A Smart Streetlight Dynamic Lighting Control Method Based on Multi-Source Sensor Fusion and Traffic Flow Prediction
The intelligent street light control method, which integrates multi-source sensor fusion and hierarchical traffic flow prediction, solves the problems of rigid control, single sensing, lag, and insufficient coordination in traditional street light systems, and achieves efficient, energy-saving, and safe dynamic lighting control.
Patent Information
- Application Number
- CN202611161264.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-03
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional street light control systems suffer from rigid control modes, limited sensing methods, outdated control strategies, lack of coordination between street lights, and weak emergency response capabilities, leading to energy waste and safety hazards.
Holographic perception is achieved by fusion of multiple sensors (millimeter-wave radar, visual sensor, illuminance sensor, environmental monitoring sensor and road condition sensor), combined with hierarchical traffic flow prediction model and multi-objective collaborative optimization, and dynamic lighting control is realized. Closed-loop feedback ensures control accuracy and emergency response.
It improves perception accuracy and control foresight, achieves a comprehensive energy saving rate of 20%-40%, provides continuous light band following effect and rapid response to emergencies, and improves road safety.
Smart Images

Figure CN123124171A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart city lighting control technology, specifically to a method for dynamic lighting control of smart streetlights based on multi-source sensor fusion and traffic flow prediction. Background Technology
[0002] With the accelerating pace of urbanization, the energy consumption and intelligent management level of urban road lighting systems are receiving increasing attention. Statistics show that urban public lighting accounts for approximately 15%-20% of total urban energy consumption, and a significant portion of this lighting is ineffective or inefficiently supplied during periods of low traffic volume. How to minimize lighting energy consumption while ensuring road safety has become a crucial issue in the construction of smart cities.
[0003] Currently, traditional street light control systems mainly suffer from the following technical defects:
[0004] First, the control mode is rigid and lacks dynamic response capability. Most existing street light systems rely on preset on / off schedules or simple photosensitive switches for control, and cannot dynamically adjust according to real-time traffic conditions, environmental changes, and emergencies. During low traffic periods at night, street lights still operate at full power, resulting in serious energy waste; while in rainy, foggy weather or in emergency scenarios, fixed illuminance is insufficient to meet safety requirements.
[0005] Second, the perception methods are limited and the ability to understand the scene is insufficient. Some improved solutions attempt to deploy microwave radar or video cameras for vehicle detection, but these typically rely on only a single type of sensor, making it difficult to comprehensively acquire multi-dimensional information about the road environment. Pure vision solutions experience a sharp decline in detection performance at night in low light or in rainy or foggy weather, while pure radar solutions cannot accurately identify target types. A single sensor is insufficient to meet the needs of holographic perception in complex scenarios.
[0006] Third, the control strategies suffer from lag. Most existing systems adjust brightness reactively based on the currently detected target location, resulting in lighting adjustments lagging behind target movement and failing to achieve smooth, following lighting effects. Some systems incorporate traffic flow statistics for coarse brightness adjustments, but these are based solely on current traffic volume or simple historical averages, lacking the ability to accurately predict the spatiotemporal evolution of traffic flow and thus failing to achieve proactive control.
[0007] Fourth, there is a lack of effective coordination among streetlights. Existing systems mostly adopt independent control of individual lights or coarse-grained unified adjustment of areas. There is a lack of information exchange and collaborative decision-making mechanisms between streetlights, making it difficult to form a continuous light strip effect that follows moving targets, and also unable to achieve globally optimized illuminance distribution in multi-target scenarios.
[0008] Fifth, the emergency response capability is weak. For emergencies such as traffic accidents, severe weather, and emergency vehicle passage, the existing system lacks the ability to automatically detect and intelligently respond, usually requiring manual intervention, resulting in slow response speed and significant safety hazards.
[0009] To address the aforementioned issues, this invention proposes a smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction. It achieves holographic perception through deep sensor fusion, forward-looking control through hierarchical traffic flow prediction, global optimization through multi-objective collaborative optimization, control accuracy through closed-loop feedback, and operational safety through emergency response. Summary of the Invention
[0010] (a) Technical problems to be solved
[0011] In view of the above-mentioned defects or improvement needs of existing technologies, this invention provides a smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction, aiming to solve the above-mentioned technical problems of traditional street light control systems.
[0012] (II) Technical Solution
[0013] To achieve the above objectives, the present invention provides the following technical solution: a smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction, comprising the following steps:
[0014] Step S1: Deploy multi-source sensor nodes on each smart street light pole in the target road area. Each sensor synchronously collects sensing data in the monitoring area at a preset sampling frequency. Use a multi-modal data fusion algorithm to perform spatiotemporal synchronization and feature-level fusion of the multi-source sensing data to generate holographic perception information of the road scene.
[0015] Step S2: Based on the holographic perception information and historical traffic flow data obtained in Step S1, construct a hierarchical traffic flow prediction model to predict the traffic flow and composition status of each road segment within the future time window T.
[0016] Step S3: Based on the real-time holographic perception information from Step S1 and the traffic flow prediction results from Step S2, establish a dynamic lighting demand model and calculate the target illuminance demand value of each street light node within the future time window T.
[0017] Step S4: Based on the target illuminance requirement values of each street light node obtained in Step S3, and combined with the real-time position and predicted trajectory of each moving target, a multi-objective collaborative decision-making model is used to generate an optimized control instruction set for each street light.
[0018] Step S5: Execute the lighting instruction set to control the corresponding streetlights to adjust their brightness, and perform closed-loop calibration of the lighting effect based on the real-time feedback from the illuminance sensor;
[0019] Step S6: Monitor road emergencies in real time, and generate emergency lighting strategies when abnormal events are detected.
[0020] As a further preferred embodiment of the present invention, the multi-source sensor nodes in step S1 include at least: a millimeter-wave radar sensor, a visual sensor, an illuminance sensor, an environmental monitoring sensor, and a road surface condition sensor; the holographic perception information includes the type label, real-time position coordinates, instantaneous velocity, direction of motion, and relative distance between targets for each moving target within the monitoring area; the multimodal data fusion algorithm includes: filtering and clustering radar point cloud data to extract target position and velocity vectors, performing target detection and recognition on visual image data to extract target type labels, associating and matching radar targets and visual targets based on timestamps and spatial coordinates, and using an extended Kalman filter algorithm to fuse and estimate the state of the associated multi-source targets.
[0021] As a further preferred embodiment of the present invention, the hierarchical traffic flow prediction model in step S2 includes: a short-term traffic flow prediction sub-model based on graph attention network, used to predict the traffic flow change trend in the next 5-15 minutes; and a medium-term traffic flow prediction sub-model based on long short-term memory network, used to predict the traffic flow evolution trend in the next 30-60 minutes; the prediction results of the two sub-models are dynamically fused through an adaptive weighted fusion mechanism, and the weight coefficients are adjusted in real time according to the prediction error.
[0022] As a further preferred embodiment of the present invention, the dynamic lighting demand model in step S3 comprehensively considers three types of factors—basic environmental lighting demand, traffic flow lighting demand, and real-time target lighting demand—to generate a comprehensive target illuminance; the basic lighting demand index... The calculation formula is as follows.
[0023] Where L amb For ambient light intensity, L vehicle For vehicle headlight interference intensity, L ref For reference illuminance, C PM C represents the concentration of airborne particulate matter. ref For reference concentration, R road R is the pavement condition coefficient. ref For reference road surface condition coefficient, The weighting coefficients and .
[0024] As a further preferred embodiment of the present invention, in step S4, the multi-objective collaborative decision-making model includes: constructing a collaborative lighting topology network with streetlights as nodes and defining the lighting influence matrix between adjacent streetlights; predicting the motion trajectory based on the real-time position and velocity of the target using Kalman filtering; calculating the lighting demand weight of each target within the coverage area of each streetlight using the target trajectory sequence as a guide; constructing a multi-objective collaborative optimization objective function, wherein the objective function includes a deviation term between actual illuminance and target illuminance, a total system power consumption term, and an illuminance smoothing term between adjacent streetlights; and using a distributed optimization algorithm to solve for the optimal brightness control value of each streetlight, forming a lighting instruction set.
[0025] As a further preferred embodiment of the present invention, the multi-objective collaborative optimization objective function is:
[0026]
[0027] Where N is the total number of streetlights. Let i be the actual output illuminance of the i-th street light. Let be the power consumption of the i-th street light. Let be the set of adjacent streetlights of the i-th streetlight. These are the weighting coefficients.
[0028] As a further preferred embodiment of the present invention, the lighting instruction set in step S4 includes brightness adjustment instructions for each street lamp in a continuous time sequence, used to form a dynamic light strip that follows the movement of the moving target; the formation rule of the dynamic light strip is as follows: street lamps within a preset distance in front of the target are lit in advance with predicted brightness, street lamps at the target location are lit with real-time calculated brightness, and street lamps within a preset distance behind the target gradually reduce brightness; the pre-lighting distance of the street lamps in front of the target is positively correlated with the target speed, and the fading distance of the street lamps behind the target is negatively correlated with the target speed.
[0029] As a further preferred embodiment of the present invention, the closed-loop feedback calibration in step S5 includes: each street light node collects the illuminance value of the actual lighting area in real time, compares it with the target illuminance value to calculate the illuminance deviation, inputs the deviation into the PID controller to generate a brightness compensation correction amount, and adjusts the street light drive signal according to the correction amount to achieve real-time calibration of the actual output illuminance; the calibrated data is fed back to the dynamic lighting demand model in step S3 for online parameter updates.
[0030] As a further preferred embodiment of the present invention, the emergency lighting strategy in step S6 includes: when a traffic accident or road congestion is detected, increasing the brightness of streetlights in the accident area and upstream road sections to a preset emergency illuminance value and providing warning lighting; when an emergency vehicle is detected passing through, pre-brightening the corresponding road section streetlights along the emergency vehicle's route to form a guide light strip; and when environmental monitoring data exceeds the severe weather threshold, automatically adjusting the lighting brightness and color temperature according to visibility.
[0031] (III) Beneficial Effects
[0032] This invention provides a smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction. It has the following beneficial effects:
[0033] Comprehensive and precise perception capabilities. By integrating deep feature-level fusion of millimeter-wave radar and visual sensors, combined with environmental monitoring and road condition sensors, a multi-dimensional holographic perception system is constructed. This overcomes the limitations of single sensors in low-light conditions or adverse weather, significantly improving target recognition accuracy and tracking stability.
[0034] The control strategy is forward-looking. A hierarchical traffic flow prediction model based on graph attention network and long short-term memory network is introduced to incorporate future traffic conditions into lighting decisions, enabling streetlights to anticipate lighting demand and adjust in advance, overcoming the lag defect of traditional reactive control.
[0035] Multi-objective collaborative optimization, balancing safety and energy saving. A collaborative street lighting topology network is constructed, and the multi-objective illuminance allocation problem is solved through distributed optimization. While meeting the lighting needs of each objective, the total system power consumption is minimized, and a smooth transition in illuminance space is ensured, achieving a comprehensive energy saving rate of 20%-40%.
[0036] Dynamic light strip following provides an excellent user experience. Based on target trajectory prediction, a dynamic light strip lighting strategy is generated, with the area in front of the target pre-lit and the area behind gradually dimming, forming a continuous and smooth light strip following effect. This ensures lighting continuity while avoiding energy waste in ineffective lighting areas.
[0037] Closed-loop adaptive calibration ensures strong robustness. Real-time feedback from the illuminance sensor and PID closed-loop control effectively suppress environmental interference and system drift, ensuring accurate tracking of the target value for the actual lighting effect and excellent long-term system stability.
[0038] Intelligent response to abnormal events ensures outstanding safety. It integrates detection and emergency lighting strategies for various emergencies such as traffic accidents, congestion, severe weather, and emergency vehicles, automatically switching emergency modes for rapid response and effectively improving road safety. Attached Figure Description
[0039] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation
[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0041] Please see Figure 1 This invention provides a technical solution: a smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction, specifically including the following steps: Step S1: Multi-source sensor data acquisition and fusion perception, deploying multi-source sensor nodes on each smart street light pole in the target road area. The multi-source sensor nodes include at least: millimeter-wave radar sensors, vision sensors, illuminance sensors, environmental monitoring sensors, and road surface condition sensors. The millimeter-wave radar is used to detect the distance, speed, and azimuth of the target; the vision sensor is used to identify the target type and acquire texture features; the illuminance sensor is used to measure the ambient natural light intensity and the actual illuminance of the road surface; the environmental monitoring sensor is used to collect air particulate matter concentration and temperature and humidity data; and the road surface condition sensor is used to detect the dryness and wetness of the road surface and water accumulation. Each sensor synchronously collects sensor data within the monitoring area at a preset sampling frequency, specifically including: radar point cloud data, video image data, ambient light intensity data, air particulate matter concentration data, and road surface condition data. The sensors achieve time synchronization through a precise clock synchronization protocol and coordinate system registration through a pre-calibrated spatial transformation matrix.
[0042] A multimodal data fusion algorithm is employed to perform spatiotemporal synchronization and feature-level fusion of the multi-source sensor data, generating holographic perception information of the road scene. The specific fusion process is as follows:
[0043] S101: Radar Target Extraction. The point cloud data output from the millimeter-wave radar is filtered to remove ground clutter and static background noise. A density-based spatial clustering algorithm is used to cluster and segment the filtered point cloud, aggregating point clouds belonging to the same target into a target cluster. For each target cluster, its centroid coordinates are calculated as the target's spatial position, and its mean Doppler velocity is calculated as the target's radial velocity, thereby extracting the three-dimensional spatial position and velocity vector of each moving target.
[0044] S102: Visual Target Detection. Target detection and recognition are performed on image data acquired by a visual sensor. An improved deep learning target detection network is used to extract bounding boxes, category labels, and confidence scores of targets in the image. This improved network, based on a standard detection network, embeds an attention mechanism module in the feature extraction layer to enhance its feature response capability for small targets under low-light conditions at night. The network outputs the type label (pedestrian, non-motorized vehicle, small vehicle, medium vehicle, large vehicle) and its position information in the image coordinate system for each target.
[0045] S103: Target Association Matching. Radar targets and visual targets are associated and matched based on timestamps and spatial coordinates. First, the visual targets are transformed from the image coordinate system to the radar coordinate system using a projection matrix. Then, the comprehensive association cost between the radar and visual targets is calculated. This comprehensive association cost considers Euclidean distance, velocity consistency, and target size matching degree. The Hungarian algorithm is used to solve for the globally optimal association matching scheme, constructing a target association matrix to determine the correspondence between radar and visual targets.
[0046] S104: Multi-source state fusion estimation. For successfully associated targets, the extended Kalman filter algorithm is used to fuse and estimate the target state from radar and vision. The state transition model of the extended Kalman filter adopts a uniformly accelerated motion model to adapt to the actual situation of target variable-speed motion in urban roads. The observation models correspond to radar ranging and velocity measurements and visual position observations, respectively. The filter output is the optimized state vector of each target, including the target's three-dimensional position coordinates, three-dimensional velocity, acceleration, and type confidence distribution. For targets that fail to be successfully associated, the independent detection results from radar or vision are output as supplementary information, including the basic lighting demand index. The calculation formula is as follows.
[0047] Where L amb For ambient light intensity, L vehicle For vehicle headlight interference intensity, L ref For reference illuminance, C PM C represents the concentration of airborne particulate matter. ref For reference concentration, R road R is the pavement condition coefficient. ref For reference road surface condition coefficient, The weighting coefficients and This index is used to quantify the impact of environmental factors on basic lighting requirements, providing input parameters for subsequent steps. Through the above fusion processing, the system obtains complete holographic perception information within the monitoring area, including: type labels, real-time position coordinates, instantaneous velocity, motion direction vectors, and relative distances between targets for each moving target.
[0048] Step S2: Spatiotemporal prediction of traffic flow
[0049] Based on the holographic perception information obtained in step S1 and the stored historical traffic flow data, a hierarchical traffic flow prediction model is constructed to predict the traffic flow and composition of each road segment within a future time window T. In this embodiment, the prediction time window T is set to 15 minutes, which can be dynamically adjusted according to the actual road type and management needs.
[0050] The hierarchical traffic flow prediction model includes the following sub-modules:
[0051] S201: Short-Term Traffic Flow Prediction. A short-term traffic flow prediction sub-model based on a graph attention network is constructed. Using the road topology as the graph network foundation, street light monitoring points are treated as graph nodes, road connectivity as edges, and real-time traffic flow data (volume, average speed, vehicle type distribution) as node features. The graph attention layer adaptively learns the attention weights between adjacent nodes to capture the spatiotemporal propagation characteristics of traffic flow in the road network. This sub-model uses time-series data from the 15 minutes preceding the current moment as input to predict traffic flow trends over the next 5-15 minutes.
[0052] S202: Medium-term traffic flow prediction. A medium-term traffic flow prediction sub-model based on a Long Short-Term Memory (LSTM) network is constructed. Using historical traffic flow data from the same period (e.g., the same time in the previous 7 days, the same time in the same week of the previous 4 weeks) and time pattern features such as daily and weekly cycles as input, an LSTM network is used to model the long-range dependencies of the time series, predicting the traffic flow evolution trend within the next 30-60 minutes.
[0053] S203: Prediction Result Fusion. An adaptive weighted fusion mechanism is employed to dynamically weight and fuse the prediction results from S201 and S202 to obtain the final traffic flow prediction result. The fusion weights of each sub-model are adjusted in real time based on its recent prediction errors, using a gradient descent approach to gradually reduce the prediction error after fusion. This mechanism ensures that the medium-term prediction weights are relatively high when traffic flow is stable, and the short-term prediction weights increase rapidly when traffic flow changes abruptly.
[0054] S204: Predicted state output. Based on the above fusion prediction results, a predicted traffic state vector for each road segment within a future time window is generated. The predicted traffic state vector includes at least: predicted vehicle flow, predicted pedestrian flow, predicted vehicle type distribution, and predicted average speed distribution, providing input for subsequent dynamic lighting demand calculations.
[0055] Step S3: Dynamic Lighting Demand Modeling
[0056] Based on the real-time holographic perception information obtained in step S1 and the traffic flow prediction results obtained in step S2, a dynamic lighting demand model is established to calculate the target illuminance demand value of each street light node at each moment within the future time window T.
[0057] The dynamic lighting demand model comprehensively considers the following three types of demand factors:
[0058] Basic ambient lighting requirements. The basic lighting requirement index calculated in step S1. Ebase Characterization reflects the impact of ambient light intensity, air quality, and road surface conditions on baseline illuminance.
[0059] Second, traffic flow lighting requirements. Based on the predicted traffic flow Q... pred Predicted traffic speed V pred The traffic flow lighting demand factor F is calculated based on vehicle type. traffic :
[0060]
[0061] Among them, D max Let D0 be the maximum target density, k be the sensitivity coefficient, and μ1 and μ2 be the weighting coefficients. The physical meaning of this factor is: the denser the target, the stronger the demand for real-time lighting, and the demand growth tends to saturate after the target density exceeds the threshold.
[0062] Based on the above three types of requirements, the comprehensive target illuminance of the i-th street lamp at time t+Δt is generated as follows:
[0063] .
[0064] Among them, I min and I max E represents the minimum and maximum safe output illuminance of the streetlight, respectively. max The maximum basic lighting demand index is defined by ω1, ω2, and ω3, which are weighting coefficients, and ω1 + ω2 + ω3 = 1. The weighting coefficients for each road segment can be differentiated according to the road grade and regional safety level. For example, the value of ω3 can be appropriately increased for important intersections and densely populated pedestrian areas.
[0065] Step S4: Multi-objective collaborative decision-making and lighting strategy generation
[0066] Based on the target illuminance demand values of each street light node obtained in step S3, and combined with the real-time position and predicted trajectory of each moving target, a multi-objective collaborative decision-making model is used to generate an optimized control instruction set for each street light in a continuous time series.
[0067] S401: Collaborative Topology Construction. Construct a collaborative lighting topology network with streetlights as nodes. The edges between nodes represent the lighting influence relationships between adjacent streetlights. Define a lighting influence matrix between adjacent streetlights to quantify the contribution of changes in the brightness of each streetlight to the illuminance of the area covered by the adjacent streetlights. This matrix is obtained through pre-simulation or on-site calibration.
[0068] S402: Target Trajectory Prediction. Based on the real-time position and velocity of the target in step S1, the trajectory of each moving target is predicted. A Kalman filter is used to recursively estimate the target state, and the predicted trajectory is corrected by combining road geometric constraints to generate a sequence of predicted trajectories for each target within a future time window.
[0069] S403: Lighting Demand Weight Calculation. Guided by the predicted target trajectory, calculate the demand weight of each target on the lighting output of each street light. The closer the target is to the street light and the higher the safety level of the target type, the greater its weight on the lighting demand.
[0070] S404: Construction of Multi-Objective Optimization Function. Constructing a multi-objective collaborative optimization problem, where the objective function contains three terms. ,
[0071] The first term is the illuminance demand deviation penalty, which reflects the weighted sum of squares of the deviations between the actual output illuminance of each street light and its target illuminance, ensuring that lighting requirements are met. The second term is the energy consumption penalty, reflecting the total power consumption of the system, ensuring energy-saving targets are met. The third term is the illuminance smoothing penalty, which reflects the sum of squares of the differences in actual output illuminance between adjacent street lights, ensuring a smooth transition of lighting in space. The constraint is that the output illuminance of each street light is between its minimum and maximum values.
[0072] S405: Distributed Optimization Solution. The distributed alternating direction multiplier method is used to solve the above optimization problem. The global optimization problem is decomposed into subproblems for each street light node. Each node performs local calculations based only on its own information and limited information from its neighbors. By iteratively exchanging dual variables, the global optimum is gradually approximated. This method has good scalability and convergence, and is suitable for real-time control of large-scale street light networks.
[0073] S406: Lighting Command Generation. After solving, the optimal brightness control values for each street light in a continuous time series are obtained, forming a lighting command set. The core control logic of this command set is: D in front of the target front Streetlights within range are pre-lit to predict brightness, creating a pre-illumination effect; streetlights at the target location are illuminated at the target brightness calculated in step S3; streetlights behind the target (D) illuminate the target. rear The streetlights within the distance gradually reduce their brightness to the base level in a linear gradient manner, and the three together form a dynamic light strip that follows the moving target.
[0074] The pre-lighting distance D front With the target velocity v target Positive correlation, fading distance D rear Negatively correlated with target speed:
[0075] Among them, D min and D max These represent the minimum and maximum effective distances, respectively, k. v The value represents the speed coefficient. This design ensures that high-speed targets have a longer forward pre-illuminated field of view, while low-speed targets have a smaller illumination range to save energy.
[0076] Step S5: Closed-loop feedback calibration
[0077] Each street light node performs brightness adjustment according to the lighting instruction set generated in step S4, and performs closed-loop calibration of the lighting effect based on the real-time measurement value of the illuminance sensor.
[0078] The specific process is as follows:
[0079] S501: The illuminance sensor at each street light node collects the actual illuminance value on the road reference surface within its illumination coverage area in real time using a high-frequency sampling rate.
[0080] S502: Compare the actual illuminance value with the target illuminance value given in step S4, and calculate the illuminance deviation.
[0081] S503: The illuminance deviation is input into the PID feedback controller, which outputs a luminance compensation correction. The parameters of the PID controller are pre-tuned according to the response characteristics of the road lighting system and are adaptively adjusted based on the system response during operation.
[0082] S504: Adjust the duty cycle of the PWM signal of the street light LED driver module according to the compensation correction amount to achieve real-time calibration of the actual output illuminance, so that the actual illuminance can quickly and stably approach the target illuminance.
[0083] S505: Feedback the calibrated actual illuminance data, illuminance deviation information and controller status to the dynamic lighting demand model in step S3 and the collaborative decision-making model in step S4 for adaptive online updates of model parameters, forming a complete closed-loop control loop of "perception-decision-execution-feedback".
[0084] Step S6: Abnormal Event Response and Emergency Lighting
[0085] The system monitors road conditions and meteorological data in real time. When an abnormal event is detected, it automatically generates and executes an emergency lighting strategy.
[0086] The sources of abnormal event detection include: detecting traffic accidents, obstacles, or abnormal pedestrian behavior through joint analysis of visual sensor and radar data; detecting sudden congestion through residual analysis of traffic flow prediction and real-time data; detecting severe weather conditions through environmental monitoring sensor data; and receiving emergency vehicle priority passage requests through V2X communication.
[0087] The system implements differentiated emergency lighting strategies for different types of abnormal events:
[0088] S601: Traffic Accident Emergency Response. When a traffic accident is detected, the system automatically increases the brightness of all streetlights in the accident area and within 200 meters upstream to 100% of their rated illuminance, and flashes warning lights at a specific frequency. At the same time, it sends an event notification to the traffic management department through the system platform.
[0089] S602: Road Congestion Emergency. When road congestion is detected, the system appropriately increases the illumination of the congested road section according to the congestion level to improve the driver's visual comfort and safety perception under congested conditions.
[0090] S603: Severe Weather Emergency. When PM2.5 concentration exceeds the threshold or visibility falls below the set value, the system automatically switches to severe weather lighting mode. In this mode, the target illuminance of each road section is significantly increased compared to normal conditions, the color temperature of streetlights is uniformly adjusted to a warm yellow temperature to reduce glare caused by atmospheric scattering, and the illuminance level is dynamically adjusted in real time according to visibility.
[0091] S604: Emergency Vehicle Guidance. When receiving priority passage requests and route information from emergency vehicles (fire trucks, ambulances, police cars), the system switches streetlights to high-brightness guidance mode segment by segment along the vehicle's expected route in advance, forming a dynamic light strip with directional indication function to assist emergency vehicles in passing quickly and safely.
[0092] Example 1: Application on urban main roads
[0093] The system described in this invention is deployed along a six-lane dual carriageway in a city. The road's design speed is 60 km / h, the streetlight spacing is 30 meters, the light pole height is 12 meters, and the single light power is 150W (LED light source). Each streetlight pole integrates a 77GHz millimeter-wave radar, a 4-megapixel starlight-level network camera, an illuminance sensor, a PM2.5 / PM10 air quality sensor, and a road surface temperature and humidity sensor. The edge computing gateway is configured with a GPU acceleration module for running deep learning object detection and fusion algorithms.
[0094] During system operation, radar and visual sensors synchronously acquire data at a frequency of 20Hz. The millimeter-wave radar's detection range covers 200 meters ahead and four lanes laterally. The visual sensors can still effectively detect pedestrians and small targets even at night with an illuminance of 0.1 lux.
[0095] In actual testing during the evening rush hour (6:00 PM - 7:00 PM), the multi-source fusion perception module achieved a vehicle detection accuracy of 98.7% and a pedestrian detection accuracy of 95.2%, representing improvements of 12.3% and 18.7% respectively compared to the single-vision solution. The hierarchical traffic flow prediction model had an average absolute percentage error of 7.8% in predicting traffic flow for the next 15 minutes, which is better than the 12.9% of the single LSTM model. The dynamic lighting control system, while ensuring road lighting standards (average illuminance not less than 20 lux), achieved energy savings of 36.8% compared to traditional timed control solutions and 21.5% compared to dimming solutions based on single-sensor detection.
[0096] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.
[0097] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.
Claims
1. A smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction, characterized in that, Includes the following steps: Step S1: Deploy multi-source sensor nodes on each smart street light pole in the target road area. Each sensor synchronously collects sensing data in the monitoring area at a preset sampling frequency. Use a multi-modal data fusion algorithm to perform spatiotemporal synchronization and feature-level fusion of the multi-source sensing data to generate holographic perception information of the road scene. Step S2: Based on the holographic perception information and historical traffic flow data obtained in Step S1, construct a hierarchical traffic flow prediction model to predict the traffic flow and composition status of each road segment within the future time window T. Step S3: Based on the real-time holographic perception information from Step S1 and the traffic flow prediction results from Step S2, establish a dynamic lighting demand model and calculate the target illuminance demand value of each street light node within the future time window T. Step S4: Based on the target illuminance requirement values of each street light node obtained in Step S3, and combined with the real-time position and predicted trajectory of each moving target, a multi-objective collaborative decision-making model is used to generate an optimized control instruction set for each street light. Step S5: Execute the lighting instruction set to control the corresponding streetlights to adjust their brightness, and perform closed-loop calibration of the lighting effect based on the real-time feedback from the illuminance sensor; Step S6: Monitor road emergencies in real time, and generate emergency lighting strategies when abnormal events are detected.
2. The smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 1, characterized in that: The multi-source sensor nodes in step S1 include at least: millimeter-wave radar sensors, visual sensors, illuminance sensors, environmental monitoring sensors, and road surface condition sensors; the holographic perception information includes the type labels, real-time position coordinates, instantaneous velocity, direction of motion, and relative distance between targets for each moving target within the monitoring area; the multimodal data fusion algorithm includes: filtering and clustering radar point cloud data to extract target position and velocity vectors, performing target detection and recognition on visual image data to extract target type labels, associating and matching radar targets and visual targets based on timestamps and spatial coordinates, and using an extended Kalman filter algorithm to fuse and estimate the state of the associated multi-source targets.
3. The intelligent street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 1, characterized in that: The hierarchical traffic flow prediction model described in step S2 includes: a short-term traffic flow prediction sub-model based on graph attention network, used to predict the traffic flow change trend in the next 5-15 minutes; and a medium-term traffic flow prediction sub-model based on long short-term memory network, used to predict the traffic flow evolution trend in the next 30-60 minutes. The prediction results of the two sub-models are dynamically fused through an adaptive weighted fusion mechanism, and the weight coefficients are adjusted in real time according to the prediction error.
4. The smart street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 1, characterized in that: The dynamic lighting demand model described in step S3 comprehensively considers three factors—basic environmental lighting demand, traffic flow lighting demand, and real-time target lighting demand—to generate a comprehensive target illuminance; the basic lighting demand index The calculation formula is as follows. ; Where L amb For ambient light intensity, L vehicle For vehicle headlight interference intensity, L ref For reference illuminance, C PM C represents the concentration of airborne particulate matter. ref For reference concentration, R road R is the pavement condition coefficient. ref For reference road surface condition coefficient, The weighting coefficients and .
5. The intelligent street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 1, characterized in that: In step S4, the multi-objective collaborative decision-making model includes: constructing a collaborative lighting topology network with streetlights as nodes and defining the lighting influence matrix between adjacent streetlights; predicting the motion trajectory using Kalman filtering based on the real-time position and velocity of the target; calculating the lighting demand weight of each target within the coverage area of each streetlight using the target trajectory sequence as a guide; constructing a multi-objective collaborative optimization objective function, which includes a deviation term between actual illuminance and target illuminance, a total system power consumption term, and an illuminance smoothing term between adjacent streetlights; and using a distributed optimization algorithm to solve for the optimal brightness control value of each streetlight, forming a lighting instruction set.
6. The intelligent street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 5, characterized in that, The objective function for the multi-objective collaborative optimization is: ; Where N is the total number of streetlights. Let i be the actual output illuminance of the i-th street light. Let be the power consumption of the i-th street light. Let be the set of adjacent streetlights of the i-th streetlight. These are the weighting coefficients.
7. The intelligent street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 1, characterized in that, The lighting instruction set mentioned in step S4 includes brightness adjustment instructions for each street light in a continuous time sequence, which are used to form a dynamic light strip that follows the movement of a moving target. The formation rules of the dynamic light strip are as follows: street lights within a preset distance in front of the target are lit in advance with predicted brightness, street lights at the target location are illuminated with real-time calculated brightness, and street lights within a preset distance behind the target gradually reduce their brightness. The pre-lighting distance of the streetlights in front of the target is positively correlated with the target speed, while the fading distance of the streetlights behind the target is negatively correlated with the target speed.
8. The intelligent street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 1, characterized in that, The closed-loop feedback calibration in step S5 includes: each street light node collects the illuminance value of the actual lighting area in real time, compares it with the target illuminance value to calculate the illuminance deviation, inputs the deviation into the PID controller to generate a brightness compensation correction amount, and adjusts the street light drive signal according to the correction amount to achieve real-time calibration of the actual output illuminance; the calibrated data is fed back to the dynamic lighting demand model in step S3 for online parameter updates.
9. The intelligent street light dynamic lighting control method based on multi-source sensor fusion and traffic flow prediction according to claim 1, characterized in that, The emergency lighting strategy described in step S6 includes: when a traffic accident or road congestion is detected, increasing the brightness of streetlights in the accident area and upstream road sections to a preset emergency illuminance value and providing warning lighting; when an emergency vehicle is detected passing through, pre-brightening the corresponding road sections of streetlights along the emergency vehicle's route to form a guide light strip; and when environmental monitoring data exceeds the severe weather threshold, automatically adjusting the lighting brightness and color temperature according to visibility.
