Control method and device of intelligent atomization window system, computer equipment and storage medium
By combining multimodal sensors and building information models, real-time fogging control of intelligent fogging windows is achieved, solving the problems of high false trigger rate and response delay in existing technologies, and providing precise proactive privacy protection.
Patent Information
- Application Number
- CN202511367035.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-24
- Publication Date
- 2025-11-25
AI Technical Summary
Existing fogging window control methods rely on manual operation, timed operation, or simple infrared sensing, resulting in a high false trigger rate and response delay. They also lack intelligent learning and predictive capabilities, and traditional video surveillance systems lack proactive protection mechanisms.
Initial multimodal data is acquired through multimodal sensors, and the YOLOv7 target detection model is used for identification and trajectory prediction. The building information model is then combined to make fogging decisions, thereby achieving real-time fogging action.
It achieves intelligent atomization window linkage control with low false triggering and high response speed, provides proactive privacy protection, accurately identifies and predicts target trajectory, and drives the atomization window to move in real time in advance.
Smart Images

Figure CN121008511A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of automation control, and in particular to a control method and device of an intelligent fogging window system, a computer device, and a storage medium. BACKGROUND
[0002] The fogging window is a product that adjusts the light transmittance of glass (switches from transparent to fogging state) through electric control or light control, can flexibly protect indoor privacy and adjust the amount of light entering, and has the advantages of energy saving, sound insulation, and convenient operation.
[0003] However, the control method of the fogging glass (fogging window) of the building in the related art mainly depends on manual operation, timing, or simple infrared induction, has the disadvantages of high false triggering rate or delayed response. The traditional video monitoring system only has the functions of recording, storing, and retrieving, and lacks an active protection mechanism. SUMMARY
[0004] The present application aims to at least partly solve one of the technical problems in the related art. To this end, the present application provides a control method and device of an intelligent fogging window system, a computer device, and a storage medium. The main technical solution adopted by the present application includes:
[0005] In a first aspect, the present application provides a control method of an intelligent fogging window system, the intelligent fogging window system being arranged on a target building; the method comprises: determining a target object in a target range of the target building, and obtaining initial multi-modal data of the target object; performing target recognition and trajectory prediction processing based on the initial multi-modal data to obtain target prediction data of the target object; wherein the target prediction data comprises a relative distance between the target object and the target building and a motion trajectory of the target object within a preset time range; calling a building information model of the target building, determining a fogging decision result based on the relative distance and the motion trajectory, and driving the intelligent fogging window system to perform a real-time fogging action.
[0006] Optionally, the initial multi-modal data is obtained by: obtaining life attribute data, three-dimensional space feature data, and motion feature data; wherein the life attribute data is used to describe the infrared radiation intensity and the contour range of the target object; the three-dimensional space feature data is used to describe the appearance details and depth information of the target object; the motion feature data is used to describe the real-time motion trajectory and motion trend of the target object; and performing time synchronization and data fusion processing on the life attribute data, the three-dimensional space feature data, and the motion feature data to obtain the initial multi-modal data.
[0007] Optionally, the target recognition and trajectory prediction processing are performed based on the initial multi-modal data to obtain target prediction data of the target object, including: performing target recognition processing on the initial multi-modal data by using a target detection model based on a YOLOv7 architecture to obtain a target recognition result of the target object; wherein the target recognition result includes an object category of the target object; and performing trajectory prediction processing based on the target recognition result and the initial multi-modal data to obtain the target prediction data of the target object.
[0008] Optionally, the trajectory prediction processing is performed based on the target recognition result and the initial multi-modal data to obtain a target prediction result of the target object, including: in a case where the object category indicates that the target object is a vehicle, obtaining real-time motion state data of the vehicle based on the initial multi-modal data; wherein the real-time motion state data includes a distance, a speed and an azimuth angle of the vehicle; and performing steering prediction on the vehicle by using a Kalman filter model based on the distance, the speed and the azimuth angle to obtain the target prediction result.
[0009] Optionally, the trajectory prediction processing is performed based on the target recognition result and the initial multi-modal data to obtain a target prediction result of the target object, including: in a case where the object category indicates that the target object is a pedestrian, determining a direction judgment result of the pedestrian based on the initial multi-modal data; and in a case where the direction judgment result indicates that the pedestrian is heading towards the target building, performing motion trajectory prediction on the target object to obtain the target prediction result.
[0010] Optionally, the motion trajectory prediction is performed on the target object to obtain the target prediction result, including: performing motion trajectory prediction on the target object by using a long short-term memory network model based on the skeletal key point time series data to obtain the target prediction result; wherein the skeletal key point time series data is determined based on the initial multi-modal data.
[0011] Optionally, the building information model of the target building is called, and a fogging decision result is determined based on the relative distance and the motion trajectory to drive the intelligent fogging window system to perform a real-time fogging action, including: determining a predicted line-of-sight area of the target object based on the relative distance and the motion trajectory; determining a target fogging window unit that needs to perform a fogging action and a privacy exposure risk index of the target fogging window unit from the intelligent fogging window system by using spatial attribute data of the building information model and the predicted line-of-sight area; determining a fogging decision result of the target fogging window unit in a case where the privacy exposure risk index exceeds a preset risk threshold; and driving the target fogging window unit to perform a real-time fogging operation by using the fogging decision result.
[0012] Secondly, embodiments of this application provide a control device for an intelligent atomizing window system, wherein the intelligent atomizing window system is installed on a target building; the device includes: an object determination module, used to determine a target object within a target area of the target building and acquire initial multimodal data of the target object; an identification and prediction module, used to perform target identification and trajectory prediction processing based on the initial multimodal data to obtain target prediction data of the target object; wherein the target prediction data includes the relative distance between the target object and the target building and the movement trajectory of the target object within a preset time range; and an atomization execution module, used to call the building information model of the target building, determine the atomization decision result based on the relative distance and movement trajectory, so as to drive the intelligent atomizing window system to perform real-time atomization actions.
[0013] Thirdly, this application also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the above methods.
[0014] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.
[0015] Fifthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above methods.
[0016] In the above embodiments, by first locking the target and fusing multimodal data to complete identification and trajectory prediction, and then calling the building information model to convert the relative distance and motion trajectory into fogging decisions, the system can drive the intelligent fogging window to act in real time before the target enters the visible range, thereby achieving proactive privacy protection with advance shielding, low accidental touch, and accurate window area response. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0018] Figure 1 This is a flowchart of a control method for an intelligent atomizing window system according to an embodiment of this application;
[0019] Figure 2 This is a flowchart of a vehicle trajectory prediction according to an embodiment of this application;
[0020] Figure 3This is a flowchart of pedestrian trajectory prediction according to an embodiment of this application;
[0021] Figure 4 This is a deployment topology diagram of an intelligent atomizing window linkage control system provided according to a scenario example of this application;
[0022] Figure 5 This is a structural block diagram of a control device for an intelligent atomizing window system according to an embodiment of this application;
[0023] Figure 6 This is an internal structural diagram of a computer device provided according to an embodiment of this application. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0025] In related technologies, the control methods for building frosted glass (frosted windows) mostly rely on manual operation, timed operation, or simple infrared sensing, lacking intelligent learning and predictive capabilities, resulting in a high false trigger rate or delayed response. Furthermore, traditional video surveillance systems only have recording, storage, and retrieval functions, lacking proactive protection mechanisms.
[0026] Specifically, the control system of the related technology for atomized windows (taking mainstream electronically controlled atomized glass as an example) consists of an atomizing layer, a conductive layer, a substrate layer, a control terminal, a central controller, sensor expansion, and a power supply module. Its control methods include manual control, timed control, sensor linkage, intelligent central control integration, or mobile terminal control. However, the above control processes have some limitations. For example, relying on simple infrared sensors or timers makes it impossible to distinguish between people / vehicles / animals. Furthermore, there is a response delay, making it unable to handle fast-moving targets and lacking predictive and intelligent decision-making capabilities. In addition, the system is also closed, making it difficult to deeply integrate with intelligent devices such as security cameras. Therefore, a smart atomized window linkage control scheme with accurate target recognition and trajectory prediction capabilities, achieving low false triggering and high response speed, is needed.
[0027] According to an embodiment of this application, a control method embodiment for an intelligent atomizing window system is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.
[0028] This embodiment provides a control method for an intelligent fogging window system, which is installed on a target building. For example... Figure 1 As shown, the process includes the following steps:
[0029] S110. Determine the target object within the target area of the target building and obtain the initial multimodal data of the target object.
[0030] The target building can refer to the main building requiring privacy protection and fogging control. The target area can refer to the peripheral area around the target building at a certain distance that may pose a potential observation threat to the privacy inside the building; the size of the area can be freely set according to the building's privacy sensitivity. The target object can refer to various moving targets entering the target area. For example, the target building can be a residence, office, hotel room, hospital clinic, or other building that requires protection of internal privacy. The target object can be a pedestrian, vehicle occupant, animal, or other entity that may observe the interior, such as a passerby loitering outside a residence window or people inside a vehicle parked downstairs in an office building. The initial multimodal data can refer to the raw, unprocessed set of sensor data directly acquired in real time from various types of sensing devices. For example, the initial multimodal data can be video data such as RGB video streams or image data such as depth image sequences.
[0031] Specifically, after identifying the target building requiring privacy protection, the first step is to determine a distance threshold based on the building's privacy requirements to delineate the target area, ensuring that all areas potentially subject to observation are covered. Initial multimodal data can then be obtained through the following methods:
[0032] First, acquire the organism's attribute data, three-dimensional spatial feature data, and motion feature data. Then, perform time synchronization and data fusion processing on the organism's attribute data, three-dimensional spatial feature data, and motion feature data to obtain the initial multimodal data.
[0033] Specifically, the above three types of data can be obtained through multimodal sensors. Multimodal sensors may include thermal imaging cameras with a detection range of 15 to 50 meters that can capture infrared radiation signals of the target object in real time, red-green-blue-depth (RGB-D) cameras that can simultaneously acquire color and depth images of the target object, and millimeter-wave radars with an operating frequency of 24 GHz ± 1 GHz that can detect parameters such as distance, speed, and direction of motion of the target object in real time.
[0034] Living organism attribute data can refer to data acquired through thermal imaging cameras that describes the infrared radiation intensity and contour range of a target object. For example, this could be thermal images acquired by a thermal imaging camera or temperature matrix data output by the camera. Three-dimensional spatial feature data can refer to data acquired through RGB-D cameras that describes the appearance details and depth information of a target object. For example, this could be color images and depth images acquired by an RGB-D camera. For instance, color images can determine appearance details such as clothing color or vehicle brand logos, while depth images can determine the actual distance between the target and a window, as well as depth information such as the target's height and width. Motion feature data can be point cloud data obtained through millimeter-wave radar that describes the real-time motion trajectory and movement trends of a target object. For example, point cloud data can reflect, for instance, the real-time motion trajectory of a target moving from the south side of a building to the east side, and the movement trend of the target moving closer to or away from a window.
[0035] It should be noted that, in order to achieve data fusion and collaborative analysis of multimodal sensors, a unified digital three-dimensional spatial coordinate system based on the target building needs to be constructed. This coordinate system takes a fixed point on the target building as its origin (which could be the center point of the outer wall), thereby mapping the position information of the target object in physical space into this spatial coordinate system.
[0036] For example, the thermal imaging camera first detects a suspected target heat source in its own coordinate system, and then outputs temperature matrix data containing the heat source temperature and coordinate information (e.g., 36.2℃, x=12.3m, y=5.7m) based on the thermal image data, as living organism attribute data. The millimeter-wave radar also detects a moving object suspected to be the target and outputs point cloud data containing motion parameters (e.g., speed 1.2m / s, azimuth angle 42°, etc.), as motion feature data. Simultaneously, the RGB-D camera, combining the point cloud data, confirms the target object's outline height as 1.72m (consistent with human characteristics) and gait frequency as 1.8Hz (normal walking frequency), and finally outputs a color image and depth image containing distance and outline (distance 10.5m, outline features), as three-dimensional spatial feature data. Finally, the multiple coordinate data from different sensors describing the same target are converted to a three-dimensional spatial coordinate system based on the target building, providing a consistent positional reference for subsequent target recognition and trajectory prediction.
[0037] Furthermore, after obtaining the organism's attribute data, three-dimensional spatial feature data, and motion feature data, these can be uniformly transmitted to the core processing unit responsible for data integration within the system. This unit may include FPGA chips, capable of extracting features from the transmitted data and aligning it using a Precision Time Protocol (PTP) to generate a fused data packet. The time synchronization error of the data processed by the FPGA chip must be less than 10ms. This fused data packet can then be used as the initial multimodal data.
[0038] Thus, by setting up different types of multimodal detectors, initial multimodal data containing three aspects—life attributes, spatial shape, and motion vectors—was obtained, enhancing the ability to distinguish targets and predict trajectories, so as to enable subsequent millisecond-level linkage fogging decisions.
[0039] S120. Based on the initial multimodal data, target recognition and trajectory prediction processing are performed to obtain target prediction data of the target object.
[0040] Among them, target prediction data can refer to a set of data that predicts the behavior trend of a target object and its relative relationship with buildings over a period of time by analyzing the target object's motion state and spatial location.
[0041] Specifically, the target prediction data includes the relative distance between the target object and the target building, as well as the movement trajectory of the target object within a preset time range.
[0042] The relative distance between the target object and the target building can refer to the actual spatial distance between the target object within the target area of the target building and the outer wall of the target building or a specific reference point. For example, the relative distance can refer to the straight-line distance of the target object along the normal vector of the outer wall of the target building. The movement trajectory of the target object within a preset time range can refer to the path of its position change within the target area over a predicted period of time based on the target object's current movement state. The preset time range can be a duration of several seconds set by the system based on factors such as the target object's movement speed and the size of the target area. This duration ensures that atomization preparation is initiated before the target enters the visible range, meeting real-time protection requirements.
[0043] Specifically, the process begins by extracting image sequences and motion parameters containing the target object from the initial multimodal data, and then preliminarily classifying the target object's category attributes. Next, combining the motion parameters from the initial multimodal data, the current direction and speed of the target object's movement are analyzed. Subsequently, an appropriate prediction method is selected based on the target object's category attributes to estimate its trajectory and distance changes over a future period. Finally, the calculated relative distance and trajectory are integrated to form target prediction data, which is used to subsequently determine whether the target object is likely to enter a privacy-sensitive area.
[0044] For example, target recognition and trajectory prediction processing are performed based on initial multimodal data to obtain target prediction data of the target object. A target detection model based on the YOLOv7 architecture can be used to perform target recognition processing on the initial multimodal data to obtain the target recognition result of the target object first.
[0045] Specifically, the initial multimodal data, including color, depth, and thermal images, is first input into a target detection model based on the YOLOv7 architecture. The model then extracts features from the input data to capture key features such as the object's outline and temperature distribution. The extracted features are then matched against the model's preset feature templates to identify potential targets in the image, and the model outputs their bounding boxes, confidence scores, category labels, and preliminary motion parameters to obtain the target recognition result. The target recognition result refers to a set of information that clearly defines the core attributes of the target object, including its position, size, speed, direction, and category within the image. The object category can refer to the type of target object classified based on its physical characteristics and motion properties; specifically, object categories can include common moving targets such as people, vehicles, or animals.
[0046] Subsequently, trajectory prediction processing is performed based on the target recognition results and the initial multimodal data to obtain the target prediction data of the target object.
[0047] Specifically, the target object category can be determined based on the target recognition results, and corresponding trajectory prediction logic can be selected for different categories. First, the current motion velocity and orientation angle parameters of the target object are extracted from the initial multimodal data as the basic input for prediction. Then, combined with the spatial coordinate system of the target area, the position offset of the target object per unit time is calculated. Subsequently, the predicted position of the target object at each time node is gradually calculated to form the motion trajectory. At the same time, based on the position offset and the initial distance, the relative distance between the target object and the target building at each prediction time node is calculated. Finally, the motion trajectory and relative distance are integrated to obtain the target prediction data.
[0048] Thus, by using YOLOv7 to perform joint feature extraction and template matching on color-infrared dual-mode images, the target recognition result is first output. Then, based on this, velocity and orientation angles are extracted from the initial multimodal data and coupled with a spatial coordinate system to calculate the continuous predicted position and the distance to the relative building. This yields target prediction data containing both motion trajectory and relative distance, enabling subsequent trajectory prediction and fogging triggering before the target enters the visible area, overcoming the lag mode of traditional fogging windows that responds only after being seen. In other words, by using the YOLOv7 target detection model to classify people / vehicles / animals, the problem of existing fogging window control technologies relying on simple infrared sensors or timers, being unable to distinguish between people, vehicles, or animals, and having a high false trigger rate is solved.
[0049] S130: Call the building information model of the target building, determine the atomization decision result based on the relative distance and motion trajectory, and drive the intelligent atomization window system to perform real-time atomization actions.
[0050] Building Information Modeling (BIM) can refer to a three-dimensional model database that contains the digital physical and functional characteristics of a target building. For example, the BIM of the target building stores data such as the window ID, three-dimensional coordinates, dimensions, associated rooms, and maximum fogging response time for each glass unit. The fogging response delay for each glass unit from receiving an instruction to initiating fogging is less than 200ms.
[0051] Furthermore, BIM is deeply integrated with the window control matrix in the intelligent fogging window system. As the core unit connecting BIM and the intelligent fogging window execution system, the window control matrix synchronizes key information from each glass unit from BIM and feeds back the status data of the intelligent fogging window to BIM, achieving bidirectional collaboration between model data and actual equipment. The core execution component in the intelligent fogging window system is the electrically controlled dimming glass unit. This unit uses polymer dispersed liquid crystal (PDLC) technology to achieve a response time of ≤300ms from fogging activation to fogging completion, and supports 0-100% stepless dimming, accurately responding to fogging decision commands and meeting real-time protection requirements.
[0052] Specifically, by calling the building information model of the target building and determining the fogging decision based on the relative distance and motion trajectory, the intelligent fogging window system can execute real-time fogging actions. First, it is necessary to determine the predicted viewing area of the target object based on the relative distance and motion trajectory.
[0053] The predicted viewing area refers to the area of the building facade that the target object may observe within a preset future time window. For example, the predicted position of the target object at various future time points can be deduced based on its movement trajectory, and the straight-line distance between the target and various areas of the building can be calculated by combining the real-time relative distance. Further, based on the human viewing angle or the typical observation perspective of vehicle occupants, a cone-shaped viewing area is constructed with the predicted target position as the viewpoint. This cone-shaped area is then projected onto the building facade to generate one or more polygonal areas, which serve as the predicted viewing area of the target object.
[0054] Subsequently, using spatial attribute data from the Building Information Modeling (BIM) and predicted viewing areas, the target fogging window unit that needs to perform fogging actions and the privacy exposure risk index of the target fogging window unit are determined from the intelligent fogging window system.
[0055] Spatial attribute data can refer to data such as the window ID, three-dimensional coordinates, dimensions, and associated rooms of each fogging window stored in the building information model.
[0056] Specifically, the system can perform spatial relationship calculations and projection analysis, based on the predicted viewing area and the building geometry data in the Building Information Model (BIM), to first determine which facade areas are likely to be seen by the target. Then, using the window ID, 3D coordinates, and dimensions from the spatial attribute data, a spatial query is performed in the BIM to find all windows that intersect with the predicted viewing area or are within a set threshold (e.g., 1m), and these are designated as candidate fogging window units. Next, the system can further filter based on the associated rooms in the spatial attribute data, according to the functional attributes of the associated rooms. For example, if a candidate fogging window unit corresponds to a corridor, it does not need to be fogged and is not considered a target fogging window unit; if a candidate fogging window unit corresponds to a conference room, it needs to be fogged and can be considered a target fogging window unit.
[0057] After identifying the target fogging window units, it is necessary to determine the privacy exposure risk index for each target fogging window unit. The privacy exposure risk index can be a numerical indicator used to quantify the degree of risk of being spied on by the target fogging window unit. Specifically, this index can also be calculated based on multiple variables such as spatial attribute data from the Building Information Modeling (BIM) and predicted viewing areas.
[0058] For example, the privacy exposure risk index R can be calculated based on the following formula:
[0059] R = e -λd ·ω·cosθ
[0060] In the formula, d represents the straight-line distance between the target object and the target atomized window unit; θ represents the angle between the line of sight and the window normal vector; ω represents the weight of the object category of the target object, which can be 1.0 for pedestrians, 0.8 for vehicles, and 0.3 for animals; λ represents the distance attenuation coefficient, which can be 0.2.
[0061] Among them, the distance attenuation coefficient is a coefficient used to reflect the degree of influence of the straight-line distance between the target object and the target fogging window unit on the privacy exposure risk index. The greater the distance, the more obvious the risk attenuation with the effect of this coefficient.
[0062] If the privacy exposure risk index exceeds the preset risk threshold, the fogging decision result of the target fogging window unit can be further determined.
[0063] The preset risk threshold can be a critical value set in advance to determine whether to activate fogging. Exceeding this value indicates that indoor details may be clearly identifiable by external targets, requiring the fogging response to be activated. For example, based on the principle of visual discernibility, the critical point R = 0.7 of the privacy exposure risk index can be used as the preset risk threshold.
[0064] The fogging decision result can refer to the instructions and corresponding parameters that control the target fogging window unit to perform fogging operations. Specifically, when the privacy exposure risk index exceeds a preset risk threshold, the system can generate fogging instructions containing control parameters such as fogging level, fogging area, and duration based on the privacy exposure risk index of different target objects. Finally, it decides whether to fog up and how to fog up within 200ms before the target reaches the visual threshold, thereby achieving precise privacy protection.
[0065] Ultimately, the atomization decision results are used to drive the target atomization window unit to perform real-time atomization operations.
[0066] Real-time atomization operation refers to the process by which the target atomization window unit adjusts its light transmission state within a very short time before the target enters the visible range, based on the received atomization command. Specifically, the system can send the atomization decision result to the control module of the target atomization window unit via a CAN bus. After receiving the command, the control module drives the PDLC film voltage of the electronically controlled dimming glass unit to achieve a fast and accurate atomization effect.
[0067] Thus, by calling the Building Information Model and integrating the predicted viewing area with BIM spatial attributes, the system first quickly locks the target fogged window unit facing the risk of being spied on, then calculates the privacy exposure risk index. Once the threshold is exceeded, a decision result containing the overall or zone fogging level is generated. The PDLC layer is driven in real time via the CAN bus to complete gradient fogging before the target enters the visible range, achieving the beneficial effects of accurate risk quantification, accurate window response, and advance privacy shielding.
[0068] Optionally, to further improve data accuracy and system reliability, the intelligent fogging window system also includes an anti-false triggering unit, which specifically comprises an environment adaptation module and a whitelist filtering module. The environment adaptation module can dynamically adjust the detection threshold based on weather data; for example, it increases the trigger threshold of each sensor by 30% during rainy or foggy weather to reduce interference from adverse weather conditions. The whitelist filtering module integrates facial recognition and license plate recognition technologies to verify the identity of targets entering the target area, thereby filtering out residents, property vehicles, and other objects without privacy threats, reducing the probability of false triggering. Geofencing can also be set, allowing users to define sensitive areas requiring protection, such as bathrooms or meeting rooms, according to their needs.
[0069] Optionally, the intelligent atomizing window system is also equipped with arbitration rules and false trigger suppression steps in the event of conflicts between multimodal sensing sensors, in order to deal with inconsistent detection results from different sensors.
[0070] Arbitration rules can employ a type priority mechanism.
[0071] For example, if the thermal imaging camera alarms but the RGB-D camera detects no target, the thermal imaging is trusted first, but the following verification is initiated: checking whether the RGB-D camera is in low light or obstructed and whether the millimeter-wave radar detects a metallic object. When the millimeter-wave radar detects motion but the thermal imaging camera does not respond, the RGB-D camera is required to verify through contour features. If a suspected target is detected as a metallic object without life characteristics, the suspected target is marked as a non-living entity such as a vehicle or drone. When the RGB-D camera detects a suspected target but the thermal imaging camera does not respond, either of the following must be met: the millimeter-wave radar simultaneously detects the movement of the suspected target, or the suspected target is within the BIM whitelist area, before it can be marked as a non-living moving object.
[0072] The false trigger suppression steps are as follows: when the difference between the detection results of the millimeter-wave radar and the RGB-D camera, or the millimeter-wave radar and the thermal imaging camera, exceeds 15%, the system initiates a verification mechanism to re-acquire and analyze the data. This difference can refer to variations in any dimension, such as position coordinates, movement speed, or target size. If a suspected object is not detected by any sensor for five consecutive frames, the system ignores the corresponding fogging command to avoid false triggers caused by brief interference.
[0073] Optionally, the fogging window control method also supports automatic restoration of transparency after fogging operation. The conditions for restoring transparency include at least one of the following: the distance between the target object and the building is greater than 20 meters as detected by millimeter-wave radar; the target object remains in front of the sensitive area for a duration exceeding a system preset (default setting is 180 seconds); or the system receives a user-issued command to terminate fogging.
[0074] In the above implementation, by first locking the target and fusing multimodal data to complete identification and trajectory prediction, and then calling the building information model to convert the relative distance and motion trajectory into fogging decisions, the system can drive the intelligent fogging window to move in real time before the target enters the visible range, thereby achieving proactive privacy protection with advance shielding, low accidental touch, and accurate window area response.
[0075] In some implementation methods, please refer to the appendix. Figure 2 Based on the target recognition results and initial multimodal data, trajectory prediction processing is performed to obtain the target prediction results of the target object, including:
[0076] S210. If the object category indicates that the target object is a vehicle, obtain the real-time motion state data of the vehicle based on the initial multimodal data.
[0077] Real-time motion state data refers to a set of parameters that reflect the instantaneous motion characteristics of a vehicle, obtained through real-time acquisition and fusion processing by a sensor array. Specifically, real-time motion state data may include the vehicle's distance, speed, and azimuth angle.
[0078] For example, the system can calculate the vehicle's precise real-time motion state data by fusing radial velocity and distance information measured by millimeter-wave radar with lateral displacement and angle changes calculated by RGB-D camera through visual ranging and feature point tracking.
[0079] S220. Based on distance, speed, and azimuth, a Kalman filter model is used to predict the vehicle's steering to obtain the target prediction result.
[0080] Specifically, the initial state vector of the Kalman filter model can be constructed based on the real-time motion state data of the vehicle. Through the prediction and update cycle of the model, the error of the vehicle's motion state can be corrected in real time. Combined with the vehicle's steering dynamics characteristics, the position change in the next 3-5 seconds can be estimated. Finally, the target prediction result containing the predicted motion trajectory of the vehicle and the relative distance to the target building can be output.
[0081] For example, taking a vehicle driving on the road outside a villa and turning towards the living room window (its three-dimensional coordinates are stored in the BIM model) as an example, the execution process of Kalman filter prediction is as follows:
[0082] First, a vehicle dynamics model is constructed and initial states are set. Specifically, based on the Ackerman steering principle, a Kalman filter state vector is constructed, containing the vehicle's real-time position (X, Y), moving speed v, and azimuth angle α. The azimuth angle is the azimuth angle of the vehicle's heading; it can be set to 0° when heading due east and 180° when heading due west. Then, combined with the real-time motion state data obtained from the initial multimodal data, the initial values are set as follows: the vehicle's current position (15m, 8m), the initial speed corresponding to the distance is 8m / s, and the azimuth angle is 0°.
[0083] Real-time perception and data fusion correction are then performed. Specifically, at t=1.5s, the millimeter-wave radar detects that the vehicle's steering wheel is turning right and outputs a front wheel steering angle δ=30°. At the same time, the RGB-D camera calculates the actual wheel angle as 29.8° using the three-dimensional point cloud of the vehicle's wheel rim. At this point, the system fuses the two sets of data and takes a weighted average to eliminate the detection error of a single sensor, which is then used as the observation input for the Kalman filter.
[0084] Next, the prediction and update cycle of Kalman filtering is performed. Specifically, in the prediction phase, the vehicle's turning radius is first calculated based on the front wheel steering angle of 30°. For example, according to the Ackerman steering formula, the turning radius can be estimated to be approximately 4.33 meters. Combined with the current speed of 8 m / s, the vehicle's position change within the next 0.2 seconds is predicted. For example, the result is that the vehicle may move from (15m, 8m) to (14.8m, 7.8m), and due to the turning, the vehicle will slightly decelerate to 7.9 m / s, and the vehicle's azimuth angle will also change from 0° to 15°. Subsequently, in the update phase, the vehicle position acquired in real time by the millimeter-wave radar and RGB-D camera is compared with the predicted value. For example, if the measured vehicle position is (14.82m, 7.81m) and the speed is 7.88 m / s, the state vector can be corrected by calculating the Kalman gain to reduce the prediction error. Furthermore, the above prediction and update process is executed every 0.2 seconds for 8 cycles to dynamically track the state changes of the vehicle during the turning process.
[0085] The final output is the target prediction result. For example, at t = 1.8s (0.3 seconds after prediction begins), the system outputs the prediction result based on the latest corrected state vector: the vehicle will arrive at position (10.3m, 3.6m) at t = 3.0s (1.2 seconds later), at which point the relative distance to the villa's living room window W-01 will be 5.2m, and the trajectory will be directly opposite W-01. This result is the vehicle's target prediction result.
[0086] In the above implementation, the real-time motion state of the vehicle is obtained by fusing data from millimeter-wave radar and RGB-D camera. Then, the Kalman filter is combined with Ackerman steering dynamics to continuously correct and extrapolate the future trajectory, so that the system outputs the target prediction result containing the predicted position and the relative distance to the building in advance. This provides the prerequisite for triggering the corresponding fogging window decision 3-5 seconds in advance, and realizes the active pre-privacy protection function in the vehicle steering scenario.
[0087] In some implementation methods, please refer to the appendix. Figure 3 Based on the target recognition results and initial multimodal data, trajectory prediction processing is performed to obtain the target prediction results of the target object, including:
[0088] S310. When the object category indicates that the target object is a pedestrian, determine the pedestrian's orientation based on the initial multimodal data.
[0089] The orientation judgment result can refer to the conclusion of the pedestrian's facing direction or the direction the body tends to be, derived by analyzing the pedestrian's body geometric features and movement direction. This can include different results such as the pedestrian facing the target building, the pedestrian moving away from the target building, or the pedestrian moving sideways to the target building. Specifically, when determining the orientation judgment result based on initial multimodal data, the pedestrian's color image and 3D point cloud data acquired by the RGB-D camera can be extracted first from the initial multimodal data. Then, the pedestrian's facial features are identified to determine the facial orientation and the spatial posture of the shoulders and torso. Finally, the overall orientation of the pedestrian is comprehensively judged by combining the thermal distribution features of the pedestrian's torso acquired by the thermal imaging camera, thus obtaining the pedestrian's orientation judgment result. For example, a human posture estimation algorithm can be used to accurately locate the key points of the human's shoulders based on the initial multimodal data, and then connect the left and right shoulder key points to form a shoulder line vector. Since the human body's orientation is usually perpendicular to the shoulder line, the positive direction of the normal vector of the shoulder line vector can be used as the pedestrian's orientation judgment result. Alternatively, a vector machine can be used for orientation determination. First, feature joints related to the human skeletal posture, such as the hip, wrist, and ankle joints, are extracted from the initial multimodal data. Then, a unit vector perpendicular to the plane defined by the waist and hip joints is calculated. This vector, along with the coordinates of the feature joints, is used to calculate a feature vector for orientation determination. Subsequently, a vector machine algorithm is used for discrimination to obtain the pedestrian's orientation determination result. S320: If the orientation determination result indicates that the pedestrian is facing the target building, the trajectory of the target object is predicted to obtain the target prediction result.
[0090] Specifically, if the orientation judgment result indicates that the pedestrian is facing the target building, the motion trajectory of the target object can be predicted using a long short-term memory network model based on the temporal data of the skeletal key points, so as to obtain the target prediction result.
[0091] Among them, skeletal keypoint time-series data refers to a continuous data sequence extracted from initial multimodal data that reflects the position and movement state of skeletal joints of a pedestrian at different time points within a time series. It can reflect the gait characteristics and movement trends of the pedestrian, such as stride length and body tilt angle. Specifically, skeletal keypoint time-series data can include key data such as hip orientation angle, center of mass shift, and knee flexion angle, and this data is determined and tracked in real time based on the initial multimodal data through human pose estimation algorithms.
[0092] For example, taking a pedestrian walking on the sidewalk outside the target building and facing the villa as an example, the trajectory prediction process of the Long Short-Term Memory Network model can be performed as follows:
[0093] First, temporal data of skeletal key points are extracted and constructed. Then, a Long Short-Term Memory (LSTM) network model is input and its features are learned. Specifically, the constructed temporal data of skeletal key points is input into the LTM model. The model first performs feature analysis on the data, learning the basic gait patterns of pedestrians and capturing their movement trends. Simultaneously, the model utilizes the spatial coordinate system of the target building, combined with the relative distance between the pedestrian and the villa measured in the initial multimodal data, to establish a spatial reference for subsequent trajectory calculation.
[0094] Next, the pedestrian's trajectory is predicted and calculated. Specifically, the Long Short-Term Memory (LSTM) network model, based on learned gait patterns and movement trends, combined with the villa's spatial coordinate system, calculates the pedestrian's future position step by step along the time dimension. Simultaneously, the model's temporal prediction capabilities are used to confirm the direction vector of the pedestrian's trajectory.
[0095] The final output is the pedestrian target prediction result. For example, the system generates structured target prediction data based on trajectory calculation results. Specifically, it predicts the next step towards the window rather than the door by using the hip orientation angle and center of mass offset. For instance, with a hip orientation angle of 15°±3° and a center of mass offset towards the window by 0.2m / step, the system predicts the next step towards the window. This result is the pedestrian target prediction result.
[0096] Therefore, when a pedestrian is identified as a target object, the system first determines the pedestrian's orientation based on initial multimodal data. Only if the pedestrian is facing the target building will subsequent trajectory prediction processing be performed, thereby reducing the system's computational load and lowering the false trigger rate. During trajectory prediction, temporal data of skeletal key points are extracted and input into a long short-term memory network. The model learns pedestrian gait patterns and movement trends in a unified spatial coordinate system, outputting target prediction results containing movement trends and directions. This provides a prerequisite for subsequent fogging decisions, achieving proactive privacy protection in pedestrian scenarios.
[0097] It should be understood that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple steps or stages, which are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.
[0098] The control method of the proposed intelligent atomizing window system is illustrated below through a scenario example. Please refer to... Figure 4This control method can be applied to Figure 4 The intelligent fogging window linkage control system shown is based on video monitoring.
[0099] Specifically, the intelligent atomizing window linkage control system based on video surveillance includes a multimodal perception module, edge computing nodes, window database, trajectory prediction engine, multi-sensor fusion module, cloud management platform, electronically controlled dimming glass matrix, and atomization control feedback unit.
[0100] The multimodal perception module consists of a sensor array comprised of a thermal imaging camera, millimeter-wave radar, and an RGB-D depth camera, which detects pedestrians and vehicles outside the building in real time. The edge computing node utilizes the Jetson AGX Orin platform, running a lightweight YOLOv7-tiny model. It receives sensor data and uses the YOLOv7 object detection model to identify the object category (person or vehicle), then outputs a trajectory prediction engine for the next 3-5 seconds with a processing latency of less than 100ms, supporting 30FPS real-time video stream analysis. The window database connects to the building's BIM database, storing the physical coordinates and topological relationships of each glass unit, providing basic spatial data for trajectory prediction. The trajectory prediction engine, relying on the edge computing node and combining information from the window database, predicts the trajectory of the target object. The multi-sensor fusion module fuses various data from the multimodal perception module and provides calibration signals to the trajectory prediction engine, optimizing the trajectory prediction results for pedestrians, vehicles, and other target objects. The cloud management platform connects with the edge computing node via 5G / Wi-Fi, enabling remote system management and data storage backup. The electrically controlled dimming glass matrix consists of electrically controlled dimming glass units employing polymer dispersed liquid crystal (PDLC) technology, which can receive control commands to adjust the degree of fogging. The fogging control feedback unit collects the actual fogging state of the electrically controlled dimming glass units and feeds it back to the edge computing node, forming a closed-loop control.
[0101] During system operation, the multimodal perception module detects target objects such as pedestrians or vehicles. The multi-sensor fusion module fuses the data and sends it to the edge computing node. The edge computing node performs target detection and trajectory prediction. The dynamic fogging control module calculates a privacy exposure risk index based on the prediction results and window database data. When the index exceeds a threshold, a zoned fogging command is generated and executed through the electrically controlled dimming glass matrix. Simultaneously, the fogging control feedback unit provides real-time feedback, while the cloud management platform performs overall control, ensuring efficient and accurate system operation. Thus, through multi-sensor fusion and intelligent algorithms, accurate identification and trajectory prediction of moving targets such as people and vehicles outside the building are achieved. The system can proactively trigger gradient fogging through the electrically controlled dimming glass matrix, achieving all-weather, low-latency, and accurate privacy protection. Furthermore, cloud and edge collaboration ensures efficient system management and intelligent optimization.
[0102] This specification also provides a control device 500 for an intelligent atomizing window system, such as... Figure 5 As shown, it includes: an object determination module 510, a recognition and prediction module 520, and a fogging execution module 530, wherein:
[0103] The object determination module 510 is used to determine the target object within the target area of the target building and to obtain the initial multimodal data of the target object.
[0104] The identification and prediction module 520 is used to perform target identification and trajectory prediction processing based on the initial multimodal data to obtain target prediction data of the target object; wherein, the target prediction data includes the relative distance between the target object and the target building and the movement trajectory of the target object within a preset time range.
[0105] The atomization execution module 530 is used to call the building information model of the target building, determine the atomization decision result based on the relative distance and motion trajectory, and drive the intelligent atomization window system to perform real-time atomization actions.
[0106] In some implementations, the object determination module 510 is further used to acquire living organism attribute data, three-dimensional spatial feature data, and motion feature data; wherein, the living organism attribute data is used to describe the infrared radiation intensity and contour range of the target object; the three-dimensional spatial feature data is used to describe the appearance details and depth information of the target object; the motion feature data is used to describe the real-time motion trajectory and motion trend of the target object; and the living organism attribute data, three-dimensional spatial feature data, and motion feature data are subjected to time synchronization and data fusion processing to obtain initial multimodal data.
[0107] In some implementations, the identification and prediction module 520 is further configured to perform target identification processing on the initial multimodal data using a target detection model based on the YOLOv7 architecture to obtain the target identification result of the target object; wherein, the target identification result includes the object category of the target object;
[0108] Trajectory prediction is performed based on the target recognition results and initial multimodal data to obtain target prediction data for the target object.
[0109] In some embodiments, the control device 500 of an intelligent fogging window system further includes a trajectory prediction module, which is used to acquire real-time motion state data of the vehicle based on initial multimodal data when the object category indicates that the target object is a vehicle; wherein, the real-time motion state data includes the vehicle's distance, speed and azimuth angle; based on the distance, speed and azimuth angle, a Kalman filter model is used to predict the vehicle's steering to obtain the target prediction result.
[0110] In some implementations, the trajectory prediction module is also used to determine the pedestrian's orientation based on initial multimodal data when the object category indicates that the target object is a pedestrian; and to predict the target object's motion trajectory when the orientation judgment result indicates that the pedestrian is facing the target building, so as to obtain the target prediction result.
[0111] In some implementations, the trajectory prediction module is also used to predict the motion trajectory of the target object based on the temporal data of skeletal key points using a long short-term memory network model, so as to obtain the target prediction result; wherein, the temporal data of skeletal key points is determined based on the initial multimodal data.
[0112] In some implementations, the atomization execution module 530 is further configured to determine the predicted viewing area of the target object based on relative distance and motion trajectory; determine the target atomization window unit to be atomized and the privacy exposure risk index of the target atomization window unit from the intelligent atomization window system using the spatial attribute data of the building information model and the predicted viewing area; determine the atomization decision result of the target atomization window unit when the privacy exposure risk index exceeds a preset risk threshold; and drive the target atomization window unit to perform real-time atomization operation using the atomization decision result.
[0113] For specific limitations regarding the control device of an intelligent atomizing window system, please refer to the limitations of the control method for an intelligent atomizing window system described above, which will not be repeated here. Each module in the control device of the aforementioned intelligent atomizing window system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.
[0114] In this embodiment, the control device for an intelligent atomizing window system is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.
[0115] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application, such as... Figure 6As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 6 Take a processor 10 as an example.
[0116] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.
[0117] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.
[0118] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0119] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.
[0120] The computer device also includes an input device 30 and an output device 40. The processor 10, memory 20, input device 30, and output device 40 can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.
[0121] Input device 30 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 40 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.
[0122] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.
[0123] This application provides a computer program product including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method of any embodiment of this application.
[0124] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.
[0125] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.
[0126] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.
[0127] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0128] 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 1 One or more processes and / or boxes Figure 1 The 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 operate 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 functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxesFigure 1 The steps of the function specified in one or more boxes.
[0129] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0130] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.
[0131] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
[0132] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.
Claims
1. A control method for an intelligent atomizing window system, characterized in that, The intelligent atomizing window system is installed on the target building; the method includes: Within the target area of the target building, a target object is identified, and initial multimodal data of the target object is obtained. Based on the initial multimodal data, target recognition and trajectory prediction processing are performed to obtain target prediction data of the target object; wherein, the target prediction data includes the relative distance between the target object and the target building and the movement trajectory of the target object within a preset time range; The building information model of the target building is invoked, and the atomization decision result is determined based on the relative distance and the motion trajectory to drive the intelligent atomization window system to perform real-time atomization actions.
2. The method according to claim 1, characterized in that, The initial multimodal data is obtained through the following methods: Acquire biological entity attribute data, three-dimensional spatial feature data, and motion feature data; wherein, the biological entity attribute data is used to describe the infrared radiation intensity and contour range of the target object; the three-dimensional spatial feature data is used to describe the appearance details and depth information of the target object; and the motion feature data is used to describe the real-time motion trajectory and motion trend of the target object. The life form attribute data, the three-dimensional spatial feature data, and the motion feature data are time-synchronized and fused to obtain the initial multimodal data.
3. The method according to claim 1, characterized in that, The target recognition and trajectory prediction processing based on the initial multimodal data to obtain target prediction data for the target object includes: The initial multimodal data is processed using a target detection model based on the YOLOv7 architecture to obtain the target recognition result of the target object; wherein, the target recognition result includes the object category of the target object; Based on the target recognition results and the initial multimodal data, trajectory prediction processing is performed to obtain the target prediction data of the target object.
4. The method according to claim 3, characterized in that, The trajectory prediction processing based on the target recognition result and the initial multimodal data to obtain the target prediction result of the target object includes: If the object category indicates that the target object is a vehicle, real-time motion state data of the vehicle is obtained based on the initial multimodal data; wherein, the real-time motion state data includes the vehicle's distance, speed, and azimuth angle; Based on the distance, speed, and azimuth angle, a Kalman filter model is used to predict the vehicle's steering direction in order to obtain the target prediction result.
5. The method according to claim 3, characterized in that, The trajectory prediction processing based on the target recognition result and the initial multimodal data to obtain the target prediction result of the target object includes: If the object category indicates that the target object is a pedestrian, the orientation judgment result of the pedestrian is determined based on the initial multimodal data; If the orientation determination result indicates that the pedestrian is facing the target building, the motion trajectory of the target object is predicted to obtain the target prediction result.
6. The method according to claim 5, characterized in that, The step of predicting the motion trajectory of the target object to obtain the target prediction result includes: Based on the skeletal key point temporal data, a long short-term memory network model is used to predict the motion trajectory of the target object to obtain the target prediction result; wherein, the skeletal key point temporal data is determined based on the initial multimodal data.
7. The method according to claim 1, characterized in that, The step of calling the building information model of the target building, determining the atomization decision result based on the relative distance and the movement trajectory, and driving the intelligent atomization window system to perform real-time atomization actions includes: The predicted line-of-sight area of the target object is determined based on the relative distance and the motion trajectory; Using the spatial attribute data of the building information model and the predicted viewing area, the target fogging window unit that needs to perform fogging action and the privacy exposure risk index of the target fogging window unit are determined from the intelligent fogging window system; If the privacy exposure risk index exceeds a preset risk threshold, the fogging decision result of the target fogging window unit is determined. The atomization decision result is used to drive the target atomization window unit to perform real-time atomization operation.
8. A control device for an intelligent atomizing window system, characterized in that, The intelligent atomizing window system is installed on the target building; the device includes: An object determination module is used to determine a target object within the target area of the target building and to acquire initial multimodal data of the target object; The identification and prediction module is used to perform target identification and trajectory prediction processing based on the initial multimodal data to obtain target prediction data of the target object; wherein, the target prediction data includes the relative distance between the target object and the target building and the movement trajectory of the target object within a preset time range; The atomization execution module is used to call the building information model of the target building, determine the atomization decision result based on the relative distance and the motion trajectory, and drive the intelligent atomization window system to perform real-time atomization actions.
9. A computer device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.
Citation Information
Cited By
Remote control method and system for man-machine interaction type gradient dimming film
CN121857214A
