Oilfield pipeline micro-leakage detection method and device based on ultra-wideband radar and diffusion model

Abstract

The invention discloses an oil field pipeline micro-leakage detection method and device based on an ultra-wideband radar and a diffusion model, and relates to the technical field of oil field safety production and intelligent monitoring. The invention provides a scheme integrating radar preprocessing, image generation and deep learning detection for the problems of weak micro-leakage characteristics, difficulty in direct visualization of radar signals, sample scarcity, scattered detection process and the like. According to the method, pipeline reflection signals are acquired by setting ultra-wideband radar parameters, and band-pass filtering, distance compression, clutter removal and image normalization are performed in sequence to obtain a gray radar image; inputting the image into an improved diffusion model to generate an image containing micro-leakage features; and inputting the generated or real image into a YOLOv8 detection model to identify a leakage position and confidence, and carrying out visualization and data output on a result. The method can enhance weak leakage characteristic expression and improve detection precision, and is suitable for the fields of on-line monitoring, risk early warning and the like of oil and gas transportation and other long-distance pipelines.

Description

Technical Field

[0001] This invention belongs to the field of oilfield safety production and intelligent monitoring technology, specifically involving ultra-wideband radar signal processing, diffusion model image generation, and micro-leakage target detection technology. Background Technology

[0002] Oilfield pipelines, as critical infrastructure for energy transportation, are prone to micro-leaks due to their high degree of concealment, delayed impact, and difficulty in detection. Micro-leaks typically involve leaks of less than 0.1 m³. 3 The leakage hole diameter is often less than 5mm, resulting in extremely weak physical signals. If not detected in time, it may lead to serious accidents such as soil pollution, fire, and explosion.

[0003] To reduce accident risks, the industry has gradually shifted from traditional manual inspections to intelligent monitoring systems relying on multi-source sensors in recent years. Among them, ultra-wideband radar has attracted attention because it can penetrate soil and cover layers and achieve non-contact detection. However, its output is usually a raw time-domain or frequency-domain signal, which lacks an intuitive way of expression, making it difficult for frontline personnel to judge the spatial characteristics of minor leaks based on radar signals.

[0004] Existing monitoring methods still face several limitations in engineering practice. On the one hand, micro-leakage signals are extremely weak and easily obscured by environmental noise, surface clutter, and equipment reflections during long-distance transmission. Traditional detection methods based on pressure and flow differences are insensitive to micro-leakage and struggle to make effective early assessments, resulting in delayed detection of potential hazards. On the other hand, oilfield sites are affected by geological conditions, electromagnetic environments, and construction equipment, causing radar signal waveforms to vary significantly with the environment. Conventional preprocessing methods lack specificity and often fail to extract stable visual information from complex backgrounds.

[0005] Meanwhile, deep learning models are gradually becoming an important direction for leak detection. However, in real-world oilfield scenarios, the incidence of micro-leakage events is extremely low, the number of available samples is limited, and the distribution of image features is uneven. Directly training detection models often leads to overfitting, resulting in a large number of missed detections. In addition, most existing systems consist of several scattered modules, with radar parameter adjustment, signal processing, and image analysis steps operating independently, failing to form a coherent data link. This makes the entire detection process time-consuming and difficult to meet the real-time requirements of the field.

[0006] Based on the aforementioned technological status, the industry urgently needs an integrated technical solution capable of rapidly and stably converting ultra-wideband radar signals into visual images, while maintaining micro-leakage identification accuracy even under conditions of scarce samples. Particularly in long-distance pipelines and complex terrain environments, a system with comprehensive capabilities including signal enhancement, image generation, automatic detection, and result output is required to improve the reliability and intelligence level of oilfield pipeline operation and maintenance. Summary of the Invention

[0007] To address the problems in existing oilfield pipeline micro-leak monitoring, such as the difficulty in visualizing radar signals, weak micro-leak features easily obscured by complex backgrounds, an extreme shortage of micro-leak samples for training, and fragmented detection processes with long response times, this invention proposes a micro-leak detection method combining ultra-wideband radar and a diffusion model. This method enhances radar signal representation in complex environments, stably converting the original signal into identifiable image features. Furthermore, it overcomes the identification difficulties caused by sample scarcity through a generative model, thereby achieving automatic detection and visualization of micro-leak locations, improving detection accuracy and real-time performance. The specific scheme is as follows: A method for detecting micro-leakage in oilfield pipelines based on ultra-wideband radar and a diffusion model includes the following steps: S1. Set the parameters of the ultra-wideband radar and collect the reflected signals of the oilfield pipeline based on the parameters; S2. The reflected signal is sequentially subjected to bandpass filtering, range compression, clutter removal and image normalization to obtain a grayscale radar image; S3. Input the grayscale radar image into the diffusion model, introduce a time embedding layer and residual connection into the backbone network of the diffusion model, and generate a radar image with micro-leakage features through forward diffusion noise addition and backward diffusion noise removal. S4. Input the radar image containing micro-leakage features into the YOLOv8 micro-leakage detection model, and output the location and confidence level of the micro-leakage in the oilfield pipeline; S5. Visualize and output the location and confidence level of the micro-leakage.

[0008] Furthermore, the bandpass filter employs a Butterworth fourth-order bandpass filter, whose cutoff frequency is automatically matched according to the parameters of the ultra-wideband radar.

[0009] Furthermore, the distance compression focuses the reflected signals from the oilfield pipeline by generating a Chirp reference signal and performing Hilbert transform and convolution processing.

[0010] Furthermore, the clutter removal is based on calculating the background mean value to eliminate clutter using the reflected signals from 100 frames of leak-free oilfield pipelines.

[0011] Furthermore, the image normalization employs logarithmic compression and 1%-99% quantile normalization to output a grayscale radar image of uniform size.

[0012] Furthermore, the forward diffusion noise addition simulates the signal degradation process by gradually adding noise to the radar image during the forward diffusion process.

[0013] Furthermore, the reverse diffusion denoising method uses an improved UNet model to predict noise, takes a 1-channel radar image as input, gradually recovers micro-leakage features, and outputs a micro-leakage image of the same size.

[0014] Based on the same inventive concept, this invention also proposes an oilfield pipeline micro-leakage detection device based on ultra-wideband radar and a diffusion model, comprising: The ultra-wideband radar parameter configuration and signal acquisition module is used to set the parameters of the ultra-wideband radar and acquire the reflected signals of oilfield pipelines based on the parameters. The signal preprocessing module is used to sequentially perform bandpass filtering, range compression, clutter removal and image normalization on the reflected signal to obtain a grayscale radar image; The diffusion model image generation module is used to input the grayscale radar image into the diffusion model and generate a radar image containing micro-leakage features through forward diffusion noise addition and backward diffusion noise removal. The micro-leakage detection module is used to input radar images containing micro-leakage features into the YOLOv8 micro-leakage detection model and output the location and confidence level of micro-leakage in the oilfield pipeline. The GUI module is used to visualize and output data on the location and confidence level of the micro-leakage.

[0015] Based on the same inventive concept, the present invention also proposes a computer storage medium for storing a computer program, wherein when the computer program is read by a computer, the computer executes the method described in the present invention.

[0016] Based on the same inventive concept, the present invention also proposes a computer, including a processor and a storage medium, wherein when the processor reads a computer program stored in the storage medium, the computer executes the method described in the present invention.

[0017] Based on the same inventive concept, the present invention also proposes a computer program product, which, when read, implements the method described in the present invention.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The oilfield pipeline micro-leakage detection method described in this invention introduces a parameter-adaptive radar preprocessing mechanism, which allows steps such as bandpass filtering, range compression, and clutter removal to be automatically adjusted according to changes in ultra-wideband radar parameters. This avoids the instability problems caused by manual adjustment and processing due to changes in bandwidth or center frequency in traditional methods, thereby maintaining high signal processing quality under different operating conditions and enabling weak leakage characteristics to be presented more clearly.

[0019] The oilfield pipeline micro-leakage detection method of the present invention uses a diffusion model that includes time embedding and residual connection to generate radar images containing micro-leakage features. Under the condition of limited real leak samples, it can still obtain diverse and structurally stable micro-leakage images, overcome the limitation of deep learning detection models that rely on a large number of measured samples, and help reduce the risk of overfitting and improve the reliability of micro-leakage identification.

[0020] The oilfield pipeline micro-leakage detection method of the present invention performs grayscale adaptation and anchor frame optimization on the YOLOv8 detection model according to the characteristics of ultra-wideband radar grayscale images, so that the model can better capture the detailed changes in the micro-leakage area. Compared with directly using the general RGB model, the detection accuracy is significantly improved, especially when identifying micro-leakage with small pixel scale, it has better robustness.

[0021] The oilfield pipeline micro-leakage detection method of the present invention integrates radar parameter configuration, signal preprocessing, diffusion model image generation, and micro-leakage detection into an end-to-end integrated processing flow. This enables changes in front-end parameters to automatically link with subsequent algorithm modules, avoiding the problems of scattered processing chains and manual connection between steps in existing solutions. This effectively shortens the detection time and improves the overall stability of the system.

[0022] This invention has the ability to stably extract and visualize ultra-wideband radar signals, generate micro-leakage feature images, and achieve automated and accurate identification in complex environments. It can maintain reliable micro-leakage detection results in oilfields with scarce samples, weak signals, and heavy clutter, and is suitable for real-time monitoring and risk warning of long-distance pipeline networks in oilfields such as oil pipelines and gas pipelines. Attached Figure Description

[0023] Figure 1 This is a flowchart of the oilfield pipeline micro-leakage detection method described in the implementation method. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Implementation Method 1 like Figure 1 As shown, a method for detecting micro-leakage in oilfield pipelines based on ultra-wideband radar and a diffusion model includes the following steps: S1. Set the parameters of the ultra-wideband radar and collect the reflected signals of the oilfield pipeline based on the parameters; S2. The reflected signal is sequentially subjected to bandpass filtering, range compression, clutter removal and image normalization to obtain a grayscale radar image; S3. Input the grayscale radar image into the diffusion model, introduce a time embedding layer and residual connection into the backbone network of the diffusion model, and generate a radar image with micro-leakage features through forward diffusion noise addition and backward diffusion noise removal. S4. Input the radar image containing micro-leakage features into the YOLOv8 micro-leakage detection model, and output the location and confidence level of the micro-leakage in the oilfield pipeline; S5. Visualize and output the location and confidence level of the micro-leakage.

[0026] By setting the parameters of the ultra-wideband radar and acquiring the reflected signal, and then sequentially performing bandpass filtering, range compression, removal of fixed clutter, and image conversion, weak leakage features can be stably extracted from the original high-noise radar signal and presented in image form, providing clear and structured input data for subsequent automatic identification.

[0027] The bandpass filter used is a Butterworth fourth-order bandpass filter, whose cutoff frequency is automatically matched according to the parameters of the ultra-wideband radar.

[0028] By employing a fourth-order Butterworth bandpass filter and automatically matching the cutoff frequency with radar parameters, the signal maintains stable filtering quality under different bandwidths and center frequencies, avoiding processing deviations that may be caused by manual adjustment. This improves the retention of effective reflective components, making micro-leakage features easier to highlight in subsequent imaging.

[0029] The distance compression focuses the reflected signals from oilfield pipelines by generating a Chirp reference signal and performing Hilbert transform and convolution processing.

[0030] By utilizing the Chirp reference signal for Hilbert transform and convolution processing, the echo is focused in the distance dimension, enabling subtle waveform changes caused by micro-leakage to be concentrated in the distance domain, enhancing the resolution of weak reflection energy, and improving the clarity of subsequent image features.

[0031] Furthermore, the clutter removal is based on calculating the background mean value to eliminate clutter using the reflected signals from 100 frames of leak-free oilfield pipelines.

[0032] By using a large number of leak-free background frames to calculate a stable background mean, the reflection of fixed structures can be effectively canceled, reducing constant clutter from the ground and equipment in complex environments, making local changes caused by micro-leakage more prominent, and providing a cleaner input image for the detection model.

[0033] Furthermore, the image normalization employs logarithmic compression and 1%-99% quantile normalization to output a grayscale radar image of uniform size.

[0034] By combining logarithmic compression with 1%-99% quantile normalization, the grayscale dynamic range between strong and weak reflections is compressed to a range suitable for model processing, making the image brightness distribution more balanced and avoiding the masking of weak leakage features due to extreme brightness differences, thereby improving the model's ability to identify subtle anomalies.

[0035] Furthermore, the forward diffusion noise addition simulates the signal degradation process by gradually adding noise to the radar image during the forward diffusion process.

[0036] By progressively adding noise to the radar image during the forward diffusion process, the model can learn the true distribution of the radar image degradation process, improve the structural recovery capability in the subsequent reverse diffusion stage, and make the generated micro-leakage image more consistent with the noise characteristics of real working conditions.

[0037] Furthermore, the reverse diffusion denoising method uses an improved UNet model to predict noise, takes a 1-channel radar image as input, gradually recovers micro-leakage features, and outputs a micro-leakage image of the same size.

[0038] By utilizing an improved UNet that incorporates residual structure and temporal embedding to predict noise and gradually recover the structure during the backdiffusion stage, the micro-leakage region is reconstructed in a stable form in the generated image, improving the accuracy and consistency of leakage feature generation and enabling the detection model to obtain more reliable training samples.

[0039] Implementation Method 2 This embodiment combines the technical solutions described in the foregoing embodiments, and, in conjunction with actual application scenarios and the usage process of computer program products that implement the method of this invention, further verifies and explains the technical effects of this invention through specific examples.

[0040] The processing described in this embodiment relies on the graphical interface provided by this invention. The system integrates radar parameter configuration, signal preprocessing, diffusion model inference, and micro-leakage detection into a visualized operation chain, enabling users to complete the entire process on a unified platform. The execution method and working principle of each step will be explained in detail below in conjunction with the interface logic.

[0041] S1: Parameter Configuration and Signal Acquisition After starting the system, the operator first accesses the "Radar Parameter Configuration" option in the interface. Key parameters of the ultra-wideband radar, including center frequency, bandwidth, sampling rate, and scan line length, are set according to the diameter range of the oilfield pipeline and the actual detection distance requirements. The size and laying depth of the pipeline directly affect the propagation conditions of the radar pulse; therefore, different frequency bands need to be adjusted according to the detection task when configuring parameters. In this implementation, when the pipeline diameter is set to 500mm, the system recommends a combination of a center frequency of approximately 5.8GHz and a bandwidth of approximately 3GHz to balance penetration capability and resolution requirements.

[0042] After setting the parameters, the user clicks "Save," and the system immediately performs adaptive calculations, estimating the range resolution based on the set bandwidth. The calculation of range resolution stems from the fundamental principle of radar imaging: resolution is inversely proportional to transmission bandwidth. Therefore, the larger the bandwidth, the more finely the system can distinguish adjacent reflective structures. In this embodiment, the system calculates a result of approximately 0.05 meters. The smaller this value, the better it is at revealing local reflection changes caused by micro-leakage.

[0043] After completing the resolution calculation, the system automatically updates the cutoff frequency of the bandpass filter in the preprocessing module to ensure that subsequent signal processing is consistent with the radar configuration, guaranteeing that the processing link does not require manual adjustment under different operating conditions. The operator then activates the ultra-wideband radar for on-site data acquisition. The radar system continuously scans the surrounding medium structure of the oilfield pipeline according to the set number of scan lines, recording the reflected waveforms and storing them in a matrix format in a file to form the raw radar data file for subsequent processing. Typically, the raw radar data file contains hundreds of scan lines, each containing thousands of sampling points, thus presenting the continuous reflection changes of the pipeline along the detection direction.

[0044] S2: Adaptive preprocessing of radar signals In the "Signal Preprocessing" module, the user loads the raw radar data file acquired by S1, and the system automatically executes a series of adaptive processing procedures according to the configuration. First, the preprocessing module performs bandpass filtering using the cutoff frequency automatically updated by S1 to retain the effective components in the radar's operating frequency band and suppress noise interference from peripheral frequencies. Since ultra-wideband radar covers a wide frequency range, using filters with fixed parameters may lead to unstable filtering effects under different mission scenarios. Therefore, this invention achieves adaptive adjustment of the filtering process through parameter linkage, ensuring that the filter always matches the radar's transmission characteristics and improving signal-to-noise ratio stability.

[0045] After filtering, the system proceeds to the range compression step. Range compression generates a chirp reference signal consistent with the transmitted signal and performs Hilbert transform and convolution operations on each received scan line, concentrating the echo energy into a narrower time window, thereby enhancing the target's range resolution. For localized and weak reflective structures like microleakage, range compression can significantly improve its discernibility in the reflected signal, making it more prominent in subsequent images.

[0046] The system then performs clutter removal processing. This implementation first calculates the background mean from 100 frames of leak-free pipeline reflection signals acquired on-site. The background represents the stable reflection pattern of the pipeline structure and surrounding medium under leak-free conditions. By subtracting the background mean from the current image, the system can effectively eliminate fixed reflection components, making the subtle incremental structures caused by leaks more apparent in the subtracted data.

[0047] After clutter removal, the system enters the image normalization stage. The grayscale distribution of the original radar reflection signal typically exhibits a large dynamic range between strong and weak reflections. If directly converted into an image, subtle leakage features would be masked by the strong reflection region. Therefore, this invention employs logarithmic compression and 1%-99% quantile normalization to compress the grayscale dynamic range and map it to a range suitable for visual and model input, outputting a uniformly sized grayscale radar image.

[0048] S3: Micro-leakage image generation In real oilfield environments, micro-leakage is a low-probability event, and the number of real leak radar images that can be collected on-site is extremely limited. Deep learning detection models typically require a large number of leak samples covering different operating conditions during the training phase. To address the problem of sample scarcity, this invention introduces an improved diffusion model in the image generation stage, constructing high-quality micro-leakage training images by simulating the radar image generation process.

[0049] After loading the grayscale radar image output by S2 in the "Image Generation" module of the interface, the user can initiate the reverse diffusion process of the diffusion model. The core of the diffusion model is a generative structure for progressive denoising. Its basic idea is to gradually add noise to the original image through forward diffusion steps until it approaches a state of random noise, and then gradually restore the image structure through reverse diffusion steps. In this embodiment, the forward diffusion process is set to 1000 steps, and the noise intensity is... By gradually increasing the value linearly from extremely low values, the model can learn the continuous evolution of radar image degradation, thereby obtaining a more stable recovery path that better conforms to physical characteristics during the reverse generation stage.

[0050] The improved diffusion model's main network employs an optimized UNet structure. Compared to the original UNet, this implementation introduces a temporal embedding layer, enabling the model to more accurately perceive the current diffusion stage when predicting noise at each step, thus improving the temporal consistency of the denoising process. Simultaneously, residual connections are added to the network's feature extraction and reconstruction paths, allowing the model to better preserve multi-scale structures when processing complex radar textures, improving the stability and detail representation of the generated images. This network supports single-channel radar image input and output, ensuring the generated image size matches the input image, facilitating direct use for training subsequent detection models.

[0051] During the backdiffusion process, the model generates radar images with different leakage morphologies through progressive denoising and structure recovery. The generated micro-leakage features can exhibit random variations in location, size, and shape, resulting in high diversity of generated results. The system supports generating 1 to 100 images at a time, with output sizes ranging from 256×256 to 640×640, adaptable to different model training needs. The generated micro-leakage images achieve a peak signal-to-noise ratio (PSNR) of over 35dB and a structural similarity of over 90% with real leak images, effectively simulating local reflection changes caused by leaks while maintaining the radar imaging style.

[0052] By adopting the improved diffusion model described above, this implementation method can continuously supplement high-quality microleakage samples, significantly enhance the generalization ability of the detection model in different scenarios, and provide reliable sample support for the identification of microleakage in complex oilfield environments.

[0053] S4: Precise Detection of Micro-Leaks In the "Leak Detection" module, users can load real-time radar images or the aforementioned generated micro-leak sample images and launch the YOLOv8 detection model, which has been optimized with grayscale image adaptation and anchor boxes. This model can effectively identify subtle abnormal areas on a single-channel grayscale image and output the location and confidence level of the leak when the confidence level exceeds a set threshold. The system automatically annotates the detection results in the image and provides audio and interface pop-up prompts when a suspected leak is detected, so that on-site personnel can respond to potential risks in a timely manner.

[0054] S5: Results Output and Report Generation After the inspection task is completed, the system automatically calculates the number of leaks, confidence level, and high-frequency leakage areas within the current inspection cycle, generating statistical data that can be used for on-site maintenance and risk analysis. Users can export annotated image files, inspection data files, and statistical reports that can be used for traceability analysis, facilitating the construction of pipeline operation history.

[0055] The above detailed description of the technical solution provided by the present invention is intended to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above detailed embodiments are not intended to limit the scope of protection of the present invention. Any reasonable modifications and improvements to the present invention, recombination of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0056] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims disclosed in the present invention can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the disclosure of the present invention. Although the present invention 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 substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle scope of the present invention should be considered to fall within the protection scope of the present invention.

[0057] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for detecting micro-leakage in oilfield pipelines based on ultra-wideband radar and a diffusion model, characterized in that, include: S1. Set the parameters of the ultra-wideband radar and collect the reflected signals of the oilfield pipeline based on the parameters; S2. The reflected signal is sequentially subjected to bandpass filtering, range compression, clutter removal and image normalization to obtain a grayscale radar image; S3. Input the grayscale radar image into the diffusion model, introduce a time embedding layer and residual connection into the backbone network of the diffusion model, and generate a radar image with micro-leakage features through forward diffusion noise addition and backward diffusion noise removal. S4. Input the radar image containing micro-leakage features into the YOLOv8 micro-leakage detection model, and output the location and confidence level of the micro-leakage in the oilfield pipeline; S5. Visualize and output the location and confidence level of the micro-leakage.

2. The method according to claim 1, characterized in that, The bandpass filter used is a Butterworth fourth-order bandpass filter, whose cutoff frequency is automatically matched according to the parameters of the ultra-wideband radar.

3. The method according to claim 1, characterized in that, The distance compression focuses the reflected signals from oilfield pipelines by generating a Chirp reference signal and performing Hilbert transform and convolution processing.

4. The method according to claim 1, characterized in that, The clutter removal is based on the background mean calculation of 100 frames of reflected signals from leak-free oilfield pipelines to eliminate clutter.

5. The method according to claim 1, characterized in that, The image normalization employs logarithmic compression and 1%-99% quantile normalization to output a uniform-sized grayscale radar image.

6. The method according to claim 1, characterized in that, The forward diffusion noise addition is achieved by gradually adding noise to the radar image during the forward diffusion process to simulate the signal degradation process.

7. The method according to claim 1, characterized in that, The reverse diffusion noise reduction method uses an improved UNet model to predict noise, takes a 1-channel radar image as input, gradually recovers micro-leakage features, and outputs a micro-leakage image of the same size.

8. A micro-leakage detection device for oilfield pipelines based on ultra-wideband radar and a diffusion model, characterized in that, include: The ultra-wideband radar parameter configuration and signal acquisition module is used to set the parameters of the ultra-wideband radar and acquire the reflected signals of oilfield pipelines based on the parameters. The signal preprocessing module is used to sequentially perform bandpass filtering, range compression, clutter removal and image normalization on the reflected signal to obtain a grayscale radar image; The diffusion model image generation module is used to input the grayscale radar image into the diffusion model and generate a radar image containing micro-leakage features through forward diffusion noise addition and backward diffusion noise removal. The micro-leakage detection module is used to input radar images containing micro-leakage features into the YOLOv8 micro-leakage detection model and output the location and confidence level of micro-leakage in the oilfield pipeline. The GUI module is used to visualize and output data on the location and confidence level of the micro-leakage.

9. A computer storage medium for storing computer programs, characterized in that, When the computer program is read by the computer, the computer executes the method according to any one of claims 1-7.

10. A computer program product, as a computer program, is characterized by: When the computer program is read, the method of any one of claims 1-7 is implemented.

Images