Control method and system based on environmental perception

Abstract

This invention relates to the field of signage control technology, and discloses a control method and system based on environmental perception. The method includes: collecting effective point cloud data of the viewing area in front of the smart signage; identifying the human target of the viewer based on the effective point cloud data, and simultaneously calculating the viewer's behavior prediction data; detecting the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light based on the behavior prediction data, and performing event trigger condition judgment to obtain a trigger control signal; calculating display control parameters based on the trigger control signal, synchronously adjusting the pulse width modulation brightness of the smart signage and switching the display content. This invention can predict future behavior based on the viewer's historical movement trajectory, effectively shortening the response time of the smart signage, and achieving an optimal balance between display effect and energy consumption.

Description

Technical Field

[0001] This invention relates to the field of signage control technology, and in particular to a control method and system based on environmental perception. Background Technology

[0002] Traditional smart signage control systems rely on simple sensor detection and fixed threshold judgment mechanisms, employing a passive response display control mode. This makes them unable to accurately perceive the location, number, and behavioral state of viewers, resulting in a temporal and spatial misalignment between the displayed content and the actual needs of the viewers. Furthermore, traditional smart signage control systems lack the ability to predict viewer behavior, only adjusting display parameters after the viewer reaches the designated location. This delayed response mode leads to significant display adjustment delays, failing to provide viewers with timely and accurate information services. Summary of the Invention

[0003] This invention provides a control method and system based on environmental perception. This invention can predict future behavior based on the viewer's historical movement trajectory, effectively shorten the response time of smart signs, and achieve an optimal balance between display effect and energy consumption.

[0004] In a first aspect, the present invention provides a control method based on environmental perception, the control method based on environmental perception comprising: Collect effective point cloud data of the viewing area in front of the smart signage; The viewer's human body targets are identified based on the effective point cloud data, and the viewer's behavior prediction data is calculated simultaneously. Based on the behavior prediction data, the rate of change in pedestrian density, the rate of change in viewing distance, and the rate of change in ambient light are detected, and event triggering conditions are judged to obtain triggering control signals; Based on the trigger control signal, the display control parameters are calculated, and the pulse width modulation brightness of the smart sign is adjusted synchronously and the display content is switched.

[0005] In conjunction with the first aspect, in a first implementation of the first aspect of the present invention, the acquisition of effective point cloud data of the viewing area in front of the smart signage includes: Laser pulses are emitted and distances are measured in the viewing area in front of the smart sign, reflected laser signals are received, and raw point cloud data is calculated based on the reflected laser signals. For each coordinate point in the original point cloud data, a search range for neighboring points is established, and the average distance from each coordinate point to its neighboring points is calculated based on the search range. Based on the average distance, noise points are removed from the original point cloud data to obtain denoised point cloud data; The denoised point cloud data is input into a three-dimensional pass-through filter for spatial range limitation processing to obtain spatially clipped point cloud data. The spatially clipped point cloud data is then compressed to obtain effective point cloud data.

[0006] In conjunction with the first aspect, in a second implementation of the first aspect of the present invention, the step of identifying the viewer's human body target based on the effective point cloud data and simultaneously calculating the viewer's behavior prediction data includes: Cluster analysis is performed on the effective point cloud data to obtain point cloud cluster groups, and the height span, width span, depth span, and point cloud density value of each cluster in the point cloud cluster group are calculated. The height span value, the width span value, the depth span value, and the point cloud density value are input into a support vector machine classifier to perform human target determination and identify the human target of the viewer; Assign an identification code to each human target of the viewer, and extract the three-dimensional coordinate position and movement speed components of each human target to form an initial state vector; The initial state vector is substituted into the extended Kalman filter algorithm for state prediction and observation update to obtain motion state data for each viewer. A predicted trajectory is generated based on the motion state data, and the viewer's behavior prediction data is calculated based on the predicted trajectory.

[0007] In conjunction with the first aspect, in a third implementation of the first aspect of the present invention, the step of inputting the height span value, the width span value, the depth span value, and the point cloud density value into a support vector machine classifier to perform human target determination and identify the viewer's human target includes: The height span, width span, depth span, and point cloud density values ​​of each cluster are arranged in order to form a standardized feature vector; The standardized feature vector is input into the radial basis function kernel of the support vector machine classifier for high-dimensional space mapping to obtain a high-dimensional feature space vector; The classification decision function value is calculated based on the distance between the high-dimensional feature space vector and the pre-stored support vector, and a binary classification decision is performed on the classification decision function value through sign judgment to obtain the classification result of each cluster. Based on the classification results, clusters of positive categories are selected as candidate targets, and the geometric dimensions of the candidate targets are verified to conform to the range of human body characteristics, thus obtaining the human body target of the viewer.

[0008] In conjunction with the first aspect, in a fourth implementation of the first aspect of the present invention, the step of generating a predicted trajectory based on the motion state data and calculating the viewer's behavior prediction data based on the predicted trajectory includes: The three-dimensional coordinate position and movement speed components of each viewer in the motion state data are used to form a feature input sequence; Motion state prediction is performed based on the aforementioned feature input sequence to obtain a prediction vector; The three-dimensional position coordinates of the viewer are parsed based on the predicted vector, and a predicted trajectory is formed based on the three-dimensional position coordinates. The velocity component of the viewer's movement on the horizontal plane is calculated based on the predicted trajectory, and the velocity vector synthesis operation is used to obtain the predicted velocity parameters of the viewer's movement toward the smart signboard. The speed prediction parameters are substituted into the exponential decay model to calculate the dwell time, and combined with the arrival time of the viewer to the optimal viewing position in the predicted trajectory for comprehensive analysis to obtain behavioral prediction data.

[0009] In conjunction with the first aspect, in a fifth implementation of the first aspect of the present invention, the step of detecting the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light based on the behavior prediction data, and performing event triggering condition judgment to obtain a triggering control signal, includes: Based on the behavioral prediction data, the total number of viewers in the viewing area at the current moment is counted and compared with the number of viewers at the previous moment to calculate the rate of change in crowd density; The straight-line distance from each viewer to the smart signboard is calculated based on the three-dimensional position coordinates of each viewer in the behavior prediction data, and the viewing distance change rate is calculated based on the straight-line distance. The current ambient light intensity data is collected by a light sensor, and the rate of change of ambient light intensity is calculated by comparing it with historical light intensity. Based on the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light, the event triggering condition is determined to obtain the triggering control signal.

[0010] In conjunction with the first aspect, in the sixth implementation of the first aspect of the present invention, the step of performing event triggering condition judgment based on the change rate of pedestrian density, the change rate of viewing distance, and the change rate of ambient light to obtain a triggering control signal includes: A first threshold parameter is set for the rate of change of pedestrian density, a second threshold parameter is set for the rate of change of viewing distance, and a third threshold parameter is set for the rate of change of ambient light. The rate of change of pedestrian density is compared with the first threshold parameter to obtain a first comparison result; the rate of change of viewing distance is compared with the second threshold parameter to obtain a second comparison result; and the rate of change of ambient light is compared with the third threshold parameter to obtain a third comparison result. Based on the first comparison result, the second comparison result and the third comparison result, perform a logical OR operation to determine the output of a trigger state flag or a stable state flag. When the trigger status flag is valid, it switches to fast response mode and outputs a high-frequency sampling signal as the trigger control signal. When the stable status flag is valid, it maintains the standard control mode and outputs a regular sampling signal as the trigger control signal.

[0011] In conjunction with the first aspect, in the seventh implementation of the first aspect of the present invention, the step of calculating display control parameters based on the trigger control signal, synchronously adjusting the pulse width modulation brightness of the smart signboard, and switching the display content includes: Based on the trigger control signal, the display brightness value, display content identifier and refresh rate parameter of the current smart sign are collected and combined to form a display status vector; Substitute the displayed state vector into the preset state transition equation to predict the state, and combine the control input matrix and output matrix to establish a state space model. Based on the state-space model, a weighted quadratic cost function containing visibility error, power consumption cost, and response time cost is constructed. The display brightness range constraint, refresh rate range constraint, and total power consumption constraint of the weighted quadratic cost function are set to obtain the constrained optimization problem. The constrained optimization problem is solved to obtain the explicit control parameters; Adjust the pulse width modulation brightness of the smart signboard and switch the displayed content according to the display control parameters.

[0012] In conjunction with the first aspect, in the eighth implementation of the first aspect of the present invention, the step of adjusting the pulse width modulation brightness of the smart signboard and switching the display content according to the display control parameters includes: The brightness adjustment amount is extracted from the display control parameters and weighted by the sum of viewer weights, ambient light correction coefficient, and viewer number correction coefficient to obtain the target brightness value. Based on the target brightness value, it is converted into a pulse width modulation duty cycle according to a linear mapping relationship, and a pulse width modulation waveform signal is generated based on the pulse width modulation duty cycle and input to the LED driver to adjust the current magnitude, thereby obtaining the display output after brightness adjustment; Based on the content switching instructions in the display control parameters, the product of importance, urgency and viewer relevance is calculated for each candidate display content to obtain a priority score. The display content with the highest priority score is selected and the display buffer is switched through the digital signal processor to obtain the updated display output.

[0013] Secondly, the present invention provides an environment-aware control system, the environment-aware control system comprising: The data acquisition module is used to collect effective point cloud data of the viewing area in front of the smart signage; The recognition module is used to identify the human body target of the viewer based on the effective point cloud data, and at the same time calculate the viewer's behavior prediction data; The judgment module is used to detect the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light based on the behavior prediction data, and to perform event trigger condition judgment to obtain a trigger control signal; The adjustment module is used to calculate display control parameters based on the trigger control signal, and synchronously adjust the pulse width modulation brightness of the smart sign and switch the display content.

[0014] The technical solution provided by this invention employs LiDAR for 3D point cloud data acquisition. Compared to traditional 2D visual perception methods, it can acquire accurate depth information and 3D spatial coordinates, eliminating the impact of lighting conditions on detection accuracy and achieving all-weather, high-precision environmental perception capabilities. Through the combined application of the DBSCAN density clustering algorithm and support vector machine classifier, it can accurately distinguish human targets from other objects, avoiding false detections and missed detections in traditional methods. Simultaneously, the introduction of an extended Kalman filter algorithm enables stable tracking of multiple targets and can predict future behavior based on the viewer's historical movement trajectory, effectively shortening the response time of smart signage and improving the matching degree between displayed content and viewer needs. A multi-parameter fusion dynamic event triggering system is established. By monitoring the combined changes in pedestrian density, viewing distance, and ambient light, adaptive switching of control strategies is achieved, avoiding the limitations of fixed threshold triggering methods and improving adaptability to complex dynamic environments. A multi-objective optimization framework including visibility, power consumption, and response time is established, achieving an optimal balance between display effect and energy consumption while meeting system constraints. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram illustrating the steps of the control method based on environmental perception in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the environmental perception-based control system in an embodiment of the present invention. Detailed Implementation

[0017] This invention provides a control method and system based on environmental perception. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the control method based on environmental perception in this invention includes: Step S1: Collect valid point cloud data of the viewing area in front of the smart signboard; Specifically, a LiDAR sensing module with 3D perception capabilities is constructed. The LiDAR continuously emits laser pulses at a preset scanning frequency towards the viewing area in front of the smart signage. By measuring the time difference between the laser pulse emission and the received reflected signal, combined with the laser propagation speed, the spatial distance of the laser beam hitting the surface of the environmental object is calculated in real time, constructing an original point cloud data frame containing a large number of 3D coordinate points. A k-neighborhood search range is established for each coordinate point in the original point cloud data. During implementation, the number of neighboring points k is set to 50. The average Euclidean distance from each point to all neighboring points is calculated to obtain the average neighborhood distance of the current point. The average distance distribution characteristics of all points are statistically analyzed globally. Based on the average distance, a statistical outlier removal algorithm is used with a standard deviation threshold. When the average neighborhood distance of a point exceeds the overall average distance plus a certain multiple of the standard deviation, the current point is determined to be a noise point and removed. The denoised point cloud data after noise removal is input into a 3D pass-through filter, with constraints set in the spatial dimension, with the x-axis direction... A finite perception space is constructed with a distance of 6 meters to 6 meters, 0 meters to 3 meters along the y-axis, and 0 meters to 10 meters along the z-axis. All point cloud data outside this range is filtered out, and only point cloud points located within the three-dimensional solid area are retained. After spatial clipping, the point cloud data is compressed using uniform downsampling or voxel meshing algorithms to reduce the number of redundant points while preserving the original structure and morphological features, thus obtaining effective point cloud data.

[0019] Step S2: Identify the viewer's human body target based on the effective point cloud data, and simultaneously calculate the viewer's behavior prediction data; Specifically, valid point cloud data is used as input, and spatial structure analysis is performed on the point cloud using density clustering algorithms. Density-based spatial clustering methods such as DBSCAN are used, leveraging DBSCAN's ability to effectively identify irregularly shaped high-density regions and remove isolated noise points. This automatically clusters different dense regions in the point cloud data, generating multiple point cloud clusters. Geometric parameters are calculated for each cluster. The maximum coordinate difference in the vertical direction is extracted as the height span, and the maximum coordinate differences in the horizontal and depth directions are used as the width span and depth span, respectively. The point cloud density per unit volume is also calculated. These four parameters together constitute a feature vector describing the spatial morphology of each cluster. This geometric feature vector is then input into a support vector machine (SVM) classifier. The SVM classifier, trained on a large amount of human and non-human point cloud data, uses nonlinear kernel function mapping to perform high-dimensional discrimination and classification of the input data, determining whether each cluster represents a human target and identifying clusters belonging to the viewer. For each successfully identified human target, a unique identification code is assigned, and the three-dimensional coordinate center position and velocity changes between consecutive frames are extracted to form an initial state vector containing six dimensions of position and velocity. This initial state vector is used as initial input and fed into an extended Kalman filter algorithm. Through two stages—state prediction and observation update—the algorithm iteratively calculates the target's motion state data for future moments. State prediction estimates future position and velocity based on a system state transition model, while observation update integrates current observations to correct prediction biases and obtain a current state estimate. Based on the filtered motion state data, a behavior prediction module is constructed. A continuous state sequence is input into a trajectory prediction model, such as an LSTM neural network, to generate a predicted motion trajectory for a future period. Based on this, behavior prediction data is calculated, including approach speed, expected dwell time, and estimated time to reach the optimal viewing position.

[0020] Step S3: Detect the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light based on the behavior prediction data, and perform event trigger condition judgment to obtain the trigger control signal; Specifically, based on behavioral prediction data, all human targets within the viewing area of ​​the smart signage are statistically analyzed. According to each target's identification code and valid presence status, the total number of people identified in the current frame is aggregated and calculated by differentiating it from the total number of people in the previous moment. Combined with the sampling interval, the rate of change in crowd density is obtained, quantifying the changing trend of crowd numbers per unit time. Based on the three-dimensional position coordinates of each viewer, using the spatial reference point of the smart signage as the origin, the Euclidean distance formula is used to calculate the straight-line distance between each target and the signage. The arithmetic mean of all viewer distances is taken to form the average viewing distance at the current moment. This average viewing distance is then calculated by differentiating it from the average viewing distance in the previous moment to obtain the rate of change in viewing distance, measuring whether the viewing crowd is moving forward or backward. Simultaneously, a light sensor integrated into the signage structure collects the real-time light intensity of the current environment and reads the light value from the previous moment from the environmental monitoring cache to calculate the current rate of change in light intensity, reflecting the rapid changes in external ambient brightness. The system compares each rate of change with a preset trigger threshold. When any one of the following is detected to exceed the threshold: the rate of change in pedestrian density exceeds 5 people per second, the rate of change in viewing distance exceeds 0.5 meters per second, or the rate of change in ambient light intensity exceeds 500 lux per second, the event determination function is triggered. The control system issues a trigger control signal, instructing the system to switch from the normal stable mode to the fast response mode, and initiate actions such as high-frequency sampling, short-term prediction, and high-priority content scheduling.

[0021] Step S4: Calculate the display control parameters based on the trigger control signal, and synchronously adjust the pulse width modulation brightness of the smart sign and switch the display content.

[0022] Specifically, based on the trigger control signal, the system collects the current display brightness value of the smart signage, the identifier code of the currently displayed content, and the corresponding refresh rate parameter. These three parameters are then assembled into a display state vector, which serves as the initial state for the model's predictive control input. The display state vector is input into a preset state transition mechanism to predict the next state. A state-space model that dynamically describes the evolutionary relationship between the display state and control behavior is established by combining the control input matrix and output matrix. Based on the state-space model, a weighted quadratic cost function is constructed, incorporating visibility error, power consumption cost, and response time cost. This weighted quadratic cost function assigns differentiated weights to the three performance indicators during design and introduces control variation and relaxation terms to suppress abrupt changes and enhance constraint handling capabilities. Furthermore, multiple strong constraints are set for the optimization objective. Display brightness is limited to within the range of human visual comfort, the refresh rate is limited to a reasonable range that conforms to visual smoothness, and the overall system power consumption must not exceed a set maximum power consumption threshold. Based on the constraint optimization solution module, the cost function and constraints are solved iteratively to output a set of optimal display control parameters, including brightness adjustment, content switching commands, and refresh rate adjustment. According to the control parameters, the duty cycle signal of the LED array is adjusted using pulse width modulation to regulate the display brightness. Simultaneously, the content switching control signal is transmitted to the display content buffer unit, and a gradual transition strategy is used to smoothly switch to new content, achieving rapid response and multi-objective collaborative control of the smart signage display system.

[0023] In one specific embodiment, the process of performing step S1 may specifically include the following steps: Laser pulse emission and distance measurement are performed on the viewing area in front of the smart sign, the reflected laser signal is received, and the original point cloud data is calculated based on the reflected laser signal; For each coordinate point in the original point cloud data, establish the search range of neighboring points, and calculate the average distance from each coordinate point to its neighboring points based on the search range; Based on the average distance, noise points are removed from the original point cloud data to obtain denoised point cloud data. The denoised point cloud data is input into a 3D pass-through filter for spatial range limitation processing to obtain spatially clipped point cloud data. The spatially clipped point cloud data is then compressed to obtain effective point cloud data.

[0024] Specifically, an active spatial scanning process is performed using a lidar device deployed on the smart signage. The lidar device emits high-density laser pulse signals at a set frequency towards the target area and continuously monitors the laser echo signals reflected from the surfaces of objects in front. By measuring the round-trip flight time of each laser beam from emission to reception of the reflected signal, and combining this with the laser propagation speed, distance is calculated to determine the spatial position of each illuminated object in the environment. The three-dimensional coordinate information is then integrated into a raw point cloud data frame containing several discrete points, forming the original spatial structure representation of the environment in front of the smart signage. For each three-dimensional coordinate point in the raw point cloud data, a neighborhood search range is sequentially established. The neighborhood range is determined by setting a fixed number of nearest neighbors or all neighboring points within a specified radius, employing a fixed k-value neighborhood strategy and using a k-dimensional spatial index structure to accelerate retrieval efficiency. After establishing the neighborhood set, the Euclidean distance from the current point to all points within the neighborhood is calculated, and their average value is obtained to obtain the average neighborhood distance characteristic of the current point. After calculating the average distance of all points, noise points and outliers in the point cloud are identified based on global statistical characteristics. An overall distribution model of the average neighborhood distance values ​​of all points is constructed, and a dynamically adjustable outlier threshold is set. When the average distance of a point deviates from the overall distribution mean and exceeds the set upper limit of the standard deviation, the current point is determined to be a noise point or an isolated point, and is removed from the point cloud data. This achieves preliminary purification of the original data, outputting denoised point cloud data that retains the main spatial structure information. The denoised point cloud data is then input into a 3D pass-through filter for spatial clipping. The value range of the entire point cloud data is limited by preset upper and lower boundary conditions along the coordinate axes. The x-axis direction is set as... The distance is set from 6 meters to 6 meters, with the y-axis range set from 0 to 3 meters and the z-axis range set from 0 to 10 meters. Only point cloud points within the effective spatial area are retained, while all points outside the visible range of the sign are filtered out, resulting in spatially cropped point cloud data that matches the spatial structure of the smart sign's viewing area. The spatially cropped point cloud data undergoes data compression processing. A uniform voxel grid method or other voxel downsampling method is used to grid the original point cloud space. A representative point is selected within each grid as a sample output, effectively reducing the number of point clouds while preserving the spatial distribution morphology and boundary features, thus forming effective point cloud data.

[0025] In one specific embodiment, the process of performing step S2 may specifically include the following steps: Cluster analysis is performed on the effective point cloud data to obtain point cloud cluster groups, and the height span, width span, depth span and point cloud density values ​​of each cluster in the point cloud cluster group are calculated; The height span value, width span value, depth span value, and point cloud density value are input into the support vector machine classifier to perform human target determination and identify the human target of the viewer; Assign an identification code to each human target of the viewer, and extract the three-dimensional coordinate position and movement speed components of each human target to form an initial state vector; The initial state vector is substituted into the extended Kalman filter algorithm for state prediction and observation update to obtain the motion state data of each viewer. Predicted trajectories are generated based on motion state data, and viewer behavior prediction data is calculated based on the predicted trajectories.

[0026] Specifically, cluster analysis is performed on the effective point cloud data to distinguish different spatial target distributions. Density-based clustering algorithms such as DBSCAN are used. DBSCAN has the advantages of not requiring a pre-set number of clusters and automatically identifying noise points, making it suitable for processing irregularly shaped human contour point cloud structures in smart signage scenarios. During the clustering operation, a neighborhood relationship is constructed for each point cloud point in three-dimensional space, and point groups are classified according to local density and a minimum point count threshold, forming multiple point cloud clusters with clear boundaries and independence. For each cluster, its geometric morphological features are calculated. The maximum and minimum coordinate difference in the z-axis direction is used as the height span value to describe the target's body shape distribution in the vertical direction. The coordinate span values ​​in the x and y-axis directions are used as the width span value and depth span value, respectively, to characterize the target's lateral width and depth dimensions on the horizontal plane. Simultaneously, the ratio of the number of points in the cluster to the volume space is calculated to obtain the point cloud density value, reflecting the density of the target in a unit space. These four parameters together form the feature vector. The aforementioned feature vectors are input one by one into a pre-trained Support Vector Machine (SVM) classifier for classification. The SVM classifier constructs a training set using a large number of real human and non-human point cloud samples, and achieves high-dimensional feature space mapping based on a nonlinear kernel function, thereby enabling the discrimination of complex point cloud structures. The classifier outputs a human or non-human label for each cluster, thus identifying all human targets in the current point cloud data. For each identified human target, a unique identification code is assigned to support continuous tracking at the individual level, and the three-dimensional coordinates of representative center points are extracted from its cluster as spatial position data. Combined with the position change between the current frame and the previous frame, the velocity vector is estimated to obtain an initial state vector containing six dimensions of position and velocity information. The initial state vector is input into the Extended Kalman Filter (EKF) algorithm to perform two processes: state prediction and observation update. State prediction calculates the human target position information for the next time step based on the current state estimate, while observation update uses the current frame observations to correct the prediction values, forming a real-time motion state estimation result, and outputting the three-dimensional coordinate position and three-dimensional velocity components of each individual at the current moment. Based on dynamic state data sequences, a behavior prediction module is constructed. By using a sliding time window to collect the historical state sequence of each individual, and by inputting it into a trajectory prediction model, such as an LSTM neural network, a predicted motion trajectory curve for several future time steps is generated. Multiple key behavioral indicators are extracted from the predicted motion trajectory, including whether the individual continues to approach the sign, the time required to reach the best viewing position, and whether there is a tendency to stay.

[0027] In one specific embodiment, the execution step of inputting the height span value, width span value, depth span value, and point cloud density value into a support vector machine classifier to perform human target determination, and the process of identifying the viewer's human target, can specifically include the following steps: The height span, width span, depth span, and point cloud density values ​​of each cluster are arranged in order to form a standardized feature vector; The standardized feature vectors are input into the radial basis function kernel of the support vector machine classifier for high-dimensional space mapping to obtain high-dimensional feature space vectors; The classification decision function value is calculated based on the distance between the high-dimensional feature space vector and the pre-stored support vector, and a binary classification decision is performed on the classification decision function value through sign judgment to obtain the classification result of each cluster. Based on the classification results, clusters of positive categories are selected as candidate targets, and the geometric dimensions of the candidate targets are verified to conform to the range of human body characteristics, thus obtaining the human body target for the viewer.

[0028] Specifically, from the point cloud data after clustering analysis, the height span, width span, depth span, and point cloud density values ​​of each cluster are extracted. These four parameters correspond to the dimensional characteristics of the cluster in the vertical, horizontal, and depth dimensions, as well as the spatial density of the points. These four values ​​are arranged in a uniform order to form the original feature vector. All feature vectors are then standardized using mean normalization or range normalization to achieve uniform scaling, ensuring that each feature value falls within a specified range, thus forming a standardized feature vector set. These standardized feature vectors are then input one by one into a support vector machine (SVM) classifier. Inside the classifier, a radial basis function kernel performs a nonlinear transformation on each input vector, mapping it from the original four-dimensional feature space to a higher-dimensional feature space. In the higher-dimensional space, the SVM constructs an optimal hyperplane to divide the data into two classes with the maximum margin, thus modeling the boundaries of complex morphological categories. For each input vector, its mapping result in high-dimensional space is distance-calculated with the support vector set formed during training. A classification decision function is constructed based on the kernel function value, reflecting the distribution position of the current input vector relative to each type of support vector in the feature space. After the decision function calculation is completed, a sign function is used to perform logical judgment on the result. When the output of the decision function is positive, the input cluster is determined to be a positive category, i.e., a human target; if it is negative, it is determined to be a non-human cluster, thus completing the binary classification process. The classification result serves as the initial recognition output. A second-stage geometric size verification is performed on all clusters determined to be positive, i.e., comparing whether their height span, width span, and depth span are within the preset reasonable range of human body dimensions. For example, whether the height span is between 1.3 meters and 2.2 meters, whether the width is between 0.3 meters and 1 meter, and whether the point cloud density meets the spatial filling level required for human body structure. Only when a cluster simultaneously satisfies the conditions of a positive classification result and geometric size parameters falling within the standard range of human body dimensions is it confirmed as a human target for the viewer.

[0029] In one specific embodiment, the process of generating a predicted trajectory based on motion state data and calculating the viewer's behavior prediction data based on the predicted trajectory may specifically include the following steps: The three-dimensional coordinate position and movement speed components of each viewer in the motion state data are used to form a feature input sequence; Motion state prediction is performed based on the feature input sequence to obtain the prediction vector; The viewer's three-dimensional position coordinates are parsed from the prediction vector, and a predicted trajectory is formed based on the three-dimensional position coordinates. The velocity component of the viewer's movement on the horizontal plane is calculated based on the predicted trajectory, and the predicted velocity parameters of the viewer's movement toward the smart sign are obtained through velocity vector synthesis calculation. By substituting the speed prediction parameters into the exponential decay model to calculate the dwell time, and combining this with the arrival time of the viewer to the optimal viewing position in the predicted trajectory for comprehensive analysis, behavioral prediction data is obtained.

[0030] Specifically, the three-dimensional coordinates and velocity components of each identified viewer are extracted from the motion state data. These two six-dimensional parameters form a temporally arranged feature input sequence, organized chronologically to create a dynamic feature stream describing the spatial position evolution and velocity changes of each target individual over consecutive time points. This feature input sequence is then fed into the state prediction model, where a recurrent neural network structure with time-series modeling capabilities, such as a two-layer LSTM network, is selected. Through its memory units and gating mechanisms, the model performs deep abstraction and dynamic feature extraction of historical state sequences. Since parameter optimization has been completed based on a large number of real trajectory samples during the training phase, it can make highly reliable predictions of the input sequence, outputting a predicted state vector sequence for several future time points. The three-dimensional coordinates corresponding to each time point are progressively parsed from the prediction output and connected chronologically to form a predicted trajectory with clear temporal attributes and spatial continuity. Based on the predicted trajectory, distance differences are calculated for the horizontal plane projections between each predicted point. Combined with the sampling time interval, the horizontal velocity components within each time period are calculated. Then, through velocity vector synthesis, the horizontal velocity vectors are modulo-processed to extract the overall movement velocity prediction parameters of the viewer relative to the smart signage on the horizontal plane. This reflects whether the individual is actively approaching the signage and measures the speed level of their approach. The velocity prediction parameters are input into a preset exponential decay model. This model uses velocity as the input variable, mapping higher speeds to shorter expected dwell times and lower speeds to longer expected dwell times, establishing a functional relationship between velocity and dwell intention. Based on this, the corresponding predicted dwell time value is calculated. Simultaneously, the spatial distance between the viewer's current position and the optimal viewing point in the predicted trajectory is divided by the predicted velocity to obtain the estimated arrival time, reflecting when the viewer will enter the optimal viewing distance area. A comprehensive analysis of the expected dwell time and arrival time determines whether the viewer has a stable viewing behavior tendency, is about to enter a key visible area, or is likely to quickly leave the signage area. The output includes behavioral prediction data including viewing intention intensity, dwell intention level, estimated arrival time, and viewing persistence score.

[0031] In one specific embodiment, the process of performing step S3 may specifically include the following steps: The rate of change in crowd density is calculated by statistically analyzing the total number of viewers in the viewing area at the current moment and comparing it with the number of viewers at the previous moment. The straight-line distance from each viewer to the smart sign is calculated based on the three-dimensional position coordinates of each viewer in the behavior prediction data, and the rate of change of viewing distance is calculated based on the straight-line distance. The current ambient light intensity data is collected by a light sensor, and the rate of change of ambient light intensity is calculated by comparing it with historical light intensity. Event triggering conditions are determined based on the rate of change in pedestrian density, the rate of change in viewing distance, and the rate of change in ambient light, respectively, to obtain triggering control signals.

[0032] Specifically, the system counts all viewers currently within the visible area in front of the smart signage. Based on the list of valid individual states output by the behavior prediction module, each active target is identified and numbered, and the total number of people is counted in the time series. This count is then compared with the number of viewers recorded at the previous sampling time, and the difference is calculated by dividing by the sampling time interval to determine the rate of change in crowd density, reflecting the increase or decrease in the number of people in the viewing area per unit time. Spatial location data is extracted from the three-dimensional coordinates of each viewer, and the reference point of the smart signage display panel is set as the target benchmark. The straight-line distance between each viewer and the signage is calculated using the geometric distance formula, and the distance values ​​of all viewers are statistically averaged to form the average viewing distance for the current frame. The difference between the current average viewing distance and the average value at the previous moment, combined with the time interval, yields the rate of change in viewing distance, used to determine whether the crowd is generally approaching the signage area or is dispersing away. Simultaneously, an ambient light sensor integrated into the signage's housing acquires current ambient brightness data and calculates the time difference by accessing historical brightness values ​​stored in the internal cache to obtain the ambient light change rate. This reflects whether there has been a sudden change in ambient lighting conditions, such as a shift from sunny to cloudy or from daylight to nighttime. The change rates of pedestrian density, viewing distance, and light intensity are compared with preset response thresholds. If any change rate exceeds its corresponding upper limit—for example, a change rate of pedestrian density exceeding 5 people per second, a change rate of viewing distance greater than 0.5 meters per second, or a change rate of light intensity greater than 500 lux per second—it is considered that there is a dynamic change in the current external environment. In this case, the logic judgment function outputs true, generating a trigger control signal and immediately switching the control mode to a fast response state. This increases the sampling frequency, shortens the prediction time domain, and dynamically adjusts the display parameters and updates the content according to the control commands.

[0033] In one specific embodiment, the process of determining the event triggering conditions based on the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light, and obtaining the triggering control signal, can specifically include the following steps: Set a first threshold parameter for the rate of change in human traffic density, a second threshold parameter for the rate of change in viewing distance, and a third threshold parameter for the rate of change in ambient light. The rate of change of pedestrian density is compared with the first threshold parameter to obtain the first comparison result; the rate of change of viewing distance is compared with the second threshold parameter to obtain the second comparison result; and the rate of change of ambient light is compared with the third threshold parameter to obtain the third comparison result. Based on the first comparison result, the second comparison result, and the third comparison result, perform a logical OR operation to determine the output of a trigger state flag or a stable state flag. When the trigger status flag is valid, it switches to fast response mode and outputs a high-frequency sampling signal as the trigger control signal. When the stable status flag is valid, it maintains the standard control mode and outputs a regular sampling signal as the trigger control signal.

[0034] Specifically, a first threshold parameter is set for the rate of change in crowd density, with a value of 5 people per second, to define whether the number of people changes drastically within a unit of time. A second threshold parameter is set for the rate of change in viewing distance, typically set to 0.5 meters per second, to assess whether there is a trend of rapid approach or rapid dispersal among the crowd. A third threshold parameter is set for the rate of change in ambient light, with a value of 500 lux per second, to determine whether changes in external light due to weather, time, or geographical conditions cause a sudden change in the display environment. During operation, the above three rate of change parameters are calculated in real time. The current rate of change in crowd density is compared with the first threshold parameter to obtain the first comparison result, i.e., whether it is greater than the preset threshold; if it is, it is recorded as true. The current rate of change in viewing distance is compared with the second threshold parameter to form the second comparison result. The current rate of change in ambient light is compared with the third threshold parameter to obtain the third comparison result. These three comparison results are uniformly logically processed using a logical OR operation rule, i.e., if any one of the three conditions is true, it is considered that the external environment has changed, and a system operation status flag is generated based on this judgment result. When the logical OR operation output is true, the current state is determined to be a triggered state, a trigger state flag is output, and the control mode is switched to fast response mode. This increases the operating frequency of each processing module, synchronously increasing the point cloud data acquisition frequency, the behavior prediction module update frequency, the display parameter adjustment frequency, and the event judgment cycle to form a high-frequency response link. The trigger control signal in this state is set to a high-frequency sampling signal. When all three comparison results are false, it indicates that the crowd size is changing gradually, the viewing distance is stable, and the ambient lighting is not fluctuating drastically. At this point, a stable state is determined, a stable state flag is generated, and the current standard control mode remains unchanged. Data acquisition, trajectory prediction, and display control continue at the normal frequency, while the output trigger control signal is a normal sampling signal to maintain a balance between system resource consumption and response load.

[0035] In one specific embodiment, the process of performing step S4 may specifically include the following steps: Based on the trigger control signal, the display brightness value, display content identifier and refresh rate parameter of the current smart sign are collected and combined to form a display status vector; The displayed state vector will be substituted into the preset state transition equation to predict the state, and a state space model will be established by combining the control input matrix and the output matrix. Based on the state-space model, a weighted quadratic cost function containing visibility error, power consumption cost, and response time cost is constructed. The display brightness range constraint, refresh rate range constraint, and total power consumption constraint are set for the weighted quadratic cost function to obtain the constrained optimization problem. Solve the constrained optimization problem to obtain the explicit control parameters; Adjust the pulse width modulation brightness of the smart signage according to the display control parameters and switch the displayed content.

[0036] Specifically, based on trigger control signals, the system collects the current display brightness value, display content identifier, and refresh rate parameters of the smart signage, and combines them into a multi-dimensional display state vector according to a preset order. The display state vector is then input into a preset state transition equation to predict the state at the next moment. Simultaneously, a state-space model is established by combining the control input matrix and output matrix. This model describes the mapping relationship between state evolution and input / output in the display control system, achieving quantitative control of the system state by the control behavior. While establishing the state-space model, a weighted quadratic cost function for multi-objective optimization is introduced. This cost function consists of three parts: the first part reflects the deviation error between the current display state and the ideal visual output; the second part represents the cost burden caused by changes in each control variable to avoid over-adjustment; and the third part introduces slack variables to absorb deviations when constraints are infeasible and ensure the problem is solvable. By assigning different weight factors to visibility error, power consumption cost, and response time cost, the optimization objectives are guided to make synergistic trade-offs among these three types of indicators. To ensure the physical feasibility of the optimization results and user visual comfort, constraints are set for the cost function. These include limiting the brightness output to a range that is comfortable for the human eye, maintaining the refresh rate within a range that ensures visual continuity, and ensuring that the overall system power consumption does not exceed the hardware's capacity limit. This constitutes a typical constrained multi-objective optimization problem. Based on the built-in optimization solver module, the constrained multi-objective optimization problem is solved using a quadratic programming method. The output is a set of optimal display control parameters, indicating how to adjust the brightness, whether to perform content switching, and whether refresh rate fine-tuning is needed in the current trigger state. This achieves the comprehensive optimal control objectives for display quality, energy consumption control, and response speed. Based on the solution results, control execution commands are issued, converting the brightness adjustment parameters into corresponding pulse width modulation (PWM) signals. These PWM signals are then applied to the display module through the LED driver circuit to adjust the brightness. Simultaneously, if the displayed content changes, a buffer switching mechanism is used to switch the content, and the refresh rate parameters are fine-tuned to ensure display continuity and visual comfort.

[0037] In one specific embodiment, the process of adjusting the pulse width modulation brightness of the smart signboard and switching the displayed content according to the display control parameters may specifically include the following steps: The brightness adjustment amount is extracted from the display control parameters and weighted by the sum of viewer weights, ambient light correction coefficient, and viewer number correction coefficient to obtain the target brightness value. Based on the target brightness value, it is converted into a pulse width modulation duty cycle according to a linear mapping relationship. Based on the pulse width modulation duty cycle, a pulse width modulation waveform signal is generated and input to the LED driver to adjust the current magnitude, thereby obtaining the display output after brightness adjustment. Based on the content switching instructions in the display control parameters, the product of importance, urgency and viewer relevance is calculated for each candidate display content to obtain a priority score. The display content with the highest priority score is selected and the display buffer is switched through the digital signal processor to obtain the updated display output.

[0038] Specifically, the required brightness adjustment for the current operating state is extracted from the display control parameters. This brightness adjustment represents the system's need for fine-tuning the existing display brightness, reflecting the control response generated by the combined effects of behavioral prediction data, environmental conditions, and optimization objectives. Using the brightness adjustment as a baseline value, three key weighting factors are introduced: the total viewer weight, the ambient light correction coefficient, and the viewer number correction coefficient. The total viewer weight is calculated by considering the spatial distance and viewing angle of all individuals within the effective viewing area. Individuals closer to the sign and with more direct angles are assigned greater weight, reflecting the concentration of current viewing behavior. The ambient light correction coefficient is adjusted based on real-time light intensity values ​​collected by the light sensor, enabling brightness output compensation in strong or weak light environments. The viewer number correction coefficient adjusts brightness based on changes in the current crowd size to enhance the collective viewing experience. The total viewer weight, ambient light correction coefficient, and viewer number correction coefficient are weighted and fused with the brightness adjustment to obtain the target brightness value, which is the most suitable display brightness setting for the current environment and crowd conditions. The target brightness value is converted into a corresponding pulse width modulation (PWM) duty cycle through a preset linear mapping relationship. The PWM duty cycle defines the proportion of the high-level duration in the PWM signal within one cycle; a higher value indicates a longer output current duration and higher LED brightness. A corresponding PWM waveform signal is generated based on the duty cycle and input to the LED driver module. The driver controls the current of the LED array to adjust and stabilize the brightness output, ensuring that the actual display effect is highly consistent with the system control target. Simultaneously, the content decision module is activated based on the content switching command in the display control parameters. All current candidate display content undergoes multi-dimensional scoring and evaluation. Each candidate content corresponds to three core evaluation indicators: content importance, content urgency, and relevance to the current viewer. Content importance reflects the fixed weight of the current information in the smart identification system; for example, legal announcements or emergency warnings have a higher weight than advertisements. Content urgency is assessed based on the event's occurrence time or time limit; for example, temporary traffic control or sudden weather warnings have a higher degree of urgency. Viewer relevance is dynamically matched by analyzing the current viewer's behavioral characteristics, interests, or regional attributes to improve content targeting and viewing efficiency. For each candidate item, the product of importance, urgency, and relevance is calculated to obtain a priority score. The item with the highest score is selected as the display object for the current round. The display content is written to the display buffer via a digital signal processor, and the buffer content switching operation is performed to make the content update process have a gradual and smooth characteristic, avoiding visual discomfort or reading interruption caused by content jumps, and completing a comprehensive display output update process including brightness adjustment and content refresh.

[0039] The control method based on environmental perception in the embodiments of the present invention has been described above. The control system based on environmental perception in the embodiments of the present invention will be described below. Please refer to [link / reference]. Figure 2 One embodiment of the environmental perception-based control system of the present invention includes: The data acquisition module is used to collect effective point cloud data of the viewing area in front of the smart signage; The recognition module is used to identify the human body target of the viewer based on effective point cloud data, and at the same time calculate the viewer's behavior prediction data; The judgment module is used to detect the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light based on behavior prediction data, and to perform event trigger condition judgment to obtain trigger control signals; The adjustment module is used to calculate display control parameters based on the trigger control signal, and synchronously adjust the pulse width modulation brightness of the smart sign and switch the display content.

[0040] Through the collaborative efforts of the aforementioned components, LiDAR is used for 3D point cloud data acquisition. Compared to traditional 2D visual perception methods, this approach can acquire precise depth information and 3D spatial coordinates, eliminating the impact of lighting conditions on detection accuracy and achieving all-weather, high-precision environmental perception capabilities. The combined application of the DBSCAN density clustering algorithm and support vector machine classifier accurately distinguishes human targets from other objects, avoiding false detections and missed detections in traditional methods. Furthermore, the introduction of an extended Kalman filter algorithm enables stable tracking of multiple targets and can predict future behavior based on the viewer's historical movement trajectory, effectively shortening the response time of smart signage and improving the matching degree between displayed content and viewer needs. A multi-parameter fusion dynamic event triggering system was established. By monitoring the combined changes in pedestrian density, viewing distance, and ambient light, adaptive switching of control strategies was achieved, avoiding the limitations of fixed threshold triggering methods and improving adaptability to complex dynamic environments. A multi-objective optimization framework incorporating visibility, power consumption, and response time was established, achieving an optimal balance between display effect and energy consumption while meeting system constraints.

[0041] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0042] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0043] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A control method based on environmental perception, characterized in that, include: Collect effective point cloud data of the viewing area in front of the smart signage; The viewer's human body targets are identified based on the effective point cloud data, and the viewer's behavior prediction data is calculated simultaneously. Based on the behavior prediction data, the rate of change in pedestrian density, the rate of change in viewing distance, and the rate of change in ambient light are detected, and event triggering conditions are judged to obtain triggering control signals; Based on the trigger control signal, the display control parameters are calculated, and the pulse width modulation brightness of the smart sign is adjusted synchronously and the display content is switched.

2. The control method based on environmental perception according to claim 1, characterized in that, The effective point cloud data of the viewing area in front of the smart signage includes: Laser pulses are emitted and distances are measured in the viewing area in front of the smart sign, reflected laser signals are received, and raw point cloud data is calculated based on the reflected laser signals. For each coordinate point in the original point cloud data, a search range for neighboring points is established, and the average distance from each coordinate point to its neighboring points is calculated based on the search range. Based on the average distance, noise points are removed from the original point cloud data to obtain denoised point cloud data; The denoised point cloud data is input into a three-dimensional pass-through filter for spatial range limitation processing to obtain spatially clipped point cloud data. The spatially clipped point cloud data is then compressed to obtain effective point cloud data.

3. The control method based on environmental perception according to claim 1, characterized in that, The step of identifying the viewer's human body target based on the effective point cloud data, and simultaneously calculating the viewer's behavior prediction data, includes: Cluster analysis is performed on the effective point cloud data to obtain point cloud cluster groups, and the height span, width span, depth span, and point cloud density value of each cluster in the point cloud cluster group are calculated. The height span value, the width span value, the depth span value, and the point cloud density value are input into a support vector machine classifier to perform human target determination and identify the human target of the viewer; Assign an identification code to each human target of the viewer, and extract the three-dimensional coordinate position and movement speed components of each human target to form an initial state vector; The initial state vector is substituted into the extended Kalman filter algorithm for state prediction and observation update to obtain motion state data for each viewer. A predicted trajectory is generated based on the motion state data, and the viewer's behavior prediction data is calculated based on the predicted trajectory.

4. The control method based on environmental perception according to claim 3, characterized in that, The step of inputting the height span value, the width span value, the depth span value, and the point cloud density value into a support vector machine classifier to perform human target determination and identify the viewer's human target includes: The height span, width span, depth span, and point cloud density values ​​of each cluster are arranged in order to form a standardized feature vector; The standardized feature vector is input into the radial basis function kernel of the support vector machine classifier for high-dimensional space mapping to obtain a high-dimensional feature space vector; The classification decision function value is calculated based on the distance between the high-dimensional feature space vector and the pre-stored support vector, and a binary classification decision is performed on the classification decision function value through sign judgment to obtain the classification result of each cluster. Based on the classification results, clusters of positive categories are selected as candidate targets, and the geometric dimensions of the candidate targets are verified to conform to the range of human body characteristics, thus obtaining the human body target of the viewer.

5. The control method based on environmental perception according to claim 4, characterized in that, The step of generating a predicted trajectory based on the motion state data and calculating the viewer's behavior prediction data based on the predicted trajectory includes: The three-dimensional coordinate position and movement speed components of each viewer in the motion state data are used to form a feature input sequence; Motion state prediction is performed based on the aforementioned feature input sequence to obtain a prediction vector; The three-dimensional position coordinates of the viewer are parsed based on the predicted vector, and a predicted trajectory is formed based on the three-dimensional position coordinates. The velocity component of the viewer's movement on the horizontal plane is calculated based on the predicted trajectory, and the velocity vector synthesis operation is used to obtain the predicted velocity parameters of the viewer's movement toward the smart signboard. The speed prediction parameters are substituted into the exponential decay model to calculate the dwell time, and combined with the arrival time of the viewer to the optimal viewing position in the predicted trajectory for comprehensive analysis to obtain behavioral prediction data.

6. The control method based on environmental perception according to claim 1, characterized in that, The process of detecting the rate of change in pedestrian density, the rate of change in viewing distance, and the rate of change in ambient light based on the behavior prediction data, and performing event trigger condition judgment to obtain a trigger control signal, includes: Based on the behavioral prediction data, the total number of viewers in the viewing area at the current moment is counted and compared with the number of viewers at the previous moment to calculate the rate of change in crowd density; The straight-line distance from each viewer to the smart signboard is calculated based on the three-dimensional position coordinates of each viewer in the behavior prediction data, and the viewing distance change rate is calculated based on the straight-line distance. The current ambient light intensity data is collected by a light sensor, and the rate of change of ambient light intensity is calculated by comparing it with historical light intensity. Based on the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light, the event triggering condition is determined to obtain the triggering control signal.

7. The control method based on environmental perception according to claim 6, characterized in that, The event triggering condition judgment is performed based on the change rate of pedestrian density, the change rate of viewing distance, and the change rate of ambient light, respectively, to obtain the triggering control signal, including: A first threshold parameter is set for the rate of change of pedestrian density, a second threshold parameter is set for the rate of change of viewing distance, and a third threshold parameter is set for the rate of change of ambient light. The rate of change of pedestrian density is compared with the first threshold parameter to obtain a first comparison result; the rate of change of viewing distance is compared with the second threshold parameter to obtain a second comparison result; and the rate of change of ambient light is compared with the third threshold parameter to obtain a third comparison result. Based on the first comparison result, the second comparison result and the third comparison result, perform a logical OR operation to determine the output of a trigger state flag or a stable state flag. When the trigger status flag is valid, it switches to fast response mode and outputs a high-frequency sampling signal as the trigger control signal. When the stable status flag is valid, it maintains the standard control mode and outputs a regular sampling signal as the trigger control signal.

8. The control method based on environmental perception according to claim 1, characterized in that, The process of calculating display control parameters based on the trigger control signal, synchronously adjusting the pulse width modulation brightness of the smart signboard, and switching the display content includes: Based on the trigger control signal, the display brightness value, display content identifier and refresh rate parameter of the current smart sign are collected and combined to form a display status vector; Substitute the displayed state vector into the preset state transition equation to predict the state, and combine the control input matrix and output matrix to establish a state space model. Based on the state-space model, a weighted quadratic cost function containing visibility error, power consumption cost, and response time cost is constructed. The display brightness range constraint, refresh rate range constraint, and total power consumption constraint of the weighted quadratic cost function are set to obtain the constrained optimization problem. The constrained optimization problem is solved to obtain the explicit control parameters; Adjust the pulse width modulation brightness of the smart signboard and switch the displayed content according to the display control parameters.

9. The control method based on environmental perception according to claim 8, characterized in that, The step of adjusting the pulse width modulation brightness of the smart signboard and switching the displayed content according to the display control parameters includes: The brightness adjustment amount is extracted from the display control parameters and weighted by the sum of viewer weights, ambient light correction coefficient, and viewer number correction coefficient to obtain the target brightness value. Based on the target brightness value, it is converted into a pulse width modulation duty cycle according to a linear mapping relationship, and a pulse width modulation waveform signal is generated based on the pulse width modulation duty cycle and input to the LED driver to adjust the current magnitude, thereby obtaining the display output after brightness adjustment; Based on the content switching instructions in the display control parameters, the product of importance, urgency and viewer relevance is calculated for each candidate display content to obtain a priority score. The display content with the highest priority score is selected and the display buffer is switched through the digital signal processor to obtain the updated display output.

10. A control system based on environmental perception, characterized in that, For performing the environment-aware control method as described in any one of claims 1-9, comprising: The data acquisition module is used to collect effective point cloud data of the viewing area in front of the smart signage; The recognition module is used to identify the human body target of the viewer based on the effective point cloud data, and at the same time calculate the viewer's behavior prediction data; The judgment module is used to detect the rate of change of pedestrian density, the rate of change of viewing distance, and the rate of change of ambient light based on the behavior prediction data, and to perform event trigger condition judgment to obtain a trigger control signal; The adjustment module is used to calculate display control parameters based on the trigger control signal, and synchronously adjust the pulse width modulation brightness of the smart sign and switch the display content.

Images