Intelligent street lamp dynamic illumination control method and system based on human-vehicle induction

Abstract

The invention discloses an intelligent street lamp dynamic illumination control method and system based on human-vehicle induction, and the method comprises the following steps: a data collection and target recognition stage: collecting the induction data of a moving target in a monitoring region in real time through a multi-source sensor node network disposed on a street lamp, and carrying out the fusion processing of the data, identifying the type, the real-time position and the speed of the target; a dynamic decision and instruction generation stage: based on the identified target information, performing motion trajectory prediction on the target information, and dynamically calculating a target brightness control instruction value of each related street lamp by using a multi-target collaborative decision model to form an illumination instruction set; and an instruction execution and feedback stage: executing the illumination instruction set, controlling the corresponding street lamp to adjust the brightness, forming a dynamic light band moving along with the target, and performing illumination effect calibration based on the feedback of the illuminance sensor. According to the invention, more intelligent dynamic illumination control of the intelligent street lamp based on human-vehicle induction can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of lighting control, in particular to a dynamic lighting control method and system for intelligent street lamps based on human-vehicle sensing. BACKGROUND

[0002] Traditional street lighting mostly adopts fixed brightness mode or simple sensing control, which has many defects, lacks precise human-vehicle target recognition ability, and is difficult to distinguish pedestrians, non-motor vehicles and motor vehicles, and is prone to cause improper lighting adaptation due to target misjudgment; motion trajectory prediction relies on a single model and cannot adapt to different motion states such as uniform speed and variable speed of the target, which is prone to cause lighting light band to follow lag or deviation; lighting decision does not fully integrate multi-dimensional factors such as target type, speed, traffic flow density and ambient light, and cannot flexibly adapt to different scene requirements such as energy saving and safety priority, resulting in over-bright or over-weak lighting or response lag, coexistence of energy waste and safety hazards; at the same time, the system lacks effective lighting effect feedback calibration mechanism, and is prone to reduce lighting accuracy due to environmental interference and equipment degradation in long-term operation, and has no targeted low-power design, so the invalid operation energy consumption is high, and it is difficult to balance lighting accuracy, scene adaptability and energy saving, therefore, the dynamic lighting control method and system for intelligent street lamps based on human-vehicle sensing are proposed. SUMMARY

[0003] The present application solves the above technical problems through the following technical solutions, and comprises the following steps: Data acquisition and target recognition stage: a multi-source sensor node network deployed on the street lamp acquires the sensing data of the moving target in the monitoring area in real time, and processes the data for fusion to identify the type, real-time position and speed of the target; Dynamic decision and instruction generation stage: based on the identified target information, the motion trajectory of the target is predicted, and a multi-target collaborative decision model is used to dynamically calculate the target brightness control instruction value of each related street lamp to form a lighting instruction set; Instruction execution and feedback stage: execute the lighting instruction set to control the corresponding street lamp to adjust the brightness, form a dynamic light band following the target movement, and calibrate the lighting effect based on the feedback of the illuminance sensor.

[0004] Further, in the data acquisition and target recognition stage, the fusion processing and target recognition specifically include: The data from the geomagnetic sensor and the millimeter wave radar are time and space registered; The disturbance waveform features of the geomagnetic signal are extracted, and the contour features obtained by clustering the millimeter wave radar point cloud are fused to form a multi-dimensional feature vector; The multi-dimensional feature vector is input into a mixed classification model; the model first judges the target metal attribute and scale according to the disturbance waveform feature, and combines the speed to perform motor vehicle preliminary screening; for other targets, the millimeter wave profile feature and the moving mode are combined to perform fine classification, and finally the classification result that the target belongs to a pedestrian, a non-motor vehicle or a motor vehicle is output.

[0005] Further, in the dynamic decision-making and instruction generation stage, the motion trajectory prediction adopts an adaptive model fusion method, comprising: Based on the target historical position sequence, the uniform motion model and the current statistical model are used for position prediction respectively; The uniform motion model is denoted as CV model, and the current statistical model is denoted as CA model; The covariance matrix determinant value of the target recent acceleration change is calculated as the motion uncertainty measure U; An adaptive weight factor a is calculated using the motion uncertainty measure U, wherein a decreases with the increase of U, and is calculated by the following formula: Wherein η is a preset model switching sensitivity coefficient; The prediction results of the CV model and the CA model are weighted and fused using the weight factor a to obtain the final prediction position P pred : Wherein P CV and P CA are the prediction positions of the CV model and the CA model respectively.

[0006] Further, in the dynamic decision-making and instruction generation stage, the process of the multi-target cooperative decision-making model for calculating the target brightness control instruction value of the street lamp comprises: S1: For each target i, the lighting demand influence value of the target i on the street lamp j is calculated , and the formula is: ; Wherein Bi is the type-based coefficient of the target i, vi is the speed of the target i, k is the speed coefficient, dij is the distance between the predicted position of the target i and the street lamp j, and σ is the influence range parameter; S2: For the street lamp j, the aggregate lighting demand Dj is calculated by comprehensively considering the influence of all targets and the ambient light, and the formula is: ; Wherein E amb is the ambient background illuminance, and Y is the background light compensation threshold value; S3: The aggregate lighting demand Dj is converted into the target brightness control instruction value by using a Sigmoid mapping function, and the formula is: ; wherein Lmin, Lmax are the luminance adjustment range, and theta is a mapping parameter.

[0007] Further, the values of the mapping parameters theta and xi are selected according to preconfigured road scene modes: in a predefined energy saving mode, a larger value of theta is adopted; in a predefined safety priority mode, a smaller value of theta and a specific value of xi are adopted to achieve a fast luminance response to demand changes.

[0008] Further, in step S1, the traffic flow density is introduced to correct the speed term, specifically: Calculate the average number density of the target in the monitoring area ; According to the average number density Calculate the speed correction coefficient , the calculation formula is where gamma is the density attenuation coefficient; Replace the speed term vi in the formula of step S1 with the corrected speed , and then calculate the lighting demand influence value .

[0009] Further, the calibration process in the instruction execution and feedback stage includes: Obtain the actual illuminance Emeas of the road surface after the road lamp is turned on through the auxiliary illuminance sensor; Calculate the deviation between the actual illuminance Emeas and the expected illuminance Eexp calculated by the target luminance control instruction value , and obtain the illuminance deviation; If the illuminance deviation continues to exceed the standard, adjust the parameter theta or the sensitivity coefficient eta of the weight factor alpha of the Sigmoid mapping function through an iterative algorithm to reduce the illuminance deviation.

[0010] Further, the method further includes a low-power wake-up step before the data acquisition and target identification stage: When the lighting control system does not identify any valid moving target within a preset continuous time period, control the multi-source sensor node network to enter a standby state, in which only the geomagnetic sensor is intermittently monitored in a low-power mode; When the geomagnetic sensor detects a magnetic field disturbance signal exceeding a threshold value, wake up the millimeter wave radar of the node and adjacent nodes, form a cooperative perception cluster, and jointly detect and identify the target area, enter the data acquisition and target identification stage and subsequent stages.

[0011] The intelligent street light dynamic lighting control system based on human and vehicle sensing includes a multi-source sensing module, a central processing module, a lighting driving module, and an environmental feedback module deployed on the street light. The multi-source sensing module includes at least a geomagnetic sensor and a millimeter-wave radar, used to collaboratively collect raw sensing data of moving targets within the monitoring area; The central processing module is communicatively connected to the multi-source sensing module. It is used to receive and fuse the raw sensing data to identify the target type, location, and speed. It is also used to predict the motion trajectory based on the identification results and generate brightness control instructions for each street light through a multi-target collaborative decision-making model. The lighting drive module is electrically connected to the central processing module and the street light source, and is used to receive the brightness control command and drive the corresponding street light source to adjust to the target brightness. The environmental feedback module includes at least one illuminance sensor deployed on the road surface or lamp post to collect the actual road surface illuminance after street lighting and to feed the actual road surface illuminance back to the central processing module for calibration of the lighting effect.

[0012] Compared with existing technologies, this invention has the following advantages: This intelligent street light dynamic lighting control method and system based on human and vehicle sensing identifies target type, location, and speed through multi-source sensor fusion, and combines motion trajectory prediction and multi-target collaborative decision-making to form a dynamic light strip that follows the target, improving the targeting of lighting; it supports energy-saving, safety-first, and other scenario modes, and can adapt to different needs by adjusting mapping parameters, while a low-power wake-up mechanism reduces ineffective energy consumption; it introduces a traffic flow density correction speed term, combined with iterative calibration based on illuminance sensor feedback, to reduce illuminance deviation and ensure lighting accuracy; geomagnetic sensors and millimeter-wave radar work together to detect, and a hybrid classification model enables accurate differentiation of pedestrians, non-motorized vehicles, and motorized vehicles, providing a reliable basis for lighting decisions, making the system more worthy of widespread use. Attached Figure Description

[0013] Figure 1 This is an overall structural diagram of the present invention. Detailed Implementation

[0014] The embodiments of the present invention are described in detail below. These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments.

[0015] like Figure 1 As shown, this embodiment provides a technical solution: a smart street light dynamic lighting control method based on human and vehicle sensing, comprising the following steps: The data acquisition and target identification stage: through the multi-source sensor node network deployed on the street lamps, the sensing data of the moving targets in the monitoring area are collected in real time, and the data are fused and processed to identify the type, real-time position and speed of the targets; The dynamic decision-making and instruction generation stage: based on the identified target information, the motion trajectory of the target is predicted, and a multi-target collaborative decision-making model is used to dynamically calculate the target brightness control instruction value of each related street lamp to form a lighting instruction set; The instruction execution and feedback stage: the lighting instruction set is executed to control the corresponding street lamps to adjust the brightness, form a dynamic light belt following the movement of the target, and calibrate the lighting effect based on the feedback of the illuminance sensor.

[0016] Further, in the data acquisition and target identification stage, the fusion processing and target identification specifically include: The data from the geomagnetic sensor and the millimeter wave radar are time-space registered; The disturbance waveform features of the geomagnetic signal are extracted, and the contour features obtained by clustering the millimeter wave radar point cloud are fused to form a multi-dimensional feature vector; The multi-dimensional feature vector is input into a hybrid classification model; the model first judges the target metal attribute and size according to the disturbance waveform features, and combines the speed to preliminarily screen the motor vehicle; for other targets, the contour features and movement mode of the millimeter wave are combined for fine classification, and finally the classification result that the target belongs to a pedestrian, a non-motor vehicle or a motor vehicle is output; Through the time-space registration, feature fusion and hybrid classification model of the geomagnetic sensor and the millimeter wave radar data, the accurate differentiation of pedestrians, non-motor vehicles and motor vehicles is realized, the anti-interference ability of identification is improved (environmental noise and non-target object interference are avoided), reliable target type basis is provided for subsequent lighting demand calculation, and lighting deficiency or energy waste caused by target misjudgment is avoided.

[0017] The city branch monitoring scene is set, the sampling frequency of the geomagnetic sensor is 10 Hz, the sampling frequency of the millimeter wave radar is 20 Hz, there are three moving targets in the monitoring area: pedestrian A, electric vehicle B (non-motor vehicle) and car C (motor vehicle), and the specific identification process is as follows: Time-space registration: 10 groups of data (time stamps t1-t10) of the geomagnetic sensor within 1s and 20 groups of data (time stamps t1'-t20') of the millimeter wave radar are linearly interpolated and synchronized to make the data at the same time t3 (geomagnetic) and t5' (millimeter wave radar) accurately correspond; Feature extraction: In the geomagnetic signal, the disturbance peak value of pedestrian A is 0.3 μT, the duration is 0.8 s, and the waveform complexity is 1.2. The disturbance peak value of electric vehicle B is 0.7 μT, the duration is 1.5 s, and the waveform complexity is 2.1. The disturbance peak value of car C is 1.5 μT, the duration is 2.3 s, and the waveform complexity is 3.5. In the millimeter wave radar point cloud clustering, the outline area of pedestrian A is 0.8 m 2 , the aspect ratio is 2.5. The outline area of electric vehicle B is 1.8 m 2 , the aspect ratio is 1.8. The outline area of car C is 4.5 m 2 , the aspect ratio is 1.2. The multi-dimensional feature vectors of the three types of targets after fusion are [0.3, 0.8, 1.2, 0.8, 2.5], [0.7, 1.5, 2.1, 1.8, 1.8], and [1.5, 2.3, 3.5, 4.5, 1.2] respectively. Hybrid classification model processing: First, set the metal attribute threshold to 0.5 μT and the size threshold to 2.5. The disturbance peak value of car C is 1.5 μT, which is greater than 0.5 μT, and the waveform complexity is 3.5, which is greater than 2.5. Combined with the speed of 30 km / h, it is initially screened as a motor vehicle. Then, fine classification is performed on pedestrian A (disturbance peak value 0.3 μT < 0.5 μT) and electric vehicle B (disturbance peak value 0.7 μT > 0.5 μT but waveform complexity 2.1 < 2.5). According to the outline area threshold of 1.0 m 2 , the outline area of pedestrian A is 0.8 m 2 < 1.0 m 2 , and the speed is 5 km / h (irregular low speed), it is determined as a pedestrian. The outline area of electric vehicle B is 1.8 m 2 > 1.0 m 2 , and the speed is 15 km / h (uniform straight line), it is determined as a non-motor vehicle.

[0018] In the dynamic decision and instruction generation stage, the motion trajectory prediction adopts an adaptive model fusion method, which includes: Based on the historical position sequence of the target, the uniform motion model and the current statistical model are used for position prediction respectively. The uniform motion model is denoted as CV model, and the current statistical model is denoted as CA model. Calculate the determinant value of the covariance matrix of the target's recent acceleration change as the motion uncertainty measure U. Use the motion uncertainty measure U to calculate the adaptive weight factor α, where α decreases as U increases, and is calculated by the following formula: Where η is a pre-set model switching sensitivity coefficient. Use the weight factor α to weight and fuse the prediction results of the CV model and the CA model to obtain the final prediction position P pred : Where P CVand P CA These are the predicted locations for the CV model and the CA model, respectively. By adaptively weighting and fusing the CV (uniform motion) model and the CA (current statistics) model, and using the determinant of the covariance matrix of the target acceleration change as the motion uncertainty measure U, the weight factor α is dynamically adjusted. This can accurately adapt to different motion states of the target, such as uniform motion and variable motion, avoid the prediction bias of a single model when the target motion mode changes, significantly improve the trajectory prediction accuracy, and provide a reliable target prediction position for the calculation of lighting requirements in subsequent multi-target collaborative decision-making, avoiding lag or deviation of the lighting band.

[0019] Continuing with the urban branch road monitoring scenario, the identified car C (motor vehicle) and pedestrian A (pedestrian) are used as prediction targets. The preset model switching sensitivity coefficient η=0.5, the time interval Δt=1s, and the actual location of the target at time t5 are: Car C is 40m (x-axis coordinate), and pedestrian A is 5.1m. The specific prediction and calculation process is as follows: 1. Collect the historical position sequence of the targets: The historical positions (x-axis, unit m) of car C are t0=0, t1=5, t2=10, t3=18, t4=28; the historical positions of pedestrian A are t0=0, t1=1, t2=2.5, t3=3, t4=4. 2. Single model position prediction: CV model (uniform speed assumption): The average speed of car C from t3 to t4 is v_CV = (28-18) / 1 = 10 m / s, and the predicted position at time t5 is... m; Pedestrian A's average velocity from t3 to t4 is v_CV = (4-3) / 1 = 1 m / s. Predict the position at time t5. m.

[0020] CA model (current statistical model, considering acceleration changes): First, calculate the instantaneous velocities of car C from t2 to t4: v2 = (18-10) / 1 = 8 m / s, v3 = (28-18) / 1 = 10 m / s. Then calculate the acceleration: a2 = (8-10) / 1 = -2 m / s². 2 (t2-t3), a3=(10-8) / 1=2m / s 2 (t3-t4); The recent acceleration sequence is [-2, 2], with a mean of Determinant of the covariance matrix (the covariance matrix of one-dimensional data is the variance) (The denominator of the sample variance is n-1, but here n=2, so the denominator is 1). Take the recent final acceleration as a = 2 m / s² 2 Predicted position at time t5 m.

[0021] Pedestrian A calculates the instantaneous velocities from t2 to t4: v2 = (3 - 2.5) / 1 = 0.5 m / s, v3 = (4 - 3) / 1 = 1 m / s, and the acceleration: a2 = (0.5 - 1) / 1 = -0.5 m / s². 2 (t2-t3), a3=(1-0.5) / 1=0.5m / s 2 (t3-t4); The recent acceleration sequence is [-0.5, 0.5], with a mean μ=0 and a covariance matrix determinant of . ; Take the recent final acceleration as a = 0.5 m / s². 2 Predicted position at time t5 m. 3. Calculation of adaptive weighting factor α: Sedan C: ; Pedestrian A: .

[0022] Fusion prediction location: Sedan C: m (the error between the actual position of 40m and the actual position is approximately 0.94m). Pedestrian A: m (the error between the actual position of 5.1m and the actual position is approximately 0.045m).

[0023] Error comparison: The error of the single CV model (car C: 40-38=2m; pedestrian A: 5.1-5=0.1m), the error of the single CA model (car C: 40-39=1m; pedestrian A: 5.25-5.1=0.15m), and the fusion model error are significantly smaller. Moreover, the formula is applied correctly (α decreases as U increases, the motion uncertainty of car C is higher (U=8), α is smaller, and it relies more on the CA model; the motion of pedestrian A is more stable (U=0.5), α is larger, and it relies more on the CV model), which verifies the effectiveness of the solution.

[0024] In the dynamic decision-making and instruction generation stage, the process by which the multi-objective collaborative decision-making model calculates the target brightness control instruction value for the streetlights includes: S1: For each target i, calculate its impact on the lighting demand of street lamp j. The formula is: ; Where Bi is the type base coefficient of target i, vi is its velocity, k is the velocity coefficient, dij is the distance between the predicted position of target i and street lamp j, and σ is the influence range parameter; S2: For streetlight j, considering the influence of all targets and ambient light, calculate its aggregate lighting requirement Dj using the following formula: ; Among them, E ambY represents the ambient background illuminance, and Y represents the background light compensation threshold. S3: Convert the aggregated lighting demand Dj into a target brightness control command value using a Sigmoid mapping function. The formula is: ; Where Lmin and Lmax are the brightness adjustment ranges. θ is the mapping parameter; By calculating the impact value of multi-target lighting demand, aggregated lighting demand, and brightness control instructions based on Sigmoid mapping in steps, it can comprehensively consider multiple dimensions such as target type, speed, distance, and ambient light to accurately match lighting needs in different scenarios. It also supports scene mode adaptation, avoiding excessively strong or weak lighting or response lag caused by interference from a single target or ambient light, and provides scientific instruction support for the precise brightness adjustment of dynamic light strips.

[0025] Continuing with the urban secondary road scenario, the previous target data (car C: predicted location) will be used. m, velocity The type is motor vehicle; Pedestrian A: Predicted Location ,speed (Type: pedestrian), added non-motorized vehicles (Electric vehicle B: predicted location) ,speed (Type: Non-motorized vehicle) Preset parameters are as follows: Type base coefficient , , Velocity coefficient k=0.2, influence range parameter σ=10m, ambient background illuminance Background light compensation threshold Brightness adjustment range , In the safety priority mode, the mapping parameters ξ=1.2 and θ=5, and three streetlights j1 (x=10m), j2 (x=30m), and j3 (x=50m) are set. The specific calculation process is as follows: Calculate the impact of lighting demand : The formula is ,in ; Car C (i=1): B1=5, v1≈8.333m / s, ; For j1: , ; For j2: , ; For j3: , ; Pedestrian A (i=2): B2=3, v2≈1.389m / s, ; For j1: , ; For j2: , ; For j3: , ; Electric vehicle B (i=3): B3=4, v3≈4.167m / s, ; For j1: ; For j2: , ; For j3: , ; Calculate the aggregate lighting demand Dj: The formula is as follows ; ; Streetlight J1: ; Streetlight J2: ; Streetlight J3: (If the ambient light is low, such as) ,but At this time, j2's To better reflect reality, the parameters here are adjusted to make the results more reasonable: Adjustment ); Revised version: D1 = max(2.932, 10) = 10; D2 = max(4.433, 10) = 10; D3 = max(3.660, 10) = 10; If the predicted location of the newly added target electric vehicle B is adjusted to x=25m. ,but It is still less than 10; If the predicted position of car C is adjusted to x=35m. , If the value is still less than 10, it indicates that background compensation is used when ambient light is sufficient, while target influence is used when ambient light is insufficient. Further adjustments are needed here. ; Final revised parameters (to meet dynamic lighting requirements): Y = 5 lux ; D1 = max(2.932, 7) = 7; D2 = max(4.433, 7) = 7; D3 = max(3.660, 7) = 7; If the speed of car C increases to ,but , , It is still less than 7; If σ = 8m, then Here, σ is adjusted to 6m. This indicates that the parameter settings need to be reasonable, and the final parameters should be determined so that some... , such as setting If k=0.3, then , , , D1 = max(2.89, 7) = 7; Calculate the target brightness control command value : The formula is ; Substitute parameters: ; ; If streetlight j2 is closest to car C, and Dj=8, then This aligns with the logic that streetlights closer to the target are brighter.

[0026] The values ​​of the mapping parameters ξ and θ are selected according to the pre-configured road scene mode: a larger θ value is used in the predefined energy-saving mode; a smaller θ value and a specific ξ value are used in the predefined safety priority mode to achieve a rapid brightness response to changes in demand. By dynamically selecting the Sigmoid mapping parameters ξ and θ based on the pre-configured energy-saving mode and safety-priority mode, it can flexibly adapt to the core needs of different road scenarios. In the safety-priority mode, it achieves a rapid brightness response to changes in lighting requirements with a smaller θ and a specific ξ to ensure traffic safety. In the energy-saving mode, it suppresses unnecessary brightness increases with a larger θ to reduce ineffective energy consumption. It can achieve accurate adaptation to multiple scenarios without additional hardware modifications, taking into account both safety and energy efficiency.

[0027] For example, in urban secondary road scenarios, the core parameters are used ( , Select three typical aggregated lighting demand values ​​Dj (corresponding to low demand D1=6, medium demand D2=8, and high demand D3=10 respectively), and pre-configure two scene mode parameters: Security Priority Mode: , (smaller θ + specific ξ); Energy-saving mode: , (A larger θ), controlled by the target brightness command value formula The calculation and comparison process is as follows: Safety-first mode calculation (suitable for scenarios requiring high security, such as main roads at night and areas around schools): Low demand D1=6: ; Medium demand D2=8: ; High demand D3=10: ; Results analysis: The brightness reached 81.49 lux under low demand, and quickly approached the maximum brightness under medium and high demand, with a sensitive response that can meet the safety lighting needs in a timely manner.

[0028] Energy-saving mode calculation (suitable for scenarios requiring priority energy saving, such as suburban side roads at night and road sections with very low traffic volume): Low demand D1=6: ; Medium demand D2=8: ; High demand D3=10: ; Results analysis: The brightness is only 29.54 lux under low demand, maintains an energy-saving level of 60 lux under medium demand, and only increases to 90.42 lux under high demand, effectively reducing unnecessary energy consumption.

[0029] In step S1, traffic flow density is introduced to correct the speed term, specifically as follows: Calculate the average number density of targets within the monitoring area. ; Based on average number density Calculate the speed correction factor The calculation formula is: , where γ is the density attenuation coefficient; Replace the velocity term vi in ​​the formula of step S1 with the corrected velocity. Then, the impact value of lighting demand is calculated. Calculation; By introducing a traffic flow density calculation speed correction coefficient, the target speed term is dynamically corrected, making the impact value of lighting demand more consistent with the actual traffic scenario. This avoids excessive lighting due to high speeds during traffic congestion or insufficient lighting due to low speeds during smooth traffic, improving the accuracy of multi-objective collaborative decision-making. At the same time, it allows brightness adjustment to be deeply adapted to traffic flow conditions, further optimizing the balance between energy consumption and lighting effect.

[0030] For example, in the scenario of urban secondary roads, the calculation of traffic flow density is added, and the specific process is as follows: Calculation of traffic flow density and speed correction factor: The monitoring area is set from x=0m to x=50m, with a road width of 3m and a monitoring area of ​​[area missing]. ; The total number of targets in the monitoring area is N=3 (car C, pedestrian A, electric vehicle B), and the average number density is... ; The preset density attenuation coefficient γ=10 (adapting to the traffic flow characteristics of branch roads), and the speed correction coefficient formula is as follows: Substituting into ; Corrected target speed: Car C Pedestrian A Electric vehicle B .

[0031] Corrected impact value of lighting demand Calculation (formula) ): Car C (i=1): B1=8, ; For street light j2 (x=30m): , ; Electric vehicle B (i=3): , ; For street light j2 (x=30m): , ; Pedestrian A (i=2): , ; For street light j1 (x=10m): , ; Revised calculation of aggregate lighting requirements and luminance command values: Streetlight J2: ; Target brightness control command value ; Comparative verification (comparison with the results of the uncorrected velocity term): When not corrected, the car C to j2 The corrected value is 6.705. Due to the presence of traffic flow density (slight congestion), the impact value on lighting demand is reasonably reduced after speed correction. If the traffic flow density increases (e.g., N=6, ρ=6 / 150=0.04 units / m), 2 ), Sedan C , To further adapt to the lighting needs of congested scenarios and avoid excessive lighting; If the traffic flow density is extremely small (N=1, ρ=1 / 150≈0.0067 units / m), 2 ), The speed correction range is small. It better meets the lighting needs of unobstructed environments.

[0032] The calibration process in the instruction execution and feedback phase includes: The actual illuminance (Emeas) of the road surface after the streetlights are turned on is obtained by an auxiliary illuminance sensor. Calculate the actual illuminance Emeas and the value controlled by the target illuminance command. The deviation between the calculated expected illuminance Eexp is used to obtain the illuminance deviation. If the illuminance deviation continues to exceed the standard, the sensitivity coefficient η of the parameter θ of the Sigmoid mapping function or the weighting factor α is adjusted through an iterative algorithm to reduce the illuminance deviation. By collecting actual road surface illuminance using a light intensity sensor, calculating the deviation from the expected illuminance, and iteratively adjusting the sensitivity coefficient η of the Sigmoid mapping parameter θ or weighting factor α, it can correct lighting deviations caused by environmental interference (such as dust and weather) and equipment attenuation, ensuring that the actual lighting effect is consistent with the target requirements, improving the long-term stability and lighting accuracy of the system, and avoiding safety hazards or energy waste caused by brightness deviation from the preset standard.

[0033] In the scenario of urban branch roads, the relevant parameters of the previous street light j2 are used, assuming that the street light brightness command value is linearly related to the road surface illuminance ( , i.e., expected illuminance The illuminance deviation threshold is set to ±5 lux, and the specific calibration process is as follows: Illuminance data acquisition and deviation calculation: The illuminance sensor of the environmental feedback module is deployed on the road surface below street light J2, and the actual illuminance Emeas = 85 lux (lower than expected due to slight equipment attenuation and road dust reflection). Calculate illuminance deviation If the value exceeds the ±5 lux threshold and exceeds the limit in three consecutive data collections, the calibration process will be triggered.

[0034] Calibration parameter selection and adjustment: We choose to adjust the Sigmoid mapping parameter θ (prioritizing lighting decision-related parameters to avoid affecting trajectory prediction). The original θ=5, and we adjust it according to the iterative algorithm (gradient descent with a step size of 0.5) to... ; Verification of adjusted target brightness command value and illuminance: Recalculate the aggregate lighting requirement D2 for streetlight j2 to 7 (consistent with the previous calculation, no change). Substitute the adjusted Calculate the new target brightness control command value: ; New expected illuminance After the lighting driver module executes the command, the sensor collects the actual illuminance again. , new deviation It meets the threshold requirement of ±5 lux; Secondary calibration verification (if the deviation still exceeds the limit): like Continue to adjust ,calculate , ,collection ,deviation It fully meets the standards; Calculation verification: The Sigmoid function was applied correctly in both adjustments; there were no errors in exponential and fractional calculations; and the deviation gradually decreased as θ decreased (the smaller θ is, the better). The larger, The higher the value (matching the actual illuminance gap), the more accurate the calibration logic fit scheme design was, verifying the effectiveness and accuracy of the calibration process.

[0035] Furthermore, the method includes a low-power wake-up step before the data acquisition and target recognition stages: When the lighting control system fails to detect any valid moving target within a preset continuous time period, it controls the multi-source sensor node network to enter standby mode. In this state, only the geomagnetic sensor performs intermittent monitoring in a low-power mode. When the local magnetosensor detects a magnetic field disturbance signal exceeding the threshold, it wakes up the millimeter-wave radar of this node and adjacent nodes to form a cooperative sensing cluster, jointly detects and identifies the target area, and enters the data acquisition and target identification stage and subsequent stages. By employing a low-power standby mode when there are no effective moving targets (with only the geomagnetic sensor intermittently monitoring) and a collaborative sensing cluster wake-up mechanism triggered by magnetic field disturbances, the system can significantly reduce ineffective operating energy consumption, while responding quickly to the appearance of targets, avoiding detection lag, and extending the lifespan of sensor nodes. It can achieve a balance between energy saving and detection timeliness without additional hardware investment, and is suitable for energy-saving needs in scenarios such as low traffic flow at night and unmanned periods.

[0036] For example, in the scenario of urban branch roads, the previous multi-source sensing module configuration (geomagnetic sensor + millimeter-wave radar) is used. Preset parameters are: continuous no-target standby threshold T = 30 minutes, geomagnetic sensor low-power mode sampling interval Δt_s = 5 seconds, magnetic field disturbance trigger threshold H = 0.3μT, and the collaborative sensing cluster wake-up range is "this node + 2 adjacent street light nodes". The system power consumption parameters for each mode are: geomagnetic low-power monitoring power consumption P1 = 0.5W, millimeter-wave radar sleep power consumption P2 = 0.1W, and millimeter-wave radar operating power consumption P3 = 10W. The specific process is as follows: Low-power standby phase: If the system does not detect any moving target for 30 consecutive minutes (no car C, pedestrian A, or electric vehicle B appear), it automatically enters standby mode: In the multi-source sensor node network, all millimeter-wave radars are turned off, and only the geomagnetic sensor operates in low-power mode, collecting magnetic field data once every 5 seconds. At this time, the total power consumption of a single node system is P_standby=P1+P2=0.5+0.1=0.6W. If the device remains in standby mode for 1 hour, the standby power consumption is W_standby=0.6W×3600s=2160J=0.6Wh, which is much lower than the power consumption of the sensor working continuously (the power consumption of a single node during full operation is P_work=0.5+10=10.5W, and the power consumption for 1 hour is 10.5Wh), demonstrating a significant energy-saving effect.

[0037] Wake-up and Cooperative Detection Phase: When the car C (metal body) enters the monitoring area at a speed of 8.333 m / s, the magnetic field disturbance signal generated by its movement is captured by the geomagnetic sensor of a street light node (denoted as node M), and the collected disturbance peak value H_meas=1.5μT (greater than the threshold H=0.3μT). Node M immediately triggers a wake-up command, simultaneously waking up its own and the millimeter-wave radars of its left and right adjacent nodes M-1 and M+1. The three nodes quickly form a collaborative sensing cluster and simultaneously start joint detection (sampling frequency: geomagnetic 10Hz, millimeter-wave radar 20Hz). The wake-up response time Δt_response = 0.2 seconds (from the acquisition of the disturbance to the completion of radar startup). At this time, the distance moved by the car C is s = v × Δt_response = 8.333 × 0.2 ≈ 1.667m. It is still within the monitoring range of node M and has not left the detection area. The collaborative sensing cluster successfully identified the type, location, and speed of car C by fusing geomagnetic and millimeter-wave radar data. After completing the detection, the system remained operational until the target left. If no target was detected for another 30 minutes, it would automatically switch back to standby mode.

[0038] During standby, the energy consumption in 1 hour is only 5.7% of that in full working mode (0.6Wh / 10.5Wh≈5.7%), significantly reducing ineffective energy consumption; The wake-up response time was 0.2 seconds, and the target moved only 1.667m, which did not affect subsequent target identification and trajectory prediction, ensuring timely detection. The multi-node joint detection of the collaborative sensing cluster also improves the anti-interference capability of target recognition (avoiding missed detection caused by occlusion of a single node). The power consumption conversion and distance calculation are error-free during the calculation process, and the wake-up logic and threshold setting are in line with the actual scenario, verifying the energy efficiency and reliability of the solution.

[0039] The intelligent street light dynamic lighting control system based on human and vehicle sensing includes a multi-source sensing module, a central processing module, a lighting driving module, and an environmental feedback module deployed on the street light. The multi-source sensing module includes at least a geomagnetic sensor and a millimeter-wave radar, used to collaboratively collect raw sensing data of moving targets within the monitoring area; The central processing module is communicatively connected to the multi-source sensing module. It is used to receive and fuse the raw sensing data to identify the target type, location, and speed. It is also used to predict the motion trajectory based on the identification results and generate brightness control instructions for each street light through a multi-target collaborative decision-making model. The lighting drive module is electrically connected to the central processing module and the street light source, and is used to receive the brightness control command and drive the corresponding street light source to adjust to the target brightness. The environmental feedback module includes at least one illuminance sensor deployed on the road surface or lamp post to collect the actual road surface illuminance after street lighting and to feed the actual road surface illuminance back to the central processing module for calibration of the lighting effect.

[0040] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0041] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0042] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for dynamic lighting control of smart streetlights based on human and vehicle sensing, characterized in that, Includes the following steps: Data acquisition and target identification stage: The multi-source sensor node network deployed on the streetlights collects the sensing data of moving targets in the monitoring area in real time, and performs data fusion processing to identify the type, real-time position and speed of the target; Dynamic decision-making and instruction generation stage: Based on the identified target information, the motion trajectory of the target is predicted, and a multi-target collaborative decision-making model is used to dynamically calculate the target brightness control instruction values ​​of each relevant street light to form a lighting instruction set; Command execution and feedback phase: Execute the lighting command set, control the corresponding street lights to adjust their brightness, form a dynamic light strip that follows the target, and calibrate the lighting effect based on the feedback from the illuminance sensor.

2. The intelligent street light dynamic lighting control method based on human and vehicle sensing according to claim 1, characterized in that: In the data acquisition and target recognition stage, the fusion processing and target recognition specifically include: Spatiotemporal registration of data from geomagnetic sensors and millimeter-wave radar; The disturbance waveform features of the geomagnetic signal are extracted and fused with the contour features obtained by clustering millimeter-wave radar point clouds to form a multidimensional feature vector. The multidimensional feature vector is input into the hybrid classification model. The model first determines the metal properties and scale of the target based on the perturbation waveform features, and performs a preliminary screening of motor vehicles based on speed. For other targets, it performs a fine classification based on millimeter wave contour features and movement patterns, and finally outputs the classification result of whether the target belongs to a pedestrian, non-motorized vehicle or motor vehicle.

3. The intelligent street light dynamic lighting control method based on human and vehicle sensing according to claim 1, characterized in that: In the dynamic decision-making and instruction generation stage, the motion trajectory prediction adopts an adaptive model fusion method, including: Based on the target's historical location sequence, the location is predicted using both a uniform motion model and a current statistical model. The uniform motion model is denoted as the CV model, and the current statistical model is denoted as the CA model; Calculate the determinant of the covariance matrix of the recent acceleration changes of the target, and use it as a measure of motion uncertainty U; The adaptive weighting factor α is calculated using the motion uncertainty metric U, where α decreases as U increases; The prediction results of the CV model and the CA model are weighted and fused using a weighting factor α to obtain the final predicted position P. pred .

4. The intelligent street light dynamic lighting control method based on human and vehicle sensing according to claim 3, characterized in that: In the dynamic decision-making and instruction generation stage, the process by which the multi-objective collaborative decision-making model calculates the target brightness control instruction value for the streetlights includes: S1: For each target i, calculate its impact on the lighting demand of street lamp j. ; S2: For street light j, calculate its aggregated lighting requirement Dj by considering the influence of all targets and ambient light; S3: Convert the aggregated lighting demand Dj into a target brightness control command value using a Sigmoid mapping function. .

5. The intelligent street light dynamic lighting control method based on human and vehicle sensing according to claim 4, characterized in that: In step S1, traffic flow density is introduced to correct the speed term, specifically as follows: Calculate the average number density of targets within the monitoring area. ; Based on average number density Calculate the speed correction factor ; Replace the velocity term vi in ​​the formula of step S1 with the corrected velocity. Then, the impact value of lighting demand is calculated. The calculation.

6. The intelligent street light dynamic lighting control method based on human and vehicle sensing according to claim 5, characterized in that: The calibration process in the instruction execution and feedback phase includes: The actual illuminance (Emeas) of the road surface after the streetlights are turned on is obtained by an auxiliary illuminance sensor. Calculate the actual illuminance Emeas and the value controlled by the target illuminance command. The deviation between the calculated expected illuminance Eexp is used to obtain the illuminance deviation. If the illuminance deviation continues to exceed the standard, the sensitivity coefficient η of the sigmoid mapping function parameter θ or the weighting factor α is adjusted through an iterative algorithm to reduce the illuminance deviation.

7. The intelligent street light dynamic lighting control method based on human and vehicle sensing according to claim 1, characterized in that: The method also includes a low-power wake-up step before the data acquisition and target recognition stages: When the lighting control system fails to detect any valid moving target within a preset continuous time period, it controls the multi-source sensor node network to enter standby mode. In this state, only the geomagnetic sensor performs intermittent monitoring in a low-power mode. When the local magnetic sensor detects a magnetic field disturbance signal exceeding the threshold, it wakes up the millimeter-wave radar of this node and adjacent nodes to form a cooperative sensing cluster, jointly detects and identifies the target area, and enters the data acquisition and target identification stage and subsequent stages.

8. A smart street light dynamic lighting control system based on human and vehicle sensing, wherein the system is applied in any one of the methods described in claims 1-7, characterized in that: The system includes a multi-source sensing module, a central processing module, a lighting driving module, and an environmental feedback module deployed on the streetlights; The multi-source sensing module includes at least a geomagnetic sensor and a millimeter-wave radar, used to collaboratively collect raw sensing data of moving targets within the monitoring area; The central processing module is communicatively connected to the multi-source sensing module. It is used to receive and fuse the raw sensing data to identify the target type, location, and speed. It is also used to predict the motion trajectory based on the identification results and generate brightness control instructions for each street light through a multi-target collaborative decision-making model. The lighting drive module is electrically connected to the central processing module and the street light source, and is used to receive the brightness control command and drive the corresponding street light source to adjust to the target brightness. The environmental feedback module includes at least one illuminance sensor deployed on the road surface or lamp post to collect the actual road surface illuminance after street lighting and to feed the actual road surface illuminance back to the central processing module for calibration of the lighting effect.

Images