High-efficiency energy-saving intelligent street lamp control system based on road flow AI identification analysis

Abstract

The invention discloses a high-efficiency energy-saving intelligent street lamp control system based on road flow AI identification analysis, and relates to the technical field of intelligent illumination, and the method comprises the steps: a road flow analysis module carries out the prediction through the link ratio and year-on-year flow time sequence information, and generates a target road prediction flow; and the street lamp intelligent dimming module formulates a preset street lamp control strategy according to the predicted flow. The local communication module monitors the road flow in real time, the private network communication module compares the deviation between the real-time flow and the predicted flow, and when the deviation is within a threshold range, the street lamp dimming control module adopts a preset control strategy; and when the deviation exceeds a threshold value, the brightness control optimization module performs optimization adjustment according to the real-time flow, generates a real-time control strategy and executes dimming. The technical problems that an existing street lamp control system depends on real-time control, so that efficiency is low, and dynamic changes of road flow cannot be responded in time are solved, and the technical effect of giving consideration to precise illumination and efficient energy saving is achieved.

Description

Technical Field

[0001] The present application relates to the field of smart lighting technology, and in particular to an efficient and energy-saving smart street lamp control system based on AI recognition and analysis of road traffic. Background Art

[0002] With the advancement of smart city construction, road lighting, as an important part of urban infrastructure, is gradually developing in the direction of intelligence. However, traditional street light control systems usually adopt time-sharing dimming or real-time sensor control, which has the problems of slow response speed and low recognition accuracy. Especially in high lamp pole scenes and complex weather conditions, traditional sensors are difficult to accurately identify traffic and pedestrian flows, resulting in dimming lag or false triggering, and it is difficult to effectively adapt to the needs of dynamic traffic flow changes.

[0003] At present, there are technical problems in relevant technologies, such as street lamp control systems relying on real-time control, resulting in low efficiency and inability to respond to dynamic changes in road traffic in a timely manner. Summary of the invention

[0004] This application provides an efficient and energy-saving smart street light control system based on AI recognition and analysis of road traffic. The road traffic analysis module is used to predict the target road predicted traffic through month-on-month and year-on-year traffic time series information; the street light intelligent dimming module formulates a preset street light control strategy based on the predicted traffic. The local communication module monitors the road traffic in real time, and the private network communication module compares the deviation between the real-time traffic and the predicted traffic. When the deviation is within the threshold range, the street light dimming control module adopts a preset control strategy; when the deviation exceeds the threshold, the brightness control optimization module optimizes and adjusts according to the real-time traffic, generates a real-time control strategy and executes dimming, so as to achieve the technical effect of both precise lighting and efficient energy saving by quickly adjusting the street light brightness according to the dynamic changes of road traffic.

[0005] This application provides an efficient and energy-saving smart street light control system based on road traffic AI recognition and analysis, including:

[0006] A road traffic analysis module, the road traffic analysis module is used to predict traffic based on the target road month-on-month traffic timing information and the target road year-on-year traffic timing information in the target time zone, and obtain the predicted traffic of the target road; a street lamp intelligent dimming module, the street lamp intelligent dimming module is used to optimize the street lamp brightness control according to the predicted traffic of the target road, and obtain the preset road street lamp control strategy; a local communication module, the local communication module is used to monitor the target road traffic information in the target time zone; a private network communication module, the private network communication module is used to determine whether the road traffic deviation between the target road traffic information and the predicted traffic of the target road is less than or equal to the road traffic deviation threshold; a street lamp dimming control module, the street lamp dimming control module is used to perform street lamp dimming control according to the preset road street lamp control strategy when the road traffic deviation is less than or equal to the road traffic deviation threshold; a brightness control optimization module, the brightness control optimization module is used to optimize the street lamp brightness control according to the target road traffic information when the road traffic deviation is greater than the road traffic deviation threshold, and obtain a real-time road street lamp control strategy for street lamp dimming control.

[0007] The present application also provides an electronic device, including:

[0008] A memory is used to store executable instructions; a processor is used to implement an efficient and energy-saving smart street lamp control system based on road traffic AI recognition and analysis when executing the executable instructions stored in the memory.

[0009] The present application also provides a computer-readable storage medium, comprising:

[0010] A computer program is stored thereon, which, when executed by the processor, implements an efficient and energy-saving smart street lamp control system based on AI recognition and analysis of road traffic.

[0011] The proposed high-efficiency and energy-saving intelligent street light control system based on AI recognition and analysis of road traffic proposed in this application, firstly, the road traffic analysis module predicts the target road predicted traffic through the month-on-month and year-on-year traffic time series information; the street light intelligent dimming module formulates a preset street light control strategy based on the predicted traffic. The local communication module monitors the road traffic in real time, and the private network communication module compares the deviation between the real-time traffic and the predicted traffic. When the deviation is within the threshold range, the street light dimming control module adopts the preset control strategy; when the deviation exceeds the threshold, the brightness control optimization module optimizes and adjusts according to the real-time traffic, generates a real-time control strategy and executes dimming, so as to achieve the technical effect of combining precise lighting and high-efficiency energy saving by quickly adjusting the street light brightness according to the dynamic changes of road traffic. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0013] Figure 1 A schematic diagram of the structure of an efficient and energy-saving smart street lamp control system based on road traffic AI recognition and analysis provided in an embodiment of the present application;

[0014] Figure 2 A schematic diagram of the structure of a street lamp intelligent dimming module of an efficient and energy-saving smart street lamp control system based on road traffic AI recognition and analysis provided in an embodiment of the present application;

[0015] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0016] Explanation of the reference numerals: road traffic analysis module 10 , street lamp intelligent dimming module 20 , local communication module 30 , private network communication module 40 , street lamp dimming control module 50 , brightness control optimization module 60 . DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0020] The present application embodiment provides an efficient and energy-saving smart street light control system based on road traffic AI recognition and analysis, such as Figure 1 As shown, the system comprises:

[0021] The road traffic analysis module 10 is used to predict traffic according to the target road month-on-month traffic time series information and the target road year-on-year traffic time series information in the target time zone, and obtain the predicted traffic of the target road. Specifically, the road traffic analysis module 10 first collects the target road month-on-month traffic time series information (including vehicles and people, such as obtaining and sorting traffic data at different time intervals through sensors, databases, and mobile application data) and year-on-year traffic time series information (obtaining the data of the same time zone in the same year corresponding to the past year from historical records) from multiple data sources, and then performs adjacent time series fitting on the vehicle month-on-month traffic time series information (calculating the adjacent time domain traffic deviation, merging and updating if it is less than or equal to the threshold, until the deviation is greater than the threshold to obtain a sequence), obtains the vehicle year-on-year traffic sequence (similar to month-on-month processing), calculates the month-on-month and year-on-year traffic change rate sequence, and inputs these sequences into the road traffic prediction model based on the data set constructed by the collected data (group sampling training data set) to train the long short-term memory neural network (multiple sampling training multiple models after the mean is fully connected) to obtain the vehicle traffic prediction value, and the crowd traffic prediction is the same, and finally the vehicle and crowd traffic prediction values ​​are combined to obtain the target road predicted traffic.

[0022] In a possible implementation, the road traffic analysis module 10 further includes: a month-on-month traffic time series information composition unit, and the month-on-month traffic time series information composition unit is used for the target road month-on-month traffic time series information, including vehicle month-on-month traffic time series information and crowd month-on-month traffic time series information. Specifically, the month-on-month traffic time series information composition unit collects the data source of the vehicle month-on-month traffic time series information, and the data source covers various types of vehicle detection sensors installed on the target road (such as induction coils using electromagnetic induction to record vehicle passing, geomagnetic sensors, radar speed meters, etc.), the vehicle traffic statistics database of the traffic management department, and the video monitoring system based on image processing technology, etc. The sensor detects the situation of vehicles passing through specific sections of the road in real time and records relevant information, and then organizes the vehicle traffic data at set time intervals (such as every hour, every day, every week, etc.), and arranges them in chronological order to form vehicle month-on-month traffic time series information, which is helpful to observe the short-term change trend of vehicle traffic. The collection of crowd flow time series information mainly relies on equipment technology that detects pedestrians, such as pedestrian counters beside roads or at intersections (using infrared sensing, video analysis, etc.), smart surveillance cameras, and Wi-Fi hotspot detection technology in some places. Crowd flow data is sorted and sorted at predetermined time intervals to form crowd flow time series information, which is of great significance for analyzing the patterns of crowd activities. The information can provide a basis for subsequent traffic forecasts and street lamp dimming control, ensuring lighting needs while achieving energy conservation.

[0023] The year-on-year flow time series information composition unit is used for the year-on-year flow time series information of the target road, including the year-on-year flow time series information of vehicles and the year-on-year flow time series information of the crowd. Specifically, the year-on-year flow time series information composition unit extracts the vehicle flow records of the target road corresponding to the same time period in the past years from the historical data repository of the traffic management department. These historical data are rich in accumulation and are cleaned and sorted to ensure accuracy and completeness. Missing values, abnormal values, etc. need to be processed, such as correcting abnormal data caused by sensor failure, etc., to reflect the long-term change trend of vehicle flow. The acquisition method of the crowd year-on-year flow time series information is similar, which comes from historical pedestrian flow monitoring projects, traffic planning research data, etc. After acquisition, it is also cleaned and pre-processed to unify the data format and statistical caliber of different monitoring methods to ensure data quality, so as to accurately reflect the crowd flow in the same period in the past, analyze the long-term change pattern of crowd flow, and provide a comprehensive decision-making basis for intelligent dimming of street lights to adapt to the needs of crowd activities in different periods.

[0024] A traffic prediction unit, the traffic prediction unit is used to perform traffic prediction based on the vehicle month-on-month traffic time series information and the vehicle year-on-year traffic time series information to obtain the vehicle traffic on the target road. Specifically, the traffic prediction unit first performs an adjacent time series fitting operation on the vehicle month-on-month traffic time series information, reads the adjacent time domain vehicle traffic data to calculate the deviation, and compares it with the fitting traffic deviation threshold. If it is less than or equal to the threshold, the time domains are merged to take the average value and update the time series information until the adjacent time domain traffic deviation is greater than the threshold, thereby obtaining a smooth vehicle month-on-month traffic sequence. For the vehicle year-on-year traffic time series information, similar operations are performed in the same time period in different years to obtain the year-on-year traffic sequence, reflecting the long-term change trend. Then, the change rate sequences of the vehicle month-on-month and year-on-year traffic sequences are calculated respectively to provide dynamic change information input for subsequent predictions. After that, historical vehicle flow data is collected to construct a data set and divided into k groups (k≥5). After k times of sampling with replacement, a training data set is obtained. Each training data set is trained with LSTM, and the Mth road flow prediction model is trained according to the Mth training data set (M≥7). Finally, a mean full connection operation is performed on multiple models, and the prediction results of each model are weighted averaged to obtain a road flow prediction model that can more accurately predict the vehicle flow on the target road.

[0025] A crowd-road vehicle flow acquisition unit, the crowd-road vehicle flow acquisition unit is used to perform flow prediction based on the crowd month-on-month flow time series information and the crowd year-on-year flow time series information to obtain crowd-road vehicle flow. Specifically, the crowd-road vehicle flow acquisition unit predicts crowd flow based on the crowd month-on-month flow time series information and the crowd year-on-year flow time series information, and its process is similar to vehicle flow prediction. First, the crowd month-on-month flow time series information is fitted with adjacent time series, the crowd flow deviation in adjacent time domains is calculated and compared with the threshold, and the time domain is merged or retained accordingly to obtain a smooth crowd month-on-month flow sequence to reduce the impact of short-term fluctuations. The year-on-year flow time series information of the crowd is also processed with a similar method to process the year-on-year flow sequence of data in the same time period of different years, and analyze the long-term change trend, such as the change of road crowd flow near tourist attractions affected by multiple factors. Then, the change rate sequence of the crowd month-on-month and year-on-year flow sequences is calculated respectively, and the calculation method is similar to the vehicle flow change rate sequence, which can reflect the increase and decrease speed of crowd flow and provide dynamic information for the prediction model. Then the month-on-month crowd flow change rate sequence, month-on-month crowd flow sequence, year-on-year crowd flow change rate sequence and year-on-year crowd flow sequence are input into the crowd flow prediction model (the principle may be similar to the vehicle flow prediction model), and the crowd road vehicle flow prediction value is obtained after model calculation and processing. This value, combined with the vehicle flow prediction value, can comprehensively evaluate the road traffic flow situation, provide accurate data support for applications such as intelligent dimming of street lights, ensure that the lighting system is reasonably dimmed according to the situation of traffic participants, and improve energy utilization efficiency, traffic safety and pedestrian experience.

[0026] In a possible implementation, the traffic prediction unit further includes: an adjacent time series fitting subunit, which is used to perform adjacent time series fitting on the vehicle ring-by-ring traffic time series information to obtain a vehicle ring-by-ring traffic sequence. Specifically, the vehicle ring-by-ring traffic sequence acquisition process is as follows: read data from the vehicle ring-by-ring traffic time series information that is arranged in chronological order and originates from the road vehicle detection sensor. Before reading, check and preprocess the data to ensure accuracy and completeness, and handle problems such as missing data or abnormal values. Then select the vehicle flow data of adjacent time domains (such as adjacent two hours or two days, etc.) in order, and calculate its deviation. The formula is deviation = (Qi+1-Qi) / Qi. For example, when the flow rates of two time domains are 100 and 120 respectively, the deviation is 20%. Then compare the deviation with the fitting traffic deviation threshold determined by empirical analysis. If the deviation is less than or equal to the threshold, merge the time domains. The merged time domain range covers the original two time domains, and the flow is averaged, such as (Qi+Qi+1) / 2. Then update the vehicle ring-by-ring traffic time series information, and the number of time domains is reduced but the representativeness is enhanced. Repeat this process until all adjacent time domains are analyzed, and finally obtain a vehicle traffic sequence that can more smoothly reflect the short-term trend of vehicle traffic and reduce local fluctuation interference, laying the foundation for grasping the rules for subsequent traffic forecasts.

[0027] The year-on-year traffic sequence acquisition subunit is used to perform adjacent time series fitting on the vehicle year-on-year traffic time series information to obtain the vehicle year-on-year traffic sequence. Specifically, when the year-on-year traffic sequence acquisition subunit processes the vehicle year-on-year traffic time series information, it first extracts the vehicle traffic records of the same time period in the past years from the historical data repository to ensure data reliability. For example, to obtain the year-on-year traffic of a specific period this year, it is necessary to accurately find the data of the same period last year. Then, the adjacent time series fitting method is used to select the adjacent year-on-year time domains (such as the adjacent two days of the same week last year and this year) in chronological order to calculate the deviation of vehicle traffic data. The formula is deviation = (Qi”-Qi') / Qi'. For example, when the vehicle flow on a certain day last year was 80 vehicles and the same day this year was 90 vehicles, the deviation was 12.5%. Then, the deviation is compared with the fitting traffic deviation threshold. If it is less than or equal to the threshold, the year-on-year time domains are merged, and the traffic is averaged (Qi'+Qi”) / 2, and then the year-on-year traffic time series information is updated. This process is continued until all adjacent year-on-year time domains are processed, and finally a vehicle year-on-year traffic sequence that can reflect the long-term (annual) change trend of vehicle traffic is obtained.

[0028] The month-on-month traffic flow rate change sequence acquisition subunit is used to obtain the month-on-month traffic flow rate sequence of the vehicle month-on-month traffic flow sequence. Specifically, after obtaining the vehicle month-on-month traffic flow sequence, the month-on-month traffic flow rate change sequence acquisition subunit proceeds to calculate the month-on-month traffic flow rate sequence. First of all, it is necessary to ensure that the vehicle month-on-month traffic flow sequence data is accurate and arranged in time series. This sequence is obtained by fitting adjacent time series and reflects the change trend of vehicle traffic after time domain merging. Then select adjacent data points (except the first point) from the sequence and calculate according to the formula month-on-month traffic flow rate = (Qi-Qi-1) / Qi-1. For example, when the adjacent data points are 105 and 100, the change rate is 5%, indicating an increase in traffic. In this way, the month-on-month traffic flow rate of each data point (except the first point) is calculated and arranged into a sequence in time series. The sequence can intuitively present the short-term increase and decrease rate of vehicle traffic and provide dynamic change information for the traffic prediction model. If the sequence shows continuous large positive values, it means that the traffic is increasing rapidly; if the sequence shows continuous large negative values, it means that the traffic is decreasing. This information helps traffic management departments to respond in advance, adjust the brightness of street lights, ensure safety and improve energy efficiency.

[0029] The year-on-year flow rate change sequence acquisition subunit is used to obtain the year-on-year flow rate change sequence of the vehicle year-on-year flow sequence. Specifically, before the year-on-year flow rate change sequence acquisition subunit calculates the year-on-year flow rate change sequence, it is necessary to ensure that the vehicle year-on-year flow sequence data is accurate and the time range corresponds correctly. The sequence is obtained by fitting the adjacent time series of vehicle flow data in the same period of different years by the year-on-year flow sequence acquisition subunit, reflecting the long-term change trend. When calculating, except for the first point, the formula of year-on-year flow rate change = (Qi”-Qi') / Qi' is used for calculation. For example, if the vehicle flow on a certain day this year is 120 vehicles and on the same day last year is 100 vehicles, the change rate is 20%, indicating an increase this year compared with the same period last year. The year-on-year flow rate change rate of each data point (except the first point) is calculated in this way and arranged into a sequence in chronological order. This sequence can clearly show the increase or decrease of vehicle flow over a long span, and complements the month-on-month flow rate change rate sequence, providing more comprehensive information for the road flow prediction model. If the year-on-year flow rate change rate sequence shows a stable increase in the same period of each year in a certain period of time, combined with the recent rapid increase in the month-on-month flow rate change rate, future flow can be predicted more accurately, providing accurate data support for intelligent dimming of street lamps, etc., so that lighting can adapt to flow changes, ensure safety and energy-saving optimization.

[0030] The target road vehicle flow acquisition subunit is used to input the month-on-month flow rate change sequence, the vehicle month-on-month flow sequence, the year-on-year flow rate change sequence and the vehicle year-on-year flow sequence into the road flow prediction model embedded in the road flow analysis module to obtain the target road vehicle flow. Specifically, the target road vehicle flow acquisition subunit first collects data sets for building the road flow prediction model, which covers data such as the vehicle month-on-month flow record sequence, the year-on-year flow record sequence, the month-on-month flow rate change record sequence and the year-on-year flow rate change record sequence. The data are widely sourced and include past subunit processing data and historical database supplements to comprehensively and accurately present the target road vehicle flow change characteristics. Subsequently, the data set is equally divided into k groups (k≥5), and a training data set is obtained by k times of sampling with replacement. For example, after being divided into 5 groups, multiple random samplings are performed to obtain multiple different training data sets, thereby enhancing data diversity and improving the generalization ability of the model. Then, for each training data set, a long short-term memory neural network (LSTM) training is applied. Each sequence is used as input, and the model adjusts the neuron weights and continuously iterates and optimizes to make the predicted traffic close to the actual historical traffic. After each training, the performance and generalization ability are evaluated on the validation data, and the training is continued until the Mth road traffic prediction model is obtained based on the Mth training data set (M≥7). Finally, after obtaining multiple training models (such as 7 or more), the weighted average of the vehicle traffic prediction values ​​of each model at a given prediction time point is fully connected by the mean, such as calculating the average prediction value of the 7 models to be about 103.14, thereby obtaining the final road traffic prediction model that combines the advantages of each model.

[0031] In a possible implementation, the target road vehicle flow acquisition subunit further includes: a data set construction micro unit, which is used to collect vehicle month-on-month flow record sequence, vehicle year-on-year flow record sequence, month-on-month flow change rate record sequence and year-on-year flow change rate record sequence, and set it as a road flow prediction model construction data set. Specifically, the dataset construction micro-unit is responsible for collecting data for building the road traffic prediction model. The vehicle month-on-month traffic record sequence is derived from the continuous monitoring data of the target road at short time intervals (such as daily, weekly, etc.) (after adjacent time series fitting processing), such as the original sequence formed by recording the vehicle passing volume in different periods of several months through sensors, and then processing the month-on-month traffic record sequence; the vehicle year-on-year traffic record sequence is mined from the historical traffic flow database, extracting the data of the same time period in the past years, and finding the traffic data of specific periods of each year in the past five years to form the year-on-year traffic record sequence; the month-on-month traffic change rate record sequence is calculated by the adjacent month-on-month traffic data points (the next data point minus the previous data point divided by the previous data point) and the change rate of the adjacent points of the entire sequence is calculated in turn, which can show the short-term change speed of vehicle traffic; the year-on-year traffic change rate record sequence is similarly constructed, and the change rate of traffic calculation in the same period of different years is compared and sorted, which is complementary to the month-on-month traffic change rate record sequence. Finally, these sequences are integrated into the road traffic prediction model construction dataset, which is a collection of multi-dimensional vehicle traffic information. One record covers the vehicle month-on-month traffic, year-on-year traffic, month-on-month traffic change rate, year-on-year traffic change rate, etc. on a certain day, providing a rich data basis for model training.

[0032] The first training data set acquisition micro-unit is used to divide the data set for building the road traffic prediction model into k groups, perform k times of sampling with replacement, and obtain the first training data set, where k≥5. Specifically, after the first training data set acquisition micro-unit obtains the data set for building the road traffic prediction model, it proceeds to perform a grouping operation and divides it into k groups (k≥5). The grouping is for obtaining different training data sets for subsequent sampling with replacement. For example, when k=5, the data set will be divided into five equal parts, each containing a certain number of vehicle traffic data records. Although these records are random, they can represent the target road traffic flow situation as a whole. The grouping method can be divided in chronological order or adjusted after random division to ensure that the statistical characteristics of each group are relatively balanced and avoid special data affecting the representativeness of the sampling. After grouping is completed, k samplings with replacement are performed. In the first sampling, one group is randomly selected from the k groups (such as group 3), and its data records are entered into the first training data set. Due to sampling with replacement, group 3 still has the opportunity to be selected again later. The first training data set obtained after k sampling operations is a random subset of the original data set, which is used to train the first road flow prediction model. Its diversity helps the model learn different traffic flow patterns and improve the generalization ability to better adapt to the actual complex and changeable traffic flow conditions.

[0033] The replacement sampling micro-unit is used to divide the data set for building the road flow prediction model into k groups, perform k replacement samplings, and obtain the Mth training data set, where M≥7. Specifically, after the first training data set acquisition micro-unit completes the first k replacement samplings to obtain the first training data set, the replacement sampling micro-unit starts the subsequent work, and also divides the data set for building the road flow prediction model into k groups (the grouping method is the same as before), and then performs k replacement samplings to obtain a new training data set, which is marked as the Mth training data set (M starts from 7). If M=7, the seventh sampling is performed, and each group is randomly selected from the k groups, and its data is added to the Mth training data set, and each group has the same probability of being selected each time, so as to ensure the randomness of sampling and data diversity. The process will be repeated continuously with the replacement sampling micro-unit until the predetermined M value (M≥7) is reached. The purpose is to obtain enough different training data sets for training multiple road flow prediction models. The data sets are random and unique, and can reflect the characteristics of traffic flow from different angles. Subsequently, they can be used to train long and short-term memory neural networks to obtain models with different learning results, laying the foundation for model integration to improve prediction accuracy. If M=10 is set, ten samplings will be performed to obtain ten different training data sets for training ten models, making full use of data diversity to improve system performance.

[0034] The first road traffic prediction model obtains a micro-unit, and the first road traffic prediction model obtains a micro-unit for training a long short-term memory neural network according to the first training data set to obtain a first road traffic prediction model. Specifically, after the first road traffic prediction model obtains the micro-unit to obtain the first training data set, it starts to train the long short-term memory neural network (LSTM) to build a model. First, the LSTM model is initialized to determine structural parameters such as the number of hidden layers and the number of neurons, and reasonable adjustments are made according to the size of the data set and the characteristics of the traffic flow data. If the data set is large and the traffic flow changes are complex, the parameters can be appropriately increased to improve the learning ability. At the same time, the data such as the month-on-month and year-on-year traffic record series and the change rate record series in the first training data set are preprocessed, such as normalization processing to map them to a specific interval (such as [-1,1] or [0,1]), and the numerical range can be narrowed by dividing by the historical highest traffic value to meet the input requirements. Then the preprocessed data set is input into the initialized LSTM model to start training. The model calculates the predicted vehicle flow value based on the input data and compares it with the actual value to calculate the error. The mean square error (MSE) and mean absolute error (MAE) are commonly used. For example, the error is obtained by subtracting the predicted value from the actual value at a certain moment. Then, the back propagation algorithm is used to update the neuron weights from the output layer to the input layer according to the error to optimize the model. This continues for multiple rounds of training. As the number of rounds increases, the model's understanding of the traffic pattern becomes deeper and the error decreases. The training ends when the predetermined stopping condition is reached (such as the error is less than a threshold or the number of rounds reaches a maximum value, etc.). The first road flow prediction model obtained can preliminarily predict vehicle flow, and its learned feature patterns provide a basis for subsequent improved integration.

[0035] A neural network training micro-unit is used to train the long short-term memory neural network according to the Mth training data set to obtain the Mth road traffic prediction model. Specifically, the neural network training micro-unit operates on the Mth training data set (M≥7), and after the generation of the replacement sampling micro-unit, it is used to train the long short-term memory neural network to obtain the Mth road traffic prediction model. First, the LSTM model is initialized (the structure can be fine-tuned as needed but it is often kept the same as the previous model to ensure comparability) and the Mth training data set is preprocessed to meet the input requirements, and then the preprocessed data is input into the LSTM model to start training, calculate the error between the predicted value and the actual value, and use the back propagation algorithm to update the weight optimization model to minimize the error. Continue multiple rounds of training until the predetermined stop condition is met (such as setting an upper limit of 100 rounds, and the error in the 80th round is stable and minimum and can be stopped), so that each Mth training data set can obtain a corresponding model, which learns different traffic feature patterns and provides diverse prediction capabilities for integration. The neural network training micro-unit continues this process, starting from M=7 and increasing, and uses each Mth training data set to train the LSTM model. If 10 models are to be obtained (M=10), four models are trained using the 7th to 10th data sets, and the first model trained by the micro-unit is obtained by adding the first road flow prediction model. A total of 10 models are used to construct the final road flow prediction model, ensuring that there are enough different models involved in the integration to improve the prediction accuracy and stability.

[0036] A road flow prediction model acquisition unit, the road flow prediction model acquisition unit is used to perform mean full connection on the first road flow prediction model to the Mth road flow prediction model to obtain the road flow prediction model. Specifically, after the first road flow prediction model acquisition micro-unit and the neural network training micro-unit respectively train the first road flow prediction model to the Mth road flow prediction model (M≥7), the mean full connection operation is started. First, for a given prediction time point (such as 8-9 p.m. tomorrow), let each model output a prediction value based on the learned traffic flow pattern characteristics (assuming that the prediction values ​​of the seven models are 100, 110, 95, 105, 102, 108, 98, etc.), and then determine the weighted average weight based on the model performance evaluation results. When the performance is similar, the simple average can be used. For example, the above example uses a simple average calculation, and the mean (100+110+95+105+102+108+98) / 7≈103.14 is the final prediction result at that time point. Then, the prediction value acquisition and weighted average calculation process are repeated for different prediction time points, and the prediction values ​​of each time point of the final road flow prediction model are calculated in turn, thereby constructing a complete final prediction model.

[0037] In a possible implementation, the adjacent time series fitting subunit further includes: a time domain traffic acquisition micro unit, which is used to obtain the i-th time domain traffic and the i+1-th time domain traffic, wherein the i-th time domain and the i+1-th time domain are adjacent time domains. Specifically, the work of the time domain traffic acquisition micro-unit begins with obtaining the i-th time domain traffic and the i+1-th time domain traffic from the collected vehicle traffic data. These data are composed of the number of vehicles passing through recorded by various sensors such as induction coils, geomagnetic sensors, and cameras installed on the target road at different time points. Although they may have been preliminarily organized in chronological order, it is necessary to accurately extract the adjacent time domain traffic data. For example, when recording every hour, the i-th time domain may be 9:00-10:00 in the morning, and the i+1-th time domain is 10:00-11:00 in the morning. The unit reads its vehicle traffic value from the database or data structure to prepare for subsequent analysis. At the same time, when reading the data, it is clear that the i-th time domain and the i+1-th time domain are adjacent time domains, that is, they are continuous and have no intervals in time. For example, when taking days as the unit, the i-th day and the i+1-th day are adjacent time domains, and when taking minutes as the unit, two adjacent minute periods are adjacent time domains. The selection depends on the data collection frequency and analysis accuracy requirements. Paying attention to the changes in traffic in adjacent time domains can capture short-term fluctuations and trends in traffic flow, and provide a basis for subsequent merging or retention decisions.

[0038] A time domain merging micro-unit is used to merge the i-th time domain and the i+1-th time domain to obtain a fitting time domain when the flow deviation between the i-th time domain flow and the i+1-th time domain flow is less than or equal to a fitting flow deviation threshold. Specifically, after obtaining the i-th time domain flow and the i+1-th time domain flow, the time domain merging micro-unit uses the formula deviation = (i+1-th time domain flow - i-th time domain flow) / i-th time domain flow to calculate the flow deviation between the two. For example, when the vehicle flow in the i-th time domain is 100 vehicles and the vehicle flow in the i+1-th time domain is 120 vehicles, the deviation is (120-100) / 100=0.2 (i.e., 20%). Then, the deviation is compared with the fitting flow deviation threshold determined based on the experience and data analysis of the target road traffic flow change law. If the flow deviation is less than or equal to this threshold, it means that the vehicle flow changes in the two time domains are small and similar, and a merge operation can be performed. The merged time domain range covers the original two time domains. When analyzing the flow of two adjacent days in units of days, the merged time domain can be expressed as the i-th day-i+1-th day. The merge operation can reduce short-term data fluctuations and make the vehicle month-on-month flow time series information smoother, thereby more accurately reflecting the long-term trend and change law of traffic flow, and providing a more reliable data basis for subsequent flow prediction and analysis.

[0039] A timing information updating micro-unit, the timing information updating micro-unit is used to use the average of the i-th time domain flow and the i+1-th time domain flow as the fitted time domain flow, and update the vehicle ratio flow timing information according to the fitted time domain and the fitted time domain flow. Specifically, when it is determined that a time domain merger is to be performed, the timing information updating micro-unit uses the average of the i-th time domain flow and the i+1-th time domain flow as the fitted time domain flow, and its calculation formula is (i-th time domain flow + i+1-th time domain flow) / 2. For example, when the i-th time domain flow is 80 vehicles and the i+1-th time domain flow is 90 vehicles, the fitted time domain flow is (80+90) / 2=85 vehicles. This average value, as the representative value of the flow in the merged time domain, can comprehensively reflect the traffic flow level by integrating the two time domain flow information, and then the vehicle ratio flow timing information is updated according to this fitted time domain and the fitted time domain flow. Information is obtained, that is, the i-th time domain and the i+1-th time domain and their corresponding traffic data recorded in the original vehicle month-on-month traffic time series information are replaced with the merged fitting time domain and fitting time domain traffic data. In this way, the number of time domains in the vehicle month-on-month traffic time series information will be reduced by one, and the traffic flow characteristics represented by each time domain will be more stable and representative. With the continuous updating operation, the vehicle month-on-month traffic time series information will gradually change from the original data with more short-term fluctuations to a data sequence that can clearly show the long-term change trend of traffic flow, providing more accurate and valuable input data for the subsequent traffic prediction model.

[0040] A cyclic analysis micro-unit, the cyclic analysis micro-unit is used for cyclic analysis until the traffic deviation between any two adjacent time domains is greater than the fitting traffic deviation threshold, and the vehicle ratio traffic sequence is output. Specifically, the cyclic analysis micro-unit dominates the entire cyclic analysis process, first calling the time domain traffic acquisition micro-unit to obtain the initial i-th time domain traffic and i+1-th time domain traffic, and then passing it to the time domain merging micro-unit for deviation calculation and merging decision. If the time domain merging micro-unit determines that merging is not necessary (that is, the traffic deviation is greater than the fitting traffic deviation threshold), the cycle stops. If merging is required, after the timing information update micro-unit completes the information update, the cyclic analysis micro-unit calls the time domain traffic acquisition micro-unit again to obtain new adjacent time domain traffic (the i-th time domain and i+1-th time domain are now updated), and continues to perform deviation calculation and merging decisions. This cycle continues until the stop condition is met. When the cycle stops, that is, when the flow deviation between any two adjacent time domains is greater than the fitting flow deviation threshold, the cyclic analysis micro-unit outputs a vehicle month-on-month flow sequence. This sequence is subjected to a series of time domain merging and information updating operations to remove some short-term fluctuations in the original data and more smoothly present the changing trend of vehicle flow over a longer time range. It serves as the key data basis for the subsequent calculation of the month-on-month flow change rate sequence and the input road flow prediction model, which is of great significance for the accurate prediction of vehicle flow on the target road. It helps traffic management departments or related systems to deeply understand the dynamic changes of traffic flow and make reasonable decisions, such as adjusting the brightness of street lights to adapt to different traffic flows and improve road safety and energy efficiency.

[0041] The intelligent dimming module 20 of street lamps is used to optimize the brightness control of street lamps according to the predicted traffic flow of the target road, and obtain the control strategy of street lamps on preset roads. Specifically, after receiving the predicted traffic flow data of the target road from the road traffic analysis module, the intelligent dimming module 20 of street lamps first checks and organizes it to ensure that the data is complete and accurate. If problems are found, appropriate methods will be used to repair and adjust. Then, based on these data, the optimization work of street lamp brightness control is carried out, and the brightness requirements of street lamps under different traffic conditions are comprehensively considered. The relationship between traffic flow and brightness setting in historical data, relevant lighting standards and specifications, and the performance parameters of street lamps themselves are referred to. Possible optimization algorithms are used to determine the optimal street lamp brightness value under different predicted traffic flows under the premise of meeting lighting requirements, so as to obtain the control strategy of street lamps on preset roads, provide a basis for subsequent street lamp dimming control, realize intelligent matching of street lamp brightness and traffic flow, and balance traffic safety, lighting quality and energy efficiency. For example, when the traffic flow is low in the middle of the night, it is adjusted to the lowest energy-saving brightness, and when the peak hours in the morning and evening are adjusted to the highest standard brightness.

[0042] In one possible implementation, Figure 2As shown, the street lamp intelligent dimming module 20 further includes: a target road predicted flow composition unit, and the target road predicted flow composition unit is used for the target road predicted flow including vehicle predicted flow and crowd predicted flow. Specifically, the target road predicted flow composition unit is responsible for integrating the predicted flow information from different data sources, and dividing it into two parts: vehicle predicted flow and crowd predicted flow. The vehicle predicted flow is mainly based on data such as the vehicle month-on-month flow record sequence, the vehicle year-on-year flow record sequence, the month-on-month flow change rate record sequence, and the year-on-year flow change rate record sequence. It is obtained through a series of complex data processing and analysis (such as the previously mentioned data set construction, model training and prediction processes). The data reflects the traffic flow change rules of vehicles on different time scales, and can provide an important basis for the subsequent street lamp brightness control. For example, when a large number of vehicles are predicted to pass during peak traffic hours. , street lamps need to provide higher brightness to ensure driving safety, and the predicted crowd flow is obtained by analyzing the time series information of the crowd's month-on-month flow and the crowd's year-on-year flow. The collection of crowd flow data may involve pedestrian counters installed around roads or intersections, pedestrian recognition systems based on image processing technology, and some intelligent monitoring equipment. The equipment can monitor the number and flow direction of pedestrians in real time. After data processing and analysis, the predicted crowd flow is obtained. Changes in crowd flow will also affect the demand for street lamp brightness. For example, in crowded areas or periods of frequent pedestrian activities, the brightness of street lamps needs to be appropriately increased to ensure pedestrian safety and visual comfort.

[0043] A multi-level adjacency analysis unit, the multi-level adjacency analysis unit is used to perform a multi-level adjacency analysis according to the predicted vehicle flow and the predicted crowd flow, and obtain a set of street lamp brightness record values. Specifically, after obtaining the predicted vehicle flow and the predicted crowd flow, the multi-level adjacency analysis unit starts to perform a multi-level adjacency analysis. First, a correlation relationship between the flow and the street lamp brightness must be established. According to historical data and relevant lighting standards, different traffic flows (including vehicles and crowds) correspond to different optimal street lamp brightness requirements. For example, when the traffic flow is large, in order to ensure that traffic participants can clearly see the road conditions and the surrounding environment, a higher street lamp brightness is required; when the traffic flow is small, the brightness can be appropriately reduced to achieve energy saving. The analysis process may take into account a variety of factors, such as road type (main roads, secondary roads, branch roads, etc.), surrounding environment (commercial areas, residential areas, industrial areas, etc.) and time factors (daytime, night, different seasons, etc.). For different types of roads, their traffic flow characteristics and lighting requirements are different. Main roads usually have large traffic volumes and fast speeds, requiring higher brightness and better lighting uniformity; while roads in residential areas have more pedestrians at night, so the road traffic volume is relatively large. The requirement for street lamp brightness may focus more on the visual experience and safety of pedestrians. Different surrounding environments will also affect the demand for street lamp brightness. Commercial areas may require brighter lights to create a good business atmosphere, while industrial areas may adjust the brightness of street lamps according to production operation time and safety requirements. The multi-level adjacency analysis unit determines the corresponding street lamp brightness value according to different combinations of vehicle predicted flow and crowd predicted flow. The brightness value will form a street lamp brightness record value set. For example, in a certain time period, the predicted vehicle flow is 500 vehicles per hour and the crowd flow is 200 people per hour. After analysis, it is determined that the street lamp brightness at this time should be 80% (assuming that the street lamp brightness can be adjusted in the range of 0-100%). The brightness value will be recorded in the set. With the analysis of different flow combinations, multiple street lamp brightness record values ​​will gradually accumulate in the set. The record value reflects the recommended street lamp brightness under various possible traffic flow scenarios, providing a data basis for subsequent central trend evaluation.

[0044] A central tendency evaluation unit, the central tendency evaluation unit is used to evaluate the central tendency of the street light brightness record value set to obtain the street light brightness recommended interval. Specifically, the central tendency evaluation unit evaluates the central tendency of the street light brightness record value set, and uses statistical methods to analyze the data in the set. Common central tendency indicators include mean, median and mode, etc. Calculating these indicators can help understand the distribution center position of the street light brightness record values ​​in the data set. For example, calculating the mean of the street light brightness record values ​​can obtain an average brightness recommendation value, which reflects the overall brightness tendency; the median can avoid the influence of extreme values ​​on the results and provide a more robust intermediate brightness value; the mode can show the brightness value with the highest frequency in the data set, which may represent the street light brightness requirements under the most common traffic flow conditions.

[0045] The recommended interval setting unit is used to set the street lamp brightness recommended interval to the preset road street lamp control strategy. Specifically, according to the calculated central tendency index, the street lamp brightness recommended interval is determined in combination with the actual situation. The determination of the interval needs to take into account the lighting safety and comfort, as well as the energy-saving requirements. For example, if the mean is 70% and the standard deviation is 10% (the standard deviation is used to measure the discreteness of the data), the recommended interval can be determined according to certain rules (such as the mean plus or minus a certain multiple of the standard deviation), such as [60%, 80%]. The street lamp brightness within this interval can meet the lighting needs in most cases.

[0046] In a possible implementation, the multi-level adjacency analysis unit further includes: a street light brightness record value acquisition subunit, the street light brightness record value acquisition subunit is used to perform adjacency retrieval of at least 5 samples according to the vehicle predicted flow and the crowd predicted flow, and obtain a first-level street light brightness record value, wherein the first-level street light brightness record value has a first-level vehicle record flow identifier and a first-level crowd record flow identifier. Specifically, the street light brightness record value acquisition subunit first obtains the vehicle predicted flow and crowd predicted flow data obtained through complex analysis and prediction, which contains estimated information on the number of vehicles and people on the target road at different time periods in the future. When performing an adjacency search, it is necessary to set appropriate search conditions, such as setting a flow range centered on the current predicted flow as a boundary condition. This range setting needs to take into account the traffic flow change law and data distribution to ensure that a sufficient number of representative samples are obtained. Then, according to the set conditions, an adjacent search is performed for at least 5 samples in the database or data warehouse that stores information such as past vehicle flow, crowd flow and corresponding street lamp brightness settings to find the street lamp brightness value corresponding to the historical flow conditions similar to the current vehicle and crowd predicted flow, that is, the first-level street lamp brightness record value. Each record value will be marked with the corresponding first-level vehicle record flow identifier (such as vehicle flow of 300-350 vehicles per hour) and first-level crowd record flow identifier (such as crowd flow of 150-200 people per hour) to facilitate subsequent analysis and tracing, and to clarify the traffic flow scenario generated by each brightness value.

[0047] The adjacent search subunit is used to traverse the first-level street light brightness record value according to the first-level vehicle record flow identifier and the first-level crowd record flow identifier to perform adjacent search for at least 5 samples to obtain the second-level street light brightness record value. Specifically, after the adjacent search subunit receives the first-level street light brightness record value and its corresponding first-level vehicle and crowd record flow identifier, it adjusts the search conditions based on these identifiers. In view of the fact that the first-level street light brightness record value is derived from historical traffic conditions similar to the current predicted traffic, resetting the search range with its flow identifier as the center can more accurately find similar samples. For example, when the first-level vehicle record flow identifier shows that the vehicle flow is 300-350 vehicles per hour, the second-level search can narrow the vehicle flow range to about 320-330 vehicles per hour. At the same time, the crowd flow search range is adjusted accordingly according to the first-level crowd record flow identifier, and the conditions are refined to improve the accuracy and relevance of the search results. Then, based on the adjusted conditions, the first-level street light brightness record values ​​are traversed to perform adjacent searches for at least 5 samples. This process is similar to the search of the street light brightness record value acquisition sub-unit but with a more precise range. The second-level street light brightness record values ​​are obtained through a secondary search. These record values ​​carry second-level vehicle and crowd record flow identifiers, recording more accurate flow conditions and providing more detailed information for subsequent analysis.

[0048] A loop analysis subunit, the loop analysis subunit is used for loop analysis until the L-level street light brightness record value is obtained. Specifically, the loop analysis subunit first starts the street light brightness record value acquisition subunit to obtain the first-level street light brightness record value, and then passes it to the adjacent search subunit to obtain the second-level street light brightness record value, and then determines whether the stop condition is met. If the predetermined analysis level L is not reached, the second-level street light brightness record value and its identifier are used as new inputs, and the adjacent search subunit is called again to start the next level (third level) search. This cycle is repeated, and each cycle will optimize the next level search range based on the previous level street light brightness record value and its identifier, so that the search conditions are more accurate and a hierarchical progressive analysis is achieved. Through continuous and in-depth retrieval, street light brightness record values ​​that match the current predicted traffic at different accuracies are gradually found, so as to comprehensively consider various possible traffic flow scenarios. In each cycle, it is determined whether the current analysis level reaches the L level. Once it reaches, the loop search is stopped. At this time, the street light brightness record value obtained is the L-level street light brightness record value. After multiple cycles of optimization and progressive analysis, the recorded values ​​of the L-level street lamp brightness are highly correlated and accurate with the current predicted vehicle flow and crowd flow, and can provide a reliable data basis for the final determination of the recommended range of street lamp brightness.

[0049] The brightness record value adding subunit is used to add the first-level street light brightness record value, the second-level street light brightness record value, and the L-level street light brightness record value into the street light brightness record value set. Specifically, before performing the record value adding operation, first check whether the street light brightness record value set exists. If it does not exist, create a new empty set to store the subsequent street light brightness record values ​​of each level. Then, the first-level street light brightness record value obtained by the street light brightness record value acquisition subunit is added to the set. This is the starting data for analysis and can be used as a reference for subsequent analysis. Subsequently, when the adjacent retrieval subunit obtains the second-level, third-level, and L-level street light brightness record values ​​in turn, they are added to the set one by one to ensure that the set covers all record values ​​from the first level to the L level, forming a complete data set, and providing comprehensive data support for subsequent operations such as centralized trend evaluation. At the same time, during the adding process, the record value is always associated with its corresponding vehicle and crowd record flow identifier, so that the traffic flow scene and meaning corresponding to each record value can be accurately understood in the subsequent analysis.

[0050] In a possible implementation, the street lamp brightness record value acquisition subunit further includes: a deviation threshold configuration microunit, which is used to configure the vehicle flow deviation threshold and the crowd flow deviation threshold. Specifically, the deviation threshold configuration microunit collects historical vehicle and crowd flow data covering different time periods, dates, weather conditions, etc., as well as street lamp brightness setting data, such as relevant data for different time periods every day in the past year, in the data collection and analysis preparation stage, so as to fully grasp the traffic flow change law. Then, the data is preprocessed to check the integrity and accuracy, and the missing values ​​and abnormal values ​​are processed. The interpolation method or the mean value filling method can be used to process the missing data, and the abnormal data can be verified and corrected to provide a reliable basis for threshold analysis. In the threshold calculation and configuration link, the vehicle and crowd flow deviation thresholds are calculated by comprehensively considering the data distribution characteristics, the degree of discreteness and the actual scene requirements. For example, the vehicle flow deviation threshold can be adjusted with reference to the standard deviation or its multiple and combined with experience and road characteristics. The threshold of the road with stable traffic flow is set narrower, and the fluctuation is large. Finally, the threshold is configured as the basis for sample analysis and judgment to ensure that the street lamp brightness reference data is effectively obtained in different scenarios, taking into account accuracy, flexibility and adaptability.

[0051] The sample analysis micro-unit is used to add the first-level street lamp brightness record value when the sample to be analyzed satisfies the vehicle flow deviation threshold and the crowd flow deviation threshold at the same time. Specifically, the sample analysis micro-unit obtains the sample to be analyzed from the data set containing information such as vehicle predicted flow, crowd predicted flow and street lamp brightness in the sample acquisition and flow deviation calculation link, and then calculates the deviation according to the method of vehicle flow deviation = (vehicle flow of the sample to be analyzed - benchmark vehicle flow) / benchmark vehicle flow, crowd flow deviation = (crowd flow of the sample to be analyzed - benchmark crowd flow) / benchmark crowd flow. The benchmark can be the historical average flow or the reference flow for a specific time period, such as using the average flow of the past month as the benchmark to calculate the current sample deviation. In the deviation threshold judgment and sample screening stage, the calculated vehicle and crowd flow deviations are compared with the configured corresponding thresholds. When the vehicle flow deviation meets the threshold range and the crowd flow deviation also meets the requirements, such as the vehicle flow deviation threshold is ±20% and the crowd flow deviation threshold is ±15%, and a sample vehicle flow deviation of 15% and crowd flow deviation of 10% meet the requirements, the sample is added to the first-level street lamp brightness record value set, providing accurate data support for the reasonable setting of street lamp brightness, helping to achieve accurate matching of street lamp lighting and traffic flow, and improving energy utilization efficiency and lighting effects.

[0052] The local communication module 30 is used to monitor the target road traffic information in the target time zone. Specifically, when the local communication module 30 monitors the target road traffic information in the target time zone, it first completes the deployment and initialization of various sensors and communication equipment, establishes a stable wireless communication connection with the centralized controller, and ensures the normal operation of the equipment and smooth communication. Then, within the target time zone, the vehicle and crowd flow data are continuously collected from the sensors, including the number of vehicles, passing time, speed and number of pedestrians, flow direction, etc., and collected and preliminarily processed and integrated at preset time intervals. Finally, the collected and processed data is transmitted to the centralized controller in a timely manner, and encryption technology is used to ensure data security. It has a cache function to cope with communication interruptions, ensure data continuity and integrity, and provide accurate data support for decisions such as intelligent dimming of the street light control system.

[0053] The private network communication module 40 is used to determine whether the road flow deviation between the target road flow information and the target road predicted flow is less than or equal to the road flow deviation threshold. Specifically, when the private network communication module 40 is working, it first receives the target road flow information and the target road predicted flow data, ensures that the data is complete and accurate, and performs verification analysis and alignment to make the two comparable. Then the road flow deviation is calculated, and the formula is (target road flow information-target road predicted flow) / target road predicted flow, and at the same time, a preset road flow deviation threshold is obtained, which is determined by multiple factors. Finally, the calculated deviation is compared with the threshold. If the deviation is less than or equal to the threshold, a lighting state maintenance instruction is sent to the street light dimming module 50 and data is uploaded; if the deviation is greater than the threshold, an adjustment lighting instruction is sent to the street light dimming module 50 and data is reported. The server recalculates the dimming instruction accordingly and sends it, so as to achieve dynamic matching of street light lighting and traffic flow, and improve lighting effects and energy efficiency.

[0054] The street light dimming control module 50 is used to perform street light dimming control according to the preset road street light control strategy when the road flow deviation is less than or equal to the road flow deviation threshold. Specifically, when the street light dimming control module 50 is working, it first receives the instruction of the private network communication module 40. When the road flow deviation is less than or equal to the threshold, it obtains the preset road street light control strategy, and parses and verifies to ensure its completeness and correctness. Then, according to the strategy, the brightness value that the street light should be set is determined, and the dimmable power supply is controlled to adjust the output parameters to achieve precise dimming. During the process, the actual brightness is monitored in real time. If there is a deviation, it is calibrated in time, and the dimming data is recorded for analysis and optimization. After the dimming is completed, the status information is fed back to the private network communication module 40 to facilitate the server to record and analyze, and then the road flow changes are continuously monitored, waiting for new instructions, and the process is repeated to dynamically adjust the street light brightness to ensure that the road lighting is suitable, improve the efficiency of traffic and energy utilization, and reduce maintenance costs and energy consumption.

[0055] The brightness control optimization module 60 is used to optimize the street light brightness control according to the target road traffic information when the road traffic deviation is greater than the road traffic deviation threshold, and obtain a real-time road street light control strategy to perform street light dimming control. Specifically, the brightness control optimization module 60 receives the road traffic deviation information and the target road traffic information from the private network communication module 40. When the road traffic deviation is greater than the threshold, the target road traffic information (including vehicle traffic, crowd traffic, vehicle type, pedestrian direction, etc.) is deeply analyzed, and a lighting demand model is constructed based on the classification and evaluation of the different road users' demand for lighting. Then, combined with the actual parameters of the street light system, an optimization algorithm (such as a genetic algorithm, a particle swarm optimization algorithm, etc.) is used to optimize the street light brightness control. In the process, constraints such as the upper and lower limits of street light brightness and the coordination of adjacent street light brightness are considered, and a real-time road street light control strategy is obtained through iterative calculation. Then, according to this strategy, control instructions are sent to the street lamp dimming control module 50 to adjust the brightness of the street lamp. During the dimming process, the actual brightness of the street lamp and the road traffic conditions are continuously monitored, the actual brightness is compared with the target brightness and the deviation is calibrated in time, and the lighting demand model and control strategy are updated according to the traffic changes to realize the intelligence and efficiency of the street lamp lighting system, and the relevant data are fed back to other modules to support system optimization.

[0056] The embodiment of the present application adopts a road traffic analysis module to predict the target road traffic through the month-on-month and year-on-year traffic time series information, and generate the target road predicted traffic; the street lamp intelligent dimming module formulates a preset street lamp control strategy based on the predicted traffic. The local communication module monitors the road traffic in real time, and the private network communication module compares the deviation between the real-time traffic and the predicted traffic. When the deviation is within the threshold range, the street lamp dimming control module adopts the preset control strategy; when the deviation exceeds the threshold, the brightness control optimization module optimizes and adjusts according to the real-time traffic, generates a real-time control strategy and executes dimming, so as to achieve the technical effect of combining precise lighting and efficient energy saving by quickly adjusting the street lamp brightness according to the dynamic changes of road traffic.

[0057] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0058] Based on the foregoing embodiments, the embodiments of the present application further provide an electronic device and a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of the electronic device, it can implement the method described in any of the foregoing embodiments.

[0059] Figure 3 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing an implementation of the present invention. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention. The electronic device is in the form of a general computing device, and its components may include but are not limited to an input device 301, a processor 302, a memory 303, and an output device 304. Among them, the processor 302 may be one or more; the memory 303 may include a computer-readable medium and at least one program product, and the program product has a set (at least one) of program modules, which are configured to perform the functions of each embodiment of the present application.

[0060] The memory 303 shown in the embodiment of the present invention may adopt any combination of one or more computer-readable media; the computer-readable storage medium may be, but is not limited to, an infrared, semiconductor system, device or component, or any combination of the above, for storing software programs, computer executable programs and modules, such as the program instructions / modules corresponding to a multi-scenario multi-language automatic recognition method in the embodiment of the present invention. The processor 302 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 303, thereby realizing the above-mentioned high-efficiency and energy-saving smart street lamp control system based on road traffic AI recognition and analysis.

[0061] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An efficient and energy-saving intelligent street light control system based on road traffic AI recognition and analysis, characterized by: include: A road traffic analysis module, the road traffic analysis module is used to perform traffic forecasting based on the target road's quarter-on-quarter traffic time series information and the target road's year-on-year traffic time series information in the target time zone, to obtain the target road's predicted traffic; A street lamp intelligent dimming module, which is used to optimize the street lamp brightness control according to the predicted traffic flow of the target road and obtain a preset road street lamp control strategy; A local communication module, the local communication module is used to monitor target road traffic information in the target time zone; A private network communication module, the private network communication module is used to determine whether a road flow deviation between the target road flow information and the target road predicted flow is less than or equal to a road flow deviation threshold; A street light dimming control module, the street light dimming control module is used to perform street light dimming control according to the preset road street light control strategy when the road flow deviation is less than or equal to the road flow deviation threshold; The brightness control optimization module is used to optimize the brightness control of street lamps according to the target road traffic information when the road traffic deviation is greater than the road traffic deviation threshold, and obtain a real-time road lamp control strategy to perform street lamp dimming control.

2. The system according to claim 1, characterized in that The road flow analysis module comprises: A ring-to-month traffic flow time series information composition unit, wherein the ring-to-month traffic flow time series information composition unit is used for the target road ring-to-month traffic flow time series information including vehicle ring-to-month traffic flow time series information and crowd ring-to-month traffic flow time series information; A year-on-year traffic flow time series information composition unit, wherein the year-on-year traffic flow time series information composition unit is used for the target road year-on-year traffic flow time series information including vehicle year-on-year traffic flow time series information and crowd year-on-year traffic flow time series information; A traffic flow prediction unit, the traffic flow prediction unit is used to perform traffic flow prediction based on the vehicle month-on-month traffic flow time series information and the vehicle year-on-year traffic flow time series information to obtain the vehicle flow of the target road; A crowd-road-vehicle flow acquisition unit is used to perform flow prediction based on the crowd month-on-month flow time series information and the crowd year-on-year flow time series information to obtain the crowd-road-vehicle flow.

3. The system according to claim 2, characterized in that The traffic prediction unit comprises: An adjacent time series fitting subunit, the adjacent time series fitting subunit is used to perform adjacent time series fitting on the vehicle cycle flow time series information to obtain a vehicle cycle flow sequence; A year-on-year traffic sequence acquisition subunit, the year-on-year traffic sequence acquisition subunit is used to perform adjacent time series fitting on the vehicle year-on-year traffic time series information to obtain the vehicle year-on-year traffic sequence; A month-on-month flow rate change sequence acquisition subunit, the month-on-month flow rate change sequence acquisition subunit is used to obtain the month-on-month flow rate change sequence of the vehicle month-on-month flow sequence; A year-on-year flow rate change sequence acquisition subunit, the year-on-year flow rate change sequence acquisition subunit is used to obtain the year-on-year flow rate change sequence of the vehicle year-on-year flow sequence; A target road vehicle flow acquisition subunit is used to input the month-on-month flow change rate sequence, the vehicle month-on-month flow sequence, the year-on-year flow change rate sequence and the vehicle year-on-year flow sequence into a road flow prediction model embedded in the road flow analysis module to obtain the target road vehicle flow.

4. The system according to claim 3, characterized in that: The target road vehicle flow acquisition subunit includes: A data set construction micro-unit is used to collect a vehicle month-on-month flow record sequence, a vehicle year-on-year flow record sequence, a month-on-month flow change rate record sequence, and a year-on-year flow change rate record sequence, and is set as a road flow prediction model construction data set; A first training data set acquisition micro-unit, wherein the first training data set acquisition micro-unit is used to divide the road flow prediction model construction data set into k groups, perform k times of sampling with replacement, and obtain a first training data set, wherein k ≥ 5; Replace the sampling micro-unit, the replacement sampling micro-unit is used to divide the road flow prediction model construction data set into k groups, perform k times of replacement sampling, and obtain the Mth training data set, where M≥7; a first road flow prediction model acquisition micro-unit, wherein the first road flow prediction model acquisition micro-unit is used to train a long short-term memory neural network according to the first training data set to obtain a first road flow prediction model; A neural network training micro-unit, wherein the neural network training micro-unit is used to train a long short-term memory neural network according to the Mth training data set to obtain an Mth road flow prediction model; Perform mean full connection on the first road flow prediction model to the Mth road flow prediction model to obtain the road flow prediction model.

5. The system according to claim 3, characterized in that The adjacent timing fitting subunit comprises: A time domain traffic acquisition micro-unit, wherein the time domain traffic acquisition micro-unit is used to obtain the i-th time domain traffic and the i+1-th time domain traffic, wherein the i-th time domain and the i+1-th time domain are adjacent time domains; A time domain merging micro-unit, wherein the time domain merging micro-unit is used to merge the i-th time domain and the i+1-th time domain to obtain a fitting time domain when the flow deviation between the i-th time domain flow and the i+1-th time domain flow is less than or equal to a fitting flow deviation threshold; A timing information updating micro-unit, wherein the timing information updating micro-unit is used to take the average of the i-th time domain flow and the i+1-th time domain flow as the fitting time domain flow, and update the vehicle ratio flow timing information according to the fitting time domain and the fitting time domain flow; A cyclic analysis micro-unit is used for cyclic analysis until the flow deviation between any two adjacent time domains is greater than a fitting flow deviation threshold, and the vehicle ring flow sequence is output.

6. The system according to claim 1, wherein: The street lamp intelligent dimming module comprises: A target road predicted flow composition unit, wherein the target road predicted flow composition unit is used for the target road predicted flow to include vehicle predicted flow and crowd predicted flow; A multi-level adjacency analysis unit, the multi-level adjacency analysis unit is used to perform a multi-level adjacency analysis according to the predicted vehicle flow and the predicted crowd flow to obtain a set of street lamp brightness record values; A central tendency evaluation unit, the central tendency evaluation unit is used to perform a central tendency evaluation on the street lamp brightness record value set to obtain a street lamp brightness recommended interval; A recommended interval setting unit, the recommended interval setting unit is used to set the street lamp brightness recommended interval as the preset road street lamp control strategy.

7. The system according to claim 6, characterized in that The multi-level adjacency analysis unit comprises: A street light brightness record value acquisition subunit, the street light brightness record value acquisition subunit is used to perform an adjacent search for at least 5 samples according to the vehicle predicted flow and the crowd predicted flow to obtain a first-level street light brightness record value, wherein the first-level street light brightness record value has a first-level vehicle record flow identifier and a first-level crowd record flow identifier; An adjacency search subunit, the adjacency search subunit is used to traverse the first-level street lamp brightness record value to perform adjacency search for at least 5 samples according to the first-level vehicle record flow identifier and the first-level crowd record flow identifier, and obtain the second-level street lamp brightness record value; A cyclic analysis subunit, the cyclic analysis subunit is used for cyclic analysis until the L-level street lamp brightness record value is obtained; A brightness record value adding subunit is used to add the first-level street lamp brightness record value, the second-level street lamp brightness record value, and up to the L-level street lamp brightness record value into the street lamp brightness record value set.

8. The system according to claim 7, characterized in that The street lamp brightness record value acquisition subunit includes: A deviation threshold configuration micro unit, wherein the deviation threshold configuration micro unit is used to configure a vehicle flow deviation threshold and a crowd flow deviation threshold; The sample analysis micro unit is used to add the first-level street lamp brightness record value when the sample to be analyzed satisfies both the vehicle flow deviation threshold and the crowd flow deviation threshold.

9. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; The processor is used to implement the efficient and energy-saving smart street lamp control system based on road traffic AI recognition and analysis as described in any one of claims 1 to 7 when executing the executable instructions stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, an efficient and energy-saving smart street lamp control system based on road traffic AI recognition and analysis as described in any one of claims 1 to 7 is implemented.

Images