Deicing height prediction method and system for milling cutter of deicing vehicle and deicing vehicle

Abstract

The invention relates to the technical field of road ice and snow removal, and discloses a method and system for predicting the deicing height of a milling cutter of a deicing vehicle and the deicing vehicle, and the method comprises the steps: obtaining the thickness of a current pavement ice layer in a preset range in front of the deicing vehicle, the preset range comprises a plurality of preset unit ranges divided in the width direction and the length direction of the road; a road surface ice layer thickness matrix is constructed based on the road surface ice layer thickness, the road surface ice layer thickness matrix is a two-dimensional matrix, a transverse sequence in the road surface ice layer thickness matrix is the ice layer thickness in each preset unit range in the road width direction, and a vertical sequence in the road surface ice layer thickness matrix is the ice layer thickness in each preset unit range in the road length direction; acquiring the current environment temperature and the longitude and latitude of the current position of the deicing vehicle; and according to the pavement ice layer thickness matrix, the environment temperature and the longitude and latitude, predicting the optimal ice layer cutting height of the milling cutter on the deicing vehicle. The prediction system comprises a data acquisition module and an analysis module. The deicing effect is good, and the safety to the road surface is high.

Description

Technical Field

[0001] The present invention relates to the technical field of road snow and ice removal, and in particular to a method and system for predicting the deicing height of a milling cutter of a deicing vehicle, and a deicing vehicle. Background Art

[0002] With the progress and development of scientific society, the smoothness of roads and driving safety are increasingly concerned with the quality of human production and life and daily operations. Among them, in extreme weather, roads will be covered with snow and ice, which greatly affects travel efficiency and safety. In traditional road deicing, most of them are deiced by spreading salt, snow melting agents or thermal deicing. Spreading salt will have a certain negative impact on the environment, spreading snow melting agents is limited in application, and thermal deicing consumes a lot of energy, which is more suitable for small-surface deicing. At present, in order to improve the efficiency and effect of deicing, mechanical methods are used, such as deicing machines or deicing vehicles, to save labor and time. However, in actual work, the existing deicing machines or deicing vehicles are prone to incomplete deicing or damage to the road surface due to the random initial cutting height of the milling cutter or relying on manual experience, which brings considerable negative obstacles to deicing.

[0003] Therefore, there is an urgent need for a method and system for predicting the deicing height of a deicing vehicle milling cutter and a deicing vehicle to solve the above problems. Summary of the invention

[0004] Based on the above, the purpose of the present invention is to provide a method and system for predicting the deicing height of a deicing vehicle milling cutter, and a deicing vehicle, which have better deicing effect and higher safety for the road surface.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for predicting the deicing height of a deicing vehicle milling cutter, comprising:

[0007] Obtaining the current road ice thickness in a preset range in front of the de-icing vehicle, wherein the preset range includes a plurality of preset unit ranges divided along the width direction of the road and along the length direction of the road;

[0008] Constructing a road surface ice layer thickness matrix based on the road surface ice layer thickness, wherein the road surface ice layer thickness matrix is ​​a two-dimensional matrix, the horizontal sequence in the road surface ice layer thickness matrix is ​​the ice layer thickness within each preset unit range along the road width direction, and the vertical sequence is the ice layer thickness within each preset unit range along the road length direction;

[0009] Obtaining the current ambient temperature and the longitude and latitude of the current location of the de-icing vehicle;

[0010] The optimal height at which the milling cutter on the deicing vehicle cuts into the ice layer is predicted according to the road ice layer thickness matrix, the ambient temperature and the longitude and latitude.

[0011] As a preferred solution of a method for predicting the deicing height of a milling cutter on a deicing vehicle, predicting the optimal height of the milling cutter on the deicing vehicle to cut into the ice layer according to the road ice thickness matrix, the temperature and the longitude and latitude includes:

[0012] According to the road ice thickness matrix, the ambient temperature and the longitude and latitude, the optimal height of the milling cutter to cut into the ice layer is predicted by a milling cutter height prediction model preset on the deicing vehicle.

[0013] As a preferred solution of a method for predicting the deicing height of a milling cutter of a deicing vehicle, the milling cutter height prediction model includes a convolutional neural network and a fully connected neural network, the convolutional neural network is used to extract the characteristics of the road ice thickness matrix, and the milling cutter height prediction model preset on the deicing vehicle predicts the optimal height of the milling cutter cutting into the ice layer, including:

[0014] The road surface ice layer thickness matrix is ​​input into the milling cutter height prediction model, and the convolutional neural network extracts the features of the road surface ice layer thickness matrix;

[0015] Flattening the features of the road ice thickness matrix extracted by the convolutional neural network into a first vector;

[0016] The ambient temperature and the longitude and latitude are added to the first vector to form a second vector. The fully connected neural network extracts features of the second vector and outputs a predicted height of the milling cutter, where the predicted height of the milling cutter is the optimal height for the milling cutter to cut into the ice layer.

[0017] As a preferred solution of a method for predicting the deicing height of a deicing vehicle milling cutter, the method further includes:

[0018] The current height of the milling cutter is obtained, and the predicted adjustment value of the milling cutter is obtained according to the current height of the milling cutter and the predicted optimal height of the milling cutter for cutting into the ice layer.

[0019] As a preferred solution of a deicing vehicle milling cutter deicing height prediction method, a road surface ice thickness matrix is ​​constructed based on the road surface ice thickness within a preset range in front of the current deicing vehicle, including:

[0020] Obtaining the original height of the road surface within the preset range in front of the deicing vehicle, obtaining the current ice surface height within the preset range in front of the deicing vehicle through laser radar detection preset on the deicing vehicle, and obtaining the ice layer thickness point cloud within the preset range according to the original height of the road surface and the current ice surface height;

[0021] The ice thickness point cloud is converted into the road ice thickness matrix using a two-dimensional matrix.

[0022] A milling cutter height prediction model training method, comprising:

[0023] Acquire training samples, the training samples include a first input data set and a first output data set, the data in the first output data set are several first preset cutting-in heights of the milling cutter, wherein after the milling cutter de-ices at the first preset cutting-in height, the ice thickness of the road surface is within a preset thickness range and the road surface is not damaged; the data in the first input data set are the ambient temperature corresponding to the milling cutter working at each of the first preset cutting-in heights, the longitude and latitude of the location of the de-icing vehicle, and a road surface ice thickness matrix, wherein the road surface ice thickness matrix is ​​constructed based on the current road surface ice thickness within a preset range in front of the de-icing vehicle, the preset range includes several preset unit ranges divided along the width direction of the road and along the length direction of the road, the road surface ice thickness matrix is ​​a two-dimensional matrix, the horizontal sequence in the road surface ice thickness matrix is ​​the ice thickness within each preset unit range along the width direction of the road, and the vertical sequence is the ice thickness within each preset unit range along the length direction of the road;

[0024] Construct an initial neural network model for predicting milling cutter height, use the training samples to train the initial neural network model for predicting milling cutter height, use mean square error as the loss function, optimize and update the parameters in the initial neural network model for predicting milling cutter height, wherein the parameters include weight parameters and bias parameters, and obtain the milling cutter height prediction model, which is used to predict the optimal height of the milling cutter cutting into the ice layer according to the road ice thickness matrix, the current ambient temperature and the longitude and latitude.

[0025] As a preferred solution of a milling cutter height prediction model training method, the milling cutter height prediction initial neural network model includes a convolutional neural network and a fully connected neural network, the convolutional neural network is used to extract the characteristics of the road ice thickness matrix and output a first vector, the fully connected layer is used to extract the characteristics of the first vector, the ambient temperature and the longitude and latitude, and output the predicted milling cutter height;

[0026] According to the predicted milling cutter height and the first preset cutting height, the parameters are updated through loss function optimization to obtain the milling cutter height prediction model.

[0027] As a preferred solution of a milling cutter height prediction model training method, obtaining the training samples further includes:

[0028] Data enhancement processing is performed on the data in the first input data set to obtain several second input data sets, each of which corresponds to the same output data set, and the data enhancement processing includes rotating, translating and / or scaling the road ice thickness matrix in the first input data set; and / or adding noise, scaling and / or translating the ambient temperature in the first input data set; and / or rotating, translating and / or scaling the longitude and latitude in the first input data set.

[0029] A de-icing vehicle milling cutter height prediction system, comprising:

[0030] A data acquisition module is used to collect the current ambient temperature, the longitude and latitude of the current location of the de-icing vehicle, and the current road ice thickness within a preset range in front of the de-icing vehicle;

[0031] The analysis module is used to construct a road surface ice layer thickness matrix according to the road surface ice layer thickness, and output a milling cutter height prediction value according to the road surface ice layer thickness matrix, the ambient temperature and the longitude and latitude.

[0032] A deicing vehicle comprises a vehicle body and a deicing vehicle cutter height prediction system as claimed in claim 9 arranged on the vehicle body.

[0033] The beneficial effects of the present invention are:

[0034] The present invention provides a method for predicting the deicing height of a milling cutter of a deicing vehicle. The method predicts the optimal height at which the milling cutter will cut into the ice layer, i.e., the operating height, according to the real-time detected road ice layer thickness, the current ambient temperature, and the longitude and latitude of the deicing vehicle. The method enables the milling cutter to efficiently remove ice and snow on the road surface during operation. The method combines a variety of factors that affect the state of the ice layer, thereby achieving the best deicing and snow removal effect of the milling cutter, while avoiding damage to the road surface and improving operation safety.

[0035] The present invention also provides a method for training a milling cutter height prediction model, which trains the initial neural network model for milling cutter height prediction through training samples to obtain a milling cutter height prediction model that can predict the optimal height of the milling cutter penetrating the ice layer based on the road ice thickness matrix, current ambient temperature and longitude and latitude, thereby improving the efficiency, effect and safety of deicing operations.

[0036] The present invention also provides a deicing vehicle cutter height prediction system, which collects data in real time through a data acquisition module, and then analyzes and processes the data through an analysis module to predict the cutter height, thereby improving the efficiency, effect and safety of deicing operations.

[0037] The present invention also provides an ice-removing vehicle, which has high ice-removing efficiency, good ice-removing effect and low probability of damaging the road surface. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the contents of the embodiments of the present invention and these drawings without paying any creative work.

[0039] Figure 1 is a schematic diagram of a de-icing vehicle cutter height prediction system provided by an embodiment of the present invention;

[0040] Figure 2 is a schematic diagram of a data acquisition module provided by an embodiment of the present invention;

[0041] Figure 3 is a flow chart of a method for predicting deicing height of a deicing vehicle milling cutter provided in an embodiment of the present invention;

[0042] Figure 4 is a schematic diagram of a preset unit range in a preset range provided by an embodiment of the present invention;

[0043] Figure 5 It is a flow chart of the milling cutter height prediction model training method provided by an embodiment of the present invention.

[0044] In the figure:

[0045] 1. Data acquisition module; 2. Analysis module; 3. Control module; 4. Drive module; 5. Storage module; 11. GPS positioning device; 12. Temperature sensor; 13. LiDAR. DETAILED DESCRIPTION

[0046] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0047] In the embodiments of the present application, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, "plurality" means two or more.

[0048] The term "at least one" in this application means one or more, and the term "multiple" in this application means two or more, for example, multiple second messages means two or more second messages. The terms "system" and "network" are often used interchangeably herein.

[0049] It should be understood that the terms used in the description of the various examples herein are only for describing specific examples and are not intended to be limiting. As used in the description of the various examples and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0050] It should also be understood that the term "and / or" used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. The term "and / or" is a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this application generally indicates that the associated objects before and after are in an "or" relationship.

[0051] It should also be understood that in the various embodiments of the present application, the size of the serial number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0052] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0053] It should also be understood that the term “comprise” (also known as “includes,” “including,” “comprises” and / or “comprising”) when used in this specification specifies the presence of stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0054] It should also be understood that the term "if" may be interpreted to mean "when" or "upon" or "in response to determining" or "in response to detecting." Similarly, the phrase "if it is determined that ..." or "if [a stated condition or event] is detected" may be interpreted to mean "upon determining that ..." or "in response to determining that ..." or "upon detecting [a stated condition or event]" or "in response to detecting [a stated condition or event]," depending on the context.

[0055] It should be understood that the “one embodiment”, “an embodiment”, and “a possible implementation method” mentioned throughout the specification are

[0056] It means that a specific feature, structure or characteristic related to an embodiment or implementation is included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" or "a possible implementation" appearing in various places throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner.

[0057] In extreme weather, roads will accumulate snow and ice, which greatly affects travel efficiency and safety. In traditional road deicing, most of the deicing is done by spreading salt, snow-melting agents or thermal deicing. Spreading salt will have a certain negative impact on the environment, spreading snow-melting agents is limited in application, and thermal deicing consumes a lot of energy and is more suitable for small-surface deicing. At present, in order to improve the efficiency and effect of deicing, mechanical methods are used, such as deicing machines or deicing vehicles, to save labor and time. However, in actual work, the existing deicing machines or deicing vehicles are prone to incomplete deicing or damage to the road surface because the initial cutting height of the milling cutter is relatively random or relies on manual experience, which brings considerable negative obstacles to deicing.

[0058] like Figure 1 and Figure 2 As shown, based on the above technical problems, the present embodiment provides an ice-removing vehicle, which uses a milling cutter on the ice-removing vehicle to remove ice and snow on the road. Specifically, the ice-removing vehicle includes a vehicle body and a milling cutter height prediction system for the ice-removing vehicle arranged on the vehicle body. When the ice and snow are removed by the milling cutter, the optimal cutting height of the milling cutter into the ice layer is predicted by the milling cutter height prediction system of the ice-removing vehicle, so that the ice and snow on the road can be removed at one time, while ensuring that the milling cutter will not cause damage to the road surface, that is, to improve work efficiency, safety and reliability.

[0059] In this embodiment, the de-icing vehicle milling cutter height prediction system includes a data acquisition module 1 and an analysis module 2. The data acquisition module 1 is used to collect the current ambient temperature, the longitude and latitude of the current location of the de-icing vehicle, and the current road ice thickness within a preset range in front of the de-icing vehicle. The analysis module 2 is used to construct a road ice thickness matrix based on the road ice thickness, and output the milling cutter height prediction value based on the road ice thickness matrix, the ambient temperature, and the longitude and latitude. When performing ice and snow removal operations, the current ambient temperature is collected through the data acquisition module 1, and the melting state of ice and snow on the current pre-treated road surface can be understood in real time based on the ambient temperature, so that the milling cutter cutting height can be accurately predicted in combination with the collected road ice thickness; at the same time, the longitude and latitude of the location of the de-icing vehicle, that is, the longitude and latitude of the location of the de-icing vehicle, are collected through the data acquisition module 1 to ensure that the collected data has geographic reference and improve the prediction accuracy.

[0060] The data acquisition module 1 includes a laser rangefinder sensor, a GPS positioning device 11 and a temperature sensor 12. The laser rangefinder sensor, namely a laser radar 13, is installed at the front or bottom of the deicing vehicle and is used to collect the road surface height in real time and generate a road surface height map. The laser radar 13 measures the laser reflection time by emitting a laser beam and calculates the distance between the road surface and the sensor, thereby obtaining accurate road surface information. In order to obtain the thickness of the ice layer, the laser radar 13 is used to measure the original height of the road surface in advance and store the measurement data; during the deicing operation, the laser radar 13 is used to measure the height of the ice layer in real time, and the difference between the ice layer height and the original road surface height is the ice layer thickness. The GPS positioning device 11 is installed on the top of the deicing vehicle and is used to accurately locate the geographical location of the deicing vehicle. The GPS positioning device 11 provides real-time location information by receiving satellite signals to ensure that the collected data has a geographical reference, so that an accurate road surface model can be established in different sections. The temperature sensor 12 is installed at the front of the deicing vehicle and is used to collect ambient temperature data. The temperature sensor 12 assists in judging the road surface conditions, such as the melting state and thickness changes of ice and snow, and provides additional parameters for the analysis module 2.

[0061] Preferably, the data acquisition module 1 can also collect road types, such as highway types, rural road types, bridge and elevated road types, etc., and use the road type as one of the prediction conditions to further improve the accuracy of the milling cutter height prediction. The road type can be manually judged and input, or it can be obtained through the GPS positioning device 11, and the road type is preset with a number, such as highway type 1, rural road type 2, bridge and elevated road type 3, as a subsequent prediction factor input into the deicing vehicle milling cutter height prediction system.

[0062] Among them, the analysis module 2 analyzes and processes the data collected by the data acquisition module 1 through the preset trained milling cutter height prediction model to obtain the milling cutter height prediction value. Specifically, the milling cutter height prediction model includes a convolutional neural network and a fully connected neural network. The data acquisition module 1 collects the road type, the current ambient temperature, the longitude and latitude of the location of the de-icing vehicle, and the current road ice thickness in real time. The analysis module 2 receives these data and establishes a road ice thickness matrix based on the road ice thickness. The convolutional neural network extracts the features of the road ice thickness matrix, and then flattens the extracted features into a first vector. The temperature, longitude and latitude, and road type are merged into the first vector to obtain the second vector, and then the fully connected neural network extracts the features, and finally outputs the milling cutter height prediction value, which is the optimal cutting height of the milling cutter.

[0063] In order to facilitate direct adjustment of the milling cutter height, the analysis module 2 is also used to output the milling cutter height adjustment value according to the current height of the milling cutter and the predicted value of the milling cutter height prediction model. In this embodiment, the deicing vehicle also includes a control module 3 and a drive module 4. The control module 3 is used to control the drive module 4 to drive the milling cutter to move according to the analysis result of the analysis module 2, that is, after the analysis module 2 outputs the milling cutter height adjustment value to the control module 3, the control module 3 controls the drive module 4, and the drive module 4 drives the milling cutter to move according to the milling cutter height adjustment value.

[0064] The driving module 4 is a hydraulic driving component or an electric actuator, and the control module 3 drives the milling cutter to rise and fall by controlling the hydraulic driving component or the electric actuator. The hydraulic driving component has a large thrust and precise control capability, and the electric actuator is more flexible and easy to maintain.

[0065] After receiving the height adjustment value of the milling cutter from the analysis module 2, the control module 3 calculates the required hydraulic pressure or electric force, and the drive module responds to adjust the height of the milling cutter. For example, the height adjustment value of the milling cutter output by the analysis module 2 is +5 or -5 cm, and the control module controls the drive module to drive the milling cutter to rise 5 cm or fall 5 cm.

[0066] The data acquisition module 1 collects the current data in real time, and the analysis module 2 performs real-time analysis and the control module 3 performs real-time control, so that the height of the milling cutter can be adjusted in real time to ensure the accuracy and working reliability of the milling cutter during operation.

[0067] Preferably, the deicing vehicle cutter height prediction system further includes a storage module 5, which is used to store the collected data of the data collection module 1 and the output data of the analysis module 2. For example, the original road surface height, the ice surface height before deicing, the ice surface height during deicing, the road surface height after deicing, the position information and the temperature information collected by the data collection module 1 are stored to form a complete database, which can be used for statistical analysis of deicing effects and can also be used as historical data for reference in subsequent deicing tasks.

[0068] It is understandable that the deicing vehicle cutter height prediction system can be selectively placed on the body of the deicing vehicle, or installed on other deicing devices, such as a deicer, to meet different needs.

[0069] The method provided in the embodiments of the present application is described below with reference to the accompanying drawings.

[0070] refer to Figure 3 , Figure 3 The schematic diagram of the flow chart of the deicing height prediction method of the deicing cutter provided in the embodiment of the present application is shown. The deicing height prediction method of the deicing cutter of the deicing cutter is used to predict the optimal height of the cutter to cut into the ice layer.

[0071] The method may include the following steps:

[0072] S100: Obtaining the current road ice thickness in a preset range in front of the deicing vehicle, wherein reference Figure 4 , the preset range includes a plurality of preset unit ranges divided along the width direction of the road and along the length direction of the road;

[0073] During deicing, the original height of the road surface within the preset range in front of the current deicing vehicle is obtained, and the laser radar 13 set on the deicing vehicle is used to measure the road surface height within the preset range. The road surface height here is the ice surface height, and the ice layer thickness point cloud within the preset range is obtained according to the original road surface height and the current ice surface height. That is, each point in the point cloud contains the position coordinates of the point relative to the laser radar and the ice layer thickness at the point.

[0074] It should be understood that the width of the preset range refers to the length in the direction consistent with the road width, and the length of the preset range refers to the distance from the cutting position of the milling cutter of the deicing vehicle to the driving direction of the deicing vehicle to the preset length, or from the preset distance from the cutting position of the milling cutter to the driving direction of the deicing vehicle to the preset length. The preset length is usually equal to the width of the road to be deiced or the width of the milling cutter.

[0075] S200: constructing a road surface ice thickness matrix based on the road surface ice thickness, wherein the road surface ice thickness matrix is ​​a two-dimensional matrix, wherein the horizontal sequence in the road surface ice thickness matrix is ​​the ice thickness within each preset unit range along the road width direction, and the vertical sequence is the ice thickness within each preset unit range along the road length direction;

[0076] Among them, a two-dimensional matrix is ​​used to convert the ice thickness point cloud into a road surface ice thickness matrix.

[0077] As an example, the road width of the road to be de-iced is 6.4m, the preset range is the 6.4m×6.4m road surface in front of the de-icing vehicle, a 64×64 road surface ice thickness matrix is ​​pre-established, and each 10cm×10cm road surface range is a preset unit range. The ice thickness within each preset unit range is calculated by the point cloud matrix conversion method, that is, the road surface ice thickness matrix is ​​obtained, the horizontal sequence is the ice thickness within each preset unit range along the road width direction, and the vertical sequence is the ice thickness within each preset unit range along the road length direction. The point cloud matrix conversion method is described below:

[0078] The laser radar 13 measures the road surface height within a preset range to generate three-dimensional point cloud data, where each point includes (x, y, z) coordinates, x, y represent the position of the point on the horizontal plane (i.e., the distance of the point from the laser radar in the x and y directions, where the x and y directions can be understood as the width and length directions of the road), and z represents the thickness of the ice layer at the point. The three-dimensional point cloud data is projected onto a two-dimensional plane, i.e., the influence of the z coordinate is ignored, and x, y are retained as plane coordinates; the projected point cloud is mapped onto a preset fixed two-dimensional grid, such as Figure 4 As shown, for example, a 64×64 matrix (i.e. Figure 4 In the figure, n is 64), each cell of the matrix (i, j in the figure) represents a certain physical area, and the grid size is defined. For example, each 10cm×10cm road surface range is a preset cell range; the x and y coordinates of each point are converted into a grid index, and the grid index i and j of the point are determined by the following formula:

[0079] i=floor((x-xmin) / Δx)

[0080] j = floor((y-ymin) / Δy)

[0081] Among them, x min and min is the minimum coordinate within the preset range (for example, the lower left corner of the preset range area in front of the de-icing vehicle); Δx and Δy are the size of the grid unit (that is, the size of the physical space represented by each unit, 1cm or other values); floor(·) means rounding off, that is, rounding the coordinate to the nearest grid unit;

[0082] For each point, after calculating its corresponding grid unit position i and j, the thickness value of the point is filled into the corresponding position h(i, j) of the matrix, which is h described later. xiyi ;

[0083] It should be noted that if multiple points fall in the same grid unit, the final thickness value of the grid can be determined by taking the average value or the maximum value, for example, h(i,j)=(z 1 +z 2 +...+z m ) / m; where m is the number of points falling on the grid and z is the ice thickness value at the corresponding point.

[0084] Specifically, in the two-dimensional matrix generated above, there may be some empty cells (no lidar points fall into these grids). In order to fill these empty cells, the nearest neighbor interpolation method is used:

[0085] For each empty grid cell h(i, j), find the nearest non-empty cell;

[0086] The distance to the nearest neighbor can be determined by calculating the Euclidean distance between two grid cells:

[0087] d = sqrt((i1-i2)^2+(j1-j2)^2)

[0088] Where (i 1 , j 1 ) is the position of the current empty cell, (i 2 , j 2 ) is the position of some non-empty cell;

[0089] After finding the nearest non-empty cell, fill its ice thickness value into the current empty cell;

[0090] After nearest neighbor interpolation, all grid cells have been filled, and the final matrix is ​​the road ice thickness matrix:

[0091] Among them, h xiyi is the ice thickness within the i-th and j-th preset units within the preset range.

[0092] S300: Obtaining the current ambient temperature and the longitude and latitude of the current location of the de-icing vehicle;

[0093] Optionally, the current ambient temperature is detected in real time by a temperature sensor 12 disposed on the deicing vehicle, and the longitude and latitude of the deicing vehicle is detected in real time by a GPS positioning device 11. Of course, in other embodiments, the current ambient temperature can also be obtained in real time from the Internet, which is not limited here.

[0094] During de-icing operations, road conditions such as the thickness, hardness, and melting state of the ice and snow layer are affected by multiple factors. By real-time detection of the current ambient temperature, the degree of melting and hardness of the ice and snow can be obtained in real time, and the longitude and latitude can ensure that the data obtained in different geographical environments can be correctly mapped to specific road sections, facilitating decision-making and analysis of historical road data. Combining different factors such as temperature, location longitude and latitude, and ice thickness can improve the prediction accuracy of the optimal height for the milling cutter to cut into the ice layer. For example, in low temperatures and thick snow, the milling cutter may need to cut deeper, and in a state close to melting, the milling cutter may need to work at a higher height to avoid damaging the road surface.

[0095] Preferably, the factors affecting the height of the milling cutter also include adding road types, that is, during deicing, the road type can also be obtained in real time, such as highway type, rural road type, bridge and elevated road type, etc., and the road type can be used as one of the prediction conditions to further improve the accuracy of the milling cutter height prediction. Among them, the road type can be manually judged and input, or it can be obtained through the GPS positioning device 11, and the road type is preset with a number, such as highway type 1, rural road type 2, bridge and elevated road type 3, as a subsequent prediction factor input into the deicing vehicle milling cutter height prediction system.

[0096] Exemplarily, the temperature affects the milling cutter operating height as follows: When the temperature is between 0℃ and -1℃, there may be supercooled water, that is, some ice and snow are still in liquid state; the surface begins to freeze, but the overall snow structure is not completely frozen. At this temperature, the ice and snow may partially melt, especially under traffic pressure or mechanical action. At this time, it is necessary to slightly adjust the milling cutter height to avoid damaging the road surface and maintain the appropriate height of the milling cutter to clean up local water accumulation and partially frozen ice and snow. When the temperature is between -1℃ and -5℃, the ice and snow gradually freeze and harden, and the compacted snow layer becomes more solid, but it does not reach the extremely hard state, and there is almost no melting. The road surface remains covered with ice and snow, and may be partially loosened under mechanical action. At this time, when clearing the snow layer, keep the milling cutter at a low height to effectively clean the frozen road surface snow layer. When the temperature is between -5°C and -10°C, the snow and ice are completely frozen, and the snow becomes very solid, similar to frozen ice. The surface is compacted and frozen, and there is no melting. The snow and ice remain frozen. The milling cutter should be kept at a low height to remove the hard ice layer, and other equipment (such as heaters) may be needed to assist in cleaning. When the temperature is between -10°C and -20°C, the snow and ice are extremely frozen, and the snow becomes very brittle. Although hard, it is easy to break into small pieces under mechanical action. There is no melting. The snow is in an extremely stable frozen state. In a low temperature environment, the milling cutter should be kept at a low height to clean the brittle ice and snow layer, and other equipment may be needed to handle the hard snow. When the temperature is below -20°C, in the extremely frozen state, the snow is very fragile and easy to break, but it remains hard and frozen, and there is no melting. The milling cutter should be kept at a low height to handle the fragile ice layer, and pay attention to avoid excessive damage to the road surface. When the temperature is between 0℃ and 1℃, the ice and snow begin to melt slowly, and there may be some water accumulation or slippery areas on the road surface. The ice and snow may melt partially under the action of local environmental pressure (such as traffic) or mechanical operation (such as milling cutter). The height of the milling cutter needs to be adjusted frequently to ensure that it responds to the melting of ice and snow in different sections of the road to avoid excessive damage to the road surface. When the temperature is between 1℃ and 3℃, the ice and snow begin to melt gradually, the thickness decreases, the road surface becomes slippery and the water accumulation increases. The ice and snow melt slowly, melt water forms on the surface, and the snow accumulation gradually decreases. The height of the milling cutter should be appropriately lowered to ensure that the residual snow layer can be thoroughly cleaned and the damage to the road surface is reduced. When the temperature is between 3℃ and 5℃, the melting of ice and snow accelerates, the road surface begins to be exposed and there is more water accumulation, the melting is accelerated, the snow accumulation is significantly reduced, and the surface is exposed. As the melting accelerates, the height of the milling cutter can be gradually increased to reduce unnecessary wear on the road surface, but it is still necessary to ensure that the snow is cleared. When the temperature is above 5°C, the ice and snow disappear quickly, the road surface is exposed, and a small amount of residual water accumulates. The ice and snow melt completely, and the residual water evaporates or forms water accumulation. At this time, the milling cutter's work is focused on dealing with the residual water, keeping the equipment clean, and adjusting it to a higher height when necessary to avoid additional damage to the road surface.

[0097] Exemplarily, the influence of location information on the working height of the milling cutter is as follows: in mountainous areas, due to the high altitude, relatively low temperature, frequent snowfall, thick snow, and more complicated deicing work, the mountainous area has a cold climate, large temperature difference, uneven road surface, and snow easily accumulates into ice; during deicing, the roads in the mountains may require more milling cutter adjustments because the road surface is undulating and the thickness of ice and snow is uneven. In addition, the mountainous roads are tortuous, and more accurate location information is needed for automatic adjustment to avoid excessive wear of the milling cutter. When deicing in the mountains, the milling cutter should adjust its height according to the slope and snow thickness of different sections of the road to ensure effective handling of the thick ice layer and avoid damage to the equipment due to uneven road surface. In plain areas, snowfall is relatively stable and the road surface is relatively flat, so snow accumulation is usually more uniform, and de-icing is relatively easy. The road surface in plain areas is flat, snow accumulation is relatively uniform, and temperature changes are relatively mild. Due to the flat terrain, the thickness of snow is generally more consistent. The role of location information is more reflected in the combination of historical data, such as special adjustments when there is more snow in certain areas. In plain areas, the height of the milling cutter can be relatively fixed and only needs to be adjusted according to the thickness of the snow. Location information can help the system identify sections of roads that often accumulate snow and make height adjustments in advance. In urban areas, de-icing work in urban areas is more complicated because of the dense road network and large flow of people. De-icing work must not only ensure the removal of snow, but also avoid damage to urban facilities. The road network in urban areas is complex, there are many buildings, dense traffic, and uneven snow accumulation and melting speed. In urban areas, location information can help the system identify different road types (such as main roads, secondary roads, and sidewalks), so as to make different adjustments to the height of the milling cutter. Snow on main roads is usually cleared quickly, while secondary roads and sidewalks may require more sophisticated de-icing strategies; in cities, the milling cutter needs to adjust its height frequently to ensure that the main roads can be de-iced quickly, and in places with heavy traffic, the milling cutter needs to avoid damaging ground facilities (such as gutters and manhole covers). In coastal areas, the climate in coastal areas is relatively humid, and the temperature is relatively high due to the influence of sea breeze, but it is also more likely to have a mixture of rain and snow, which makes de-icing work challenging; the climate in coastal areas is humid, and it is easy to have a mixture of ice and snow, and the wind speed is high, which makes it difficult for snow to accumulate. Location information can help the system identify special weather conditions in coastal areas and adjust the height of the milling cutter based on data such as wind speed and humidity. Due to the less snow in coastal areas, the height of the milling cutter needs to be adjusted more flexibly during the de-icing process. The height of the milling cutter needs to be flexibly adjusted according to the changes in wind speed and humidity along the coast, especially when there is snow, it needs to be cleared quickly to avoid the formation of ice.

[0098] For example, on highways, which are usually wide and straight, and are the type of roads with heavy traffic, deicing work needs to ensure traffic safety. The thickness of snow on highways may be uneven, and snow is easily blown away when the wind speed is high. Vehicles traveling at high speeds will make the road surface more prone to ice. On highways, the milling cutter should maintain a relatively fixed height, but in sections with thick snow or prone to ice, the height needs to be adjusted quickly to ensure the flatness of the road surface. On rural roads, rural roads are usually narrow and tortuous, with complex snow accumulation conditions and thick snow layers. The milling cutter work on rural roads needs to be more flexible, especially when clearing narrow bends and irregular roads. The height of the milling cutter needs to be dynamically adjusted according to the position information to ensure cleaning efficiency and road safety. Bridges and elevated roads are more susceptible to icing because the surface temperature of these structures is usually lower than the surrounding air temperature, resulting in the rapid formation of snow and ice. The milling cutter height and frequency can be adjusted in advance on these sections to cope with roads that are more prone to icing. On bridges and elevated sections, the milling cutter should be kept at a low height and the ice should be cleared regularly to avoid traffic accidents caused by excessively thick ice.

[0099] In this embodiment, the data acquisition module's acquisition frequency, the milling cutter's adjustment frequency, the milling cutter's minimum and maximum height limits, etc. can be pre-set based on the collected road type, location information, ambient temperature information, etc., to ensure optimal operational accuracy and safety.

[0100] S400: Predicting the optimal ice-cutting height of the milling cutter on the deicing vehicle according to the road ice thickness matrix, ambient temperature, and longitude and latitude. Of course, the optimal ice-cutting height of the milling cutter on the deicing vehicle can also be predicted according to the road type, road ice thickness matrix, ambient temperature, and longitude and latitude, and the specific setting is based on actual needs.

[0101] Specifically, the cutter height prediction model preset on the de-icing vehicle predicts the optimal cutting height of the cutter into the ice layer. That is, through the data acquisition module or by inputting the data of the above-mentioned influencing factors, the cutter height prediction model can output the cutter height prediction value.

[0102] It should be understood that the prediction can also be made by a milling cutter height prediction model that is not pre-installed on the de-icing vehicle. The data acquisition module or the data of the above-mentioned influencing factors is transmitted to the network terminal. The network terminal is provided with a milling cutter height prediction model. The milling cutter height prediction model predicts the optimal height of the milling cutter to cut into the ice layer.

[0103] The cutter height prediction model includes a convolutional neural network and a fully connected neural network. The convolutional neural network is used to extract the characteristics of the road ice thickness matrix. The cutter height prediction model preset on the deicing vehicle predicts the optimal height of the cutter to cut into the ice layer, including:

[0104] S401: inputting the road surface ice thickness matrix into the milling cutter height prediction model, and extracting the features of the road surface ice thickness matrix through a convolutional neural network;

[0105] Specifically, the convolutional neural network includes an input layer, a convolution layer, a pooling layer and an output layer, wherein the number of convolution layers is, for example, 2 or 3 layers, and the number of pooling layers is, for example, 2 or 3 layers, which is specifically set according to actual needs or determined through iterative training optimization.

[0106] For example, the input road ice thickness matrix is ​​as follows:

[0107] Among them, h xiyi is the ice thickness within the i-th and j-th preset units within the preset range.

[0108] The features of the road ice thickness matrix are extracted through 16 different convolution windows with a kernel size of 3 and a step size of 1. Among them, the convolution kernel size of 3 means that the convolution kernel is a 3×3 matrix. The values ​​in the matrix are the weights of the convolution layer, and the convolution layer also has a bias parameter. After one convolution layer, 16 stacked first intermediate two-dimensional matrices are obtained, that is, the first tensor (n-2, 16). The amount of data in each first intermediate two-dimensional matrix is ​​smaller than the road ice thickness matrix, which is specifically determined by the size and step size of the convolution kernel.

[0109] The 16 stacked first intermediate two-dimensional matrices are pooled through the first pooling layer. For example, maximum pooling is selected to pool each first intermediate two-dimensional matrix. The optional pooling size is 2 and the step size is 2. The data amount of the first tensor is reduced by pooling to obtain the second tensor ((n-2) / 2, 16).

[0110] The second tensor after pooling is input into the second convolution layer, and features are extracted from the second tensor through 32 different convolution windows with a kernel size of 3 and a step size of 1, obtaining 32 stacked second intermediate two-dimensional matrices, i.e., the third tensor ((n-4) / 2, 32).

[0111] The third tensor is pooled through the second pooling layer. For example, the maximum pooling is selected to pool each second intermediate two-dimensional matrix. The optional pooling size is 2 and the step size is 2. The data amount of the third tensor is reduced by pooling to obtain the fourth tensor ((n-4) / 4, 32).

[0112] The fourth tensor after pooling is input into the third convolution layer, and features are extracted from the fourth tensor through 64 different convolution windows with a kernel size of 3 and a step size of 1, obtaining 64 stacked third intermediate two-dimensional matrices, i.e., the fifth tensor ((n-6) / 4, 64).

[0113] The fifth tensor is pooled through the third pooling layer. For example, the maximum pooling is selected to pool each third intermediate two-dimensional matrix. The optional pooling size is 2 and the step size is 2. The data amount of the fifth tensor is reduced by pooling to obtain the sixth tensor ((n-6) / 8, 64).

[0114] S402: Flattening the features of the road ice thickness matrix extracted by the convolutional neural network into a first vector;

[0115] The sixth tensor is flattened into a first vector, and the length of the first vector is 64×[(n-6) / 8].

[0116] S403: The ambient temperature and longitude and latitude are added to the first vector to form a second vector, and a fully connected neural network extracts features of the second vector and outputs a predicted height of the milling cutter, which is the optimal height for the milling cutter to cut into the ice layer.

[0117] Among them, the second vector is obtained by merging the ambient temperature and longitude and latitude, and the length of the second vector is 64×[(n-6) / 8]+3. Of course, the road type can also be added, and the length of the second vector is 64×[(n-6) / 8]+4.

[0118] For example, the number of layers of the fully connected neural network is 2, the number of neurons in the input layer is 64×[(n-6) / 8]+3 or 64×[(n-6) / 8]+4; the number of output neurons in the first layer is 128, the activation function is ReLU, and the formula is y 1 =ReLU(W 1 *x+b 1 ), where x refers to the input data, i.e., the second vector, W 1 is the weight of the first layer, b 1 is the bias; the number of input neurons in the second layer is 128, the number of output neurons is 64, the activation function is ReLU, and the formula is y 2 =ReLU(W 2 *y 1 +b 2 ), where W 2 is the weight of the second layer, b 2 is the bias; the number of input neurons in the output layer is 64, the number of output neurons is 1, there is no activation function, and the formula is y 3 =W 3 *y 2 +b 3 , where y 3 is the predicted value of the final milling cutter height, W 3 is the weight of the output layer, b 3 For bias.

[0119] After the predicted height of the milling cutter is obtained through the milling cutter height prediction module, the height of the milling cutter can be adjusted to the target working height. In order to facilitate adjustment, it also includes:

[0120] S500: Obtain the current height of the milling cutter, and obtain a predicted adjustment value of the milling cutter according to the current height of the milling cutter and the predicted optimal height of the milling cutter for cutting into the ice layer.

[0121] For example, if the predicted height of the milling cutter is +5 or -5 cm, the milling cutter is controlled to rise 5 cm or fall 5 cm.

[0122] It should be understood that the height adjustment frequency of the milling cutter can be set in advance, that is, the frequency of the data acquisition module collecting data and the frequency of the analysis module outputting the predicted value of the milling cutter height can be set in advance. For example, after the first prediction, the milling cutter works and the deicing vehicle advances the first preset distance, such as 1m, 2m or 3m, etc., and then the data acquisition module re-collects data, and the analysis module outputs the predicted value of the milling cutter height again. After adjusting the milling cutter height, the milling cutter works and the deicing vehicle advances the second preset distance again. The frequency can be adjusted in real time according to road conditions and other conditions to ensure the best deicing effect and efficiency.

[0123] The deicing height prediction method for the deicing vehicle milling cutter provided in this embodiment can efficiently and quickly obtain the optimal operating height of the milling cutter, so that the milling cutter can efficiently and accurately remove the ice and snow layer or the accumulated water layer during operation, while also preventing the milling cutter from damaging the road surface.

[0124] like Figure 5 As shown, this embodiment also provides a milling cutter height prediction model training method, through which the milling cutter height prediction model is trained, so that the milling cutter height prediction model can quickly and accurately predict the milling cutter operating height during actual application.

[0125] Specifically, the milling cutter height prediction model training method includes:

[0126] S10: Obtain training samples, the training samples include a first input data set and a first output data set, the data in the first output data set are several first preset cutting-in heights of the milling cutter, wherein after the milling cutter de-ices at the first preset cutting-in height, the ice thickness of the road surface is within the preset thickness range and the road surface is not damaged; the data in the first input data set are the ambient temperature corresponding to the milling cutter working at each first preset cutting-in height, the longitude and latitude of the location of the de-icing vehicle, the road type, and the road surface ice thickness matrix, wherein the road surface ice thickness matrix is ​​constructed based on the current road surface ice thickness within a preset range in front of the de-icing vehicle, the preset range includes several preset unit ranges divided along the width direction of the road and along the length direction of the road, the road surface ice thickness matrix is ​​a two-dimensional matrix, the horizontal sequence in the road surface ice thickness matrix is ​​the ice thickness within each preset unit range along the width direction of the road, and the vertical sequence is the ice thickness within each preset unit range along the length direction of the road.

[0127] Among them, the acquisition of the first input data set and the first output data set can be data collected by the de-icing vehicle during the actual working process, such as the data stored in the storage module mentioned above, or it can be data obtained through simulation tests. For example, different road types, different temperatures, different ice thicknesses, etc. in different location environments are simulated on a test road, and the optimal height of the milling cutter cutting into the ice layer under different conditions is obtained through experiments.

[0128] The establishment of the road ice thickness matrix refers to the above description and will not be repeated here.

[0129] Preferably, in order to expand the training samples and obtain more effective and reliable data, it is also necessary to process the data in the first input data set, including performing data enhancement processing on the data in the first input data set to obtain several second input data sets, each second input data set corresponding to the same output data set, and the data enhancement processing includes rotating, translating and / or scaling the road ice thickness matrix in the first input data set; and / or adding noise, scaling and / or translating the ambient temperature in the first input data set; and / or rotating, translating and / or scaling the longitude and latitude in the first input data set.

[0130] It should be understood that rotating the ice thickness matrix can change the spatial relationship of the data and improve the robustness of the milling cutter height prediction model to different viewing angles. Scaling can simulate different distance scales and enhance the adaptability of the milling cutter height prediction model to scale changes. Translation can simulate different position offsets and improve the positioning robustness of the milling cutter height prediction model.

[0131] It should be understood that by rotating the position information, the position of the de-icing vehicle in different directions can be simulated, by scaling the position information or temperature information, different geographical range scaling and temperature range changes can be simulated, and by translating the position information or temperature information, the position offset and systematic temperature offset can be simulated.

[0132] Preferably, the data processing further includes adding noise to the data to increase the randomness of the data and simulate the measurement error in a real environment, such as adding Gaussian noise to the temperature data.

[0133] Preferably, the data processing also includes data denoising and outlier processing, wherein the data in the first input data set in the original training sample is denoised by a moving average method or a weighted average method, and the outlier processing uses a Z-score method or an IQR method to identify outliers, and after identifying the outliers, the outliers are removed from the training sample. Since the moving average method and the weighted average method for data denoising and the Z-score method and the IQR method for identifying abnormal data are relatively mature existing technologies, they will not be described one by one here.

[0134] Preferably, the data processing also includes standardization of the data, for example, standardization of data samples such as ice thickness, temperature, longitude and latitude, etc.:

[0135] Normalize for ice thickness:

[0136] Compute the maximum and minimum values ​​in the sample data:

[0137] X_min=min(data)

[0138] X_max=max(data),

[0139] Normalization:

[0140] X_norm=(X-X_min) / (X_max-X_min)

[0141] Calculate the mean (μ) and standard deviation (σ) of the data:

[0142] μ=sum(data) / len(data)

[0143] σ=sqrt(sum((x-μ)**2forx in data) / len(data))

[0144] Data Standardization:

[0145] X_z=(X-μ) / σ;

[0146] Normalize the location data:

[0147] Normalize the longitude data:

[0148] Longitude_norm=(Longitude+180) / 360

[0149] Normalize the latitude data:

[0150] Latitude_norm=(Latitude+90) / 180

[0151] Calculate the mean (μ_longitude, μ_latitude) and standard deviation (σ_longitude, σ_latitude) of the longitude and latitude data:

[0152] μ_longitude=sum(longitude_data) / len(longitude_data)

[0153] σ_longitude=sqrt(sum((x-μ_longitude)**2for x in longitude_data) /

[0154] len(longitude_data))

[0155] μ_latitude=sum(latitude_data) / len(latitude_data)

[0156] σ_latitude=sqrt(sum((x-μ_latitude)**2for x in latitude_data) /

[0157] len(latitude_data))

[0158] Data Standardization:

[0159] Longitude_z=(Longitude-μ_longitude) / σ_longitude

[0160] Latitude_z=(Latitude-μ_latitude) / σ_latitude;

[0161] Standardize the temperature data:

[0162] Calculate the minimum value (T_min) and maximum value (T_max) in the data set:

[0163] T_min=min(data)

[0164] T_max=max(data)

[0165] Apply linear normalization:

[0166] T_norm=(T-T_min) / (T_max-T_min)

[0167] Calculate the mean (μ) and standard deviation (σ) of the temperature data:

[0168] μ=sum(data) / len(data)

[0169] σ=sqrt(sum((x-μ)**2forx in data) / len(data))

[0170] Data Standardization:

[0171] T_z=(T-μ) / σ.

[0172] S20: Construct an initial neural network model for predicting the milling cutter height, use training samples to train the initial neural network model for predicting the milling cutter height, use mean square error as the loss function, optimize and update the parameters in the initial neural network model for predicting the milling cutter height, wherein the parameters include weight parameters and bias parameters, and obtain a milling cutter height prediction model. The milling cutter height prediction model is used to predict the optimal height of the milling cutter for cutting into the ice layer according to the road ice thickness matrix, the current ambient temperature, and the longitude and latitude.

[0173] Among them, the structure of the initial neural network model for predicting the milling cutter height refers to the milling cutter height prediction model described in the above-mentioned deicing vehicle milling cutter deicing height prediction method. During training, the convolution kernel size, convolution step size, bias, activation function, pooling layer number, size, pooling step size, and activation function, weights and biases, and number of layers of the fully connected neural network are given, and the training sample is brought in. The convolution neural network extracts the characteristics of the road ice thickness matrix and outputs the first vector. The fully connected layer extracts the first vector, ambient temperature, road type, and longitude and latitude characteristics, and outputs the predicted height of the milling cutter; the predicted height of the milling cutter is compared with the data in the first output data set, and the mean square error is used as the loss function. The optimization algorithms such as stochastic gradient descent (SGD) or adaptive moment estimation (Adam) are used to update the parameters, and the parameters in the convolution neural network, such as the convolution kernel size, bias, weights and bias of the fully connected neural network, are iteratively optimized to obtain a milling cutter height prediction model with higher accuracy.

[0174] In order to ensure that the final milling cutter height prediction model has a high prediction accuracy, a verification data set and a test data set are also constructed. The method of obtaining the verification data set and the test data set refers to S10 and will not be repeated here. For example, 100 copies of data are obtained, 60 of which are used as training samples, 20 as verification sets, and the remaining 20 as test sets. The data in the training samples, verification data set, and test data set do not intersect with each other.

[0175] The validation data set can select the milling cutter height prediction model with different hyperparameters, and select the milling cutter height prediction model with the highest accuracy as the final milling cutter height prediction model. The hyperparameters are, for example, the number of layers and the number of neurons of the convolutional neural network and the full neural network.

[0176] Among them, the test dataset is used to evaluate the generalization ability of the final milling cutter height prediction model.

[0177] Through the milling cutter height prediction model training method provided in this embodiment, the trained milling cutter height prediction model can reflect different current ambient temperatures, different longitudes and latitudes, and different road types in the prediction results. That is, in actual operations, based on the collected current ambient temperature, longitude and latitude, and road type, the predicted milling cutter height integrates multiple factors such as ambient temperature, location, and road type. The predicted milling cutter height is more accurate, so that the height at which the milling cutter cuts into the ice layer is optimal, which can not only ensure the best ice and snow removal effect, but also avoid damage to the road surface.

[0178] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for predicting the deicing height of a deicing vehicle milling cutter, characterized in that: include: Obtaining the current road ice thickness in a preset range in front of the de-icing vehicle, wherein the preset range includes a plurality of preset unit ranges divided along the width direction of the road and along the length direction of the road; Constructing a road surface ice layer thickness matrix based on the road surface ice layer thickness, wherein the road surface ice layer thickness matrix is ​​a two-dimensional matrix, the horizontal sequence in the road surface ice layer thickness matrix is ​​the ice layer thickness within each preset unit range along the road width direction, and the vertical sequence is the ice layer thickness within each preset unit range along the road length direction; Obtaining the current ambient temperature and the longitude and latitude of the current location of the de-icing vehicle; The optimal height at which the milling cutter on the deicing vehicle cuts into the ice layer is predicted according to the road ice layer thickness matrix, the ambient temperature and the longitude and latitude.

2. The deicing height prediction method of the deicing vehicle milling cutter according to claim 1 is characterized in that: Predicting the optimal height of the milling cutter on the deicing vehicle to cut into the ice layer according to the road ice thickness matrix, the temperature and the longitude and latitude includes: According to the road ice thickness matrix, the ambient temperature and the longitude and latitude, the optimal height of the milling cutter to cut into the ice layer is predicted by a milling cutter height prediction model preset on the deicing vehicle.

3. The deicing height prediction method of the deicing cutter of a deicing vehicle according to claim 2, characterized in that: The milling cutter height prediction model includes a convolutional neural network and a fully connected neural network, the convolutional neural network is used to extract the characteristics of the road ice thickness matrix, and the milling cutter height prediction model preset on the deicing vehicle predicts the optimal height of the milling cutter to cut into the ice layer, including: The road surface ice layer thickness matrix is ​​input into the milling cutter height prediction model, and the convolutional neural network extracts the features of the road surface ice layer thickness matrix; Flattening the features of the road ice thickness matrix extracted by the convolutional neural network into a first vector; The ambient temperature and the longitude and latitude are added to the first vector to form a second vector. The fully connected neural network extracts features of the second vector and outputs a predicted height of the milling cutter, where the predicted height of the milling cutter is the optimal height for the milling cutter to cut into the ice layer.

4. The deicing height prediction method of the deicing vehicle milling cutter according to claim 1 is characterized in that: Also includes: The current height of the milling cutter is obtained, and the predicted adjustment value of the milling cutter is obtained according to the current height of the milling cutter and the predicted optimal height of the milling cutter for cutting into the ice layer.

5. The deicing height prediction method of the deicing vehicle milling cutter according to claim 1 is characterized in that: The road ice thickness matrix is ​​constructed based on the road ice thickness within the preset range in front of the current de-icing vehicle, including: Obtaining the original height of the road surface within the preset range in front of the deicing vehicle, obtaining the current ice surface height within the preset range in front of the deicing vehicle through laser radar detection preset on the deicing vehicle, and obtaining the ice layer thickness point cloud within the preset range according to the original height of the road surface and the current ice surface height; The ice thickness point cloud is converted into the road ice thickness matrix using a two-dimensional matrix.

6. A milling cutter height prediction model training method, characterized in that: include: Acquire training samples, the training samples include a first input data set and a first output data set, the data in the first output data set are several first preset cutting-in heights of the milling cutter, wherein after the milling cutter de-ices at the first preset cutting-in height, the ice thickness of the road surface is within a preset thickness range and the road surface is not damaged; the data in the first input data set are the ambient temperature corresponding to the milling cutter working at each of the first preset cutting-in heights, the longitude and latitude of the location of the de-icing vehicle, and a road surface ice thickness matrix, wherein the road surface ice thickness matrix is ​​constructed based on the current road surface ice thickness within a preset range in front of the de-icing vehicle, the preset range includes several preset unit ranges divided along the width direction of the road and along the length direction of the road, the road surface ice thickness matrix is ​​a two-dimensional matrix, the horizontal sequence in the road surface ice thickness matrix is ​​the ice thickness within each preset unit range along the width direction of the road, and the vertical sequence is the ice thickness within each preset unit range along the length direction of the road; Construct an initial neural network model for predicting milling cutter height, use the training samples to train the initial neural network model for predicting milling cutter height, use mean square error as the loss function, optimize and update the parameters in the initial neural network model for predicting milling cutter height, wherein the parameters include weight parameters and bias parameters, and obtain the milling cutter height prediction model, which is used to predict the optimal height of the milling cutter cutting into the ice layer according to the road ice thickness matrix, the current ambient temperature and the longitude and latitude.

7. A milling cutter height prediction model training method according to claim 6, characterized in that: The initial neural network model for predicting the milling cutter height includes a convolutional neural network and a fully connected neural network, wherein the convolutional neural network is used to extract the features of the road ice thickness matrix and output a first vector, and the fully connected layer is used to extract the features of the first vector, the ambient temperature, and the longitude and latitude, and output the predicted milling cutter height; According to the predicted milling cutter height and the first preset cutting height, the parameters are updated through loss function optimization to obtain the milling cutter height prediction model.

8. A milling cutter height prediction model training method according to claim 6, characterized in that: Acquiring the training sample further includes: Data enhancement processing is performed on the data in the first input data set to obtain several second input data sets, each of which corresponds to the same output data set, and the data enhancement processing includes rotating, translating and / or scaling the road ice thickness matrix in the first input data set; and / or adding noise, scaling and / or translating the ambient temperature in the first input data set; and / or rotating, translating and / or scaling the longitude and latitude in the first input data set.

9. A de-icing vehicle milling cutter height prediction system, characterized in that: include: A data acquisition module is used to collect the current ambient temperature, the longitude and latitude of the current location of the de-icing vehicle, and the current road ice thickness within a preset range in front of the de-icing vehicle; The analysis module is used to construct a road surface ice layer thickness matrix according to the road surface ice layer thickness, and output a milling cutter height prediction value according to the road surface ice layer thickness matrix, the ambient temperature and the longitude and latitude.

10. A de-icing vehicle, characterized in that: The invention comprises a vehicle body and a de-icing vehicle cutter height prediction system as claimed in claim 9 which is arranged on the vehicle body.

Images