A liquid hydrogen leak remote rapid detection method, system, medium and program product

Abstract

The present application provides a method, system, medium and program product for remote rapid detection of liquid hydrogen leakage, which relates to the field of chemical fuel leakage detection technology. The method includes using a low-light video acquisition device to collect a target characteristic cloud image of a liquid hydrogen leakage gas mass, and extracting the cloud characteristic information, while obtaining the distance information between it and the leakage gas mass through a ranging sensor. The cloud characteristic information and distance information are then input into a leakage source judgment model based on multi-source data and trained by a domestic AI intelligent processor module to obtain the leakage source location information. Subsequently, the target characteristic cloud range boundary information is determined, and the leakage area is clarified in combination with the leakage source location. Finally, the liquid hydrogen leakage dedicated disposal system is automatically associated and started, so that the leakage source can be quickly and accurately located and disposed of, effectively reducing the hazards of liquid hydrogen leakage.

Description

Technical Field

[0001] The present application relates to the technical field of chemical fuel leakage detection, and in particular to a method, system, medium and program product for remote rapid detection of liquid hydrogen leakage. Background Art

[0002] Hydrogen fuel leakage is a safety hazard that has emerged in recent years with the rapid development of hydrogen energy technology. During the storage, transportation, and use of liquid hydrogen fuel, leakage can occur due to aging equipment, operational errors, or external interference. Once a leak occurs, hydrogen rapidly diffuses into the air, mixing with oxygen in the air to form an explosive mixture. Within a specific concentration range, exposure to open flames or high temperatures can cause fires or explosions, posing a serious threat to personnel safety and the environment. Therefore, the issue of liquid hydrogen fuel leakage needs to be addressed urgently.

[0003] At present, the existing liquid hydrogen fuel leak detection method is: step 1, pre-simulate a leakage event on site, test the near-end and far-end hydrogen concentration detectors to obtain prior data; step 2, obtain data through leakage event testing, and determine the leakage detection alarm threshold and event level estimation threshold; step 3, during the operation phase, the near-end and far-end detectors continuously output and detect and record, and this method outputs the final result after conditional judgment based on the alarm threshold and event level estimation threshold.

[0004] However, the existing methods require multiple threshold judgments, and the response speed of liquid hydrogen fuel leak detection is slow and the safety and reliability are low. Summary of the Invention

[0005] The present application provides a method, system, medium and program product for remote rapid detection of liquid hydrogen leakage, which is used to efficiently, accurately and in real time monitor and locate areas where liquid hydrogen leakage may occur in liquid hydrogen storage facilities, and to deal with liquid hydrogen leakage without contact.

[0006] In a first aspect, the present application provides a method for remote and rapid detection of liquid hydrogen leaks, which is applied to a detection system. The method includes: real-time acquisition of a target characteristic cloud image of a liquid hydrogen leaking gas mass through a low-light video acquisition device; extraction of cloud characteristic information based on the target characteristic cloud image; acquisition of distance information between the ranging sensor and the liquid hydrogen leaking gas mass through a ranging sensor; inputting the cloud characteristic information and the distance information into a preset leakage source judgment model to obtain liquid hydrogen leakage source location information, wherein the leakage source judgment model is based on cloud sample characteristic data under multiple known liquid hydrogen leakage scenarios, multiple distance data between the ranging sensor and the cloud samples, and measured and labeled leakage source location data, and is obtained by iterative training using a domestic AI intelligent processor module; after determining the range boundary information of the target characteristic cloud based on the target characteristic cloud image, the leakage area is determined in combination with the liquid hydrogen leakage source location information; and a dedicated disposal system for liquid hydrogen leakage is controlled to efficiently dispose of the leakage area.

[0007] By adopting the above technical solutions, it is possible to quickly and accurately locate the leakage area and deal with it efficiently, effectively reducing the hazards of liquid hydrogen leakage, improving detection and treatment efficiency, and ensuring the safety of liquid hydrogen-related scenarios.

[0008] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0009] 1. Due to the collaborative operation of low-light video acquisition equipment and ranging sensors, combined with a leakage source judgment model based on iterative training of multi-source data to locate the leakage area, and then control the liquid hydrogen leakage special disposal system to treat the leakage area, the technical problems of inaccurate liquid hydrogen leakage positioning and untimely treatment in the existing technology are effectively solved, thereby achieving the technical effect of quickly and accurately locating the liquid hydrogen leakage area and efficiently handling the leakage, reducing the hazards of liquid hydrogen leakage, and ensuring the safe operation of liquid hydrogen-related scenarios.

[0010] 2. Due to the technical means of using the pre-order layer, merging layer and decoding layer in the target detection algorithm to collaboratively process the target feature cloud image in sequence to accurately determine the range boundary information, it effectively solves the technical problem in the existing technology that it is difficult to accurately define the range of the liquid hydrogen leakage cloud from the image, resulting in large deviations in the determination of the leakage area, thereby achieving improved accuracy in determining the leakage area, enabling the liquid hydrogen leakage dedicated disposal system to act more accurately on the leakage area and improve the treatment effect.

[0011] 3. Due to the technical means of using a ranging sensor to determine the laser beam parameters at the center point of the leakage area to guide the liquid hydrogen leakage special disposal system for disposal, it effectively solves the technical problem in the existing technology that the disposal system has no precise parameter basis and is difficult to effectively cover the leakage area. In turn, the liquid hydrogen leakage special disposal system is achieved to accurately cover the leakage area, effectively block the spread of liquid hydrogen, improve the accuracy and effectiveness of emergency treatment, and reduce environmental pollution and facility damage. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a flow chart of a method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application;

[0013] Figure 2 Schematic diagram of detecting liquid hydrogen leakage gas mass in an embodiment of the present application;

[0014] Figure 3 Schematic diagram of the leakage area of ​​the liquid hydrogen leakage gas mass in the embodiment of the present application;

[0015] Figure 4 This is a working diagram of the method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application;

[0016] Figure 5 This is another working schematic diagram of the method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application;

[0017] Figure 6 This is another flow chart of the method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application;

[0018] Figure 7 This is a schematic diagram of the network structure of the target detection algorithm in the embodiment of the present application;

[0019] Figure 8 It is a schematic diagram of the structure of a physical device of the detection system in the embodiment of the present application. DETAILED DESCRIPTION

[0020] The terms used in the following examples of the present application are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in the specification and appended claims of the present application, the singular expressions "a," "an," "said," "above," "the," and "this" are intended to include plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present application refers to and encompasses any or all possible combinations of one or more of the listed items.

[0021] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of the technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of this application, unless otherwise specified, "plurality" means two or more.

[0022] For ease of understanding, the following describes the process of the method provided by this implementation. Figure 1 , is a flow chart of a method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application.

[0023] S101, using a low-light video acquisition device to collect a target characteristic cloud image of the liquid hydrogen leak in real time;

[0024] When liquid hydrogen leaks, due to its extremely low temperature (-252.87°C), the water vapor in the surrounding air will quickly condense to form water mist, and then form an air mass with specific characteristics. The low-light video acquisition equipment in the detection system plays a key role in this step. In this application, the low-light video acquisition equipment may include a low-light camera. The low-light camera uses advanced imaging sensor technology with a resolution of not less than 1920×1080. This high resolution can capture the subtle details of the liquid hydrogen leakage air mass and its surrounding environment. For the liquid hydrogen leakage air mass, there are differences in temperature, density, etc. between it and the surrounding environment. Such differences will appear in the image as different brightness, contrast and other characteristics. High-resolution imaging can present these characteristics more accurately; the response frequency of the low-light camera is 30Hz, which enables it to capture 30 frames of images per second. In the liquid hydrogen leakage scenario, the state of the air mass may change rapidly over time, such as the diffusion rate of the air mass, shape changes, and interaction with the surrounding environment. A 30Hz response frequency ensures the camera captures these dynamic changes promptly, avoiding missing important information due to slow response. In the early stages of a liquid hydrogen leak, a cloud of gas may spread outward at a rapid rate. A higher response frequency allows the camera to record the changes in the cloud's position and shape at different moments. By analyzing multiple consecutive frames of imagery, the cloud's spread speed and direction can be accurately calculated, providing critical data for assessing the leak's development trends. Low-light-level cameras with a minimum pixel count of 4 million provide rich image information. More pixels mean greater detail and color gradation in the image. In liquid hydrogen leak scenarios, the characteristics of a cloud of gas may manifest as subtle changes in color and brightness, and sufficient pixels allow for more nuanced capture of these changes. The photosensors in low-light-level cameras respond strongly to weak light. When the faint light surrounding the leaking cloud strikes the sensor, the photosensors convert the optical signal into an electrical signal. The device is equipped with a specially designed optical lens system featuring low dispersion and high light transmittance. The optical lens system focuses and refracts light, ensuring that light reflected or scattered from the liquid hydrogen leak is accurately projected onto the imaging sensor, minimizing light loss and image distortion. For example, the use of special optical coating technology enhances the transmittance of light of specific wavelengths while suppressing interference from stray light, thereby improving image clarity and contrast.

[0025] The low-light video capture device also integrates an intelligent image preprocessing module. This module processes the raw image data in real time at the moment of image capture. It automatically adjusts the image's brightness, contrast, and color balance to highlight the characteristics of the liquid hydrogen leak plume. For example, an algorithm enhances the contrast between the plume and its surroundings, making its outline more distinct. Simultaneously, it performs image noise reduction, removing noise caused by low light levels and environmental interference, ensuring that the captured image quality meets the requirements of subsequent analysis.

[0026] In actual operation, the low-light video acquisition equipment continuously captures images at a preset frame rate (e.g., 30 frames per second). By analyzing multiple consecutive frames, it can capture the dynamic characteristics of the liquid hydrogen leak gas cloud, such as its diffusion speed and shape changes, providing rich data support for accurate leakage assessment.

[0027] S102, extracting cloud feature information based on the target feature cloud image;

[0028] After acquiring an image of the target characteristic cloud, the detection system uses advanced image processing algorithms and deep learning techniques to extract cloud feature information. First, an edge detection algorithm is used to identify the contour features of the cloud. This algorithm determines the location of the object's edge by calculating the grayscale change rate of pixels in the image. For a liquid hydrogen leak, the cloud differs from its surroundings in temperature, density, and other aspects, resulting in distinct edge features in the image. Accurate edge detection accurately delineates the shape of the cloud, such as circular, elliptical, or irregular. This shape information is crucial for subsequently determining the type and status of the leak source. Next, a texture analysis algorithm is used to extract internal texture features of the cloud. The texture within a liquid hydrogen leak typically reflects internal airflow, temperature distribution, and interaction with the surrounding environment. This texture analysis algorithm statistically analyzes the spatial distribution of pixel grayscale values ​​in the image to extract characteristic parameters such as texture direction, texture roughness, and texture periodicity. For example, if there is a vortex-like texture inside the air mass, it may indicate that the liquid hydrogen jet at the leak point has certain rotational characteristics, which helps to further locate the specific location and leakage method of the leak source.

[0029] Color features are also an important element to extract. Although image color information is relatively weak in low-light environments, the color characteristics of the air mass can still be obtained through specific color space conversion and enhancement algorithms. For example, converting an image from the common RGB color space to the HSV color space, where the hue component more effectively reflects the color characteristics of an object, can be used. Liquid hydrogen leaks may exhibit a specific hue range in the image due to a unique optical phenomenon caused by condensation of surrounding water vapor at low temperatures. Analysis of this hue range can assist in determining the composition and state of the air mass.

[0030] S103, obtaining distance information between the distance measuring sensor and the liquid hydrogen leakage gas mass through a distance measuring sensor;

[0031] The distance measuring sensor in the detection system uses high-precision laser ranging technology to obtain the distance information between the gas mass and the liquid hydrogen leakage. The distance measuring sensor in the present application can adopt a high-precision laser ranging sensor. The laser emitting module inside the high-precision laser ranging sensor works based on semiconductor laser technology. When the sensor receives a measurement instruction, it precisely controls the current pulse of the laser diode to make the laser diode generate a high-intensity laser pulse signal. The laser emitting module can emit a laser beam of a specific wavelength, and the laser beam has a high degree of directionality and energy concentration. When the laser emitting module is started, a high-intensity laser pulse signal is generated by precisely controlling the current pulse of the laser diode. After the laser beam is emitted toward the area where the liquid hydrogen leakage gas mass is located, it will be reflected when it encounters the surface of the gas mass. The reflected light signal is received by the receiving module of the distance measuring sensor. The receiving module uses a highly sensitive photodetector that can convert the weak reflected light signal into an electrical signal.

[0032] The high-precision laser ranging sensor features low-temperature resistance and explosion-proof design, enabling stable operation in the unique and hazardous environment of liquid hydrogen leaks. This low-temperature design ensures the sensor's proper operation in the low temperatures and large ambient temperature fluctuations that can occur with liquid hydrogen leaks. Its internal electronic and optical components are specially selected and packaged to withstand the effects of low temperatures and temperature fluctuations. Its explosion-proof design prevents explosions caused by sparks or other factors generated by the sensor's operation in environments where hydrogen leaks could create explosive mixtures. The sensor housing is constructed of explosion-proof materials, and its internal circuitry is explosion-proof, complying with relevant explosion-proof standards. The high-precision laser ranging sensor also boasts a low latency response time of less than 2 seconds. In liquid hydrogen leak detection, where leak conditions can change rapidly, quickly acquiring accurate distance information is crucial for timely assessment. This low latency response time ensures the sensor can quickly transmit laser pulses and receive echoes within a very short timeframe, reducing measurement errors caused by time delays and ensuring that measurement results more accurately reflect the changing location of the leaking gas mass.

[0033] In order to accurately calculate the distance, the timing circuit inside the ranging sensor plays a key role. The timing circuit records the time interval from the moment the laser pulse is emitted to the moment the reflected light signal is received. According to the speed of light (approximately 3x10 8 m / s), and then use the distance calculation formula to calculate the distance information.

[0034] In practical applications, to ensure measurement accuracy and reliability, distance sensors use an averaging method. Multiple laser pulses are emitted continuously over a short period of time, and the corresponding reflection times are recorded. The resulting distance values ​​are then averaged to reduce measurement errors caused by environmental interference (such as airflow fluctuations and dust particle scattering).

[0035] S104. Inputting the cloud characteristic information and the distance information into a preset leakage source determination model to obtain liquid hydrogen leakage source location information. The leakage source determination model is based on characteristic data of cloud samples in multiple known liquid hydrogen leakage scenarios, distance data between multiple ranging sensors and cloud samples, and measured and labeled leakage source location data, and is obtained by iterative training using a domestically produced AI intelligent processor module.

[0036] After acquiring cloud feature information and distance information, the detection system inputs them into a pre-trained leak source identification model. This leak source identification model is built based on a deep learning framework. Its architecture primarily comprises a combination of a convolutional neural network (CNN) and a recurrent neural network (RNN). The CNN component extracts features from the input cloud feature image and consists of multiple convolutional, pooling, and fully connected layers. The convolutional layers slide convolution kernels of varying sizes across the image to extract local features such as edges and texture. The pooling layers downsample the features, reducing the data volume while retaining key features. The fully connected layers integrate the previously extracted features and output a feature vector. The RNN component processes the distance information sequence captured by the ranging sensor. It learns how distance information changes over time and exploits its temporal characteristics. Through its internal recurrent structure, the RNN unit receives the current distance information input at each time step and updates the current hidden state based on the hidden state from the previous time step, thereby capturing the dynamic changes in distance information. During the model training phase, a large amount of cloud sample feature data from known liquid hydrogen leak scenarios, distance data between ranging sensors and cloud samples, and precisely measured and annotated leak source location data were used. This data was divided into training, validation, and test sets. During training, the cloud sample feature image data was first input into the CNN for feature extraction, generating an image feature vector. Simultaneously, the corresponding distance data sequence was input into the RNN for processing, generating a distance feature vector. These two feature vectors were then concatenated and passed as the final input features to the subsequent fully connected layer. The fully connected layer predicts the leak source location based on the input features, compares the prediction with the actual annotated location, and calculates a loss function. A backpropagation algorithm was used to update and optimize the model parameters based on the loss function, continuously adjusting parameters such as the convolutional kernel weights and the fully connected layer weights in the CNN and RNN to minimize the loss function. During training, the validation set was used to monitor model performance to prevent overfitting. Training was terminated when the model performance on the validation set no longer improved.

[0037] In actual operation, when the model inputs the cloud characteristics and distance information to be detected, it goes through the forward propagation process described above and outputs the predicted location of the liquid hydrogen leak source, including the coordinates (x, y, z) of the leak source in three-dimensional space. To improve the model's robustness and accuracy, the model also adopts an ensemble learning strategy. This involves training multiple models with different structures or parameter initializations and then combining their outputs during prediction, such as through weighted averaging, to obtain the final leak source location information, further reducing prediction errors.

[0038] S105, after determining the range boundary information of the target characteristic cloud according to the target characteristic cloud image, determine the leakage area in combination with the location information of the liquid hydrogen leakage source;

[0039] The detection system first processes the target cloud image to determine its boundaries. This system uses image segmentation algorithms, such as threshold segmentation, region growing, or clustering. For example, threshold segmentation analyzes the grayscale distribution of pixels in the image and selects an appropriate threshold to classify the pixels into two categories: those belonging to the air mass region and those belonging to the background region. This threshold can be determined using various methods, such as the maximum inter-class variance method, which automatically finds a threshold that maximizes the inter-class variance, thereby separating the air mass from the background as much as possible.

[0040] The region growing algorithm begins with one or more seed points in the image and, based on certain similarity criteria (such as pixel grayscale value similarity or color similarity), continuously merges adjacent similar pixels into a growing region until a stop-growth condition is met, such as when the growing region stops expanding or reaches a preset size. Ultimately, the air mass boundary is determined. After determining the air mass boundary, the leak region is determined by combining it with the previously obtained liquid hydrogen leak source location information. Because the leak location information is represented by coordinates in three-dimensional space, while the air mass boundary in the image is a two-dimensional region, coordinate transformation and spatial mapping are required. By establishing a transformation relationship between the image coordinate system and the real-world coordinate system, the coordinates of the boundary points in the two-dimensional image are transformed into three-dimensional space. For example, given the intrinsic parameters (such as focal length) and extrinsic parameters (such as the camera's position and posture in space) of the image acquisition device, coordinate transformation can be achieved using methods such as perspective transformation. Then, with the leak source location as the center, a leak region is determined based on the air mass boundary in three-dimensional space, encompassing the leak source and its potential impact area. This region can be an irregular three-dimensional volume, its shape and size determined by the actual distribution of the air mass and the location of the leak source. For example, the spatial portion within a certain range from the leakage source (set based on experience or actual conditions) and within the air mass boundary can be defined as the leakage source area, providing an accurate target area for the subsequent efficient disposal of the liquid hydrogen leakage dedicated disposal system.

[0041] S106. Control the dedicated disposal system for liquid hydrogen leakage to efficiently dispose of the leakage area.

[0042] The detection system and dedicated disposal system are connected and control commands are transmitted via wireless communication modules (such as Wi-Fi, Bluetooth, or a dedicated wireless communication protocol). Once the leak area is determined, the detection system generates control commands based on the area's location, size, and other information and sends them to the dedicated disposal system.

[0043] In the above embodiments, the advantage of the present invention compared with the prior art is that, by adopting advanced AI algorithms, the specific location of the liquid hydrogen leak can be quickly located, and the leakage distance can be measured using an integrated automation system; this method not only greatly shortens the time for leak detection and response, but also improves the safety and accuracy of the treatment process. Subsequently, a professional disposal system was deployed to effectively control potential risks. Compared with traditional leak handling methods, this method not only optimizes the leak emergency response process, but also reduces the impact on the environment while ensuring the safety of personnel. This efficient processing method that combines AI technology and automated machinery is a major breakthrough in the field of liquid hydrogen safety management, and provides strong technical support for the safe operation of the hydrogen energy industry in the future.

[0044] For easier understanding, see Figure 2 , Figure 2 Schematic diagram of detecting liquid hydrogen leakage gas mass in an embodiment of the present application.

[0045] exist Figure 2 The figure shows the characteristic air mass produced by the leakage of liquid hydrogen. Liquid hydrogen has an extremely low temperature. When liquid hydrogen leaks, it will quickly change from liquid to gas. This process is a strong endothermic process, which will cause the temperature of the surrounding environment to drop sharply. When the water vapor in the surrounding air is cooled, the molecular thermal motion slows down, the distance between the water vapor molecules decreases, and they gradually condense into tiny water droplets. These tiny water droplets gather together to form a characteristic air mass. Through dynamic observation of the characteristic air mass, such as observing the area with the highest concentration of the characteristic air mass, the center direction of diffusion, etc., the direction of the leak source can be preliminarily inferred. When combined with other detection methods (such as ranging sensors, low-light video acquisition equipment, etc.), the temperature-changing characteristic air mass can serve as an important reference to help further determine the precise location of the leakage area.

[0046] The following combination Figure 2 , see Figure 3 , Figure 3 Schematic diagram of the leakage area of ​​the liquid hydrogen leakage gas mass in the embodiment of the present application.

[0047] exist Figure 3In the process, the leakage area of ​​the liquid hydrogen leakage gas mass can be displayed on the infrared display screen. When liquid hydrogen leaks, the temperature of its surrounding environment will change significantly. The infrared display screen senses this temperature change through infrared detectors. Infrared detectors work based on the principle of thermal radiation. All objects emit infrared radiation, and its intensity is related to the temperature of the object. When liquid hydrogen leaks, the temperature of the leakage area is significantly different from that of the surrounding environment due to the low temperature characteristics of liquid hydrogen. This temperature difference will cause the intensity of infrared radiation emitted by this area to be different from that of the surrounding environment. The sensor in the infrared detector can capture this change in infrared radiation intensity and convert it into an electrical signal. After processing and amplification, these electrical signals are transmitted to the display module of the infrared display screen. After receiving the processed electrical signal, the display module of the infrared display screen presents an image of the liquid hydrogen leakage area on the screen according to the signal strength and location information. In the process, the infrared display screen displays the image of the liquid hydrogen leakage area on the screen according to the signal strength and location information. Figure 3 In the figure, a quadrilateral is used to represent the liquid hydrogen leakage area, and the "+" sign in the middle may represent the center of the leakage area. This display method intuitively presents the location and approximate range of the leakage. The display module uses different grayscale or color to distinguish the leakage area from the surrounding environment. For example, in Figure 3 In the image, the leak area is represented in gray, while the background is dark gray, which can clearly identify the leak area. In this way, operators can quickly and intuitively obtain key information about liquid hydrogen leaks, making it easier to take timely countermeasures.

[0048] In some embodiments, a distance sensor on the detection device emits a laser beam toward the center of the leak area to determine the center-to-center distance d. This principle is similar to the previously described method for obtaining distance information between the sensor and the liquid hydrogen leaking gas mass and will not be further elaborated here. The distance sensor is equipped with an internal angle measurement device, such as a gyroscope or tilt sensor. When emitting the laser beam, the angle measurement device simultaneously measures the horizontal angle α between the laser beam and the horizontal plane. This angle information can be used to determine the vertical position of the leak area relative to the detection device. In practical scenarios, the liquid hydrogen leak area may be located at different heights, such as a high-altitude pipeline leak or a valve leak near the ground. Combining the horizontal angle α with the center-to-center distance d can accurately determine the location of the leak area in three-dimensional space. Using trigonometric relationships, the horizontal distance L from the leak source to the detection device can be calculated as d × cosα. Determining the horizontal distance L helps assess the horizontal separation between the leak source and the detection device, which is important for rationally arranging equipment layout and planning emergency response routes. For example, during on-site rescue operations, the horizontal distance can be used to determine a safe approach route to prevent rescuers from getting too close to the leak source. At the same time, the height between the leak source and the liquid hydrogen leaking gas cloud is h = d × sinα. This height information is crucial for fully understanding the leak situation. It can help determine the possible spread direction and range of the liquid hydrogen leak, as well as the degree of impact on surrounding facilities at different heights. For example, if the leak source is at a high altitude, the liquid hydrogen gas cloud may be more likely to spread upward, affecting ventilation equipment or electrical equipment at high altitudes. This needs to be taken into account when formulating emergency response plans, and appropriate protective and treatment measures must be implemented.

[0049] The dedicated liquid hydrogen leak disposal system precisely adjusts its response parameters based on the centerline distance d and horizontal angle α provided by the detection device. First, the system's outlet water pressure is adjusted based on the centerline distance d to ensure the disposal system can reach the leaking area. At longer distances, the outlet water pressure is increased to ensure the disposal system has sufficient kinetic energy to reach the target location. At closer distances, the pressure is appropriately reduced to avoid waste and unnecessary impacts caused by excessive disposal. The system's disposal angle is then adjusted based on the horizontal angle α to ensure the disposal system accurately covers the leak source.

[0050] During the disposal process, the detection system continuously monitors the leak area. After a set duration, the target cloud image captured by the low-light video acquisition device is analyzed again to extract cloud characteristics, primarily including its shape, size, texture, color, and brightness. Regarding shape, if the cloud gradually changes from its initial irregular, diffuse shape to a relatively regular and stable one, it may indicate that the leak has been contained. Regarding size, a significant decrease in the area of ​​the cloud indicates a decrease in hydrogen release from the leak source, indicating that the disposal system has effectively suppressed its spread. Changes in texture can reflect changes in airflow and composition within the cloud. For example, a smoother, previously chaotic texture may indicate a decrease in hydrogen concentration and a stabilization of airflow. Color and brightness are also important. A lighter or lowered color or brightness may indicate a decrease in hydrogen concentration or changes in optical properties caused by reactions with the disposal system. These characteristics are interrelated, and comprehensive analysis can provide a more comprehensive assessment of leak control effectiveness. The cloud characteristics obtained from the recaptured image are then compared against the pre-set characteristics. If all or most of the characteristic information meets the preset standards, such as the shape, size, texture, color, and brightness of the air mass, the leak is considered to have been handled. If some characteristics are not met, further monitoring, adjustment of treatment measures, or extension of the assessment time may be required until all preset characteristics are met to ensure that the liquid hydrogen leak is effectively controlled and site safety is guaranteed.

[0051] For ease of understanding, the following describes the working scenario between the liquid hydrogen leakage area detection and the detection device provided in this application. Figure 4 , Figure 4 This is a working diagram of the method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application:

[0052] exist Figure 4The figure shows a scenario where the leak area is higher than the detection device, which may include a coordinate anchoring device. The figure shows the line connecting the liquid hydrogen leak detection and anchoring device to the center of the leak area as the center straight line distance d. The horizontal angle α is the angle between the laser beam emitted by the coordinate anchoring device and the horizontal plane. The horizontal distance from the leak source to the detection device, L = d × cosα, and the height between the leak source and the liquid hydrogen leaking gas mass, h = d × sinα, can be further calculated. The figure shows a vehicle, indicating that the detection device may be mounted on a mobile platform, such as a vehicle. This design allows the detection device to be flexibly moved to areas where liquid hydrogen leaks may occur for detection. At large liquid hydrogen storage sites or transportation routes, a mobile detection platform can quickly reach suspected leak locations and promptly identify and locate the source. The detection device on the vehicle can transmit detection data in real time to a control center via wireless communication technology. Control center personnel can use this data to analyze and make decisions, such as dispatching emergency response personnel and adjusting on-site equipment operations, to ensure that liquid hydrogen leaks are promptly and effectively addressed.

[0053] In some embodiments, when the leakage area is below or equal to the detection device, refer to Figure 5 , Figure 5 This is a working diagram of the method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application:

[0054] exist Figure 5 In the figure, based on the measured center straight line distance d and the horizontal angle β between the center straight line and the leakage area, the horizontal distance L can be calculated as d×cosβ. The height h is calculated as h=d×sinβ. When the leakage area is lower than or equal to the detection device, the value of the height h may be zero or negative, which helps to determine the vertical position of the leakage area relative to the detection device. Similarly, this liquid hydrogen leakage area detection and coordinate anchoring device is usually installed on a mobile platform (such as a vehicle). Figure 5 The truck shown in the figure allows the detection device to be flexibly moved to areas where liquid hydrogen leaks may occur for detection.

[0055] After combining the above content, the following is a more detailed description of the process of the method provided by this implementation. Figure 6 , is another flow chart of the method for remote rapid detection of liquid hydrogen leakage in an embodiment of the present application.

[0056] S601. Extract hydrogen cluster feature information from the target feature cloud image through the pre-processing layer of the target detection algorithm. The pre-processing layer of the target detection algorithm in the detection system primarily uses a deep learning convolutional neural network (CNN) architecture to extract hydrogen cluster feature information. First, the convolution layer in the pre-processing layer plays a key role. A convolution layer contains multiple convolution kernels with different weight parameters, similar to a set of filters. When the target feature cloud image is input into the convolution layer, each convolution kernel slides across the image, performing a convolution operation with the pixels in the image. For example, a 3x3 convolution kernel performs a weighted sum operation with each 3x3 pixel neighborhood in the image to extract local image features. By operating with multiple convolution kernels, image features at different scales and directions, such as edges, textures, and corners, can be simultaneously acquired. These features are crucial for identifying the presence and morphology of hydrogen clusters in an image. During the training phase, the parameters of the pre-processing layer (such as the convolution kernel weights) are optimized using a backpropagation algorithm. A large amount of image data labeled with hydrogen clouds and backgrounds is fed into the network. Based on the error between the network's predictions and the actual annotations, the convolution kernel weights are continuously adjusted, enabling the network to better extract the characteristics of hydrogen clouds. After multiple iterations of training, the pre-processing layer learns representative hydrogen cloud feature patterns, enabling it to accurately extract relevant hydrogen cloud information from target cloud images during actual detection, providing a solid foundation for subsequent processing.

[0057] S602, summarizing the characteristic information of the hydrogen cluster through the convergence layer of the target detection algorithm;

[0058] The pooling layer of the object detection algorithm plays a key role in integrating information throughout the feature extraction and object localization process. The pooling layer receives multiple feature maps from the preceding layers as input. These feature maps contain characteristic information of hydrogen clusters at different levels and scales, such as edges and textures at different resolutions. The pooling layer first concatenates these feature maps along the channel dimension, stacking them together to form a new feature tensor. This operation aims to integrate feature information from different levels into a unified representation for subsequent comprehensive analysis. The pooling layer then processes the concatenated feature tensor using global average pooling. Global average pooling calculates the average of all pixel values ​​in each feature channel, compressing the feature information for each channel into a single value. This operation effectively reduces the dimensionality of the feature tensor while preserving the overall feature information of each channel, avoiding the risk of overfitting and making the feature representation more robust to spatial variations in the input image. After global average pooling, the pooling layer can also introduce additional fully connected layers. The fully connected layer further integrates and transforms feature information. By adjusting the weight parameters of the fully connected layer, the correlations and combination patterns between different feature channels are learned. Each neuron in the fully connected layer is connected to all neurons in the previous layer. Through weighted summation and activation function calculation, the input feature information is mapped to a new feature space, thereby more effectively representing the overall characteristics of the hydrogen cluster.

[0059] Through these operations in the merging layer, the scattered hydrogen cluster feature information from the previous layer is aggregated and refined to form a more representative and discriminative feature representation, which provides strong support for the subsequent accurate determination of the bounding box and confidence value of the hydrogen cluster.

[0060] S603: Determine multiple initial bounding boxes and confidence values ​​corresponding to the initial bounding boxes based on the summarized characteristic information of the hydrogen cloud. The initial bounding boxes are used to preliminarily define the area suspected to be the hydrogen cloud in the target characteristic cloud image. The confidence values ​​are used to indicate the degree of confidence that the bounding boxes accurately define the hydrogen cloud.

[0061] After aggregating the hydrogen cluster feature information through the pooling layer, the detection system utilizes a region proposal network (RPN) to determine multiple initial bounding boxes and corresponding confidence scores. The RPN first generates a series of anchor boxes on the feature map. These anchor boxes have pre-defined scales and aspect ratios. For example, five scales (from smallest to largest: s1, s6, s3, s4, and s5) and three aspect ratios (1:1, 1:2, and 2:1) are set, resulting in 15 anchor boxes for each location in the feature map. The centers of the anchor boxes correspond one-to-one with pixels on the feature map, and their sizes and shapes cover the various possible situations in which hydrogen clusters may appear in the image. For each anchor box, the RPN further processes the aggregated hydrogen cluster feature information through a convolutional layer to obtain a feature vector corresponding to the anchor box. This feature vector is then input into two fully connected layers. One of the fully connected layers predicts the probability that the anchor box contains a hydrogen cluster, i.e., the confidence score. The output of this fully connected layer is mapped through a logistic regression function, resulting in a confidence score ranging from 0 to 1. For example, an output of 0.8 indicates an 80% probability that the anchor box contains a hydrogen cluster. Another fully connected layer predicts the parameters for adjusting the anchor box, including the position (x and y offsets) and size (scaling of width w and height h). These adjustment parameters are learned from a large dataset of known hydrogen cluster locations and shapes. During the training phase, the RPN network is trained using image data annotated with the true locations and boundaries of hydrogen clusters. For each anchor box, a cross-entropy loss is calculated between its predicted confidence score and the true label (whether it contains a hydrogen cluster), as well as a regression loss based on the difference in position and size between the predicted adjustment parameters and the true hydrogen cluster bounding box and the anchor box. Using backpropagation, these losses are used to optimize the parameters of the convolutional and fully connected layers in the RPN network, enabling the network to more accurately predict initial bounding boxes and confidence scores.

[0062] After training, the RPN network can generate corresponding confidence values ​​and adjustment parameters for each anchor box in actual detection based on the input summary hydrogen cluster feature information. For anchor boxes with higher confidence values ​​(for example, greater than the set threshold of 0.7), they are used as candidates for the initial bounding box. Then, these candidate anchor boxes are corrected according to the predicted adjustment parameters to obtain an initial bounding box that is closer to the real hydrogen cluster boundary. In this way, multiple initial bounding boxes that preliminarily frame the suspected hydrogen cluster area in the target feature cloud image are obtained, as well as the confidence value corresponding to each bounding box that indicates the degree of credibility of its accurate framing of the hydrogen cluster.

[0063] S604: The decoding layer based on the target detection algorithm determines a final hydrogen characteristic cloud target bounding box and a corresponding category probability according to the initial bounding box and the confidence value, where the category probability indicates the probability that the bounded area belongs to the hydrogen characteristic cloud;

[0064] After receiving the initial bounding box and confidence value, the decoding layer of the target detection algorithm further processes it to determine the final hydrogen feature cloud target bounding box and category probability. The decoding layer first screens the initial bounding box. According to the set confidence threshold (such as 0.5), the bounding boxes with confidence values ​​greater than the threshold are retained, and those bounding boxes with low confidence and unlikely to contain hydrogen clusters are removed. This step can effectively reduce the subsequent calculation amount while improving the accuracy of the detection results. For the retained initial bounding boxes, the decoding layer uses the non-maximum suppression (NMS) algorithm for further processing. The principle of the NMS algorithm is: when multiple bounding boxes have a high degree of overlap in the image (that is, multiple bounding boxes of the same hydrogen cluster are detected), only the bounding box with the highest confidence is retained, and other bounding boxes with high overlap are suppressed. The specific operation process is as follows: First, the bounding box with the highest confidence score is selected as the currently retained bounding box. Then, the intersection of union (IoU) of other bounding boxes with the currently retained bounding box is calculated. If the IoU is greater than a set threshold (such as 0.5), these bounding boxes are considered to have a high degree of overlap with the currently retained bounding box and are suppressed (i.e., removed from the list of candidate bounding boxes). This process is repeated until all bounding boxes have been processed. The NMS algorithm can produce a more accurate and refined set of bounding boxes, reducing redundant detection results.

[0065] After determining the final bounding box, the decoding layer uses a classification network to predict the class probability of each object within the bounding box belonging to a hydrogen cloud. This classification network, based on a convolutional neural network architecture, learns the characteristic patterns of different objects by iteratively training on a large number of images annotated with hydrogen clouds and other objects (such as background and interference). The classification network's input is the image region within the bounding box processed by the decoding layer (which can be obtained by cropping the original image), and its output is the probability that the region belongs to a hydrogen cloud. For example, an output of 0.95 indicates that there is a 95% probability that the object within the bounding box is a hydrogen cloud. To improve classification accuracy, the decoding layer can also utilize multi-scale feature fusion. This involves fusing feature maps from different layers (such as shallow and deep layers). The shallow feature maps contain more detailed information, while the deep feature maps contain more abstract semantic information. By fusing these features, the classification network can better identify the characteristics of the hydrogen cloud, thereby more accurately predicting the class probability.

[0066] S605: If the category probability is greater than a preset threshold, the hydrogen characteristic cloud target bounding box is determined as the final range boundary information.

[0067] After obtaining the corresponding class probability for each hydrogen cloud target bounding box, the detection system compares it with a preset threshold. This threshold is set based on actual application requirements and detection accuracy requirements, for example, 0.8. If the class probability for a bounding box is greater than the preset threshold, it indicates that the bounding box is likely to contain a hydrogen cloud, and the detection system determines this as the final boundary information. This final boundary information is used to subsequently determine the leak area and guide the liquid hydrogen leak disposal system's response. For example, if the class probability for a bounding box is 0.9, which is greater than the preset threshold of 0.8, the area enclosed by the bounding box is considered a valid hydrogen cloud area, and its boundary coordinates serve as an important basis for subsequent processing. If the class probability is less than or equal to the preset threshold, the detection system deems the bounding box unlikely to contain a hydrogen cloud, possibly due to false detection or other interference factors. The detection system excludes these bounding boxes from subsequent operations such as leak area determination. This avoids unnecessary processing caused by false detections and improves the accuracy and effectiveness of the entire detection and processing process.

[0068] In the embodiments of this application, the detection system can accurately screen the true hydrogen characteristic cloud target bounding box based on the category probability, thereby determining reliable final range boundary information, providing accurate target area positioning for subsequent liquid hydrogen leak treatment, ensuring that the liquid hydrogen leak dedicated treatment system can accurately treat the leak area, and effectively reducing the risk of liquid hydrogen leaks. At the same time, in actual applications, the preset thresholds can be adjusted and optimized according to different scenarios and needs to balance the relationship between detection accuracy and missed detection rate, adapting to various complex liquid hydrogen leak detection environments.

[0069] For ease of understanding, the following is a detailed introduction to the structure of the target detection algorithm. Figure 7 , Figure 7 This is a schematic diagram of the network structure of the target detection algorithm in the embodiment of the present application:

[0070] exist Figure 7In the figure, the entire network structure exhibits a multi-layered feature, containing multiple modules and connection paths. This multi-layered structure helps to gradually extract and process feature information in the image, from low-level features (such as edges and textures) to high-level features (such as object shape and category). The network contains various types of modules, such as convolutional layers (Conv) and connection layers (Concat). There are multiple feature extraction paths in the network structure, which extract features of different levels and types through different module combinations and connection methods. For example, starting from the left side of the figure, there are several paths that go through operations such as convolution, connection, and upsampling to gradually extract and process features. This complex network structure ultimately achieves precise positioning and accurate classification of the target through multi-level, multi-module feature extraction and processing. In the target detection task, the network needs to determine whether the target object exists in the image and determine its position (usually represented by a bounding box) and category.

[0071] The following describes the detection system in the embodiment of the present invention from the perspective of hardware processing. Figure 8 , is a schematic diagram of the physical device structure of the detection system in an embodiment of the present application.

[0072] It should be noted that Figure 8 The structure of the detection system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0073] like Figure 8 As shown, the detection system includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 802 or the program loaded from the storage part 808 into the random access memory (RAM) 808, such as the method described in the above embodiment. In the RAM 808, various programs and data required for system operation are also stored. The CPU 801, ROM 802 and RAM 808 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0074] The following components are connected to the I / O interface 805: an input section 806 including an audio input device, a push button switch, and the like; an output section 807 including a liquid crystal display (LCD), an audio output device, an indicator light, and the like; a storage section 808 including a hard disk and the like; and a communication section 809 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like, is installed in the drive 810 as needed so that a computer program read therefrom can be installed into the storage section 808 as needed.

[0075] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 809 and / or installed from a removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the various functions defined in the present invention are performed.

[0076] It should be noted that specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0077] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings.

[0078] Specifically, the detection system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the remote rapid detection method for liquid hydrogen leakage provided by the above embodiment is implemented.

[0079] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the detection system described in the above embodiments, or may exist independently and not be incorporated into the detection system. The storage medium carries one or more computer programs, which, when executed by a processor of the detection system, enable the detection system to implement the remote rapid detection method for liquid hydrogen leaks provided in the above embodiments.

[0080] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0081] As used in the above embodiments, the term “when…” may be interpreted to mean “if…” or “after…” or “in response to determining…” or “in response to detecting…”, depending on the context. Similarly, the phrases “upon determining…” or “if (stated condition or event) is detected” may be interpreted to mean “if determining…” or “in response to determining…” or “upon detecting (stated condition or event)” or “in response to detecting (stated condition or event)”, depending on the context.

[0082] Those skilled in the art will appreciate that all or part of the process steps in the above-described method embodiments can be implemented by a computer program instructing the relevant hardware. The program can be stored in a computer-readable storage medium, and when executed, the program can include the process steps in the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for rapid remote detection of liquid hydrogen leakage, applied to a detection system, characterized in that: The method comprises: The target characteristic cloud image of the liquid hydrogen leakage gas cloud is collected in real time through low-light video acquisition equipment; extracting cloud feature information according to the target feature cloud image; Acquiring distance information between the distance measuring sensor and the liquid hydrogen leakage gas mass through a distance measuring sensor; Inputting the cloud characteristic information and the distance information into a preset leakage source judgment model to obtain liquid hydrogen leakage source location information, wherein the leakage source judgment model is based on cloud sample characteristic data under multiple known liquid hydrogen leakage scenarios, distance data between multiple ranging sensors and cloud samples, and measured and labeled leakage source location data, and is obtained through iterative training using an AI intelligent processor module; After determining the range boundary information of the target characteristic cloud according to the target characteristic cloud image, determining the leakage area in combination with the liquid hydrogen leakage source location information; Controlling a dedicated disposal system for liquid hydrogen leakage to efficiently dispose of the leakage area; After determining the range boundary information of the target characteristic cloud according to the target characteristic cloud image, and before determining the leakage area in combination with the liquid hydrogen leakage source location information, the method further includes: Extracting hydrogen cluster feature information from the target feature cloud image through the preceding layer of the target detection algorithm; Aggregating the hydrogen cluster feature information through the merging layer of the target detection algorithm; Determining a plurality of initial bounding boxes and confidence values ​​corresponding to the initial bounding boxes based on the aggregated hydrogen cloud feature information, wherein the initial bounding boxes are used to preliminarily define an area suspected to be a hydrogen cloud in the target feature cloud image, and the confidence values ​​are used to indicate the degree of confidence that the bounding boxes accurately define the hydrogen cloud; The decoding layer based on the target detection algorithm determines a final hydrogen characteristic cloud target bounding box and a corresponding category probability according to the initial bounding box and the confidence value, wherein the category probability indicates the probability degree that the framed area belongs to the hydrogen characteristic cloud; If the category probability is greater than a preset threshold, the hydrogen characteristic cloud target bounding box is determined as the final range boundary information.

2. The method according to claim 1, characterized in that The steps of collecting a target characteristic cloud image of a liquid hydrogen leaking gas cloud in real time using a low-light video acquisition device specifically include: Real-time monitoring and identification of characteristic gas masses corresponding to the liquid hydrogen leak by a low-light-level camera, wherein the low-light-level camera has a resolution of not less than 1920×1080, a response frequency of 30 Hz, a pixel of not less than 4 million, and supports a zoom of not less than 20 times; A target characteristic cloud image is determined according to the characteristic air mass changes.

3. The method according to claim 1, characterized in that The step of obtaining the distance information between the distance measuring sensor and the liquid hydrogen leakage gas mass by using the distance measuring sensor specifically includes: A high-precision laser ranging sensor is used to emit a laser towards the liquid hydrogen leaking gas mass and record the laser pulse emission time. The high-precision laser ranging sensor has a low-temperature resistance and explosion-proof design and a low-latency response time of less than 2 seconds; Determining an echo reception time when the high-precision laser ranging sensor receives an echo; The distance information between the distance measuring sensor and the liquid hydrogen leakage gas mass is determined according to the speed of light, the laser pulse emission time, and the echo reception time.

4. The method according to claim 1, wherein After determining the range boundary information of the target characteristic cloud according to the target characteristic cloud image, and determining the leakage area in combination with the liquid hydrogen leakage source location information, the method further includes: emitting a laser beam to the center point of the leakage area by the distance measuring sensor on the detection device to determine the center straight line distance d; Determine the horizontal angle α between the laser beam and the horizontal plane by the distance measuring sensor; The liquid hydrogen leakage special treatment system on the detection device is controlled to perform efficient treatment according to the center straight line distance d and the horizontal angle α.

5. The method according to claim 4, characterized in that After the step of controlling the liquid hydrogen leakage treatment system on the detection device to perform efficient treatment according to the center straight line distance and the horizontal angle α, the method further includes: Obtaining a horizontal distance L from the leakage source device to the detection device, wherein the horizontal distance L=d×cosα; The height h between the leakage source device and the liquid hydrogen leakage gas mass is determined, where the height h=d×sinα.

6. The method according to claim 1, characterized in that After the step of controlling the dedicated disposal system for liquid hydrogen leakage to efficiently dispose of the leakage area, the method further includes: After the set time, the target characteristic cloud image of the liquid hydrogen leakage gas mass is collected again by the low-light video acquisition device; If the cloud feature information corresponding to the target feature cloud image is smaller than the preset feature, it is determined that the leakage processing has met the standard.

7. A server, characterized in that: The server includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the server to execute the method according to any one of claims 1 to 6.

8. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a server, the server is caused to execute the method according to any one of claims 1 to 6.

9. A computer program product, characterized in that When the computer program product is run on a server, the server is caused to perform the method according to any one of claims 1 to 6.

Images