Intelligent Lighting Control Method in Harsh Environments Based on Multi-Sensor Fusion

Abstract

The present invention relates to the field of electric lighting control, and discloses an intelligent lighting control method based on multi-sensor fusion for improving the performance of intelligent lighting systems in harsh environments. It can be applied to complex and harsh working environments such as underground mining. By arranging a multi-modal sensor network, the present invention can monitor and fuse various environmental parameters in real time, such as light intensity, temperature, pressure, vibration, and corrosive substances. Using these data, the actual state of the current environment can be evaluated more accurately, and future change trends can be predicted. By analyzing the importance of sensor data through an entropy weight model and combining it with fuzzy logic control, the system can output optimized lighting control results. In addition, through kernel density estimation and the GBM gradient boosting machine model, the distribution of future lighting requirements in harsh environments can be predicted and the lighting requirement values in harsh environments can be quantified, so as to formulate a more accurate intelligent lighting control strategy for harsh environments.

Description

Technical Field

[0001] The present invention relates to the field of electric lighting control, and particularly to an intelligent lighting control method based on multi-sensor fusion for harsh environments. Background Art

[0002] With the rapid development of technology and the continuous improvement of people's requirements for the quality of life, intelligent lighting systems play an increasingly important role in modern buildings and home environments. Especially in harsh environments, such as extreme weather, complex working scenarios, or emergency situations, the stability and intelligence level of intelligent lighting systems are particularly important. These systems not only need to be able to automatically adapt to changes in the external environment, but also need to meet the personalized lighting needs of users while ensuring the efficient utilization of energy.

[0003] However, existing intelligent lighting control technologies still have many deficiencies when facing harsh environments. First of all, most systems are relatively simple in data fusion processing and fail to fully utilize the information redundancy and complementarity of multi-modal sensor networks, resulting in insufficient accuracy and robustness of environmental perception. Secondly, existing technologies often ignore the importance and weight allocation of sensor data when formulating lighting control strategies and cannot dynamically adjust according to the reliability and effectiveness of different sensors in harsh environments. In addition, existing systems do not consider the personalized lighting needs of users sufficiently, lacking flexibility and self-adaptability, and it is difficult to meet the diverse lighting needs of users.

[0004] To solve the above problems, this research proposes an intelligent lighting control method based on multi-sensor fusion for harsh environments. It can significantly improve the performance of intelligent lighting systems in harsh environments and provide a more comfortable, energy-saving, and safe lighting environment for users. Summary of the Invention

[0005] The present invention provides an intelligent lighting control method based on multi-sensor fusion for harsh environments, which is used to improve the performance of intelligent lighting systems in harsh environments and provide a more comfortable, energy-saving, and safe lighting environment for users.

[0006] The first aspect of the present invention provides an intelligent lighting control method for harsh environments based on multi-sensor fusion. The intelligent lighting control method for harsh environments based on multi-sensor fusion includes: collecting and inputting control information on harsh environments based on a multi-modal sensor network, obtaining and fusing the input control information on harsh environments in the data of the multi-modal sensor network, including monitoring data on light intensity, color, images, vibrations, temperature, pressure, and corrosive substances, to obtain the fused state information on harsh environments; using a preset entropy weight model, combining the fused state information on harsh environments, analyzing the data importance of different sensors in the multi-modal sensor network, and performing weight allocation according to their reliability and effectiveness in harsh environments to obtain the weights and readings of each sensor's data; adopting fuzzy logic control, using the readings of each sensor's data as input variables, and adjusting the membership function of the input using the weights of each sensor's data to obtain the output result of the harsh environment lighting control system; extracting lighting demand-related features based on the fused state information on harsh environments, including: based on the data of the light sensor, combining factors such as time, the contrast of light inside and outside the working environment, and special lighting conditions in harsh environments, where the special lighting conditions in harsh environments include the dimness and uneven light conditions in underground mining environments, predicting the most suitable light intensity through a machine learning model to obtain the first lighting demand-related feature; using the data of the color sensor and the camera, analyzing the environmental color, human activities, and special requirements for color and color temperature in harsh environments, where the special requirements for color and color temperature in harsh environments include the demand for the anti-corrosion performance of equipment in corrosive environments, and using pattern recognition technology to determine the color and color temperature combination to obtain the second lighting demand-related feature; combining the human movement detection data, historical behavior patterns, and the impact of harsh environments on human movement and lighting demand, where the impact of harsh environments on human movement and lighting demand includes the limited personnel flow in high-temperature and high-pressure environments, and using time series analysis or machine learning algorithms to predict the change trend of lighting demand in the next period of time, including the personnel flow path and work area conversion, to obtain the third lighting demand-related feature; annotating the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature, marking the changes in lighting demand under special events, as well as the user preference pattern; adopting data enhancement techniques, including synthetic data generation and time series extension, to enrich the data set and improve the generalization ability of the model; constructing a historical data set of harsh environments based on the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature; performing probability density function estimation on the historical data set of harsh environments according to the preset kernel density estimation, and obtaining the distribution estimation of future lighting demand in harsh environments by selecting the kernel function and bandwidth parameters, and obtaining the quantified lighting demand value in harsh environments by combining the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature with the preset GBM gradient boosting machine model;An intelligent lighting control strategy for harsh environments is obtained based on the output results of the harsh environment lighting control system, the distribution estimation of future harsh environment lighting requirements, and the quantification of harsh environment lighting requirement values.

[0007] Optionally, in the first implementation manner of the first aspect of the present invention, the multi-modal sensor network includes a light sensor, a temperature sensor, a pressure sensor, a vibration sensor, and a corrosion monitoring sensor. Using the heat and electrical signals generated by each sensor itself as additional resources, an additional resource dataset is obtained. Each sensor collects environmental data according to a preset sampling frequency and packs the data. The sink node in the sensor network collects data from each sensor to form a multi-modal input control environment sensor dataset; a preset dynamic sensor activation strategy is designed to dynamically select and activate the most relevant sensors according to the current environmental conditions and lighting requirements. When the daylight is sufficient during the day, the sampling frequency of the light intensity sensor is reduced or turned off, while when it is night or the light is dim, the sampling frequency is increased; the sensor data is fused on different time scales, and the time scales include seconds, minutes, and hours to capture the environmental change characteristics on different time scales and obtain the harsh environment fusion state information.

[0008] Optionally, in the second implementation manner of the first aspect of the present invention, the weight assignment of the entropy weight model is combined with an online learning algorithm to update the weight assignment in real time; a multi-level fuzzy logic control is designed, where each level is responsible for processing sensor data and lighting control tasks with different granularities. The low-level controller is responsible for basic brightness adjustment, and the high-level controller is responsible for more complex scene recognition and lighting strategy adjustment; using a fuzzy inference mechanism, based on the adjusted membership function and fuzzy rule base, fuzzy inference is performed to obtain the output results of the harsh environment lighting control system.

[0009] Optionally, in the fourth implementation manner of the first aspect of the present invention, a multi-dimensional kernel density estimation method is adopted, including a multi-dimensional kernel density estimation based on Copula, to more accurately capture the joint distribution between features and obtain their respective probability density functions; using the probability density functions of the harsh environment historical dataset, the sliding window method is used to estimate the distribution of future harsh environment lighting requirements: where, is the conditional probability of the future harsh environment lighting requirement x t+1 is the conditional probability, is the indicator function, N is the number of sliding windows in the harsh environment historical dataset, is the lighting requirement sequence in the harsh environment historical dataset; the distribution estimation of the future harsh environment lighting requirement is obtained.

[0010] Optionally, in the fifth implementation manner of the first aspect of the present invention, the output result of the harsh environment lighting control system is evaluated. By comparing the actual output of the lighting system with the preset lighting standard or user expectations, it is determined whether the current lighting system meets the lighting requirements, and whether there are problems of over-brightness, over-darkness, or uneven light distribution, and the harsh environment lighting demand value is quantified; an environment-demand-response dataset is formed; the environment-demand-response dataset is deeply analyzed to identify the change trend, outliers, and periodic patterns of the lighting demand, and an intelligent lighting control strategy for the harsh environment is obtained.

[0011] Optionally, in the sixth implementation manner of the first aspect of the present invention, it further includes: setting a benchmark lighting control strategy according to the harsh environment historical dataset and user preferences, including default light intensity, color temperature, and dynamic adjustment rules; formulating a dynamic adjustment strategy based on the distribution estimation of future harsh environment lighting demands. When it is predicted that the future harsh environment lighting demand will increase, the light intensity or color temperature is adjusted in advance to meet user needs; the feedback data of the lighting control system is collected in real time, including actual light intensity, color temperature, and energy consumption, and compared with the expected values, and adjustments are made in a timely manner when deviations are found; the control strategy is evaluated and optimized regularly, and according to user feedback, environmental changes, and technological progress, the control strategy is continuously adjusted and improved to improve the intelligent level of lighting control and user satisfaction.

[0012] The second aspect of the present invention provides an intelligent lighting control device for harsh environments based on multi-sensor fusion. The intelligent lighting control device for harsh environments based on multi-sensor fusion includes: an acquisition module, configured to collect and input control harsh environment information based on a multi-modal sensor network, acquire and fuse the input control harsh environment information in the multi-modal sensor network data, including light intensity, temperature, pressure, vibration, and corrosive substance monitoring data, to obtain harsh environment fusion state information; a processing module, configured to use a preset entropy weight model, combine the harsh environment fusion state information, analyze the data importance of different sensors in the multi-modal sensor network, and perform weight allocation according to their reliability and effectiveness in harsh environments to obtain the weight and reading of each sensor data; adopt fuzzy logic control, use the reading of each sensor data as an input variable, and adjust the membership function of the input using the weight of each sensor data to obtain the output result of the harsh environment lighting control system; an extraction module, configured to extract first lighting requirement-related features, second lighting requirement-related features, and third lighting requirement-related features based on the harsh environment fusion state information, and construct a harsh environment historical data set based on the first lighting requirement-related features, second lighting requirement-related features, and third lighting requirement-related features; a quantization module, configured to perform probability density function estimation on the harsh environment historical data set based on a preset kernel density estimation, obtain a distribution estimation of future harsh environment lighting requirements by selecting a kernel function and a bandwidth parameter, and obtain a quantized harsh environment lighting requirement value by combining the first lighting requirement-related features, second lighting requirement-related features, and third lighting requirement-related features with a preset GBM gradient boosting machine model; an allocation module, configured to obtain a harsh environment intelligent lighting control strategy based on the output result of the harsh environment lighting control system, the distribution estimation of future harsh environment lighting requirements, and the quantized harsh environment lighting requirement value.

[0013] The third aspect of the present invention provides an intelligent lighting control device for harsh environments based on multi-sensor fusion, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor invokes the instructions in the memory to cause the intelligent lighting control device for harsh environments based on multi-sensor fusion to execute the above-mentioned intelligent lighting control method for harsh environments based on multi-sensor fusion.

[0014] The fourth aspect of the present invention provides a computer-readable storage medium, in which instructions are stored, and when it runs on a computer, it causes the computer to execute the above-mentioned intelligent lighting control method for harsh environments based on multi-sensor fusion.

[0015] In the technical solution provided by the present invention, environmental information is collected through a multi-modal sensor network, and a Kalman filter is used for data fusion to obtain the integrated state information of the harsh environment. Then, the importance of sensor data is analyzed in combination with the entropy weight model to provide a basis for adjusting fuzzy logic control. By extracting features related to lighting requirements and constructing a historical dataset of the harsh environment, the distribution of future lighting requirements in the harsh environment is predicted using kernel density estimation, and the lighting requirement value in the harsh environment is quantified using the GBM gradient boosting machine model. Finally, an intelligent lighting control strategy is formulated based on the above information to achieve intelligent and precise control of the lighting system.

[0016] Through multi-sensor fusion technology, the present invention can more comprehensively perceive environmental changes, adjust lighting parameters in real time, and meet lighting requirements in different scenarios.

[0017] This technology particularly considers the impact of harsh environment factors. By formulating an emergency lighting control strategy, it ensures the reliable operation of the lighting system under harsh conditions.

[0018] By accurately predicting and adjusting lighting requirements, the present invention avoids unnecessary lighting waste, effectively reduces energy consumption, and realizes a green and environmentally friendly lighting solution.

[0019] In summary, the present invention combines a variety of advanced algorithms and models, can self-learn and self-optimize, continuously adapt to environmental changes and user needs changes, and improve the intelligent level of the lighting system. Brief Description of the Drawings

[0020] Figure 1 It is a schematic diagram of an embodiment of the intelligent lighting control method for harsh environments based on multi-sensor fusion in an embodiment of the present invention;

[0021] Figure 2 It is a schematic diagram of another embodiment of the intelligent lighting control method for harsh environments based on multi-sensor fusion in an embodiment of the present invention;

[0022] Figure 3 It is a schematic diagram of an embodiment of the intelligent lighting control device for harsh environments based on multi-sensor fusion in an embodiment of the present invention;

[0023] Figure 4 It is a schematic diagram of an embodiment of the intelligent lighting control equipment for harsh environments based on multi-sensor fusion in an embodiment of the present invention. Detailed Description of the Invention

[0024] An embodiment of the present invention provides a method for intelligent lighting control in harsh environments based on multi-sensor fusion, which is used to improve the performance of the intelligent lighting system in harsh environments and provide a more comfortable, energy-saving and safe lighting environment for users. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present invention are used to distinguish similar objects and do not necessarily describe a specific order or sequence.

[0025] For ease of understanding, the specific process of the embodiment of the present invention will be described below. Please refer to Figure 1 , an embodiment of the method for intelligent lighting control in harsh environments based on multi-sensor fusion in the embodiment of the present invention includes:

[0026] 101. Collect and input control harsh environment information based on the multi-modal sensor network, obtain and fuse the input control harsh environment information in the multi-modal sensor network data, including light intensity, temperature, pressure, vibration and corrosive substance monitoring data, to obtain the harsh environment fusion state information;

[0027] It can be understood that the execution subject of the present invention can be an intelligent lighting control device based on multi-sensor fusion, or a terminal or a server. Specifically, it is not limited here. The embodiment of the present invention takes the server as the execution subject for illustration.

[0028] It should be noted that considering an underground mining environment, where lighting control is crucial, not only related to work efficiency but also directly affecting the safety of workers. Due to the harsh underground environment, there are challenges such as high temperature, high pressure, vibration and corrosion. Therefore, it is necessary to build an intelligent lighting control system that can cope with these challenges.

[0029] First, a multi-modal sensor network is arranged in key areas of the underground mining environment (such as the working face, intersection of channels, etc.). These sensors include but are not limited to light sensors, temperature sensors, pressure sensors, vibration sensors and corrosion monitoring sensors. Each sensor is optimized for specific challenges in the underground environment to ensure reliable operation under harsh conditions. Light sensor: used to monitor the ambient light intensity in real time and provide a direct input control signal for the lighting system. Temperature sensor: can withstand high temperature environments, such as using platinum-based sensors, whose working range can cover -70°C to 500°C to monitor the temperature changes underground. Pressure sensor: used to monitor the changes in underground air pressure or water pressure to ensure the safe operation of the lighting system in high-pressure environments. Vibration sensor: can detect and quantify the vibration level underground and provide stability feedback for the lighting system. Corrosion monitoring sensor: monitors the corrosive substances that may exist in the environment to protect the lighting equipment and sensor network from damage.

[0030] Data acquisition and preprocessing: Each sensor continuously collects environmental data at a set frequency (such as per second or per minute). The raw data collected is first subjected to preprocessing steps, including signal amplification, analog-to-digital conversion, noise filtering, and interference compensation, etc., to ensure the accuracy and reliability of the data.

[0031] Multi-modal input control for environmental sensor data fusion processing: The preprocessed multi-modal sensor data is transmitted to the central control unit for data fusion processing. Here, various data fusion methods can be adopted, such as the Kalman filtering method or the Bayesian estimation method, to synthesize the information of each sensor and obtain the fused state information of the harsh environment. Taking the Kalman filtering method as an example, this method can recursively determine the estimated value of the fused data, making the estimated value optimal in the statistical sense. By fusing the data of multiple sensors, the system can more accurately estimate the actual state of the current environment (such as light intensity, temperature, pressure, etc.) and predict future change trends.

[0032] Application of the fused state information of the harsh environment and lighting control: According to the fused environmental state information, the intelligent lighting control system can adjust parameters such as the brightness and color temperature of the lamps in real time to provide the best lighting effect and ensure the safety of workers. For example, when it is detected that the environmental light intensity is too low, the system will automatically increase the brightness of the lamps; when the temperature is too high, the working state of the lamps can be adjusted to reduce heat generation.

[0033] Data example: At a certain moment, the light sensor detects that the environmental light intensity is 50 lux (lx), the temperature sensor shows that the current temperature is 45°C, the pressure sensor reads 1.2 atmospheres (atm), the vibration sensor detects a slight vibration (amplitude < 0.5 mm), and the corrosion monitoring sensor does not detect corrosive substances. After these data are fused and processed, the system determines that the current environment requires an appropriate increase in lighting brightness to ensure that workers can clearly see the working area. Therefore, the system automatically increases the brightness of the lamps to 100 lux (lx) and at the same time adjusts the color temperature to a warmer tone to reduce visual fatigue.

[0034] 102. Using a preset entropy weight model, combined with the fused state information of the harsh environment, analyze the data importance of different sensors in the multi-modal sensor network, and perform weight allocation according to their reliability and effectiveness in the harsh environment to obtain the weight and reading of each sensor data. Adopt fuzzy logic control, take the reading of each sensor data as an input variable, and use the weight of each sensor data to adjust the membership function of the input to obtain the output result of the harsh environment lighting control system;

[0035] It should be noted that in the weight distribution and fuzzy logic control of the intelligent lighting control system for underground mining environments, the information fusion state of harsh environments and the entropy weight model: First, the environmental information is obtained and fused through the multi-modal sensor network described previously to obtain the information fusion state of harsh environments. This information includes multiple dimensions such as light intensity, temperature, pressure, vibration, etc.

[0036] Next, a preset entropy weight model is used to analyze the importance of these sensor data. The entropy weight method is a method for determining objective weights based on the variability of indicators. Here, we evaluate the importance based on the historical variability and current stability of the sensor data.

[0037] For example, we have the following set of sensor readings (for illustration purposes only, actual data may vary in real time according to the underground environment):

[0038] Light sensor reading: 80 lx

[0039] Temperature sensor reading: 40°C

[0040] Pressure sensor reading: 1.5 atm

[0041] Vibration sensor reading: 0.3 mm / s

[0042] Through the analysis of the entropy weight model, we can obtain the preliminary weights of each sensor data:

[0043] Weight of light sensor: 0.4

[0044] Weight of temperature sensor: 0.3

[0045] Weight of pressure sensor: 0.2

[0046] Weight of vibration sensor: 0.1

[0047] Weight distribution and adjustment: Based on the preliminary weights obtained from the entropy weight model, we can further adjust according to the real-time performance and historical performance of the sensors. For example, if a certain sensor shows a high failure rate or unstable data in a recent period, its weight may be appropriately reduced.

[0048] After adjustment, we obtain the final weights of each sensor data, and these weights will be used in the subsequent fuzzy logic control.

[0049] Fuzzy logic control implementation. In fuzzy logic control, we take the readings of each sensor as input variables. At the same time, the weights of the data of each sensor are used to adjust the membership functions of these inputs. The membership function is a function in fuzzy logic that describes the degree to which an input variable belongs to a certain fuzzy set. For the light sensor with a reading of 80 lx, we can define a fuzzy set "suitable lighting" and calculate the degree to which this reading belongs to the "suitable lighting" set according to its weight and the preset membership function. Similarly, similar calculations are performed for other sensors. Finally, through the fuzzy inference and defuzzification steps, we obtain the output result of the harsh environment lighting control system. This output result may be a specific lighting brightness value, color temperature value or other control instructions for adjusting the lighting equipment in the mine to meet the lighting needs and safety requirements in the current environment.

[0050] 103. Extract the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature based on the harsh environment fusion state information, and construct a harsh environment historical data set based on the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature;

[0051] It should be noted that for the acquisition of the harsh environment fusion state information: First, obtain and fuse the multi-dimensional information of the underground environment through the previously described multi-modal sensor network, including but not limited to light intensity, temperature, pressure, vibration, etc. This information constitutes the harsh environment fusion state information and provides a basis for subsequent extraction of lighting demand features.

[0052] Based on the harsh environment fusion state information, we can extract features closely related to lighting demand. These features reflect the actual demand of the underground environment for the lighting system and help to achieve refined lighting control.

[0053] The first lighting demand-related feature: lighting intensity demand, extraction method: According to the lighting intensity data in the harsh environment fusion state information, combined with the lighting demand standard of the underground working face (such as the minimum illuminance value specified in the national work safety standard), calculate the required lighting intensity in the current environment.

[0054] Specific data: At a certain moment, the actual lighting intensity of the underground working face is 100 lx, while the specified minimum illuminance value is 200 lx. Therefore, the extracted lighting intensity demand feature value is 200 lx.

[0055] The second lighting demand-related feature: color temperature demand, extraction method: According to the temperature data in the harsh environment fusion state information and the comfortable perception range of the color temperature of the human body at different temperatures (such as preferring a low color temperature in a warm environment and a high color temperature in a cold environment), determine the appropriate color temperature value in the current environment.

[0056] Specific data: The underground temperature is 30°C, which belongs to a warm environment. According to the human comfort perception range, the appropriate color temperature value is determined to be 4000K (warm tone). Therefore, the extracted color temperature requirement characteristic value is 4000K.

[0057] The third lighting requirement-related characteristic: stability requirement, extraction method: According to the vibration and pressure data in the harsh environment fusion state information, evaluate the stability of the underground environment. In areas with frequent vibration or large pressure changes, the stability requirements for the lighting system are higher.

[0058] Specific data: The reading of the vibration sensor in a certain area is 0.5 mm / s (slight vibration), and the reading of the pressure sensor is 1.2 atm (relatively stable). Based on these data, the stability requirement of this area is evaluated to be at a medium level. Therefore, the extracted stability requirement characteristic value can be quantified as a value between 0 and 1, such as 0.6 (indicating medium stability requirement).

[0059] Construct a harsh environment historical data set, and organize and store the above-extracted first lighting requirement-related characteristics (light intensity requirement), second lighting requirement-related characteristics (color temperature requirement), and third lighting requirement-related characteristics (stability requirement) in time series to form a harsh environment historical data set. These data sets can be used to train and optimize the algorithm model of the intelligent lighting control system to improve the adaptability and accuracy of the system in future operations.

[0060] 104. Estimate the probability density function of the harsh environment historical data set based on the preset kernel density estimation. By selecting the kernel function and bandwidth parameters, obtain the distribution estimation of the future harsh environment lighting requirements. Based on the first lighting requirement-related characteristics, second lighting requirement-related characteristics, and third lighting requirement-related characteristics, combine with the preset GBM gradient boosting machine model to obtain the quantified harsh environment lighting requirement value;

[0061] It should be noted that for the intelligent lighting requirement prediction and quantification in the underground mining environment, the environmental parameters of this area are as follows: average temperature: 45°C; average pressure: 1.5 atm; vibration frequency: 0.4 mm / s; corrosive gas concentration: low; These environmental parameters are continuously monitored by a multi-modal sensor network and form a harsh environment historical data set.

[0062] Harsh environment historical data set preparation, based on the data of the past month, we constructed a harsh environment historical data set containing the following characteristics:

[0063] The first lighting requirement-related characteristic: light intensity requirement (unit: lx)

[0064] Example data: [150, 160, 170, 180, ...]

[0065] Second lighting demand - related feature: Color temperature demand (unit: K)

[0066] Example data: [4200, 4300, 4100, 4250, ...]

[0067] Third lighting demand - related feature: Stability demand (a value between 0 - 1)

[0068] Example data: [0.7, 0.75, 0.65, 0.72, ...]

[0069] Kernel Density Estimation (KDE) implementation, selecting the kernel function: We select the Gaussian kernel function as the kernel function for KDE. Determining the bandwidth parameter: Through cross - validation, the optimal bandwidth parameter is determined to be 0.1. Probability density function estimation: Apply KDE to estimate the probability density function of the light intensity demand in the historical dataset of harsh environments, obtaining the distribution estimate of future harsh environment lighting demands. Estimation result: The light intensity demand is mainly distributed between 150 - 200 lx, and the peak appears near 170 lx.

[0070] GBM Gradient Boosting Machine model application, feature input: Use the first, second, and third lighting demand - related features as the input of the GBM model. Model training: Use the historical dataset of harsh environments to train the GBM model and optimize the model parameters. Quantifying the prediction of harsh environment lighting demand values: Apply the trained GBM model, combined with the current harsh environment fusion status information, to predict future quantified harsh environment lighting demand values.

[0071] Prediction result: The current harsh environment fusion status information shows that the light intensity demand is 165 lx, the color temperature demand is 4200 K, and the stability demand is 0.7. Based on these inputs, the GBM model predicts that the future quantified harsh environment lighting demand value is 172 lx (indicating that the system should be adjusted to this light intensity under the premise of meeting safety and work efficiency).

[0072] Based on the output result of the harsh environment lighting control system, the distribution estimate of future harsh environment lighting demands, and the quantified harsh environment lighting demand values, obtain the intelligent lighting control strategy for harsh environments;

[0073] It should be noted that in the underground mining environment, lighting is crucial for work efficiency and worker safety. Considering harsh conditions such as high temperature, high pressure, vibration, and corrosion, we need to build an intelligent lighting control system that can optimize lighting strategies based on multi - sensor fusion technology.

[0074] The current output result of the lighting control system shows: Current light intensity: 200 lx; Current color temperature: 4000 K;

[0075] Distribution estimation of future lighting requirements in harsh environments. The distribution of future lighting requirements in harsh environments is estimated by the kernel density estimation method:

[0076] Estimation result: The future lighting requirements in harsh environments are mainly distributed between 150 - 250 lx, and the requirements in the range of 180 - 220 lx are the most concentrated.

[0077] Quantify the lighting requirement value in harsh environments. The quantified lighting requirement value in harsh environments predicted according to the GBM gradient boosting machine model: Prediction result: The quantified lighting requirement value in harsh environments is 210 lx.

[0078] Formulate intelligent lighting control strategies. Compare the output result of the lighting control system in harsh environments with the distribution of future lighting requirements in harsh environments: The current light intensity is 200 lx, while the future lighting requirements in harsh environments are mainly distributed between 150 - 250 lx, and the predicted requirement value is 210 lx.

[0079] Adjustment strategy: Since the current light intensity is slightly lower than the predicted requirement value, the control system should gradually adjust the light intensity to about 210 lx to meet the future lighting requirements in harsh environments. At the same time, considering work efficiency and the visual comfort of workers, maintain the color temperature between 4000 - 4500 K.

[0080] Implementation details: The lighting system should gradually increase the light intensity to avoid sudden changes that may impact the vision of workers. Adjust the light intensity and color temperature by adjusting the power and color ratio of LED lamps. The monitoring system should detect environmental parameters and lighting effects in real time to ensure the effectiveness of the lighting strategy.

[0081] Specific data:

[0082] Before adjustment: Light intensity 200 lx, Color temperature 4000 K

[0083] After adjustment: Light intensity 210 lx, Color temperature 4200 K (fine-tuned according to actual requirements)

[0084] Through the above embodiments, we have demonstrated how to formulate intelligent lighting control strategies in harsh environments based on the output result of the lighting control system in harsh environments, the distribution estimation of future lighting requirements in harsh environments, and the quantified lighting requirement value in harsh environments. These strategies can ensure that the lighting in the underground mining environment meets both work efficiency requirements and the safety of workers.

[0085] In the embodiments of the present invention, in harsh environments such as underground mining, appropriate lighting is a key factor in ensuring the safety of workers. By continuously monitoring environmental parameters and adjusting the lighting strategy, the present invention can ensure that workers can obtain sufficient lighting in any environment, thereby reducing safety accidents caused by insufficient lighting. Appropriate lighting conditions not only help workers better perform their work tasks, but also reduce visual fatigue and improve work efficiency. By intelligently adjusting the lighting brightness and color temperature, the present invention creates a comfortable and efficient working environment for workers. By precisely controlling the brightness and color temperature of the lighting system, the present invention can avoid unnecessary energy waste and achieve the goal of energy conservation and environmental protection. While meeting the lighting requirements, it reduces energy consumption. The present invention realizes the intelligent management of lighting. By automatically adjusting the lighting strategy, it reduces the need for manual intervention and improves the management efficiency. At the same time, through continuous monitoring and data analysis, the system can continuously optimize the lighting strategy to adapt to changing environmental conditions. The multi-sensor fusion technology improves the stability and reliability of the lighting control system. Even in harsh environments, the system can accurately sense environmental changes and make corresponding adjustments to ensure the stability and continuity of lighting.

[0086] Please refer to Figure 2 , another embodiment of the intelligent lighting control method for harsh environments based on multi-sensor fusion in the embodiments of the present invention includes:

[0087] 201. Acquire and fuse the input control harsh environment information in the multi-modal sensor network data according to the input control harsh environment information collected by the multi-modal sensor network, including the monitoring data of light intensity, temperature, pressure, vibration and corrosive substances, to obtain the harsh environment fusion state information;

[0088] Specifically, based on the multi-modal sensor network to collect input control information of the harsh environment, the multi-modal sensor network includes a light intensity sensor, a human movement sensor, a temperature and humidity sensor, and a sound sensor. Using the heat and electrical signals generated by each sensor itself as additional resources, an additional resource dataset is obtained. Each sensor collects environmental data according to a preset sampling frequency and packs the data. The sink node in the sensor network collects data from each sensor to form a multi-modal input control environment sensor dataset; a preset dynamic sensor activation strategy is set. According to the current environmental conditions and lighting requirements, the most relevant sensors are dynamically selected and activated. When there is sufficient daylight, the sampling frequency of the light intensity sensor is reduced or turned off, while when it is night or the light is dim, the sampling frequency is increased; the multi-modal input control environment sensor dataset is cleaned, noise and outliers are removed, and normalization processing is performed to make the data of different sensors comparable and consistent; a Kalman filter model is constructed for each sensor, and the state estimate is updated using the estimated value of the previous moment and the observed value of the current moment. According to the characteristics of the sensor and the requirements of lighting control, the initial state, state transition equation, observation equation, and noise covariance matrix parameters of the Kalman filter are set. At least two groups of multi-modal input control environment sensor data are input into the Kalman filter for data fusion processing. The Kalman filter iteratively updates the state estimate and error covariance matrix according to the preset model parameters and input data. An initial state of the Kalman filter is set for each sensor, including the initial state estimate value and the initial error covariance matrix P0, and the state transition equation is set as F, then: Wherein, is the current moment state prediction based on the estimated value of the previous moment;

[0089] The observation matrix is H, and the observation equation is: Wherein, z k is the observed value of the current moment, x k is the true state of the current moment, v k is the observation noise;

[0090] The process noise covariance matrix is Q, and the observation noise covariance matrix is R;

[0091] The recursive update process of the Kalman filter includes two steps: prediction and update: Predict the error covariance matrix: Update step:

[0092] Calculate the Kalman gain: Update the state estimate: Update the error covariance matrix: Wherein, K kis a key parameter in the Kalman filter, which determines how to fuse the observation information into the state estimate. Its calculation depends on the prediction error covariance matrix P k∣k−1 , the observation matrix H, and the observation noise covariance matrix R. Specifically, K k is obtained by solving an optimization problem, aiming to minimize the mean square error of the state estimate. The role of K k can be regarded as a weight factor, which balances the contributions of the predicted value and the observed value in updating the state estimate. When the observation noise is large, K k will be small, meaning that the predicted value dominates in the update; when the prediction error is large, K k will be large, meaning that the observed value dominates in the update; through data fusion processing, the fused state information in a harsh environment containing information from multiple sensors is obtained; and / or the sensor data is fused at different time scales, where the time scales include seconds, minutes, and hours, to capture the environmental change characteristics at different time scales and obtain the fused state information in a harsh environment.

[0093] It should be noted that considering the working environment in coal mines, the characteristics include high humidity, a lot of dust, poor ventilation, the possible presence of harmful gases such as gas, and extremely insufficient light. The intelligent lighting control system is crucial for ensuring the safety of miners and improving work efficiency.

[0094] Implementation steps: Multi-modal sensor network data collection: Deploy light intensity sensors, human movement sensors, temperature and humidity sensors, and sound sensors. Set the sampling frequencies: light intensity sensor (1 Hz), human movement sensor (0.5 Hz), temperature and humidity sensor (0.2 Hz), sound sensor (2 Hz).

[0095] Specific data: Light intensity sensor: Initial reading 50 Lux (extremely weak light underground); Human movement sensor: Reading 0 when there is no movement, reading 1 when there is movement; Temperature and humidity sensor: Temperature 28 °C, humidity 80%RH; Sound sensor: Initial reading 40 dB (background noise); Dynamic sensor activation strategy. Due to the extremely weak light underground, maintain a high sampling frequency for the light intensity sensor. When the human movement sensor detects movement, temporarily increase the sampling frequency of the sound sensor to capture possible abnormal sounds.

[0096] Clean the collected data to remove outliers caused by equipment failures or temporary interferences. Normalization processing: Normalize the temperature and humidity data to the [0,1] interval for subsequent data processing. Specific data (after normalization):

[0097] Temperature: 28°C is normalized to (28 - (-40)) / (85 - (-40)) ≈ 0.73

[0098] Humidity: 80%RH is normalized to 80 / 100 = 0.8

[0099] Kalman filter model construction and data fusion. A Kalman filter model is constructed for each sensor, and parameters such as the initial state, error covariance matrix, and state transition equation are set.

[0100] Use the Kalman filter for data fusion, combine the readings of multiple sensors, and obtain a more accurate estimate of the environmental state.

[0101] Specific parameters (taking the light intensity sensor as an example):

[0102] Initial state estimate value = 50 Lux; Initial error covariance matrix P0 = 25 (assuming an error of ±5 Lux in the initial estimate)

[0103] State transition equation F = 1 (assuming little change in light intensity in a short period)

[0104] Observation matrix H = 1

[0105] Process noise covariance matrix Q = 4 (representing the uncertainty in the state transition process)

[0106] Observation noise covariance matrix R = 9 (representing the uncertainty in the observation process)

[0107] Data fusion result: Through the iterative update of the Kalman filter and the combination of data from multiple sensors, the system obtains a more accurate and comprehensive estimate of the underground environmental state, including information such as light intensity, temperature and humidity, human movement, and sound intensity. This information is used to intelligently adjust the brightness and working mode of the lighting system to ensure the safety of miners and improve work efficiency.

[0108] Multi-time scale data fusion fuses sensor data at different time scales such as seconds, minutes, and hours to capture the long-term and short-term characteristics of environmental changes. Adjust the lighting strategy according to the fusion result, for example, reduce the lighting intensity to save energy when there is no human activity for a long time.

[0109] 202. Using a preset entropy weight model, combined with the information on the fusion state of the harsh environment, analyze the data importance of different sensors in the multi-modal sensor network, and perform weight allocation according to their reliability and effectiveness in the harsh environment to obtain the weight and reading of each sensor's data. Adopt fuzzy logic control, use the reading of each sensor's data as an input variable, and use the weight of each sensor's data to adjust the membership function of the input to obtain the output result of the harsh environment lighting control system;

[0110] Specifically, based on the information on the fusion state of the harsh environment, construct an evaluation matrix containing the data of different sensors in the multi-modal sensor network. Each row of the evaluation matrix represents a sensor, and each column represents the reading of the sensor at different times or under different conditions. Set the evaluation matrix as R, then: R is an m×n matrix, where m is the number of sensors and n is the reading at different times or under different conditions, and r ij represents the reading of the i-th sensor at the j-th time or under the j-th condition; Input the evaluation matrix into the entropy weight model to calculate the entropy value of each sensor. The lower the entropy value, the more reliable and effective the information provided by the sensor. Calculate the weight of each sensor according to the entropy value. The weight is inversely proportional to the entropy value, that is, the smaller the entropy value, the larger the weight. Set the entropy value as e i , then: where p ij is the normalized sensor reading; k is a constant, taking such that 0 ≤ e i ≤ 1; w i is the weight of the i-th sensor; Introduce an online learning algorithm, which includes incremental learning and online gradient descent, so that the entropy weight model can be dynamically updated according to real-time environmental data and lighting effect feedback; Construct a membership function for each sensor, map the sensor reading in the information on the fusion state of the harsh environment to the input domain of fuzzy logic control, and combine the weight of each sensor's data to adjust the membership function. Use the weight as the coefficient of the membership function, or multiply the weight by the output of the membership function to obtain the adjusted membership function; Fuzzy logic control includes a fuzzy rule base and a fuzzy inference mechanism; Design a fuzzy rule base, and the rule base contains the corresponding lighting control strategies under different combinations of sensor readings; and / or design a multi-level fuzzy logic control, where each level is responsible for processing sensor data and lighting control tasks with different granularities. The low-level controller is responsible for basic brightness adjustment, and the high-level controller is responsible for more complex scene recognition and lighting strategy adjustment; Use the fuzzy inference mechanism, based on the adjusted membership function and the fuzzy rule base, perform fuzzy inference to obtain the output result of the harsh environment lighting control system.

[0111] It should be noted that in an intelligent lighting control system for harsh environments based on multi-sensor fusion, we have adopted four sensors: a light intensity sensor (S1), a human movement sensor (S2), a temperature and humidity sensor (S3), and a sound sensor (S4). These sensors collect data under different times and environmental conditions. According to the readings of the four sensors at five different time points (T1, T2, T3, T4, T5), an evaluation matrix R is constructed: Among them, each row of the matrix represents the evaluation data of a sensor. Specifically: the first row (S1) represents the evaluation data of light intensity; the second row (S2) represents the evaluation data of human movement (represented by 0 and 1, which may be some binary detection results); the third row (S3) represents the evaluation data of temperature; the fourth row (S4) represents the evaluation data of sound intensity. Here, m = 4 (the number of sensors), and n = 5 (the number of time points). The entropy value e of each sensor is calculated using the entropy weight model i and the weight w i : The calculated weights are set as: w1 = 0.4, w2 = 0.2, w3 = 0.1, w4 = 0.3. As new data arrives, the parameters in the entropy weight model are updated using incremental learning and online gradient descent methods to adapt to environmental changes. A membership function is constructed for each sensor and adjusted according to the weight. For example, for the light intensity sensor S1, a triangular membership function can be constructed to represent the fuzzy concept of "moderate brightness", and its weight w1 is used as a coefficient.

[0112] A fuzzy rule base is designed. For example: If S1 (light intensity) is low and S2 (human movement) has activity, then increase the brightness. If S3 (temperature) is high and S4 (sound intensity) is high, then adjust to a cold tone lighting. A two-level fuzzy logic controller is designed: The low-level controller adjusts the basic brightness according to the readings of S1 and S2; the high-level controller adjusts the lighting strategy and scene according to the readings of S3 and S4.

[0113] Using the fuzzy inference mechanism, combined with the adjusted membership function and the fuzzy rule base, reasoning is carried out to obtain the output result of the lighting control system for harsh environments. For example, at a certain time point, according to the sensor readings and fuzzy rules, the system decides to increase the brightness and adjust to warm tone lighting to provide a more comfortable environment.

[0114] 203. Extract first lighting demand-related features, second lighting demand-related features, and third lighting demand-related features based on the harsh environment fusion state information, and construct a harsh environment historical data set based on the first lighting demand-related features, second lighting demand-related features, and third lighting demand-related features;

[0115] Specifically, according to the lighting control requirements, three main feature dimensions are defined: the first lighting requirement-related feature is the light intensity requirement, the second lighting requirement-related feature is the color and color temperature requirement, and the third lighting requirement-related feature is the dynamic adjustment requirement; based on the light sensor data, combined with time, the contrast between the internal and external light in the working environment, and the special lighting conditions in the harsh environment, the special lighting conditions in the harsh environment include the dimness and uneven lighting in the underground mining environment. By using a machine learning model, the most suitable light intensity is predicted to obtain the first lighting requirement-related feature; using color sensor and camera data, the environmental color, human activities, and the special requirements of the harsh environment for color and color temperature are analyzed. The special requirements of the harsh environment for color and color temperature include the demand for the anti-corrosion performance of equipment in a corrosive environment. Pattern recognition technology is used to determine the color and color temperature combination to obtain the second lighting requirement-related feature; combined with human movement detection data, historical behavior patterns, and the impact of the harsh environment on human movement and lighting requirements, the impact of the harsh environment on human movement and lighting requirements includes the limited personnel flow in high-temperature and high-pressure environments. Time series analysis or machine learning algorithms are used to predict the change trend of lighting requirements in the future period, including the personnel flow path and work area conversion, to obtain the third lighting requirement-related feature; design a comprehensive data table including timestamps, environmental parameters, lighting settings, and user feedback, and collect the first lighting requirement-related feature, the second lighting requirement-related feature, the third lighting requirement-related feature, and the corresponding lighting control results and user feedback extracted in the above steps in real time; label the first lighting requirement-related feature, the second lighting requirement-related feature, and the third lighting requirement-related feature, mark the changes in lighting requirements under special events, and the user preference patterns; use data enhancement techniques, including synthetic data generation and time series extension, to enrich the dataset and improve the generalization ability of the model; obtain the historical dataset of the harsh environment.

[0116] It should be noted that in the working environment of coal mines, to ensure the safety of miners and improve work efficiency, we implemented an intelligent lighting control method for harsh environments based on multi-sensor fusion. This method mainly realizes intelligent lighting control by defining and extracting three main feature dimensions.

[0117] The first lighting requirement-related feature: Light intensity requirement: Data collection: Collect the light intensity data of different areas in the underground mine through light sensors, and record the timestamp, the contrast between the internal and external light in the working environment, and the special lighting conditions in the underground mining environment (such as dimness, uneven lighting, etc.). Feature extraction: Combine time factors (such as day, night, dusk, etc.), the contrast between the internal and external light in the working environment, and the special lighting conditions in the underground mining environment, and use a machine learning model (such as random forest, neural network, etc.) to predict the most suitable light intensity.

[0118] Sample data:

[0119] Timestamp: 2024-09-24 12:00:00

[0120] Illumination intensity in the working environment: 10 Lux

[0121] Illumination intensity outside the working environment: 5000 Lux

[0122] Description of the underground environment: Dim, uneven illumination

[0123] Predicted most suitable illumination intensity: 500 Lux.

[0124] Second lighting demand-related features: Color and color temperature requirements: Data collection: Use color sensors and cameras to capture the color information of the underground environment and the activities of miners. Feature extraction: Analyze the environmental color, human activities, and special requirements for color and color temperature in harsh environments (such as the demand for equipment corrosion resistance in corrosive environments) through pattern recognition technologies (such as support vector machines, deep learning models, etc.).

[0125] Sample data:

[0126] Main color of the environment: Gray

[0127] Activity status of miners: At work

[0128] Special requirements for harsh environments: Corrosive environment, need to consider equipment corrosion protection.

[0129] Recommended color and color temperature combination: Cool color tone, 5000K color temperature.

[0130] Third lighting demand-related features: Dynamic adjustment demand: Data collection: Record the movement data of miners through human movement detection sensors and analyze it in combination with historical behavior patterns. Feature extraction: Combine human movement detection data, historical behavior patterns, and the impact of harsh environments on human movement and lighting demand (such as restricted personnel flow in high-temperature and high-pressure environments), and use time series analysis or machine learning algorithms to predict the change trend of lighting demand in the future period, including personnel movement paths, work area conversions, etc.

[0131] Sample data:

[0132] Current location of miners: Area A

[0133] Predicted location in the next time period: Area B

[0134] Impact of harsh environment: High-temperature environment, restricted personnel flow

[0135] Lighting adjustment strategy: Gradually increase the lighting brightness in Area B and reduce the lighting brightness in Area A.

[0136] Furthermore, in harsh environments, the intelligent lighting control system needs to comprehensively consider various factors to accurately extract features related to lighting requirements. In addition to conventional environmental parameters (such as light intensity, temperature, pressure, vibration), the monitoring data of corrosive substances is particularly added, which is crucial for formulating more precise control strategies.

[0137] The light intensity requirement is not only affected by the current light intensity, time, and the contrast of light inside and outside the working environment, but also needs to consider the potential damage that corrosive substances may cause to lighting equipment. In a highly corrosive environment, lighting equipment may age or be damaged due to corrosion, resulting in a decrease in the actual light intensity. Therefore, when predicting the light intensity requirement, the concentration of corrosive substances and their impact on equipment must be considered.

[0138] Data example analysis:

[0139] Light sensor data: The current light intensity is 100 lx.

[0140] Time factor: 3 pm, the light outside the working environment is moderate.

[0141] Corrosive substance data: The concentration of sulfuric acid mist is 5 ppm (the threshold is 10 ppm).

[0142] Prediction of light intensity requirement: Although the current light intensity is low, considering that the concentration of sulfuric acid mist is low and does not exceed the threshold, the damage to the equipment is limited. Therefore, the predicted light intensity requirement is 200 lx (standard requirement). However, if the concentration of sulfuric acid mist approaches or exceeds the threshold, it may be necessary to increase the light intensity to compensate for potential equipment damage.

[0143] Second lighting requirement-related feature: Color and color temperature requirements

[0144] Color and color temperature requirements are not only related to the environmental color and human activities, but also affected by corrosive substances. Certain corrosive gases may affect workers' perception of color and color temperature. Therefore, it is necessary to adjust the color temperature according to the concentration of corrosive substances to reduce the irritation to workers' eyes.

[0145] Data example analysis:

[0146] Environmental color: Mainly gray tone.

[0147] Human activity situation: Workers are performing delicate operations.

[0148] Corrosive substance data: The concentration of hydrogen chloride is 3 ppm (the threshold is 5 ppm).

[0149] Color and Color Temperature Requirements: When performing delicate operations in a gray-toned environment, it is recommended to use a cool color temperature (such as 6500K) to improve visual clarity. At the same time, considering that the concentration of hydrogen chloride is low and does not exceed the threshold, the current color temperature setting should be able to reduce the irritation to the workers' eyes.

[0150] Third Lighting Requirement-related Feature: Dynamic Adjustment Requirement

[0151] The dynamic adjustment requirement involves predicting the future trend of lighting demand changes and must comprehensively consider factors such as personnel flow, work area conversion, and changes in corrosive substances. Especially when the concentration of corrosive substances suddenly increases, it is necessary to immediately adjust the lighting strategy to reduce potential hazards to workers.

[0152] Data Example Analysis:

[0153] Personnel Flow Data: It is expected that workers will transfer from Area A to Area B within the next hour.

[0154] Work Area Conversion: Area B is a high-risk area for corrosive substances.

[0155] Corrosive Substance Data: The concentration of nitric acid mist in Area B is monitored in real time and has rapidly increased from 2 ppm to 8 ppm (the threshold is 10 ppm).

[0156] Dynamic Adjustment Strategy: Given that the concentration of nitric acid mist in Area B has rapidly increased and is approaching the threshold, the protection level of lighting equipment in Area B should be immediately increased, and the lighting strategy should be adjusted to a low brightness and high color temperature mode. This can reduce the time workers are exposed to the highly corrosive environment and improve visual alertness at the same time.

[0157] Design a comprehensive data table that includes timestamps, environmental parameters, lighting settings, and user feedback to collect the first lighting requirement-related feature, the second lighting requirement-related feature, the third lighting requirement-related feature, and the corresponding lighting control results and user feedback extracted in the above steps in real time.

[0158] Data table fields include: Timestamp: Records the specific time of data collection; Environmental parameters: Light intensity, color, color temperature, special underground lighting conditions, corrosive environment, etc.; Lighting settings: The set values of light intensity, color, and color temperature; User feedback: Miners' satisfaction feedback on the current lighting settings

[0159] Data annotation and enhancement, annotate the collected data to mark changes in lighting requirements and user preference patterns under special events. Use data enhancement techniques (such as synthetic data generation, time series extension, etc.) to enrich the dataset and improve the generalization ability of the model. For example, different underground environmental changes at different time periods can be simulated to generate more diverse data samples.

[0160] Construct a historical dataset for harsh environments, integrate the labeled and enhanced data to form a historical dataset for harsh environments that includes features related to the first lighting requirement, the second lighting requirement, and the third lighting requirement. This dataset will be used for subsequent model training and the formulation of intelligent lighting control strategies.

[0161] 204. Perform probability density function estimation on the historical dataset for harsh environments according to a preset kernel density estimation. By selecting a kernel function and bandwidth parameters, obtain the distribution estimation of future lighting requirements in harsh environments. Combine the features related to the first lighting requirement, the second lighting requirement, and the third lighting requirement with a preset GBM gradient boosting machine model to obtain a quantified lighting requirement value for harsh environments.

[0162] Specifically, in combination with context recognition technology, dynamically select a kernel function according to the current context. The current context includes weather, season, and activity type. The kernel functions include Gaussian kernel and Epanechnikov kernel. Use cross-validation, interpolation method, or empirical formula to determine the bandwidth parameters. Perform kernel density estimation on each feature dimension separately to obtain their respective probability density functions; and / or use a multi-dimensional kernel density estimation method, including Copula-based multi-dimensional kernel density estimation, to more accurately capture the joint distribution between features and obtain their respective probability density functions. Use the probability density function of the historical dataset for harsh environments and adopt a sliding window method to estimate the distribution of future lighting requirements in harsh environments: Among them, is the conditional probability of future lighting requirements x in harsh environments t+1 is the indicator function, N is the number of sliding windows in the historical dataset for harsh environments, is the lighting requirement sequence in the historical dataset for harsh environments; obtain the distribution estimation of future lighting requirements in harsh environments; preprocess the features related to the first lighting requirement, the second lighting requirement, and the third lighting requirement at the current moment, including data cleaning and normalization; construct a GBM gradient boosting machine model, select a loss function and weak learners. The loss function uses the squared error, and the weak learners include decision trees and linear regression; train the GBM gradient boosting machine model according to the features and corresponding lighting requirement values in the historical dataset for harsh environments, and obtain the optimal model parameters through iterative optimization. For the squared error loss function, use the following formula to update the model: Among them, F m (x) is the GBM gradient boosting machine model after the m-th iteration, h (x; a m ) is the weak learner in the m-th iteration, βm is the weight of the weak learner, and a m is the parameter of the weak learner; m

[0163] Input the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature at the current moment into the trained GBM (Gradient Boosting Machine) model: Among them, is the predicted lighting demand value, and F M (x') is the trained GBM model, and x' is the preprocessed feature value; the quantized lighting demand value is obtained.

[0164] It should be noted that considering the working environment in coal mines, its characteristics include high humidity, a lot of dust, poor ventilation, etc., and the light is extremely insufficient. To ensure the safety of miners and improve work efficiency, we implemented an intelligent lighting control method for harsh environments based on multi-sensor fusion. This embodiment will detail how to dynamically select a kernel function in combination with context recognition technology and use the GBM model to predict lighting demand.

[0165] Context recognition and kernel function selection, current context: weather - cloudy, season - autumn, activity type - tunneling operation. Dynamically select the kernel function: According to the current context, select the Gaussian kernel for kernel density estimation.

[0166] Kernel density estimation, data: Use features such as light intensity, color and color temperature requirements, and dynamic adjustment requirements in the historical dataset of harsh environments. Kernel density estimation: Perform Gaussian kernel density estimation on each feature dimension respectively to obtain their respective probability density functions.

[0167] Estimation of future lighting demand distribution in harsh environments, sliding window method: Use the sliding window method to estimate the distribution of future lighting demand in harsh environments. Parameter settings: Set the number of sliding windows N = 100, and the length of the lighting demand sequence in the historical dataset of harsh environments is n = 10. Result: Obtain the conditional probability distribution of future lighting demand in harsh environments.

[0168] Feature preprocessing, data cleaning: Remove outliers and missing values. Normalization: Normalize the first lighting demand-related feature (light intensity demand), the second lighting demand-related feature (color and color temperature demand), and the third lighting demand-related feature (dynamic adjustment demand).

[0169] GBM model construction and training, model parameters: Select the mean squared error as the loss function, and use decision trees as weak learners. Training data: Use the features and corresponding lighting demand values in the historical dataset of harsh environments for model training. Iterative optimization: Obtain the optimal model parameters through iterative optimization, including the weights and parameters of the weak learners.

[0170] Illumination demand prediction, input features: Input the first, second, and third illumination demand-related features after preprocessing at the current moment into the trained GBM model. Prediction result: Obtain the quantified illumination demand value. For example, the predicted light intensity is 600 Lux, the recommended color and color temperature combination is warm color tone and 4000K color temperature, and the dynamic adjustment strategy is to gradually increase the illumination brightness of the working area.

[0171] 205. Obtain the intelligent lighting control strategy for the harsh environment based on the output result of the harsh environment lighting control system, the distribution estimation of future harsh environment light demand, and the quantified harsh environment lighting demand value;

[0172] Specifically, integrate the output result of the harsh environment lighting control system, the distribution estimation of future harsh environment light demand, and the quantified harsh environment lighting demand value; use the distribution estimation of future harsh environment light demand to predict the change trend of light demand in the future for a period of time by analyzing the probability density function, peak value, and valley value information in the distribution estimation; evaluate the output result of the harsh environment lighting control system, and judge whether the current lighting system meets the lighting demand and whether there are problems of over-brightness, over-darkness, or uneven light distribution by comparing the actual output of the lighting system with the preset lighting standard or user expectations, combined with the quantified harsh environment lighting demand value; form an environment-demand-response dataset; conduct in-depth analysis on the environment-demand-response dataset to identify the change trend, outliers, and periodic patterns of light demand, and obtain the intelligent lighting control strategy for the harsh environment.

[0173] It should be noted that in the harsh environment of the coal mine underground, the core task of the intelligent lighting control system is to ensure the safety of miners and improve work efficiency. This embodiment will elaborate on how to specifically implement the intelligent lighting control method for the harsh environment with multi-sensor fusion, including integrating the system output result, the distribution estimation of future harsh environment light demand, and the quantified harsh environment lighting demand value, and forming an environment-demand-response dataset to formulate the control strategy.

[0174] Integrate the system output result and the light demand. System output result: The light intensity of the current lighting system is 500 Lux, the color temperature is 4500K, and the uniformity reaches 0.8. Distribution estimation of future harsh environment light demand: The peak value of light demand in the future for a period of time obtained by kernel density estimation is 600 Lux, and the valley value is 400 Lux. Quantified harsh environment lighting demand value: According to the prediction of the GBM model, the quantified lighting demand value is 550 Lux.

[0175] Analyze the changing trend of lighting requirements in future harsh environments. Probability density function analysis: The lighting requirements in future harsh environments are mainly distributed in the range of 450 - 650 Lux, with a relatively high probability density. Peak and valley value information: It is predicted that the lighting requirements in future harsh environments will peak at around 600 Lux and reach a trough at around 400 Lux. Changing trend prediction: It is expected that the lighting requirements will gradually increase and then tend to be stable in the next period of time.

[0176] Evaluate the output results of the lighting control system in harsh environments. Preset lighting standards: Illuminance 500 - 700 Lux, color temperature 4000 - 5000K, uniformity ≥ 0.7. Comparison of actual output with the standard: The current system output meets the preset standard, but considering the upward trend of lighting requirements in future harsh environments, appropriate adjustments may be needed. Problem identification: There are no problems of over - brightness, over - darkness or uneven light distribution.

[0177] Form an environment - demand - response dataset. Environmental data: Humidity 80%, dust concentration 0.5mg / m³, gas concentration 0.1%, temperature 25°C. Demand data: Lighting intensity demand 550 Lux, color temperature demand 4500K, uniformity demand ≥ 0.8. Response data: The actual output of the system is lighting intensity 500 Lux, color temperature 4500K, uniformity 0.8; The future adjustment strategy is to gradually increase the lighting intensity to 550 Lux.

[0178] Deeply analyze the dataset and formulate control strategies. Changing trend identification: The lighting requirements show periodic fluctuations with time and operating activities. Outlier detection: No obvious outliers are found. Periodic pattern: The lighting requirements are relatively high from 10 am to 2 pm every day and relatively low from 8 pm to 4 am the next day. Control strategy formulation: According to the periodic pattern, formulate a lighting adjustment strategy by time period to meet the lighting requirements in different time periods.

[0179] 206. Set a benchmark lighting control strategy based on the harsh environment historical dataset and user preferences, including default light intensity, color temperature, and dynamic adjustment rules; formulate a dynamic adjustment strategy based on the distribution estimation of future harsh environment lighting requirements. When it is predicted that the future harsh environment lighting requirements will increase, adjust the light intensity or color temperature in advance to meet user needs; consider the impact of the harsh environment and formulate an emergency lighting control strategy to ensure the stability and safety of the lighting system in the harsh environment; formulate a personalized control strategy according to the user's personalized lighting requirements, and adjust the light intensity and color temperature according to the user's activity habits and emotional state; collect the feedback data of the lighting control system in real time, including actual light intensity, color temperature, and energy consumption, compare it with the expected value, and make timely adjustments when deviations are found; regularly evaluate and optimize the control strategy, and continuously adjust and improve the control strategy according to user feedback, environmental changes, and technological progress to improve the intelligent level of lighting control and user satisfaction.

[0180] It should be noted that the working environment in coal mines is harsh, and an intelligent lighting control system is required to ensure the safety of miners and improve work efficiency. This embodiment will specifically illustrate how to implement an intelligent lighting control method in a harsh environment based on multi-sensor fusion, including setting a benchmark lighting control strategy, formulating a dynamic adjustment strategy, an emergency lighting control strategy, a personalized control strategy, as well as real-time feedback adjustment and strategy evaluation and optimization.

[0181] Set a benchmark lighting control strategy. Default light intensity: Based on historical data, set the default light intensity underground to 500 Lux. Default color temperature: Considering the visual comfort of miners, set the default color temperature to 4500K. Dynamic adjustment rule: The light intensity is automatically adjusted every 30 minutes, and the adjustment range each time does not exceed ±10%.

[0182] Formulate a dynamic adjustment strategy. Distribution estimation of future harsh environment lighting requirements: Through sensor data prediction, the lighting requirements will increase to 600 Lux within the next 2 hours. Advance adjustment: The system starts to gradually increase the light intensity 15 minutes in advance to ensure that it reaches 600 Lux at the demand peak.

[0183] Emergency lighting control strategy. Harsh environment detection: When the sensor detects that the gas concentration exceeds 0.5%, trigger the emergency lighting. Emergency lighting parameters: The light intensity is automatically increased to 800 Lux, and the color temperature is adjusted to 5000K to improve vigilance. Stability guarantee: In the emergency lighting mode, the system automatically switches to the backup power supply to ensure continuous lighting for more than 6 hours.

[0184] Personalized control strategy, user activity habits: According to the miner's work shift, the light intensity is set to 550 Lux during the day and reduced to 450 Lux at night. Emotional state adjustment: Detect the emotional state through the wearable device worn by the miner. When fatigue is detected, the color temperature is automatically adjusted to a warmer 3500K.

[0185] Real-time feedback adjustment, actual light intensity and color temperature monitoring: The system real-time monitors that the current light intensity is 490 Lux and the color temperature is 4400K. Comparison with the expected value: It is found that the actual light intensity is lower than the set value of 500 Lux. Timely adjustment: The system automatically increases the power of the lighting equipment to restore the light intensity to the set value.

[0186] Strategy evaluation and optimization, user feedback collection: Regularly conduct lighting satisfaction surveys on miners and collect feedback. Environmental change monitoring: Continuously monitor changes in underground environmental parameters such as humidity and dust concentration. Integration of technological progress: With the development of LED lighting technology, regularly update more efficient and stable lighting equipment. Strategy optimization implementation: According to user feedback, environmental changes, and technological progress, conduct a comprehensive evaluation and optimization of the control strategy every quarter.

[0187] In the embodiments of the present invention, by real-time monitoring of underground environmental parameters and miner activities, the lighting system is intelligently adjusted, effectively avoiding potential safety hazards caused by insufficient lighting or over-illumination, and ensuring the safe operation of miners; according to the miner's activity habits and emotional state, the lighting brightness and color temperature are dynamically adjusted to create a more comfortable working environment, thereby improving the work efficiency and work quality of miners; by accurately predicting and dynamically adjusting lighting requirements, unnecessary energy waste is avoided, and the goals of energy conservation and emission reduction are achieved. Automatically reduce the lighting brightness in areas with no personnel activity for a long time to reduce energy consumption; adopt a variety of advanced algorithms and technical means (such as Kalman filtering, entropy weight model, GBM gradient boosting machine, etc.) to improve the intelligent level and adaptive ability of the system, ensuring stable operation of the system in different environments and requirements; by real-time collecting miners' feedback on the lighting system and optimizing the strategy according to the feedback, improve miners' satisfaction and acceptance of the lighting system, and enhance miners' work enthusiasm and sense of belonging.

[0188] In summary, the present invention can not only effectively cope with the challenges of harsh environments such as underground coal mines, improve the safety and work efficiency of miners, but also achieve multiple goals such as energy conservation and emission reduction and improving the intelligent level of the system, with significant beneficial effects and broad application prospects.

[0189] The above describes the intelligent lighting control method for harsh environments based on multi-sensor fusion in the embodiments of the present invention. Next, the intelligent lighting control device for harsh environments based on multi-sensor fusion in the embodiments of the present invention will be described. Please refer to Figure 3, an embodiment of the intelligent lighting control device based on multi-sensor fusion in the embodiments of the present invention includes: an acquisition module 301, configured to acquire input control harsh environment information, collect the input control harsh environment information based on a multi-modal sensor network, obtain at least two groups of multi-modal input control environment sensor data, and perform data fusion processing on the at least two groups of multi-modal input control environment sensor data to obtain harsh environment fusion state information; a processing module 302, configured to analyze the data importance of different sensors in the multi-modal sensor network according to the harsh environment fusion state information in combination with a preset entropy weight model, perform weight allocation according to its reliability and effectiveness in the harsh environment to obtain the weight of each sensor data and the reading of each sensor, the input variable of the fuzzy logic control is the reading of each sensor, and use the weight of each sensor data as the membership function for adjusting these inputs to obtain the output result of the harsh environment lighting control system; an extraction module 303, configured to extract lighting demand-related features according to the harsh environment fusion state information to obtain first lighting demand-related features, second lighting demand-related features, and third lighting demand-related features, and construct a harsh environment historical data set according to the first lighting demand-related features, second lighting demand-related features, and third lighting demand-related features; a quantization module 304, configured to perform probability density function estimation on the harsh environment historical data set according to a preset kernel density estimation, obtain the distribution estimation of future harsh environment lighting demand by selecting a kernel function and a bandwidth parameter, and obtain a quantified harsh environment lighting demand value according to the first lighting demand-related features, second lighting demand-related features, and third lighting demand-related features in combination with a preset GBM gradient boosting machine model; an allocation module 305, configured to obtain a harsh environment intelligent lighting control strategy according to the output result of the harsh environment lighting control system, the distribution estimation of future harsh environment lighting demand, and the quantified harsh environment lighting demand value.

[0190] In the embodiments of the present invention, through the multi-sensor fusion technology, the device can comprehensively perceive the environmental state, improve the accuracy and adaptability of lighting control. The application of the entropy weight model enables the device to automatically evaluate the importance of sensor data and effectively filter unreliable data, enhancing the robustness of the system. Based on the harsh environment historical data set and advanced algorithms, the device can predict future harsh environment lighting demand and dynamically adjust the lighting strategy to meet the real-time changing lighting demand. Through accurate prediction and intelligent control, the device can effectively avoid unnecessary lighting waste, improve energy utilization efficiency, and reduce energy consumption costs. The device can provide personalized lighting services according to user preferences and activity habits, improving user satisfaction and comfort.

[0191] Above Figure 3The embodiments of the present invention for the intelligent lighting control device based on multi-sensor fusion in harsh environments will be described in detail from the perspective of modular functional entities. Next, the intelligent lighting control device based on multi-sensor fusion in harsh environments in the embodiments of the present invention will be described in detail from the perspective of hardware processing.

[0192] Figure 4 FIG. 4 is a schematic structural diagram of an intelligent lighting control device based on multi-sensor fusion provided by an embodiment of the present invention. The intelligent lighting control device 400 based on multi-sensor fusion may vary greatly due to configuration or performance, and may include one or more processors (central processing units, CPUs) 410 (for example, one or more processors) and a memory 420, and one or more storage media 430 for storing application programs 433 or data 432 (for example, one or more mass storage devices). Among them, the memory 420 and the storage media 430 may be transient storage or persistent storage. The program stored in the storage media 430 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations for the intelligent lighting control device 400 based on multi-sensor fusion. Further, the processor 410 may be configured to communicate with the storage media 430 and execute a series of instruction operations in the storage media 430 on the intelligent lighting control device 400 based on multi-sensor fusion.

[0193] The intelligent lighting control device 400 based on multi-sensor fusion may further include one or more power supplies 440, one or more wired or wireless network interfaces 450, one or more input / output interfaces 460, and / or one or more operating systems 431, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, and so on. Those skilled in the art can understand that Figure 4 the shown structure of the intelligent lighting control device based on multi-sensor fusion does not limit the intelligent lighting control device based on multi-sensor fusion, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0194] The present invention also provides an intelligent lighting control device based on multi-sensor fusion. The intelligent lighting control device based on multi-sensor fusion includes a memory and a processor. When the computer-readable instructions stored in the memory are executed by the processor, the processor executes the steps of the intelligent lighting control method based on multi-sensor fusion in the above embodiments.

[0195] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the intelligent lighting control method based on multi-sensor fusion.

[0196] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0197] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the present invention can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0198] In summary, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for intelligent lighting control in harsh environments based on multi-sensor fusion, characterized in that: The harsh environment intelligent lighting control method based on multi-sensor fusion includes: According to the multimodal sensor network, the harsh environment information of the input control is collected and integrated in the multimodal sensor network data, including the light intensity, color, image, vibration, temperature, pressure and corrosive material monitoring data, to obtain the harsh environment fusion status information; Using the preset entropy weight model and combining the fusion state information of the harsh environment, the importance of the data of different sensors in the multimodal sensor network is analyzed, and the weights are assigned according to their reliability and effectiveness in the harsh environment to obtain the weight and reading of each sensor data; Fuzzy logic control is adopted, the reading of each sensor data is taken as the input variable, and the input membership function is adjusted by using the weight of each sensor data to obtain the output result of the harsh environment lighting control system; Extracting lighting demand-related features based on the harsh environment fusion state information includes: Based on the light sensor data, combined with the time, the contrast of light inside and outside the working environment, and the special light conditions in the harsh environment, the special light conditions in the harsh environment include the dimness and uneven light conditions in the underground mining environment, the most suitable light intensity is predicted through the machine learning model to obtain the first lighting demand related characteristics; Using color sensors and camera data, analyze the environmental color, human activities, and special requirements of harsh environments for color and color temperature, including the requirements for equipment anti-corrosion performance in corrosive environments, and use pattern recognition technology to determine the combination of color and color temperature to obtain the second lighting demand-related features; Combined with human movement detection data, historical behavior patterns, and the impact of harsh environments on human movement and lighting needs, including the restricted flow of people under high temperature and high pressure environments, time series analysis or machine learning algorithms are used to predict the changing trend of lighting needs in the future, including personnel flow paths and work area conversions, to obtain third lighting demand-related features; The first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature are marked to mark changes in lighting demand under special events and user preference modes; Use data enhancement techniques, including synthetic data generation and time series expansion, to enrich the dataset and improve the generalization ability of the model; Constructing a harsh environment historical data set according to the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature; The probability density function of the harsh environment historical data set is estimated according to the preset kernel density estimation, and the distribution estimation of the future harsh environment lighting demand is obtained by selecting the kernel function and bandwidth parameters, including: Multidimensional kernel density estimation methods, including Copula-based multidimensional kernel density estimation, are used to more accurately capture the joint distribution between features and obtain their respective probability density functions; Using the probability density function of the harsh environment historical data set, the sliding window method is used to estimate the distribution of future harsh environment lighting requirements: in, Is the future harsh environment lighting demand x t+1 The conditional probability of is the indicator function, N is the number of sliding windows in the harsh environment historical data set, is the sequence of light demand in the harsh environment historical dataset; Get the distribution estimate of future harsh environment lighting demand; According to the first lighting demand related feature, the second lighting demand related feature, and the third lighting demand related feature combined with a preset GBM gradient boosting machine model, a quantified harsh environment lighting demand value is obtained; According to the output result of the harsh environment lighting control system, the distribution estimation of future harsh environment lighting demand and the quantified harsh environment lighting demand value, a harsh environment intelligent lighting control strategy is obtained.

2. The method for intelligent lighting control in harsh environments based on multi-sensor fusion according to claim 1 is characterized in that: The multimodal sensor network includes a light sensor, a temperature sensor, a pressure sensor, a vibration sensor and a corrosion monitoring sensor. The heat and electrical signals generated by each sensor are used as additional resources to obtain an additional resource data set. Each sensor collects environmental data according to a preset sampling frequency and packages the data. The aggregation node in the sensor network collects data from each sensor to form a multimodal input control environment sensor data set. Preset dynamic sensor activation strategy to dynamically select and activate the most relevant sensors based on current environmental conditions and lighting requirements. When there is sufficient light during the day, the sampling frequency of the light intensity sensor is reduced or turned off, while the sampling frequency is increased at night or when the light is dim. The sensor data are fused at different time scales, including seconds, minutes, and hours, to capture the environmental change characteristics at different time scales and obtain the fusion status information of harsh environments.

3. The method for intelligent lighting control in harsh environments based on multi-sensor fusion according to claim 1 is characterized in that: The weight distribution of the entropy weight model is combined with an online learning algorithm to update the weight distribution in real time; Design multi-level fuzzy logic control, where each level is responsible for processing sensor data and lighting control tasks of different granularity, with low-level controllers responsible for basic brightness adjustment and high-level controllers responsible for more complex scene recognition and lighting strategy adjustment; By using the fuzzy reasoning mechanism, fuzzy reasoning is performed according to the adjusted membership function and fuzzy rule base to obtain the output result of the harsh environment lighting control system.

4. The method for intelligent lighting control in harsh environments based on multi-sensor fusion according to claim 1, characterized in that: Evaluate the output results of the harsh environment lighting control system, and judge whether the current lighting system meets the lighting needs by comparing the actual output of the lighting system with the preset lighting standards or user expectations, and whether there are problems such as too bright, too dark, or uneven light distribution, combined with quantifying the harsh environment lighting demand value; Forming environment-demand-response data sets; An in-depth analysis is conducted on the environment-demand-response data set to identify the changing trends, abnormal values, and periodic laws of lighting demand, and to obtain intelligent lighting control strategies for harsh environments.

5. The method for intelligent lighting control in harsh environments based on multi-sensor fusion according to claim 1, characterized in that: Also includes: Set a baseline lighting control strategy based on the harsh environment historical data set and user preferences, including default light intensity, color temperature, and dynamic adjustment rules; Based on the estimated distribution of lighting demand in harsh environments in the future, a dynamic adjustment strategy is formulated. When it is predicted that lighting demand in harsh environments will increase in the future, the lighting intensity or color temperature is adjusted in advance to meet user needs; Collect feedback data from the lighting control system in real time, including actual light intensity, color temperature, and energy consumption, compare them with expected values, and make timely adjustments when deviations are found; Regularly evaluate and optimize control strategies, and continuously adjust and improve control strategies based on user feedback, environmental changes, and technological advances to improve the intelligence level of lighting control and user satisfaction.

6. An intelligent lighting control device for harsh environments based on multi-sensor fusion, characterized in that: The harsh environment intelligent lighting control device based on multi-sensor fusion includes: An acquisition module is used to collect input control harsh environment information based on the multimodal sensor network, acquire and fuse the input control harsh environment information in the multimodal sensor network data, including light intensity, temperature, pressure, vibration and corrosive substance monitoring data, and obtain harsh environment fusion state information; The processing module is used to use a preset entropy weight model, combined with the harsh environment fusion state information, to analyze the importance of data from different sensors in the multimodal sensor network, and to assign weights according to their reliability and effectiveness in the harsh environment, so as to obtain the weight and reading of each sensor data; fuzzy logic control is used, the reading of each sensor data is used as an input variable, and the input membership function is adjusted using the weight of each sensor data to obtain the output result of the harsh environment lighting control system; an extraction module, configured to extract lighting demand-related features according to the harsh environment fusion state information to obtain first lighting demand-related features, second lighting demand-related features, and third lighting demand-related features, and to construct a harsh environment historical data set according to the first lighting demand-related features, the second lighting demand-related features, and the third lighting demand-related features; A quantization module is used to estimate the probability density function of the harsh environment historical data set according to a preset kernel density estimation, obtain a distribution estimate of the future harsh environment lighting demand by selecting a kernel function and a bandwidth parameter, and obtain a quantized harsh environment lighting demand value according to the first lighting demand-related feature, the second lighting demand-related feature, and the third lighting demand-related feature combined with a preset GBM gradient boosting machine model, including: Multidimensional kernel density estimation methods, including Copula-based multidimensional kernel density estimation, are used to more accurately capture the joint distribution between features and obtain their respective probability density functions; Using the probability density function of the harsh environment historical data set, the sliding window method is used to estimate the distribution of future harsh environment lighting requirements: in, Is the future harsh environment lighting demand x t+1 The conditional probability of is the indicator function, N is the number of sliding windows in the harsh environment historical data set, is the sequence of light demand in the harsh environment historical dataset; Get the distribution estimate of future harsh environment lighting demand; The allocation module is used to obtain a harsh environment intelligent lighting control strategy based on the output result of the harsh environment lighting control system, the distribution estimation of future harsh environment lighting demand and the quantified harsh environment lighting demand value.

7. A harsh environment intelligent lighting control device based on multi-sensor fusion, characterized in that: The harsh environment intelligent lighting control device based on multi-sensor fusion includes: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the harsh environment intelligent lighting control device based on multi-sensor fusion to execute the harsh environment intelligent lighting control method based on multi-sensor fusion as described in any one of claims 1-5.

8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the harsh environment intelligent lighting control method based on multi-sensor fusion as described in any one of claims 1-5 is implemented.

Images