Water supply pipeline health evaluation method based on image algorithm

Abstract

The invention relates to a water supply pipeline health evaluation method based on an image algorithm, and aims to solve the problem that the health condition of a pipeline is difficult to comprehensively and quickly evaluate in the prior art. According to the invention, the image feature indexes are rapidly obtained by collecting the image data of the inner wall and the water quality of the pipeline and carrying out inner wall defect identification and water quality state detection on the pipeline by using an image identification algorithm. Meanwhile, a water supply pipeline health evaluation index system fusing pipeline environment, operation and historical data is constructed, comprehensive calculation and evaluation are carried out in combination with image feature indexes, and comprehensive and quantitative evaluation of the pipeline health level is achieved. According to the method, various defects of the pipeline can be effectively identified, the water quality condition is accurately evaluated, a scientific decision basis is provided for maintenance, repair and management of the water supply pipeline, the pipeline safety risk is effectively reduced, and the water supply safety is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of pipeline detection, and in particular to a water supply pipeline health assessment method based on an image algorithm. Background Art

[0002] In recent years, water supply pipeline inspection technology based on image algorithms has gradually emerged, bringing new breakthroughs in the inspection and maintenance of water supply pipelines. This technology mainly uses advanced image acquisition equipment, such as pipeline robots and pipeline periscopes, to obtain high-definition image information deep inside the water supply pipeline. These images can intuitively reflect the condition of the inner wall of the pipeline, including various defects such as cracks, corrosion, scaling, and deformation.

[0003] Image algorithms play a core role in water supply pipeline inspection. By analyzing the collected pipeline inner wall images and water quality image data, it is possible to identify pipeline inner wall defects and detect water quality conditions, thereby obtaining image feature indicators.

[0004] However, the current water supply pipeline detection technology based on image algorithms still faces some challenges. On the one hand, the internal environment of the water supply pipeline is complex, and water flow, dirt, sediment, etc. will affect the image quality, bringing difficulties to algorithm analysis; on the other hand, the image characteristics of pipelines of different types, materials and diameters are different, and the adaptability of the algorithm needs to be improved. Moreover, the huge amount of image data generated by large-scale detection places high demands on data transmission and processing speed. At the same time, the high cost of professional equipment and software also limits the widespread application of this technology. In addition, how to effectively combine image recognition results and other pipeline-related information to conduct a comprehensive health evaluation is still a problem that existing technologies need to solve. Summary of the invention

[0005] In order to solve the above problems, the purpose of the present invention is to provide a water supply pipeline health assessment method based on image algorithm to solve the problem in the prior art that it is difficult to comprehensively and quickly evaluate the health status of water supply pipelines, to achieve accurate identification and assessment of pipeline inner wall defects and water quality status, and to combine multiple pipeline-related indicator data to conduct a comprehensive evaluation of the pipeline health level.

[0006] In order to achieve the above object, the present invention designs a water supply pipeline health assessment method based on image algorithm, comprising the following steps:

[0007] Collect the inner wall image and water quality image data of the water supply pipe;

[0008] Based on the image data, an image recognition algorithm is used to identify inner wall defects and detect water quality status of the pipeline to obtain image feature indicators;

[0009] Based on various types of pipeline-related indicator data, a water supply pipeline health evaluation index system is constructed;

[0010] Based on the image feature indicators and other types of pipeline-related indicator data, the indicator system is calculated and evaluated to comprehensively evaluate the pipeline health level.

[0011] Furthermore, the image recognition algorithm includes:

[0012] Pipe wall defect recognition algorithm: collect pipe wall image data, import it into the pipe wall detection model for analysis and judgment, and obtain the pipe wall detection results;

[0013] Water quality status detection algorithm: Collect water quality image data, import it into the water quality detection model for analysis and judgment, and obtain water quality detection results.

[0014] The calculation and evaluation of the indicator system includes:

[0015] Indicator system construction steps: Based on pipeline environmental data, operation data and historical data, a water supply pipeline health evaluation indicator system is constructed;

[0016] Indicator data acquisition step: Based on the indicator system, various data are collected and corresponding indicator values ​​are calculated, wherein the indicator data includes image feature indicators obtained by an image recognition algorithm;

[0017] Health evaluation calculation steps: Based on the obtained indicator data values, the pipeline health level is comprehensively calculated.

[0018] Furthermore, the pipe wall defect recognition algorithm includes the following steps:

[0019] Database construction steps: Collect water supply network defect image samples and annotate them according to defect categories;

[0020] Model training steps: import sample data into the initial pipe wall detection model, perform training and adjust parameters;

[0021] Image acquisition steps: using image acquisition equipment to acquire images of the pipe wall at multiple locations;

[0022] Pipe wall defect recognition step: based on the pre-processed pipe wall image data, import it into the pipe wall detection model for analysis and judgment to obtain the pipe wall image detection result;

[0023] Index calculation steps: Calculate the corresponding index value based on the pipe wall image detection results.

[0024] Furthermore, the water quality status detection algorithm includes the following steps:

[0025] Database construction steps: Collect water quality image datasets of water supply networks and annotate them according to different turbidity levels;

[0026] Model training steps: import sample data into the initial water quality detection model, train it and adjust the parameters;

[0027] Image acquisition steps: using image acquisition equipment to collect water quality images in the pipeline at multiple locations;

[0028] Water quality status detection step: based on the pre-processed water quality image data, import it into the water quality detection model for analysis and judgment to obtain the water quality image detection result;

[0029] Index calculation steps: Calculate the corresponding index value based on the water quality image detection results.

[0030] Furthermore, the water supply pipeline health evaluation index system includes image feature index items, physical information index items and historical maintenance index items;

[0031] The image feature index items include pipe scale, cracks, corrosion, and water quality;

[0032] The physical information indicators include pipe diameter, pipe material, pipe age, and buried depth;

[0033] The historical maintenance index items include the number of maintenance times, the last maintenance time, the maintenance reasons and results.

[0034] The health evaluation calculation step includes the image feature index item score R I , Physical information index score R P , Historical maintenance data index score R N and calculation of pipeline health comprehensive score R;

[0035] The image feature index item score R I The calculation formula is:

[0036] R I =ω G H G +ω C H C +ω R H R +ω W H W

[0037] Among them, H G Score the pipe scale index, with a weight of ω G ;H C Score the crack index with weight ω C ;H R Score the corrosion index with weight ω R;H W Score the water turbidity index, with a weight of ω W ; and satisfy ω G +ω C +ω R +ω W =1

[0038] The physical information index item score R P The calculation formula is:

[0039] R P =ω D H D +ω M H M +ω L H L +ω B H B

[0040] Among them, H D Score the pipe diameter index, with a weight of ω D ;H M Score the pipe index with a weight of ω M ;H L Score the management age index, with a weight of ω L ;H B Score the depth index, with a weight of ω B ; and satisfy ω D +ω M +ω L +ω B =1;

[0041] The historical maintenance data index item score R N The calculation formula is:

[0042] R N =ω N H N +ω P H P +ω T H T

[0043] Among them, H N Score the maintenance frequency indicator, with a weight of ω N ;H P It is the score of the maintenance cause and treatment result index, and the weight is ω P ;H T Score the last maintenance time indicator with a weight of ω T ; and satisfy ω N +ω P +ω T =1;

[0044] The calculation formula of the pipeline health comprehensive score R is:

[0045] R=ω i R I +ω p R P +ω n R N

[0046] Among them, the weight of the image feature index item is ω i ; The weight of the physical information index item is ω p ; The weight of historical maintenance data index item is ω n ; and satisfy ω i +ω p +ω n =1.

[0047] The pipe wall defect recognition algorithm requires the collection of water supply pipe image data in advance. Different defect types need to be manually labeled for each image: pipe scale, crack, corrosion and normal pipe. When the proportion of some defect samples is small, the samples are expanded by angle rotation and noise addition to obtain the sample library required for model training. By applying the convolutional neural network ResNet-18, Con v 2. Con v 3. Con v 4. Con v The 5-layer features f2, f3, f4, and f5 are fused using the following formula:

[0048] F 5 =Conv 1*1 (f 5 )

[0049] F i =Conv 1*1 (f i )+Upsampling(F i+1 )

[0050] P i =Conv 3*3 (F i )

[0051] Where: i = 2, 3, 4, Conv 1*1 The upsampling is 2 times the sampling for 1*1 convolution; Conv 3*3 is a 3*3 convolution; P i is the feature parameter after fusion;

[0052] During the pipe wall defect recognition step, a multi-task learning method is used to simultaneously perform defect recognition and bounding box regression, where the loss function L is:

[0053] L=L cls +L reg ;

[0054]

[0055] Among them, L cls is the loss function for defect recognition; L reg is the loss function of bounding box regression; n is the sample size; n p is the defect sample size. smooth L 1 For smoothing L 1 Loss function; y i is the predicted label; y′ i is the true label; t i is the predicted bounding box parameter; t′ i is the true bounding box parameter.

[0056] The water quality status detection algorithm needs to collect water quality image data of the water supply network in advance. Each picture needs to be manually labeled according to different turbidity levels, which are excellent, good, qualified, and unqualified. The data set contains 1,000 water quality labeled pictures, of which 800 are used as training pictures and 200 are used as test pictures. In the model construction process, the convolutional neural network ResNet-18 is used. The network starts with a 7*7 convolution layer, followed by a 3*3 maximum pooling layer, followed by a typical Conv 2 、Conv 3 、Conv 4 、Conv 5 The residual structure is followed by an average pooling sampling and a fully connected layer with softmax as the output.

[0057] The present invention provides a water supply pipeline health assessment method based on image algorithm. The method collects the inner wall image and water quality image data of the water supply pipeline and uses advanced image recognition algorithm to quickly and accurately detect and identify the inner wall defects and water quality status of the pipeline. The image feature indicators obtained by this method can provide reliable data support and reference basis for pipeline health assessment.

[0058] The innovation of this method lies in the construction of a complete water supply pipeline health evaluation index system, which comprehensively considers image feature indicators as well as pipeline environment, operating status, historical data and other factors, thereby achieving a comprehensive assessment of the pipeline health status. By importing the collected pipe wall image data and water quality image data into the specially constructed pipe wall detection model and water quality detection model for analysis and judgment, various defects on the inner wall of the pipeline can be accurately and efficiently identified, and the status of the water quality of the pipeline can be effectively evaluated.

[0059] In addition, this method also constructs a water supply pipeline health evaluation index system based on pipeline environmental data, operation data and historical data, and improves the accuracy and reliability of pipeline health evaluation through comprehensive analysis of multi-dimensional indicators. At the same time, by collecting various pipeline status data and calculating the corresponding indicator values, and then performing comprehensive calculations based on the obtained indicator data values, the quantitative evaluation results of the pipeline health level are finally obtained.

[0060] In order to further improve the accuracy and robustness of the image recognition algorithm, this method also constructed a water supply network defect image sample database and a water quality image dataset, and used a multi-task learning method to simultaneously perform defect recognition and bounding box regression, greatly improving the efficiency and accuracy of pipe wall defect recognition. At the same time, by pre-collecting a large amount of water supply pipeline image data and manually annotating it, high-quality data support is provided for the training of the image recognition model.

[0061] In general, the water supply pipeline health assessment method based on image algorithm provided by the present invention comprehensively uses a variety of technical means such as image processing and machine learning to achieve a rapid, accurate and comprehensive assessment of the health status of water supply pipelines, providing a reliable decision-making basis for pipeline maintenance and management, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is the overall framework and flow chart of water supply pipeline health assessment in Example 1.

[0063] Figure 2 This is a flow chart of the pipe wall defect recognition algorithm of Example 1.

[0064] Figure 3 This is a flow chart of the water quality status identification algorithm of Example 1.

[0065] Figure 4 This is a schematic diagram of comprehensive index evaluation of water supply pipelines in Example 1. DETAILED DESCRIPTION

[0066] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0067] Example 1.

[0068] like Figure 1 As shown, the water supply pipeline health assessment method based on image algorithm provided in this embodiment mainly includes two core links: image data acquisition, image recognition, and indicator system calculation and evaluation.

[0069] Image data acquisition is the basis for subsequent analysis, mainly collecting the inner wall image and water quality image data of the water supply pipeline. These data can be obtained through image acquisition equipment such as pipeline robots and pipeline periscopes.

[0070] The collected image data then enters the image recognition stage. Figure 1 As shown in the figure, the goal of image recognition is to identify inner wall defects and water quality status of the pipeline, so as to obtain image feature indicators. The specific image recognition process is implemented by the pipe wall defect recognition algorithm and the water quality status recognition algorithm, as shown in Figure 2 and Figure 3 shown.

[0071] like Figure 2 As shown in the figure, the process of pipe wall defect recognition algorithm is as follows:

[0072] First, we build a database, that is, collect defect image samples of the water supply network and annotate them according to the defect category, such as marking image samples as pipe scale, cracks, corrosion or normal pipes. Considering that some defect samples may be small, we can use angle rotation, noise addition and other methods to enhance the data and expand the sample library.

[0073] Next, the model is trained, and the constructed sample data is imported into the initial pipe wall detection model for training and parameter adjustment. In this embodiment, the convolutional neural network ResNet-18 is used as the basic model, and Conv 2 、Conv 3 、Conv 4 、Conv 5 The feature f of the layer 2 、f 3 、f 4 、f 5 , and use the following formula for feature fusion:

[0074] F 5 =Conv 1*1 (f 5 )

[0075] F i =Conv 1*1 (f i )+Upsampling(F i+1 )

[0076] P i =Conv 3*3 (F i )

[0077] Where i = 2, 3, 4; Conv 1 * 1is 1*1 convolution; Upsampling is 2 times sampling; Conv 3 *3 is 3*3 convolution; P i is the feature parameter after fusion.

[0078] After completing the model training, image acquisition can be carried out. The image acquisition equipment can be used to collect images of the pipe wall at multiple locations inside the pipeline.

[0079] After preprocessing, the collected pipeline inner wall image data enters the pipe wall defect recognition step, that is, it is imported into the trained pipe wall detection model for analysis and judgment to obtain the pipe wall image detection results, such as identifying whether there are defects such as pipe scale, cracks or corrosion in the image, and the location and size of the defects can be measured. In the defect detection process, this embodiment adopts a multi-task learning method to perform defect recognition and bounding box regression at the same time, and improves the adaptability of the model to different data by comprehensively considering the loss functions of different tasks and their weights. The calculation formula of the loss function L is as follows:

[0080] L=L cls +L reg

[0081]

[0082] Among them, L cls is the loss function for defect recognition; L reg is the loss function of bounding box regression; n is the sample size; n p is the defect sample size. smooth L 1 For smoothing L 1 Loss function; y i is the predicted label; y′ i is the true label; t i is the predicted bounding box parameter; t′ i is the true bounding box parameter.

[0083] Finally, the index calculation is performed, and the corresponding index value is calculated according to the pipe wall image detection result. For example, the pipe wall defect degree index can be calculated according to the type, number and area of ​​the defects.

[0084] like Figure 3 As shown in the figure, the process of the water quality status detection algorithm is similar to that of the pipe wall defect identification algorithm, including:

[0085] Database construction: Collect water quality image datasets of water supply networks and annotate them according to different turbidity levels, for example, annotate them as excellent, good, qualified, and unqualified. In this embodiment, a dataset containing 1,000 water quality annotated images is used, of which 800 are used as training images and 200 are used as test images.

[0086] Model training: Import sample data into the initial water quality detection model, train it and adjust parameters. Similar to pipe wall defect recognition, this embodiment also uses the convolutional neural network ResNet-18 as the basic model. The network structure starts with a 7*7 convolution layer, followed by a 3*3 maximum pooling layer, followed by a typical Conv 2 、Conv 3 、Conv 4 、Conv 5 The residual structure is then followed by an average pooling sampling and a fully connected layer, and then used so The ftmax function is used as the output layer for classification prediction.

[0087] Image acquisition: Use image acquisition equipment to collect water quality images in multiple locations in the pipeline.

[0088] Water quality status detection: Based on the preprocessed water quality image data, it is imported into the trained water quality detection model to analyze and judge the water quality status and obtain the water quality image detection results, such as judging the turbidity level of water quality.

[0089] Index calculation: Calculate the corresponding index value based on the water quality image detection results. For example, the water quality cleanliness index can be calculated based on the turbidity level.

[0090] Back to Figure 1 ,After completing image recognition, the obtained image feature indicators will be fed back to the ,calculation and evaluation link of the indicator system.

[0091] like Figure 4 As shown in the figure, this link first constructs the indicator system, that is, based on the pipeline environment data, operation data and historical data, a water supply pipeline health evaluation indicator system is constructed. The indicator system includes image feature indicators, physical information indicators and historical maintenance indicators. Among them, the image feature indicators include indicators such as pipe scale, cracks, corrosion, and water quality obtained by the image recognition algorithm; the physical information indicators include information such as pipe diameter, pipe material, pipe age, and burial depth; the historical maintenance indicators include information such as the number of maintenance, the last maintenance time, the reasons for maintenance and results.

[0092] Then, the index data acquisition step is entered, and based on the constructed index system, various data are collected and the corresponding index values ​​are calculated. It should be noted that the image feature index value can be directly obtained by the aforementioned image recognition algorithm.

[0093] Finally, the health evaluation calculation is performed, such as Figure 1 and Figure 4 As shown, based on the obtained data values ​​of various indicators, the pipeline health level is comprehensively calculated.

[0094] In this embodiment, the health evaluation calculation step specifically includes the image feature index item score R I , physical information index score RP, historical maintenance data index score R N and pipeline health comprehensive score R.

[0095] Image feature index score R I The calculation formula is:

[0096] R I =ω G H G +ω C H C +ω R H R +ω W H W

[0097] Among them, H G Score the scale index, with a weight of ω G ;H C Score the crack index with weight ω C ;H R Score the corrosion index with weight ω R ;H w Score the water turbidity index, with a weight of ω W ; and satisfy ω G +ω C +ω R +ω W =1

[0098] The calculation formula of the physical information index item score RP is:

[0099] R P =ω D H D +ω M H M +ω L H L +ω B H B

[0100] Among them, H D Score the pipe diameter index, with a weight of ω D ;H M Score the pipe index with a weight of ω M ;H L Score the management age index, with a weight of ω L ;H B Score the depth index, with a weight of ω B ; and satisfy ω D +ω M +ω L +ωB =1;

[0101] The historical maintenance data index item score R N The calculation formula is:

[0102] R N =ω N H N +ω P H P +ω T H T

[0103] Among them, H N Score the maintenance frequency indicator, with a weight of ω N ;H P It is the score of the maintenance cause and treatment result index, and the weight is ω P ;H T Score the last maintenance time indicator with a weight of ω T ; and satisfy ω N +ω P =1;

[0104] The calculation formula of the pipeline health comprehensive score R is:

[0105] R=ω i R I +ω p R P +ω n R N

[0106] Among them, the weight of the image feature index item is ω i ; The weight of the physical information index item is ω p ; The weight of historical maintenance data index item is ω n ; and satisfy ω i +ω p +ω n =1.

[0107] Through the above embodiments, a comprehensive assessment of the health of the water supply pipeline can be achieved, providing a scientific decision-making basis for the maintenance and management of the pipeline. It should be noted that the weight of each evaluation index can be adjusted according to actual needs.

[0108] In the description of the present invention, it should be noted that the terms "vertical", "up", "down", "horizontal", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0109] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0110] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A water supply pipeline health assessment method based on image algorithm, characterized in that: The following steps are involved: Collect the inner wall image and water quality image data of the water supply pipe; Based on the image data, an image recognition algorithm is used to identify inner wall defects and detect water quality status of the pipeline to obtain image feature indicators; Based on various types of pipeline-related indicator data, a water supply pipeline health evaluation index system is constructed; Based on the image feature indicators and other types of pipeline-related indicator data, the indicator system is calculated and evaluated to comprehensively evaluate the pipeline health level.

2. The water supply pipeline health assessment method according to claim 1, characterized in that: The image recognition algorithm includes: Pipe wall defect recognition algorithm: collect pipe wall image data, import it into the pipe wall detection model for analysis and judgment, and obtain the pipe wall detection results; Water quality status detection algorithm: Collect water quality image data, import it into the water quality detection model for analysis and judgment, and obtain water quality detection results.

3. The water supply pipeline health assessment method according to claim 1, characterized in that: The calculation and evaluation of the indicator system includes: Indicator system construction steps: Based on pipeline environmental data, operation data and historical data, a water supply pipeline health evaluation indicator system is constructed; Indicator data acquisition step: Based on the indicator system, various data are collected and corresponding indicator values ​​are calculated, wherein the indicator data includes image feature indicators obtained by an image recognition algorithm; Health evaluation calculation steps: Based on the obtained indicator data values, the pipeline health level is comprehensively calculated.

4. The water supply pipeline health assessment method according to claim 2, characterized in that: The pipe wall defect recognition algorithm comprises the following steps: Database construction steps: Collect water supply network defect image samples and annotate them according to defect categories; Model training steps: import sample data into the initial pipe wall detection model, perform training and adjust parameters; Image acquisition steps: using image acquisition equipment to acquire images of the pipe wall at multiple locations; Pipe wall defect recognition step: based on the pre-processed pipe wall image data, import it into the pipe wall detection model for analysis and judgment to obtain the pipe wall image detection result; Index calculation steps: Calculate the corresponding index value based on the pipe wall image detection results.

5. The water supply pipeline health assessment method according to claim 2, characterized in that: The water quality status detection algorithm comprises the following steps: Database construction steps: Collect water quality image datasets of water supply networks and annotate them according to different turbidity levels; Model training steps: import sample data into the initial water quality detection model, train it and adjust the parameters; Image acquisition steps: using image acquisition equipment to collect water quality images in the pipeline at multiple locations; Water quality status detection step: based on the pre-processed water quality image data, import it into the water quality detection model for analysis and judgment to obtain the water quality image detection result; Index calculation steps: Calculate the corresponding index value based on the water quality image detection results.

6. The water supply pipeline health assessment method according to claim 3, characterized in that: The water supply pipeline health evaluation index system includes image feature index items, physical information index items and historical maintenance index items; The image feature index items include pipe scale, cracks, corrosion, and water quality; The physical information indicators include pipe diameter, pipe material, pipe age, and buried depth; The historical maintenance index items include the number of maintenance times, the last maintenance time, the maintenance reasons and results.

7. The water supply pipeline health assessment method according to claim 3 or 6, characterized in that: The health evaluation calculation step includes the image feature index item score R I , Physical information index score R P , Historical maintenance data index score R N and calculation of pipeline health comprehensive score R; The image feature index item score R I The calculation formula is: R I =ω G H G +oh C H C +oh R H R +oh W H W Among them, H G Score the scale index, with a weight of ω G ;H C Score the crack index with weight ω C ;H R Score the corrosion index with weight ω R ;H W Score the water turbidity index, with a weight of ω W ; and satisfy ω G +ω C +ω R +ω W =1 The physical information index item score R P The calculation formula is: R P =ω D H D +oh M H M +oh L H L +oh B H B Among them, H D Score the pipe diameter index, with a weight of ω D ;H M Score the pipe index with a weight of ω M ;H L Score the management age indicator, with a weight of ω L ,H B Score the depth index, with a weight of ω B ; and satisfy ω D +ω M +ω L +ω B =1; The historical maintenance data index item score R N The calculation formula is: R N =ω N H N +oh P H P +oh T H T Among them, H N Score the maintenance frequency indicator, with a weight of ω N ;H P It is the score of the maintenance cause and treatment result index, and the weight is ω P , H T Score the last maintenance time indicator with a weight of ω T ; and satisfy ω N +ω P +ω T =1; The calculation formula of the pipeline health comprehensive score R is: R=ω i R I +oh p R P +oh n R N Among them, the weight of the image feature index item is ω i ; The weight of the physical information index item is ω p ; The weight of historical maintenance data index item is ω n ; and satisfy ω i +ω p +ω n =1.

8. The water supply pipeline health assessment method according to claim 4, characterized in that: The pipe wall defect recognition algorithm requires the collection of water supply pipe image data in advance. Different defect types need to be manually labeled for each image: pipe scale, cracks, corrosion and normal pipes. When the proportion of some defect samples is small, the samples are expanded by angle rotation and noise addition to obtain the sample library required for model training. By applying the convolutional neural network ResNet-18, the features f2, f3, f4, and f5 of the Conv2, Conv3, Conv4, and Conv5 layers are extracted, and the following formula is used for feature fusion: F5=Conv 1*1 (f5) F i =Conv 1*1 (f i )+Upsampling(F i+1 ) P i =Conv 3*3 (F i ) Among them: i = 2, 3, 4, Conv1*1 is 1*1 convolution; Upsampling is 2 times sampling; Conv3*3 is 3*3 convolution; Pi is the feature parameter after fusion; During the pipe wall defect recognition step, a multi-task learning method is used to simultaneously perform defect recognition and bounding box regression, where the loss function L is: L=L cls +L reg ; Among them, L cls is the loss function for defect recognition; L reg is the loss function of bounding box regression; n is the sample size; np is the defect sample size. smooth L1 is the smooth L1 loss function; y i is the predicted label; y' i is the true label; t i is the predicted bounding box parameter; t' i is the true bounding box parameter.

9. The water supply pipeline health assessment method according to claim 5, characterized in that: The water quality status detection algorithm needs to collect water quality image data of the water supply network in advance. Each picture needs to be manually labeled according to different turbidity levels, namely excellent, good, qualified, and unqualified. The data set contains 1,000 water quality labeled pictures, of which 800 are used as training pictures and 200 are used as test pictures. In the model construction process, the convolutional neural network ResNet-18 is used. The network starts with a 7*7 convolution layer, followed by a 3*3 maximum pooling layer, followed by a typical Conv2, Conv3, Conv4, and Conv5 residual structure, followed by an average pooling sampling and a fully connected layer, with softmax as the output.

Images