Highway agglomerate fog prediction method based on multi-source data fusion and electronic equipment

Abstract

The invention provides a road agglomerate fog prediction method based on multi-source data fusion and electronic equipment. The method comprises the following steps: calculating thermal inversion layer structure parameters through a gradient analysis algorithm; calculating dew point temperature sequence data based on the temperature vertical sequence data and the humidity vertical sequence data; on the basis of the humidity vertical sequence data and the dew point temperature sequence data, saturation layer structure parameters are calculated through a saturation criterion algorithm; calculating the overlapping degree of the thermal inversion layer and the saturation layer in the vertical direction; and constructing vector features by using the thermal inversion layer structure parameters, the saturation layer structure parameters, the overlapping degree, the near-surface meteorological data and the underlying surface data, and inputting the vector features into a trained agglomerate fog prediction model to obtain an agglomerate fog occurrence probability. According to the method, through space-air-ground multi-source data fusion, the vertical overlapping degree of a thermal inversion layer and a saturation layer which are closely related to formation of agglomerate fog along a highway is quantified, the formation and dissipation process of the agglomerate fog is reflected, and through deep learning model learning, prediction of the occurrence probability of the agglomerate fog is carried out.

Description

Technical Field

[0001] This application belongs to the technical field of fog monitoring, and particularly relates to a method and electronic device for predicting fog on highways based on multi-source data fusion. Background Technology

[0002] Highway fog poses a significant threat to road traffic safety, and the prediction and detection of fog patches present enormous technical challenges. The occurrence of fog patches is closely related to meteorological factors such as temperature and humidity, wind speed and direction, and air pressure, as well as environmental factors such as underlying surface type and the topography along the highway. Changes in these factors along the highway can lead to the formation of fog patches in new areas or increase the probability of fog patches occurring in local areas. Due to insufficient spatial density of monitoring equipment, monitoring fog patches in these areas presents significant challenges.

[0003] Currently, there are three main technical solutions. The first utilizes advanced sensor technology, primarily high-precision instruments such as next-generation laser visibility meters and miniature weather stations. These sensors can measure key meteorological parameters in the air in real time, such as visibility, temperature, and humidity, thus providing detailed meteorological information and data support for the formation of fog patches. They can detect visibility within a 50m to 4km area, measuring the visibility distribution at various points within the radar's perimeter. Based on the visibility distribution, computer-programmed algorithms determine whether fog patches are forming. However, this system can only monitor fog patches and is unlikely to provide early warnings. With the development of ground-based remote sensing and artificial intelligence technologies, new possibilities have emerged for accurate prediction of fog patches. Ground-based microwave radiometers can continuously monitor the vertical structure of atmospheric temperature and humidity, providing crucial data for understanding the physical mechanisms of fog patch formation.

[0004] However, how to effectively integrate multi-source data and establish reliable prediction models remains a technical challenge that urgently needs to be addressed in this field. Summary of the Invention

[0005] This application provides a method for predicting highway fog based on multi-source data fusion, including the following steps: Using ground-based microwave radiometers deployed along highways, discrete vertical temperature series data of atmospheric temperature variation with altitude and discrete vertical humidity series data of atmospheric humidity variation with altitude are obtained above preset points. Acquire near-surface meteorological data measured by ground meteorological stations; Register underlying surface data using remote sensing satellite monitoring and geographic data along highways; Based on the aforementioned temperature vertical sequence data, the structural parameters of the inversion layer are identified and calculated using a gradient analysis algorithm. Based on the temperature vertical sequence data and humidity vertical sequence data, the dew point temperature sequence data is calculated. Based on humidity vertical sequence data and dew point temperature sequence data, the saturation layer structure parameters are calculated using a saturation criterion algorithm; based on the inversion layer structure parameters and the saturation layer structure parameters, the vertical overlap between the inversion layer and the saturation layer is calculated. The inversion layer structural parameters, saturation layer structural parameters, overlap, near-surface meteorological data, and underlying surface data are used to construct a vector feature. The feature vector is input into the trained fog prediction model to obtain the probability of fog occurrence on the target road section of the highway.

[0006] Furthermore, the near-surface meteorological data includes at least near-surface air temperature, dew point temperature, wind speed, and wind direction.

[0007] Furthermore, the underlying surface data includes at least water bodies, vegetation, bridges, river valleys, and built-up areas.

[0008] Furthermore, the step of identifying and calculating the inversion layer structural parameters through gradient analysis algorithm includes: Scan the temperature vertical sequence data and locate the critical height at which the temperature vertical gradient changes from negative to positive as the bottom height of the inversion layer; Scan upwards from the bottom height of the inversion layer to locate the critical height at which the vertical temperature gradient changes from a positive to a negative value, and take this as the top height of the inversion layer. The temperature difference between the top and bottom of the inversion layer is calculated and used as the inversion layer strength.

[0009] Furthermore, the calculation of the saturated layer structure parameters includes: Scan and compare the temperature vertical sequence data with the dew point temperature sequence data, and locate the height at which the temperature-dew point temperature difference first equals or is less than a preset saturation threshold as the bottom height of the saturation layer. Scan upwards from the bottom height of the saturation layer, and determine the height at which the temperature-dew point temperature difference first exceeds the preset saturation threshold as the top height of the saturation layer; Calculate the height difference between the top and bottom of the saturated layer, and use it as the thickness of the saturated layer.

[0010] Furthermore, the degree of overlap conforms to the following formula: ; Where η is the overlap degree; H ib H is the height of the inversion layer bottom. it H is the height of the inversion layer top. sb H is the height of the saturation layer bottom. st Determine the height of the saturation layer; D s This represents the thickness of the saturated layer.

[0011] Furthermore, the method also includes: Multiplying the overlap by the stability factor yields the fusion feature; The inversion layer structural parameters, saturation layer structural parameters, fusion features, near-surface meteorological data, and underlying surface data are used to construct a vector feature; The formula for calculating the stability factor is: ; Where W is the stability factor; I Strength denoted as the inversion layer intensity; C is a scale constant, ranging from 1.0 to 3.0°C.

[0012] Furthermore, before calculating the structural parameters of the inversion layer, the following steps are also included: The temperature vertical sequence data, dew point temperature sequence data, near-surface meteorological data, and underlying surface data are spatiotemporally registered and unified into the same spatiotemporal coordinate system to form a spatiotemporally unified multi-source dataset.

[0013] Furthermore, the fog prediction model is a deep neural network model, and the training process includes: The feature vector is labeled with a binary fog patch label, which is determined by laser visibility meter data based on a visibility threshold. Initialize the weight parameters of the multi-branch neural network; The weight parameters are iteratively optimized using the backpropagation algorithm with the goal of minimizing the binary cross-entropy loss between the predicted probability and the true label. Once the loss function converges or reaches the predetermined number of training rounds, the final weight parameters are saved to obtain the trained fog prediction model.

[0014] Another aspect of this application proposes an apparatus comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the highway fog prediction method based on multi-source data fusion as described in any of the above technical solutions.

[0015] The above-described technical solution of the present invention has at least the following beneficial technical effects: This invention provides a method for predicting fog patches on highways based on multi-source data fusion. By integrating ground-based microwave radiometer, ground meteorological station and satellite remote sensing data, a comprehensive fog patch monitoring and prediction system is constructed.

[0016] This application uses multi-source data fusion from air, space, and ground to quantify the vertical overlap of the inversion layer and saturation layer, which are closely related to the formation of fog patches along highways, to reflect the formation and dissipation process of fog patches, and uses a deep learning model to learn and predict the probability of fog patch occurrence. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or in the conventional technology, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a highway fog prediction method based on multi-source data fusion in one embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the various embodiments of this application to facilitate a better understanding of the application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the various embodiments below is for ease of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0020] like Figure 1 As shown, this application provides a method for predicting highway fog based on multi-source data fusion, including the following steps: Obtain multi-source data within historical time periods, specifically including: S1. Using ground-based microwave radiometers deployed along highways, discrete vertical temperature sequence data and discrete vertical humidity sequence data of atmospheric temperature variation with altitude are acquired above preset points. Multiple preset points can be identified, which are the locations of the microwave radiometers. Existing microwave radiometers directly measure the microwave radiation intensity of spontaneous emission from the atmosphere at different frequency channels. This radiation intensity can be converted into brightness temperature. Because microwave radiometers have a certain resolution (vertical direction) within their detection range, the temperatures obtained at different altitudes are discrete, not continuous. Existing microwave radiometers have accompanying software that integrates a forward model and an inversion algorithm. The microwave radiometer calculates the sensitivity distribution of atmospheric humidity at different altitude levels for each water vapor detection frequency point by using the brightness temperatures at the 22-30 GHz water vapor absorption channel, wing region frequency point, and window region frequency point. Then, using a final estimation method or a neural network method, iterative optimization is employed to minimize the difference between the simulated and measured brightness temperature values, thereby solving for the atmospheric humidity vertical sequence data.

[0021] Preferably, ground-based microwave radiometers are deployed along representative key road sections, such as areas with a history of frequent fog, bridges with complex terrain that are prone to fog formation, and river valley sections. The microwave radiometers can monitor temperature and relative humidity from the ground to a height of 10 kilometers above it, with a vertical resolution of 50-100 meters.

[0022] S2. Obtain near-surface meteorological data measured by a ground meteorological station; preferably, the near-surface meteorological data records climate data at a certain height (e.g., 1.5 meters) above the ground surface, including at least near-surface air temperature, dew point temperature, wind speed, and wind direction. Simultaneously, visibility measured by a forward-scattering laser visibility meter at the ground meteorological station also needs to be obtained.

[0023] S3. Register underlying surface data using remote sensing satellite monitoring and highway-side geographic data; the underlying surface data should include at least water bodies, vegetation, bridges, river valleys, and high-density building areas. The vegetation and river width data in the highway-side geographic data published by government departments are affected by seasonal climate; therefore, data identification and calibration from remote sensing satellite monitoring are necessary. Optionally, underlying surface data can also be obtained solely from remote sensing satellite monitoring or solely from highway-side geographic data published by government departments.

[0024] Preferably, before calculating the inversion layer structure parameters, the method further includes: spatiotemporally registering the vertical temperature sequence data, vertical humidity sequence data, near-surface meteorological data, and underlying surface data to a unified spatiotemporal coordinate system, forming a spatiotemporally unified multi-source dataset. The underlying surface data is processed using one-heat encoding to convert it into data that the model can recognize and calculate, such as water bodies as 0, vegetation as 1, bridges as 3, river valleys as 4, and high-density building areas as 5, or directly converted into binary code. Through this processing, the subsequent model can automatically discover the complex nonlinear relationship between different underlying surfaces and fog formation during training, reducing reliance on manually preset weights.

[0025] S4. Based on vertical temperature sequence data, the structural parameters of the inversion layer are identified and calculated using a gradient analysis algorithm. The specific process is as follows: Scan the vertical temperature sequence data and locate the critical height at which the vertical temperature gradient changes from negative to positive as the bottom height of the inversion layer. Scan upwards from the bottom of the inversion layer and locate the critical height at which the vertical temperature gradient changes from positive to negative as the top height of the inversion layer. The temperature difference between the top and bottom of the inversion layer is calculated and used as the inversion layer strength.

[0026] S5. Based on temperature and humidity vertical series data, the dew point temperature at each altitude layer is calculated using the Magnus formula recommended by the World Meteorological Organization, thus forming dew point temperature series data. ; in, T d This is the dew point temperature, measured in °C. T Temperature, in °C; RH Humidity (%).

[0027] S6. Based on humidity vertical sequence data and dew point temperature sequence data, the saturation layer structure parameters are calculated using a saturation criterion algorithm. The calculation and determination process is as follows: Scan and compare the vertical temperature sequence data with the dew point temperature sequence data, and locate the height at which the temperature-dew point temperature difference first equals or is less than a preset saturation threshold as the bottom height of the saturation layer; the preset saturation threshold can be 1℃.

[0028] Scan upwards from the bottom of the saturation layer and determine the height at which the temperature-dew point temperature difference first exceeds a preset saturation threshold as the top of the saturation layer. Calculate the height difference between the top and bottom of the saturated layer, and use it as the thickness of the saturated layer.

[0029] S7. Based on the structural parameters of the inversion layer and the saturation layer, calculate the overlap between the inversion layer and the saturation layer in the vertical direction. The degree of overlap conforms to the following formula: ; in, This is used to precisely align the bottom of the inversion layer and the saturation layer, indicating whether the bottom of the inversion layer and the bottom of the saturation layer are at the same height in the vertical direction. H ib and H sb The larger the absolute value of the difference, the more severe the misalignment between the inversion layer and the saturation layer.

[0030] The effective coverage rate is the ratio of the effective coverage thickness to the saturation layer thickness. A ratio approaching 1 indicates that the entire saturation layer is almost completely covered by the inversion layer; a ratio approaching 0 indicates that the inversion layer and the saturation layer do not overlap vertically.

[0031] or Overlap degree; H ib This refers to the height of the bottom of the inversion layer. H it This is the height of the top of the inversion layer; H sb This is the height of the bottom of the saturation layer; H st Determine the height of the saturation layer; D s This represents the thickness of the saturated layer.

[0032] orWhen the value approaches 1, it indicates that saturated air is covered by an inversion layer, and the upward vertical diffusion of low temperature is suppressed, allowing water vapor near the ground to accumulate rapidly and reach supersaturation, thus drastically increasing the probability of fog outbreaks in a short period of time. or When the temperature approaches 0°C, there is no inversion layer covering the saturated air, so water vapor can easily diffuse vertically and does not easily condense into dense fog.

[0033] Preferably, the overlap or Multiplying by the stability factor W yields the fusion feature S: S= or ×W.

[0034] The formula for calculating the stability factor is: ; Where W is the stability factor; I Strength is the inversion layer strength; C is a scale constant, ranging from 1.0 to 3.0℃, preferably 2.0. I Strength Temperatures below 1.0℃ indicate a weak temperature inversion, with generally poor atmospheric stability that is easily disrupted. I Strength The temperature inversion range of 1.0-3.0℃ indicates a moderate temperature inversion, and the atmosphere exhibits good stability. I Strength Temperatures above 3.0℃ constitute a strong inversion, where the atmosphere is very stable and conducive to fog retention. Therefore, option C is chosen within the moderate inversion range. When the inversion intensity reaches a moderate level, the stability factor begins to play a significant role, avoiding oversensitivity to weak inversions and under-response to strong inversions.

[0035] When the inversion layer intensity is 0, W=1, indicating no enhancement effect; as the inversion layer intensity increases, W increases logarithmically, reflecting diminishing marginal effects.

[0036] The overlap between the inversion layer and the saturation layer provides the potential for the fog prediction model to determine the formation of fog, while the stability factor related to the intensity of the inversion layer provides the model with the potential for maintaining fog.

[0037] S8. Construct vector features by combining inversion layer structure parameters, saturation layer structure parameters, overlap (preferably fused features), near-surface meteorological data, and underlying surface data. Before constructing the vector features, the data needs to be preprocessed: interpolate all data to a unified time node spatial matching, and divide the data into road segment units, each road segment unit being 10×10 square meters; associate point data with the nearest road segment unit; and convert all data into a unified tensor format for easy model processing.

[0038] S9. Input the feature vector into the trained fog prediction model to obtain the probability of fog occurrence in the target section of the highway. The training process includes: Binary fog labels are assigned to the feature vectors. The fog labels are determined by the laser visibility meter data based on visibility. For example, the feature vector corresponding to visibility less than 200 meters is a positive sample with a label of 1; the feature vector corresponding to visibility greater than 800 meters is a negative sample with a label of 0; the remaining feature vectors are not labeled.

[0039] Initialize the weight parameters of the multi-branch neural network; The backpropagation algorithm is used to iteratively optimize the weight parameters with the goal of minimizing the binary cross-entropy loss between the predicted probability and the true label. Once the loss function converges or reaches the predetermined number of training rounds, the final weight parameters are saved to obtain the trained fog prediction model.

[0040] Input the vertical sequence data of temperature and humidity at different altitudes detected at the current road segment and time, the surface meteorological data and the corresponding underlying surface data into the trained fog prediction model, and output the prediction results.

[0041] Another aspect of this application proposes an apparatus comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the highway fog prediction method based on multi-source data fusion as described in any of the above technical solutions. Therefore, it possesses all the advantages and beneficial effects of the above technical solutions, which will not be elaborated upon here.

[0042] This application uses multi-source data fusion from air, space, and ground to quantify the vertical overlap of the inversion layer and saturation layer, which are closely related to the formation of fog patches along highways, to reflect the formation and dissipation process of fog patches. It also uses a deep learning model to learn and predict the probability of fog patch occurrence.

[0043] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of this application and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of this application should be included within the protection scope of this application. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A method for predicting highway fog based on multi-source data fusion, characterized in that, Includes the following steps: Using ground-based microwave radiometers deployed along highways, discrete vertical temperature series data of atmospheric temperature variation with altitude and discrete vertical humidity series data of atmospheric humidity variation with altitude are obtained above preset points. Acquire near-surface meteorological data measured by ground meteorological stations; Register underlying surface data using remote sensing satellite monitoring and geographic data along highways; Based on the aforementioned temperature vertical sequence data, the structural parameters of the inversion layer are identified and calculated using a gradient analysis algorithm. Based on the temperature vertical sequence data and humidity vertical sequence data, the dew point temperature sequence data is calculated. Based on humidity vertical sequence data and dew point temperature sequence data, the saturation layer structure parameters are calculated using a saturation criterion algorithm. Based on the structural parameters of the inversion layer and the saturation layer, the overlap between the inversion layer and the saturation layer in the vertical direction is calculated. The inversion layer structural parameters, saturation layer structural parameters, overlap, near-surface meteorological data, and underlying surface data are used to construct a vector feature. The feature vector is input into the trained fog prediction model to obtain the probability of fog occurrence on the target road section of the highway.

2. The method according to claim 1, characterized in that, The near-surface meteorological data includes at least near-surface air temperature, dew point temperature, wind speed, and wind direction.

3. The method according to claim 1, characterized in that, The underlying surface data includes at least water bodies, vegetation, bridges, river valleys, and built-up areas.

4. The method according to claim 1, characterized in that, The process of identifying and calculating the inversion layer structural parameters using a gradient analysis algorithm includes: Scan the temperature vertical sequence data and locate the critical height at which the temperature vertical gradient changes from negative to positive as the bottom height of the inversion layer; Scan upwards from the bottom height of the inversion layer to locate the critical height at which the vertical temperature gradient changes from a positive to a negative value, and take this as the top height of the inversion layer. The temperature difference between the top and bottom of the inversion layer is calculated and used as the inversion layer strength.

5. The method according to claim 4, characterized in that, The calculation of saturated layer structure parameters includes: Scan and compare the temperature vertical sequence data with the dew point temperature sequence data, and locate the height at which the temperature-dew point temperature difference first equals or is less than a preset saturation threshold as the bottom height of the saturation layer. Scan upwards from the bottom height of the saturation layer, and determine the height at which the temperature-dew point temperature difference first exceeds the preset saturation threshold as the top height of the saturation layer; Calculate the height difference between the top and bottom of the saturated layer, and use it as the thickness of the saturated layer.

6. The method according to claim 4, characterized in that, The degree of overlap conforms to the following formula: ； Where η is the overlap degree; H ib H is the height of the inversion layer bottom. it H is the height of the inversion layer top. sb H is the height of the saturation layer bottom. st Determine the height of the saturation layer; D s This represents the thickness of the saturated layer.

7. The method according to claim 6, characterized in that, Also includes: Multiplying the overlap by the stability factor yields the fusion feature; The inversion layer structural parameters, saturation layer structural parameters, fusion features, near-surface meteorological data, and underlying surface data are used to construct a vector feature; The formula for calculating the stability factor is: ； Where W is the stability factor; I Strength denoted as the inversion layer intensity; C is a scale constant, ranging from 1.0 to 3.0°C.

8. The method according to claim 4, characterized in that, Before calculating the structural parameters of the inversion layer, the following steps are also included: The temperature vertical sequence data, humidity vertical sequence data, near-surface meteorological data, and underlying surface data are spatiotemporally registered and unified into the same spatiotemporal coordinate system to form a spatiotemporally unified multi-source dataset.

9. The method according to claim 1, characterized in that, The fog prediction model is a deep neural network model, and the training process includes: The feature vector is labeled with a binary fog patch label, which is determined by laser visibility meter data based on a visibility threshold. Initialize the weight parameters of the multi-branch neural network; The weight parameters are iteratively optimized using the backpropagation algorithm with the goal of minimizing the binary cross-entropy loss between the predicted probability and the true label. Once the loss function converges or reaches the predetermined number of training rounds, the final weight parameters are saved to obtain the trained fog prediction model.

10. A device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the highway fog prediction method based on multi-source data fusion as described in any one of claims 1 to 9 when executing the computer program.

Images