Beacon light intelligent control method and system based on visible light communication

Abstract

The invention relates to the field of beacon light control, in particular to a beacon light intelligent control method and system based on visible light communication. The method comprises the following steps: acquiring a weather type, a weather type grade, ambient light brightness, visibility and a wind speed of a target area; monitoring the target area based on the ambient light brightness to obtain a target ship; constructing a beacon light power prediction model, and predicting the beacon light power by using the beacon light power prediction model according to the weather type, the weather type grade, the ambient light brightness, the visibility and the wind speed; and obtaining a visible light communication protocol, and controlling the beacon light to flicker at the beacon light power based on the visible light communication protocol and the target ship. According to the invention, the problem that the beacon light is difficult to automatically adjust according to the environmental information in the prior art is solved, the real-time performance, the reliability and the self-adaptability of the working of the beacon light in a complex environment are improved, and the safety of ship navigation is further improved.

Description

Technical Field

[0001] The present invention relates to the field of navigation light control, and in particular to a navigation light intelligent control method and system based on visible light communication. Background Art

[0002] As a critical navigational facility, navigation lights' signal stability and energy efficiency directly impact navigation safety and energy consumption. Traditional navigation light control methods are relatively simple, employing a fixed power output mode that struggles to adapt to dynamic changes in complex environments. In inclement weather such as fog and heavy rain, fixed brightness levels not only result in insufficient signal coverage, making it difficult for ships to accurately identify the location of navigation marks, but also lead to unnecessary energy waste and increased operational costs.

[0003] With the rapid development of smart shipping technology, visible light communication (VLC) has become a key area for the intelligent upgrade of navigation lights, thanks to its advantages in high-speed data transmission and precise light intensity control. This technology enables real-time information transmission through the flashing of lights. However, VLC still faces significant challenges in practical applications. On the one hand, severe weather conditions such as fog, rain, and snow can cause severe attenuation of optical signals, and ambient light interference can also cause signal distortion, directly affecting the accuracy of navigation light power regulation and causing power prediction models to deviate from actual requirements. On the other hand, navigation lights may experience abnormal brightness due to issues such as light source aging, component failure, or insufficient power. However, traditional control methods make it difficult to detect faults in a timely manner, resulting in signal blind spots for ship navigation and posing a potential threat to navigation safety. Therefore, a solution that integrates environmental perception, real-time feedback, and intelligent control is urgently needed to improve the reliability and safety of navigation lights in complex scenarios. Summary of the Invention

[0004] In response to the defects in the prior art, the present invention provides a method and system for intelligent control of navigation lights based on visible light communication, which solves the problem in the prior art that navigation lights are difficult to automatically adjust according to environmental information.

[0005] In order to achieve the above-mentioned purpose, one aspect of the present invention provides a method for intelligent control of navigation lights based on visible light communication, the method comprising: obtaining the weather type, weather type level, ambient light brightness, visibility and wind speed of the target area; monitoring the target area based on the ambient light brightness to obtain the target ship; constructing a navigation light power prediction model, and using the navigation light power prediction model to predict the navigation light power according to the weather type, the weather type level, the ambient light brightness, the visibility and the wind speed; obtaining a visible light communication protocol, and controlling the navigation light to flash at the navigation light power based on the visible light communication protocol and the target ship.

[0006] The present invention acquires multi-dimensional environmental parameters such as weather type, level, and ambient light brightness in the target area in real time, and combines it with a beacon light power prediction model to accurately predict the power value that adapts to the current environment, change the traditional fixed power output mode, solve the problems of insufficient signal coverage and energy waste in bad weather, improve the adaptability of beacon lights in complex environments, and ensure the stability of ship navigation signals. Based on the monitoring of target ships based on ambient light brightness, the beacon lights can be started under appropriate conditions to achieve on-demand operation and improve energy utilization efficiency. The use of visible light communication protocols to control the flashing of beacon lights can not only emit light at predicted power, but also convert the coordinates of dangerous areas into flashing frequencies, convey more information to ships, and enhance navigation safety and accuracy. By combining multi-dimensional environmental parameters with intelligent models, a closed-loop system from environmental perception, power prediction to precise control is formed, which improves the intelligence of beacon light control.

[0007] Optionally, the constructing of the beacon light power prediction model includes: obtaining the rated power of the beacon light, and using the rated power to set a power penalty term; setting an initial loss function with a mean square error, and adding the power penalty term to the initial loss function to obtain an optimized loss function; and constructing the beacon light power prediction model based on the optimized loss function; By obtaining the rated power of the beacon light and setting a power penalty term, the present invention can effectively limit the predicted power from exceeding the safety range of the equipment, avoid device damage or excessive energy consumption due to excessive power, and ensure the safety of the beacon light operation from the model level. The initial loss function is set with the mean square error and the power penalty term is incorporated into the optimized loss function. The deviation between the predicted power and the actual power is measured by the mean square error to ensure the model prediction accuracy, and the power penalty term is used to implement the constraint on the power upper limit, so that the model can strike a balance between accurate prediction and safe operation. The model constructed based on the optimized loss function can fully learn the nonlinear relationship between multi-dimensional parameters such as weather type and ambient light brightness and power demand, thereby improving the performance of the beacon light power prediction model.

[0008] Optionally, controlling the navigation light to flash at the navigation light power based on the visible light communication protocol and the target ship includes: obtaining the coordinates of the dangerous area, and converting the coordinates of the dangerous area into the navigation light flashing frequency according to the visible light communication protocol; and based on the target ship, controlling the navigation light to flash according to the navigation light power and the navigation light flashing frequency.

[0009] The present invention converts the coordinates of dangerous areas into the flashing frequency of navigation lights through the visible light communication protocol, and combines it with predicted power to control the operation of navigation lights, thereby improving the visual transmission of navigation information. Ships can obtain the location of dangerous areas by identifying the flashing frequency, thereby improving navigation safety.

[0010] Optionally, the intelligent control method of the beacon light based on visible light communication also includes: monitoring the performance of the beacon light according to the flicker and the beacon light power to obtain a monitoring result; adjusting the beacon light power based on the monitoring result to obtain an adjusted power; and updating and training the beacon light power prediction model using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility and the wind speed.

[0011] The present invention forms a closed-loop optimization mechanism through real-time monitoring of navigation light performance and dynamic model updating. By collecting the actual brightness of flashing lights and comparing them with theoretical values, it can accurately judge abnormalities such as light source aging and device failure, avoid navigation blind spots caused by brightness attenuation, and improve navigation safety. Based on the monitoring results, the power is dynamically adjusted, and the environmental requirements can be adapted in real time according to the ship's response time. By adjusting the power and updating the training model with environmental parameters, the model can continuously learn the deviation characteristics in actual operation and gradually improve the prediction accuracy.

[0012] Optionally, the monitoring of the performance of the beacon light according to the flicker and the beacon light power to obtain a monitoring result includes: obtaining the actual brightness of the light according to the flicker; constructing a beacon light brightness prediction model, inputting the ambient light brightness and the adjusted power into the beacon light brightness prediction model to obtain the theoretical brightness of the light; and comparing the actual brightness of the light with the theoretical brightness of the light to obtain a monitoring result.

[0013] This invention establishes a precise navigation light status assessment system by comparing and analyzing measured brightness with theoretical brightness. By collecting the measured brightness in real time, the current luminous intensity of the navigation light can be intuitively reflected. A brightness prediction model is constructed using ambient light brightness and adjusted power to scientifically deduce the theoretical brightness value. Comparing the two allows rapid identification of anomalies such as light source aging and component failure, thus improving the automatic detection capability of navigation marks and, in turn, enhancing the safety of ship navigation.

[0014] Optionally, constructing a beacon light brightness prediction model includes: obtaining historical ambient light brightness, historical beacon light power and historical measured brightness of beacon lights based on lighting tests on multiple beacon lights; constructing training samples using the historical ambient light brightness, the historical beacon light power and the historical measured brightness of beacon lights; constructing a beacon light brightness prediction model, and training the beacon light brightness prediction model using the training samples.

[0015] This method obtains historical data through multiple beacon lighting tests, ensuring the richness and reliability of training samples. Constructing training samples based on historical ambient light levels, beacon light power, and measured brightness accurately captures the mapping relationship between environmental parameters and brightness output, making the model more tailored to actual operating scenarios. After sample training, the constructed model can accurately predict theoretical brightness values ​​based on real-time ambient light levels and adjusted power, providing a scientific benchmark for beacon light performance monitoring and improving the accuracy of brightness predictions.

[0016] Optionally, adjusting the power of the beacon light based on the monitoring result to obtain the adjusted power includes: obtaining the time length for the target ship to respond to the flashing based on the detection result; adjusting the horizontal scattering angle and the vertical scattering angle of the beacon light according to the time length to obtain the adjusted scattering angle; based on the adjusted scattering angle, adjusting the power of the beacon light using the time length and the power threshold to obtain the adjusted power.

[0017] By obtaining the target ship's response time to the flashing signal, the signal effectiveness at the current power can be intuitively reflected. If the response time is too long, it means that the signal strength is insufficient. At this time, first adjust the horizontal and vertical scattering angles of the navigation light. If the adjustment fails, increase the power with the power threshold as the upper limit until the ship accurately receives the signal. This can avoid navigation blind spots caused by insufficient power and prevent excessive power from damaging equipment through threshold limitations, thereby improving the scientific nature of power adjustment.

[0018] Optionally, the updating and training of the beacon light power prediction model using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility and the wind speed includes: constructing a real-time feedback reinforcement term using the difference between the beacon light power and the adjusted power; adding the real-time feedback reinforcement term to the optimization loss function to obtain an enhanced loss function; constructing an enhanced sample set using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility and the wind speed; setting sample weights for the enhanced sample set based on the difference; and updating and training the beacon light power prediction model based on the enhanced loss function using the enhanced sample set and sample weights.

[0019] By using the difference between the adjusted power and the predicted power to construct an enhancement term and add a loss function, the actual operation deviation can be converted into a model optimization signal, prompting the model to focus on correcting the power prediction error; based on the multi-dimensional environmental parameters and the adjusted power, an enhanced sample set is constructed, and the sample weight is set according to the difference, so that samples with large deviations obtain a higher training priority, realizing targeted optimization of power regulation in complex scenarios, breaking the traditional static training mode, and improving the adaptability and practical application capabilities of the beacon light power prediction model through a closed-loop process of error feedback-weight allocation-model iteration.

[0020] Optionally, setting a sample weight for the enhanced sample set based on the difference includes: calculating a sum of differences of the enhanced sample set according to the difference; and setting a ratio of the difference to the sum of differences as the sample weight.

[0021] The present invention calculates the sum of differences by difference and sets sample weights by ratio, which can dynamically give higher weights to samples with large deviations, focus model training on scenarios with significant power prediction errors, enhance the ability to learn power regulation deviations in complex environments, and further improve the performance of the beacon light power prediction model.

[0022] Another aspect of the present invention provides a navigation light intelligent control system based on visible light communication, comprising: a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute a navigation light intelligent control method based on visible light communication as described in any one of the previous aspects of the present invention.

[0023] The present invention provides an intelligent control system for beacon lights based on visible light communication, which has a compact structure, stable performance, high integration and simple composition. It can stably execute an intelligent control method for beacon lights based on visible light communication provided in the previous aspect of the present invention, further improving the overall applicability and practical application capabilities of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of a method for intelligent control of navigation lights based on visible light communication according to an embodiment of the present invention; Figure 2 The figure is a schematic structural diagram of an intelligent control system for navigation lights based on visible light communication according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] Specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the present invention. In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, it will be apparent to one of ordinary skill in the art that these specific details are not necessarily required to practice the present invention. In other instances, well-known circuits, software, or methods are not specifically described to avoid obscuring the present invention.

[0026] Throughout this specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present invention. Therefore, appearances of the phrases "in one embodiment," "in an embodiment," "an example," or "an example" in various places throughout this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics may be combined in any suitable combinations and / or subcombinations in one or more embodiments or examples. Furthermore, those of ordinary skill in the art will appreciate that the figures provided herein are for illustrative purposes only and are not necessarily drawn to scale.

[0027] See Figure 1 In an alternative embodiment, as Figure 1 The method for intelligently controlling a navigation light based on visible light communication includes the following steps: Step S1, obtaining the weather type, weather type level, ambient light brightness, visibility and wind speed of the target area.

[0028] In this embodiment, a network of meteorological sensors deployed in the target area (such as a multi-factor weather station) collects real-time atmospheric data. This data is then combined with infrared remote sensing, humidity, and pressure sensors to identify the target area's weather type. The system pre-configures typical weather scenarios, including sunny, cloudy, rainy, foggy, and snowy. Finally, a numerical mapping is used to map sunny, cloudy, rainy, foggy, and snowy days to numbers 1-5, respectively.

[0029] Weather type classification compares sensor-collected data such as temperature and humidity, air pressure change rate, and precipitation particle characteristics with a database model. Using a pattern recognition algorithm, the system automatically determines the current weather type. For example, when precipitation intensity exceeds 0.1 mm / h and air humidity exceeds 90%, it is considered rainy; when horizontal visibility is less than 1 km and air humidity exceeds 95%, it is considered foggy. Identified weather types are classified based on meteorological industry standards, such as the China Meteorological Administration's "Weather Phenomenon Classification Standards." Differentiated classification indicators are set for different weather types: rainy days are classified based on 24-hour rainfall into light rain (less than 10 mm), moderate rain (10 mm-25 mm), and heavy rain (25 mm-50 mm). Fog days are classified based on visibility into light fog (1 km-10 km), heavy fog (0.5 km-1 km), and dense fog (greater than 0.5 km). Snow levels can be divided according to the cumulative snowfall in 24 hours, with light snow ranging from 0.1mm-2.4mm, moderate snow ranging from 2.5mm-4.9mm, heavy snow ranging from 5.0mm-9.9mm, blizzard ranging from 10.0mm-19.9mm, severe blizzard ranging from 20.0mm-29.9mm, and extremely severe blizzard ranging from more than 29.9mm. Finally, the weather type levels corresponding to rain, fog and snow weather are also converted into corresponding numerical types. There is no need to set the weather type level to 0 in sunny and cloudy scenes.

[0030] High-precision light sensors installed around navigation lights monitor the ambient light intensity in the target area in real time. The sensors collect light signals with a sampling period of 100ms and convert them into digital signals through an analog-to-digital converter. If higher-precision ambient light intensity is required, a Kalman filter algorithm can be used to eliminate transient interference from strong light (such as ship lights or lightning) and output a stable ambient light brightness value.

[0031] When using the Kalman filter algorithm to eliminate instantaneous strong light interference, we first establish a filtering model containing a state equation and an observation equation based on the light signal collected by the light sensor within a 100ms sampling period, predict the ambient light brightness state at the current moment, then compare the measured light intensity signal with the predicted value, calculate the error covariance and update the state estimate, and continuously optimize the prediction results through recursive iteration, thereby effectively filtering out instantaneous strong light interference such as ship lights and lightning, and outputting a stable ambient light brightness value.

[0032] By deploying transmissive or scattering visibility meters, the atmospheric transparency of the target area is monitored in real time based on the principle of optical attenuation. The visibility meter emits an infrared light signal from the transmitter to the receiver, and the visibility distance is calculated based on the degree of light scattering and absorption by aerosol particles in the atmosphere.

[0033] A three-cup wind speed sensor or ultrasonic anemometer is used and installed at a suitable location in the target area (avoiding obstructions) to collect horizontal wind speed data in real time. The sensor records the wind cup rotation frequency or ultrasonic time difference by pulse counting and converts it into wind speed value through the signal processing module.

[0034] Step S2: monitoring the target area based on the ambient light brightness to obtain a target ship.

[0035] In this embodiment, an ambient light brightness threshold is pre-set, and the threshold is used to determine the working conditions of the navigation light. If the ambient light brightness is high, the navigation light is not needed to assist the ship. The ambient light brightness threshold is determined based on the actual working environment of the target area. The critical value of the ambient light brightness for safe navigation of a ship is determined by expert experience, which is the ambient light brightness threshold.

[0036] When the ambient light brightness is lower than the ambient light brightness threshold, the ship monitoring equipment deployed in the target area is activated. For example, the radar transmits and receives electromagnetic wave signals, and determines whether there is a ship target based on the time, frequency and other information of the echo. The photoelectric sensor uses the principle of light reflection to capture the changes in light signals caused by ship occlusion and reflection in the target area. The video surveillance system can also be used in combination with the image recognition algorithm to analyze the collected video frames, identify the outline and motion characteristics of the ship, and conduct all-round and multi-dimensional scanning and monitoring of the target area. After data processing and feature extraction, the target that meets the characteristics of the ship is screened out, thereby accurately identifying the target ship.

[0037] Step S3: constructing a navigation light power prediction model, and using the navigation light power prediction model to predict the navigation light power according to the weather type, the weather type level, the ambient light brightness, the visibility and the wind speed.

[0038] The construction of the navigation light power prediction model specifically includes the following sub-steps: Step S301: Acquire the rated power of the navigation light, and use the rated power to set a power penalty item.

[0039] In this embodiment, the power penalty term satisfies the following formula: , in, is the power penalty term, is the sample size, For the The prediction power of samples, is the rated power of the navigation light.

[0040] In the above formula, It represents the difference between the predicted power and the rated power. Clearly punish only the predicted behavior of exceeding the rated power of the beacon light, Indicates converting the absolute power difference into a relative proportion to avoid imbalance in penalty intensity due to different rated powers. Squaring is used for nonlinear penalty. The larger the excess ratio, the squared increase in penalty intensity. This forces the model to avoid large excesses. The penalties for all samples are averaged to avoid a single extreme sample dominating the loss function.

[0041] Step S302 : setting an initial loss function with a mean square error, and adding the power penalty term to the initial loss function to obtain an optimized loss function.

[0042] In this embodiment, the loss function satisfies the following formula: , The optimization loss function satisfies the following formula: , , in, is the loss function, is the sample size, For the The predicted power value of samples, For the The actual power value of the samples, To optimize the loss function, is the power penalty term.

[0043] In the above formula, the optimization loss function combines the initial loss function and the power penalty term. The initial loss function is used to measure the error between the predicted power and the actual power to ensure the accuracy of the model's power prediction. The power penalty term is also used to limit the situation where the predicted power exceeds the rated power, thereby preventing the beacon light from being damaged due to excessive power or excessive energy consumption. While pursuing accurate predictions, the model also takes into account the power constraints of the actual operation of the beacon light, thereby improving the practicality of the model and the reliability of the beacon light control, so that the trained model can better adapt to the beacon light power prediction and control scenarios.

[0044] Step S303, constructing a navigation light power prediction model based on the optimized loss function; In this embodiment, it is first necessary to obtain sample data including historical navigation light power, historical weather type, historical weather type level, historical ambient light brightness, historical visibility and historical wind speed.

[0045] Flashing feedback tests are conducted in the target area under various weather conditions and at various time periods. The operating range of the navigation light is the range of the target area. The ship arrives at various critical positions within the target area in advance, and the optical signal receiver on the ship is on standby, ready to receive the optical signal of the navigation light at any time.

[0046] First, the beacon light needs to set the horizontal and vertical scattering angles according to the size of the target area, and there is room for contraction in the horizontal and vertical scattering angles. Then, ensure that the target ship is in the middle of the beacon light signal coverage. The power of the beacon light is gradually increased from low to high. After receiving the light signal of the beacon light, the ship sends a radio signal, indicating that the ship confirms that it has received the light signal of the beacon light. After receiving the radio signal, the beacon light stops increasing the power. At this time, the final power, as well as the weather type, weather type level, ambient light brightness, visibility and wind speed at that time are collected and stored. Finally, sample data of historical beacon light power, historical weather type, historical weather type level, historical ambient light brightness, historical visibility and historical wind speed are formed.

[0047] A navigation light power prediction model was constructed based on a deep neural network. The sample data includes multi-dimensional information such as weather type, weather level, light brightness, visibility, and wind speed. The multi-layer nonlinear transformation capabilities of the deep neural network effectively handle the coupling of different types of features. The complex nonlinear relationship between navigation light power and environmental parameters, such as the exponential relationship between visibility and power demand in foggy weather, can be learned by the hidden layer of the deep neural network through activation functions.

[0048] In the input layer of the deep neural network, the weather type, weather level, brightness, visibility, and wind speed features are normalized and used as input. A 2-3-layer fully connected layer is used, and the number of neurons in each layer can be set to 1.5 times to 2 times the feature dimension. A single neuron linear output is used to predict the beacon light power value. The activation function can be a linear function. Finally, sample data is used for training to obtain the beacon light power prediction model.

[0049] The weather type, weather type level, ambient light brightness, visibility and wind speed are input into the trained beacon light power prediction model to obtain the predicted beacon light power.

[0050] Step S4: obtaining a visible light communication protocol, and controlling the navigation light to flash at the navigation light power based on the visible light communication protocol and the target ship.

[0051] The step of controlling the navigation light to flash at the navigation light power based on the visible light communication protocol and the target ship specifically includes the following sub-steps: Step S401: Acquire the coordinates of a dangerous area, and convert the coordinates of the dangerous area into a flashing frequency of a navigation light according to the visible light communication protocol.

[0052] In this embodiment, the visible light communication protocol is a set of rules and standards for transmitting information using visible light. According to the visible light communication protocol, the coordinates of the hazardous area are first encoded and converted into a specific digital signal. Then, according to the pre-defined mapping rules in the protocol, the encoded hazardous area coordinate digital signal is mapped to the flashing frequency of the navigation light. For example, different coordinate intervals correspond to different flashing frequency ranges. Finally, a flashing frequency command corresponding to the hazardous area coordinates is generated to control the navigation light to flash at that frequency, thereby transmitting the hazardous area's location information to the ship.

[0053] Step S402: Based on the target ship, the navigation light is controlled to flash according to the navigation light power and the navigation light flashing frequency.

[0054] In this embodiment, the horizontal azimuth of the ship relative to the beacon light is calculated through the real-time positioning data of the target ship (such as radar echo), and the preset horizontal scattering angle and vertical scattering angle are used as fixed coverage ranges, so that the real-time position of the ship is always in the geometric center area of ​​the horizontal scattering angle and the vertical scattering angle. After obtaining the coordinates of the dangerous area and converting them into the beacon light flashing frequency according to the visible light communication protocol, the beacon light is flashed in combination with the power value output by the beacon light power prediction model. The working power of the beacon light must be ensured to be within the rated power range to avoid equipment damage or excessive energy consumption due to excessive power. At the same time, different frequency combinations of the flashing frequencies corresponding to the dangerous area coordinates represent different dangerous area information. Finally, the beacon light is controlled to flash according to the determined beacon light power and flashing frequency, which reflects the real-time nature of the danger warning of the present invention.

[0055] Step S5: monitoring the performance of the navigation light according to the flicker and the power of the navigation light to obtain a monitoring result.

[0056] The step of monitoring the performance of the beacon light according to the flicker and the power of the beacon light to obtain a monitoring result specifically includes the following sub-steps: Step S501: obtaining the actual brightness of the light according to the flicker.

[0057] In this embodiment, high-precision light sensors are rationally arranged around the beacon lights to collect light signals emitted when the beacon lights flash in real time with a fixed sampling period. The sensors convert the light signals into electrical signals, which are then converted into digital signals through an analog-to-digital conversion module. Finally, the processed digital signals are converted into the actual measured brightness value of the light according to the calibration parameters and conversion formula of the sensors.

[0058] Step S502: constructing a beacon light brightness prediction model, inputting the ambient light brightness and the adjustment power into the beacon light brightness prediction model to obtain the light theoretical brightness.

[0059] The construction of the beacon light brightness prediction model specifically includes the following sub-steps: Step S50201, based on the illumination test of multiple navigation lights, obtain the historical ambient light brightness, historical navigation light power and historical navigation light measured brightness.

[0060] In this example, several navigation lights of the same model and in good condition were selected and tested under varying ambient light conditions. High-precision light sensors installed around the lights collected ambient light data in real time with a sampling period of 100ms. The lights were then controlled to operate at varying power levels, and the same type of light sensors were used to measure the actual brightness of the lights as they flashed. During the test, the ambient light brightness corresponding to each power level was recorded simultaneously with the measured brightness of the lights. Anomalous data caused by sensor failure or sudden strong light interference (such as ship lights or lightning) was eliminated.

[0061] Step S50202: construct a training sample using the historical ambient light brightness, the historical navigation light power, and the historical navigation light measured brightness.

[0062] In this embodiment, the historical ambient light brightness, historical beacon light power and historical beacon light measured brightness are normalized to ensure the consistency and reliability of the data. The processed data are then combined, and the historical ambient light brightness and historical beacon light power are marked as input, and the historical beacon light measured brightness is marked as output, finally forming a training sample.

[0063] Step S50203: construct a beacon light brightness prediction model, and use the training samples to train the beacon light brightness prediction model.

[0064] In this example, a deep neural network architecture is used to construct a beacon light brightness prediction model. Feature data, such as historical ambient light levels and historical beacon light power levels, is normalized and used as input. Nonlinear features are extracted through two or three fully connected hidden layers. Reluctant Unit (ReLU) activation functions are used to enhance the model's expressiveness. A linear activation function is used in the output layer to predict the measured beacon light brightness. Training is performed using a mean squared error loss function, the Adam optimizer is used to iteratively update parameters, and an early stopping strategy is introduced to prevent overfitting.

[0065] The training samples were divided into a training set and a validation set in an 8:2 ratio and fed into a deep neural network model. During training, the mean squared error (MSE) between the predicted and actual brightness values ​​was calculated through forward propagation. Combined with the loss function, the weight parameters were iteratively updated using the Adam optimizer with a learning rate of 0.001. After each round of training, model performance was evaluated on the validation set. An early stopping strategy was triggered when the validation set loss stopped decreasing for five consecutive rounds to prevent overfitting. This ultimately resulted in a trained beacon light brightness prediction model.

[0066] When inputting ambient light brightness and adjusted power into the beacon light brightness prediction model, these two data items are first normalized and preprocessed to ensure their numerical range matches the model input requirements. The data then enters the trained beacon light brightness prediction model, with ambient light brightness and adjusted power serving as feature vectors in the input layer. This data is processed through two to three fully connected hidden layers, each of which uses a ReLU activation function to extract nonlinear features of the effects of ambient light and power on brightness, such as the attenuation relationship between beacon light power and actual brightness in strong light environments. Finally, the output layer generates a theoretical light brightness value through a linear transformation. This value, based on the historical data mapping relationship learned by the model, accurately deduces the theoretical luminous intensity under the current parameter combination, providing a scientific benchmark for comparison with actual measured brightness.

[0067] Step S503: Compare the measured light brightness with the theoretical light brightness to obtain a monitoring result.

[0068] In this embodiment, a validation set is obtained, and the mean and standard deviation of the prediction error of the beacon light brightness prediction model are statistically calculated based on the validation set; the brightness error ratio of the beacon light is obtained, and the performance fluctuation threshold is calculated based on the brightness error ratio, the mean and the standard deviation; the difference between the measured brightness of the light and the theoretical brightness of the light is compared with the performance fluctuation threshold, and the performance of the beacon light is determined based on the comparison result.

[0069] First, an independent validation set is created from historical data. For each sample in the validation set, the prediction error of the navigation light brightness prediction model is calculated—that is, the absolute difference between the predicted brightness and the measured brightness. The mean and standard deviation of this error series are then calculated. Next, the brightness error ratio of the navigation light is determined, for example, ±10% of the rated brightness, as referenced by industry standards. The performance fluctuation threshold is calculated, and the difference between the measured and theoretical brightness is compared to the performance fluctuation threshold (total tolerable brightness error). A large difference indicates that the navigation light is aging or malfunctioning and should be reported for repair. A small difference indicates that the performance of the navigation light is normal and can be used normally.

[0070] The performance fluctuation threshold satisfies the following formula: , in, is the performance fluctuation threshold, is the mean, is the standard deviation, is the rated brightness value, is the brightness error ratio, The brightness attenuation value allowed for the navigation light.

[0071] The above formula It refers to the use of the statistical properties of the normal distribution to cover the 99.7% confidence interval of the model prediction error, that is, the upper limit of the vast majority of normal prediction deviations. Quantify the inherent brightness error of navigation light hardware, including the deviation from rated brightness caused by production tolerance and power supply fluctuation. Due to the limitations of the test, the training samples cannot fully cover the rated brightness deviation. Reserve space for brightness attenuation during long-term operation of the equipment, including the allowable variation due to slow-changing factors such as light source aging, dust accumulation, and occlusion.

[0072] Step S6: adjusting the power of the navigation light based on the monitoring result to obtain an adjusted power.

[0073] The step of adjusting the power of the navigation light based on the monitoring result to obtain the adjusted power specifically includes the following sub-steps: Step S601: obtaining the time length for the target ship to respond to the flashing based on the detection result.

[0074] In this embodiment, after determining that there is no problem with the performance of the navigation light, the response of the ship to the flashing is waited (described in detail in step S303 ), and the time length of the response is obtained.

[0075] Step S602: adjusting the horizontal scattering angle and the vertical scattering angle of the navigation light according to the time length to obtain an adjusted scattering angle.

[0076] In this embodiment, first, based on corresponding tests of the beacon light and the ship, the minimum horizontal scattering angle and the minimum vertical scattering angle of the beacon light (the angle size at which the light signal can be normally received) are set, and the response time of the target ship to the flashing signal is compared with a preset time threshold (such as 1 minute, which depends on the water environment). If the response time is too long, it indicates that the signal strength is insufficient. The horizontal scattering angle and the vertical scattering angle are adjusted according to a certain ratio (such as the horizontal scattering angle and the vertical scattering angle are reduced by 1% for every second that the time threshold is exceeded). When one of the horizontal scattering angle and the vertical scattering angle reaches the minimum horizontal scattering angle or the minimum vertical scattering angle, the adjustment is stopped.

[0077] Step S603: Based on the adjusted scattering angle, the power of the navigation light is adjusted using the time length and the power threshold to obtain an adjusted power.

[0078] In this embodiment, after adjusting the horizontal and vertical scattering angles to no avail, the power of the navigation light is adjusted. First, a power threshold is set based on the light's rated power, equipment safety standards, and actual environmental requirements. This threshold must be within the power range required for safe operation of the light. Next, the target ship's response time to the flashing signal is compared with a pre-set time threshold. If the response time is too long, indicating insufficient signal strength, the current power is increased proportionally (e.g., a 1% power increase for every second the time threshold is exceeded) using the power threshold as the upper limit until feedback from the ship is received. At this point, the light's power is adjusted. If the response time is within a reasonable range, the current power is maintained. During the adjustment process, the adjusted power is ensured to remain within the power threshold range to balance signal effectiveness and equipment safety.

[0079] Step S7: updating and training the navigation light power prediction model using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility, and the wind speed.

[0080] The updating and training of the navigation light power prediction model using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility, and the wind speed specifically includes the following sub-steps: Step S701: construct a real-time feedback reinforcement item using the difference between the navigation light power and the adjustment power.

[0081] In this example, the difference between the beacon light power (predicted by the model) and the adjusted power is calculated. This difference is then squared to amplify the effect of the deviation. Samples are then weighted according to their importance. Finally, the weighted squared differences across all samples are summed and averaged to create a real-time feedback enhancement term. This enhancement term quantifies the accuracy of power adjustment and provides real-time feedback for optimizing the beacon light power prediction model. This allows the model to more accurately adjust the beacon light power under different environmental conditions, ensuring the effectiveness of the beacon light signal and energy conservation requirements.

[0082] The real-time feedback reinforcement item satisfies the following formula: , in, For time feedback reinforcement, To increase the number of sample sets, To strengthen the sample set The weight of the samples, To strengthen the sample set The adjusted power value of samples, To strengthen the sample set The predicted power value of each sample.

[0083] The above formula quantifies the degree of deviation between the actual adjusted power of the beacon light and the predicted power in a weighted average manner. By squaredly summing the power difference of each sample in the enhanced sample set according to the sample weight and then taking the average, the impact of the deviation is amplified, the contribution of samples of different importance is highlighted, and the degree of fit between the actual power control value and the predicted value is reflected. This provides quantitative feedback for dynamically optimizing the power of the beacon light and adapting to environmental changes (such as weather and visibility), helping to improve the accuracy and stability of the power adjustment of the beacon light.

[0084] Step S702: Add the real-time feedback enhancement term to the optimization loss function to obtain an enhanced loss function.

[0085] The reinforcement loss function satisfies the following formula: , in, To strengthen the loss function, is the initial loss function, is the power penalty term, It is a timely feedback reinforcement item.

[0086] The above formula integrates the initial loss function, power penalty term, and real-time feedback enhancement term. It constrains the optimization direction of the model from multiple dimensions. The initial loss ensures the achievement of the basic goal, the power loss focuses on the rationality of the power regulation of the beacon light, and the real-time feedback enhancement term introduces the deviation feedback between the actual power adjustment and the prediction. This allows the model to follow the basic rules and adapt to the dynamic power regulation requirements during optimization. It can also make precise corrections based on real-time feedback, improve the accuracy and adaptability of the prediction and adjustment of the beacon light power, and help the beacon light operate stably and efficiently in complex environments. Step S703 : constructing an enhanced sample set using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility, and the wind speed.

[0087] The acquired data such as adjustment power, weather type, weather type level, ambient light brightness, visibility and wind speed are pre-processed to ensure the accuracy and consistency of the data. Among them, weather type and weather type level need to be standardized and coded according to the preset standards, and numerical data such as ambient light brightness, visibility and wind speed are normalized. Then, taking each independent monitoring cycle or event as a unit, the adjustment power within the same cycle is combined with the corresponding weather type, weather type level, ambient light brightness, visibility and wind speed into a sample. Each sample needs to clearly define the correspondence between each characteristic variable and the adjustment power. Finally, all samples that meet the requirements are sequentially integrated to form an enhanced sample set containing multi-dimensional environmental characteristics and corresponding adjustment power, providing structured data support for the subsequent update training of the beacon light power prediction model.

[0088] Step S704: setting sample weights for the enhanced sample set based on the difference.

[0089] Wherein, setting the sample weight for the enhanced sample set based on the difference specifically includes the following sub-steps: Step S70401: Calculate the sum of the differences of the enhanced sample set according to the differences.

[0090] In this embodiment, the differences of all samples are accumulated to obtain the sum of the differences. During the calculation process, the accuracy of each sample difference must be ensured to avoid calculation errors. By summing all sample differences, the overall deviation between the power adjustment and prediction of the enhanced sample set is fully reflected, providing reliable data support for subsequent analysis or model optimization based on the sum of differences.

[0091] Step S70402: Set the ratio of the difference to the sum of the differences as the sample weight.

[0092] The sample weight satisfies the following formula: , in, To strengthen the sample set The sample weight of the sample, To strengthen the sample set The adjusted power of samples, To strengthen the sample set The prediction power of samples, To increase the number of sample sets, To strengthen the sample set The adjusted power of samples, To strengthen the sample set The prediction power of samples.

[0093] In the above formula, the sample weights are dynamically allocated based on the difference between the adjusted power and the predicted power of each sample in the enhanced sample set. This factor reflects the degree of deviation between the actual power adjustment and prediction for a single sample. The numerator highlights the impact of the sample's own deviation, while the denominator sums the deviations of all samples, normalizing the weights to a reasonable value. This setting assigns higher weights and greater attention to samples with significant deviations during subsequent model optimization. This allows the model to focus on samples with significant deviations between power prediction and adjustment, enabling targeted optimization and improving the accuracy of navigation light power control. This meets the need for differentiated treatment of samples with varying deviations in real-world scenarios.

[0094] Step S706 : Based on the enhanced loss function, the enhanced sample set and sample weights are used to update and train the navigation light power prediction model.

[0095] In this embodiment, based on an enhanced loss function, a sample set containing adjusted power, weather type, weather type level, ambient light level, visibility, and wind speed is fed into the navigation light power prediction model, along with sample weights set based on power differences. During training, the model prioritizes samples with large deviations based on the sample weights. A backpropagation algorithm continuously adjusts model parameters, gradually optimizing the model's ability to predict navigation light power while taking into account multiple factors such as prediction error, power constraints, and real-time feedback deviations. This allows for dynamic model training in complex environments, improving prediction accuracy and adaptability.

[0096] like Figure 2 As shown, another intelligent control system for navigation lights based on visible light communication includes: a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute relevant steps of a relevant embodiment of a method for intelligent control of navigation lights based on visible light communication of the present invention.

[0097] The present invention provides a beacon light intelligent control system based on visible light communication. Each functional component can be integrated into a single processing unit, each component can exist physically separately, or two or more components can be integrated into a single unit. These integrated components can be implemented as either hardware or software functions.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A method for intelligent control of navigation lights based on visible light communication, characterized in that: The method comprises: Obtain the weather type, weather type level, ambient light brightness, visibility and wind speed of the target area; Monitoring the target area based on the ambient light brightness to obtain a target ship; Constructing a navigation light power prediction model, and predicting the navigation light power using the navigation light power prediction model according to the weather type, the weather type level, the ambient light brightness, the visibility, and the wind speed; A visible light communication protocol is acquired, and based on the visible light communication protocol and the target ship, the navigation light is controlled to flash at the navigation light power.

2. The method for intelligent control of navigation lights based on visible light communication according to claim 1, characterized in that: The construction of the navigation light power prediction model includes: Obtaining the rated power of the navigation light, and setting a power penalty item using the rated power; Setting an initial loss function with a mean square error, and adding the power penalty term to the initial loss function to obtain an optimized loss function; A navigation light power prediction model is constructed based on the optimized loss function.

3. The method for intelligent control of navigation lights based on visible light communication according to claim 1, characterized in that: The controlling the navigation light to flash at the navigation light power based on the visible light communication protocol and the target ship comprises: Acquire the coordinates of the dangerous area, and convert the coordinates of the dangerous area into a flashing frequency of a navigation light according to the visible light communication protocol; Based on the target ship, the navigation light is controlled to flash according to the navigation light power and the navigation light flashing frequency.

4. The method for intelligent control of navigation lights based on visible light communication according to claim 1, characterized in that: The method for intelligently controlling a navigation light based on visible light communication further includes: Monitoring the performance of the navigation light according to the flickering and the power of the navigation light to obtain a monitoring result; Adjusting the power of the navigation light based on the monitoring result to obtain an adjusted power; The navigation light power prediction model is updated and trained using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility, and the wind speed.

5. The method for intelligent control of navigation lights based on visible light communication according to claim 4, characterized in that: The monitoring of the performance of the beacon light according to the flicker and the power of the beacon light to obtain a monitoring result includes: Obtaining the measured brightness of the light according to the flickering; Constructing a beacon light brightness prediction model, inputting the ambient light brightness and the adjustment power into the beacon light brightness prediction model to obtain the light theoretical brightness; The measured brightness of the light is compared with the theoretical brightness of the light to obtain a monitoring result.

6. The method for intelligent control of navigation lights based on visible light communication according to claim 5, characterized in that: The construction of the beacon light brightness prediction model includes: Based on the illumination test of multiple beacon lights, the historical ambient light brightness, historical beacon light power and historical measured brightness of beacon lights are obtained; Constructing a training sample using the historical ambient light brightness, the historical navigation light power, and the historical navigation light measured brightness; A beacon light brightness prediction model is constructed, and the beacon light brightness prediction model is trained using the training samples.

7. The method for intelligent control of navigation lights based on visible light communication according to claim 4, characterized in that: The adjusting the power of the navigation light based on the monitoring result to obtain the adjusted power includes: Obtaining, based on the detection result, a time length for the target ship to respond to the flashing; Adjusting the horizontal scattering angle and the vertical scattering angle of the navigation light according to the time length to obtain an adjusted scattering angle; Based on the adjusted scattering angle, the power of the navigation light is adjusted using the time length and the power threshold to obtain an adjusted power.

8. The method for intelligent control of navigation lights based on visible light communication according to claim 4, characterized in that: The updating and training of the navigation light power prediction model using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility, and the wind speed includes: A real-time feedback reinforcement item is constructed using the difference between the navigation light power and the adjustment power; Adding the real-time feedback reinforcement term to the optimization loss function to obtain a reinforcement loss function; constructing an enhanced sample set using the adjusted power, the weather type, the weather type level, the ambient light brightness, the visibility, and the wind speed; Setting a sample weight for the enhanced sample set based on the difference; Based on the enhanced loss function, the beacon light power prediction model is updated and trained using the enhanced sample set and sample weights.

9. The method for intelligent control of navigation lights based on visible light communication according to claim 8, characterized in that: The setting of sample weights for the enhanced sample set based on the difference comprises: Calculating a sum of differences of the enhanced sample set according to the differences; The ratio of the difference value to the sum of the differences is set as the sample weight.

10. An intelligent control system for navigation lights based on visible light communication, characterized in that: include: A processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute any one of claims 1 to 9 of the method for intelligent control of navigation lights based on visible light communication.

Images