Spray simulation method and device for augmented reality and social distance keeping method
Through augmented reality technology, simulate droplet effects on AR devices and track head posture in real time, solving the problem that existing equipment is difficult to encourage users to maintain social distance, achieving natural and effective reminder effects, and reducing costs.
Patent Information
- Application Number
- CN202410927203.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-08
AI Technical Summary
Existing equipment is difficult to effectively encourage users to actively avoid others and maintain social distance, and the commonly used reminder methods are unpopular and costly.
Using augmented reality technology, people's head images in the environment are collected through AR devices, and a three-dimensional rendering engine is used to simulate droplet effects, track head postures in real time, display breathing effects, and remind users to maintain social distance.
Through natural and visual reminders, help users develop the habit of maintaining social distance, reduce costs and improve the effectiveness of reminders.
Smart Images

Figure CN118918292A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of augmented reality (AR) and virus protection technology, and relates to an augmented reality droplet simulation method, device and social distance keeping method. Background Art
[0002] Humans naturally need to communicate with each other in the real world, and people will unconsciously forget social distance in daily communication because it is counterintuitive to always maintain a distance of about 1.8 meters during social activities. Even if people realize the importance of social distance in slowing the spread of the virus and reducing the possibility of infection, it is still difficult to maintain a certain social distance through strong self-awareness. Especially during an infectious disease outbreak, the importance of maintaining social distance becomes increasingly apparent, and an effective and natural way to maintain social distance is needed.
[0003] The current common method is to regularly remind people in public places to maintain social distance by repeatedly playing recorded reminders through social workers, volunteers or even loudspeakers. Many researchers have been working on inventing devices that encourage people to maintain social distance. Combining the latest advances in robotics and computer vision, researchers have even developed robots that move around the streets and remind people to maintain social distance. For example, a fully automatic surveillance robot based on a four-legged platform, equipped with multiple cameras and three-dimensional light detection and ranging sensors, has been developed. The robot can move freely in dynamic scenes and issue friendly voice prompts to advise crowded people to disperse. Others have developed wearable smart devices that alert users when someone is detected within 6 feet of the user. These devices have been tested and found to be useful in maintaining social distance.
[0004] Most of these devices use sensors to detect distance and provide warnings to users, but there is relatively little research on the technology needed to encourage users to actively avoid others. Generally speaking, these reminder methods are not popular with people because people tend to feel forced to obey others' orders, and the reminder effect is not ideal. In addition, it is costly to hire workers or robots to repeat the reminder task regularly. Summary of the invention
[0005] Technical purpose: In response to the above technical problems, the present invention proposes an augmented reality droplet simulation method, device and social distance keeping method, which stimulate users to maintain an appropriate social distance from others through visually prominent breathing effects, help users develop the habit of keeping a certain distance from others, and even if they are not wearing a device, users can remember to keep a distance.
[0006] Technical solution: To achieve the above technical objectives, the present invention adopts the following technical solution:
[0007] A droplet simulation method for augmented reality, comprising the steps of:
[0008] S1. Using a camera of an AR device worn by a user, collecting head images of one or more other persons in the environment where the user is located;
[0009] S2, processing the collected head image to obtain the processing result of the head image, including the head posture and the face position and orientation;
[0010] S3. The three-dimensional rendering engine renders a droplet special effect according to the processing result of the head image, aligns the droplet special effect with the head image collected by the AR device, forms an AR view, and displays it on the display of the AR device.
[0011] Preferably, in step S2, the collected head image is processed in the following manner:
[0012] Use face recognition technology to identify and locate the collected head images and mark the face area;
[0013] The marked face area is processed using a data-driven regression method to obtain the three-dimensional pose of the head and determine the position of the head in the pixel coordinate system and the orientation of the face.
[0014] Preferably, in step S3, the three-dimensional rendering engine simulates the droplet particles in the following manner:
[0015] Pre-calculate the motion field with physical significance, that is, solve the Navier-Stokes equations on the three-dimensional grid, and pre-calculate the dynamic velocity field of the fluid at different initial velocities v0 when it moves at position p at time t
[0016] The precomputed dynamic velocity field Rotate and translate to face position (o i ,R i ) and transform the particle system using the translated dynamic velocity field;
[0017] The particle system randomly generates particles, sets the initial parameters of the particles, including initial position, initial velocity, initial lifespan, initial size and initial opacity, and modifies the parameters of each particle in the next frame of the image in chronological order.
[0018] Preferably, the particle system sets the initial parameters of the particles and modifies the parameters of each generated particle in the following manner:
[0019] Particle speed: If the particle is generated at position p0 at time t0, then its speed is corrected to And move to the new position at this speed in the time step δt The particle system queries the dynamic velocity field after translation and obtains the value of the velocity field of the particle at p1 Then update the velocity of the particle at the next moment, that is, the velocity of the particle in the next frame of image is
[0020] Particle lifespan: The lifespan of each particle is set to t=1 when it is generated. The particle's current speed v and the speed change Δv of the previous frame are calculated. Check whether the particle's current speed value |v| is less than the set speed threshold, and the particle's speed change |Δv| is less than the set speed change threshold. If so, the particle's lifespan is set to decay according to the proportional coefficient β. If not, the particle's speed continues to be updated.
[0021] Particle size and opacity: The particle size and opacity are set to change with the life of the particle, size s = ts0, opacity o = to0, where s0 and o0 are the set initial size and initial opacity, and the size and opacity of each particle are perturbed on the initial size and initial opacity according to Gaussian distribution.
[0022] An augmented reality droplet simulation device, comprising:
[0023] An AR device is provided with a camera and a head mounted display, wherein the camera is used to capture head images of one or more other persons in the environment where the user is located, and the head mounted display is used to display an AR view;
[0024] The data processing module is used to receive the head image uploaded by the AR device, process the collected head image, and obtain the processing result of the head image, including the head posture and the face position and orientation;
[0025] The three-dimensional rendering engine is used to render the droplet special effect according to the processing result of the head image, and align the droplet special effect with the head image collected by the AR device to form an AR view.
[0026] Preferably, the AR device and the three-dimensional rendering engine are implemented using a Hololens 2 device;
[0027] The data processing module is implemented by setting up a background computer, and the background computer is wirelessly connected to the Hololens 2 device.
[0028] Preferably, the data processing module includes:
[0029] A face recognition unit is used to identify and locate the collected head image and mark the face area;
[0030] The convolutional neural network uses a regression-based method to process the marked face area, obtain the three-dimensional pose of the head, and determine the position of the head in the pixel coordinate system and the position and orientation of the face.
[0031] Preferably, the three-dimensional rendering engine comprises:
[0032] A data receiving module, used to receive the head image sent by the AR device and the processing result of the head image sent by the data processing module;
[0033] The motion field pre-calculation module is used to pre-calculate the motion field with physical significance, that is, to solve the Navier-Stokes equations on the three-dimensional grid and pre-calculate the dynamic velocity field of the fluid at different initial velocities v0 when it moves at position p at time t.
[0034] Conversion module, used to convert pre-computed dynamic velocity fields Rotate and translate to face position (o i ,R i ) and transform the particle system using the translated dynamic velocity field;
[0035] The particle system is used to randomly generate particles, set the initial parameters of the particles, including the initial position, initial velocity, initial life, initial size and initial opacity, and modify the parameters of each particle in the next frame of the image in chronological order;
[0036] The rendering module is used to align the droplet special effects generated by the particle system with the head image collected by the AR device to form an AR view.
[0037] A method for maintaining social distance, wherein a user wears the augmented reality droplet simulation device and adjusts the social distance with others according to the direction and range of the droplet special effects display.
[0038] Beneficial effects: Due to the adoption of the above technical solution, the present invention has the following beneficial effects:
[0039] The present invention designs an augmented reality prototype that combines real-time head pose tracking with a particle system, automatically aligns rendering special effects with objects in the real world through AR devices such as AR glasses, highlights the visualization of respiratory droplets to enhance daily conversation scenarios, and reminds people to effectively maintain social distance in daily life. The effectiveness of the architecture is demonstrated by conducting user studies, and quantitative and qualitative comparisons can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 A flowchart of an augmented reality droplet simulation method proposed by the present invention;
[0041] Figure 2 A schematic diagram of the simulation of droplets in a scene where a person is speaking according to the present invention;
[0042] Figure 3 It is a schematic diagram of the present invention applied to droplet simulation in various scenarios;
[0043] Figure 4 This is a schematic diagram of the social distance maintenance method based on the droplet simulation proposed in the present invention. DETAILED DESCRIPTION
[0044] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0045] Embodiment 1
[0046] like Figure 1 As shown, this embodiment proposes an augmented reality droplet simulation method, comprising the steps of:
[0047] Using a camera of an AR device worn by a user, a head image of one or more other persons in the environment where the user is located is collected;
[0048] Processing the collected head image to obtain the processing result of the head image, including the head posture, the position of the head in the pixel coordinate system, and the position and orientation of the face;
[0049] The three-dimensional rendering engine renders the droplet special effects according to the processing result of the head image and displays it on the display of the AR device.
[0050] The droplet simulation method of augmented reality proposed in this embodiment uses an AR-based interface to display virtual effects, by identifying individuals in a public space, tracking their head postures to simulate the breathing effect well, aligning the virtual effect with each individual and rendering it using a head-mounted display (HMD) to highlight the breathing gas in the real world. All the calculations involved are completed in real time, and the augmentation of the real world will be natural for each subject.
[0051] Figure 2 The process of the augmented reality droplet simulation method is shown:
[0052] (1) The camera on the AR head-mounted display captures the image and uploads it;
[0053] (2) Use face recognition technology to locate the face in the collected image, identify and mark the pixel position of the face area;
[0054] (3) Return the position of the detected face to the ResNet50 network, and regress the three posture angles of the annotated area to regress the posture angle of the head;
[0055] (4) The detected face and its orientation are rendered using a 3D rendering engine, and the user can see the displayed virtual effects by wearing an AR headset.
[0056] In step (1), the camera view of the Hololens 2 device is used as the visual input. Finally, the collected information is rendered as special effects by the 3D rendering engine integrated in the Hololens 2 device and displayed on the Hololens 2 device, creating an augmented reality (AR) effect that is continuously updated in the camera video stream. Throughout the implementation process, Hololens is used as an image acquisition and display device to render AR effects. The computing task is performed on a computer connected to Hololens2. In addition to processing the collected data, the computer also needs to calculate other parameters related to AR display. The computing task is performed on the connected computer after Hololens transmits data to it. This approach is necessary due to the limited computing power of Hololens.
[0057] The key technologies involved in the method of the present invention are described in detail below.
[0058] 1. Head tracking
[0059] Identifying each object in the field of view of the AR headset and tracking their head pose is accomplished using advanced learning-based computer vision techniques. The proposed method faces several challenging technical issues, including tracking the pose of the head in extreme orientations or when it is occluded. It also involves handling situations when there are multiple people present in the scene.
[0060] In order to achieve robust positioning and tracking of head pose, the present invention designs a face tracking step and a head pose estimation step. For the face tracking step, the present invention uses a face recognition library [1] (Geitgey A (2022) face recognition), which is a fast pre-trained deep learning model. The present invention uses the tracked face to clip the original image and uses the cropped area to solve the head pose. For head pose estimation, the traditional method usually detects facial landmarks first and then uses the PnP method to recover the 3D pose from the 2D information. However, when a small number of feature points are found in the view, the stability of this method is poor. Therefore, the present invention uses this deep learning-based method to directly regress the head pose (including the three-dimensional position and three-dimensional orientation in the camera coordinate system). Finally, the present invention uses a model trained on the 300W-LP dataset to solve the head pose [2] (Zhu X, Lei Z, Liu X, et al (2016) Face alignment across large poses: A 3d solution.). That is, for the multiple face sub-images selected in the face tracking step, the pre-trained convolutional neural network is used to regress the facial pose information of different individuals.
[0061] 2. AR Effects
[0062] After all the head pose information is obtained, the virtual effects are prepared to be aligned with the images captured from the camera on the HMD to form an AR view. That is, after collecting the head pose data, it is sent to the 3D rendering engine together with the position of the head in the pixel coordinate system and the 3D orientation estimated by the learning-based algorithm, and the particle system is transformed using the obtained head pose. If multiple objects are detected, the particle system is replicated and transformed according to each head pose to generate the AR effect. The visual effect is refreshed as the detected head pose is updated.
[0063] The present invention simulates respiratory droplets by using a particle system prepared in a three-dimensional rendering engine. Higher quality breathing effects based on (for example, fluid simulation) can also be used with current graphics tools. However, in order to maintain high operating performance, the present invention weighs the quality of dynamic effects and uses a particle system to highlight respiratory droplets at appropriate calculation and rendering costs. The present invention preferably uses Microsoft's Hololens2 as an AR device, and Hololens2 is provided with a camera three-dimensional rendering engine, a communication module, and a head display HMD, etc. After all programs are deployed on the AR device, the user can see the visualization effect when wearing the HMD. If there are multiple faces in the scene, the same method is used to generate and render droplet special effects for different faces, and they are displayed on the screen at the same time. Figure 3Shows the AR effects of different scenarios: (a), a person talking; (b), two friends talking.
[0064] 3. Data-driven augmented reality droplet particle system visualization module
[0065] Through the above computer processing, the face position orientation (o i ,R i ) information.
[0066] In addition to the face position and orientation tracked by the augmented reality system (o i ,R i ), the droplet visualization effect depends on a large number of adjustable parameters in the particle system, including particle system speed m, particle size s, particle generation time interval, particle life cycle T, particle opacity, etc. Users can adjust according to their needs to further control the movement effect and intensity of virtual droplets, and use droplet special effects to remind and control the degree of social distance.
[0067] The properties of each randomly generated droplet particle include:
[0068] The emission source center p0 is set to o i +∈, where ∈ is a random offset that follows a Gaussian distribution;
[0069] The emission speed v = md, d is the direction vector, set to the direction d facing the detected face i A random vector in a conical area with a center and an angle of θ, whose size is m;
[0070] The initial droplet special effect motion has high running efficiency, but due to the lack of physical model, its motion realism is poor. In order to enhance the authenticity of the droplet visualization effect, the present invention performs data-driven physical effect enhancement on the droplet particle system and corrects the current motion speed of each particle in each time frame.
[0071] The present invention adopts a data-driven method to correct the particle motion and life state by precalculating a physically meaningful motion field. The specific scheme is as follows:
[0072] 1) Pre-calculation of the velocity field of the physical droplet motion effect, solving the Navier-Stokes equation on the three-dimensional grid, and obtaining the dynamic velocity field of the fluid motion at different initial velocities v0 at time t and position p
[0073] 2) Rotate and translate the pre-calculated dynamic velocity field to the face position (o i ,R i ) In the local coordinate system defined by , each particle randomly generated by the particle system moves under the calculated dynamic velocity field.
[0074] If a particle is generated at p0 at time t0, its velocity is corrected to And move to the new position at this speed in the time step δt The system will continue to query the dynamic velocity field to obtain the value of the velocity field at p1, and then update the velocity of the particle at the next moment;
[0075] 3) A dynamic velocity field with physical meaning will decay to 0 after a certain period of time. Setting the particle life cycle to a form related to the velocity will be more physically meaningful than manually setting the life cycle. First, set the life cycle of each particle to t=1 when it is generated. Calculate the velocity v of the particle movement and the change Δv from the previous frame at each step. Set a threshold to check if the velocity value |v| is not large and the velocity change |Δv| is not large, indicating that the particle has begun to remain stationary. For example, |v| is less than 0.1cm / s, |Δv| is less than 0.05cm^2 / s, and set the particle life cycle to decay according to the proportional coefficient β.
[0076] 4) When drawing, the particle size and particle opacity are set to change with the particle lifespan, size s = ts0, opacity o = to0, where s0 and o0 are the initial set particle size and opacity. To increase the randomness of the droplet particle system, the size and opacity of each particle will be perturbed according to the Gaussian distribution on the preset s0 and o0 values to further improve the randomness of the particle system.
[0077] Compared with the traditional particle system in which the movement speed of each droplet particle is constant, the movement speed field of the droplet particles in the present invention is obtained by solving the fluid motion equation, which will have a more realistic physical meaning; at the same time, the position, size, opacity and other properties of each particle in the particle system are variable, which retains the randomness of the particle system; in addition, this scheme avoids the computational overhead of real-time physical solution of the augmented reality system by pre-calculating the physical velocity field, and can maintain the efficiency of presenting the virtual animation effect of the droplets.
[0078] 5. Other implementation details
[0079] Considering the limited computing resources and requirements of AR devices, in terms of real-time augmented reality effects, the present invention does not use marker recognition when tracking the position of human heads. Instead, the present invention only identifies the position information of all heads in the field of view, without providing the identity of each person. This method may result in the inability to correspond to the position changes of different individuals between different frames. To solve this problem, the present invention allows Unity to process as many frames as possible per unit time and only generates special effects for the current frame. When the frame rate reaches a certain value, the effect can, to a certain extent, give a person's visual system a coherent sense of position.
[0080] Embodiment 2
[0081] This embodiment proposes a method based on the first embodiment to achieve a method of maintaining social distance, using an AR interface to help maintain social distance. In the proof-of-concept demonstration, a user wearing AR glasses is aware of the visualized respiratory droplets when communicating with a friend. Humans are sensitive to secretions or expired gases produced by other objects, so they will consider maintaining an appropriate social distance during a conversation, such as Figure 4 shown.
[0082] Embodiment 3
[0083] This embodiment provides an augmented reality droplet simulation device, including:
[0084] An AR device is provided with a camera and a head mounted display, wherein the camera is used to capture head images of one or more other persons in the environment where the user is located, and the head mounted display is used to display an AR view;
[0085] The data processing module is used to receive the head image uploaded by the AR device, process the collected head image, and obtain the processing result of the head image, including the head posture and the face position and orientation;
[0086] The three-dimensional rendering engine is used to render the droplet special effect according to the processing result of the head image, and align the droplet special effect with the head image collected by the AR device to form an AR view.
[0087] Specifically, the AR device and the 3D rendering engine are implemented using a Hololens 2 device. The data processing module is implemented by setting up a background computer, and the background computer is wirelessly connected to the Hololens 2 device.
[0088] The data processing module includes:
[0089] A face recognition unit is used to identify and locate the collected head image and mark the face area;
[0090] The convolutional neural network uses a regression-based method to process the marked face area, obtain the three-dimensional pose of the head, and determine the position of the head in the pixel coordinate system and the position and orientation of the face.
[0091] The 3D rendering engine includes:
[0092] A data receiving module, used to receive the head image sent by the AR device and the processing result of the head image sent by the data processing module;
[0093] The motion field pre-calculation module is used to pre-calculate the motion field with physical significance, that is, to solve the Navier-Stokes equations on the three-dimensional grid and pre-calculate the dynamic velocity field of the fluid at different initial velocities v0 when it moves at position p at time t.
[0094] Conversion module, used to convert pre-computed dynamic velocity fields Rotate and translate to face position (o i ,R i ) and transform the particle system using the translated dynamic velocity field;
[0095] The particle system is used to randomly generate particles, set the initial parameters of the particles, including the initial position, initial velocity, initial life, initial size and initial opacity, and modify the parameters of each particle in the next frame of the image in chronological order;
[0096] The rendering module is used to align the droplet special effects generated by the particle system with the head image collected by the AR device to form an AR view.
[0097] The above description is only a preferred specific implementation manner of the present invention. It should be pointed out that the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A droplet simulation method for augmented reality, characterized in that: Includes steps: S1. Using a camera of an AR device worn by a user, collecting head images of one or more other persons in the environment where the user is located; S2, processing the collected head image to obtain the processing result of the head image, including the head posture and the face position and orientation; S3. The three-dimensional rendering engine renders a droplet special effect according to the processing result of the head image, aligns the droplet special effect with the head image collected by the AR device, forms an AR view, and displays it on the display of the AR device.
2. The method for simulating droplets of augmented reality according to claim 1, characterized in that: In step S2, the collected head image is processed in the following manner: Use face recognition technology to identify and locate the collected head images and mark the face area; The marked face area is processed using a data-driven regression method to obtain the three-dimensional pose of the head and determine the position of the head in the pixel coordinate system and the orientation of the face.
3. The method for simulating droplets of augmented reality according to claim 1, characterized in that: In step S3, the three-dimensional rendering engine simulates the droplet particles in the following manner: Pre-calculate the motion field with physical significance, that is, solve the Navier-Stokes equations on the three-dimensional grid, and pre-calculate the dynamic velocity field of the fluid at different initial velocities v0 when it moves at position p at time t The precomputed dynamic velocity field Move to the face position (o i ,R i ) and transform the particle system using the translated dynamic velocity field; The particle system randomly generates particles, sets the initial parameters of the particles, including initial position, initial velocity, initial lifespan, initial size and initial opacity, and modifies the parameters of each particle in the next frame of the image in chronological order.
4. The method for simulating droplets of augmented reality according to claim 3, characterized in that: The particle system sets the initial parameters of the particles and modifies the parameters of each generated particle in the following way: Particle speed: If the particle is generated at position p0 at time t0, then its speed is corrected to And move to the new position at this speed in the time step δt The particle system queries the dynamic velocity field after translation and obtains the value of the velocity field of the particle at p1 Then update the velocity of the particle at the next moment, that is, the velocity of the particle in the next frame of image is Particle lifespan: The lifespan of each particle is set to t=1 when it is generated. The particle's current speed v and the speed change Δv of the previous frame are calculated. Check whether the particle's current speed value |v| is less than the set speed threshold, and the particle's speed change |Δv| is less than the set speed change threshold. If so, the particle's lifespan is set to decay according to the proportional coefficient β. If not, the particle's speed continues to be updated. Particle size and opacity: The particle size and opacity are set to change with the life of the particle, size s = ts0, opacity o = to0, where s0 and o0 are the set initial size and initial opacity, and the size and opacity of each particle are perturbed on the initial size and initial opacity according to Gaussian distribution.
5. An augmented reality droplet simulation device, characterized in that: include: An AR device is provided with a camera and a head mounted display, wherein the camera is used to capture head images of one or more other persons in the environment where the user is located, and the head mounted display is used to display an AR view; The data processing module is used to receive the head image uploaded by the AR device, process the collected head image, and obtain the processing result of the head image, including the head posture and the face position and orientation; The three-dimensional rendering engine is used to render the droplet special effect according to the processing result of the head image, and align the droplet special effect with the head image collected by the AR device to form an AR view.
6. The augmented reality droplet simulation device according to claim 5, characterized in that: The AR device and the three-dimensional rendering engine are implemented using a Hololens 2 device; The data processing module is implemented by setting up a background computer, and the background computer is wirelessly connected to the Hololens 2 device.
7. The augmented reality droplet simulation device according to claim 5, characterized in that: The data processing module comprises: A face recognition unit is used to identify and locate the collected head image and mark the face area; The convolutional neural network uses a regression-based method to process the marked face area, obtain the three-dimensional pose of the head, and determine the position of the head in the pixel coordinate system and the position and orientation of the face.
8. The augmented reality droplet simulation device according to claim 5, characterized in that: The three-dimensional rendering engine comprises: A data receiving module, used to receive the head image sent by the AR device and the processing result of the head image sent by the data processing module; The motion field pre-calculation module is used to pre-calculate the motion field with physical significance, that is, to solve the Navier-Stokes equations on the three-dimensional grid and pre-calculate the dynamic velocity field of the fluid at different initial velocities v0 when it moves at position p at time t. Conversion module, used to convert pre-computed dynamic velocity fields Rotate and translate to face position (o i ,R i ) and transform the particle system using the translated dynamic velocity field; The particle system is used to randomly generate particles, set the initial parameters of the particles, including the initial position, initial velocity, initial life, initial size and initial opacity, and modify the parameters of each particle in the next frame of the image in chronological order; The rendering module is used to align the droplet special effects generated by the particle system with the head image collected by the AR device to form an AR view.
9. A method for maintaining social distance, characterized in that: A user wears the augmented reality droplet simulation device described in any one of claims 5-8 and adjusts the social distance with others according to the direction and range of the droplet special effects displayed.
Citation Information
Patent Citations
Method and apparatus for stratifying respiratory infected patients
US20240118191A1
Augmented reality-based firefighter training system and method
WO2003015057A1
Cited By
Image processing method, system, device and computer program product
CN119832133A
