A method and system for maintaining and detecting water heating pipelines

Abstract

The present invention relates to the technical field of image data processing, and specifically to a method and system for maintaining and detecting water heating pipes. First, collect the sequential image data, pipe feature image data, pipe thermal imaging image data, and pipe ordinary imaging data of the pipes in the areas that need to be detected in the floor heating pipes in the building, and construct a water heating pipe image data set. Secondly, identify the water heating pipe image data set by constructing a water heating pipe abnormal state detection model to obtain a pipe deformation abnormal coefficient, a pipe texture abnormal coefficient, and a pipe temperature abnormal coefficient. Finally, according to the pipe deformation abnormal coefficient, the pipe texture abnormal coefficient, and the pipe temperature abnormal coefficient, generate a water heating pipe abnormal coefficient through multi-modal data fusion. Based on the water heating pipe abnormal coefficient, maintain the water heating pipes, thereby improving the maintenance efficiency of the water heating pipes.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and particularly to a method and system for maintaining and detecting water heating pipes. Background Art

[0002] As an important part of the modern building heating system, the operating state of water heating pipes has a crucial impact on the heating effect and safety of buildings. Due to factors such as long-term operation, external environmental changes, and material aging, water heating pipes are prone to problems such as deformation, texture damage, and temperature anomalies. If these problems are not detected and maintained in a timely manner, it may lead to serious consequences such as reduced heating efficiency, water leakage, and even pipe bursting, threatening building safety and increasing maintenance costs. Therefore, real-time monitoring and anomaly detection of the operating state of water heating pipes have become an important research direction in the maintenance of building heating systems.

[0003] Existing water heating pipe detection methods mainly rely on manual inspection or single-sensor measurement, such as observing changes on the pipe surface manually and using thermal imaging equipment to monitor temperature distribution. However, these methods have defects such as low detection accuracy, insufficient efficiency, and limited coverage, and it is difficult to comprehensively and accurately reflect the actual health state of the pipes. At the same time, due to the diversity and complexity of the surface texture changes, deformation characteristics, and temperature distribution of the pipes, a single data source or detection technology is difficult to meet the needs of comprehensive evaluation.

[0004] For this reason, a method and system for maintaining and detecting water heating pipes are proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for maintaining and detecting water heating pipes. First, collect the pipeline time-series image data, pipeline feature image data, pipeline thermal imaging image data, and pipeline ordinary imaging data of the area to be detected in the floor heating pipes in the building, and construct a water heating pipe image dataset; secondly, identify the water heating pipe image dataset by constructing a water heating pipe abnormal state detection model to obtain the pipeline deformation abnormal coefficient, pipeline texture abnormal coefficient, and pipeline temperature abnormal coefficient; finally, according to the pipeline deformation abnormal coefficient, the pipeline texture abnormal coefficient, and the pipeline temperature abnormal coefficient, generate a water heating pipe abnormal coefficient through multi-modal data fusion; based on the water heating pipe abnormal coefficient, maintain the water heating pipes.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Collect the pipeline time-series image data, pipeline feature image data, pipeline thermal imaging image data, and pipeline ordinary imaging data of the area to be detected in the floor heating pipes in the building, and construct a water heating pipe image dataset;

[0008] Build a detection model for abnormal states of water heating pipes to identify the water heating pipe image dataset, and obtain the pipe deformation abnormal coefficient, the pipe texture abnormal coefficient, and the pipe temperature abnormal coefficient;

[0009] The pipe deformation abnormal coefficient is obtained by identifying the pipe time-series image data through the local outlier factor algorithm;

[0010] The pipe texture abnormal coefficient is obtained by identifying the pipe feature image data through a texture feature analysis method;

[0011] The pipe temperature abnormal coefficient is obtained by identifying the pipe thermal imaging image data and the pipe ordinary imaging data through data fusion technology;

[0012] According to the pipe deformation abnormal coefficient, the pipe texture abnormal coefficient, and the pipe temperature abnormal coefficient, generate a water heating pipe abnormal coefficient through multi-modal data fusion; the formula for the multi-modal data fusion is:

[0013] ;

[0014] where, is the water heating pipe abnormal coefficient, is the weight of the pipe deformation abnormal coefficient, is the pipe deformation abnormal coefficient, is the weight of the pipe texture abnormal coefficient, is the pipe texture abnormal coefficient, is the weight of the pipe temperature abnormal coefficient, is the pipe temperature abnormal coefficient, ;

[0015] Based on the water heating pipe abnormal coefficient, maintain the water heating pipes.

[0016] Preferably, obtain the historical image dataset of the water heating pipes and build a detection model for abnormal states of the water heating pipes;

[0017] The detection model for abnormal states of the water heating pipes includes a pipe deformation abnormal module, a pipe texture abnormal module, and a pipe temperature abnormal module;

[0018] The pipe deformation abnormal module identifies the pipe time-series image data in the water heating pipe image dataset to obtain the abnormal deformation of the water heating pipes in the detection area;

[0019] The pipe texture abnormal module identifies the pipe feature image data in the water heating pipe image dataset to obtain the pipe texture abnormal area;

[0020] The pipeline temperature anomaly module identifies pipeline temperature anomalies by recognizing the pipeline thermal imaging image data and pipeline ordinary imaging data in the water heating pipeline image dataset.

[0021] Preferably, the pipeline deformation anomaly module includes a first image preprocessing layer, a first feature extraction layer, an outlier factor calculation layer, an anomaly detection layer, and a result output layer;

[0022] The first image preprocessing layer generates preprocessed first pipeline time-series image data by performing graying, denoising, and alignment processing on the pipeline time-series image data.

[0023] The first feature extraction layer constructs a first pipeline feature vector by extracting the edge intensity, local gradient change, and structural similarity of each frame of image from the first pipeline time-series image data.

[0024] The outlier factor calculation layer analyzes the first pipeline time-series image data using the local outlier factor algorithm based on the first pipeline feature vector to obtain local outlier factors.

[0025] The anomaly detection layer analyzes the local outlier factors and the local outlier factor threshold to identify abnormal feature points and cluster and label them as abnormal regions.

[0026] The result output layer calculates the pipeline deformation anomaly coefficient by identifying the abnormal regions.

[0027] Preferably, the pipeline texture anomaly module includes a second image preprocessing layer, a second feature extraction layer, an abnormal texture detection layer, and a texture output layer;

[0028] The second image preprocessing layer performs denoising, image segmentation, and gray normalization on the pipeline feature image data to obtain first pipeline feature image data.

[0029] The second feature extraction layer constitutes a second pipeline feature by extracting local texture features, global texture features, and detailed texture features from the first pipeline feature image data.

[0030] The abnormal texture detection layer classifies the texture features through a support vector machine classification model to obtain the distribution characteristics of the texture abnormal regions.

[0031] The texture output layer identifies the distribution characteristics of the texture abnormal regions and calculates and obtains the pipeline texture anomaly coefficient.

[0032] Preferably, the pipeline temperature anomaly module includes a third image preprocessing layer, a multi-modal image registration layer, a feature extraction and fusion layer, and a temperature anomaly detection layer;

[0033] The third image preprocessing layer performs denoising, contrast enhancement, and normalization on the pipeline thermal imaging image data and pipeline ordinary imaging data respectively, to obtain the preprocessed first thermal imaging image and first ordinary imaging image;

[0034] The multimodal image registration layer performs feature point matching on the preprocessed thermal imaging image and ordinary imaging image to generate a multimodal image;

[0035] The feature extraction and fusion layer extracts temperature gradient features from the thermal imaging image in the multimodal image and extracts geometric edge features from the ordinary imaging image in the multimodal image to generate a multimodal feature vector;

[0036] The temperature anomaly detection layer analyzes based on the multimodal feature vector and the standard pipeline multimodal feature vector, and calculates and generates a pipeline temperature anomaly coefficient.

[0037] Preferably, the specific calculation formula for the pipeline deformation anomaly coefficient is:

[0038] ;

[0039] where, is the pipeline deformation anomaly coefficient, is the total number of abnormal feature points, is the local outlier factor of the abnormal feature point , is the preset local outlier factor threshold, is the gradient intensity of the abnormal feature point , is the gradient intensity of all abnormal feature points, is the weight of the abnormal feature point ;

[0040] The specific calculation formula for the pipeline texture anomaly coefficient is:

[0041] ;

[0042] where, is the pipeline texture anomaly coefficient, is the number of abnormal texture regions, is the weight of the local texture feature value of the th abnormal texture region, is the local texture feature value of the th abnormal texture region, is the weight of the global texture feature value of the th abnormal texture region, is the global texture feature value of the th abnormal texture region, is the The weight of the detailed texture feature value of the abnormal texture area is the detailed texture feature value of the abnormal texture area is the confidence coefficient of the pipeline texture abnormal area; ;

[0043] The specific calculation formula of the pipeline temperature anomaly coefficient is as follows:

[0044] ;

[0045] Among them, is the pipeline temperature anomaly coefficient, is the number of detected temperature regions, is the weight of the temperature gradient deviation value of the th detection region, is the actually measured temperature gradient of the th detection region, is the temperature gradient of the standard pipeline in the th detection region, is the weight of the geometric edge deviation value of the th detection region, is the total number of edge points in the th detection region, is the detection value of the th edge point in the th detection region, is the standard value of the th edge point in the th detection region, is the confidence coefficient of the pipeline temperature abnormal area, 。

[0046] A water heating pipeline maintenance detection system includes:

[0047] A data acquisition unit that acquires the pipeline sequential image data, pipeline feature image data, pipeline thermal imaging image data, and pipeline ordinary imaging data of the areas that need to be detected in the water heating pipes in the building, and constructs a water heating pipeline image dataset;

[0048] A related abnormal coefficient acquisition unit that constructs a water heating pipeline abnormal state detection model to identify the water heating pipeline image dataset, and obtains a pipeline deformation abnormal coefficient, a pipeline texture abnormal coefficient, and a pipeline temperature abnormal coefficient;

[0049] A water heating pipeline abnormal coefficient acquisition unit that generates a water heating pipeline abnormal coefficient through multi-modal data fusion according to the pipeline deformation abnormal coefficient, the pipeline texture abnormal coefficient, and the pipeline temperature abnormal coefficient; The formula for the multi-modal data fusion is:

[0050] ;

[0051] Among them, is the abnormal coefficient of the water heating pipeline, is the weight of the pipeline deformation abnormal coefficient, is the pipeline deformation abnormal coefficient, is the weight of the pipeline texture abnormal coefficient, is the pipeline texture abnormal coefficient, is the weight of the pipeline temperature abnormal coefficient, is the pipeline temperature abnormal coefficient, ;

[0052] Maintain the water heating pipeline according to the abnormal coefficient of the water heating pipeline.

[0053] Preferably, the pipeline deformation abnormal coefficient is obtained by identifying the pipeline time-series image data through the local outlier factor algorithm;

[0054] The pipeline texture abnormal coefficient is obtained by identifying the pipeline feature image data through the texture feature analysis method;

[0055] The pipeline temperature abnormal coefficient is obtained by identifying the pipeline thermal imaging image data and the pipeline ordinary imaging data through the data fusion technology.

[0056] Preferably, the water heating pipeline abnormal state detection model includes a pipeline deformation abnormal module, a pipeline texture abnormal module and a pipeline temperature abnormal module;

[0057] The pipeline deformation abnormal module identifies the pipeline time-series image data in the water heating pipeline image dataset to obtain the abnormal deformation of the water heating pipeline in the detection area;

[0058] The pipeline texture abnormal module identifies the pipeline feature image data in the water heating pipeline image dataset to obtain the pipeline texture abnormal area;

[0059] The pipeline temperature abnormal module identifies the pipeline thermal imaging image data and the pipeline ordinary imaging data in the water heating pipeline image dataset to obtain the pipeline temperature abnormality.

[0060] Preferably, the pipeline deformation abnormal module includes a first image preprocessing layer, a first feature extraction layer, an outlier factor calculation layer, an abnormal detection layer and a result output layer;

[0061] The pipeline texture abnormal module includes a second image preprocessing layer, a second feature extraction layer, an abnormal texture detection layer and a texture output layer;

[0062] The pipeline temperature anomaly module includes a third image preprocessing layer, a multi-modal image registration layer, a feature extraction and fusion layer, and a temperature anomaly detection layer.

[0063] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0064] 1. The present invention constructs an abnormal state detection model for water heating pipelines. By analyzing the deformation characteristics, texture characteristics, and temperature anomalies of water heating pipelines, the abnormal state of the water heating pipelines is obtained; the deformation abnormal area of the pipeline is located by identifying the deformation characteristics of the pipeline time-series image data; the texture abnormal characteristics on the pipeline surface are effectively identified through the texture analysis of the pipeline feature image data; combined with the multi-modal registration and fusion of thermal imaging image data and ordinary imaging data, the temperature distribution anomaly of the pipeline is quantitatively detected; through the multi-modal data fusion technology, it adapts to different environments and detection requirements, and improves the safety and maintenance efficiency of the operation of water heating pipelines.

[0065] 2. The pipeline deformation anomaly module of the present invention achieves precise detection through multi-level processing. The image noise is effectively reduced and the images are aligned through the preprocessing layer of the pipeline time-series image data, improving the quality of the image data; the key edges and gradient features are extracted from the preprocessed time-series images to ensure the comprehensiveness of the deformation characteristics; the outlier factor calculation layer combines the local outlier factor algorithm to quantify the abnormal feature points; finally, through clustering and region marking, the positioning and quantification of the deformation anomaly are realized, thereby improving the accuracy and reliability of the deformation anomaly detection and providing a scientific basis for pipeline maintenance.

[0066] 3. The pipeline texture anomaly module of the present invention ensures clear and stable data by denoising, image segmentation, and gray normalization of the pipeline feature image data; the local, global, and detailed texture features are extracted from the processed pipeline feature images to enhance the accuracy of anomaly detection; the support vector machine classification model is used to classify the texture features to achieve precise identification of the abnormal area; further, by quantifying the characteristics of the abnormal area, the texture anomaly coefficient is calculated, thereby improving the comprehensiveness and accuracy of the texture anomaly detection.

[0067] 4. The pipeline temperature anomaly module of the present invention ensures the image quality by denoising, contrast enhancement, and normalization of the pipeline thermal imaging and ordinary imaging data; the multi-modal image registration layer fuses multi-source data to improve the matching accuracy of the pipeline temperature characteristics; the feature extraction and fusion layer fully extracts the temperature gradient and geometric edge features to realize the construction of multi-modal feature vectors; the temperature anomaly detection layer accurately calculates the temperature anomaly coefficient through multi-modal feature analysis, thereby improving the accuracy and robustness of the pipeline temperature anomaly detection. Brief Description of the Drawings

[0068] Figure 1Schematic flow chart of a method for maintaining and detecting water heating pipes provided by the present invention;

[0069] Figure 2 Schematic structural diagram of a water heating pipe maintenance and detection system provided by the present invention;

[0070] Figure 3 Schematic structural diagram of a detection model for abnormal states of water heating pipes provided by an embodiment of the present invention;

[0071] Figure 4 Schematic diagram of an abnormal state of a water heating pipe provided by an embodiment of the present invention.

[0072] In the figure: 1. Abnormal deformation of the water heating pipe; 2. Abnormal texture of the water heating pipe. Specific embodiments

[0073] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0074] Please refer to Figure 1 , the present invention provides a method for maintaining and detecting water heating pipes, and the technical solution is as follows:

[0075] Industrial cameras are arranged around the water heating pipes to obtain pipe images;

[0076] Referring to Figure 1 in S10, the pipeline time-series image data, pipeline feature image data, pipeline thermal imaging image data, and pipeline ordinary imaging data of the area to be detected in the floor heating pipes in the building are collected, and a water heating pipe image dataset is constructed;

[0077] Referring to Figure 1 in S20, a detection model for abnormal states of water heating pipes is constructed to identify the water heating pipe image dataset, and an abnormal deformation coefficient of the pipe, an abnormal texture coefficient of the pipe, and an abnormal temperature coefficient of the pipe are obtained;

[0078] Furthermore, a historical image dataset of water heating pipes is obtained, and a detection model for abnormal states of water heating pipes is constructed;

[0079] The detection model for abnormal states of water heating pipes includes an abnormal deformation module of the pipe, an abnormal texture module of the pipe, and an abnormal temperature module of the pipe. Refer to Figure 3 ;

[0080] The pipeline deformation anomaly module identifies the pipeline time-series image data in the water heating pipeline image dataset to obtain the deformation anomaly of the water heating pipeline in the detection area; refer to Figure 4 , the water heating pipeline deformation anomaly 1 is analyzed and identified through the pipeline time-series image data;

[0081] The pipeline texture anomaly module identifies the pipeline feature image data in the water heating pipeline image dataset to obtain the pipeline texture anomaly area; refer to Figure 4 , the water heating pipeline texture anomaly 2 is identified through the pipeline feature image;

[0082] The pipeline temperature anomaly module identifies the pipeline thermal imaging image data and the pipeline ordinary imaging data in the water heating pipeline image dataset to obtain the pipeline temperature anomaly.

[0083] In this embodiment, the industrial camera acquires the pipeline time-series image data, the pipeline feature image data, the pipeline thermal imaging image data, and the pipeline ordinary imaging data of the water heating pipeline, and analyzes the water heating pipeline image data through the water heating pipeline abnormal state detection model. In addition, this embodiment also uses two other models to analyze the water heating pipeline image data. Each model conducts 3 repeated experiments, and calculates the accuracy of the water heating pipeline abnormal area. The specific results are shown in Table 1.

[0084] Model 1 is the water heating pipeline abnormal state detection model proposed in this invention; Model 2 does not add temperature anomaly recognition based on Model 1; Model 3 only analyzes texture anomalies based on Model 1. As can be seen from Table 1, the water heating pipeline abnormal state detection model provided in this embodiment can effectively improve the safety of the water heating pipeline.

[0085] Table 1 Comparison table of the accuracy of pipeline abnormal areas of different models

[0086]

[0087] In this embodiment, by constructing a water heating pipeline abnormal state detection model, multi-dimensional accurate identification of pipeline deformation, texture, and temperature anomalies is achieved. The pipeline deformation anomaly module captures deformation features based on time-series images, the pipeline texture anomaly module extracts texture features to locate abnormal areas, and the pipeline temperature anomaly module fuses thermal imaging and ordinary imaging data to detect temperature anomalies. The cooperation of multiple modules improves the detection accuracy and efficiency, providing reliable support for the early warning of pipeline abnormal states and maintenance decisions.

[0088] Furthermore, the pipeline deformation anomaly module includes a first image preprocessing layer, a first feature extraction layer, an outlier factor calculation layer, an anomaly detection layer, and a result output layer;

[0089] The first image preprocessing layer generates preprocessed first pipeline time-series image data by performing grayscale conversion, denoising, and alignment on the pipeline time-series image data;

[0090] The first feature extraction layer constructs a first pipeline feature vector by extracting the edge intensity, local gradient change, and structural similarity of each frame of image from the first pipeline time-series image data;

[0091] The outlier factor calculation layer analyzes the first pipeline time-series image data using the local outlier factor algorithm based on the first pipeline feature vector to obtain local outlier factors;

[0092] The formula for calculating the local outlier factor is:

[0093] ;

[0094] where, is the local outlier factor of the feature point , is the feature point, is the feature point 's nearest neighbor set, is the local reachability density of the feature point , is the local reachability density of the feature point , is the number of all nearest neighbors of the feature point , is the reachability distance between the feature point and the feature point , is the maximum value function, is the feature point and the feature point 's Euclidean distance, is the distance from the feature point to its th nearest neighbor;

[0095] The anomaly detection layer analyzes the local outlier factor and the local outlier factor threshold to identify anomaly feature points and cluster and label them as anomaly regions;

[0096] When the local outlier factor is greater than the local outlier factor threshold, the feature point at this time is regarded as an anomaly feature point; the local outlier factor threshold is obtained through training with historical plumbing pipeline data;

[0097] The result output layer calculates the pipeline deformation anomaly coefficient by identifying the anomaly region.

[0098] In this embodiment, the accurate detection of pipeline deformation characteristics is achieved through the pipeline deformation anomaly module. Image quality is improved by grayscale conversion, denoising, and alignment processing. Feature vectors are constructed by combining edge intensity, local gradient change, and structural similarity. The local outlier factor algorithm is used to accurately calculate abnormal feature points and cluster and mark abnormal regions. By training the threshold with historical data, the detection sensitivity and adaptability are improved, thereby efficiently generating the pipeline deformation anomaly coefficient, providing a scientific basis for the early warning and maintenance of deformation faults.

[0099] Furthermore, the pipeline texture anomaly module includes a second image preprocessing layer, a second feature extraction layer, an abnormal texture detection layer, and a texture output layer;

[0100] The second image preprocessing layer performs denoising, image segmentation, and grayscale normalization on the pipeline feature image data to obtain the first pipeline feature image data;

[0101] The second feature extraction layer extracts local texture features, global texture features, and detail texture features from the first pipeline feature image data to form the second pipeline feature;

[0102] The local texture features are extracted by local binary pattern, extracting the local texture patterns around each pixel to reflect the roughness of the local area; the global texture features are extracted by gray-level co-occurrence matrix, calculating the co-occurrence relationship of pixel grayscales in the image to extract texture features such as contrast, correlation, and entropy; the detail texture features are extracted by multi-resolution wavelet transform, decomposing the image into frequency components of different scales to extract detail texture features;

[0103] The abnormal texture detection layer classifies the texture features through a support vector machine classification model to obtain the distribution characteristics of the texture abnormal region;

[0104] The texture output layer identifies the distribution characteristics of the texture abnormal region and calculates and obtains the pipeline texture anomaly coefficient.

[0105] In this embodiment, the present invention achieves the comprehensive detection of pipeline texture features through the pipeline texture anomaly module, and improves data quality by denoising, image segmentation, and grayscale normalization. Local, global, and detail texture features are extracted by combining local binary pattern, gray-level co-occurrence matrix, and multi-resolution wavelet transform, and the support vector machine classification model accurately identifies the texture abnormal region. Through classification analysis and abnormal coefficient calculation, the pipeline texture anomaly is accurately located, providing a reliable basis for pipeline status evaluation and maintenance, and significantly improving the detection accuracy and efficiency.

[0106] Furthermore, the pipeline temperature anomaly module includes a third image preprocessing layer, a multi-modal image registration layer, a feature extraction and fusion layer, and a temperature anomaly detection layer;

[0107] The third image preprocessing layer performs denoising, contrast enhancement, and normalization on the pipeline thermal imaging image data and the pipeline ordinary imaging data respectively, to obtain the preprocessed first thermal imaging image and the first ordinary imaging image;

[0108] The multi-modal image registration layer performs feature point matching on the preprocessed thermal imaging image and the ordinary imaging image, to generate a multi-modal image;

[0109] The feature extraction and fusion layer extracts temperature gradient features from the thermal imaging image in the multi-modal image, and extracts geometric edge features from the ordinary imaging image in the multi-modal image, to generate a multi-modal feature vector;

[0110] The temperature anomaly detection layer analyzes based on the multi-modal feature vector and the standard pipeline multi-modal feature vector, and calculates and generates a pipeline temperature anomaly coefficient.

[0111] In this embodiment, the present invention realizes precise detection of the temperature anomaly of the water heating pipeline through the pipeline temperature anomaly module, uses the multi-modal fusion technology of thermal imaging and ordinary imaging data, and combines denoising, contrast enhancement, and normalization processing to improve the image quality. Through the extraction and fusion of temperature gradient features and geometric edge features, it ensures multi-dimensionality and high precision of anomaly detection. The temperature anomaly detection layer effectively compares the actual and standard multi-modal feature vectors, calculates the pipeline temperature anomaly coefficient, provides a scientific basis for pipeline heat loss assessment and maintenance strategies, and greatly improves the detection reliability and accuracy.

[0112] Further, the specific calculation formula of the pipeline deformation anomaly coefficient is:

[0113] ;

[0114] where, is the pipeline deformation anomaly coefficient, is the total number of abnormal feature points, is the local outlier factor of the abnormal feature point , is the preset local outlier factor threshold, is the gradient intensity of the abnormal feature point , is the gradient intensity of all abnormal feature points, is the weight of the abnormal feature point ;

[0115] The specific calculation formula of the pipeline texture anomaly coefficient is:

[0116] ;

[0117] where, is the pipeline texture anomaly coefficient, is the number of anomalous texture regions, is the local texture eigenvalue weight of the th anomalous texture region, is the local texture eigenvalue of the th anomalous texture region, is the global texture eigenvalue weight of the th anomalous texture region, is the global texture eigenvalue of the th anomalous texture region, is the detailed texture eigenvalue weight of the th anomalous texture region, is the detailed texture eigenvalue of the th anomalous texture region; ;

[0118] The specific calculation formula for the pipeline temperature anomaly coefficient is:

[0119] ;

[0120] where, is the pipeline temperature anomaly coefficient, is the number of temperature detection regions, is the temperature gradient deviation value weight of the th detection region, is the actually measured temperature gradient of the th detection region, is the temperature gradient of the standard pipeline in the th detection region, is the geometric edge deviation value weight of the th detection region, is the total number of edge points in the th detection region, is the detection value of the th edge point in the th detection region, is the standard value of the th edge point in the th detection region, is the pipeline temperature anomaly region confidence coefficient, .

[0121] In this embodiment, the abnormal coefficient of pipeline deformation, the abnormal coefficient of pipeline texture, and the abnormal coefficient of pipeline temperature are calculated through specific formulas to achieve a quantitative evaluation of the abnormal state of the water heating pipeline. The deformation abnormal coefficient combines the outlier factor and the gradient intensity to improve the accuracy of deformation detection; the texture abnormal coefficient integrates multi-dimensional texture features and the confidence coefficient to enhance the ability to locate texture abnormalities; the temperature abnormal coefficient is based on the temperature gradient and the geometric edge deviation to accurately reflect the characteristics of the temperature abnormal area.

[0122] Referring to Figure 1 S30 in, according to the abnormal coefficient of pipeline deformation, the abnormal coefficient of pipeline texture, and the abnormal coefficient of pipeline temperature, through multi-modal data fusion, an abnormal coefficient of the water heating pipeline is generated; the formula for the multi-modal data fusion is:

[0123] ;

[0124] Where is the abnormal coefficient of the water heating pipeline, is the weight of the abnormal coefficient of pipeline deformation, is the abnormal coefficient of pipeline deformation, is the weight of the abnormal coefficient of pipeline texture, is the abnormal coefficient of pipeline texture, is the weight of the abnormal coefficient of pipeline temperature, is the abnormal coefficient of pipeline temperature, ; The data is shown in Table 2;

[0125] When the abnormal coefficient of the water heating pipeline is greater than the abnormal coefficient threshold of the water heating pipeline 0.5, by analyzing the abnormal coefficient of pipeline deformation, the abnormal coefficient of pipeline texture, and the abnormal coefficient of pipeline temperature, the abnormal area and the abnormal type are obtained in a timely manner;

[0126] Table 2 Water heating pipeline abnormal data fusion table

[0127]

[0128] According to the abnormal coefficient of the water heating pipeline, the water heating pipeline is maintained.

[0129] Through the abnormal state detection model of water heating pipelines and the multi-modal data fusion technology, the precise evaluation of pipeline deformation, texture and temperature abnormalities is realized, and a comprehensive water heating pipeline abnormality coefficient is generated. The deformation abnormality coefficient combines the outlier factor and the gradient intensity to enhance the deformation recognition ability; the texture abnormality coefficient comprehensively analyzes local, global and detailed texture features to enhance the localization of texture abnormalities; the temperature abnormality coefficient fuses the thermal imaging and ordinary imaging features to accurately reflect the characteristics of the temperature abnormal area. Finally, through the multi-modal data fusion formula, each abnormality coefficient is weighted and integrated to provide a comprehensive and scientific abnormality evaluation result, providing strong support for the efficient maintenance and accurate detection of water heating pipelines.

[0130] Embodiment 2

[0131] In a water heating pipeline system, the long-term operation of the pipeline may be affected by the external environment and internal media, resulting in problems such as deformation, texture abnormality and temperature abnormality, which directly affect the service life and safety of the pipeline. In view of the above problems, the present invention provides a water heating pipeline maintenance and detection system to efficiently and accurately identify pipeline abnormalities.

[0132] The present invention provides a method for maintaining and detecting water heating pipelines. This method is applied to a water heating pipeline maintenance and detection system. For the specific method flow chart and system structure diagram, refer to Figure 1 and Figure 2 In the present invention, all weights are obtained through machine learning training of historical data of water heating pipelines.

[0133] As an implementation manner of the present invention, referring to S10 in Figure 1 S10 is applied to the data acquisition unit of a water heating pipeline maintenance and detection system. The data acquisition unit constructs a water heating pipeline image dataset by collecting pipeline time-series image data, pipeline feature image data, pipeline thermal imaging image data and pipeline ordinary imaging data in the water heating pipes in the building that need to be detected.

[0134] Referring to S20 in Figure 1 S20 is applied to the relevant abnormality coefficient acquisition unit of a water heating pipeline maintenance and detection system. The relevant abnormality coefficient acquisition unit identifies the water heating pipeline image dataset by constructing an abnormal state detection model of the water heating pipeline, and obtains a pipeline deformation abnormality coefficient, a pipeline texture abnormality coefficient and a pipeline temperature abnormality coefficient.

[0135] Furthermore, a historical image dataset of water heating pipelines is obtained, and an abnormal state detection model of the water heating pipeline is constructed;

[0136] The abnormal state detection model of the water heating pipeline includes a pipeline deformation abnormality module, a pipeline texture abnormality module and a pipeline temperature abnormality module;

[0137] The pipeline deformation anomaly module identifies the pipeline time-series image data in the water heating pipeline image dataset to obtain the deformation anomaly of the water heating pipeline in the detection area;

[0138] The pipeline texture anomaly module identifies the pipeline feature image data in the water heating pipeline image dataset to obtain the pipeline texture anomaly area;

[0139] The pipeline temperature anomaly module identifies the pipeline thermal imaging image data and the pipeline ordinary imaging data in the water heating pipeline image dataset to obtain the pipeline temperature anomaly.

[0140] Further, the pipeline deformation anomaly module includes a first image preprocessing layer, a first feature extraction layer, an outlier factor calculation layer, an anomaly detection layer, and a result output layer;

[0141] The first image preprocessing layer performs grayscale conversion, denoising, and alignment processing on the pipeline time-series image data to generate the preprocessed first pipeline time-series image data;

[0142] The first feature extraction layer extracts the edge intensity, local gradient change, and structural similarity of each frame of image from the first pipeline time-series image data to construct a first pipeline feature vector;

[0143] The outlier factor calculation layer analyzes the first pipeline time-series image data using the local outlier factor algorithm based on the first pipeline feature vector to obtain the local outlier factor;

[0144] The anomaly detection layer analyzes the local outlier factor and the local outlier factor threshold to identify the abnormal feature points and cluster and label them as abnormal areas;

[0145] The result output layer identifies the abnormal area and calculates the pipeline deformation anomaly coefficient.

[0146] Use time series analysis algorithms to dynamically monitor pipeline deformation, extract time-related features of pipeline deformation through this analysis, including but not limited to the bending degree, sag degree, and deformation rate of the pipeline, and evaluate the structural changes of the pipeline based on these features. The severity of pipeline deformation is judged according to the deformation rate, bending degree, and amplitude of deformation. Larger bending degrees and higher deformation rates may cause greater pressure on the pipeline structure and a higher risk of failure.

[0147] Further, the pipeline texture anomaly module includes a second image preprocessing layer, a second feature extraction layer, an abnormal texture detection layer, and a texture output layer;

[0148] The second image preprocessing layer performs denoising, image segmentation, and gray normalization on the pipeline feature image data to obtain the first pipeline feature image data;

[0149] The second feature extraction layer constitutes the second pipeline feature by extracting local texture features, global texture features, and detail texture features from the first pipeline feature image data.

[0150] The abnormal texture detection layer classifies the texture features through a support vector machine classification model to obtain the distribution characteristics of the texture abnormal area.

[0151] The texture output layer identifies the distribution characteristics of the texture abnormal area and calculates and obtains the pipeline texture abnormal coefficient.

[0152] By analyzing the texture features of the pipeline image, information such as the roughness, crack width, crack length, color change, edge sharpness, and texture fineness of the pipeline surface is extracted. The texture feature analysis method automatically identifies the corrosion area of the pipeline based on the texture distribution of the image and quantifies the degree of corrosion. The severity of corrosion is evaluated according to the fineness of the texture features, crack width, and length. Areas with wider and longer cracks and smaller texture fineness may indicate more severe corrosion and a higher failure risk of the pipeline.

[0153] Furthermore, the pipeline temperature anomaly module includes a third image preprocessing layer, a multi-modal image registration layer, a feature extraction and fusion layer, and a temperature anomaly detection layer.

[0154] The third image preprocessing layer performs denoising, contrast enhancement, and normalization processing on the pipeline thermal imaging image data and pipeline ordinary imaging data respectively to obtain the preprocessed first thermal imaging image and first ordinary imaging image.

[0155] The multi-modal image registration layer performs feature point matching on the preprocessed thermal imaging image and ordinary imaging image to generate a multi-modal image.

[0156] The feature extraction and fusion layer extracts temperature gradient features from the thermal imaging image in the multi-modal image and extracts geometric edge features from the ordinary imaging image in the multi-modal image to generate a multi-modal feature vector.

[0157] The temperature anomaly detection layer analyzes based on the multi-modal feature vector and the standard pipeline multi-modal feature vector and calculates and generates the pipeline temperature anomaly coefficient.

[0158] Combined with thermal imaging and ordinary imaging image data, the leakage area of the pipeline is detected through data fusion technology. Thermal imaging images are used to detect areas with abnormal temperature distributions, and ordinary imaging images are used to exclude interference from external environmental factors and accurately locate the position of the leakage point. By analyzing the temperature change amplitude in the thermal imaging image and the water stain signs in the ordinary image, the size of the leakage area and the degree of temperature change are related to the severity of the leakage. Areas with a larger temperature change amplitude and a larger leakage area may indicate a more serious leakage problem.

[0159] Furthermore, the specific calculation formula for the pipeline deformation anomaly coefficient is:

[0160] ;

[0161] Where, is the pipeline deformation anomaly coefficient, is the total number of abnormal feature points, is the local outlier factor of the abnormal feature point , is the preset local outlier factor threshold, is the gradient intensity of the abnormal feature point , is the gradient intensity of all abnormal feature points, is the weight of the abnormal feature point ;

[0162] The specific calculation formula for the pipeline texture anomaly coefficient is:

[0163] ;

[0164] Where, is the pipeline texture anomaly coefficient, is the number of abnormal texture regions, is the weight of the local texture feature value of the th abnormal texture region, is the th local texture feature value of the abnormal texture region, is the weight of the global texture feature value of the th abnormal texture region, is the th global texture feature value of the abnormal texture region, is the weight of the detailed texture feature value of the th abnormal texture region, is the th detailed texture feature value of the abnormal texture region, is the confidence coefficient of the pipeline texture abnormal region; ;

[0165] The specific calculation formula for the pipeline temperature anomaly coefficient is as follows:

[0166] ;

[0167] Among them, is the pipeline temperature anomaly coefficient, is the number of temperature detection regions, is the weight of the temperature gradient deviation value of the th detection region, is the actually measured temperature gradient of the th detection region, is the temperature gradient of the standard pipeline in the th detection region, is the weight of the geometric edge deviation value of the th detection region, is the total number of edge points in the th detection region, is the detection value of the th edge point in the th detection region, is the standard value of the th edge point in the th detection region, is the confidence coefficient of the pipeline temperature anomaly region, .

[0168] Refer to Figure 1 S30 in. S30 is applied to the plumbing pipeline anomaly coefficient acquisition unit of a plumbing pipeline maintenance detection system. The plumbing pipeline anomaly coefficient acquisition unit generates a plumbing pipeline anomaly coefficient through multi-modal data fusion based on the pipeline deformation anomaly coefficient, the pipeline texture anomaly coefficient, and the pipeline temperature anomaly coefficient.

[0169] The formula for the multi-modal data fusion is as follows:

[0170] ;

[0171] Among them, is the plumbing pipeline anomaly coefficient, is the weight of the pipeline deformation anomaly coefficient, is the pipeline deformation anomaly coefficient, is the weight of the pipeline texture anomaly coefficient, is the pipeline texture anomaly coefficient, is the weight of the pipeline temperature anomaly coefficient, is the pipeline temperature anomaly coefficient, ; Refer to Table 3 for data;

[0172] Table 3 Plumbing Pipeline Anomaly Data Table

[0173]

[0174] Maintain the plumbing pipes according to the abnormal coefficient of the plumbing pipes.

[0175] A plumbing pipe maintenance and detection system of the present invention uses a plumbing pipe abnormal state detection model and a multi-modal data fusion method to comprehensively analyze deformations, textures, and temperature anomalies, and generate an abnormal coefficient of the plumbing pipes. The deformation abnormal coefficient is based on the outlier factor and the gradient intensity to accurately identify pipe deformation problems; the texture abnormal coefficient fuses local, global, and detailed texture features to accurately locate the abnormal area; the temperature abnormal coefficient combines thermal imaging and ordinary imaging features to comprehensively reflect the temperature change characteristics. The multi-modal data fusion formula weights and integrates each abnormal coefficient to provide a scientific and comprehensive pipe anomaly assessment result, providing a reliable basis for the maintenance and fault detection of plumbing pipes, and improving the detection efficiency and maintenance accuracy.

[0176] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A water heating pipeline maintenance and detection method, comprising: Collect pipeline time-series image data, pipeline feature image data, pipeline thermal imaging image data, and pipeline general imaging data from the plumbing pipeline inspection area in the building to construct a plumbing pipeline image dataset; Constructing a water heating pipe abnormal state detection model to identify the water heating pipe image dataset, and obtaining a pipe deformation abnormality coefficient, a pipe texture abnormality coefficient, and a pipe temperature abnormality coefficient; The pipeline deformation anomaly coefficient is obtained by identifying the pipeline time series image data using a local outlier factor algorithm; The pipeline texture anomaly coefficient is obtained by identifying the second pipeline feature of the pipeline feature image data; The pipeline temperature anomaly coefficient is obtained by identifying the multimodal feature vector of the multimodal image; According to the pipeline deformation anomaly coefficient, the pipeline texture anomaly coefficient and the pipeline temperature anomaly coefficient, a water heating pipeline anomaly coefficient is generated through multimodal data fusion; the formula for the multimodal data fusion is: in, is the abnormal coefficient of water heating pipes, is the weight of pipeline deformation anomaly coefficient, is the abnormal coefficient of pipeline deformation, is the pipeline texture anomaly coefficient weight, is the pipeline texture anomaly coefficient, is the pipe temperature anomaly coefficient weight, is the pipe temperature anomaly coefficient, The specific calculation formula of the pipeline deformation abnormal coefficient is: in, is the abnormal coefficient of pipeline deformation, is the total number of abnormal feature points, Abnormal feature points The local outlier factor, is the preset local outlier factor threshold, Abnormal feature points The gradient strength, is the gradient strength of all abnormal feature points, Abnormal feature points The weight of The specific calculation formula of the pipeline texture anomaly coefficient is: in, is the pipeline texture anomaly coefficient, is the number of abnormal texture areas, For the The weight of local texture feature value in abnormal texture area, For the Local texture feature values of abnormal texture areas, For the The global texture eigenvalue weight of the abnormal texture area, For the Global texture eigenvalues of abnormal texture regions, For the The detail texture feature value weight of abnormal texture area, For the Detailed texture feature values of abnormal texture areas, is the confidence coefficient of the abnormal area of pipeline texture; The specific calculation formula of the pipeline temperature anomaly coefficient is: in, is the pipe temperature anomaly coefficient, is the number of detection temperature zones, For the The temperature gradient deviation value weight of each detection area, For the The actual measured temperature gradient of the detection area, For the The temperature gradient of the standard pipe in each detection area, For the The geometric edge deviation weight of each detection area, For the The total number of edge points in the detection area, For the Within the detection area The detection value of edge points, For the Within the detection area The standard value of the edge point, is the confidence coefficient of the abnormal pipeline temperature area, The plumbing pipes are maintained according to the plumbing pipe abnormality coefficient.

2. A water heating pipe maintenance and detection method according to claim 1, characterized in that: Obtain a historical image dataset of plumbing pipes and build a plumbing pipe abnormality detection model; The water heating pipeline abnormal state detection model includes a pipeline deformation abnormality module, a pipeline texture abnormality module and a pipeline temperature abnormality module; The pipe deformation anomaly module obtains the water heating pipe deformation anomaly in the detection area by identifying the pipe time series image data of the water heating pipe image data set; The pipeline texture anomaly module obtains pipeline texture anomaly areas by identifying pipeline feature image data in the water heating pipeline image dataset; The pipeline temperature anomaly module obtains pipeline temperature anomalies by identifying pipeline thermal imaging image data and pipeline ordinary imaging data in the water heating pipeline image data set.

3. A water heating pipe maintenance and detection method according to claim 2, characterized in that: The pipeline deformation anomaly module includes a first image preprocessing layer, a first feature extraction layer, an outlier factor calculation layer, an anomaly detection layer and a result output layer; The first image preprocessing layer generates preprocessed first pipeline time series image data by performing grayscale conversion, denoising and alignment processing on the pipeline time series image data; The first feature extraction layer constructs a first pipeline feature vector by extracting edge strength, local gradient change and structural similarity of each frame of image from the first pipeline time series image data; The outlier factor calculation layer analyzes the first pipeline time series image data using a local outlier factor algorithm based on the first pipeline feature vector to obtain a local outlier factor; The anomaly detection layer analyzes the local outlier factor and the local outlier factor threshold, identifies abnormal feature points and clusters and marks them as abnormal areas; The result output layer calculates the pipeline deformation anomaly coefficient by identifying the abnormal area.

4. A water heating pipe maintenance and detection method according to claim 2, characterized in that: The pipeline texture anomaly module includes a second image preprocessing layer, a second feature extraction layer, an abnormal texture detection layer and a texture output layer; The second image preprocessing layer performs denoising, image segmentation and grayscale normalization on the pipeline feature image data to obtain first pipeline feature image data; The second feature extraction layer extracts local texture features, global texture features and detail texture features from the first pipeline feature image data to form a second pipeline feature; The abnormal texture detection layer classifies texture features through a support vector machine classification model to obtain the distribution characteristics of texture abnormality areas; The texture output layer identifies the distribution characteristics of the texture abnormality area and calculates and obtains the pipeline texture abnormality coefficient.

5. A water heating pipeline maintenance and detection method according to claim 2, characterized in that: The pipeline temperature anomaly module includes a third image preprocessing layer, a multimodal image registration layer, a feature extraction and fusion layer, and a temperature anomaly detection layer; The third image preprocessing layer performs denoising, contrast enhancement and normalization on the pipeline thermal imaging image data and the pipeline ordinary imaging data respectively, to obtain a preprocessed first thermal imaging image and a first ordinary imaging image; The multimodal image registration layer performs feature point matching on the pre-processed thermal imaging image and the ordinary imaging image to generate a multimodal image; The feature extraction and fusion layer extracts temperature gradient features from the thermal imaging image in the multimodal image, and extracts geometric edge features from the ordinary imaging image in the multimodal image, to generate a multimodal feature vector; The temperature anomaly detection layer performs analysis based on the multimodal feature vector and the standard pipeline multimodal feature vector to calculate and generate a pipeline temperature anomaly coefficient.

6. A water heating pipeline maintenance and detection system, characterized in that: Executing the water heating pipe maintenance and detection method according to claim 1, comprising: A data acquisition unit collects pipeline time-series image data, pipeline feature image data, pipeline thermal imaging image data, and pipeline general imaging data of the area to be inspected in the heating water pipes in the building, and constructs a water heating pipe image dataset; A related abnormality coefficient acquisition unit is configured to construct a water heating pipe abnormality state detection model to identify the water heating pipe image data set, and obtain a pipe deformation abnormality coefficient, a pipe texture abnormality coefficient, and a pipe temperature abnormality coefficient; The water heating pipe abnormality coefficient acquisition unit generates the water heating pipe abnormality coefficient through multimodal data fusion according to the pipe deformation abnormality coefficient, the pipe texture abnormality coefficient and the pipe temperature abnormality coefficient; the formula for the multimodal data fusion is: in, is the abnormal coefficient of water heating pipes, is the weight of pipeline deformation anomaly coefficient, is the abnormal coefficient of pipeline deformation, is the pipeline texture anomaly coefficient weight, is the pipeline texture anomaly coefficient, is the pipe temperature anomaly coefficient weight, is the pipe temperature anomaly coefficient, The plumbing pipes are maintained according to the plumbing pipe abnormality coefficient.

7. A water heating pipe maintenance and detection system according to claim 6, characterized in that: The pipeline deformation anomaly coefficient is obtained by identifying the pipeline time series image data using a local outlier factor algorithm; The pipeline texture anomaly coefficient is obtained by identifying the pipeline feature image data using a texture feature analysis method; The pipeline temperature anomaly coefficient is obtained by identifying the pipeline thermal imaging image data and the pipeline ordinary imaging data through data fusion technology.

8. The water heating pipe maintenance and detection system according to claim 6, characterized in that: The water heating pipeline abnormal state detection model includes a pipeline deformation abnormality module, a pipeline texture abnormality module and a pipeline temperature abnormality module; The pipe deformation anomaly module obtains the water heating pipe deformation anomaly in the detection area by identifying the pipe time series image data in the water heating pipe image data set; The pipeline texture anomaly module obtains pipeline texture anomaly areas by identifying pipeline feature image data in the water heating pipeline image data set; The pipeline temperature anomaly module obtains pipeline temperature anomalies by identifying pipeline thermal imaging image data and pipeline ordinary imaging data in the water heating pipeline image data set.

9. The water heating pipe maintenance and detection system according to claim 6, characterized in that: The pipeline deformation anomaly module includes a first image preprocessing layer, a first feature extraction layer, an outlier factor calculation layer, an anomaly detection layer and a result output layer; The pipeline texture anomaly module includes a second image preprocessing layer, a second feature extraction layer, an abnormal texture detection layer and a texture output layer; The pipeline temperature anomaly module includes a third image preprocessing layer, a multimodal image registration layer, a feature extraction and fusion layer, and a temperature anomaly detection layer.

Images