Windshield attachment processing device and method based on visual neural network

Abstract

The invention provides a windshield attachment processing device and method based on an optic neural network, and relates to the technical field of intelligent automobile safety. According to the scheme, attachments on the windshield are detected in real time through a vehicle-mounted camera and a pre-trained visual neural network model, after soft sheltering objects such as plastic bags are recognized, the boundary, the gravity center, the area and the adhesion strength of the soft sheltering objects are analyzed, and the optimal blowing angle and force are calculated; the control system drives the air pump device near the windscreen wiper on the main driving side to stretch out and draw back and adjust the angle, airflow impact is applied according to needs, and accurate removal is achieved. And if the result fails, iteratively optimizing the parameters and re-blowing until the operation is successful or the maximum number of attempts is reached. The method does not need manual intervention, effectively solves the problem that a traditional windscreen wiper cannot remove soft attachments, remarkably improves the driving safety, and is particularly suitable for high-speed driving scenes.

Description

Technical Field

[0001] This application relates to the field of intelligent vehicle safety technology, specifically to a windshield adhesion treatment device and method based on visual neural networks. Background Technology

[0002] During driving, especially at high speeds, soft objects such as plastic bags and paper can suddenly adhere to the windshield, severely affecting the driver's vision and posing a significant safety hazard. Traditional vehicles are generally equipped with windshield wipers to remove surface contaminants such as rainwater and dust. Their working principle involves the reciprocating motion of a rubber blade across the glass surface. However, windshield wipers are extremely ineffective at removing large areas of lightweight, tightly adhered soft objects (such as plastic bags). In fact, the wiping motion may even cause the adhered area to expand or generate vibration and noise, failing to effectively solve the problem.

[0003] Current technologies lack automatic identification and removal mechanisms for such sudden soft obstructions. When a plastic bag covers the windshield while driving, drivers often need to slow down or even stop to manually remove it, which not only affects traffic efficiency but also significantly increases the risk of accidents at high speeds. Therefore, there is an urgent need for a technological solution that can intelligently identify and proactively remove soft deposits from the windshield to improve driving safety and automation. Summary of the Invention

[0004] This application provides a windshield attachment removal device and method based on visual neural networks, which can solve the technical problem that soft attachments such as plastic bags obstruct the windshield during high-speed vehicle operation, resulting in obstructed vision, and that traditional wipers cannot effectively remove them.

[0005] To achieve the above objectives, this application provides the following technical solution:

[0006] The first aspect of this application provides a windshield adhesion processing device based on a visual neural network, including a visual detection subsystem, an adhesion state analysis subsystem, a parameter determination subsystem, and a control subsystem installed in an in-vehicle system;

[0007] When the vision inspection subsystem detects an attachment on the windshield, it triggers the activation of the attachment status analysis subsystem. This subsystem receives image data of the attachment transmitted from the vision inspection subsystem and calculates the attachment's boundary information, adhesion strength parameters, center of gravity data, and area data. The parameter determination subsystem, based on the boundary information, adhesion strength parameters, center of gravity data, and area data calculated by the attachment status analysis subsystem, determines the blowing angle parameters and blowing force parameters. The control subsystem receives the blowing angle parameters and blowing force parameters and controls the air pump device located near the driver's side windshield wiper for adjustment.

[0008] The visual detection subsystem is used to capture image frames of the windshield area again after the blowing operation is completed, and analyze and determine whether the adhering object has been successfully removed. If the detection result shows that the adhering object has been removed, the current removal operation is terminated. If the adhering object has not been removed, the system returns to the adhering state analysis step, recalculates the blowing parameters and performs the blowing operation until the adhering object is removed or the preset number of attempts is reached, at which point the operation is terminated and a prompt signal can be issued.

[0009] In one optional embodiment, the parameter determination subsystem includes: a first parameter determination module, used to determine the coverage area of ​​the attachment based on boundary information;

[0010] The second parameter determination module is used to determine the degree of adhesion between the adhering substance and the windshield based on the adhesion strength parameter.

[0011] The third parameter determination module is used to locate the core area of ​​the attachment based on the center of gravity data;

[0012] The fourth parameter determination module is used to quantify the size of the attachment based on the area data.

[0013] In one optional embodiment, the parameter determination subsystem includes: a first calculation module for calculating the blowing angle parameter based on the center of gravity data and boundary information of the attachment.

[0014] In an optional embodiment, the parameter determination subsystem further includes a second calculation module, used to calculate the blowing force parameter by means of the ratio between the area data of the attachment and the length of the attachment boundary.

[0015] In one optional embodiment, the control subsystem includes a telescopic mechanism and an angle adjustment mechanism; the telescopic mechanism is used to adjust the extension length of the air pump body; the angle adjustment mechanism is used to adjust the blowing direction of the air pump body so that the air pump's blowing port is aligned with the core area of ​​the attached material; the output pressure of the air pump is adjusted according to the blowing force parameters so that the airflow intensity meets the removal requirements.

[0016] In one alternative embodiment, the air pump device is used to perform a blowing operation on the attachment on the windshield according to the adjusted position, angle and pressure parameters, using the impact force of the airflow to detach the attachment from the windshield.

[0017] In one alternative embodiment, the visual inspection subsystem includes a camera and a pre-trained visual neural network model; the camera is used to capture image frames of the windshield area in real time; the visual neural network model performs cyclic analysis of the image frames to identify the presence of attachments.

[0018] The second aspect of this application provides a method for removing windshield deposits based on a visual neural network. The method includes: cyclically detecting the windshield area using a pre-trained visual neural network model to identify the presence of deposits; when deposits are detected, calculating the boundary information, adhesion strength parameters, center of gravity data, and area data of the deposits, and determining the blowing angle parameters and blowing force parameters; adjusting the position and angle of the air pump device according to the blowing angle parameters, adjusting the air pump pressure according to the blowing force parameters, and performing the blowing operation according to the adjusted parameters; detecting again whether the deposits on the windshield have been removed; if not removed, recalculating the parameters and performing the blowing operation until the deposits are removed or a preset number of attempts is reached, then terminating the operation.

[0019] A third aspect of this application provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the steps of a windshield adhesion treatment method based on a visual neural network according to the second aspect.

[0020] The fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the windshield adhesion processing method based on a visual neural network of the second aspect.

[0021] The present invention achieves the following beneficial effects through the above technical solution:

[0022] This application provides a windshield debris removal device and method based on visual neural networks. By setting up a visual detection module combined with a deep learning model, it achieves high-precision identification and positioning of debris on the windshield, solving the problem that traditional wipers cannot remove soft obstructions such as plastic bags. By analyzing the boundary, center of gravity, area, and adhesion strength of the debris, the optimal blowing strategy is dynamically generated and precisely executed by a retractable and angle-adjustable air pump device, improving the removal success rate. The closed-loop feedback mechanism supports multiple adaptive retries, enhancing the system's robustness. The overall solution operates fully automatically, responds rapidly, and significantly improves active safety performance at high speeds. Attached Figure Description

[0023] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0024] Figure 1This is a flowchart of a windshield adhesion treatment method based on a visual neural network provided by the present invention;

[0025] Figure 2 This is a schematic diagram of the windshield adhesion treatment device based on visual neural network provided by the present invention;

[0026] Figure 3 This is an internal structural diagram of the electronic device provided by the present invention. Detailed Implementation

[0027] The technical solutions in the embodiments of this invention / invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this invention / invention, and not all embodiments. Based on the embodiments of this invention / invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention / invention.

[0028] Example 1: During high-speed vehicle operation, soft or semi-solid foreign objects such as plastic bags, leaves, and insect remains may suddenly adhere to the windshield. These objects, due to their flexibility, strong adhesion, and large contact area with the glass surface, are often difficult to remove effectively with traditional windshield wipers. Especially when these objects are located in the driver's primary field of vision, they can severely obstruct vision, creating a significant safety hazard. Current technologies lack automatic identification and efficient removal methods for such sudden, non-rigid attachments, forcing drivers to manually clean them or stop the vehicle during driving, affecting driving continuity and increasing the risk of accidents. Therefore, there is an urgent need for a technical solution that can sense, intelligently analyze, and proactively remove various attachments from the windshield in real time to improve driving safety and automation.

[0029] This embodiment proposes the following components: a vision detection subsystem 210, an attachment state analysis subsystem 220, a parameter determination subsystem 230, and a control subsystem 240, all integrated into the vehicle-mounted system. Figure 2 ;

[0030] When the visual inspection subsystem detects an attachment on the windshield, it triggers the attachment status analysis subsystem to start.

[0031] The attachment status analysis subsystem is used to receive the attachment image data transmitted by the visual inspection subsystem and calculate the boundary information, attachment strength parameters, center of gravity data and area data of the attachment.

[0032] The parameter determination subsystem determines the blowing angle parameters and blowing force parameters based on the boundary information, adhesion strength parameters, center of gravity data and area data calculated by the adhesion state analysis subsystem.

[0033] The control subsystem receives blowing angle parameters and blowing force parameters, and controls the air pump device located near the windshield wiper on the driver's side of the car to make adjustments.

[0034] The air pump device blows air onto the debris on the windshield according to the adjusted position, angle and pressure parameters, using the impact force of the airflow to detach the debris from the windshield.

[0035] The visual inspection subsystem is used to recapture image frames of the windshield area after the blowing operation is completed, and analyze and determine whether the adhering objects have been successfully removed. If the detection result shows that the adhering objects have been removed, the current removal operation is terminated. If the adhering objects have not been removed, the system returns to the adhering status analysis step, recalculates the blowing parameters and performs the blowing operation until the adhering objects are removed or the preset number of attempts is reached, at which point the operation is terminated and a prompt signal can be issued.

[0036] This embodiment provides a windshield debris removal device based on a visual neural network. Its overall architecture revolves around a closed-loop control logic of "perception—analysis—decision—execution—feedback," aiming to achieve fully automatic, high-precision, and adaptive removal of debris from windshields. The device integrates an onboard visual detection module, a debris status analysis subsystem, a parameter determination subsystem, a control subsystem, and an air pump device to form a complete intelligent removal system. Its core lies in combining deep learning technology with pneumatic actuators, using a visual neural network to accurately identify and extract features of the debris, dynamically generating an optimal blowing strategy, and finally completing the physical removal action through adjustable airflow output, significantly improving removal efficiency and system robustness.

[0037] The visual detection module, installed in the vehicle system, continuously monitors changes in the windshield area. This module includes one or more high-resolution cameras, optionally mounted above the inner side of the driver's side windshield to ensure coverage of the main driver's field of vision. The cameras acquire video frame data in real time and transmit it to a pre-trained visual neural network model for iterative analysis. This model can employ deep learning architectures with instance segmentation capabilities, such as Mask R-CNN, YOLOv5-seg, or U-Net, to accurately identify the presence and outline of attached objects under complex lighting conditions and motion blur backgrounds. Once an abnormal object is detected on the windshield, subsequent processing is triggered, activating the attachment status analysis subsystem.

[0038] After receiving image data from the visual inspection subsystem, the adhesion state analysis subsystem further performs refined data analysis tasks. The calculated boundary information defines the spatial distribution range of the adherend on the windshield, typically expressed as its circumscribed rectangle or polygonal outline in pixel coordinates. The adhesion strength parameter reflects the degree of adhesion between the adherend and the glass, and can be estimated through image texture analysis, edge sharpness assessment, and local contrast changes, or by establishing a mapping relationship using historical experimental data. The centroid data is calculated using an image centroid algorithm, representing the mass concentration point of the adherend and serving as a crucial basis for determining the blowing point. The area data is obtained by statistically analyzing the number of effective pixels in the segmentation results, quantifying the overall size of the adherend. These four types of parameters collectively constitute the basic input for subsequent blowing strategy formulation.

[0039] The parameter determination subsystem, based on the boundary information, adhesion strength parameters, center of gravity data, and area data output by the adhesion state analysis subsystem, comprehensively models and determines the optimal blowing angle and blowing force parameters. The blowing angle parameter is mainly calculated based on the spatial relative position of the center of gravity of the adherend and the air pump device, with the goal of ensuring that the airflow direction is directly facing the core area of ​​the adherend, thereby maximizing the peeling effect. The blowing force parameter considers the proportional relationship between the area of ​​the adherend and its boundary adhesion length, and introduces empirical coefficients for weighted correction to ensure that the applied airflow energy effectively disrupts the adhesion balance without damaging the glass or causing secondary splashing due to excessive impact.

[0040] The control subsystem receives the blowing angle and blowing force parameters from the parameter determination subsystem and converts them into specific mechanical control commands to adjust the air pump unit located near the driver's side windshield wiper. This subsystem includes a signal processing unit, a drive circuit module, and a communication interface, ensuring compatibility with the vehicle's CAN bus or other vehicle network protocols, enabling information exchange with other electronic control units (ECUs). The control subsystem adjusts the spatial orientation of the air pump unit based on the blowing angle parameters, including its extension stroke and nozzle direction; simultaneously, it sets the air pump's operating pressure value based on the blowing force parameters, controlling the flow rate and speed of compressed air output.

[0041] The air pump device is a miniaturized, high-pressure differential pneumatic actuator, fixedly installed near the wiper arm base on the driver's side, making good use of existing structural space and not affecting the appearance design. This device features an adjustable mechanical structure, allowing it to change its position extension and nozzle pointing angle under the command of the control subsystem, thereby achieving precise alignment with objects in different locations. During operation, the air pump obtains compressed gas from the vehicle's air tank or an electric air compressor, releasing a high-speed airflow under set pressure conditions. This airflow acts directly on the surface of the object, using the shearing and impact forces generated by the airflow's kinetic energy to detach it from the glass surface.

[0042] After a complete cleaning process is completed, the visual inspection subsystem will restart the image acquisition function to obtain a new frame image of the current windshield area to evaluate the actual cleaning effect of the previous air blowing operation. If the analysis results show that the original deposits are no longer present or the residual area is below the threshold, the cleaning is considered successful and the entire process ends; otherwise, if significant residue is still detected, the system automatically returns to the deposit status analysis step, re-acquires the latest deposit status data, updates the air blowing parameters, and executes a new round of air blowing. This process can be repeated multiple times, up to a preset number of attempts (e.g., 3-5 times), to prevent ineffective loops from wasting resources. If the maximum number of attempts is reached and the deposits are still not cleared, the system can generate a warning signal and notify the driver via the instrument panel or voice prompt, suggesting other manual intervention measures.

[0043] The various components are interconnected and coordinated for data sharing and control via an in-vehicle communication network. The visual inspection module, as the front-end perception unit, provides the raw input to the entire system; the attachment state analysis subsystem, as the intermediate processing layer, extracts and quantifies key features; the parameter determination subsystem undertakes the decision-making function, generating optimal control parameters; the control subsystem, as the execution center, coordinates the action of the air pump device; the air pump device, as the end effector, completes the actual physical removal task; and the re-inspection mechanism of the visual inspection subsystem forms a closed-loop feedback channel, enabling the system to have adaptive optimization capabilities. This multi-level, closed-loop architecture design ensures that the system maintains high response accuracy and removal success rate when facing attachments of different types and states.

[0044] Through the above technical solution, this application achieves intelligent, non-contact removal of soft deposits on windshields. Utilizing high-precision recognition technology based on visual neural networks, the system can quickly detect and locate sudden deposits, avoiding reliance on manual detection. By dynamically generating a blowing strategy based on multi-dimensional feature parameters such as boundary, center of gravity, area, and adhesion strength, the system ensures the targeted and effective removal action. Combined with a closed-loop feedback mechanism, even if the initial blowing fails to completely remove the deposits, the system can autonomously adjust parameters and attempt repeated attempts, enhancing its ability to handle complex conditions. Ultimately, without driver intervention, the system significantly improves the timeliness and reliability of windshield cleaning, solving the technical challenge of removing soft foreign objects obstructing vision at high speeds using traditional wipers, effectively enhancing driving safety and the overall vehicle intelligence level.

[0045] Example 2:

[0046] Based on the above embodiments, this embodiment further provides a parameter determination subsystem;

[0047] The parameter determination subsystem includes: a first parameter determination module, used to determine the coverage area of ​​the attachment based on boundary information;

[0048] The second parameter determination module is used to determine the degree of adhesion between the adhering substance and the windshield based on the adhesion strength parameter.

[0049] The third parameter determination module is used to locate the core area of ​​the attachment based on the center of gravity data;

[0050] The fourth parameter determination module is used to quantify the size of the attachment based on the area data.

[0051] This embodiment improves the interpretability and engineering applicability of the system in converting attachment identification results into control commands by clearly defining the functional definitions of each attachment state analysis parameter. Boundary information is a geometric description of the spatial distribution contour of the attachment within the windshield imaging area. It can be obtained by extracting pixel-level masks using image segmentation algorithms (such as Mask R-CNN) and fitting them to obtain the minimum bounding rectangle, polygon, or closed curve. This is used to define the spatial location and shape characteristics of the area to be processed, providing a spatial reference for the subsequent delineation of the airflow action area. For example, in a certain implementation scenario, when a plastic bag is irregularly attached to the upper left corner of the windshield, the boundary information can accurately reflect the specific coordinate range it occupies, enabling the control system to avoid ineffective blowing of unobstructed areas and improve response accuracy. As a variant implementation, the boundary information can also be expressed in polar coordinates, suitable for angle-distance control strategies in a local coordinate system with the wiper base point as the origin.

[0052] In the above embodiments, boundary information is used to determine the coverage area of ​​the attachment, adhesion strength parameters are used to determine the degree of adhesion between the attachment and the windshield, center of gravity data is used to locate the core area of ​​the attachment, and area data is used to quantify the size of the attachment.

[0053] The adhesion strength parameter characterizes the tightness of the bond between the adhered material and the windshield surface. It is not a direct measurement but a comprehensive evaluation index derived from the fusion of visual features and prior experimental data. This parameter can be estimated by combining image features such as edge blurring, texture tightness, and dynamic jitter frequency, and by referencing typical adhesion behavior models of different materials (such as plastic film, leaves, and insect remains) under different vehicle speeds and airflow conditions from a pre-set database. For example, for a plastic bag tightly adhering to the glass surface at high speed with no obvious edge movement, the system determines its adhesion strength to be high, thus triggering a stronger airflow output. As an optional solution, the adhesion strength parameter can also incorporate data from onboard environmental sensors (such as current vehicle speed, wind pressure, and humidity) for dynamic correction to improve evaluation accuracy.

[0054] Center of gravity data refers to the center of mass or geometric center of the attachment on the two-dimensional image plane. It is usually calculated by weighted averaging of the pixel coordinates of the connected components after image segmentation and is used to identify the optimal point of impact of the airflow. Since most soft attachments tend to roll or peel off around a core area during force detachment, directing the airflow towards this core area helps to disrupt its overall stability with minimal energy. For example, when the attachment is a large area of ​​plastic sheeting with an eccentric distribution, the system prioritizes focusing the airflow on its center of gravity, causing that area to lift first, and then causing the surrounding parts to detach continuously. In an alternative embodiment, the center of gravity data can also be extended to a set of the centers of mass of multiple sub-regions, which is suitable for complex-shaped objects with multiple independent adhesion points, enabling phased, multi-point collaborative removal.

[0055] Area data is a quantified result of the number of pixels occupied by the attachment, converted into its actual physical size after camera calibration. The unit can be expressed as square centimeters or equivalent projected area, and it is used in the calculation model for blowing force. Larger attachments typically require a wider range and stronger, more sustained airflow coverage for effective removal; therefore, area data is one of the important input variables determining the air pump's output pressure and injection time. For example, when the detected attachment area exceeds a preset threshold (e.g., 500 cm²), the system automatically increases the base blowing force level. Alternatively, area data can be used in conjunction with shape factors (e.g., aspect ratio, perimeter-to-area ratio) to distinguish the different aerodynamic response characteristics of elongated strips and approximately circular objects, further optimizing the blowing strategy.

[0056] The aforementioned parameters collectively constitute a multi-dimensional representation system of the adhesion state, with complementary functions among them: boundary information provides spatial constraints, area data quantifies scale, center of gravity data indicates the target point, and adhesion strength parameters reflect the difficulty of physical interaction. These four parameters work together to support intelligent decision-making for subsequent air blowing parameters. Through this technical solution, this application achieves refined modeling of the state of soft adhesions on windshields. By clearly defining the functions of each parameter, the system can accurately understand the visual characteristics of the adhesions and transform them into executable control logic without human intervention. This solves the problem of low removal efficiency caused by the lack of state perception capabilities in traditional removal methods, thereby improving the reliability and adaptability of the automated removal process.

[0057] Example 3:

[0058] Based on the above embodiments, this embodiment further provides:

[0059] The first calculation module is used to calculate the blowing angle parameters based on the center of gravity data and boundary information of the attached object; the second calculation module is used to calculate the blowing force parameters by using the ratio between the area data of the attached object and the length of the attached boundary.

[0060] This embodiment introduces two dedicated calculation modules to achieve refined and quantifiable modeling of the blowing angle and blowing force parameters. The first calculation module converts the geometric features of the attachments output by the visual inspection system into optimal action direction commands, determining the target angle at which the airflow from the air pump nozzle should point. The second calculation module establishes a mechanical response model based on the physical properties of the attachments, outputting control parameters for the required airflow intensity. Together, these two modules constitute the core logical unit for parameter decision-making, enabling the removal strategy to be adaptive and dynamically adjust its actions according to different attachment states.

[0061] The "center of gravity data" relied upon by the first calculation module refers to the coordinate position of the center of mass calculated using the image moment algorithm after pixel-level analysis of the attachment image. This position reflects the visually concentrated area of ​​the attachment on the windshield, usually corresponding to its main obstruction area or the most difficult part to peel off. Combined with "boundary information," namely the set of vertices of the polygonal outline of the attachment segmented by the visual neural network, the module can further analyze the spatial distribution characteristics of the outline, such as the direction of the longest axis and geometric attributes such as the convex hull offset angle. Based on this information, the first calculation module uses a vector synthesis method to determine the optimal blowing angle: with the center of gravity as the reference origin, it calculates the directional weight of each point on the boundary relative to the center of gravity, and comprehensively considers the position of the weak area of ​​the attachment boundary (such as the edge lifting), finally generating a direction vector pointing to the most vulnerable to breaking the attachment balance, which is converted into the angle parameter that the air pump device needs to align with. This angle is not simply perpendicular to the glass surface, but is intelligently tilted according to the attachment shape to ensure that the airflow impact can effectively pry away rather than just push the attachment.

[0062] As an optional implementation, when the attachment exhibits irregular tearing, the first calculation module can also employ a clustering algorithm to divide the boundary into multiple sub-regions, identify multiple potential stress points, and generate a multi-stage blowing angle sequence to support step-by-step removal operations. Furthermore, in some variant solutions, a drag coefficient correction term can be introduced, and the influence of the reverse airflow can be estimated in conjunction with the vehicle's current speed to dynamically compensate for the target angle, thereby improving blowing efficiency under high-speed conditions.

[0063] The second calculation module uses "area data," which refers to the actual projected area calculated from the effective pixels occupied by the attachment in the image plane after camera calibration, reflecting its overall size. "Attachment boundary length" refers to the total length of the actual contact edge between the attachment and the windshield, obtained by extracting and accumulating the Euclidean distance between adjacent pixels using a contour tracking algorithm, reflecting the degree of adhesion. This module constructs a proportional relationship between the two—"area / attachment boundary length"—to form a physical dimensional index characterizing the peeling load that needs to be overcome per unit contact length. Based on this, multiplying by an empirical constant k1 (whose value is obtained through wind tunnel experiments and real vehicle testing calibration) outputs the corresponding blowing force parameter. This formula implies the following physical logic: large-area attachments with short contact edges (such as lightweight plastic bags stuck in the center and curled up around the edges) are easily lifted by a smaller airflow; while small-area attachments with tightly adhered edges (such as wet, sticky leaves) require stronger, continuous air pressure to separate. Therefore, this ratio reasonably reflects the essential characteristics of the removal difficulty, avoiding the energy waste or removal failure problems caused by traditional fixed-gear control.

[0064] As an alternative implementation, the second calculation module can also introduce additional factors for weighted correction, such as the glass surface wetness fed back by the ambient humidity sensor, the effect of temperature on air density, or the learning weights of historical cleaning success rates, thereby forming a more complex nonlinear function model. In other embodiments, a look-up table (LUT) can also be used instead of real-time calculation, pre-storing the recommended strength values ​​corresponding to different area-boundary combinations to improve computational efficiency.

[0065] The two modules work together, with the former determining "where to blow" and the latter determining "how hard to blow," jointly completing the mapping process from perception to decision. They receive structured data input from the attachment state analysis subsystem, process it through internal algorithms, and output standardized control commands for subsequent control subsystems to call. This modular design not only enhances the system's maintainability and scalability but also provides an interface foundation for future upgrades and optimizations.

[0066] Through the above technical solution, this application achieves scientific modeling and automated generation of blowing parameters. By employing a vector analysis method based on center of gravity and boundary information to determine the blowing angle, the airflow direction can accurately match the geometric weaknesses of the attached object, improving the initial peeling success rate. Furthermore, by introducing the ratio of area to the length of the attachment boundary as the basis for calculating the blowing force intensity, the output force can cover a wide range of attachment scenarios while also responding to local adhesion characteristics, avoiding excessive force that would waste energy or disturb the surrounding glass structure. Therefore, without manual intervention, the system can autonomously formulate a removal strategy based on the specific shape of each attachment, significantly improving the intelligence level and actual removal efficiency in handling complex soft foreign objects.

[0067] Example 4:

[0068] Based on the above embodiments, this embodiment further provides:

[0069] The control subsystem includes a telescopic mechanism and an angle adjustment mechanism;

[0070] The telescopic mechanism is used to adjust the extension length of the air pump body;

[0071] The angle adjustment mechanism is used to adjust the blowing direction of the air pump body so that the air pump nozzle is aimed at the core area of ​​the attached material; the output pressure of the air pump is adjusted according to the blowing force parameters so that the airflow intensity meets the removal requirements.

[0072] Through the above technical solution, this application realizes a pneumatic actuator with spatial positioning and power adjustment capabilities, which can perform precise and adaptive air blowing to remove foreign objects with different positions, shapes, and attachment states on the windshield. The telescopic mechanism and the angle adjustment mechanism together constitute a multi-degree-of-freedom motion platform, which, combined with a pressure-adjustable air source control system, achieves three-dimensional dynamic control of the airflow's point of action, direction, and intensity.

[0073] The telescopic mechanism adjusts the extension length of the air pump body, overcoming the spatial limitations of the vehicle's windshield wiper structure. This allows the air pump to actively extend or retract in a direction perpendicular to the windshield surface, shortening the airflow distance and improving impact efficiency. The telescopic mechanism can be implemented using linear drive components such as electric push rods, lead screw and nut pairs, or hydraulic cylinders. Optionally, a DC motor-driven lead screw structure with built-in guide rails can be used to ensure smooth operation and rapid response. The telescopic stroke can be designed from 50mm to 150mm depending on the vehicle's front compartment layout, sufficient to cover the effective working area from the wiper mounting base to the top of the inner curved surface of the windshield. In practical applications, when the object is located in a high or distant area, the telescopic mechanism automatically extends the air pump body, bringing it closer to the target surface, reducing airflow diffusion loss and enhancing local impact force.

[0074] An angle adjustment mechanism is used to adjust the air blowing direction of the air pump body, ensuring that the air nozzle can flexibly point towards the core area of ​​the adhered object. This core area is usually calculated by the adhesion state analysis subsystem and corresponds to the center of gravity of the adhered object or the area of ​​maximum adhesion concentration; it is the key point for achieving efficient peeling. The angle adjustment mechanism can use a rotary servo motor, a stepper motor with gear transmission, or a servo rotary platform to adjust the azimuth and pitch angles in a two-dimensional plane. It can also be integrated with a multi-axis linkage mechanism to adapt to the complex curvature of the windshield surface. For example, in one optional embodiment, the angle adjustment mechanism includes a horizontal rotation joint and a vertical tilt joint, each driven by an independent servo motor, forming a gimbal-like structure that can continuously adjust the air blowing direction within a horizontal range of ±60° and a vertical range of ±30°. By receiving command signals from the control subsystem, the angle adjustment mechanism drives the air pump nozzle to precisely align with the target center, ensuring that the airflow is incident directly on the adhesion interface, maximizing shear stress and peeling effect.

[0075] Furthermore, the control subsystem dynamically adjusts the air pump's output pressure based on the blowing force parameter, ensuring the airflow intensity matches the current removal requirements of the deposits. The blowing force parameter is calculated from the adhesion status analysis results, reflecting the required airflow kinetic energy. Air pump pressure regulation can be achieved through proportional solenoid valves, variable frequency drives for the air compressor, or pulse width modulation (PWM) control of the air pump motor speed. For example, when dealing with large areas of weakly adhered plastic film, the system can be set to a lower pressure, continuous blowing mode; while for small areas of tightly adhered residual adhesive tape, a high-pressure, short-time pulse blowing strategy is employed. This on-demand energy supply not only improves the removal success rate but also avoids energy waste or unnecessary mechanical impact on the glass surface caused by a fixed high-pressure output.

[0076] The aforementioned components are interconnected through electromechanical coordination: the telescopic mechanism changes the distance between the air pump and the glass, the angle adjustment mechanism determines the direction of airflow, and the pressure regulation module controls the airflow output. The combined effect of these three mechanisms enables the entire pneumatic cleaning system to operate with "close proximity, precise targeting, and controllable pressure." Especially when dealing with deposits in non-planar or edge areas, this structural design significantly improves the system's spatial adaptability and cleaning reliability.

[0077] Through the above technical solution, this application realizes an intelligent air-blowing control mechanism based on the coupling of spatial positioning and power adjustment. Due to the telescopic mechanism, the air pump can overcome the limitations of its original installation location and operate close to the windshield surface, effectively reducing energy attenuation during airflow propagation. Due to the angle adjustment mechanism, the blowing direction can dynamically track the core area of ​​the attached object, ensuring that the airflow impacts the attachment interface at the optimal angle, improving peeling efficiency. Simultaneously, the output pressure is adjusted in real time according to the blowing force parameters, matching the airflow intensity with the removal difficulty, balancing removal effectiveness and energy consumption optimization. Overall, this control subsystem enhances the response capability and removal success rate to different types and locations of attached objects, solving the technical defects of traditional fixed air-blowing devices, such as limited range, non-adjustable direction, and power mismatch. It provides key execution guarantees for achieving fully automatic and highly adaptable windshield foreign object removal.

[0078] Example 5:

[0079] Based on the above embodiments, this embodiment further provides:

[0080] The visual inspection subsystem includes a camera and a pre-trained visual neural network model;

[0081] A camera is used to capture image frames of the windshield area in real time;

[0082] The visual neural network model performs iterative analysis on image frames to identify the presence of attached objects.

[0083] The camera is a high dynamic range (HDR) industrial-grade image sensor installed inside the vehicle's front compartment, facing the windshield. It features a resolution of 1080p or higher and a frame rate of 30fps or higher, enabling stable operation under varying lighting conditions day and night. The camera connects to the main control unit via an in-vehicle Ethernet or MIPI interface, continuously transmitting video stream data to the visual neural network model. Its installation position is optically optimized to avoid image distortion caused by viewing angle shifts, and it has a self-cleaning coating to prevent dust and water damage that could affect image quality. In optional embodiments, the camera can be replaced with an infrared camera or a multispectral imaging device to enhance detection capabilities in low-visibility environments such as foggy days and rainy nights.

[0084] The pre-trained visual neural network model is a deep learning-based architecture for object detection and semantic segmentation, deployed on an embedded AI accelerator in an in-vehicle computing platform. It receives and processes image frame sequences from a camera. The model has been trained offline using a large number of labeled samples, covering image data of various typical attachment types (such as plastic bags, paper scraps, insect remains, leaves, etc.) under different lighting conditions, angles, and occlusion levels, demonstrating strong generalization capabilities. During runtime, the model performs loop inference operations on each frame, outputting a judgment result on whether it contains attachments and the corresponding pixel-level mask information. In its implementation, the neural network can employ lightweight segmentation network structures such as Mask R-CNN, YOLACT, or Lite-DeepLab, ensuring accuracy while meeting the stringent real-time and power consumption requirements of the in-vehicle system. Alternatively, a two-stage processing flow can be used: first, an efficient detection model such as YOLOv8 locates suspected regions, and then a fine-grained segmentation sub-model is called to refine the boundaries, thereby reducing the overall computational load.

[0085] A closed-loop perception link is formed between the camera and the visual neural network model: the camera continuously acquires environmental images, while the neural network model initiates analysis tasks at fixed intervals or through event triggers. Once an abnormal object is detected in the windshield area, an activation signal is generated and transmitted to the subsequent attachment status analysis subsystem. Together, they achieve all-weather, uninterrupted monitoring of the windshield surface, ensuring that the system response latency is controlled within 200 milliseconds, meeting the rapid response requirements in high-speed driving scenarios.

[0086] Through the above technical solution, this application achieves high-precision and robust identification of deposits on windshields. Due to the adoption of a hardware-software combined visual perception architecture, the system not only overcomes the problem of false detections or missed detections easily caused by traditional image processing methods under complex background interference, but also maintains stable identification performance under conditions of drastic changes in lighting, wet glass, or the presence of stains, thanks to the powerful feature extraction capabilities of deep neural networks. Simultaneously, the cyclic analysis mechanism ensures the continuity of the monitoring process, avoids transient omissions, and provides a reliable preliminary judgment basis for subsequent cleaning actions. Therefore, this implementation significantly improves the autonomous decision-making capability and practical application reliability of the intelligent cleaning system.

[0087] Example 9:

[0088] During high-speed driving, plastic bags, paper, or other flexible foreign objects may suddenly adhere to the windshield. These objects are characterized by strong adhesion, complex deformation, and a large contact area with the glass surface. Traditional windshield wipers, relying on mechanical wiping and with uniform force distribution, are ineffective in removing such soft deposits. Especially when these objects are located in the driver's primary field of vision, they can severely obstruct vision and increase driving risks. Current technologies lack automated methods for identifying and removing non-rigid foreign objects in high-speed dynamic environments, forcing drivers to intervene actively or stop the vehicle, resulting in delayed response and significant safety hazards. Therefore, there is an urgent need for a technological solution capable of autonomously detecting, intelligently deciding, and executing removal operations to address the visual obstruction caused by sudden windshield deposits.

[0089] Based on this, this application provides a method for treating windshield deposits based on a visual neural network, such as... Figure 1 As shown, the specific steps include:

[0090] Step S101: Use a pre-trained visual neural network model to perform cyclic detection on the windshield area to identify whether there are any attached objects;

[0091] The pre-trained visual neural network model refers to a deep learning model trained using supervised learning on a large dataset of labeled images (such as those containing various common windshield attachments—plastic bags, leaves, insect remains, etc.). Examples include architectures like Mask R-CNN, YOLOv5-seg, or U-Net, which possess object detection and semantic segmentation capabilities. This model is deployed in an in-vehicle computing unit (such as a domain controller or AI chip) and can receive video frame streams in real time from a high-definition camera positioned in the forward-looking direction inside the vehicle. Loop detection refers to the system continuously analyzing image sequences at a set frequency (e.g., 10-30 frames per second) to achieve uninterrupted monitoring. The detection process includes image preprocessing (denoising, illumination normalization), feature extraction, candidate region generation, and classification. When a target category with a confidence level higher than a threshold (e.g., "plastic bag") appears in the output, it is determined that an attachment exists, triggering subsequent processing. As an optional implementation, the model can integrate an attention mechanism to enhance the recognition of blurry or semi-transparent attachments; it can also continuously update model parameters through online incremental learning to adapt to newly emerging attachment types.

[0092] Step S102: When an attachment is detected, calculate the boundary information, attachment strength parameters, center of gravity data and area data of the attachment, and determine the blowing angle parameters and blowing force parameters.

[0093] Boundary information refers to the polygonal representation of the attachment's outer contour in the image coordinate system, used to define its spatial coverage area, and can be obtained through contour tracking algorithms (such as the findContours function in OpenCV). Area data is calculated by multiplying the total number of pixels within the mask by the actual physical area corresponding to each pixel (derived from the camera's focal length and installation position), used to quantify the size of the attachment. Centroid data is obtained by calculating the weighted average of the coordinates of all pixels within the mask area, representing the mass concentration point of the attachment, and plays a crucial role in subsequently locating the core area of ​​the blowing air.

[0094] The adhesion strength parameter is used to reflect the degree of adhesion between the adhering material and the glass. Its calculation method can be combined with image texture features (such as edge sharpness reduction rate and light transmission unevenness) and prior experimental database for comprehensive evaluation. For example, in wind tunnel testing, a reference table of adsorption forces corresponding to different materials (PE film, paper, etc.) at different vehicle speeds can be established, and then the approximate strength level can be matched by the deformation state of the adhering material in the current image.

[0095] The determination of the blowing angle parameter depends on the relationship between the center of gravity and the vehicle coordinate system: if the center of gravity is biased to the left, the air pump is tilted to the left at a certain angle to ensure that the airflow line passes through or is close to the center of gravity, improving the flipping and peeling efficiency; alternatively, the direction of the main axis can be fitted by combining the boundary orientation to make the blowing direction perpendicular to the longest attachment edge. The blowing force parameter is estimated based on the mechanical model to obtain the required airflow energy, for example, using the formula: Force = k1 × (area) / (attachment boundary length); where k1 is an empirical coefficient obtained through actual vehicle testing and the attachment boundary length refers to the effective component of the perimeter of the contact between the attachment and the glass (excluding free overhanging edges). All of the above parameters can be dynamically updated, supporting multiple rounds of iterative optimization.

[0096] Step S103: Adjust the position and angle of the air pump device according to the blowing angle parameter, adjust the air pump pressure according to the blowing force parameter, and perform the blowing operation according to the adjusted parameters;

[0097] The adjustment of the air pump's position and angle involves controlling the pneumatic components mounted near the driver's side wiper base to adjust their spatial attitude. This device is equipped with a telescopic mechanism (such as an electric push rod or lead screw slide) to change the distance between the nozzle and the glass, and an angle adjustment mechanism (such as a servo-driven universal joint structure) to adjust the pitch and yaw angles. The control system converts the blowing angle parameters into motor control commands, driving the actuator to the target position to ensure the nozzle is directly aligned with the center of gravity or optimal point of action of the deposit. Adjusting the air pump pressure is achieved by regulating the opening of the proportional valve at the outlet of the air source (such as an onboard air compressor or air tank), controlling the output gas flow and pressure per unit time to match the calculated blowing force parameters. During the blowing operation, the air pump briefly releases a high-pressure airflow (e.g., 0.4~0.8 MPa, lasting 0.5~2 seconds), creating a localized shock wave that disturbs the adhesion interface, disrupting the static friction balance and causing the deposit to curl and detach from the edges. As an alternative, a pulsed blowing mode can be used, using high-frequency intermittent jets to improve removal efficiency without wasting energy.

[0098] Step S104: Check again whether the deposits on the windshield have been removed; if not, recalculate the parameters and perform the blowing operation until the deposits are removed or the preset number of attempts is reached, then terminate the operation.

[0099] The re-detection step refers to the system automatically returning to step S101 after each air blowing cycle, calling the visual neural network to re-examine the latest image frame to confirm whether the same type of adhering material still exists in the original target area. The judgment criteria can be set as follows: the residual area at the original location is less than 10% of the initial area, or the classification confidence is lower than a set threshold. Recalculating parameters means entering a new round of closed-loop feedback—even if the same adhering material is not completely removed, its shape may change (e.g., partially lifted, position shifted). In this case, parameters such as boundary and center of gravity need to be re-extracted, and the air blowing strategy updated accordingly to avoid ineffective blowing caused by repeatedly using old parameters. The entire process constitutes an adaptive iterative cleaning mechanism, allowing a maximum of N consecutive executions (e.g., N=3~5 times) to prevent infinite looping and energy consumption. Once the maximum number of attempts is reached and the material is still not successfully removed, the system can issue a prompt signal (e.g., dashboard warning icon, voice reminder) to notify the driver to take manual action. This mechanism significantly improves the system's robustness and fault tolerance, especially suitable for complex weather conditions or strong adsorption scenarios.

[0100] Through the above-described steps, this application achieves a fully intelligent removal process from perception to execution to feedback. It utilizes visual neural networks to achieve high-precision identification and state analysis of adhering objects, dynamically generates optimal blowing parameters using mechanical modeling, precisely applies aerodynamic force using an adjustable air pump, and finally ensures a high removal success rate through a closed-loop verification mechanism. This method requires no driver intervention, responds rapidly, and is applicable to various soft foreign object scenarios. It effectively solves the technical bottleneck of traditional windshield wipers' inability to handle flexible adhering objects, significantly improving driving safety and the environmental adaptability of autonomous driving systems.

[0101] Example 10:

[0102] If soft foreign objects such as plastic bags suddenly adhere to the windshield while a car is traveling at high speed, it can severely obstruct the driver's view, posing a significant safety hazard. These objects are highly flexible, adherent, and easily deformed by airflow. Traditional windshield wipers, relying on mechanical wiping motions, are often ineffective at removing large areas of low-rigidity debris, and may even tear or spread the object, further expanding the obstructed area. Furthermore, current vehicles lack the intelligent sensing capabilities to detect the condition of the windshield surface, making proactive identification and response difficult. Therefore, there is an urgent need for a technological solution that can automatically detect and remove soft deposits from the windshield to improve driving safety and the level of driving automation.

[0103] In addition, this application also proposes an electronic device and a computer-readable storage medium.

[0104] In one embodiment, the provided electronic device may be a terminal, and its internal structure diagram may be as follows: Figure 3 As shown. The electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements the method described in any one of steps S101 to S104. The display screen can be a liquid crystal display (LCD) or an electronic ink display. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0105] Those skilled in the art will understand that Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0106] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any one of S101-S104.

[0107] Specifically, the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0108] The code of the computer program can be in the form of source code, object code, executable file, or some intermediate form.

[0109] Computer-readable storage media may include cache, high-speed random access memory (RAM), such as the common double data rate synchronous dynamic random access memory (DDR SDRAM), and may also include non-volatile memory (NVRAM), such as one or more read-only memory (ROM), disk storage devices, flash memory devices, or other non-volatile solid-state storage devices such as optical discs (CD-ROM, DVD-ROM), floppy disks, or data tapes.

[0110] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0111] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0113] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A windscreen accretion treatment device based on a visual neural network, characterized by, The method comprises the following steps: A visual detection subsystem, an adhesion state analysis subsystem, a parameter determination subsystem and a control subsystem are arranged in a vehicle-mounted system; When the visual detection subsystem detects that there is an adherend on the windshield, the adhesion state analysis subsystem is triggered to start; The adhesion state analysis subsystem is used to receive the adherend image data transmitted by the visual detection subsystem, calculate the boundary information, adhesion strength parameter, barycentric data and area data of the adherend; The parameter determination subsystem determines the blowing angle parameter and blowing strength parameter based on the boundary information, adhesion strength parameter, barycentric data and area data calculated by the adhesion state analysis subsystem; The control subsystem is used to receive the blowing angle parameter and blowing strength parameter, and control the air pump device arranged near the rain wiper of the driver's side of the vehicle to adjust; The visual detection subsystem is used to capture the image frame of the windshield area again after the blowing operation is completed, and analyze and judge whether the adherend is successfully removed; If the detection result shows that the adherend has been removed, the current removal operation is terminated; if the adherend has not been removed, the adhesion state analysis step is returned to, the blowing parameters are recalculated and the blowing operation is performed, and the operation is terminated after the adherend is removed or the preset number of attempts is reached, and a prompt signal can be issued.

2. The apparatus of claim 1, wherein, The parameter determination subsystem comprises a first parameter determination module for determining the coverage range of the adherend based on the boundary information; A second parameter determination module is used to judge the adsorption degree of the adherend and the windshield based on the adhesion strength parameter; A third parameter determination module is used to locate the core area of the adherend based on the barycentric data; A fourth parameter determination module is used to quantify the size of the adherend based on the area data.

3. The apparatus of claim 2, wherein, The parameter determination subsystem comprises a first calculation module for calculating the blowing angle parameter based on the barycentric data and boundary information of the adherend.

4. The apparatus of claim 3, wherein, The parameter determination subsystem further comprises a second calculation module for calculating the blowing strength parameter through the proportional relationship between the area data of the adherend and the adhesion boundary length.

5. The apparatus of claim 1, wherein, The control subsystem comprises a telescopic mechanism and an angle adjusting mechanism; The telescopic mechanism is used to adjust the extension length of the air pump body; The angle adjusting mechanism is used to adjust the blowing direction of the air pump body, so that the blowing port of the air pump is aligned with the core area of the adherend; and the output pressure of the air pump is adjusted according to the blowing strength parameter, so that the airflow intensity meets the removal requirement.

6. The apparatus of claim 5, wherein, The air pump device is used to perform the blowing operation on the adherend on the windshield according to the adjusted position, angle and pressure parameters, and use the impact force of the airflow to make the adherend separate from the windshield.

7. The apparatus of claim 1, wherein, The visual detection subsystem comprises a camera and a pre-trained visual neural network model; The camera is used to capture the image frame of the windshield area in real time; The visual neural network model analyzes the image frame in a loop to identify whether there is an adherend.

8. A method for processing windshield adherent objects based on a visual neural network, characterized by, The method comprises the following steps: The windshield area is detected in a loop by the pre-trained visual neural network model to identify whether there is an adherend; When the adherend is detected, the boundary information, adhesion strength parameter, barycentric data and area data of the adherend are calculated, and the blowing angle parameter and blowing strength parameter are determined; According to the blowing angle parameter, the position and angle of the air pump device are adjusted, according to the blowing strength parameter, the air pump pressure is adjusted, and the blowing operation is performed according to the adjusted parameters; The attachment on the windshield is detected again to determine whether it is removed; if not, the parameters are recalculated and the blowing operation is performed until the attachment is removed or the operation is terminated after a preset number of attempts.

9. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory connected to the at least one processor in communication; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the steps of the method of claim 8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method of claim 8.

Images