A contact type road surface temperature monitoring system

Abstract

The present invention provides a contact type road surface temperature monitoring system, comprising: a plurality of contact type temperature sensors. The contact type temperature sensor includes a first temperature monitoring module and a second temperature monitoring module. The first temperature monitoring module is arranged at different depths at any position of the snow-covered section, and is used for monitoring the internal temperature and pressure data at different depths of the snow-covered section. The second temperature monitoring module is arranged at a position with a target distance higher than the surface of the snow-covered section, and is used for monitoring the ambient temperature of the snow-covered position; a temperature prediction module, which is used for predicting the temperatures at different depths of multiple points of the snow-covered section according to the internal temperature data, pressure data and ambient temperature at different depths of multiple target positions of the snow-covered section; a predicted temperature display module, which is used for performing image processing on the temperatures at different depths of multiple points of the snow-covered section and displaying the three-dimensional temperature image of the whole snow-covered section. By implementing the present invention, comprehensive and effective information can be provided to relevant personnel so as to judge potential hidden dangers in advance.

Description

Technical Field

[0001] The present invention belongs to the technical field of temperature monitoring, and particularly relates to a contact-type road surface temperature monitoring system. Background Art

[0002] In winter, ice and snow on the road surface are one of the main causes of frequent traffic accidents. Accurately monitoring the road surface temperature is crucial for preventing and coping with the traffic safety risks brought by ice and snow on the road surface. Traditional road surface temperature monitoring methods are mostly point-like monitoring, which is difficult to comprehensively reflect the distribution of road surface temperature. Especially in sections with complex terrain and different road surface materials, the temperature difference is large, and point-like monitoring cannot provide sufficient information to guide road management and maintenance. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a contact-type road surface temperature monitoring system to meet the need of providing comprehensive and effective information for relevant parties to prevent risks in advance.

[0004] To achieve the above purpose, the present invention provides the following technical solutions:

[0005] The present invention provides a contact-type road surface temperature monitoring system, including: a plurality of contact-type temperature sensors arranged at a plurality of target positions in a snow-covered section. Each contact-type temperature sensor includes a first temperature monitoring module and a second temperature monitoring module. The first temperature monitoring module includes a first temperature sensing probe and a pressure monitor, and the first temperature monitoring module is arranged at different depths at any position in the snow-covered section for monitoring the internal temperature and pressure data at different depths in the snow-covered section. The second temperature monitoring module is a second temperature sensing probe arranged at a position at a target distance above the surface of the snow-covered section for monitoring the ambient temperature of the snow-covered position; a temperature prediction module for predicting the temperature at different depths at a plurality of points in the snow-covered section according to the internal temperature data, pressure data, and ambient temperature at different depths at a plurality of target positions in the snow-covered section; and a predicted temperature display module for performing image processing on the temperature at different depths at a plurality of points in the snow-covered section and displaying the three-dimensional temperature image of the entire snow-covered section.

[0006] Optionally, predict the temperatures at different depths at multiple points on the snow-covered section based on the internal temperature data, pressure data, and ambient temperature at different depths at multiple target locations on the snow-covered section, including: dividing the snow-covered section into grids, determining the grids where contact temperature sensors are not installed, and taking the center points in the grids as the points to be predicted; predicting the ambient temperature at the points to be predicted according to the pre-trained first target neural network; determining the geothermal heat of the snow-covered section according to the geothermal measurement method; determining the relationship between the pressure at the points to be predicted and the snow depth based on the positions of the points to be predicted and historical data; inputting the geothermal heat, ambient temperature, and pressures corresponding to different depths into the pre-trained second target neural network to obtain the temperatures corresponding to different snow depths on the road surface represented by the points to be predicted; wherein, the pre-trained second target neural network is trained according to the internal temperature data, pressure data, and ambient temperature at different depths at multiple target locations on the snow-covered section.

[0007] Optionally, predicting the ambient temperature at the points to be predicted according to the pre-trained first target neural network includes: obtaining weather factors and the ambient temperature obtained by contact temperature sensors at multiple target locations; inputting the ambient temperature, weather factors, and their corresponding target locations into the first target neural network; in the first target neural network, using the gradient descent method to adjust the network weights and minimize the loss function until the training is completed to obtain the trained first target neural network; inputting the location information of the points to be predicted and the weather factors into the trained first target neural network to obtain the ambient temperature at the points to be predicted.

[0008] Optionally, the training process of the second target neural network includes: taking the geothermal heat, in-snow temperature, pressure data, and ambient temperature obtained by the same contact temperature sensor as a data group, wherein the depth, ambient temperature, pressure, and geothermal heat in the data group are used as samples, and the internal temperature is used as the sample label; inputting multiple data groups into the second target neural network, and using the gradient descent method to adjust the network weights and minimize the overall loss function until the training is completed to obtain the trained second target neural network, wherein the second target neural network is a physics-informed neural network, and its overall loss function is composed of an error term between the predicted temperature and the actual temperature, a constraint error term for the heat transfer relationship between snow layers, and a constraint error term for the heat transfer relationship in the snow ground.

[0009] Optionally, the overall loss function includes: ;

[0010] wherein, represents the overall loss function, are hyperparameters, representing the weight coefficients of the error term between the predicted temperature and the actual temperature, the weight coefficient of the constraint error term for the heat transfer relationship between snow layers, and the weight coefficient of the constraint error term for the heat transfer relationship in the snow ground, respectively, represents the predicted internal temperature, the actual internal temperature, is the snow density, is the specific heat capacity of snow, is the snow layer temperature, is the snow layer conductivity, z is the depth, and t is the time, represents the geothermal heat, the heat conductivity of the ground surface.

[0011] Optionally, a contact road surface temperature monitoring system further includes: a lidar scanning device for scanning the three-dimensional coordinates of the snow cover on the snow-covered section; performing image processing on the temperatures at different depths of multiple points on the snow-covered section to display the overall three-dimensional temperature image of the snow-covered section, including: obtaining the temperatures at different depths of multiple points and binding the three-dimensional positions at random depths of any point with the temperature to form a data group, and multiple data groups form a data set; clustering multiple data groups in the data set to obtain multiple cluster centers; reading the temperature values and distance values of multiple adjacent cluster centers, and calculating the temperature change rate corresponding to the adjacent cluster centers; determining the temperature identifier at the cluster center according to the temperature of the cluster center; adjusting the change rate of the temperature identifier color according to the temperature identifier at the cluster center and the temperature change rate; performing color rendering on each cluster area in the three-dimensional image of the snow-covered section according to the positions, temperature identifier colors, and temperature identifier color change rates of each cluster center.

[0012] Optionally, reading the temperature values and distance values of multiple adjacent cluster centers and calculating the temperature change rate corresponding to the adjacent cluster centers includes: Step S1, determining the temperature values of multiple points on the line connecting the adjacent cluster centers according to the temperature difference corresponding to the adjacent cluster centers and the distance value between the adjacent cluster centers; Step S2, using the distance between each point and the corresponding cluster center as the radius to obtain each circular line; Step S3, determining the temperatures of all points in the first point set and the temperatures of all points in the second point set corresponding to each circular line, where the first point set consists of all points on the same circular line, and the second point set consists of all points with a distance less than a preset value from the circular line; Step S4, determining whether the temperature differences of all points in the first point set and the second point set do not exceed a preset threshold. If so, execute Step S5. If not, execute Step S6; Step S5, solving the average value of the temperatures of all points as the temperature values of each point on the line connecting the adjacent cluster centers, and determining the change rate of the temperature value according to the temperature value of the point and the distance between the two points; Step S6, marking the points exceeding the preset threshold as a new cluster center, performing clustering, and then repeating Step S1 - Step S4.

[0013] Optionally, according to the position of the point to be predicted and historical data, determine the relationship between the pressure at the point to be predicted and the snow depth, including: dividing the snow-covered section into a trampled part and a non-trampled part according to whether it is trampled; for the trampled part, according to the movement trajectories of pedestrians or vehicles in the historical data, divide the trampled part into trampling areas of different levels, randomly select multiple points in each trampling area, sample the pressures corresponding to different depths at multiple points, and perform linear fitting on the pressures corresponding to different depths at multiple points to obtain the pressure-depth fitting curve of the corresponding area; for the non-trampled part, solve the point pressure value according to the relationship between snow pressure, snow density, and snow depth.

[0014] An embodiment of the present invention provides a contact-type road surface temperature monitoring system. By using multiple contact-type temperature sensors to obtain the ambient temperature at different positions, as well as the temperature and pressure at different snow layer depths at different positions, the temperature prediction module performs comprehensive analysis based on multi-dimensional data such as the temperature and pressure data inside the snow layer and the ambient temperature, and can more accurately predict the temperature change trend at multiple points with different depths in the snow-covered section. Then, the visualization method using three-dimensional image display is intuitive and vivid, allowing road managers to clearly understand the temperature distribution in the snow-covered section, promptly discover temperature abnormal areas, and take corresponding measures. Compared with only monitoring the surface temperature of the snow layer, this solution predicts the temperature change inside the snow layer through the data of multiple contact-type temperature sensors, which can provide comprehensive and effective information for road section management personnel to help them judge potential safety hazards in advance. For example, when the temperature inside the snow layer approaches or is lower than the freezing point, the road surface is more likely to freeze, causing vehicles to skid and leading to safety problems, so corresponding measures can be taken in advance. In addition, in the snow-covered section, according to the change of the road surface temperature, the spreading amount and timing of the snow melting agent can be accurately controlled. For another example, during periods with lower temperatures and higher ice formation risks, increase the spreading amount of the snow melting agent; while when the temperature rises and the ice and snow start to melt, reduce the use of the snow melting agent, which can not only effectively melt the ice and snow but also avoid waste of resources. Finally, by monitoring the temperature inside the snow layer, the heat preservation effect of the snow cover can be understood, and its impact on the operating temperature of facilities and equipment can be evaluated. When the temperature inside the snow layer rises, the snow may melt, causing the facilities and equipment to be soaked by water and increasing the risk of equipment damage. By monitoring the temperature inside the snow layer, the snow melting phenomenon can be promptly discovered and corresponding protective measures can be taken.

[0015] Other advantages, objectives, and features of the present invention will be described in the subsequent specification, and to some extent, they are obvious to those skilled in the art, or those skilled in the art can obtain teachings from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following specification. Brief Description of the Drawings

[0016] To make the objectives, technical solutions, and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0017] Figure 1 It is a specific example diagram of a contact-type road surface temperature monitoring system in the present invention;

[0018] Figure 2 It is a specific example flowchart for calculating the temperature change rate corresponding to adjacent clustering centers in the present invention. Specific Embodiments

[0019] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and can also be the communication inside two components. It can be a wireless connection or a wired connection. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0021] In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0022] An embodiment of the present invention provides a contact-type road surface temperature monitoring system, as Figure 1 shown, including:

[0023] A plurality of contact-type temperature sensors 101 are arranged at a plurality of target positions in the snow-covered section. Each contact-type temperature sensor includes a first temperature monitoring module and a second temperature monitoring module. The first temperature monitoring module includes a first temperature sensing probe and a pressure monitor. The first temperature monitoring module is arranged at different depths at any position in the snow-covered section for monitoring the internal temperature and pressure data at different depths in the snow-covered section. The second temperature monitoring module is a second temperature sensing probe, which is arranged at a position at a target distance above the surface of the snow-covered section for monitoring the environmental temperature of the snow-covered position;

[0024] A temperature prediction module 102 for predicting the temperature at different depths at a plurality of points in the snow-covered section according to the internal temperature data, pressure data, and environmental temperature at different depths at a plurality of target positions in the snow-covered section;

[0025] The predicted temperature display module 103 is used to perform image processing on the temperatures at different depths of multiple points in the snow-covered section and display the overall three-dimensional temperature image of the snow-covered section.

[0026] Exemplarily, a plurality of contact temperature sensors 101 can be arranged at different positions in the snow-covered section to cover the key areas of the entire section. Since the internal temperature data of the snow reflects the heat conduction situation of the snow layer, the ambient temperature data reflects the influence of external heat sources, and the pressure data is related to the density and heat conduction characteristics of the snow layer. Therefore, in this embodiment, each contact temperature sensor includes two modules, a first temperature monitoring module and a second temperature monitoring module. The first temperature monitoring module is used to penetrate into the snow layer to monitor the temperature and pressure at different snow layer depths, and the second temperature monitoring module is used to monitor the ambient temperature at the position where the contact temperature sensor is located. The second temperature monitoring module is connected to the contact temperature sensor body in a telescopic connection manner, and the height of the second temperature monitoring module from the snow surface can be adjusted arbitrarily.

[0027] The temperature prediction module 102 predicts the temperatures at different depths of multiple points in the snow-covered section based on the internal temperature data, pressure data, and ambient temperature at different depths of multiple target positions in the snow-covered section. By comprehensively considering multi-dimensional data such as internal temperature, pressure, and ambient temperature, it can more accurately simulate the heat conduction process and temperature change law of the snow layer. Compared with the prediction model based on a single factor, the method proposed in this embodiment can improve the prediction accuracy and reliability. Compared with obtaining the internal temperature of the snow-covered section entirely by sensors, in this embodiment, the corresponding neural network model can be obtained through training of the neural network to complete the temperature prediction at different points and different depths of the snow-covered section, reducing the number of hardware devices, improving the degree of intelligence, and reducing the temperature monitoring cost.

[0028] The predicted temperature display module 103 performs image processing on the temperatures at different depths of multiple points in the snow-covered section and displays the overall three-dimensional temperature image of the snow-covered section. For the staff to view the overall snow layer temperature situation of the snow-covered section.

[0029] An embodiment of the present invention provides a contact-type road surface temperature monitoring system. By using multiple contact-type temperature sensors to obtain the ambient temperature at different positions, as well as the temperature and pressure at different snow layer depths at different positions, a temperature prediction module performs comprehensive analysis based on multi-dimensional data such as the internal temperature and pressure data of the snow layer and the ambient temperature, and can more accurately predict the temperature change trends at different depths of multiple points in the snow-covered section. Then, a visualization method using three-dimensional images is intuitive and vivid, allowing road managers to clearly understand the temperature distribution in the snow-covered section, promptly discover temperature anomaly areas, and take corresponding measures. Compared with only monitoring the surface temperature of the snow layer, this solution predicts the change of the internal temperature of the snow layer through the data of multiple contact-type temperature sensors, and can provide effective information for road section management personnel to help them judge potential safety hazards in advance. For example, when the internal temperature of the snow layer approaches or is lower than the freezing point, the road surface is more likely to freeze, causing vehicles to skid and leading to safety problems, so corresponding measures can be taken in advance. In addition, in the snow-covered section, according to the change of the road surface temperature, the spreading amount and timing of the snowmelt agent can be accurately controlled. For example, in a period with lower temperature and higher ice formation risk, increase the spreading amount of the snowmelt agent; while when the temperature rises and the ice and snow begin to melt, reduce the use of the snowmelt agent, which can not only effectively melt the snow and remove ice but also avoid waste of resources. Finally, by monitoring the internal temperature of the snow layer, the heat preservation effect of the snow cover can be understood, and its impact on the operating temperature of facilities and equipment can be evaluated. When the internal temperature of the snow layer rises, the snow may melt, causing the facilities and equipment to be soaked by water and increasing the risk of equipment damage. By monitoring the internal temperature of the snow layer, the snow melting phenomenon can be promptly discovered and corresponding protective measures can be taken.

[0030] As an optional implementation manner, predicting the temperature at different depths of multiple points in the snow-covered section according to the internal temperature data, pressure data, and ambient temperature at different depths of multiple target positions in the snow-covered section includes:

[0031] Dividing the snow-covered section into grids, determining the grids where no contact-type temperature sensors are set, taking the center points in the grids as the points to be predicted; predicting the ambient temperature of the points to be predicted according to a pre-trained first target neural network; determining the geothermal heat of the snow-covered section according to the geothermal measurement method; determining the relationship between the pressure and the snow depth of the points to be predicted according to the positions of the points to be predicted and historical data; inputting the geothermal heat, ambient temperature, and the pressures corresponding to different depths into a pre-trained second target neural network to obtain the temperatures corresponding to different snow cover depths of the road surface represented by the points to be predicted; wherein the pre-trained second target neural network is trained according to the internal temperature data, pressure data, and ambient temperature at different depths of multiple target positions in the snow-covered section.

[0032] Exemplarily, the contact temperature sensor may further include a positioning device to determine its longitude and latitude coordinates, determine the grid it is located in according to the longitude and latitude coordinates, and thus determine the grids where no contact temperature sensor is set; in addition, in the case where the contact temperature sensor does not include a positioning device, the grids where the contact temperature sensor is set can be manually marked to determine the grids where no contact temperature sensor is set.

[0033] In this embodiment, in order to obtain the temperatures corresponding to different snow cover depths on the road surface represented by the point to be predicted, first, the influence of the ambient temperature on it is considered. In order to improve the intelligence and accuracy of temperature prediction for different snow cover depths, the ambient temperature is predicted by the first target neural network in this embodiment. On the one hand, compared with directly using the weather temperature to predict the temperature of different snow cover depths, the ambient temperature predicted by this method is the ambient temperature of a specific point. Compared with the weather temperature, the mapping relationship between the ambient temperature of a specific point and the snow layer temperature is more accurate, so as to accelerate the learning speed of the second target neural network and improve the prediction accuracy; on the other hand, generally, the range of snow-covered sections is relatively large. By collecting the ambient temperature at some positions and using machine learning methods, this method can predict the temperatures of other points where no contact temperature sensor is installed, which can reduce the number of hardware devices used, save costs, and there is no need for staff to lay sensors over a large area, improving the overall prediction intelligence.

[0034] Specifically, by training the first target neural network, the ambient temperature of the point to be predicted is predicted. Then, the ambient temperature of the point to be predicted is used as one of the inputs of the second target neural network. Then, pressure and depth are also considered as the inputs of the second target neural network to obtain the temperatures of different snow cover depths on the road surface represented by the point to be predicted.

[0035] The specific implementation of predicting the ambient temperature of the point to be predicted according to the pre-trained first target neural network includes: obtaining weather factors and the ambient temperature obtained by the contact temperature sensors at multiple target positions; inputting the ambient temperature, weather factors and their corresponding target positions into the first target neural network; in the first target neural network, using the gradient descent method to adjust the network weights and minimize the loss function until the training is completed to obtain the trained first target neural network; inputting the position information and weather factors of the point to be predicted into the trained first target neural network to obtain the ambient temperature of the point to be predicted.

[0036] The first target neural network includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is consistent with the dimension of the number of positions, and the number of nodes in the output layer is one, representing the predicted environmental temperature. Weather factors can include humidity, wind direction, etc., and the weather factors can be determined according to weather forecast data. The obtained environmental temperature, weather factors, and their corresponding target positions are used as training data and input into the first target neural network. The weather factors and target positions are used as input features, and the environmental temperature is used as a label. The mean squared error can be used as the loss function to calculate the difference between the predicted value and the actual value. During the training process, the gradient descent method can be used to optimize the weights and biases of the first target neural network, calculate the gradient of the loss function with respect to each weight, and then update the weights according to the learning rate to gradually reduce the value of the loss function until the loss function converges and the training is completed. At this time, the trained first target neural network can learn the mapping relationship between the environmental temperature at the current position under the influence of different regions and different weather factors.

[0037] To further improve the prediction accuracy of the first target neural network, in this embodiment, the first target neural network can also be a network for transfer learning. Since the number of target positions and environmental temperatures as samples and sample labels in the current environment is small, the accuracy of the trained first target neural network may be low. Therefore, a neural network model pre-trained on a large-scale meteorological data set can be selected. This model has learned the complex patterns and features of temperature changes under different positions and different meteorological conditions. When applying the pre-trained model to the current task, most of the parameters of the model can be frozen, especially the parameters of the early layers, because the features learned by these layers are general in different tasks. Use a small amount of temperature data sets in the current task to fine-tune the model. During the fine-tuning process, some trainable parameters in the pre-trained model are updated. Through fine-tuning, the model can better adapt to the data distribution and features of the new task. Input the position information and weather factors of the point to be predicted into the trained first target neural network, and the environmental temperature of the point to be predicted can be obtained.

[0038] Before predicting the temperature corresponding to different snow cover depths on the road surface represented by the point to be predicted, it is also necessary to determine the relationship between the pressure and snow depth and the geothermal heat at the point to be predicted according to the location of the point to be predicted and historical data. Specifically, the snow-covered road section can be divided into a trampled part and a non-trampled part. For the trampled part, according to the movement trajectories of pedestrians or vehicles, the trampled part is divided into trampling areas of different levels, and multiple points are randomly selected in each area to sample the pressures corresponding to different depths at multiple points. The pressures corresponding to different depths at multiple points are linearly fitted to obtain the pressure-depth fitting curve of this area. For the non-trampled part, the pressure value of the point is solved according to the relationship between snow pressure, snow density, and snow depth. When it is necessary to determine the relationship between the pressure and snow depth at the point to be predicted, first, according to the location of the point to be predicted, determine the pressure-depth relationship corresponding to this point. For example, when this point belongs to a certain trampling area, the pressure-depth fitting curve of this trampling area pre-fitted is retrieved to obtain the pressures corresponding to different depths.

[0039] The geothermal heat determination method can be to select representative deep wells or geothermal wells (holes) in the snowfield for in-hole (well) temperature measurement. After obtaining the pressure value, ambient temperature value, and geothermal heat, the temperature corresponding to different snow cover depths on the road surface represented by the point to be predicted can be predicted. Specifically, the geothermal heat, ambient temperature, and pressures corresponding to different depths are input into a pre-trained second target neural network to obtain the temperature corresponding to different snow cover depths on the road surface represented by the point to be predicted.

[0040] The training process of the second target neural network includes: taking the geothermal heat, in-snow temperature, pressure data, and ambient temperature obtained by the same contact temperature sensor as a data group. Among them, the depth, ambient temperature, pressure, and geothermal heat in the data group are used as samples, and the internal temperature is used as the sample label; multiple data groups are input into the second target neural network, and the gradient descent method is used to adjust the network weights to minimize the overall loss function until the training is completed to obtain the trained second target neural network. Among them, the second target neural network is a physics-informed neural network, and its overall loss function is composed of an error term between the predicted temperature and the actual temperature, a constraint error term for the heat transfer relationship between snow layers, and a constraint error term for the heat transfer relationship in the snowfield.

[0041] The overall loss function includes: ;

[0042] Among them, represents the overall loss function, are hyperparameters, representing the weight coefficients of the error term between the predicted temperature and the actual temperature, the weight coefficient of the constraint error term for the heat transfer relationship between snow layers, and the weight coefficient of the constraint error term for the heat transfer relationship in the snowfield respectively, represents the predicted internal temperature, the actual internal temperature, is the snow density, is the specific heat capacity of snow, is the temperature of the snow layer, is the thermal conductivity of the snow layer, z is the depth, and t is the time, represents the geothermal heat, is the thermal conductivity of the ground surface.

[0043] The embodiment of the present invention provides a contact type road surface temperature monitoring system. In the prediction process, not only main factors such as environmental temperature and geothermal heat are considered, but also other influencing factors such as the relationship between the pressure at the point to be predicted and the snow depth are combined. This method of comprehensive prediction of multiple factors can more comprehensively reflect the actual situation of the temperature change of the snow-covered road section, and improve the accuracy and reliability of the prediction result.

[0044] As an optional implementation manner, a contact type road surface temperature monitoring system further includes: a lidar scanning device for scanning the three-dimensional coordinates of the snow accumulation on the snow-covered road section;

[0045] Perform image processing on the temperatures at different depths of multiple points on the snow-covered road section to display the overall three-dimensional temperature image of the snow-covered road section, including: obtaining the temperatures at different depths of multiple points, and binding the three-dimensional position and temperature at a random depth of any point to form a data group, and multiple data groups form a data set; clustering multiple data groups in the data set to obtain multiple clustering centers; reading the temperature values and distance values of multiple adjacent clustering centers, and calculating the temperature change rate corresponding to the adjacent clustering centers; determining the temperature identifier at the clustering center according to the temperature of the clustering center; adjusting the change rate of the temperature identifier color according to the temperature identifier at the clustering center and the temperature change rate; performing color rendering on each clustering area in the three-dimensional image of the snow-covered road section according to the positions, temperature identifier colors, and temperature identifier color change rates of each clustering center.

[0046] Exemplarily, the working principle of the lidar scanning device is as follows: When the laser pulse encounters the snow accumulation, it will be reflected back, and the lidar receiver will capture these reflected signals. By measuring the time difference between the emission and reception of the laser pulse, and combining the propagation speed of the laser, the distance between the lidar and the reflection point is calculated. At the same time, the angle of laser emission is recorded. According to the distance, angle, and the position and attitude information of the lidar device, the three-dimensional coordinates of each reflection point are calculated. After obtaining the three-dimensional coordinates of each reflection point, a digital model of the snow accumulation on the snow-covered road section is reconstructed according to the three-dimensional coordinates.

[0047] Based on the reconstruction of the digital model of snow accumulation on the snow-covered section, the temperature at multiple points on the snow-covered section is randomly predicted, and the predicted point temperatures are filled with colors and rendered on the digital model according to the display colors corresponding to the temperatures, so as to obtain the overall three-dimensional temperature image of the snow-covered section. Specifically, in order to avoid the problem of large computational load and waste of computing resources caused by predicting the temperature at each point on the snow-covered section, in this embodiment, multiple points are randomly selected to predict the temperature at these points to obtain temperatures at different depths. Then, any depth is randomly selected to determine its temperature, and the temperature is bound to the three-dimensional coordinates where the depth of the randomly selected point is located to form a data group. , represents the three-dimensional position of the point, T represents the temperature, and then clustering analysis is performed based on the data group. According to the clustering analysis results, the digital model is rendered. In this solution, only some points need to be randomly selected for prediction to complete the rendering of the entire digital model.

[0048] The specific method of clustering is as follows: After forming the data group, clustering is performed on multiple data groups. The specific clustering method can be clustering. Before performing clustering, the number of clusters (i.e., the number of clusters) needs to be determined. This can be achieved by the elbow method, that is, calculating the total squared error under different numbers of clusters, and selecting the point where the error begins to decrease significantly as the number of clusters. Randomly select K points as the initial cluster centers, and measure the similarity between data groups according to the distance metric. In this embodiment, the Euclidean distance can be used as the similarity measurement calculation formula. Specifically, the Euclidean distance formula can be: ;

[0049] where, respectively represent the weights of the distance difference and temperature difference between two data points on the similarity calculation, which can be adjusted as needed. This embodiment gives , , and respectively represent different data groups.

[0050] Each data group is assigned to the nearest cluster center to form K clusters, the centroid of each cluster is calculated, and the cluster center is updated to the centroid. The above assignment and update process is repeated until the cluster center no longer changes or the preset number of iterations is reached. According to the clustering results, the characteristics and distribution of different clusters are analyzed, and it can be found that data groups with similar temperatures and close distances are clustered together to form different temperature areas. Each temperature area has a cluster center. The temperature values ​​and distance values ​​of multiple adjacent cluster centers are read. The temperature change rate corresponding to the adjacent cluster centers is calculated by comparing the temperature difference with the distance value. At the same time, according to the temperature of the cluster center, the temperature and temperature identification color comparison table is searched to determine the temperature identification at the cluster center. In order to avoid single coloring of a single point each time, resulting in low coloring efficiency, this embodiment adjusts the change rate of the temperature identification color according to the temperature identification and temperature change rate at the cluster center, and then fills the temperature identification color at the cluster center position, and in the direction from the current cluster center to each other adjacent cluster center, the cluster area is filled and rendered according to the temperature identification color change rate. It can be understood that the cluster area is a closed space connected by the outermost data point positions. By connecting the cluster centers and solving the temperature change rate between the cluster centers, and filling the color according to the change rate between the cluster centers, the problem of sudden color changes and inability to connect between different cluster areas can be effectively avoided.

[0051] It should be explained in detail that in the direction extending from the current cluster center to each other adjacent cluster center, the specific process of color filling and rendering the cluster area according to the temperature identification color change rate is as follows: when there are multiple other adjacent cluster centers of the current cluster center, any two-dimensional slice of the cluster area is divided into multiple sub-areas by the line connecting the current cluster center and each adjacent cluster center, and a line is made in each sub-area, which divides the angle between the current cluster center and the two adjacent cluster centers equally, and the average value of the temperature identification color change rate corresponding to the line connecting the current cluster center and the two adjacent cluster centers is solved as the average value of the temperature identification color change rate of the line, and the entire sub-area is filled with color according to the mean value, and the above steps are repeated to obtain the coloring of multiple two-dimensional slices of the cluster area. When the number of two-dimensional slices filled with color meets the preset number, the remaining three-dimensional gaps in the cluster area are filled with color according to the color and distance of the corresponding positions of the two two-dimensional slices.

[0052] As an optional implementation, based on the above coloring process, in order to further improve the accuracy of coloring, this embodiment reads the temperature values ​​and distance values ​​of multiple adjacent cluster centers, and calculates the temperature change rate corresponding to the adjacent cluster centers, such as Figure 2 As shown, including:

[0053] Step S1: Determine the temperature values at multiple points on the line connecting adjacent cluster centers based on the temperature difference corresponding to the adjacent cluster centers and the distance value between the adjacent cluster centers.

[0054] Step S2: Use the distance between each point and the corresponding cluster center as the radius to obtain each circular line.

[0055] Step S3: Determine the temperatures of all points in the first point set and the second point set corresponding to each circular line. Here, the first point set consists of all points on the same circular line, and the second point set consists of all points whose distance from the circular line is less than a preset value.

[0056] Step S4: Determine whether the temperature differences of all points in the first point set and the second point set do not exceed the preset threshold. If so, execute Step S5; if not, execute Step S6.

[0057] Step S5: Solve the average value of the temperatures of all points as the temperature value at each point on the line connecting adjacent cluster centers, and determine the change rate of the temperature value based on the temperature value of the point and the distance between two points.

[0058] Step S6: Mark the points exceeding the preset threshold as a new cluster center, perform clustering, and then repeat Steps S1 - S4.

[0059] Exemplarily, based on the above coloring method in this embodiment, in order to further improve the coloring accuracy and coloring efficiency, the temperature values at multiple points on the line connecting adjacent cluster centers are determined according to the temperature difference corresponding to the adjacent cluster centers and the distance value between the adjacent cluster centers. The determination method of multiple points can be to divide by a preset fixed length or a fixed number of points. According to the ratio of the distance value to the number of points, the distance between points is determined, so as to determine the position of each point. This embodiment does not limit this; use the distance between each point and the corresponding cluster center as the radius to obtain each circular line. Then, since a cluster center may have multiple adjacent cluster centers, and there are multiple connection lines between the cluster center and each adjacent cluster center, for any point on the line connecting the cluster center and any adjacent cluster center, there may be other points on the corresponding circular line, and there may be other points near the circular line. Therefore, in this embodiment, the temperature values of all points on the same circular line and all points whose distance from the circular line is less than a preset value are obtained, and the average value is solved as the temperature value of the point on the line connecting the cluster center and the adjacent cluster center. Here, the preset value can be 3 cm, and the size of the preset value may be set according to needs. This embodiment does not limit this.

[0060] Determine the rate of change of the temperature value according to the temperature value of a point position and the distance between two point positions. When the temperature difference of any point position in the first point position set and the second point position set exceeds a preset threshold, mark the point position exceeding the preset threshold as a new clustering center, perform clustering, and then repeat the above steps until the rate of change of the temperature corresponding to all adjacent clustering centers is obtained.

[0061] An embodiment of the present invention provides a contact type road surface temperature monitoring system. By the temperature difference corresponding to adjacent clustering centers and the distance value between adjacent clustering centers, determine the temperature values of multiple point positions on the line connecting the adjacent clustering centers, and solve the average value of the points on the same circumferential line and the nearby points as the temperature value of the points on the connecting line. Use the difference between the temperature value determined by this point position and the adjacent temperature value as the element for solving the rate of change of the temperature. Compared with directly determining the rate of change of the temperature according to the temperature difference between two clustering centers, this solution can further manifest the subtle changes in the temperature between the clustering centers, making the displayed three-dimensional image of the snow-covered section more accurate.

[0062] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A contact road surface temperature monitoring system, characterized in that: include: A plurality of contact temperature sensors are arranged at a plurality of target positions of a snow-covered road section, each of the contact temperature sensors comprises a first temperature monitoring module and a second temperature monitoring module, the first temperature monitoring module comprises a first temperature sensing probe and a pressure monitor, the first temperature monitoring module is arranged at different depths at any position of the snow-covered road section, and is used to monitor the internal temperature and pressure data of different depths of the snow-covered road section, the second temperature monitoring module is a second temperature sensing probe, which is arranged at a position higher than the target distance on the surface of the snow-covered road section, and is used to monitor the ambient temperature of the snow-covered position; A temperature prediction module is used to predict the temperature at different depths of multiple points on the snow-covered road section based on the internal temperature data, pressure data, and ambient temperature at different depths of multiple target locations on the snow-covered road section; The predicted temperature display module is used to process the temperature at different depths at multiple points on the snow-covered road section and display a three-dimensional temperature image of the entire snow-covered road section; The above system also includes: LiDAR scanning equipment, used to scan the three-dimensional coordinates of snow on snow-covered road sections; Perform image processing on the temperature at different depths of multiple points on the snow-covered road section to display the three-dimensional temperature image of the entire snow-covered road section, including: Obtain the temperature of multiple points at different depths, and bind the three-dimensional position of any point at random depth to the temperature to form a data group. Multiple data groups form a data set. Cluster multiple data groups in the data set to obtain multiple cluster centers; Read the temperature values ​​and distance values ​​of multiple adjacent cluster centers, and calculate the temperature change rates corresponding to the adjacent cluster centers; According to the temperature of the cluster center, determine the temperature mark at the cluster center; According to the temperature mark at the cluster center and the temperature change rate, the change rate of the temperature mark color is adjusted; According to the position of each cluster center, the temperature identification color and the temperature identification color change rate, each cluster area in the three-dimensional image of the snow-covered road section is rendered in color.

2. A contact-type road surface temperature monitoring system according to claim 1, characterized in that: Based on the internal temperature data, pressure data, and ambient temperature at different depths of multiple target locations on the snow-covered road section, the temperature at different depths of multiple points on the snow-covered road section is predicted, including: Divide the snow-covered road section into grids, determine the grids where no contact temperature sensors are installed, and take the center point in the grid as the point to be predicted; Predicting the ambient temperature of the point to be predicted based on the pre-trained first target neural network; Determine the geothermal heat of the snow-covered road section based on geothermal measurement methods; According to the location of the point to be predicted and historical data, the relationship between the pressure of the point to be predicted and the snow depth is determined; The geothermal heat, ambient temperature and pressures corresponding to different depths are input into a pre-trained second target neural network to obtain the temperatures corresponding to different snow depths of the road surface represented by the predicted point; wherein the pre-trained second target neural network is trained based on the internal temperature data, pressure data and ambient temperature at different depths of multiple target positions on the snow-covered road section.

3. A contact type road surface temperature monitoring system according to claim 2, characterized in that: According to the pre-trained first target neural network, the ambient temperature of the point to be predicted is predicted, including: Obtain weather factors and ambient temperature obtained by contact temperature sensors at multiple target locations; Inputting ambient temperature, weather factors and their corresponding target positions into a first target neural network; In the first target neural network, the gradient descent method is used to adjust the network weights and minimize the loss function until the training is completed to obtain the trained first target neural network; The location information and weather factors of the point to be predicted are input into the trained first target neural network to obtain the ambient temperature of the point to be predicted.

4. A contact type road surface temperature monitoring system according to claim 2, characterized in that: The second objective neural network training process includes: The ground heat, snow temperature, pressure data, and ambient temperature obtained by the same contact temperature sensor are taken as a data group, wherein the depth, ambient temperature, pressure, and ground heat in the data group are taken as samples, and the internal temperature is taken as a sample label; Multiple data groups are input into the second target neural network, and the network weights are adjusted using the gradient descent method to minimize the overall loss function until the training is completed, thereby obtaining a trained second target neural network. The second target neural network is a physical information neural network, and its overall loss function consists of the error terms of the predicted temperature and the actual temperature, the error terms of the constraints on the heat transfer relationship between snow, and the error terms of the constraints on the heat transfer relationship between snow and ground.

5. A contact type road surface temperature monitoring system according to claim 4, characterized in that: The overall loss function includes: ; in, represents the overall loss function, are hyperparameters, representing the error term weight coefficients of the predicted temperature and the actual temperature, the error term weight coefficients of the snow-to-snow heat transfer constraint, and the error term weight coefficients of the snow-to-ground heat transfer constraint. represents the predicted internal temperature, The actual internal temperature, is the snow density, is the specific heat of snow, is the snow layer temperature, is the snow layer conductivity, z is the depth, t is the time, Indicates geothermal heat, The thermal conductivity of the ground.

6. A contact type road surface temperature monitoring system according to claim 1, characterized in that: Read the temperature values ​​and distance values ​​of multiple adjacent cluster centers, and calculate the temperature change rate corresponding to the adjacent cluster centers, including: Step S1, determining the temperature values ​​of multiple points on the line connecting the adjacent cluster centers according to the temperature difference corresponding to the adjacent cluster centers and the distance value between the adjacent cluster centers; Step S2, taking the distance between each point and the corresponding cluster center as the radius to obtain each circular line; Step S3, determining the temperatures of all points in the first point set and the temperatures of all points in the second point set corresponding to each circumferential line, wherein the first point set is composed of all points on the same circumferential line, and the second point set is composed of all points whose distance from the circumferential line is less than a preset value; Step S4, determining whether the temperature differences of all points in the first point set and the second point set do not exceed a preset threshold value, if yes, executing step S5, if no, executing step S6; Step S5, solving the average value of the temperature of all points as the temperature value of each point on the line connecting the adjacent cluster centers, and determining the rate of change of the temperature value according to the temperature value of the point and the distance between two points; Step S6, marking the point exceeding the preset threshold as a new cluster center, performing clustering, and then repeating steps S1 to S4.

7. A contact-type road surface temperature monitoring system according to claim 2, characterized in that: According to the location of the predicted point and historical data, the relationship between the pressure of the predicted point and the snow depth is determined, including: Divide the snow-covered road section into a trampled part and a non-trampled part according to whether it is trampled or not; For the trampled part, according to the movement trajectory of pedestrians or vehicles in the historical data, the trampled part is divided into trampled areas of different levels, and multiple points are randomly selected in each trampled area, and the pressures corresponding to different depths of the multiple points are sampled. The pressures corresponding to different depths of the multiple points are linearly fitted to obtain the pressure-depth fitting curve of the corresponding area; For the non-trampled part, the point pressure value is solved based on the relationship between snow pressure, snow density and snow depth.

Images