Event camera based liquid leakage metrology detection method and system
By combining event cameras with image processing technology and physical models, the problems of response lag and insufficient accuracy in liquid leak detection have been solved. High-precision liquid leakage detection has been achieved under complex lighting and high-speed leakage conditions, and it is suitable for large-scale multi-point synchronous detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DONGFANG ELECTRIC MACHINERY
- Filing Date
- 2026-03-12
- Publication Date
- 2026-05-29
AI Technical Summary
Existing technologies for liquid leak detection suffer from problems such as response lag, insufficient detection accuracy, inability to accurately quantify, and limited applicability, especially in complex lighting and high-speed leakage scenarios where effective detection and quantification are difficult.
A liquid leakage measurement and detection method based on event cameras is adopted. By acquiring asynchronous event streams through event cameras and combining image processing technology and physical models, the amount of droplet leakage can be accurately estimated in real time.
It achieves high-precision, low-latency liquid leakage detection in complex lighting and high-speed leakage scenarios, and can perform non-contact measurement without changing the liquid properties, making it suitable for large-scale, multi-point synchronous detection.
Smart Images
Figure CN122115538A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of oil and water pipeline monitoring in power plants, specifically involving a liquid leakage metering and detection method and system based on event cameras. Background Technology
[0002] In core industrial sectors such as petroleum, chemical, and energy, industrial pipelines and containers are the core carriers for the transmission and storage of media, and the liquids flowing through or stored inside them are mostly flammable, explosive, toxic, and corrosive liquids. Once a liquid leak occurs, it can not only cause serious safety accidents such as fires, explosions, and poisoning of personnel, but also cause the loss of a large amount of valuable media, environmental pollution, and production line shutdowns for maintenance, resulting in incalculable economic losses and negative social impacts. Therefore, leak detection and quantitative analysis are of paramount importance.
[0003] Currently, traditional leak detection methods in the industry mainly rely on pressure sensor monitoring and manual inspection, which have obvious technical shortcomings: pressure sensors only trigger alarms when the leakage reaches a certain threshold, resulting in significant response lag; manual inspection is limited by factors such as inspection cycle, personnel experience, and environmental visibility, resulting in a high rate of missed detections and an inability to accurately quantify the leakage amount, making it difficult to meet the needs of refined safety management.
[0004] While existing visual inspection technologies have been applied in some scenarios, their adaptability and detection accuracy are insufficient. This technology can only function in specific scenarios such as small-scale, single-light illumination, and low-speed leakage. When faced with the instantaneous nature of high-speed leakage in industrial sites, complex lighting (strong light, backlight, shadows), and background interference, it is prone to missed or false identification problems. Moreover, its core limitation lies in its inability to accurately quantify the leakage amount, making it difficult to support subsequent risk assessment and handling decisions.
[0005] Currently, there are three main types of methods for detecting liquid leaks in the industry, each with its own advantages, disadvantages, and limitations in applicable scenarios: The first is the direct measurement method using measuring cups and cylinders. This method is simple to operate and has high measurement accuracy, but it is only suitable for small-scale, single-point leak scenarios and cannot cover the detection needs of large-scale, multi-point synchronous leaks in industrial sites, thus limiting its practicality.
[0006] Secondly, indirect measurement methods such as pressure drop and flow rate comparison are widely used due to their wide applicability and the fact that they do not require contact with the leaking medium. The core principle is to indirectly calculate the leak volume by monitoring changes in system pressure or the difference in inlet and outlet flow rates. However, the conversion process is easily affected by additional error variables such as temperature fluctuations, changes in medium viscosity, and pipeline resistance, which significantly reduces the detection sensitivity and makes it difficult to detect minute leaks.
[0007] Thirdly, there are auxiliary quantitative analysis methods such as fluorescent tracer detection. These methods involve adding fluorescent tracers to the medium and using specialized equipment to capture the tracer signal to locate and quantify leaks. However, this method requires the addition of additional chemical substances, which may affect the purity of the medium. It is not suitable for scenarios involving food-grade or high-purity chemical raw materials. Furthermore, the amount of tracer added and the diffusion rate can interfere with the quantitative results, limiting its application scenarios. Summary of the Invention
[0008] This application aims to solve the above-mentioned problems existing in the prior art. It proposes a liquid leakage measurement and detection method and system based on an event camera, which can detect the amount of liquid leakage in real time without changing the liquid properties. By taking the event stream of the event camera as input and combining image processing technology and physical model, the amount of liquid leakage can be accurately estimated.
[0009] To achieve the above objectives, the technical solution of the present invention is as follows: A liquid leak measurement and detection method based on an event camera includes the following steps: Step S1. Use an event camera to continuously monitor concentrated areas of industrial pipelines or flange connections to obtain raw event data consisting of asynchronous event streams; Step S2. Perform spatiotemporal filtering on the original event data to generate event image frames, and perform denoising filtering on the event image frames to eliminate irrelevant background noise; Step S3. Input the filtered event image frame into a pre-trained target detection neural network to identify whether there is a liquid leakage target in the image. The target detection neural network adopts the YOLO architecture. During training, droplets, personnel, animals and other equipment are labeled separately to achieve accurate differentiation of droplet targets. Step S4. If a liquid leakage target is detected, the corresponding event image frame is input into the image segmentation network to obtain the pixel-level segmentation result of the droplet region; Step S5. Calculate the droplet contour area based on the segmentation result, and determine whether the current leakage state is dripping or spraying according to the preset droplet shape-volume mapping model, and then convert the droplet contour area into a single leakage volume. Step S6. Combine the timestamp information of the event camera to count the number of leakage events and the volume of a single leakage per unit time, calculate the real-time leakage rate, and integrate over time to obtain the cumulative leakage amount.
[0010] Furthermore, in step S1, the event camera is a LUCID EVS series event vision sensor, which features microsecond-level temporal resolution, high dynamic range, and robustness to changes in illumination. Without external supplementary lighting, leak detection can be completed solely relying on ambient natural light. The event camera's large dynamic range allows it to stably output valid event data even in low-light or sudden changes in strong light conditions.
[0011] Furthermore, in step S2, the spatiotemporal filtering process includes accumulating events within a set time window, which is between 0.8 milliseconds and 10 milliseconds, to balance the event density and the signal-to-noise ratio of the image frame.
[0012] Furthermore, in step S5, the droplet morphology-volume mapping model is established based on offline calibration. It is determined by collecting the contour area of various standard droplets with known volumes under the same observation perspective and fitting the nonlinear function relationship between area and volume. This nonlinear function relationship reflects the power law relationship between droplet volume and its projected contour area, that is, the power function relationship between volume and area.
[0013] Furthermore, the droplet morphology-volume mapping model converts the projected area A of the droplet into the single leakage volume V using the following formula:
[0014] Where k is a calibration coefficient related to the physical properties of the liquid and the observation perspective, and α is a power exponent describing the geometric relationship between the three-dimensional shape of the droplet and its two-dimensional projection, determined by fitting offline calibration experimental data; for the dripping mode, different sets of coefficients are used ( , For the injection mode, the corresponding coefficient set is ( , ).
[0015] Furthermore, in step S6, the real-time leakage rate Q(t) is calculated based on leakage events within a unit time Δt, and the formula is:
[0016] in, In the time window The single leakage volume of the i-th leakage event detected internally; the cumulative leakage amount. The real-time leakage rate Q(t) is obtained by integrating it over the monitoring period T. .
[0017] Furthermore, in step S4, the image segmentation network is a U-Net or Mask R-CNN structure, used to accurately extract droplet boundaries from event image frames.
[0018] A liquid leak measurement and detection system based on an event camera, comprising: At least one LUCID EVS event camera is deployed in an area with concentrated industrial piping or at flange connections to acquire asynchronous event streams; An event acquisition card, connected to the event camera, is used to receive raw event data and convert it into a standard data format; The server is communicatively connected to the event acquisition card and is equipped with a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements a liquid leakage measurement and detection method. The server supports simultaneous connection to multiple event cameras to achieve multi-point synchronous leakage monitoring.
[0019] Furthermore, the event acquisition card is connected to the event camera via the GigE Vision interface and transmits data to the server via the PCIe interface.
[0020] The server is deployed at the edge of the industrial site, has real-time processing capabilities, and a response latency of less than 50 milliseconds.
[0021] The system requires no additional lighting and utilizes the event camera's adaptability to low-light environments to accurately detect and measure liquid leaks even under complex lighting conditions such as darkness, backlighting, or flickering light sources.
[0022] The advantages of this application are: 1. This application proposes a liquid leakage measurement and detection method and system based on an event camera, which can detect the amount of liquid leakage in real time without changing the liquid properties. By taking the event stream of the event camera as input and combining image processing technology and physical model, the amount of liquid leakage can be accurately estimated.
[0023] 2. The method of this application enables direct measurement and detection of leaked liquids without contact, reducing the physical quantity conversion process and improving measurement accuracy. Compared with measuring cups and cylinders, it has the advantages of a wider measurement range and non-contact measurement. Compared with fluorescent tracer detection methods, it eliminates the need for additional substances to be added to the liquid being tested.
[0024] 3. The patented method and system can achieve non-invasive, wide-range direct detection of leaked liquid volume in narrow locations where manual measurement is inconvenient.
[0025] 4. Compared with traditional leakage detection methods based on frame-type visible light cameras, this invention utilizes the asynchronous and high temporal resolution characteristics of event cameras to significantly improve the ability to capture transient droplet events, and effectively avoids motion blur and illumination interference, thereby achieving high-precision and low-latency leakage measurement. Attached Figure Description
[0026] Figure 1 This is a flowchart of the application process. Detailed Implementation
[0027] To make the objectives, technical solutions, and advantages of the embodiments of the invention clearer, the technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, not all embodiments. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0028] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.
[0029] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0030] In the description of this invention, it should be noted that the terms "upper," "vertical," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are only used for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0031] This invention presents an aircraft surface feature segmentation method based on contour constraint optimization, which uses a deep learning network to segment targets in an image. The method learns feature information from the image through a feature extraction backbone network, then fits the outer contour constraint of the target based on this feature information, initially segmenting the target in the image. Finally, the target contour constraint is used to optimize the segmentation result, achieving high-precision segmentation of each target instance in the image.
[0032] Example 1 like Figure 1 As shown, a liquid leakage measurement and detection method based on an event camera includes the following steps: Step S1. Use an event camera to continuously monitor concentrated areas of industrial pipelines or flange connections to obtain raw event data consisting of asynchronous event streams; Step S2. Perform spatiotemporal filtering on the original event data to generate event image frames, and perform denoising filtering on the event image frames to eliminate irrelevant background noise; Step S3. Input the filtered event image frame into a pre-trained target detection neural network to identify whether there is a liquid leakage target in the image. The target detection neural network adopts the YOLO architecture. During training, droplets, personnel, animals and other equipment are labeled separately to achieve accurate differentiation of droplet targets. Step S4. If a liquid leakage target is detected, the corresponding event image frame is input into the image segmentation network to obtain the pixel-level segmentation result of the droplet region; Step S5. Calculate the droplet contour area based on the segmentation result, and determine whether the current leakage state is dripping or spraying according to the preset droplet shape-volume mapping model, and then convert the droplet contour area into a single leakage volume. Step S6. Combine the timestamp information of the event camera to count the number of leakage events and the volume of a single leakage per unit time, calculate the real-time leakage rate, and integrate over time to obtain the cumulative leakage amount.
[0033] This invention uses the high frame rate characteristics of an event camera to slice the leaked liquid into contours, and then uses a droplet morphology-volume mapping model to convert the captured contours into volumes. The single volumes are accumulated to obtain the current leak volume.
[0034] Example 2 like Figure 1 As shown, a liquid leakage measurement and detection method based on an event camera includes the following steps: Step S1. Use an event camera to continuously monitor concentrated areas of industrial pipelines or flange connections to obtain raw event data consisting of asynchronous event streams; Step S2. Perform spatiotemporal filtering on the original event data to generate event image frames, and perform denoising filtering on the event image frames to eliminate irrelevant background noise; Step S3. Input the filtered event image frame into a pre-trained target detection neural network to identify whether there is a liquid leakage target in the image. The target detection neural network adopts the YOLO architecture. During training, droplets, personnel, animals and other equipment are labeled separately to achieve accurate differentiation of droplet targets. Step S4. If a liquid leakage target is detected, the corresponding event image frame is input into the image segmentation network to obtain the pixel-level segmentation result of the droplet region; Step S5. Calculate the droplet contour area based on the segmentation result, and determine whether the current leakage state is dripping or spraying according to the preset droplet shape-volume mapping model, and then convert the droplet contour area into a single leakage volume. Step S6. Combine the timestamp information of the event camera to count the number of leakage events and the volume of a single leakage per unit time, calculate the real-time leakage rate, and integrate over time to obtain the cumulative leakage amount.
[0035] In step S1, the event camera is a LUCID EVS series event vision sensor, which features microsecond-level temporal resolution, high dynamic range, and robustness to changes in illumination. Without external supplementary lighting, leakage detection can be completed solely using ambient natural light. The event camera's large dynamic range allows it to stably output valid event data even in low-light or sudden changes in strong light conditions.
[0036] In step S2, the spatiotemporal filtering process includes accumulating events within a set time window, which is between 0.8 milliseconds and 10 milliseconds, to balance the event density and the signal-to-noise ratio of the image frame.
[0037] In step S5, the droplet morphology-volume mapping model is established based on offline calibration. It is determined by collecting the contour area of various standard droplets with known volumes under the same observation angle and fitting the nonlinear function relationship between area and volume. This nonlinear function relationship reflects the power law relationship between droplet volume and its projected contour area, that is, the power function relationship between volume and area. The droplet morphology-volume mapping model converts the projected area A of the droplet into the single leakage volume V using the following formula:
[0038] Where k is a calibration coefficient related to the physical properties of the liquid and the observation perspective, and α is a power exponent describing the geometric relationship between the three-dimensional shape of the droplet and its two-dimensional projection, determined by fitting offline calibration experimental data; for the dripping mode, different sets of coefficients are used ( , For the injection mode, the corresponding coefficient set is ( , ).
[0039] In step S6, the real-time leakage rate Q(t) is calculated based on leakage events within a unit time Δt, and the formula is:
[0040] in, In the time window The single leakage volume of the i-th leakage event detected internally; the cumulative leakage amount. The real-time leakage rate Q(t) is obtained by integrating it over the monitoring period T. .
[0041] In step S4, the image segmentation network is a U-Net or Mask R-CNN structure, used to accurately extract droplet boundaries from event image frames.
[0042] This invention uses the high frame rate characteristics of an event camera to slice the leaked liquid into contours, and then uses a droplet morphology-volume mapping model to convert the captured contours into volumes. The single volumes are accumulated to obtain the current leak volume.
[0043] Example 3 A liquid leak measurement and detection system based on an event camera, comprising: At least one LUCID EVS event camera is deployed in an area with concentrated industrial piping or at flange connections to acquire asynchronous event streams; An event acquisition card, connected to the event camera, is used to receive raw event data and convert it into a standard data format; The server is communicatively connected to the event acquisition card and is equipped with a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements a liquid leakage measurement and detection method. The server supports simultaneous connection to multiple event cameras to achieve multi-point synchronous leakage monitoring.
[0044] Furthermore, the event acquisition card is connected to the event camera via the GigE Vision interface and transmits data to the server via the PCIe interface.
[0045] The server is deployed at the edge of the industrial site, has real-time processing capabilities, and a response latency of less than 50 milliseconds.
[0046] The system requires no additional lighting and utilizes the event camera's adaptability to low-light environments to accurately detect and measure liquid leaks even under complex lighting conditions such as darkness, backlighting, or flickering light sources.
[0047] This invention uses the high frame rate characteristics of an event camera to slice the leaked liquid into contours, and then uses a droplet morphology-volume mapping model to convert the captured contours into volumes. The single volumes are accumulated to obtain the current leak volume.
Claims
1. A liquid leakage measurement and detection method based on an event camera, characterized in that, Includes the following steps: Step S1. Use an event camera to continuously monitor concentrated areas of industrial pipelines or flange connections to obtain raw event data consisting of asynchronous event streams; Step S2. Perform spatiotemporal filtering on the original event data to generate event image frames, and perform denoising filtering on the event image frames to eliminate irrelevant background noise; Step S3. Input the filtered event image frame into a pre-trained target detection neural network to identify whether there is a liquid leakage target in the image. The target detection neural network adopts the YOLO architecture and labels droplets, personnel, animals and other equipment respectively during training. Step S4. If a liquid leakage target is detected, the corresponding event image frame is input into the image segmentation network to obtain the pixel-level segmentation result of the droplet region; Step S5. Calculate the droplet contour area based on the segmentation result, and determine whether the current leakage state is dripping or spraying according to the preset droplet shape-volume mapping model, and then convert the droplet contour area into a single leakage volume. Step S6. Combine the timestamp information of the event camera to count the number of leakage events and the volume of a single leakage per unit time, calculate the real-time leakage rate, and integrate over time to obtain the cumulative leakage amount.
2. The liquid leakage measurement and detection method based on an event camera according to claim 1, characterized in that, In step S1, the event camera is a LUCID EVS series event vision sensor.
3. The liquid leakage measurement and detection method based on an event camera according to claim 1, characterized in that, In step S2, the spatiotemporal filtering process includes accumulating events within a set time window, which is between 0.8 milliseconds and 10 milliseconds.
4. The liquid leakage measurement and detection method based on an event camera according to claim 1, characterized in that, In step S5, the droplet morphology-volume mapping model is established based on offline calibration. It is determined by collecting the contour area of various standard droplets with known volumes under the same observation viewpoint and fitting the nonlinear function relationship between area and volume. This nonlinear function relationship reflects the power law relationship between droplet volume and its projected contour area, that is, the power function relationship between volume and area.
5. The liquid leakage measurement and detection method based on an event camera according to claim 4, characterized in that, The droplet morphology-volume mapping model converts the projected area A of the droplet into the single leakage volume V using the following formula: Where k is a calibration coefficient related to the physical properties of the liquid and the observation perspective, and α is a power exponent describing the geometric relationship between the three-dimensional shape of the droplet and its two-dimensional projection, determined by fitting offline calibration experimental data; for the dripping mode, different sets of coefficients are used ( , For the injection mode, the corresponding coefficient set is ( , ).
6. The liquid leakage measurement and detection method based on an event camera according to claim 1, characterized in that, In step S6, the real-time leakage rate Q(t) is calculated based on leakage events within a unit time Δt, and the formula is: in, In the time window The single leakage volume of the i-th leakage event detected internally; the cumulative leakage amount. The real-time leakage rate Q(t) is obtained by integrating it over the monitoring period T. 。 7. The liquid leakage measurement and detection method based on an event camera according to claim 1, characterized in that, In step S4, the image segmentation network is a U-Net or Mask R-CNN structure, used to accurately extract droplet boundaries from event image frames.
8. A liquid leakage measurement and detection system based on an event camera, characterized in that, include: At least one LUCID EVS event camera is deployed in an area with concentrated industrial piping or at flange connections to acquire asynchronous event streams; An event acquisition card, connected to the event camera, is used to receive raw event data and convert it into a standard data format; The server is communicatively connected to the event acquisition card and is equipped with a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements a liquid leakage measurement and detection method. The server supports simultaneous connection to multiple event cameras to achieve multi-point synchronous leakage monitoring.
9. A liquid leakage measurement and detection system based on an event camera according to claim 8, characterized in that, The event acquisition card connects to the event camera via the GigE Vision interface and transmits data to the server via the PCIe interface.
10. A liquid leakage measurement and detection system based on an event camera according to claim 8, characterized in that, The server is deployed at the edge of the industrial site, has real-time processing capabilities, and a response latency of less than 50 milliseconds.