standard calibration, the average value of all pixels in the standard calibration map after reference deviation subtraction is calculated. This average value represents the reference value level of the standard calibration component under ideal uniform response conditions. Finally, for each pixel position in the standard calibration map after reference deviation subtraction, the value of that pixel position is divided by the average value within the effective area of ​​standard calibration. This normalization operation eliminates the influence of inconsistent response sensitivity of each pixel in the IoT sensing unit, ultimately obtaining a sensing response uniformity correction map used to correct the non-uniformity of monitoring data response.

[0027] At the first IoT acquisition node, the standard calibration maps of the first monitoring dimension and the second monitoring dimension are respectively subjected to reference deviation subtraction and sensing response uniformity correction to obtain the corrected standard calibration map; the average value of the corrected standard calibration map is calculated within the effective area of ​​the standard calibration to obtain the channel gain of the first monitoring parameter and the channel gain of the second monitoring parameter. Specifically, the first and second monitoring dimensions are two different subsea pipeline monitoring parameters, such as pressure, temperature, strain, and corrosion, that are collected independently under the same acquisition geometry. These parameters are defined by the IoT sensing unit, and their monitoring accuracy meets the requirements for subsea pipeline safety monitoring and is selected within a range where the sensing unit's response sensitivity and signal transmission stability are high. The two dimensions must be different to form a dimensional difference, thereby amplifying the response of the subsea pipeline's operating status to changes in the monitoring parameters and combining it with the acquisition nodes to form a four-channel acquisition.

[0028] Specifically, selecting the first and second monitoring dimensions to collect data on the standard calibration piece at the first IoT acquisition node fixes the channel transfer coefficient, which is composed of the sensing response, signal transmission, node acquisition efficiency, and electronic gain of the two monitoring dimensions, under the same acquisition geometry. This avoids scale drift caused by the transmission characteristics of different acquisition nodes. The standard calibration piece can be regarded as an ideal standard monitoring target, whose ideal monitoring response should be constant within the monitoring field of view. The numerical unevenness in the actual data mainly comes from additive reference bias and multiplicative flattening terms. The reference bias includes dark noise and fixed pattern noise, while the flattening term includes sensing response unevenness, node transmission loss, and environmental factors. To eliminate interference, the additive term is removed by subtracting the reference deviation pixel by pixel from the standard calibration map. Then, the multiplicative term is removed by dividing the sensor response uniformity correction map pixel by pixel. At this point, the average value of the calibrated standard calibration map is calculated within the effective area of ​​the standard calibration defined by the numerical threshold in the monitoring terminal. The resulting first monitoring parameter channel gain and second monitoring parameter channel gain are the relative scale anchoring quantities of the two monitoring dimensions under this acquisition geometry and monitoring link. This can unify the values ​​of the subsequent actual monitoring data in the two dimensions to comparable dimensions, laying a stable monitoring calibration foundation for constructing state correlation terms based on the difference between the two monitoring dimensions and cooperating with the common mode term to complete the normalization.

[0029] Specifically, in the first IoT acquisition node state, standard calibration maps corresponding to the first monitoring dimension and the second monitoring dimension are acquired respectively. For the standard calibration map of the first monitoring dimension, for each pixel position, the value of the corresponding pixel position in the reference deviation map is subtracted from the pixel value to complete the reference deviation subtraction. Then, the subtracted value is divided by the correction value of the corresponding pixel position in the sensor response uniformity correction map to complete the sensor response uniformity correction, thereby obtaining the corrected standard calibration map of the first monitoring dimension. Using the same processing method, the reference deviation subtraction and sensor response uniformity correction are sequentially performed on the standard calibration map of the second monitoring dimension to obtain the corrected standard calibration map of the second monitoring dimension. For the corrected standard calibration map of the first monitoring dimension, within the determined standard calibration effective area, the average value of all pixels in the area is calculated, and this average value is used as the gain of the first monitoring parameter channel. For the corrected standard calibration map of the second monitoring dimension, the average value of all pixels in the standard calibration effective area is also calculated, and this average value is used as the gain of the second monitoring parameter channel.

[0030] Preferably, four original monitoring data images are acquired under the conditions of the first IoT acquisition node and the second IoT acquisition node, and under the conditions of the first monitoring dimension and the second monitoring dimension, respectively. The four original monitoring data images are then corrected, including: fixing the parameters of the combination of the first monitoring dimension, the second monitoring dimension, the first IoT acquisition node, and the second IoT acquisition node, and recording the acquisition time and signal gain of the monitoring dimension to obtain a metadata group containing the monitoring dimension, acquisition node, acquisition time, and signal gain; under the same pipeline monitoring posture triggering condition, based on the metadata group, four original monitoring data images corresponding to the first monitoring dimension and the first IoT acquisition node, the first monitoring dimension and the second IoT acquisition node, the second monitoring dimension and the first IoT acquisition node, and the second monitoring dimension and the second IoT acquisition node are acquired; the four original monitoring data images are respectively subjected to reference deviation subtraction and sensor response uniformity correction, and the channel gain is consistent using the channel gain of the first monitoring parameter and the channel gain of the second monitoring parameter, respectively, to obtain a corrected four-channel monitoring data stack.

[0031] Specifically, the first and second IoT acquisition nodes refer to IoT data acquisition terminals deployed in different monitoring sections of the subsea pipeline. They are used to achieve synchronous monitoring of multiple locations along the pipeline. The first acquisition node focuses on monitoring core parameters of key pipeline sections, retaining basic monitoring signals and environmental interference components related to pipeline operation. The second acquisition node, as a complementary monitoring point, strongly suppresses common environmental interference components unrelated to pipeline status, while making changes in monitoring signals caused by changes in pipeline operation status more readily apparent. The two acquisition nodes provide complementary monitoring information at the same monitoring location. The former emphasizes retaining basic monitoring components, while the latter emphasizes suppressing environmental interference components. This facilitates the subsequent construction of differential indicators that are more sensitive to the subsea pipeline's operational status and more stable to environmental interference using multi-dimensional differential and common-mode normalization methods.

[0032] Specifically, the parameters of the first monitoring dimension, the second monitoring dimension, the first IoT acquisition node, and the second IoT acquisition node are fixed, and the acquisition time and signal gain of the monitoring dimensions are recorded. This is to lock the variable factors of the monitoring link before acquisition, ensuring that the subsequent four raw monitoring data images are under the same acquisition conditions and the same electronic magnification scale. Acquiring four raw monitoring data images under the same pipeline monitoring attitude trigger can avoid the phase drift of the monitoring signal caused by the installation attitude of the acquisition node and instantaneous changes in the marine environment, thus ensuring a one-to-one correspondence of the monitoring positions. The sensor unit output can be understood as the actual monitored physical quantity multiplied by the acquisition time multiplied by the signal gain plus the reference deviation bias, where the reference deviation bias is composed of dark noise and fixed pattern noise. Therefore, the reference deviation subtraction of the four raw monitoring data images can remove the additive error term. After the reference deviation subtraction, the standard calibration image still contains multiplicative terms caused by uneven sensor response, node transmission loss, and local interference from the marine environment. By dividing the original monitoring data map pixel by pixel according to the sensor response uniformity correction map, such multiplicative distortion can be eliminated, achieving flat field correction. Since the sensor unit response, signal transmission characteristics, and acquisition node gain are different under different monitoring dimensions, the response scale of the two monitoring dimensions under the same physical quantity is inconsistent. Further, the channel gain of the first monitoring parameter channel and the channel gain of the second monitoring parameter channel are used to unify the channel gain of the corrected data map, which can unify the values ​​of the two monitoring dimensions to comparable dimensions. The acquisition node follows the signal transmission characteristics of the Internet of Things, and its acquisition intensity is inversely proportional to the loss of the transmission link. By fixing the geometry of the acquisition node and performing the above correction and unification on the two monitoring dimensions respectively, the subsequent differential operation based on the two monitoring dimensions can truly reflect the parameter-dependent characteristic terms of the subsea pipeline operation status. At the same time, the common mode term constructed by the four channels can suppress environmental interference and overall sensor gain drift, providing a traceable physical baseline for the stable calculation of the differential index of multi-node multi-dimensional monitoring.

[0033] Specifically, first, determine four combinations of monitoring dimensions and acquisition nodes: the first monitoring dimension and the first IoT acquisition node, the first monitoring dimension and the second IoT acquisition node, the second monitoring dimension and the first IoT acquisition node, and the second monitoring dimension and the second IoT acquisition node. For each of these four combinations, the parameters are fixed. The acquisition time and signal gain are recorded separately according to the monitoring dimension. That is, the two acquisition node combinations corresponding to the first monitoring dimension share the acquisition time and signal gain of the first monitoring dimension, and the two acquisition node combinations corresponding to the second monitoring dimension share the acquisition time and signal gain of the second monitoring dimension. The monitoring dimension, acquisition node, acquisition time of that dimension, and signal gain of that dimension corresponding to each combination are associated and recorded to form a metadata group containing the monitoring dimension, acquisition node, acquisition time, and signal gain. Maintaining the operational monitoring posture of the subsea pipeline unchanged, data acquisition is triggered based on the aforementioned metadata group. The original monitoring data map corresponding to the first monitoring dimension and the first IoT acquisition node, the original monitoring data map corresponding to the first monitoring dimension and the second IoT acquisition node, the original monitoring data map corresponding to the second monitoring dimension and the first IoT acquisition node, and the original monitoring data map corresponding to the second monitoring dimension and the second IoT acquisition node are acquired simultaneously in one go, ensuring that the four original monitoring data maps are common field monitoring maps under the same monitoring posture. The four original monitoring data images were processed separately. First, for each pixel position in each original monitoring data image, the value of the corresponding pixel position in the baseline deviation image was subtracted from the value of that pixel to complete the baseline deviation deduction. Then, the value obtained after baseline deviation deduction was divided by the correction value of the corresponding pixel position in the sensor response uniformity correction image to complete the sensor response uniformity correction. Subsequently, the channel gain was selected according to the monitoring dimension corresponding to the original monitoring data image. The two corrected images corresponding to the first monitoring dimension and the first IoT acquisition node, and the two corrected images corresponding to the first monitoring dimension and the second IoT acquisition node were normalized using the first monitoring parameter channel gain. The two corrected images corresponding to the second monitoring dimension and the first IoT acquisition node, and the two corrected images corresponding to the second monitoring dimension and the second IoT acquisition node were normalized using the second monitoring parameter channel gain to achieve channel gain consistency. Finally, these four images with complete correction processing were integrated to obtain the corrected four-channel monitoring data stack.

[0034] Preferably, the four original monitoring data images after correction are uniformized to obtain four uniform monitoring images, and the common mode term is calculated based on the pixel intensity of the four uniform monitoring images, including: dividing the corrected four-channel monitoring data stack by the acquisition time and signal gain respectively to obtain four uniform monitoring images; adding the pixel intensity of the same pixel position in the four uniform monitoring images and multiplying it by one-half to obtain the common mode term.

[0035] Specifically, after benchmark deviation subtraction and sensor response uniformity correction, the monitoring link can approximately follow a linear response model. That is, the pixel intensity is proportional to the actual monitored physical quantity of the subsea pipeline and is simultaneously affected by the multiplicative amplification effect of the acquisition time and signal gain. In order to restore the scale proportional to the actual monitored physical quantity, the four-channel corrected monitoring data map needs to be divided pixel by the corresponding acquisition time and signal gain, thereby eliminating the scale difference caused by the IoT monitoring terminal settings and putting different monitoring dimensions and acquisition nodes on a unified scale. On this basis, the pixel intensities of the four uniform monitoring maps at the same monitoring location are added together and halved to construct a common mode term. This term is equivalent to the weighted average of four independent observations. It can merge the overall changes in the marine environment and the overall drift of the system gain into a common component and reduce the variance of random noise with the number of observations through multi-channel averaging. At the same time, the magnitude of the denominator is matched with the typical amplitude of the multi-node, multi-dimensional difference numerator to avoid over-amplification or compression of the normalization result, thus laying the foundation for obtaining a stable, comparable, and environmentally insensitive difference index.

[0036] Specifically, the four images in the corrected four-channel monitoring data stack are first defined as corresponding to combinations of the first monitoring dimension and the first IoT acquisition node, the first monitoring dimension and the second IoT acquisition node, the second monitoring dimension and the first IoT acquisition node, and the second monitoring dimension and the second IoT acquisition node, respectively. For these four images, based on the monitoring dimension corresponding to each image, the recorded acquisition time and signal gain of that monitoring dimension are matched. For each pixel position in each image, the intensity value of that pixel is divided by the product of the acquisition time and signal gain of the corresponding monitoring dimension. This operation eliminates the influence of differences in acquisition time and signal gain under different monitoring dimensions on pixel intensity. After all four images have undergone this processing, four consistent monitoring images are obtained. For the generated four consistent monitoring images, each pixel position in the image is traversed, and the pixel intensity value of each of the four consistent monitoring images at that pixel position is extracted. These four pixel intensity values ​​are summed, and the summation result is multiplied by one-half to obtain the common mode value of that pixel position. After performing the above extraction, summation, and multiplication operations on all pixel positions in the image, a common mode term covering the entire monitoring area is finally formed.

[0037] Preferably, the difference term is calculated based on the pixel intensity of the first and second monitoring dimensions of the four consistent monitoring images, and the difference term is normalized based on the common mode term to obtain a multi-node multi-dimensional monitoring difference index map. This includes: subtracting the pixel intensity of the first and second consistent monitoring images of the first IoT acquisition node to obtain the first acquisition node multi-dimensional difference term; subtracting the pixel intensity of the first and second consistent monitoring images of the second IoT acquisition node to obtain the second acquisition node multi-dimensional difference term; subtracting the second acquisition node multi-dimensional difference term from the first acquisition node multi-dimensional difference term to obtain the difference term; and dividing the difference term by the sum of the common mode term and a small positive number to obtain the multi-node multi-dimensional monitoring difference index map.

[0038] Specifically, within the first IoT acquisition node channel, subtracting the second monitoring dimension from the first monitoring dimension-consistent monitoring map is equivalent to differentially analyzing the response of the subsea pipeline's operating status to changes in monitoring parameters. This can suppress environmental constants and slowly varying terms while highlighting characteristic terms related to the pipeline's operating status and monitoring parameters. Performing the same multi-dimensional differential within the second IoT acquisition node channel yields the corresponding state-sensitive quantities under a different acquisition geometry. Subtracting the multi-dimensional differential from the second acquisition node's multi-dimensional differential further utilizes the transmission characteristics and sensing capabilities of the IoT acquisition nodes. The response assigns different weights to the monitored components to cancel out environmental interference and transmission noise components that are unrelated to the monitoring parameters or insensitive to the acquisition nodes, while retaining the components that are most sensitive to changes in the monitoring dimensions and are related to the pipeline's operating status. Finally, the sum of the common-mode term obtained by superimposing four uniform monitoring maps and a small positive number is used to normalize the above differences. This can incorporate multiplicative common-mode disturbances such as changes in marine environmental intensity and electronic gain drift into the denominator and stabilize the numerical range. As a result, the obtained multi-node, multi-dimensional monitoring differential index has both a clear pipeline status orientation and cross-batch, cross-node comparability and noise robustness.

[0039] Specifically, first, two consistent monitoring maps corresponding to the first IoT data acquisition node are determined: the first-dimensional consistent monitoring map and the second-dimensional consistent monitoring map of the first IoT data acquisition node. For these two maps, each pixel position is traversed, and the pixel intensity value of the first-dimensional consistent monitoring map at that pixel position is subtracted from the pixel intensity value of the second-dimensional consistent monitoring map at the same pixel position. After performing this pixel-level subtraction operation for all pixel positions, the multi-dimensional difference term of the first data acquisition node is obtained. Next, two consistent monitoring maps corresponding to the second IoT data acquisition node are determined: the first-dimensional consistent monitoring map and the second-dimensional consistent monitoring map of the second IoT data acquisition node. Using the same pixel-level processing method as the multi-dimensional difference term of the first data acquisition node, all pixel positions are traversed, and the pixel intensity value of the first-dimensional consistent monitoring map at that pixel position is subtracted from the pixel intensity value of the second-dimensional consistent monitoring map at the same pixel position. After performing this pixel-level operation for all pixel positions, the multi-dimensional difference term of the second data acquisition node is obtained. Next, taking the multi-dimensional difference terms of the first and second acquisition nodes as the processing objects, for each pixel position in the image, the value in the multi-dimensional difference term of the first acquisition node is subtracted from the value in the multi-dimensional difference term of the second acquisition node. This subtraction operation is performed on all pixels in the image to obtain the difference term. To avoid division by zero during the calculation, a small positive number is introduced. For each pixel value in the difference term, the pixel value is divided by the sum of the value at the corresponding pixel position in the common mode term and the small positive number. After performing this normalization operation on all pixels, the final multi-node multi-dimensional monitoring difference index map is obtained.

[0040] Preferably, a saturated data mask is generated based on four uniform monitoring images, and confidence level calculation and robust smoothing are performed on the multi-node multi-dimensional monitoring difference index map based on the saturated data mask to obtain a confidence map and a robust multi-node multi-dimensional monitoring difference index map. This includes: comparing the pixel intensity at the same pixel position in the four uniform monitoring images, comparing the maximum pixel intensity with a saturation threshold, and marking the pixel position as a saturated data point if the maximum value is greater than or equal to the saturation threshold; otherwise, marking it as a non-saturated data point to generate a saturated data mask; calculating the median absolute deviation of the multi-node multi-dimensional monitoring difference index map within a preset size neighborhood of the non-saturated data points and performing coefficient conversion to obtain a local noise estimate; dividing the absolute value of the difference term by the sum of the local noise estimate and a small positive number to obtain a confidence map; and performing median filtering on the multi-node multi-dimensional monitoring difference index map under the constraint of the saturated data mask to obtain a robust multi-node multi-dimensional monitoring difference index map.

[0041] Specifically, when the IoT sensing unit enters the shearing zone near full scale, the output is no longer proportional to the actual monitored physical quantity. Therefore, by comparing the maximum value at the same pixel position of four consistent monitoring images with the saturation threshold and generating a saturated data mask, distorted data points can be removed from subsequent calculations to avoid the amplification of the difference and normalization by the shearing effect. For the multi-node, multi-dimensional monitoring difference index map, the median absolute deviation is used in the neighborhood of the unsaturated data points and multiplied by a coefficient of 1:4826 for conversion. This yields a local noise estimate that is insensitive to isolated anomalies and consistent with the Gaussian case, thus using the absolute value of the difference term. The confidence map obtained by dividing the value by the sum of the local noise estimate and the small positive number to prevent division by zero is the dimensionless measure of the local signal-to-noise ratio, which reflects both the strength of the differential signal and ensures numerical stability. Median filtering is applied to the multi-node multi-dimensional monitoring differential index map under the constraint of saturated data mask. This can significantly suppress salt-and-pepper noise and sparse artifacts while preserving the edges and details of the monitoring data. This makes the resulting robust multi-node multi-dimensional monitoring differential index map more spatially continuous and consistent with the changes in the operating status of the subsea pipeline, thus providing a reliable and comparable input for subsequent statistical calculations based on effective data masks.

[0042] Specifically, the process first iterates through each pixel location in the four uniform monitoring images, extracting the pixel intensity of that location in each of the four images. These four pixel intensities are then compared to determine the maximum value. This maximum value is then compared with a preset saturation threshold. If the maximum value is greater than or equal to the saturation threshold, the current pixel location is marked as a saturated data point. If the maximum value is less than the saturation threshold, the current pixel location is marked as a non-saturated data point. The saturation threshold is determined based on the maximum range output characteristics of the IoT sensor unit and can be obtained by gradient acquisition testing of standard physical quantities, taking the critical value at which the pixel intensity no longer changes with the increase of the actual physical quantity. After completing the above comparison and marking operations for all pixel locations, all marking results are integrated to generate a saturated data mask covering the entire monitoring area. For multi-node, multi-dimensional monitoring differential index maps, the focus is only on the locations marked as unsaturated data points. For each unsaturated data point, a neighborhood of a preset size is defined. The neighborhood size can be set to a positive odd number, such as 3×3 or 5×5. The median absolute deviation of all pixel values ​​within this neighborhood is calculated, and then the obtained median absolute deviation is multiplied by a preset coefficient, such as 1.4826. The preset coefficient 1.4826 is based on the statistical conversion relationship between median absolute deviation and standard deviation under a normal distribution. This coefficient is used to convert the median absolute deviation into an approximate standard deviation. After the conversion, the local noise estimate for each unsaturated data point location is obtained. The local noise estimate for saturated data point locations can be set to an invalid value. For each pixel location in the difference term, the absolute value of the pixel value at that location is first calculated, and then this absolute value is divided by the sum of the local noise estimate for the corresponding pixel location and a small positive number. The small positive number is introduced to avoid division by zero when the local noise estimate is zero. This operation is performed sequentially for all pixel locations, and finally, a confidence map is obtained. Using the generated saturated data mask as a constraint, median filtering is performed on the multi-node, multi-dimensional monitoring differential index map. During the filtering process, only the positions marked as non-saturated data points in the saturated data mask are replaced with the median value of the pixels in the surrounding preset size neighborhood. For the positions marked as saturated data points in the saturated data mask, their original pixel state is maintained or set to a preset invalid value. After the median filtering operation of the entire map is completed, a robust multi-node, multi-dimensional monitoring differential index map is obtained.

[0043] Preferably, a valid data mask is obtained by thresholding and morphological cleaning of the confidence map. Based on the valid data mask, a statistical set is obtained by statistically calculating the robust multi-node multi-dimensional monitoring difference index map. This includes: thresholding the confidence map to obtain a coarse mask, and performing opening and closing operations on the coarse mask to obtain the valid data mask; calculating the mean, standard deviation, and quantiles of preset percentiles for the robust multi-node multi-dimensional monitoring difference index map within the pixel set of the valid data mask, and combining the obtained mean, standard deviation, and quantiles of each preset percentile to form a statistical set.

[0044] Specifically, the confidence map is given by the ratio of the absolute value of the difference numerator to the sum of the local noise estimate and the small positive number to prevent division by zero. Its dimension can be regarded as the local signal-to-noise ratio. Therefore, applying a threshold to the confidence map can ensure that the signal strength of the pixels entering the statistics is sufficiently significant relative to the neighborhood noise level, thereby reducing the interference of weak signals and random disturbances on the pipeline state determination conclusion. Subsequently, opening and closing operations are performed on the coarse mask in sequence. By utilizing the properties of mathematical morphology regarding the removal of small isolated structures and the filling of narrow holes, an effective data mask with good connectivity and smooth boundaries is obtained, avoiding point artifacts or small holes from spatially destroying statistical stability. Within the pixel set defined by the mask, the mean, standard deviation, and quantiles of the preset percentiles are calculated for the robust multi-node, multi-dimensional monitoring differential index map. The mean is used to reflect the overall operational status of the subsea pipeline monitoring area, the standard deviation is used to measure the dispersion of the monitoring data and the fluctuation of the pipeline's operational status, and the quantiles are used to characterize the middle and tail shapes of the monitoring data distribution. Under the law of large numbers, the quantiles converge as the number of effective data points increases, thus forming a set of statistics that is insensitive to environmental changes and gain drift and has a monotonic correspondence with changes in the operational status of the subsea pipeline. This facilitates horizontal comparison and process traceability of monitoring data across nodes and batches.

[0045] Specifically, based on the actual monitoring data processing requirements and the numerical distribution characteristics of the confidence map, a confidence threshold is preset. This threshold is used to distinguish between pixels that meet the confidence level and pixels that do not meet the confidence level. Each pixel position in the confidence map is traversed, and the confidence value of that position is compared with the preset confidence threshold. If the confidence value of the pixel is greater than or equal to the confidence threshold, the pixel is determined to be a pixel that meets the confidence level and is marked as valid. If the confidence value of the pixel is less than the confidence threshold, the pixel is determined to be a pixel that does not meet the confidence level and is marked as invalid. After completing the above determination and marking for all pixel positions, a preliminary coarse mask is obtained. For the obtained coarse mask, an opening operation is first performed. The opening operation first uses a structuring element of a preset size to erode the coarse mask to remove isolated small noise points in the mask. Then, an expansion operation is performed on the eroded mask to restore the original shape of the main body of the mask. After the opening operation is completed, a closing operation is performed on the mask. The closing operation first uses the same or adapted structuring element to expand the mask after the opening operation to fill the small holes in the main body of the mask. Then, an erosion operation is performed on the expanded mask to trim the shape of the mask edges. After the combined processing of opening and closing operations, an effective data mask with a regular shape and no obvious noise or holes is obtained. Subsequently, using an effective data mask as the selection criterion, all pixels marked as valid states in the robust multi-node multi-dimensional monitoring difference index map are extracted to form an effective pixel set. Within this effective pixel set, the arithmetic mean of all pixel values ​​is calculated to obtain the mean of the robust multi-node multi-dimensional monitoring difference index. Simultaneously, the standard deviation of all pixel values ​​within this effective pixel set is calculated to reflect the degree of dispersion. Furthermore, based on the data distribution characteristics required for submarine pipeline safety monitoring decisions, multiple target percentiles are preset, such as the 10th percentile, 50th percentile, and 90th percentile. After sorting the pixel values ​​within the effective pixel set in ascending order, the values ​​corresponding to the preset percentiles are extracted to obtain the quantiles of each preset percentile. The calculated mean, standard deviation, and quantiles of each preset percentile are integrated to form a statistical set for cross-node comparison and safety monitoring decisions.

[0046] Preferably, the fusion and visualization of multi-parameter monitoring data of subsea pipelines is completed based on the monitoring sample data package, including: layer fusion of the robust multi-node multi-dimensional monitoring differential index map, confidence map, and effective data mask in the monitoring sample data package to generate a multi-dimensional fused monitoring base map; visual annotation of the numerical indicators in the statistical set and overlaying them with the multi-dimensional fused monitoring base map; registration of the overlaid fused monitoring map with the spatial location of the subsea pipeline based on the geographic information data of the IoT subsea pipeline to generate a three-dimensional / two-dimensional visualization map of multi-parameter monitoring data of the subsea pipeline; dynamic updating and interactive design of the visualization map to support multi-view viewing of monitoring data, highlighting of abnormal data, and retrospective comparison of historical data.

[0047] Specifically, layer fusion involves overlaying and fusing different types of monitoring data maps at the pixel level, preserving the core features of each map. A robust multi-node, multi-dimensional monitoring differential index map reflects the core differential features of pipeline operation status, a confidence map reflects the reliability of each data point, and an effective data mask defines the effective monitoring area. The fusion of these three elements creates a multi-dimensional fused monitoring base map that combines distinctiveness, reliability, and effectiveness. Numerical indicators such as mean, standard deviation, and quantiles from the statistical set are visualized using numbers, color codes, and legends, and overlaid on the corresponding positions of the fused monitoring base map, enabling monitoring personnel to intuitively obtain quantitative statistical information. Combined with geographic information system (GIS) data of the subsea pipeline, This includes information such as the spatial orientation, burial depth, pipe diameter, and node location of pipelines. The system integrates monitoring maps with the actual spatial locations of pipelines to accurately register them, generating two-dimensional or three-dimensional maps, achieving a one-to-one correspondence between monitoring data and the physical location of the pipelines. A dynamic update module processes the monitoring data collected in real-time by the Internet of Things according to the method of this invention, updating the visualized maps in real time. Interactive functions are also designed to support monitoring personnel in viewing from multiple perspectives, such as zooming in, zooming out, and panning. Abnormal data points are highlighted according to preset thresholds (e.g., marked in red), and historical monitoring data can be retrieved for retrospective comparison. This provides intuitive and efficient visualization support for safety early warning, fault diagnosis, and operation and maintenance decisions for subsea pipelines.

[0048] Preferably, the robust multi-node multi-dimensional monitoring differential index map, confidence map, effective data mask, and statistical set are packaged to obtain a monitoring sample data package. This package includes: first, confirming that all data items to be packaged correspond to the same subsea pipeline monitoring sample; the robust multi-node multi-dimensional monitoring differential index map is a pixel-level core indicator map after saturation suppression and smoothing; the confidence map is a quantitative map reflecting the reliability of each pixel indicator; the effective data mask is a binary map marking areas with acceptable confidence levels and regular shapes; and the statistical set includes the mean, standard deviation, and quantiles of preset percentiles within the effective data area. The above four data items are standardized according to preset image and data formats to ensure consistency in image size, resolution, and statistical numerical accuracy; simultaneously, the sample's acquisition metadata, including monitoring dimensions, acquisition nodes, acquisition time, signal gain, etc., is associated, allowing all data items in the data package to be traced back to the original acquisition conditions. By using data integration tools, the standardized robust multi-node multi-dimensional monitoring differential index map, confidence map, effective data mask, statistical set and associated metadata are encapsulated into a unified file format, ultimately forming a monitoring sample data package containing complete monitoring data processing results and quantitative indicators. This data package can be directly used for quality comparison and safety operation and maintenance classification decisions of subsea pipeline monitoring data of different batches and different nodes.

[0049] This invention also provides an IoT-based multi-parameter monitoring data fusion and visualization system for subsea pipelines. The system includes: a baseline calibration module for generating a baseline deviation map, a sensor response uniformity correction map, and the gains of a first and second monitoring parameter channels based on a baseline zero-value map and a standard calibration map; a node acquisition and correction module for acquiring four original monitoring data maps at a first and a second IoT acquisition node, under first and second monitoring dimensions respectively, and performing correction processing on the four original monitoring data maps; a consistency module for performing consistency processing on the corrected four original monitoring data maps to obtain four consistent monitoring maps, and calculating a common-mode term based on the pixel intensity of the four consistent monitoring maps; and a difference normalization module for calculating a difference term based on the pixel intensity of the first and second monitoring dimensions of the four consistent monitoring maps, and based on... The common mode term normalizes the difference term to obtain a multi-node, multi-dimensional monitoring difference index map. The confidence robustness module generates a saturated data mask based on four uniform monitoring maps, and performs confidence calculations and robust smoothing on the multi-node, multi-dimensional monitoring difference index map based on the saturated data mask, resulting in a confidence map and a robust multi-node, multi-dimensional monitoring difference index map. The effective domain statistics module performs threshold filtering and morphological cleansing on the confidence map to obtain an effective data mask, and performs statistical calculations on the robust multi-node, multi-dimensional monitoring difference index map based on the effective data mask to obtain a set of statistics. The data packaging module packages the robust multi-node, multi-dimensional monitoring difference index map, confidence map, effective data mask, and set of statistics to obtain a monitoring sample data package. The fusion visualization module completes the fusion visualization of multi-parameter monitoring data of the subsea pipeline based on the monitoring sample data package.

[0050] Preferably, the fusion visualization module includes: a layer fusion unit, used to fuse the robust multi-node multi-dimensional monitoring differential index map, confidence map, and effective data mask in the monitoring sample data package to generate a multi-dimensional fusion monitoring base map; an indicator annotation unit, used to visualize and annotate the numerical indicators in the statistical set and overlay them with the multi-dimensional fusion monitoring base map; a spatial registration unit, used to register the overlaid fusion monitoring map with the spatial location of the subsea pipeline based on the geographic information data of the IoT subsea pipeline to generate a three-dimensional / two-dimensional multi-parameter monitoring data visualization map of the subsea pipeline; and an interactive update unit, used to dynamically update and interactively design the visualization map, supporting multi-view viewing of monitoring data, highlighting of abnormal data, and retrospective comparison of historical data.

[0051] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things, characterized in that, include: S1. Based on the reference zero value map and the standard calibration map, generate the reference deviation map, the sensor response uniformity correction map, the first monitoring parameter channel gain and the second monitoring parameter channel gain; S2. Obtain four original monitoring data images from the first IoT acquisition node and the second IoT acquisition node, respectively, under the conditions of the first monitoring dimension and the second monitoring dimension, and perform correction processing on the four original monitoring data images; S3. Perform uniformization processing on the four original monitoring data images after correction to obtain four uniform monitoring images, and calculate the common mode term based on the pixel intensity of the four uniform monitoring images; S4. Calculate the difference term based on the pixel intensity of the first and second monitoring dimensions of the four consistent monitoring maps, and normalize the difference term based on the common mode term to obtain the multi-node multi-dimensional monitoring difference index map. S5. Generate a saturated data mask based on four consistent monitoring maps, and perform confidence calculation and robust smoothing on the multi-node multi-dimensional monitoring difference index map based on the saturated data mask to obtain a confidence map and a robust multi-node multi-dimensional monitoring difference index map. S6. Threshold filtering and morphological purification are performed on the confidence graph to obtain an effective data mask. Based on the effective data mask, statistical calculations are performed on the robust multi-node multi-dimensional monitoring difference index graph to obtain a set of statistics. S7. Package the robust multi-node multi-dimensional monitoring differential index map, confidence map, effective data mask and statistical set to obtain the monitoring sample data package, and complete the fusion visualization of multi-parameter monitoring data of the subsea pipeline based on the monitoring sample data package.

2. The method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things as described in claim 1, characterized in that, Based on the baseline zero-value plot and the standard calibration plot, a baseline deviation plot, a sensor response uniformity correction plot, and the gains of the first and second monitoring parameter channels are generated, including: Multi-frame averaging noise reduction is performed on the baseline zero-value map to obtain the baseline deviation map; Based on the effective acquisition range defined by the internal parameters of the IoT monitoring terminal and the numerical threshold of the standard calibration map, the effective area of ​​standard calibration is obtained; the reference deviation map is subtracted from the standard calibration map, the average pixel intensity of the subtracted standard calibration map within the effective area of ​​standard calibration is calculated and normalized to obtain the sensor response uniformity correction map. At the first IoT acquisition node, the standard calibration maps of the first monitoring dimension and the second monitoring dimension are respectively subjected to benchmark deviation subtraction and sensing response uniformity correction to obtain the corrected standard calibration map. The average value of the calibrated standard calibration chart within the effective area of ​​the standard calibration is calculated to obtain the gain of the first monitoring parameter channel and the gain of the second monitoring parameter channel.

3. The method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things as described in claim 1, characterized in that, Four raw monitoring data images were acquired from the first and second IoT data acquisition nodes, respectively, under the conditions of the first and second monitoring dimensions. The four raw monitoring data images were then corrected, including: The parameters of the combination of the first monitoring dimension, the second monitoring dimension, the first IoT acquisition node, and the second IoT acquisition node are fixed, and the acquisition time and signal gain of the monitoring dimension are recorded to obtain a metadata group containing the monitoring dimension, acquisition node, acquisition time, and signal gain. Under the same pipeline monitoring attitude triggering condition, based on the metadata group, four original monitoring data maps are obtained corresponding to the first monitoring dimension and the first IoT acquisition node, the first monitoring dimension and the second IoT acquisition node, the second monitoring dimension and the first IoT acquisition node, and the second monitoring dimension and the second IoT acquisition node. The four original monitoring data images were subjected to benchmark deviation subtraction and sensor response uniformity correction, and the channel gain was made consistent using the channel gain of the first monitoring parameter and the channel gain of the second monitoring parameter, respectively, to obtain the corrected four-channel monitoring data stack.

4. The method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things as described in claim 1, characterized in that, The four original monitoring data images after correction are subjected to a standardization process to obtain four standardized monitoring images. The common mode term is then calculated based on the pixel intensity of the four standardized monitoring images, including: Divide the corrected four-channel monitoring data stack by the acquisition time and signal gain respectively to obtain four uniform monitoring images; add the pixel intensity of the same pixel position in the four uniform monitoring images and multiply by one-half to obtain the common mode term.

5. The method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things as described in claim 1, characterized in that, The difference term is calculated based on the pixel intensity of the first and second monitoring dimensions of the four uniform monitoring maps, and the difference term is normalized based on the common mode term to obtain a multi-node, multi-dimensional monitoring difference index map, including: The pixel intensity of the first monitoring dimension consistent monitoring map and the second monitoring dimension consistent monitoring map of the first IoT acquisition node are subtracted to obtain the multi-dimensional difference term of the first acquisition node. The pixel intensity of the first and second monitoring dimension consistent monitoring maps of the second IoT acquisition node is subtracted to obtain the multi-dimensional difference terms of the second acquisition node. The difference term is obtained by subtracting the multi-dimensional difference term of the second acquisition node from the multi-dimensional difference term of the first acquisition node; the difference term is then divided by the sum of the common modulus term and the small positive number to obtain the multi-node multi-dimensional monitoring difference index map.

6. The method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things as described in claim 1, characterized in that, A saturated data mask is generated based on four uniform monitoring maps. Then, confidence scores and robust smoothing are performed on the multi-node, multi-dimensional monitoring difference index map based on the saturated data mask, resulting in a confidence map and a robust multi-node, multi-dimensional monitoring difference index map, including: The pixel intensity at the same pixel location in the four consistent monitoring images is compared. The maximum pixel intensity is compared with the saturation threshold. If the maximum value is greater than or equal to the saturation threshold, the pixel location is marked as a saturated data point; otherwise, it is marked as a non-saturated data point to generate a saturated data mask. The median absolute deviation of the multi-node, multi-dimensional monitoring differential index map is calculated within a preset size neighborhood of unsaturated data points, and the coefficients are converted to obtain a local noise estimate. The confidence plot is obtained by dividing the absolute value of the difference term by the sum of the local noise estimate and the small positive number. Median filtering is performed on the multi-node, multi-dimensional monitoring difference index map under the constraint of saturated data mask to obtain a robust multi-node, multi-dimensional monitoring difference index map.

7. The method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things as described in claim 1, characterized in that, An effective data mask is obtained by thresholding and morphological cleansing of the confidence plot. Based on the effective data mask, a set of statistics is obtained by statistically calculating the robust multi-node, multi-dimensional monitoring difference index plot, including: A coarse mask is obtained by thresholding the confidence map, and an effective data mask is obtained by opening and closing operations on the coarse mask. The mean, standard deviation and quantile of the preset percentile are calculated for the robust multi-node multi-dimensional monitoring difference index map within the pixel set of the effective data mask. The obtained mean, standard deviation and quantile of each preset percentile are combined to form a statistical set.

8. The method for fusing and visualizing multi-parameter monitoring data of subsea pipelines based on the Internet of Things as described in claim 1, characterized in that, Based on monitoring sample data packages, multi-parameter monitoring data of subsea pipelines are fused and visualized, including: The robust multi-node multi-dimensional monitoring differential index map, confidence map, and effective data mask in the monitoring sample data package are fused to generate a multi-dimensional fused monitoring base map. The numerical indicators in the statistical set are visualized and labeled, and then overlaid with a multi-dimensional integrated monitoring base map. Based on the geographic information data of the Internet of Things submarine pipeline, the superimposed fused monitoring map is registered with the spatial location of the submarine pipeline to generate a three-dimensional / two-dimensional visualization map of multi-parameter monitoring data of the submarine pipeline. The visualization map is dynamically updated and interactively designed, supporting multi-view viewing of monitoring data, highlighting of abnormal data, and retrospective comparison of historical data.

9. A multi-parameter monitoring data fusion and visualization system for subsea pipelines based on the Internet of Things, characterized in that, include: The baseline calibration module is used to generate a baseline deviation map, a sensor response uniformity correction map, and the gain of the first monitoring parameter channel and the gain of the second monitoring parameter channel based on the baseline zero value map and the standard calibration map. The node acquisition and correction module is used to acquire four original monitoring data images from the first IoT acquisition node and the second IoT acquisition node, under the conditions of the first monitoring dimension and the second monitoring dimension, and to perform correction processing on the four original monitoring data images. The unification module is used to unify the four original monitoring data images after correction to obtain four unified monitoring images, and to calculate the common mode term based on the pixel intensity of the four unified monitoring images; The differential normalization module is used to calculate the differential term based on the pixel intensity of the first and second monitoring dimensions of the four consistent monitoring maps, and to normalize the differential term based on the common mode term to obtain a multi-node multi-dimensional monitoring differential index map. The confidence robustness module is used to generate a saturated data mask based on four consistent monitoring maps, and to perform confidence calculation and robust smoothing on the multi-node multi-dimensional monitoring difference index map based on the saturated data mask, so as to obtain a confidence map and a robust multi-node multi-dimensional monitoring difference index map. The effective domain statistics module is used to perform threshold filtering and morphological purification on the confidence graph to obtain an effective data mask. Based on the effective data mask, statistical calculations are performed on the robust multi-node multi-dimensional monitoring difference index graph to obtain a set of statistics. The data packaging module is used to package the robust multi-node multi-dimensional monitoring difference index map, confidence map, effective data mask and statistical set to obtain the monitoring sample data package; The fusion visualization module is used to complete the fusion visualization of multi-parameter monitoring data of subsea pipelines based on monitoring sample data packets.

10. The IoT-based multi-parameter monitoring data fusion and visualization system for subsea pipelines according to claim 9, characterized in that, The fusion visualization module includes: The layer fusion unit is used to fuse the robust multi-node multi-dimensional monitoring difference index map, confidence map, and effective data mask in the monitoring sample data package to generate a multi-dimensional fused monitoring base map; the indicator annotation unit is used to visualize and annotate the numerical indicators in the statistical set and overlay them with the multi-dimensional fused monitoring base map. The spatial registration unit is used to register the superimposed fused monitoring map with the spatial location of the subsea pipeline based on the geographic information data of the Internet of Things subsea pipeline, and generate a three-dimensional / two-dimensional visualization map of multi-parameter monitoring data of the subsea pipeline. The interactive update unit is used to dynamically update and interactively design the visualization map, supporting multi-view viewing of monitoring data, highlighting of abnormal data, and retrospective comparison of historical data.

Images