AI-based self-adaptive garage lighting method

Abstract

The invention provides an AI-based self-adaptive garage illumination method, relates to the technical field of artificial intelligence, and aims to solve the problems that a parking lot illumination system cannot be flexibly adjusted according to illumination intensity and cannot guarantee sufficient brightness and energy conservation at the same time. Comprising the following steps: A1, monitoring camera data acquisition: acquiring data of vehicles entering and exiting a garage through a monitoring camera, wherein the data comprises videos and license plate numbers of the vehicles entering and exiting the garage; a2, weather sensing equipment data acquisition: acquiring weather data outside the garage through the weather sensing equipment, wherein the weather data comprises illumination intensity, visibility, temperature and humidity; aiming at the problem that the energy loss rate is too high in the existing garage lighting technology, the invention provides a new optimization scheme for the existing garage lighting system, and by introducing the artificial intelligence technology, the decision-making ability of the lighting strategy is greatly improved, and the method plays an important role in reducing the cost, saving the energy and reducing the maintenance cost.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and more specifically, particularly relates to an AI-based adaptive garage lighting method. Background Art

[0002] With the development of the economy and the improvement of people's living standards, the ownership of motor vehicles is increasing rapidly. Along with it comes the problem of parking difficulty. On the one hand, the growth of public parking facilities is slow, resulting in a shortage of parking spaces; on the other hand, the utilization rate of existing parking spaces is relatively low, and it is difficult to meet the demand during peak periods. The lighting system in the parking lot plays a very important role in guiding vehicles in and out of parking spaces, improving parking accuracy, and ensuring the safety of the parking area.

[0003] However, currently in parking lots, especially outdoor parking lots, a lighting system with a single brightness is often adopted, usually for night lighting. There is no difference in the lighting intensity on sunny days and cloudy days. However, the night lighting intensity on sunny days is much higher than that on cloudy days. Therefore, the parking lot lighting system cannot be flexibly adjusted according to the lighting intensity, and cannot save energy while ensuring sufficient brightness. At the same time, the vehicle and personnel flow patterns in the parking lot are different from those on the road. During the day, the time for people going to work to enter and leave vehicles is concentrated; at night, some vehicles are going out and some vehicles are going home. In addition, on rainy days and snowy days, the utilization rate of the parking lot drops significantly. In current intelligent road systems, etc., the vehicle and personnel flow is changing in real time, and it is often used as an input feature to control traffic signals through machine learning algorithms. In the parking lot, the utilization rate changes more with time, and there are no effective input features.

[0004] The typical practice of current intelligent lighting systems is to use light sensors or infrared sensing detectors to sense the lighting intensity or pedestrians and vehicles in real time, and then intelligently control the lighting according to the set threshold or manual observation. The defect of this kind of method lies in the complexity of setting the threshold. For example, the outdoor lighting intensity changes greatly during the day. If the threshold is set too high, the darkness may not be enough at night; if the threshold is set too low, the lighting may be turned on and off too frequently during the day. Secondly, when dealing with input features, this kind of method pays more attention to the primary and secondary relationships of each element. For example, the weights of day and night, sunny day and cloudy day are different. This is completely different from the utilization rate in different situations in the parking lot. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides an AI-based adaptive garage lighting method, a parking lot lighting system that can ensure sufficient brightness and energy saving at the same time, and this system can adapt to the characteristics of the utilization rate of the parking lot changing with time, effectively process input features in different situations, and achieve intelligent lighting control, so as to solve the following problems: 1. The lighting system of the parking lot cannot be flexibly adjusted according to the light intensity, and it is impossible to ensure sufficient brightness and energy conservation at the same time. 2. The complexity of setting the threshold of the intelligent lighting system is high, and it is difficult to adapt to the changes in the outdoor light intensity during the day. 3. When the intelligent lighting system processes input features, it attaches great importance to the primary and secondary relationships of various elements, which is completely different from the utilization rate in different situations of the parking lot, and there are no effective input features.

[0006] The AI-based adaptive garage lighting method of the present invention is achieved by the following specific technical means: An AI-based adaptive garage lighting method, characterized in that it includes the following steps: (1) Data accumulation stage: A1. Monitoring camera data collection: Collect data on vehicles entering and leaving the garage through a monitoring camera, including videos and license plate numbers of the entering and leaving vehicles; A2. Weather sensing device data collection: Collect weather data outside the garage through a weather sensing device, including light intensity, visibility, temperature, and humidity; A3. Brightness sensing device data collection: Collect brightness data inside the garage through a brightness sensing device; A4. Garage access control system data collection: Collect the number of people entering and leaving the garage through the garage access control system, including the number of people and time; A5. Vehicle flow sensing device data collection: Collect vehicle flow data entering and leaving the garage through a vehicle flow sensing device, including the number of vehicles and time; Collect the above various data to form a data sample as the basis for subsequent artificial intelligence analysis; (2) Deep learning stage: B1. Algorithm design for data analysis: Analyze the collected data samples, and through algorithm design, use the method of machine learning to analyze the relationships between various types of data and lighting parameters and garage energy consumption; B2. Lighting strategy design: Based on the analysis results of B1, design the best lighting strategy as the artificial intelligence lighting solution; (3) Strategy Update Phase: C1. Strategy Adjustment and Update: As scene data such as weather, vehicle and pedestrian flow change, the artificial intelligence strategy needs to be adjusted accordingly; C2. New Data Accumulation: As new data accumulates continuously, it further provides richer and more practical data samples for the learning of artificial intelligence, forming an ever-updating and optimizing lighting strategy. In at least some embodiments, the data accumulation phase is based on existing camera data, garage access control data, vehicle flow sensing data, lighting brightness data, and outdoor weather data of the garage. These data include: 1 - Videos and license plate numbers of vehicles entering and leaving. Collect vehicles entering and leaving the garage in the past month, record the license plate numbers, and associate them with video information; 2 - Outdoor weather sensing device data, such as light intensity, visibility, temperature, and humidity, etc. Collect relevant data in the past month; 3 - Indoor brightness sensing device data of the garage. Record the brightness of the garage and collect relevant data in the past month; 4 - Garage access control system data. Record the number of people entering and leaving the garage and collect data on entering and leaving the garage in the past month, including time periods and the number of people; 5 - Vehicle flow sensing device data. Record the vehicle flow entering and leaving the garage and collect data on entering and leaving the garage in the past month, including time periods and the number of vehicles. In at least some embodiments, in step (2), by learning and analyzing these data through algorithms, a lighting control strategy is found, including the following steps: 21. Analyze the data and establish a data set: First, analyze the data set, analyze the relationships between the data, and establish a lighting control strategy; 22. Extract features from the data using machine learning algorithms: According to different categories, extract appropriate features from the data set and extract representative data; 23. Establish a model: Use the regression model of machine learning to establish different models according to different categories and features; 24. Obtain the optimal strategy using a clustering algorithm: Use a clustering algorithm for the outputs of different models for classification to obtain different types of lighting strategies; 25. Artificial intelligence learning and simulation: Use simulation to imitate different lighting strategies and observe their effects and energy consumption; In at least some embodiments, in step (3), as the garage environment and access patterns change, the strategy also needs to be continuously adjusted. The specific process is as follows: 31. Strategy adjustment: When environmental factors such as weather and the number of people change, the artificial intelligence strategy needs to be adjusted. By adjusting parameters, better effects can be obtained; 32. New data accumulation: As new data accumulates continuously, it further provides richer and more practical data samples for the learning of artificial intelligence, forming an ever-updating and optimizing lighting strategy; 33. Through data accumulation, deep learning, and strategy update, intelligently formulate a more reasonable lighting strategy.

[0007] In at least some embodiments, the steps for the machine learning algorithm to process and analyze data samples are as follows: 1 - Construct data samples: Preprocess the collected data, including operations such as cleaning and integration, to facilitate subsequent analysis; 2 - Feature extraction: Extract appropriate features according to the type and characteristics of the data samples; 3 - Model establishment: Select an appropriate machine learning algorithm to establish a regression model; 4 - Model training and adjustment: Train the model and continuously adjust the parameters and structure of the model to obtain better performance; 5 - Model evaluation and selection: Evaluate the obtained model and select an appropriate model; In at least some embodiments, the artificial intelligence strategy obtains the optimal lighting control strategy according to the results of algorithm analysis in the deep learning stage, including the following steps: 1 - Data simulation: Use the method of mathematical simulation to simulate the analyzed data and observe the energy consumption and effects; 2 - Strategy adjustment: Fine-tune the control strategy according to the simulation results and repeatedly adjust until satisfactory results are obtained; 3 - Strategy implementation: Apply the optimal strategy to the actual garage system.

[0008] Compared with the prior art, the AI-based adaptive garage lighting method of the present invention has the following beneficial effects: (1) The advantages of this invention are as follows: By adopting algorithms such as data accumulation, deep learning, and strategy update, it fully explores the patterns in garage lighting data, formulates intelligent and dynamic lighting strategies, making the lighting more energy-efficient, reasonable, and meeting actual needs; (2) Compared with existing lighting control methods, this invention can process a large amount of complex data sets, analyze the relationships between various types of data (such as weather, vehicle in and out, pedestrian flow, etc.) and lighting parameters, and determine the optimal lighting strategy through deep learning algorithms. It can not only greatly save lighting energy consumption but also adapt to more diverse lighting needs, including different lighting conditions, vehicle in and out volumes, etc.; (3) Through algorithms such as data accumulation, deep learning, and strategy update, this invention can continuously update the lighting strategy to meet different lighting needs, including different time periods, weather conditions, and vehicle in and out volumes. This enables the lighting system to more intelligently adapt to different environments, improve lighting efficiency, reduce lighting energy consumption, and meet different customer needs; (4) Compared with traditional lighting control methods, this invention can better simulate complex garage environments, including pedestrian flow, vehicle in and out volume, weather conditions, etc., and thus better formulate the most suitable lighting strategy. Traditional lighting control methods may not be able to consider these factors, while this invention can comprehensively consider these factors to formulate a more intelligent and efficient lighting strategy; (5) Another advantage of this invention is that it can flexibly adjust the lighting strategy to adapt to different lighting needs. As time and environmental conditions change, the lighting strategy can be adjusted accordingly to ensure that the lighting system always operates optimally; (6) Another advantage of this invention is that it adopts machine learning technology, which can discover and process complex lighting data and extract useful features from it to better understand the lighting data; (7) Another advantage of this invention is that it can use data-driven intelligent methods to optimize and improve the lighting strategy instead of being based on fixed rules. This enables the lighting strategy to be continuously updated and improved to adapt to different lighting needs and environmental conditions; (8) Another advantage of this invention is that it can achieve automated operation without manual intervention. This automated operation makes the control system more efficient and accurate, without the need for staff to manually adjust the lighting strategy; (9) Another advantage of this invention is that it can control lighting based on multiple factors (such as weather conditions, vehicle in and out volume, etc.), making the lighting more in line with actual needs and environmental conditions. This ensures the practicality of lighting, reduces unnecessary or excessive lighting, and saves energy; (10) By introducing machine learning and intelligent algorithms, this invention can formulate more intelligent and adaptable lighting strategies, which is of great significance for saving energy and improving lighting efficiency; (11) By providing the best lighting strategy according to actual needs, this invention can more intelligently and effectively optimize energy and reduce energy waste.In summary, in view of the problem of excessive energy loss rate in the existing garage lighting technology, the present invention proposes a new optimization scheme for the existing garage lighting system. By introducing artificial intelligence technology, the decision-making ability for lighting strategies is greatly improved, which plays an important role in reducing costs, saving energy, and reducing maintenance fees. Detailed implementation manners

[0009] The following further describes in detail the implementation manners of the present invention.

[0010] Embodiment: An AI-based adaptive garage lighting method includes the following steps: (1) Data accumulation stage: The data accumulation stage refers to using the monitors or cameras in the garage to record video data under different time periods, different weather conditions, with or without incoming and outgoing vehicles and personnel, vehicle incoming and outgoing directions / status, etc., and storing and organizing the data feature formats into files or databases; The recording method can be video frame screenshots, video parsing or similar technologies.

[0011] Traffic flow refers to the number of vehicles and personnel entering and leaving the garage at different time periods.

[0012] Weather refers to mostly cloudy or mostly sunny. It is extracted from weather forecast data.

[0013] Vehicle incoming and outgoing directions / status refer to entering or leaving the garage, parking, or reversing into the garage, etc.

[0014] A1. Data collection by monitoring cameras: Collect data of vehicles entering and leaving the garage through monitoring cameras, including videos and license plate numbers of incoming and outgoing vehicles; A2. Data collection by weather sensing devices: Collect weather data outside the garage through weather sensing devices, including light intensity, visibility, temperature, and humidity; A3. Data collection by brightness sensing devices: Collect brightness data inside the garage through brightness sensing devices; A4. Data collection by garage access control systems: Collect the number of people and time of the flow of people entering and leaving the garage through garage access control systems; A5. Data collection by traffic flow sensing devices: Collect the number of vehicles and time of the traffic flow entering and leaving the garage through traffic flow sensing devices; Collect the above various data to form data samples as the basis for subsequent artificial intelligence analysis; (2) Deep learning stage: The deep learning stage refers to using machine learning algorithms and image analysis algorithms to analyze and train the video data recorded in the data accumulation stage under different time periods, different weather conditions, with or without incoming and outgoing vehicles and personnel, vehicle incoming and outgoing directions / status conditions, so that the system can judge and select the optimal lighting parameters according to the recorded conditions. Machine learning algorithms include logistic regression, K-nearest neighbor, random forest, neural network, etc.

[0015] B1. Algorithm design for data analysis: Analyze the collected data samples. Through algorithm design and using machine learning methods, analyze the relationships between various types of data, lighting parameters, and garage energy consumption; B2. Lighting strategy design: Based on the analysis results of B1, design the optimal lighting strategy as the artificial intelligence lighting solution; (3) Strategy Update Phase: The strategy update phase refers to dynamically adjusting the optimal lighting scheme by combining actual lighting requirements, manual setting / adjustment of lighting, etc. The lighting requirements can be light intensity requirements or lighting brightness levels. The strategy update phase mainly includes methods such as setting manual light intensity and updating lighting brightness levels. C1. Strategy Adjustment and Update: As scene data such as weather, vehicle and pedestrian flow change, the artificial intelligence strategy needs to be adjusted accordingly; C2. New Data Accumulation: As new data accumulates continuously, it further provides richer and more practical data samples for the learning of artificial intelligence, forming an ever-updating and ever-optimizing lighting strategy. Specifically, the data accumulation phase is based on existing camera data, garage access control data, vehicle flow sensing data, lighting brightness data, and outdoor garage weather data. These data include: 1 - Videos and license plate numbers of vehicles entering and leaving. Collect vehicles entering and leaving the garage in the past month, record the license plate numbers, and associate them with video information; 2 - Outdoor garage weather sensing device data, such as light intensity, visibility, temperature, and humidity, etc. Collect relevant data in the past month; 3 - Indoor garage brightness sensing device data, record the brightness of the garage, and collect relevant data in the past month; 4 - Garage access control system data, record the number of people entering and leaving the garage, and collect data on entering and leaving the garage in the past month, including time periods and numbers; 5 - Vehicle flow sensing device data, record the vehicle flow entering and leaving the garage, and collect data on entering and leaving the garage in the past month, including time periods and the number of vehicles; Specifically, in step (2), by learning and analyzing these data through algorithms, a lighting control strategy is found, including the following steps: 21. Analyze the data and establish a data set: First, analyze the data set, analyze the relationships between the data, and establish a lighting control strategy; 22. Extract features from the data using machine learning algorithms: According to different categories, extract appropriate features from the data set and extract representative data; 23. Establish a model: Use the regression model of machine learning to establish different models according to different categories and features; 24. Obtain the best strategy using a clustering algorithm: Use a clustering algorithm for the outputs of different models for classification to obtain different types of lighting strategies; 25. Artificial intelligence learning and simulation: Use simulation to imitate different lighting strategies and observe their effects and energy consumption; Specifically, in step (3), as the garage environment and access patterns change, the strategy also needs to be continuously adjusted. The specific process is as follows: 31. Strategy adjustment: When environmental factors such as weather and the number of people change, the artificial intelligence strategy needs to be adjusted. By adjusting parameters, better effects can be obtained; 32. New data accumulation: As new data accumulates continuously, it further provides richer and more practical data samples for the learning of artificial intelligence, forming an ever-updating and ever-optimizing lighting strategy; 33. Through data accumulation, deep learning, and strategy update, intelligently formulate a more reasonable lighting strategy.

[0016] Specifically, the steps for the machine learning algorithm to process and analyze data samples are as follows: 1 - Construct data samples: Preprocess the collected data, including operations such as cleaning and integration, to facilitate subsequent analysis; 2 - Feature extraction: Extract appropriate features according to the type and characteristics of the data samples; 3 - Model establishment: Select an appropriate machine learning algorithm to establish a regression model; 4 - Model training and adjustment: Train the model and continuously adjust the parameters and structure of the model to obtain better performance; 5 - Model evaluation and selection: Evaluate the obtained model and select an appropriate model. Specifically, the artificial intelligence strategy obtains the optimal lighting control strategy based on the results of algorithm analysis in the deep learning stage, including the following steps: 1 - Data simulation: Use the method of mathematical simulation to simulate the analyzed data and observe energy consumption and effects; 2 - Strategy adjustment: Fine-tune the control strategy according to the simulation results and repeatedly adjust until satisfactory results are obtained; 3 - Strategy implementation: Apply the optimal strategy to the actual garage system.

[0017] Furthermore: In the data accumulation stage, traffic flow data is calculated based on the statistical results of the number of vehicle entries and exits.

[0018] In the data accumulation stage, light intensity is calculated based on the detection results of light sensors.

[0019] In the data accumulation stage, weather data is calculated based on local historical average weather data and light sensor data.

[0020] In the data accumulation stage, vehicle entry / exit direction / status includes vehicle entry / exit, parking, reverse parking into the garage, turning around, etc.

[0021] In the deep learning stage, the neural network uses ReLU as the activation function. The number of classifications in each feature space is N, and a K (K ≤ N)-layer network is adopted. The first K - 1 layers use the relu activation function, and the last layer uses the sigmoid function for normalization. The initial parameters of the network are obtained using the ReLU activation function, and the parameters are optimized using stochastic gradient descent and dropout.

[0022] The specific network structure adopted includes: ① Input layer: 6 - dimensional; ② Hidden layer: (the first layer) 6 * 4; the second layer: 4 * 8; the third layer: 8 * 16; the fourth layer: 16 * 32; the fifth layer: 32 * 64; the sixth layer: 64 * 128; the seventh layer: 128 * 256; the eighth layer: 256 * 512; the ninth layer: 512 * 512; ③ Output layer (feature classification layer): N * (N + 1) - dimensional.

[0023] In the deep learning stage, the cross-entropy loss function is adopted.

[0024] In the deep learning stage, the cross-validation method is used for comparison.

[0025] The deep learning stage outputs two parameters: whether to turn on a certain layer of lighting parameters and the brightness level of the corresponding lighting layer.

[0026] In the policy update stage, the artificial light intensity is achieved by methods such as artificially setting / adjusting lighting, etc.

[0027] In the policy update stage, the method for updating the lighting brightness level is as follows: a. According to the currently recorded traffic flow data, determine whether to update the lighting brightness level: ① When the traffic flow is less than a certain threshold, the lighting brightness level remains unchanged; ② When the traffic flow meets a certain threshold, increase the lighting brightness level.

[0028] b. According to the current light intensity, determine whether to change the lighting brightness level: ① When [condition not specified], the lighting brightness level remains unchanged; ② When the light intensity is less than the minimum of the preset brightness level, reduce the lighting brightness level; ③ When the light intensity is greater than the maximum of the preset brightness level, increase the lighting brightness level; ④ When the light intensity is between the minimum and maximum of the preset brightness level, compare with the previous set of data. When [condition not specified], increase the lighting brightness level.

[0029] In the data accumulation stage, during the data accumulation process, the update period of each group of data can be adjusted.

[0030] In the data accumulation stage, during the data accumulation process, the amount of data accumulation can be adjusted.

[0031] In the data accumulation stage, the algorithms for video parsing and traffic flow statistics can be adjusted.

[0032] In the deep learning stage, the number of layers of the neural network can be adjusted.

[0033] In the deep learning stage, the activation function of the neural network can be adjusted.

[0034] In the deep learning stage, the parameter initialization algorithm and the optimization algorithm of the neural network can be adjusted.

[0035] In the deep learning stage, the optimized loss function can be adjusted.

[0036] In the policy update stage, the time for artificially setting / adjusting lighting can be adjusted.

[0037] During the policy update phase, the method for updating the lighting brightness level is adjustable.

[0038] During the policy update phase, the time for changing the lighting parameters is adjustable.

[0039] During the policy update phase, the principle for determining whether to update the lighting brightness level is adjustable.

[0040] During the policy update phase, the principle for determining whether to change the lighting parameters is adjustable.

[0041] During the data accumulation phase, during the data accumulation process, the correctness of the lighting state can be manually intervened.

[0042] During the deep learning phase, when manual intervention is required during the training and update processes, it can be adjusted.

[0043] During the policy update phase, when manual intervention is required during the update process, it can be adjusted.

[0044] The following are application examples of the present invention: The present invention has been actually applied in the parking lot of a commercial office building. Vehicles and personnel entering and leaving are extracted from the surveillance videos at the entrance and exit of the parking garage, and the weather status is judged according to the weather forecast database. Specifically: Data accumulation phase: Continuously accumulate historical data for a period of time, including time, weather, whether there are vehicles entering or leaving the garage, and the traffic flow during that period.

[0045] Deep learning phase: Analyze the accumulated historical data to fit the relationship curves between time, weather and vehicle entry and exit, and between traffic flow and vehicle entry and exit.

[0046] Policy update phase: Dynamically set the lighting policy.

[0047] Policy update phase: Deploy light sensors at the entrance and exit of the parking garage. If the traffic flow is greater than or equal to 5 and the light sensing intensity is lower than a certain threshold, turn on the lighting; if the traffic flow is less than or equal to 2, or the light sensing intensity exceeds a certain threshold, turn off the lighting.

[0048] Results of the deep learning phase are as follows: When the lighting brightness level is 2, when the number of vehicles entering and leaving the garage is less than 1 (i.e., the traffic flow is 0), turn on the lighting; when the light sensing is insufficient, turn on the lighting; When the lighting brightness level is 1, when the number of vehicles entering and leaving the garage is less than 2 (i.e., the traffic flow is 0), do not turn on the lighting; when the light sensing is insufficient, do not turn on the lighting; when the light sensing intensity is close to 1, do not turn on the lighting; When the lighting brightness level is 0, when the number of vehicles entering and leaving the garage is less than 2 (i.e., the traffic flow is 0), do not turn on the lighting; when the light sensing is insufficient, do not turn on the lighting; when the light sensing intensity is close to 1, do not turn on the lighting.

[0049] The present invention can make full use of historical data under conditions such as different time periods, different weather conditions, no vehicles or people entering or leaving, etc., making the lighting more in line with actual needs and optimizing the lighting strategy. This strategy is applicable to parking facilities with relatively fixed vehicle and pedestrian flows, such as shopping malls and office buildings.

[0050] In the present invention: (1) Algorithms such as data accumulation, deep learning, and strategy update are adopted to fully explore the rules in garage lighting data, and an intelligent and dynamic lighting strategy is formulated to make the lighting more energy-saving, reasonable, and in line with actual needs; (2) It can process a large number of complex data sets, analyze the relationships between various types of data and lighting parameters, and determine the optimal lighting strategy through deep learning algorithms, greatly saving lighting energy consumption and adapting to richer lighting needs; (3) Continuously update the lighting strategy to meet different lighting needs, including different time periods, weather conditions, and vehicle entry and exit volumes, improving lighting efficiency; (4) Better simulate complex garage environments, including pedestrian flow, vehicle entry and exit volume, weather conditions, etc., and formulate the most suitable lighting strategy; (5) Flexibly adjust the lighting strategy to adapt to different lighting needs, ensuring that the lighting system can always achieve the optimal effect; (6) Adopt machine learning technology to discover and process complex lighting data, and extract useful features from it to better understand lighting data; (7) Use data-driven intelligent methods to optimize and improve the lighting strategy to adapt to different lighting needs and environmental conditions; (8) Achieve automated operation without manual intervention, making the control system more efficient and accurate; (9) Control lighting according to multiple factors (such as weather conditions, vehicle entry and exit volume, etc.), making the lighting more in line with actual needs and environmental conditions, and saving energy; (10) Formulate a more intelligent and adaptable lighting strategy, saving energy and improving lighting efficiency; (11) Provide a more intelligent and effective energy optimization solution to reduce energy waste.

Claims

1. An AI-based adaptive garage lighting method, characterized by: The steps include: (1) Data accumulation stage: A1. Surveillance camera data collection: The surveillance camera collects data on vehicles entering and exiting the garage, including videos and license plate numbers of vehicles entering and exiting the garage. A2. Weather sensor data collection: The weather sensor collects weather data outside the garage, including light intensity, visibility, temperature and humidity. A3. Brightness sensor data collection: The brightness sensor collects brightness data inside the garage. A4. Garage access control system data collection: The garage access control system collects traffic data on people entering and exiting the garage, including the number of people and time. A5. Data collection of vehicle flow sensor equipment: The vehicle flow sensor equipment collects vehicle flow data in and out of the garage, including the number of vehicles and time; Collect all the above data to form data samples as the basis for subsequent artificial intelligence analysis; (2) Deep learning stage: B1. Data analysis algorithm design: Analyze the collected data samples, and use machine learning methods to analyze the relationship between various data and lighting parameters and garage energy consumption through algorithm design; B2. Lighting strategy design: Based on the analysis results of B1, design the best lighting strategy as the solution for AI lighting; (3) Strategy update stage: C1. Strategy adjustment and update: As scene data such as weather and traffic flow change, the AI ​​strategy needs to be adjusted accordingly; C2. Accumulation of new data: With the continuous accumulation of new data, it further provides richer and more realistic data samples for artificial intelligence learning, forming a constantly updated and optimized lighting strategy.

2. The AI-based adaptive garage lighting method according to claim 1, characterized in that: The data accumulation stage is based on the existing camera data, garage access control data, vehicle flow sensor data, lighting brightness data and weather data outside the garage. These data include: 1-Videos and license plate numbers of vehicles entering and exiting the garage, collecting vehicles entering and exiting the garage in the past month, recording license plate numbers, and associating them with video information; 2-Weather sensor equipment data outside the garage, such as light intensity, visibility, temperature and humidity, etc., collecting relevant data for the past month; 3-Brightness sensor equipment data inside the garage, recording the brightness of the garage, and collecting relevant data for the past month; 4-Garage access control system data, recording the flow of people entering and exiting the garage, collecting data on entering and exiting the garage in the past month, including time periods and number of people; 5-Traffic flow sensor equipment data, recording the flow of vehicles entering and exiting the garage, collecting data on entering and exiting the garage in the past month, including time periods and number of vehicles.

3. The AI-based adaptive garage lighting method according to claim 1, characterized in that: In the step (2), the lighting control strategy is found by learning and analyzing the data through an algorithm, including the following steps:

21. Analyze the data and establish a data set: first analyze the data set, analyze the relationship between the data, and establish a lighting control strategy; 22. Use a machine learning algorithm to extract features from the data: according to different categories, extract appropriate features from the data set and extract representative data; 23. Establish a model: Use a machine learning regression model to establish different models according to different categories and features; 24. Use a clustering algorithm to obtain the best strategy: Use a clustering algorithm to classify the outputs of different models to obtain different types of lighting strategies; 25. Artificial intelligence learning and simulation: Use simulation methods to imitate different lighting strategies and observe their effects and energy consumption.

4. The AI-based adaptive garage lighting method according to claim 1, characterized in that: In step (3), as the garage environment and entry and exit patterns change, the strategy also needs to be constantly adjusted. The specific process is as follows:

31. Strategy adjustment: When environmental factors such as weather and traffic flow change, the artificial intelligence strategy needs to be adjusted to achieve better results by adjusting parameters; 32. New data accumulation: With the continuous accumulation of new data, it further provides richer and more realistic data samples for artificial intelligence learning, forming a continuously updated and optimized lighting strategy; 33. Through data accumulation, deep learning and strategy updates, more reasonable lighting strategies can be formulated intelligently.

5. The AI-based adaptive garage lighting method according to claim 3, characterized in that: The steps of using the machine learning algorithm to process and analyze data samples are as follows: 1-Constructing data samples: preprocessing the collected data, including cleaning, integration and other operations, to facilitate subsequent analysis; 2-Feature extraction: extracting appropriate features based on the type and characteristics of the data sample; 3-Model building: selecting a suitable machine learning algorithm to establish a regression model; 4-Model training and adjustment: training the model, and continuously adjusting the model's parameters and structure to obtain better performance; 5-Model evaluation and selection: evaluating the obtained model and selecting a suitable model.

6. The AI-based adaptive garage lighting method according to claim 1 or 4, characterized in that: The artificial intelligence strategy obtains the best lighting control strategy based on the results of the algorithm analysis in the deep learning stage, including the following steps: 1-Data simulation: Use mathematical simulation methods to simulate the analyzed data and observe energy consumption and effects; 2-Strategy adjustment: According to the simulation results, fine-tune the control strategy and adjust it repeatedly until a satisfactory result is obtained; 3-Strategy implementation: Apply the best strategy to the actual garage system.