A method and system for monitoring night-time leakage of oil pipelines by using a UAV infrared imaging

Abstract

The application discloses a kind of unmanned aerial vehicle infrared imaging oil pipeline night leakage monitoring method and system, belong to oil pipeline safety monitoring technical field.For solving the problem of low efficiency, false alarm rate is high in night leakage monitoring, the present application will monitoring area be divided into independent monitoring unit and set up temperature threshold;Obtain the temperature of each unit by thermal imaging, mark the unit over threshold as suspected area to complete primary screening;Suspected area is continuously tracked, and dynamic leakage is determined according to area change rate;For the area that does not pass dynamic determination, static leakage is determined according to the temperature difference of multiple points;For the area that does not pass the first two determinations, hidden leakage is determined by reflight temperature rise;According to priority, determine early warning level;After triggering early warning, visible light recheck is confirmed, and false alarm is updated threshold.The present application realizes high-precision, low false alarm monitoring of oil pipeline night leakage, reduces the cost of inspection and emergency disposal risk.

Description

Technical Field

[0001] This invention relates to the field of oil pipeline safety monitoring technology, and in particular to a method and system for nighttime leak monitoring of oil pipelines using UAV infrared imaging. Background Technology

[0002] Oil pipelines are core infrastructure for energy transportation, and leaks can easily lead to safety accidents and environmental pollution. Currently, oil pipeline leak monitoring mainly relies on the following methods: first, manual inspection, where maintenance personnel walk or drive along the pipeline to detect leaks using portable detectors or visual observation; second, fixed sensor monitoring, which deploys sensors along the pipeline to measure temperature, pressure, and flow, detecting leaks through data anomalies; and third, satellite or aerial remote sensing monitoring, which uses satellites or manned aircraft equipped with optical or thermal infrared sensors for large-scale inspections. However, all of these existing technologies have significant shortcomings in nighttime scenarios: manual inspections have limited visibility at night, making it difficult to achieve full coverage, resulting in a high rate of missed detections, and require lighting and protective equipment, significantly increasing inspection costs and posing personnel safety risks; fixed sensor monitoring has limited coverage, high sensor deployment and maintenance costs, and is unsuitable for long-distance, terrain-crossing pipelines; and satellite and aerial remote sensing monitoring lacks sufficient spatiotemporal resolution, making it difficult to achieve real-time detection and continuous tracking of small-scale, slow leaks.

[0003] In recent years, drones equipped with infrared imaging technology have provided a new solution for nighttime leak monitoring. After an oil spill, the temperature is typically 2-5°C higher than the surrounding environment, and this temperature difference is even more pronounced in the low-temperature environment at night, providing a basis for locating the leak point. However, existing research still has the following shortcomings: First, monitoring often uses fixed thresholds for judgment, failing to consider the impact of environmental heterogeneity in different areas (such as equipment heat radiation, crop shading, water surface reflection, etc.) on leak signals, easily leading to false alarms or missed alarms; second, the early warning logic design is imperfect, lacking targeted identification methods for concealed leaks that reach temperature warning levels but whose area has not changed significantly; and third, a closed-loop monitoring system encompassing data acquisition, threshold judgment, leak verification, and parameter optimization has not been established.

[0004] Therefore, there is an urgent need for a method and system for monitoring nighttime leaks in oil pipelines that can adapt to complex terrain and diverse scenarios, and has multi-level judgment logic and adaptive optimization capabilities. Summary of the Invention

[0005] To address the issues of high false alarm rates for fixed thresholds and lack of targeted identification of subtle area changes in nighttime oil pipeline leak monitoring due to environmental heterogeneity, this invention provides a method and system for nighttime oil pipeline leak monitoring using UAV infrared imaging. By setting differentiated temperature and area judgment parameters for different regions, adding a re-flight temperature rise judgment step for suspected areas with unqualified areas, and performing visible light verification on each warning and feeding back false alarm data to the threshold optimization process, high-precision, low-false-alarm nighttime leak monitoring is achieved.

[0006] The technical solution of this invention is: A method for nighttime leak detection of oil pipelines using UAV infrared imaging includes the following steps: The monitoring area is divided into independent monitoring units according to size, and leakage temperature thresholds and area change rate standards are established; the leakage temperature thresholds and area change rate standards can be dynamically fine-tuned based on the environmental characteristics of different areas. The temperature of each unit is obtained by thermal imaging, and units whose temperature exceeds the leakage temperature threshold are marked as suspected leakage areas. The suspected leakage area is continuously monitored. When the rate of change of its area and the characteristics of its shape change meet the dynamic leakage conditions, it is determined to be a dynamic leakage and a first-level warning is triggered. For suspected areas that fail the dynamic leakage assessment, when the temperature difference change of the internal monitoring unit at different times meets the static leakage conditions, it is determined to be a static leakage and a level-two warning is triggered. For suspected leak areas that fail both dynamic and static leak detection, if the temperature anomaly persists and the area change is limited, the UAV will re-monitor the temperature rise value of the core area by re-flying. When the temperature rise value meets the conditions for concealed leak, it will be determined as a concealed leak and a level-two warning will be triggered. When the same area meets multiple leakage criteria at the same time, the final warning level shall be determined in the order of priority for dynamic leakage, followed by concealed leakage, and lastly for static leakage. After an alert is triggered, the suspected area is reviewed using visible light imaging. If a leak is confirmed, the location is merged; otherwise, the temperature alert conditions are updated.

[0007] Furthermore, the temperature threshold is determined by prior field tests and historical data statistics to establish the nighttime baseline temperature range for different environmental types, with a threshold of 2°C above the baseline temperature; the area change rate standard is determined by conducting simulated leakage tests on different areas, and the area change rate standard is 1.0 m² / h.

[0008] Furthermore, the continuous tracking and monitoring interval is 1 hour; the dynamic leakage condition is: analyzing its shape and area change characteristics, if the suspected area shows irregular expansion in shape after three consecutive monitoring sessions, and the area change rate is ≥1.0m...2 If / h, it is determined to be a dynamic leak.

[0009] Furthermore, the static leakage condition is as follows: extract the temperature values ​​of at least three continuous monitoring units in the suspected leakage area at different times, and calculate the temperature difference change. If the temperature difference is ≥3℃, the interference of ambient temperature fluctuation is excluded, and it is determined to be a static leakage; the time interval between the different times is 30 minutes.

[0010] Furthermore, the concealed leakage condition is as follows: the temperature of the suspected leakage area exceeds the benchmark threshold by 2°C and the duration is ≥1h, and the area change rate does not reach the standard after the corresponding area threshold is fine-tuned. An emergency re-flight monitoring is initiated for the suspected leakage area for 1h. If the temperature of the core cell of the suspected leakage area increases by ≥3°C after the re-flight monitoring, it is determined to be a concealed leakage.

[0011] Furthermore, during the visible light imaging verification, the drone's flight altitude is lowered from the normal inspection altitude to a preset low altitude, and visible light images of the suspected leak area are collected using a high-definition night vision camera.

[0012] Furthermore, the specific method of dividing the monitoring area into independent monitoring units according to size is as follows: based on the oil pipeline GIS map, the area is divided into independent monitoring units with a size of 1m×1m.

[0013] Furthermore, the dynamic fine-tuning includes conducting statistical analysis of monitoring samples, on-site simulated leakage tests, and data fitting optimization for different regions to determine the fine-tuned thresholds and standards for each region.

[0014] Another aspect of the present invention relates to an unmanned aerial vehicle (UAV) infrared imaging system for nighttime leak monitoring of oil pipelines, comprising front-end acquisition equipment and a ground control center, wherein the ground control center includes: The image display module is used to display the infrared thermal image acquired by the infrared imager in real time; A temperature data extraction module is used to extract temperature data of each monitoring unit from the infrared thermal image; The leak area marking module is used to mark units whose temperature exceeds the threshold on the infrared image; The threshold determination module is used to automatically determine the temperature threshold and the area change rate standard, and trigger corresponding warnings based on the determination logic. The anomaly marking and go-around command module is used to mark suspected leakage areas that trigger warnings and automatically send go-around commands to drones; The early warning information push module is used to push leak location, temperature data, infrared images, and area diffusion trends to operation and maintenance personnel; The data storage and backtracking module is used to support real-time analysis of monitoring data and backtracking of historical data.

[0015] Furthermore, the front-end acquisition device includes a drone and an infrared imager and a data transmission module mounted on the drone; The infrared imager is used to collect infrared thermal imaging data of the monitoring area in real time, and the data transmission module is used to transmit the infrared thermal imaging data back to the ground control center in real time.

[0016] The beneficial effects of this invention are: 1. This invention effectively solves the problems of false alarms and missed alarms in traditional methods and single-threshold drone monitoring by using multi-dimensional early warning logic and regional threshold fine-tuning. Compared with traditional manual inspection and single temperature / area threshold drone monitoring methods, the accuracy of identification is greatly improved, the false alarm rate is reduced to less than 5%, and the emergency response cost is effectively reduced. 2. This invention uses autonomous drone re-flight and intelligent early warning judgment to confirm leaks. The average time taken is only 8.3 minutes, which is more than half that of traditional manual inspection or drone monitoring methods based on single temperature or area thresholds. This saves valuable time for emergency response to crude oil leaks and can effectively reduce the losses caused by them. 3. While maintaining high-precision positioning of ≤3m, this invention significantly reduces the unit mileage inspection cost compared to manual inspection or single temperature / area threshold UAV monitoring methods by optimizing inspection routes and intelligent data processing, thus achieving the dual goals of high precision and low cost. 4. This invention formulates differentiated strategies for environmental heterogeneity in different regions, effectively solving monitoring problems in complex scenarios such as farmland shading, wetland tides, and equipment interference. The accuracy rate in each region remains above 93%, which is far higher than traditional monitoring methods and better meets the needs of actual engineering applications. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is an example image of a temperature anomaly monitoring point; Figure 3 This is an example image of an area anomaly point monitoring. Detailed Implementation

[0018] The following will be based on embodiments of the present invention. Figures 1-3 The technical solutions in the embodiments of the present invention will be clearly and completely described herein. The described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments in this application without creative effort are within the scope of protection of this application.

[0019] This invention uses temperature changes and leakage area changes captured by infrared imaging technology as the core judgment criteria. It designs a two-level early warning logic for dynamic and static leakage judgment, while a concealed leakage judgment logic forms a triple guarantee. Combined with a dynamic threshold fine-tuning mechanism and a leakage verification and correction process, it constructs a complete method for monitoring nighttime leaks in oil pipelines using unmanned aerial vehicles (UAVs). All indicators involving leakage temperature or area are uniformly expressed in ℃ / h or m². 2 / h is used as the unit of measurement to ensure data consistency and comparability.

[0020] Example 1

[0021] A method for nighttime leak detection of oil pipelines using UAV infrared imaging includes the following steps: The monitoring area is divided into independent monitoring units according to size, and leakage temperature thresholds and area change rate standards are established; the leakage temperature thresholds and area change rate standards can be dynamically fine-tuned based on the environmental characteristics of different areas. Furthermore, the temperature threshold is determined by prior field tests and historical data statistics to establish the nighttime baseline temperature range for different environmental types, with a threshold of 2°C above the baseline temperature; the area change rate standard is determined by conducting simulated leakage tests on different areas, and the standard is 1.0 m² / h.

[0022] Furthermore, the monitoring area is divided into independent monitoring units according to size, specifically: based on the GIS map of the oil pipeline, it is divided into independent monitoring units with a size of 1m×1m.

[0023] Furthermore, dynamic fine-tuning includes statistical analysis of monitoring samples in different regions, on-site simulated leakage tests, and data fitting optimization to determine the fine-tuned thresholds and standards for each region.

[0024] Specifically, firstly, based on the GIS map of the oil pipeline, the entire monitoring area was divided into several independent cells of 1m×1m, each cell serving as a basic temperature monitoring unit. This division was automatically completed by the ground control center software and visualized on the electronic map. Then, through preliminary field tests and historical data statistics, the nighttime baseline temperature range for different environmental types (such as exposed ground in oilfield operating areas, crop-covered surfaces in farmland areas, and water bodies and vegetation in riverside wetlands) was determined. More specifically, under normal operating conditions without leaks, for several consecutive nights (e.g., from 23:00 to 04:00 the next day), an infrared imager mounted on a drone was used to hover and photograph various types of areas for 10 minutes each time, collecting a large amount of temperature data. Statistical analysis yielded the environmental baseline temperature values ​​for each area (e.g., 18~22℃ in farmland areas, 20~24℃ in oilfield operating areas, and 16~20℃ in riverside wetlands). Based on this, a leakage temperature threshold was set to exceed the environmental baseline temperature of the area by 2℃. Simultaneously, simulated leakage tests were conducted in different areas to determine the area change rate standard. The method for simulating leak tests is as follows: Near typical pipe sections in each region, simulated crude oil leaks are conducted according to preset leak rates (e.g., small flow, medium flow). Unmanned aerial vehicles (UAVs) are used for continuous tracking and monitoring, recording the area change of the leak zone over time and calculating the area diffusion rate. After repeated tests, the standard for the area change rate of dynamic leak judgment is uniformly set at 1.0 m² / h. To adapt to the environmental heterogeneity of different regions, the above-mentioned leak temperature threshold and area change rate standard can be dynamically fine-tuned. The fine-tuning process includes three steps: First, statistical analysis of historical monitoring data is performed to statistically analyze the threshold distribution range of real leak signals and interference signals in each region; second, multiple different combinations of temperature and area thresholds are set in typical test sections of each region, and simulated leak tests are conducted to verify the effectiveness of different combinations with the goal of maximizing identification accuracy; third, binary regression fitting analysis is performed on the statistical data and test results, with identification accuracy as the dependent variable and temperature and area thresholds as independent variables, and the value corresponding to the peak accuracy is selected as the fine-tuned standard for that region.

[0025] It should be noted that the core of the first step in the fine-tuning process, threshold analysis, is to obtain one or more statistically significant quantiles based on the statistical distribution characteristics of the data (such as a Gaussian distribution) as the basis for judgment. In this embodiment, the system first collects infrared data of each region under leak-free conditions and analyzes the statistical distribution formed by these data. The actual leakage signal often appears as an outlier deviating from the main body in the statistical distribution, while environmental noise is concentrated in the main body. By analyzing the differences between the two in their distribution range, key data references can be provided for subsequent threshold fine-tuning, thereby more effectively separating the actual leakage signal from background interference.

[0026] Additionally, it should be noted that genuine leak signals in historical monitoring data originate from events confirmed as genuine leaks through on-site manual verification or visible light verification. These events are marked and stored in the leak sample library by maintenance personnel. Interference signals originate from events where the trigger temperature exceeds the threshold but is confirmed as non-leakage through visible light verification (such as equipment thermal radiation, water surface reflection, animal activity, etc.). These events are automatically marked and stored in the interference sample library by the system. A threshold fine-tuning process is automatically triggered every 30 samples accumulated in both types of sample libraries.

[0027] It should also be noted that binary regression is a classic binary classification statistical model used to predict the probability of an output result of "leakage" or "non-leakage". Its basic principle is to map a linear combination of two independent variables, temperature threshold and area threshold, to a probability range of 0 to 1 through a logistic function; the closer the output value is to 1, the higher the probability of leakage. In this embodiment, the system aims to maximize identification accuracy. It uses the field-verified leakage or non-leakage labeling results as the dependent variable and the temperature threshold and area threshold as independent variables. It fits the regression coefficients using maximum likelihood estimation to construct a classification model. An accuracy of 95% or higher is used as the optimal criterion. The system searches the parameter space of temperature and area thresholds for the parameter combination that results in the peak classification accuracy of the model, and uses this combination as the fine-tuned threshold for that region.

[0028] The temperature of each unit is obtained by thermal imaging, and units whose temperature exceeds the leakage temperature threshold are marked as suspected leakage areas. Specifically, a drone equipped with an infrared imager cruises along a pre-defined route on the oil pipeline, collecting real-time infrared thermal imaging data of the monitored area. Based on a pre-divided 1m×1m monitoring grid, the temperature value of each cell is extracted from the infrared thermal imaging data. The extraction method is as follows: according to the drone's current flight altitude and the infrared imager's field of view parameters, the pixels in the thermal image are mapped to actual geographic coordinates. The average temperature of all pixels within each cell is calculated by aggregating the data by grid, and this average temperature is taken as the measured temperature value of that cell. The measured temperature value of each cell is compared with a pre-set leakage temperature threshold (ambient reference temperature + 2℃) for that area. When the temperature value of any cell exceeds this threshold, that cell is marked as a suspected leakage area. Simultaneously, the geographic coordinates, measured temperature value, and corresponding infrared thermal image of the suspected leakage area are recorded.

[0029] It should be noted that the mapping relationship adopts the pinhole imaging model, and the specific steps are as follows: First, the optical center of the UAV's infrared imager is taken as the origin of the coordinate system, and the image plane of the imager is located at the focal length of the lens. According to the principles of geometric optics, the positional relationship between the ground target point and the corresponding image point on the image plane is determined by the focal length of the imager, the coordinates of the principal image point, and the distance from the target point to the center of the lens (which is the UAV's flight altitude in orthogonal inspection). Second, when calculating the local ground coordinates from the image point coordinates, given the UAV's flight altitude and lens focal length, and the fact that the optical axis is perpendicular to the ground, the image plane coordinates can be converted into local horizontal coordinates with the ground projection point of the optical axis as the origin. The X and Y axes of this coordinate system correspond to the east and north directions on the ground, respectively. Finally, when converting the local horizontal coordinates to geodetic coordinates, by combining the GNSS coordinates (east and north coordinates) acquired by the UAV in real time and the UAV's heading angle (yaw angle), the local coordinate points can be converted into actual geographic coordinates through coordinate rotation and translation. Through the above mapping process, the actual geographic coordinates of any pixel in the infrared image can be determined, and then the 1m×1m monitoring grid to which the pixel belongs can be determined.

[0030] The above steps utilize the physical characteristic that the temperature of crude oil after a leak is higher than that of the surrounding environment. Infrared imaging is used to convert the temperature difference into quantifiable data, automatically completing temperature comparison, suspected leak marking, and information recording. However, relying solely on temperature exceeding a threshold is susceptible to interference; therefore, subsequent dynamic leak determination steps are necessary for verification.

[0031] The suspected leakage area is continuously monitored. When the rate of change of its area and the characteristics of its shape change meet the dynamic leakage conditions, it is determined to be a dynamic leakage and a first-level warning is triggered. Furthermore, the continuous monitoring interval is 1 hour; even further, the dynamic leakage condition is: analyzing its shape and area change characteristics, if the suspected area shows irregular expansion in shape after three consecutive monitoring sessions, and the area change rate is ≥1.0m... 2 If / h, it is determined to be a dynamic leak.

[0032] Specifically, for suspected leak areas, continuous monitoring is conducted at 1-hour intervals. During each monitoring session, the outline and area of ​​the suspected leak area are first extracted from the infrared thermal imaging data. Area extraction employs OTSU threshold segmentation or a temperature gradient-based edge detection algorithm to segment the leak area outline from the infrared image. Then, the number of pixels within the outline is counted, and combined with the UAV's flight altitude and the infrared imager's field of view parameters, the pixel area is converted into the actual physical area, in square meters. Three consecutive monitoring sessions are conducted, and the area value for each monitoring session is recorded. The shape and area change characteristics of the leak area during the three monitoring sessions are analyzed: the ratio of the perimeter to the area of ​​the outline obtained in each monitoring session is calculated. If this ratio continuously increases across the three monitoring sessions, it is determined that the shape is undergoing irregular expansion. The area difference between two adjacent monitoring sessions is divided by 1 hour to obtain the area change rate. If both rates are positive and both reach or exceed 1.0 m² / h, the area change rate condition is met. When both conditions are met simultaneously—irregular expansion of shape and area change rate ≥ 1.0 m² / h—it is determined to be a dynamic leak, triggering a Level 1 warning, and the leak location, temperature data, infrared image, and area diffusion trend are recorded for notification to maintenance personnel.

[0033] It should be noted that when using OTSU segmentation, the gray levels of the suspected region's local infrared image are first quantized into 256 levels (gray level range 0~255), and the number of pixels at each gray level is counted to create a gray level histogram. The 256 gray levels (0~255) are traversed, and each gray level is used as a candidate segmentation threshold to divide the image pixels into background and foreground classes. The inter-class variance (the product of the background pixel ratio and the foreground pixel ratio, multiplied by the square of the difference between the two gray level means) is calculated. The gray level that maximizes the inter-class variance is selected as the final segmentation threshold, converting the original infrared image into a binary image (background as 0, foreground as 1). Then, a 3×3 morphological closing operation is used to remove small holes and noise points: a 3×3 cross-shaped structuring element (center pixel and its four neighbors above, below, left, and right) is selected. First, a dilation operation is performed on the binary image to fill small holes and edge cracks within the foreground region, followed by an erosion operation to restore the target region boundary. Finally, the largest connected component in the binary image is extracted as the leakage region contour.

[0034] This step utilizes the physical laws that crude oil leaks inevitably exhibit characteristics of continuous area expansion and irregular edges due to gravity infiltration and surface runoff. It can effectively distinguish between real leaks and non-leakage interferences (such as equipment thermal radiation, water surface reflection, etc.) and reduce the false alarm rate.

[0035] For suspected areas that fail the dynamic leakage assessment, when the temperature difference change of the internal monitoring unit at different times meets the static leakage conditions, it is determined to be a static leakage and a level-two warning is triggered. Furthermore, the static leakage condition is as follows: extract the temperature values ​​of at least three continuous monitoring units in the suspected leakage area at different times, and calculate the temperature difference change. If the temperature difference is ≥3℃, the interference of ambient temperature fluctuation is excluded, and it is determined to be a static leakage; the interval between the different times is 30 minutes.

[0036] Specifically, for suspected leak areas that fail the dynamic leak assessment, a static leak assessment is initiated. Within the suspected leak area, at least three consecutive monitoring units are selected. The temperature values ​​of these units are measured at two different times, with a 30-minute interval between the two times. The temperature difference between the two times for each monitoring unit is calculated, and the maximum temperature difference among all selected units is taken. If this maximum temperature difference reaches or exceeds 3°C, it is determined to be a static leak, triggering a level-two warning. This temperature difference condition effectively eliminates interference from global environmental temperature fluctuations such as the slow decrease in atmospheric temperature at night, because the temperature changes caused by such interference are usually gradual and uniform. However, the leak point, due to continuous crude oil seepage or heat absorption, will generate a more significant temperature difference in a shorter period. After determining it to be a static leak, a curve showing the coordinates of the leak point and the temperature change is generated to assist in formulating a response plan.

[0037] This step is applicable to scenarios where the leakage volume is small, the leakage point is blocked, or the crude oil is absorbed by the soil, resulting in an insignificant area expansion. It captures static leaks through multi-point temperature difference detection, compensating for the shortcomings of area determination in specific scenarios.

[0038] For suspected leak areas that fail both dynamic and static leak detection, if the temperature anomaly persists and the area change is limited, the UAV will re-monitor the temperature rise value of the core area by re-flying. When the temperature rise value meets the conditions for concealed leak, it will be determined as a concealed leak and a level-two warning will be triggered. Furthermore, the conditions for a concealed leak are as follows: the temperature of the suspected leak area exceeds the benchmark threshold by 2°C and lasts for ≥1 hour, and the rate of area change does not reach the standard after the corresponding area threshold is fine-tuned. An emergency re-flight monitoring is initiated for the area for 1 hour. If the temperature of the core cell of the suspected area increases by ≥3°C after the re-flight monitoring, it is determined to be a concealed leak.

[0039] Specifically, for suspected leak areas that fail both the dynamic leak assessment (area change rate < 1.0 m² / h or no irregular expansion) and the static leak assessment (multi-point temperature difference < 3℃), the following steps are used to determine concealed leaks: Timing is continuously monitored from the moment the suspected leak area is first marked. If the temperature of any monitoring cell within the area exceeds the ambient reference temperature by 2℃ for a cumulative period of 1 hour or more, the condition for sustained temperature anomaly is met. Within this 1-hour period, the area change of the suspected leak area is continuously tracked. If the area change rate remains below the dynamic leak assessment standard (i.e., less than 1.0 m² / h), it indicates that area expansion is limited by the environment (e.g., equipment obstruction, crop cover, soil adsorption, etc.), thus meeting the condition for limited area change. When both conditions are met simultaneously, an emergency re-flight command is sent to the UAV. The UAV takes off exactly 1 hour after the initial marking and conducts re-flight monitoring of the suspected leak area. The 1st hour is a relative time, with the moment the suspected leak area is first marked (the system automatically records this timestamp) as the starting point, and 60 minutes constitutes the re-flight time. During the re-flight, the UAV uses its onboard positioning system to obtain the position coordinates from the initial monitoring and navigates to the same coordinate point. The flight altitude is automatically adjusted to the same altitude (typically 100 meters) based on the barometric altimeter data recorded during the initial monitoring to ensure the comparability of temperature data. When regular patrol flights and emergency re-flight missions conflict, the system automatically interrupts the regular patrol flight, prioritizing the emergency re-flight. Regular patrol flights resume after the re-flight is completed. From the infrared image of the initial monitoring, the cell with the highest temperature within the suspected leak area is identified and defined as the core cell. The temperature value of this core cell at the time of the initial monitoring is recorded (denoted as T0). During the re-flight monitoring, the temperature value of the same core cell is extracted from the new infrared image (denoted as T1). The temperature rise ΔT = T1 - T0 is calculated. If the temperature rise ΔT reaches or exceeds 3℃, it is determined to be a concealed leak, triggering a level-two warning. If ΔT < 3℃, no warning is triggered, the anomaly is marked as an interference source, and relevant information is recorded for subsequent analysis. After determining a concealed leak, a refined inspection route is automatically planned. The route centers on the core cell, covering a radius of 10 meters around it, with the flight altitude decreasing to 50 meters and the flight speed decreasing to 3 m / s, in order to obtain higher resolution thermal and visible light images to further verify the leak.

[0040] It should be noted that, to avoid core cell drift caused by single-point temperature fluctuations, the core cell is determined as follows: In the initial three consecutive frames of infrared images, the sampling interval between the three frames is 1 second (i.e., one frame is taken every 30 frames under the condition of an infrared imager frame rate of 30fps). The top three cells with the highest temperatures in the suspected leak area are extracted from each frame (sorted from highest to lowest temperature; if multiple cells have the same temperature for the third position, all are included). The geographical intersection of the cells that appear at least twice in the three frames is taken. If the intersection is not empty, the center cell of the intersection is taken as the fixed core cell; if the intersection is empty, the cell with the highest frequency of occurrence of the highest temperature cell in the three frames is taken, and if the frequencies are the same, the cell with the highest average temperature is taken. During re-flight monitoring, the highest temperature point is not recalculated; instead, the geographical coordinates of the fixed core cell are directly located, and its temperature value is extracted.

[0041] This step is specifically designed to identify leaks that are slow to expand or are concealed due to environmental limitations. For example, in oilfield operations, leaks may occur within equipment gaps, preventing outward expansion; in farmland, crude oil may be absorbed by crop roots and soil, limiting its spread; and in wetland riverbanks, tidal fluctuations may mask area changes. In these scenarios, neither dynamic nor static detection methods can identify the leaks. However, by utilizing the triple conditions of persistent temperature anomalies, limited area, and temperature rise during re-flight detection, concealed leaks can be effectively detected, reducing the false negative rate.

[0042] When the same area meets multiple leakage criteria at the same time, the final warning level shall be determined in the order of priority for dynamic leakage, followed by concealed leakage, and lastly for static leakage. Specifically, when the same suspected leak area simultaneously meets multiple leak detection conditions—for example, triggering both dynamic and static leak detection, or simultaneously meeting both concealed leak and static leak detection—the final warning level needs to be determined according to pre-set priority rules. The priority order is: dynamic leaks have the highest priority, followed by concealed leaks, and static leaks have the lowest priority. Specifically, if an area is simultaneously identified as having both dynamic and static leaks, the dynamic leak detection takes precedence, immediately triggering a high-level alarm and initiating emergency response procedures. If it is simultaneously identified as having both static and concealed leaks, the concealed leak detection takes precedence, as this scenario tends to be a concealed leak caused by environmental limitations, requiring priority confirmation through emergency re-flight to avoid missed detections. This priority rule ensures that when multiple detection logics conflict, a clear and unified warning result is output, avoiding confusion for maintenance personnel due to multiple warnings. It also places the most dangerous and fastest-spreading dynamic leaks with the highest priority, meeting actual emergency response needs.

[0043] After an alert is triggered, the suspected area is reviewed using visible light imaging. If a leak is confirmed, the location is merged; otherwise, the temperature alert conditions are updated.

[0044] Furthermore, during visible light imaging verification, the drone's flight altitude is lowered from the regular inspection altitude to a preset low altitude, and visible light images of the suspected leak area are collected using a high-definition night vision camera.

[0045] Specifically, upon triggering any level of warning, the leak verification process is initiated. The drone lowers its flight altitude from the routine 100 meters to 50 meters, switches to visible light imaging mode, and uses its onboard high-definition night vision camera to photograph the suspected leak area, acquiring visible light images of the area. After acquiring the visible light images, leak characteristic detection is performed. Preset leak characteristics include: liquid reflection (the oil film formed on the ground by the crude oil leak exhibits reflective characteristics under night vision conditions), abnormal soil color (such as the soil color darkening due to crude oil contamination), and withered vegetation (vegetation around the leak area turns yellow or dies due to crude oil contamination). If at least one of the above characteristics appears in the visible light image, the leak is verified. After confirming the leak, the visible light image is fused with the previously acquired infrared thermal image. The fusion uses a weighted overlay method: using the infrared thermal image as the base image (highlighting areas of abnormal temperature), the texture and edge information in the visible light image are superimposed in a weighted manner. The overlay weights were determined through preliminary experiments: Ten known leak points confirmed on-site were selected, and various combinations of infrared image weights ranging from 0.3 to 0.8 (step size 0.1) and visible light image weights corresponding to the remaining proportions were tested. Three maintenance personnel scored the clarity of the leak point outline and the identifiability of surrounding features in each group of fused images from 1 to 5 points. The results showed that the combination of an infrared weight of 0.6 and a visible light weight of 0.4 had the highest average score (4.6 points), so it was used as the default fusion weight. The precise coordinates of the leak points were marked on the fused images. The fused images can simultaneously present temperature differences and visual details, facilitating quick identification of leak locations by maintenance personnel. If no leak features were detected in the visible light image, it was determined to be a false alarm. The relevant information of the anomaly point (including coordinates, temperature value, infrared image, and visible light image) was stored in the interference database. Meanwhile, the temperature warning conditions for the corresponding area are updated based on the data from this false alarm: specifically, the temperature value of the abnormal point is compared with the current threshold. If multiple false alarms are concentrated in a certain temperature range, the leakage temperature threshold for that area is appropriately increased to eliminate similar interference sources and optimize subsequent warning logic.

[0046] This step employs dual-mode verification using both infrared and visible light. By utilizing visual features in the visible light image, the authenticity of the leak is verified, reducing false alarms and improving positioning accuracy. Simultaneously, the accumulation of false alarm data and threshold updates form a closed-loop self-correction mechanism, enabling the system to continuously optimize itself during actual operation and adapt to environmental changes in different regions and seasons.

[0047] In summary, the nighttime leak monitoring method for oil pipelines provided in this embodiment, through gridded temperature monitoring and multi-level leak determination logic, combined with priority adjudication and visible light verification correction, can effectively distinguish between real leaks and non-leakage interference, significantly reducing false alarm and missed alarm rates. This method utilizes UAV infrared imaging technology to overcome nighttime lighting limitations, achieving automated, high-precision leak monitoring of the entire pipeline. It features fast leak response, high location accuracy, excellent regional adaptability, and self-correction capabilities, effectively improving the safety and maintenance efficiency of oil pipelines during nighttime operation.

[0048] Example 2

[0049] Based on the same inventive concept as the aforementioned embodiment of the UAV infrared imaging method for nighttime leak monitoring of oil pipelines, this invention also provides a UAV infrared imaging system for nighttime leak monitoring of oil pipelines. The UAV infrared imaging system for nighttime leak monitoring of oil pipelines includes: a front-end acquisition device and a ground control center. The ground control center comprises: an image display module for real-time display of infrared thermal images acquired by an infrared imager; a temperature data extraction module for extracting temperature data of each monitoring unit from the infrared thermal images; a leak area marking module for marking units with temperatures exceeding a threshold on the infrared image; a threshold determination module for automatically determining temperature thresholds and area change rate standards, and triggering corresponding early warnings based on the determination logic; an anomaly marking and re-flight command module for marking suspected leak areas that trigger early warnings and automatically sending re-flight commands to the UAV; an early warning information push module for pushing leak location, temperature data, infrared images, and area diffusion trends to maintenance personnel; and a data storage and backtracking module for supporting real-time analysis of monitoring data and historical data backtracking.

[0050] Specifically, the data flow at the ground control center is as follows: The ground control center receives infrared thermal imaging data transmitted back by the UAV in real time through the data transmission module. The image display module displays the received infrared thermal image on the monitoring interface in real time for operation and maintenance personnel to monitor intuitively. The temperature data extraction module automatically extracts the temperature value of each monitoring unit (divided into 1m×1m grids) frame by frame from the infrared thermal image and sends the extracted temperature data to the threshold determination module. The threshold determination module compares the temperature data according to the pre-set leakage temperature threshold and area change rate standard: when the temperature of any monitoring unit exceeds the sum of the ambient reference temperature and the threshold of the area, a preliminary warning is triggered, and the unit is marked as a suspected leakage area. Subsequently, the threshold determination module further performs continuous tracking monitoring and comprehensive judgment on the suspected area according to multi-level leakage judgment logic (dynamic leakage judgment, static leakage judgment, and concealed leakage judgment) to determine the leakage type and warning level. The leakage area marking module automatically marks the units with temperatures exceeding the threshold and the areas determined to be leaking on the infrared image with a bright color or border, which facilitates quick location by operation and maintenance personnel. The anomaly marking and re-flight command module marks suspected leak areas as anomalies when an alert is triggered, and automatically generates a re-flight command based on the judgment result. This command is sent to the drone via the data transmission module, controlling the drone to perform an emergency re-flight, reduce flight altitude, or plan a refined inspection route. The alert information push module sends information such as leak location, temperature data, infrared images, and area diffusion trends to maintenance personnel in real time via SMS, APP push, or monitoring interface pop-ups to facilitate timely emergency response. The data storage and backtracking module stores all raw data (infrared images, temperature data, judgment results, alert records, etc.) in a database, supporting post-event queries, playback analysis, and historical data backtracking, while also providing data support for dynamic fine-tuning of thresholds. The above modules work together to achieve fully automated closed-loop monitoring from data reception, processing, judgment to alerting, verification, and storage.

[0051] It should be noted that the image display module, temperature data extraction module, leak area marking module, threshold determination module, anomaly marking and go-around command module, early warning information push module, and data storage and backtracking module are all connected to the processor, and the temperature data extraction module, threshold determination module, leak area marking module, and image display module are connected in sequence; the threshold determination module is also connected to the anomaly marking and go-around command module, early warning information push module, and data storage and backtracking module; the data storage and backtracking module is bidirectionally connected to each of the above modules.

[0052] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.; a general-purpose processor can be a microprocessor or any conventional processor, etc.

[0053] Furthermore, the front-end acquisition device includes a drone and an infrared imager and a data transmission module mounted on the drone; the infrared imager is used to acquire infrared thermal imaging data of the monitoring area in real time, and the data transmission module is used to transmit the infrared thermal imaging data back to the ground control center in real time.

[0054] Specifically, in this embodiment, the drone used is the DJI M3TD, with a maximum takeoff weight of 5.2 kg and an endurance of 120 min. It is equipped with an anti-interference GPS / BeiDou dual-mode navigation system, achieving centimeter-level positioning accuracy. It can fly stably in environments with wind speeds ≤12 m / s and visibility ≥500 m at night, making it suitable for inspecting oil pipelines in various terrains such as oilfields, farmland, and riverside wetlands. The infrared imager uses a high-resolution uncooled infrared focal plane detector with a resolution of 1280×1024, a temperature measurement range of -20-150℃, a temperature measurement accuracy of ±1℃, and a frame rate of 30fps. It can output real-time infrared thermal images of the leak area, capturing the temperature difference between the crude oil and the surrounding environment, providing core data support for leak detection. The data transmission module integrates 4G / 5G and satellite dual-mode communication units. Under line-of-sight conditions, the maximum communication distance between the ground control center and the drone can reach 10 km, with a data transmission delay ≤3s. This enables real-time transmission of infrared thermal imaging data and GPS positioning data, ensuring the timeliness and completeness of the monitoring data. The ground control center is equipped with self-developed leak monitoring software, which has functions such as real-time display of infrared images, extraction of temperature data, marking of leak areas, and push of early warning information. It supports real-time analysis of monitoring data and historical data backtracking, and can automatically complete threshold determination, anomaly marking and re-flight command sending, providing an operating platform for early warning decision-making.

[0055] The specific example of the UAV infrared imaging oil pipeline nighttime leakage monitoring method of the aforementioned embodiment is also applicable to the UAV infrared imaging oil pipeline nighttime leakage monitoring system of this embodiment. Through the foregoing detailed description of the UAV infrared imaging oil pipeline nighttime leakage monitoring method, those skilled in the art can clearly understand the UAV infrared imaging oil pipeline nighttime leakage monitoring method system of this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0056] Example 3

[0057] To better understand this invention, this embodiment uses an oilfield as an example to describe the operation process of the invention in detail. The oilfield covers approximately 1138 square kilometers, mainly distributed across 7 townships, 3 economic development zones, 2 farms, and 28 natural villages in two cities. The core oil-producing area is located in the transition zone between the coastal plain and the sea. Considering the large area of ​​arable land and the dense distribution of waterways, strict anti-seepage measures are required during development to prevent pollution of arable land and groundwater. Furthermore, the area has a sensitive ecological environment, and oilfield development must also consider fisheries and sea salt production. The total length of oil and water pipelines in the entire area is nearly 800 km. These pipelines mostly cross rivers, irrigation ditches, and agricultural areas. Due to erosion from river water and tides, and the abundance of oil and water microorganisms within the pipelines, the pipeline facilities are easily corroded and perforated, causing oil and water leaks and environmental pollution.

[0058] The trial period is from January 1 to March 31, 2025, lasting three months. All inspections will be carried out at night, with a full-area inspection conducted twice a day at 23:00 and 04:00 the next day. The nighttime environmental baseline temperature of the three types of areas will be recorded simultaneously by collecting data by hovering a drone in each area for 10 minutes. The temperature ranges from 18 to 22°C in farmland, 20 to 24°C in oilfield operation areas, and 16 to 20°C in riverside wet areas.

[0059] Data acquisition is performed using a combination of fixed-point infrared thermal imaging and cruise tracking. The specific process is as follows: A fixed monitoring point is set up every 500m along the pipeline. The drone hovers at each monitoring point for 5 minutes and collects a 1280×1024 resolution thermal image using an infrared imager. The temperature data of all 1m×1m cells within a 20m radius of the monitoring point are recorded.

[0060] The drone flies along the pipeline's preset route at a patrol speed of 9 m / s, completing a full-area coverage within 60 minutes. It automatically marks suspected leak areas that trigger a temperature warning exceeding 2°C for the first time, while increasing the patrol frequency to once every hour. Simultaneously, it collects three-dimensional thermal images and area diffusion data of the suspected areas. All data is transmitted in real time to the oilfield's operation and maintenance center database via a 5G private network for subsequent analysis and verification.

[0061] Using a temperature change exceeding 2°C as the core early warning threshold, and combining this with the nighttime baseline temperatures of different areas within the oilfield, the triggering status and identification effectiveness of suspected areas were statistically analyzed during the test period. Comparative imaging of temperature anomaly monitoring points in the work area, farmland area, and riverside wetland area is provided. Figure 2 As shown, the temperatures before and after the two abnormal temperature readings during the patrol were -3.8℃ / 15.9℃, -2.8℃ / 13.1℃, and -10.1℃ / 23.4℃, respectively. It can be seen that the temperature differences between the suspected leak points and the environmental baseline in all three types of areas are significantly higher than the 2℃ warning threshold, reaching 19.7℃, 15.9℃, and 33.5℃, respectively. Furthermore, the temperature signals in the leak areas are stable, and the thermal distribution characteristics are significant. Even with interference from equipment heat radiation, crop obstruction, and reed cover, the temperature difference between the core leak area and the diffusion edge remains high. Slight interference such as water surface reflection can be effectively eliminated through calculation of the regional average temperature difference, demonstrating the universality and effectiveness of the 2℃ temperature warning threshold for the three types of areas.

[0062] Based on nighttime drone patrol data from January 1st to March 31st, 2025, conducted at the oilfield, the monitoring results, focusing on temperature anomaly identification, leak verification, and interference elimination, are shown in Table 1. In terms of regional distribution, the oilfield operating area was the main concentration area for temperature anomalies, totaling 74 instances (58.7%), and also exhibiting the highest monitoring accuracy (79.7%). This is primarily due to the dense pipeline network in this area (e.g., pipelines around the joint station, heating furnace connection pipes), and the abundance of exposed equipment, resulting in minimal heat loss after crude oil leaks and significant temperature signals. The interference source in this area was singular and could be quickly eliminated using the temperature difference fluctuation coefficient, thus achieving a relatively higher accuracy rate than the farmland area (69.4%) and riverside wetlands (62.5%). While the riverside wetlands had the fewest temperature anomaly triggers, they exhibited the largest average temperature difference (33.5℃). This is because the nighttime ambient temperature in the wetlands was extremely low from January to March (-10.1~-5.2℃), significantly amplifying the temperature difference after crude oil leaks, allowing even minor leaks to consistently trigger temperature anomaly alarms. From the perspective of alarm time distribution, temperature anomaly alarms are concentrated in the early to mid-night, which is consistent with environmental stability. Across the entire area, 65.1% of temperature anomalies occurred between 23:00 and 02:00, a period characterized by minimal temperature fluctuations (≤0.3℃ / h) and stable temperature anomaly signals. In riverside wetlands, where fog easily forms on the water surface after 02:00, affecting infrared imaging, 75.0% of anomalies occurred between 23:00 and 00:00. In farmland areas, where crop transpiration is reduced at night, minimizing shading interference, 63.9% of anomalies occurred between 00:00 and 02:00. In oilfield operating areas, with no significant time-based interference, 58.1% of anomalies were evenly distributed between 23:00 and 01:00.

[0063] Table 1. Monitoring Results of Temperature Anomalies During Drone Nighttime Patrols from January to March 2025 Statistical dimensions Oilfield Operation Area farmland area Riverside Wetlands total Total distance (km) 3200 2800 2000 8000 Number of temperature abnormality alarms (times) 74 36 16 126 Monthly Alarm Frequency Distribution (times) 22 / 28 / 24 10 / 15 / 11 7 / 6 / 3 39 / 49 / 38 Number of interference anomalies (times) 15 11 6 32 Accuracy (%) 79.7 69.4 62.5 74.6 Average temperature difference (°C) 19.7±3.5 15.9±2.0 33.5±3.0 21.2±4.5 Concentrated periods of anomalies 23:00-01:00（58.1%） 00:00-02:00（63.9%） 23:00-00:00（75.0%） 23:00-02:00（65.1%） Positioning error (m) ≤2.5 ≤3.0 ≤2.0 ≤3.0 In nighttime leak monitoring of oil pipelines, single temperature anomaly indicators are susceptible to interference from factors such as equipment heat radiation and environmental temperature fluctuations, leading to false alarms or missed alarms. To further improve the accuracy of UAV monitoring of crude oil pipeline leaks, leak area monitoring was conducted at three groups of leak temperature anomaly monitoring points in oilfield operation areas, farmland areas, and riverside wetlands. Imaging images of leak area changes from these three groups were used for comparison. The results are as follows: Figure 3 As shown, due to the influence of regional environmental characteristics, the diffusion rate of the leak area varies significantly in different areas. The work area, due to its open and unobstructed terrain, exhibits free diffusion with a rate reaching 5m. 2 / h. In farmland areas, due to crop shading and soil adsorption limitations, the diffusion rate is only 1.2m. 2 / h. Riverside wetlands, driven by surface runoff and with low soil infiltration resistance, exhibit a diffusion rate of up to 7.5m. 2 The data, measured in h, all exhibit a dynamic characteristic of continuous expansion over time. This demonstrates a strong correlation between the leakage area change and the regional environment, and provides measured data support for setting and dynamically fine-tuning the area threshold.

[0064] Based on nighttime drone patrol data from January 1st to March 31st, 2025, and using 126 temperature anomaly points as a foundation, a new diffusion area monitoring indicator was added to construct a dual-indicator monitoring system for initial temperature anomaly screening and diffusion area verification. The core logic is that a temperature difference ≥2℃ serves as the primary screening condition, marking suspected leak areas, with a diffusion area change ≥1.0m² as the verification criterion. 2 The continuous expansion trend of the diffusion area is used as a secondary verification condition to distinguish between actual leaks and non-leakage interference. The goal is to ultimately achieve accurate identification and low-false-and-leak monitoring, verifying the effectiveness of the monitoring system. The monitoring results are shown in Table 2. The data in the table shows that after adding the change in diffusion area as a secondary screening condition, the total number of interference anomalies decreased from 32 under the single temperature monitoring condition to 16, and the overall monitoring accuracy increased from 74.6% to 87.3%. Single temperature monitoring, relying solely on temperature difference signals, cannot eliminate temperature anomalies caused by non-leakage factors and may be misjudged as suspected leaks, ultimately leading to a high false alarm rate. Using the change in diffusion area as a secondary verification condition reflects the essential physical difference between actual crude oil leaks and non-leakage interference. After a crude oil leak, due to gravity infiltration, surface runoff, and liquid flow, it inevitably exhibits a dynamic expansion characteristic of continuously expanding spatial range and continuously extending boundary morphology, thereby effectively reducing the false alarm rate and improving the accuracy and reliability of the monitoring system.

[0065] Table 2. Monitoring Results of Temperature Anomaly and Diffusion Area in Oil Production Plant Statistical dimensions Oilfield Operation Area farmland area Riverside Wetlands total Number of abnormal alarms (times) 74 36 16 126 Number of interference anomalies (times) 6 7 3 16 Monitoring accuracy rate (%) 91.9 80.6 81.3 87.3 Change in diffusion area (m² / h) 5.1±0.5 1.3±0.3 7.6±0.8 5.3±1.2 Positioning error (m) ≤2.5 ≤3.0 ≤2.0 ≤3.0 However, the current accuracy rate, coupled with the risks of false alarms and missed alarms, and its limitations in scenario adaptability, makes it difficult to meet the actual requirements of operation and maintenance, and still falls short of the needs of engineering applications. This is mainly reflected in three practical challenges: First, the sensitivity to emergency response costs. If an ineffective emergency response is triggered, the long-term cumulative costs of equipment scheduling and personnel deployment will significantly increase the operational burden. Second, the zero-tolerance requirement for high-risk areas. In critical areas such as riverside wetlands and oilfield storage tank areas, even a single missed alarm could trigger major accidents such as water pollution and fires; the current monitoring accuracy of 87.3% is insufficient to meet safety red line requirements. Third, insufficient adaptability to complex scenarios. Extreme weather and sudden terrain changes are likely to affect monitoring accuracy. Therefore, further optimization of the dual dynamic early warning standard system based on temperature and area changes is needed.

[0066] To further improve the accuracy of the monitoring system, it is necessary to dynamically fine-tune the temperature and area monitoring thresholds for different regions. Based on the nighttime drone patrol data of the oilfield from January to March 2025, using 126 temperature anomaly points as samples, a dual dynamic monitoring system for temperature anomalies and area changes was constructed. The temperature and area change thresholds were dynamically fine-tuned by combining the environmental characteristics and diffusion patterns of three typical regions. The technological innovation of the threshold dynamic fine-tuning mechanism is reflected in the following three aspects: First, it breaks through the limitation of a universally applicable fixed threshold. Considering the environmental heterogeneity of oilfield operating areas, farmland areas, and riverside wetlands, a temperature-area dual-indicator coordinated fine-tuning scheme was formulated for the core monitoring contradictions in each region, achieving customized and precise adaptation for each region. Second, it moves away from empirical parameter adjustments and constructs a data-driven setting system. Using measured data as the core, it quantifies the temperature difference, area diffusion rate range, and main interference threshold range of actual leaks in each region, giving the fine-tuned values ​​verifiable and reproducible scientific attributes. Third, abandoning the single-index optimization approach, we achieve a balanced control through dual indices, balancing the interference elimination effect of temperature thresholds with the leakage detection effect of area thresholds. Through coordinated fine-tuning of these two indices, we balance the false alarm and false negative rates in different areas. The fine-tuning thresholds for different areas are ultimately determined through three steps: sample statistical analysis, on-site simulation experiments, and data fitting optimization. First, statistical analysis of the monitoring samples clarifies the adjustment direction. Analysis of 126 temperature anomaly points statistically identifies the characteristics of actual leakage signals and the main interference threshold ranges in three types of areas, clarifying the core direction of threshold adjustment for each area. For example, in the work area, the temperature threshold needs to be increased to eliminate heat radiation, and the area threshold needs to be decreased to detect minor leaks. Then, on-site simulation experiments are conducted to verify the rationality of the thresholds. Multiple gradient temperature and area thresholds are set in typical test sections of the three types of areas, and repeated simulated leakage experiments are carried out to verify the monitoring effectiveness of different threshold combinations, with the standard of maximizing identification accuracy and minimizing false alarm and false negative rates. Finally, the analyzed data is fitted and optimized to determine the final values. The statistical data and experimental results were subjected to binary regression fitting analysis. The accuracy rate was used as the dependent variable, and the temperature and area thresholds were used as independent variables. The value corresponding to the peak accuracy rate in the function was selected as the final fine-tuning standard. At the same time, it was verified that the false alarm and false negative rates were reduced to ≤5% under this value, which is an acceptable range for engineering.

[0067] Table 3 shows the monitoring results for different regions under the dynamic threshold fine-tuning mechanism. The data shows that, under dual early warning standard monitoring conditions, after fine-tuning the temperature and area change early warning thresholds for different regions, the total number of interfering anomalies decreased from 16 to 6, and the monitoring accuracy increased from 87.3% to 95.2%. Introducing the early warning threshold fine-tuning mechanism into the monitoring system can essentially eliminate interfering anomalies, effectively improving the identification of crude oil leak areas. The regional fine-tuning mechanism for temperature and area thresholds essentially upgrades the dual early warning triggering mechanism from a general, adaptable approach to a precisely customized one. By closely considering the interference types and leakage diffusion patterns of each region, the temperature and area thresholds are fine-tuned, ultimately achieving an engineering-level accuracy leap from 87.3% to 95.2%, fully meeting the core requirements of low false alarms, zero missed alarms, and stable operation across all scenarios in oil pipeline monitoring.

[0068] Table 3 Comparison of monitoring data in three typical regions Statistical dimensions Oilfield Operation Area farmland area Riverside Wetlands total Number of abnormal alarms (times) 74 36 16 126 Temperature threshold fine-tuning (°C) ≥2.5 ≥2.0 ≥1.8 —— <![CDATA[Area threshold fine-tuning (m 2 / h)]]> ≥0.19 ≥0.13 ≥0.31 —— Number of interference anomalies (times) 3 2 1 6 Monitoring accuracy rate (%) 95.9 94.4 93.8 95.2 Even with a monitoring accuracy rate of 95.2% after fine-tuning the dual early warning mechanism thresholds, there are still a few missed detection scenarios where temperature changes exceed 2°C but the area affected does not reach the fine-tuned threshold. These scenarios mainly manifest as temperature measurement deviations caused by equipment obstruction in the work area, leaks in equipment gaps, low re-flight efficiency due to crop interference in farmland, missed detections due to field topography, and delayed early warnings caused by tidal fluctuations in coastal wetlands, resulting in unmonitored changes in leak area. In these scenarios, it is necessary to determine concealed leaks, i.e., triggering an alarm when the temperature exceeds 3°C, re-flight monitoring of temperature changes at 1-hour intervals, and constructing a three-dimensional monitoring closed-loop strategy of dynamic area verification and static temperature enhancement. By focusing on the intensity and persistence of temperature change signals, this strategy identifies leak points where the area expansion is slow or concealed due to regional specific environmental limitations, further reducing the number of missed detections and improving monitoring accuracy.

[0069] Based on the experimental data in Tables 1 to 3, the positioning accuracy of this invention remained consistently ≤3.0m across all three monitoring stages (single temperature monitoring, temperature-area dual-index monitoring, and monitoring after threshold fine-tuning). This positioning accuracy relies on the synergy of the following technical components: the RTK-GPS / BeiDou dual-mode navigation system provides centimeter-level real-time UAV positioning; the barometric altimeter provides flight altitude data; and the pinhole imaging model converts image pixel coordinates into geographic coordinates. The data from these three components are fused and calculated in the coordinate mapping process of Example 2, ultimately outputting the geographic coordinates of each monitoring cell. During the experiment, the positioning error in each area met the design requirement of ≤3.0m, providing a reliable spatial reference for leak location and subsequent emergency repairs.

[0070] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for nighttime leak monitoring of oil pipelines using UAV infrared imaging, characterized in that, Includes the following steps: The monitoring area is divided into independent monitoring units according to size, and leakage temperature thresholds and area change rate standards are established; the leakage temperature thresholds and area change rate standards can be dynamically fine-tuned based on the environmental characteristics of different areas. The temperature of each unit is obtained by thermal imaging, and units whose temperature exceeds the leakage temperature threshold are marked as suspected leakage areas. The suspected leakage area is continuously monitored. When the rate of change of its area and the characteristics of its shape change meet the dynamic leakage conditions, it is determined to be a dynamic leakage and a first-level warning is triggered. For suspected areas that fail the dynamic leakage assessment, when the temperature difference change of the internal monitoring unit at different times meets the static leakage conditions, it is determined to be a static leakage and a level-two warning is triggered. For suspected leak areas that fail both dynamic and static leak detection, if the temperature anomaly persists and the area change is limited, the UAV will re-monitor the temperature rise value of the core area by re-flying. When the temperature rise value meets the conditions for concealed leak, it will be determined as a concealed leak and a level-two warning will be triggered. When the same area meets multiple leakage criteria at the same time, the final warning level shall be determined in the order of priority for dynamic leakage, followed by concealed leakage, and lastly for static leakage. After an alert is triggered, the suspected area is reviewed using visible light imaging. If a leak is confirmed, the location is merged; otherwise, the temperature alert conditions are updated.

2. The method according to claim 1, characterized in that, The temperature threshold was determined through preliminary field tests and historical data statistics to establish a nighttime baseline temperature range for different environmental types, with a threshold defined as exceeding the baseline temperature by 2°C. The area change rate standard was determined through simulated leakage tests in different areas, and the area change rate standard was set at 1.0m. 2 / h.

3. The method according to claim 1, characterized in that, The continuous monitoring interval is 1 hour; the dynamic leakage condition is: analyzing its shape and area change characteristics, if the suspected area shows irregular expansion in shape after three consecutive monitoring sessions, and the area change rate is ≥1.0m 2 If / h, it is determined to be a dynamic leak.

4. The method according to claim 1, characterized in that, The static leakage condition is as follows: extract the temperature values ​​of at least three continuous monitoring units in the suspected leakage area at different times, and calculate the temperature difference change. If the temperature difference is ≥3℃, the interference of ambient temperature fluctuation is excluded, and it is determined to be a static leakage; the time interval between the different times is 30 minutes.

5. The method according to claim 1, characterized in that, The conditions for a concealed leak are as follows: the temperature of the suspected leak area exceeds the baseline threshold by 2°C and lasts for ≥1 hour, and the rate of change of the area does not reach the standard after the corresponding area threshold is fine-tuned. An emergency re-flight monitoring is initiated for the suspected leak area for 1 hour. If the temperature of the core cell of the suspected leak area increases by ≥3°C after the re-flight monitoring, it is determined to be a concealed leak.

6. The method according to claim 1, characterized in that, During the visible light imaging verification, the drone's flight altitude is lowered from the normal inspection altitude to a preset low altitude, and visible light images of the suspected leak area are collected using a high-definition night vision camera.

7. The method according to claim 1, characterized in that, The specific method of dividing the monitoring area into independent monitoring units according to size is as follows: based on the GIS map of the oil pipeline, it is divided into independent monitoring units with a size of 1m×1m.

8. The method according to claim 1, characterized in that, The dynamic fine-tuning includes statistical analysis of monitoring samples in different regions, on-site simulated leakage tests, and data fitting optimization to determine the fine-tuned thresholds and standards for each region.

9. A drone-based infrared imaging system for nighttime leak monitoring of oil pipelines, comprising front-end acquisition equipment and a ground control center, characterized in that, The ground control center includes: The image display module is used to display the infrared thermal image acquired by the infrared imager in real time; A temperature data extraction module is used to extract temperature data of each monitoring unit from the infrared thermal image; The leak area marking module is used to mark units whose temperature exceeds the threshold on the infrared image; The threshold determination module is used to automatically determine the temperature threshold and the area change rate standard, and trigger corresponding warnings based on the determination logic. The anomaly marking and go-around command module is used to mark suspected leakage areas that trigger warnings and automatically send go-around commands to drones; The early warning information push module is used to push leak location, temperature data, infrared images, and area diffusion trends to operation and maintenance personnel; The data storage and backtracking module is used to support real-time analysis of monitoring data and backtracking of historical data.

10. The system according to claim 9, characterized in that, The front-end acquisition device includes a drone and an infrared imager and data transmission module mounted on the drone; The infrared imager is used to collect infrared thermal imaging data of the monitoring area in real time, and the data transmission module is used to transmit the infrared thermal imaging data back to the ground control center in real time.

Images