Municipal pipeline leakage point rapid detection method using thermal imaging and AI

Abstract

The invention discloses a municipal pipeline leakage point rapid detection method using thermal imaging and AI, and belongs to the technical field of municipal pipeline detection. The method comprises the following steps: collecting thermal imaging data of a pipeline and a surrounding environment and recording geographic position information through a thermal imaging camera carried by an unmanned aerial vehicle or an inspection vehicle; performing denoising, enhancement and calibration preprocessing on the thermal imaging data; using the trained deep learning model to identify leakage points; positioning leakage points by combining geographical location information and evaluating severity of the leakage points; and finally, storing the management data and generating an early warning and report. The municipal pipeline leakage point detection method solves the problems of low efficiency and poor accuracy of a traditional detection method, realizes rapid and accurate detection of municipal pipeline leakage points, improves the safety and reliability of an urban pipeline system, and has wide applicability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of municipal pipeline detection, and particularly relates to a municipal pipeline leakage point rapid detection method using thermal imaging and AI. BACKGROUND

[0002] As an important part of urban infrastructure, municipal pipelines bear the important task of transporting water, gas, oil and other media. However, due to the influence of geological condition changes, corrosion, external pressure and other factors, leakage accidents occur from time to time. Pipeline leakage not only causes resource waste and environmental pollution, but also may cause safety accidents, which seriously affects the normal operation of the city and the life of residents.

[0003] Traditional municipal pipeline leakage detection methods mainly include manual inspection, pressure testing, acoustic detection, etc. Manual inspection relies on manual experience, has low efficiency and strong subjectivity, and it is difficult to find the leakage point of the deeply buried pipeline; pressure testing needs to stop the pipeline operation, which affects the normal supply, and it is difficult to accurately detect small leakage; acoustic detection is greatly disturbed by environmental noise, and the precision is limited. With the continuous expansion of the city scale and the increasing complexity of the pipeline system, the traditional detection methods have been difficult to meet the demand of rapid and accurate detection of pipeline leakage points.

[0004] In recent years, thermal imaging technology and artificial intelligence technology have developed rapidly and are widely used in many fields. Thermal imaging technology can obtain the temperature distribution information of an object by detecting the infrared radiation of the object's surface, and the temperature at the leakage point of the pipeline is usually different from the surrounding environment, so thermal imaging technology provides a new way for pipeline leakage detection. At the same time, artificial intelligence technology, especially deep learning algorithms, has shown strong capabilities in image recognition, data analysis, etc., and can efficiently process and analyze thermal imaging data to realize automatic identification and positioning of pipeline leakage points. SUMMARY

[0005] The purpose of the present application is to provide a municipal pipeline leakage point rapid detection method using thermal imaging and AI to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides the following technical scheme:

[0007] The municipal pipeline leakage point rapid detection method using thermal imaging and AI comprises the following steps:

[0008] Thermal imaging data acquisition: using a UAV or a patrol vehicle equipped with a thermal imaging camera, patrol along the municipal pipeline laying route, collect thermal imaging information of the pipeline and the surrounding environment, and record the geographical position information at the time of collection;

[0009] Thermal imaging data preprocessing: the collected thermal imaging data is sequentially denoised, image enhanced and calibrated;

[0010] AI-based leak point identification: a deep learning model is constructed, the model is trained, verified and tested using labeled thermal imaging image data, and the trained model is used to identify the pipeline leak point and location information in the preprocessed thermal imaging image;

[0011] Leak point positioning and evaluation: the actual location of the leak point is determined in combination with the geographic location information, and the severity of the leak is evaluated according to the temperature characteristics and area size of the leak point in the thermal imaging image;

[0012] Data storage and management: the collected thermal imaging data, preprocessed data, model training data and detection results are uniformly stored and managed, and a data backup and encryption mechanism is established;

[0013] Early warning and report generation: after detecting the leak point, an early warning message is sent and a report containing detection-related information is automatically generated.

[0014] Preferably, in the thermal imaging data acquisition step, the parameters of the thermal imaging camera are optimized according to the type of pipeline, pipe diameter, buried depth and surrounding environment, and the unmanned aerial vehicle or inspection vehicle maintains stable operation.

[0015] Preferably, in the thermal imaging data preprocessing step, the denoising process uses Gaussian filter or median filter algorithm; the image enhancement process uses histogram equalization or contrast stretching method; the calibration process corrects the thermal imaging data according to the known standard temperature source.

[0016] Preferably, in the AI-based leak point identification step, the deep learning model uses a convolutional neural network architecture, including an improved YOLO series or FasterR-CNN model; the labeled data includes thermal imaging images of normal pipelines and leaking pipelines, and the location and type of leak points are marked, and divided into training set, validation set and test set.

[0017] Preferably, when using the training set to train the deep learning model, adjust the model parameters to optimize performance, evaluate the training effect through the validation set and adjust the training strategy, and use the test set to evaluate the accuracy and recall rate of the model.

[0018] Preferably, in the leak point positioning and evaluation step, the scale and impact range of the leak are estimated by establishing a relationship model between temperature and leak rate and a relationship model between leak area and leak volume.

[0019] Preferably, in the data storage and management step, the data is stored in a structured manner using MySQL or MongoDB database, and important data is backed up regularly.

[0020] Preferably, in the warning and report generation step, the warning information is notified to relevant personnel via SMS, email or sound and light alarm, and the report includes the detection time, detection area, pipeline information, detection results and treatment suggestions.

[0021] Preferably, the municipal pipeline includes a water supply pipeline and a gas pipeline.

[0022] Preferably, during the thermal imaging data acquisition process, the flight altitude of the UAV is set to 10 meters and the flight speed is set to 5 meters per second; the driving speed of the inspection vehicle is set to 10 kilometers per hour.

[0023] Compared with the existing technology, the present invention provides a method for rapid detection of municipal pipeline leaks using thermal imaging and AI, which has the following beneficial effects:

[0024] Improved inspection efficiency: Using drones or patrol vehicles equipped with thermal imaging cameras for rapid inspections can quickly cover large areas of municipal pipelines, significantly improving inspection efficiency compared to manual inspections. Furthermore, AI-based leak point identification algorithms can quickly process large amounts of thermal imaging data, enabling real-time detection of leaks and reducing inspection time.

[0025] Improved detection accuracy: Thermal imaging technology can intuitively reflect the temperature distribution of pipelines and their surroundings. The deep learning model, trained with large amounts of data, can accurately identify leakage characteristics in thermal imaging images, effectively avoiding misjudgments and missed detections caused by human factors and environmental interference in traditional detection methods, thereby improving detection accuracy.

[0026] Achieve rapid positioning and assessment: Combining geographic location information and thermal imaging data, it can quickly and accurately locate pipeline leaks and assess the severity of the leak, providing strong support for timely and effective repair measures and reducing the impact of leaks on urban operations and the environment.

[0027] Facilitate data management and analysis: Establish a comprehensive data storage and management system to facilitate the long-term preservation and analysis of test data. By mining and analyzing historical data, we can understand the operating conditions and leakage patterns of pipelines, provide decision-making basis for pipeline maintenance and management, and realize intelligent management of municipal pipeline systems. DETAILED DESCRIPTION

[0028] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0029] The present invention provides a method for quickly detecting leaks in municipal pipelines using thermal imaging and AI, comprising the following steps:

[0030] Thermal imaging data collection: Use drones or inspection vehicles equipped with thermal imaging cameras to patrol along municipal pipeline laying routes, collect thermal imaging information of the pipelines and surrounding environment, and record the geographic location information at the time of collection;

[0031] Thermal imaging data preprocessing: the collected thermal imaging data is subjected to denoising, image enhancement and calibration processing in sequence;

[0032] AI-based leak point identification: Build a deep learning model and use annotated thermal imaging data to train, verify, and test the model. The trained model can then identify pipeline leak points and their locations in pre-processed thermal imaging images.

[0033] Leak point location and assessment: Determine the actual location of the leak point based on geographic location information, and assess the severity of the leak based on the temperature characteristics and area size of the leak point in the thermal imaging image;

[0034] Data storage and management: uniformly store and manage collected thermal imaging data, pre-processed data, model training data, and test results, and establish data backup and encryption mechanisms;

[0035] Early warning and report generation: After a leak is detected, an early warning message is issued and a report containing detection-related information is automatically generated.

[0036] Implementation Method

[0037] Example 1: Leakage detection of urban water supply pipelines

[0038] Thermal imaging data acquisition

[0039] A drone equipped with a FLIRA655sc thermal imaging camera was used to inspect the city's water supply pipelines. This thermal imaging camera boasts high resolution (640 x 512 pixels) and high accuracy (±2°C or ±2%), enabling clear thermal imaging of the water supply pipelines and their surroundings. Based on the pipeline's diameter and burial depth, the drone's flight altitude was set to 10 meters and its speed to 5 meters per second, ensuring full coverage of the pipeline area.

[0040] During the flight, the drone uses the GPS positioning system to record the flight trajectory and geographic location information of the collected data in real time, and transmits the thermal imaging data to the ground control station in real time.

[0041] Thermal imaging data preprocessing

[0042] At the ground control station, the received thermal imaging data is pre-processed. First, a Gaussian filter algorithm is used to denoise the image, with a Gaussian kernel size of 3x3 and a standard deviation of 1.5, effectively removing noise interference in the image. Then, the image is enhanced through histogram equalization, significantly improving the contrast of temperature differences in the image. Finally, the thermal imaging data is calibrated using known standard temperature sources to ensure the accuracy of temperature measurement.

[0043] AI-based leak point identification

[0044] An improved YOLOv8-based deep learning model is constructed. 5000 thermal imaging image data containing normal water supply pipes and leaking water supply pipes are collected, including 2000 leaking images and 3000 normal images. These data are labeled to mark the location and type of leak point (water leakage). The data are divided into training set, validation set and test set according to the ratio of 70%, 15% and 15%.

[0045] The YOLOv8 model is trained using the training set, with 100 training rounds and a learning rate of 0.001. During training, data augmentation techniques such as random flipping, rotation, scaling, etc. are used to increase data diversity and improve model generalization. The training effect of the model is evaluated through the validation set, and the model's hyperparameters are adjusted according to the evaluation results. The trained model is tested using the test set, with an accuracy of 95% and a recall rate of 93%, meeting the actual application requirements.

[0046] Leak point positioning and evaluation

[0047] When the trained YOLOv8 model detects a water supply pipe leak point in the thermal imaging image, it can accurately determine the location of the leak point in the actual geographical environment by combining the geographical location information recorded by the unmanned aerial vehicle. For example, by matching with geographic information system (GIS) data, the coordinates of the leak point can be accurately marked on the map.

[0048] According to the temperature characteristics and area size of the leak point in the thermal imaging image, the severity of the leak is evaluated. Through the established relationship model between temperature and leak rate and the relationship model between leak area and leak volume, the leak rate of the leak point is estimated to be 5 liters / minute, and the leak volume is expected to reach 300 liters within 1 hour, providing an important basis for timely repair measures.

[0049] Data storage and management

[0050] The collected thermal imaging data, pre-processed data, model training data, and detection results are stored in a MySQL database. The data is stored in a structured manner, with fields such as detection time, geographic location, pipe type, thermal imaging image data, and detection results. A data backup mechanism is established, with full backups of the database taken every morning and stored in an off-site server to ensure data security.

[0051] Early warning and report generation

[0052] When a water supply pipeline leak is detected, the system immediately notifies the city's water management department and maintenance personnel via SMS and email. The SMS content includes the location of the leak, the type of leak, the severity, and other key information. At the same time, a detailed detection report is automatically generated, containing information such as detection time, detection area, water supply pipeline information, detection results, and treatment recommendations. Maintenance personnel can promptly repair the leak based on the information in the report, effectively reducing water waste and the impact on city water supply.

[0053] Example 2: City gas pipeline leak detection

[0054] Thermal imaging data collection

[0055] A patrol vehicle equipped with a FLIRE86 thermal imaging camera is used to detect city gas pipelines. The FLIRE86 thermal imaging camera has fast response and high sensitivity, making it suitable for collecting thermal imaging data of gas pipelines during movement. The patrol vehicle travels along the laid route of the gas pipeline at a speed of 10 kilometers per hour, and the thermal imaging camera collects real-time thermal imaging data of the pipeline and its surrounding environment, and transmits the data to the monitoring center through the vehicle-mounted data transmission system.

[0056] During the collection process, the vehicle-mounted GPS positioning system and inertial navigation system are used to accurately record the position and travel trajectory of the patrol vehicle, providing accurate position information for subsequent leak point positioning.

[0057] Thermal imaging data preprocessing

[0058] In the monitoring center, the transmitted thermal imaging data is pre-processed. First, the median filter algorithm is used to remove salt and pepper noise in the image, with a filter window size of 5x5. Then, the contrast stretching method is used to enhance the contrast of temperature differences in the image, making the thermal characteristics of gas leaks more prominent. Finally, the thermal imaging data is temperature calibrated to ensure the accuracy of temperature measurement.

[0059] AI-based leak point identification

[0060] FasterR-CNN deep learning model is used for gas pipeline leak point identification. 3000 thermal imaging image data containing normal gas pipelines and leaking gas pipelines are collected, including 1000 leaking images and 2000 normal images. The data is labeled to mark the location and type (gas leak) of the leak point. The data is divided into training set, validation set and test set in the ratio of 60%, 20% and 20%.

[0061] The FasterR-CNN model is trained using the training set, and the training parameters such as learning rate, iteration number, etc. are set. During the training process, the hyperparameters of the model are constantly adjusted through the validation set to improve the performance of the model. The trained model is tested using the test set, and the accuracy of the model reaches 92%, and the recall rate reaches 90%, which can meet the actual needs of gas pipeline leak detection.

[0062] Leak point positioning and evaluation

[0063] When the FasterR-CNN model detects a gas pipeline leak point in the thermal imaging image, it can accurately determine the location of the leak point in the actual geographical environment by combining the location information recorded by the inspection vehicle. For example, by matching with city map data, the specific location of the leak point can be clearly marked on the map.

[0064] According to the temperature change of the leak point in the thermal imaging image and the size of the thermal anomaly area, the severity of the leak is evaluated. Through the established related model, it is estimated that the gas leakage of this leak point is 0.5 cubic meters per hour, and the leak may have some impact on the surrounding environment and the safety of residents, and timely measures need to be taken to handle it.

[0065] Data storage and management

[0066] All related data is stored in the MongoDB database, and the document type data structure of MongoDB is suitable for storing unstructured data such as thermal imaging image data. The data is stored in a classified manner to facilitate query and management. A data encryption mechanism is established to encrypt sensitive data and ensure data security.

[0067] Early warning and report generation

[0068] Once a gas pipeline leak point is detected, the system immediately issues an audible and visual alarm, and notifies the safety management department and repair personnel of the gas company through SMS and email. The early warning information details the location of the leak point, the type of leak and the severity. At the same time, a detection report is automatically generated, which contains the data and analysis results during the detection process, as well as the handling suggestions for the leak situation. After receiving the notification, the repair personnel quickly carry professional equipment to the leak point for repair, effectively avoiding possible safety accidents.

[0069] It should be pointed out finally that the above only describes the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent replacements to some of the technical features, as long as the modifications, equivalent replacements or improvements are within the spirit and principle of the present application, and should be included in the protection scope of the present application.

Claims

1. A rapid detection method for municipal pipeline leaks using thermal imaging and AI, characterized by: The following steps are involved: Thermal imaging data collection: Use drones or inspection vehicles equipped with thermal imaging cameras to patrol along municipal pipeline laying routes, collect thermal imaging information of the pipelines and surrounding environment, and record the geographic location information at the time of collection; Thermal imaging data preprocessing: the collected thermal imaging data is subjected to denoising, image enhancement and calibration processing in sequence; AI-based leak point identification: Build a deep learning model and use annotated thermal imaging data to train, verify, and test the model. The trained model can then identify pipeline leak points and their locations in pre-processed thermal imaging images. Leak point location and assessment: Determine the actual location of the leak point based on geographic location information, and assess the severity of the leak based on the temperature characteristics and area size of the leak point in the thermal imaging image; Data storage and management: uniformly store and manage collected thermal imaging data, pre-processed data, model training data, and test results, and establish data backup and encryption mechanisms; Early warning and report generation: After a leak is detected, an early warning message is issued and a report containing detection-related information is automatically generated.

2. The method according to claim 1, characterized in that In the thermal imaging data acquisition step, the parameters of the thermal imaging camera are optimized according to the pipeline type, diameter, burial depth and surrounding environment, and the drone or inspection vehicle maintains stable operation.

3. The method according to claim 1, characterized in that In the thermal imaging data preprocessing step, the denoising process adopts Gaussian filtering or median filtering algorithm; the image enhancement process adopts histogram equalization or contrast stretching method; the calibration process corrects the thermal imaging data according to a known standard temperature source.

4. The method according to claim 1, wherein In the AI-based leak point identification step, the deep learning model adopts a convolutional neural network architecture, including an improved YOLO series or FasterR-CNN model; the labeled data includes thermal imaging images of normal pipelines and leaking pipelines, and the location and type of the leak point are marked, and divided into training set, validation set and test set.

5. The method according to claim 4, characterized in that When training a deep learning model using a training set, adjust model parameters to optimize performance, evaluate the training effect and adjust the training strategy using a validation set, and use a test set to evaluate the model's accuracy and recall.

6. The method according to claim 1, characterized in that In the leakage point location and assessment step, the scale and impact range of the leakage are estimated by establishing a relationship model between temperature and leakage rate and a relationship model between leakage area and leakage amount.

7. The method according to claim 1, characterized in that In the data storage and management step, MySQL or MongoDB database is used to perform structured storage of data, and important data is backed up regularly.

8. The method according to claim 1, characterized in that In the warning and report generation step, the warning information is notified to relevant personnel via text messages, emails or sound and light alarms, and the report includes the detection time, detection area, pipeline information, detection results and treatment suggestions.

9. The method according to claim 1, characterized in that The municipal pipeline includes a water supply pipeline and a gas pipeline.

10. The method according to claim 1, characterized in that During the thermal imaging data collection process, the drone's flight altitude was set to 10 meters and the flight speed was set to 5 meters per second; the inspection vehicle's driving speed was set to 10 kilometers per hour.