A method for converting railway multi-projection zone data

Abstract

The present invention relates to the technical field of engineering surveying, and specifically discloses a method for converting railway multi-projection zone data, including obtaining a coordinate data set of control points within the projection zone of the current railway map, and dividing data conversion partitions according to the coordinate data set of the control points; obtaining a coordinate data set of control points of the projection zone within the data conversion partition, establishing a data conversion neural network model according to the coordinate data set of the control points within the data conversion partition, performing coordinate data conversion on each point to be measured based on the data conversion neural network model, and obtaining the coordinate conversion value of the point to be measured; determining the model error value according to the coordinate conversion value of the point to be measured, and determining the final converted coordinate data sequence according to the model error value; splicing the final converted coordinate data sequence, and obtaining the final data conversion railway map according to the data splicing result. By establishing a data conversion neural network model and correcting the output result, the accuracy of railway multi-projection zone coordinate data conversion is improved.

Description

Technical Field

[0001] The present application relates to the technical field of engineering surveying, and more specifically, to a railway multi-projection band data conversion method. Background Art

[0002] China's latest national geodetic coordinate system 2000 (CGCS2000) is a high-precision geocentric dynamic geodetic coordinate system established by space technology. After its launch in July 2008, many geodetic and surveying engineering projects were carried out based on the CGCS2000 framework. Since my country used the 1954 Beijing coordinate system and the 1980 Xi'an coordinate system, there are a large number of previous coordinate system results in the field of basic surveying and engineering application in my country. These data exist in large quantities in my country's construction field and play a key role.

[0003] Since there are often multiple projection zones in railway maps, when converting previous railway information results into the CGCS2000 coordinate system, the problems of high computational intensity and low efficiency are usually faced. In addition, the coordinate errors between multiple projection zones will affect the accuracy of the coordinate conversion of the entire coordinate system, resulting in the inability to form a unified and complete railway map. Summary of the invention

[0004] The present invention provides a railway multi-projection belt data conversion method to solve the problems of high calculation intensity, low efficiency and poor accuracy in railway multi-projection belt coordinate data conversion in the prior art. It includes:

[0005] Obtain the coordinate data set of the control points in the projection zone of the current railway map, and divide the data conversion partitions according to the coordinate data set of the control points;

[0006] Acquire a control point coordinate data set of the projection zone in the data conversion partition, establish a data conversion neural network model according to the control point coordinate data set in the data conversion partition, perform coordinate data conversion on each point to be measured based on the data conversion neural network model, and obtain a coordinate conversion value of the point to be measured;

[0007] Determine the model error value according to the coordinate conversion value of the point to be measured, and determine the final conversion coordinate data sequence according to the model error value;

[0008] The final conversion coordinate data sequence is spliced, and the final data conversion railway map is obtained based on the data splicing results.

[0009] Furthermore, the data conversion partitions are divided according to the coordinate data set of the control points, including:

[0010] According to the coordinate data set of the control points, the control points whose distances to the edge of the projection zone are less than a preset distance are selected, and the control points whose distances to the edge of the projection zone are less than the preset distance are formed into a plurality of edge coordinate data sets;

[0011] Set the number of cluster centers k, cluster the edge coordinate data set based on the k-means clustering algorithm, and obtain k cluster centers;

[0012] The control point closest to the cluster center is set as the central control point, and the k central control points are connected end to end in sequence to form a data conversion partition.

[0013] Furthermore, the edge coordinate data set is clustered based on the k-means clustering algorithm, including:

[0014] Obtain the coordinate values ​​of each control point in the edge coordinate data set, and randomly select k initial cluster centers according to the coordinate values ​​of the control points;

[0015] The distance between each control point and the initial cluster center is calculated according to the coordinate value of the control point, and each control point is divided into each cluster according to the distance between each control point and the initial cluster center;

[0016] The average value of the control points is calculated according to the coordinate values ​​of the control points in each cluster after clustering, and the cluster center is updated according to the average value of the control points;

[0017] Repeat the iteration and set new cluster centers until the cluster centers no longer change, and finally obtain the k cluster centers.

[0018] Furthermore, a data conversion neural network model is established according to the control point coordinate data set in the data conversion partition, and coordinate data conversion is performed on each point to be measured based on the data conversion neural network model, including:

[0019] Obtain the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted in the data conversion partition, and calculate the difference between the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted;

[0020] A data set is established according to the difference between the coordinate value of the current coordinate system of the control point and the coordinate value of the coordinate system to be converted, and a data conversion neural network model is established according to the data set;

[0021] The data conversion neural network model is trained according to the data set to obtain a trained data conversion neural network model;

[0022] The coordinates of the points to be measured in the data conversion partition are input into the trained data conversion neural network model to obtain the coordinate conversion values ​​of the points to be measured.

[0023] Furthermore, the data conversion neural network model is trained according to the data set, including:

[0024] Calculate a first loss value according to the difference between the coordinate value of the control point in the data set in the current coordinate system and the coordinate value of the coordinate system to be converted, and calculate a second loss value according to a preset allowable distance value;

[0025] Setting weight values ​​of the first loss value and the second loss value, and determining a final loss value based on the first loss value and the second loss value;

[0026] Whether the data conversion neural network model converges is determined based on the final loss value. If the data conversion neural network model converges, the trained data conversion neural network model is output.

[0027] Further, determining a final loss value based on the first loss value and the second loss value includes:

[0028] The final loss value of the data conversion neural network model is calculated according to the final loss value calculation formula. The final loss value calculation formula is specifically as follows:

[0029] L=αL p +βL q

[0030] Among them, L is the final loss value, p i is the preset optimal coordinate difference, r i is the difference between the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted, Q is the preset allowable distance value, α and β are the preset first weight coefficient and the preset second weight coefficient respectively.

[0031] Further, whether the data conversion neural network model converges is determined according to the final loss value, including:

[0032] Obtain the number of iterations of the current data conversion neural network model, and determine whether the current number of iterations reaches the preset maximum number of iterations. If so, determine that the model has converged;

[0033] If not reached, the difference between the final loss value corresponding to the current iteration number and the final loss value corresponding to the previous iteration is calculated. If the difference between the final loss value corresponding to the current iteration number and the final loss value corresponding to the previous iteration is less than the first preset threshold, the model is determined to have converged.

[0034] Further, the final conversion coordinate data sequence is determined according to the model error value, including:

[0035] Obtain the measured value of the coordinates of the central control point, and calculate the measured value of the perimeter of the data conversion partition according to the measured value of the coordinates of the central control point;

[0036] The coordinate value to be converted of the central control point is input into the data conversion neural network model to obtain the coordinate prediction value of the central control point, and the perimeter prediction value of the data conversion partition is calculated according to the coordinate prediction value of the central control point;

[0037] Calculate the difference between the measured value of the perimeter of the data conversion partition and the predicted value of the perimeter of the data conversion partition to obtain a model error value, and determine the data conversion correction value according to the model error value;

[0038] The coordinate transformation value of the measured point is corrected according to the data transformation correction value to obtain the final transformation coordinate data sequence.

[0039] Further, determining the data conversion correction value according to the model error value includes:

[0040] Obtaining a preset model error tolerance value, and calculating a difference between the model error value and the preset model error tolerance value;

[0041] Determine whether the difference between the model error value and the preset model error tolerance value is greater than a second preset threshold value, and if the difference between the model error value and the preset model error tolerance value is greater than the second preset threshold value, set the first correction value as the data conversion correction value;

[0042] If the difference between the model error value and the preset model error tolerance value is less than or equal to the second preset threshold, determining whether the difference between the model error value and the preset model error tolerance value is greater than a third preset threshold;

[0043] If the difference between the model error value and the preset model error tolerance value is greater than a third preset threshold, setting the second correction value as the data conversion correction value;

[0044] If the difference between the model error value and the preset model error tolerance value is less than or equal to the third preset threshold, the third correction value is set as the data conversion correction value.

[0045] Furthermore, the final data conversion railway map is obtained based on the data splicing results, including:

[0046] Obtaining the coordinate conversion value of each point to be measured in the corrected data conversion partition, and drawing an electronic map of the data conversion partition according to the coordinate conversion value of each point to be measured;

[0047] Based on the averaging method, the electronic maps of each data conversion zone are processed to obtain the final data conversion map.

[0048] The beneficial effects of the present invention are:

[0049] By applying the above technical scheme, the present invention partitions the coordinate data to be converted, effectively reduces the projection deformation at the edge of the projection belt, establishes a data conversion neural network model and corrects the output result, and performs coordinate conversion on the coordinate data to be converted based on the data conversion neural network model, which greatly reduces the calculation intensity and improves the calculation efficiency. By calculating the error value of the coordinate conversion result and correcting the conversion result, the accuracy of the conversion of railway multi-projection belt coordinate data is improved, forming a unified and complete railway map. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0051] Figure 1 The figure shows an overall flow chart of a railway multi-projection band data conversion method proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] The present application embodiment provides a railway multi-projection band data conversion method, such as Figure 1 As shown, including:

[0054] S101, obtaining a coordinate data set of control points in a projection zone of a current railway map, and dividing data conversion partitions according to the coordinate data set of the control points.

[0055] In this embodiment, the coordinate data set of the control point includes the coordinates of the control point in the railway map in the current coordinate system (i.e., the coordinates in the regional independent coordinate system) and the coordinates of the control point in the coordinate system to be converted (i.e., the coordinates in the national geodetic coordinate system). By dividing the coordinate data set into several data conversion partitions, the deformation of the coordinate projection of adjacent projection bands is reduced.

[0056] In some embodiments of the present application, data conversion partitions are divided according to coordinate data sets of control points, including: filtering out control points whose distances to the edge of the projection band are less than a preset distance according to the coordinate data sets of control points, and grouping the control points whose distances to the edge of the projection band are less than the preset distance into several edge coordinate data sets; setting the number of cluster centers k, clustering the edge coordinate data sets based on the k-means clustering algorithm, and obtaining k cluster centers; setting a control point closest to the cluster center as the central control point, and connecting the k central control points end to end in sequence to form a data conversion partition.

[0057] In this embodiment, since the projection deformation of the coordinates at the edge of the projection band is large, control points whose distance from the edge of the projection band is less than a preset distance are screened out, and the central meridian of the adjacent projection band is used as the boundary of the set. The screened control points are grouped into several edge coordinate data sets, and the edge coordinate sets are clustered to form data conversion partitions, thereby minimizing the impact of the projection deformation on subsequent calculations.

[0058] In some embodiments of the present application, an edge coordinate data set is clustered based on a k-means clustering algorithm, including: obtaining the coordinate value of each control point in the edge coordinate data set, and randomly selecting k initial clustering centers according to the coordinate value of the control point; calculating the distance value between each control point and the initial clustering center according to the coordinate value of the control point, and dividing each control point into clusters according to the distance value between each control point and the initial clustering center; calculating the average value of the control point according to the coordinate value of the control point in each cluster after clustering, and updating the cluster center according to the average value of the control point; repeating the iteration and setting a new cluster center until the cluster center no longer changes, and obtaining the final k cluster centers.

[0059] In this embodiment, the number of cluster centers k is set to 4, the central control point is determined according to the final cluster center calculation result, and the four central control points are connected end to end to form a data conversion partition.

[0060] S102, obtaining a control point coordinate data set within the data conversion partition, establishing a data conversion neural network model according to the control point coordinate data set within the data conversion partition, performing coordinate data conversion on each point to be measured based on the data conversion neural network model, and obtaining a coordinate conversion value of the point to be measured.

[0061] In this embodiment, by inputting the coordinate data of the point to be measured into the data conversion neural network model, the predicted coordinate difference can be output according to the data conversion neural network model, and then the coordinate conversion value of the point to be measured can be calculated according to the predicted coordinate difference. The neural network model can greatly reduce the computational intensity of multi-projection band coordinate data conversion and effectively improve the computational efficiency.

[0062] In some embodiments of the present application, a data conversion neural network model is established according to a set of control point coordinate data within a data conversion partition, and coordinate data conversion is performed on each point to be measured based on the data conversion neural network model, including: obtaining the coordinate value of the current coordinate system of the control point within the data conversion partition and the coordinate value of the coordinate system to be converted, and calculating the difference between the coordinate value of the current coordinate system of the control point and the coordinate value of the coordinate system to be converted; establishing a data set according to the difference between the coordinate value of the current coordinate system of the control point and the coordinate value of the coordinate system to be converted, and establishing a data conversion neural network model according to the data set; training the data conversion neural network model according to the data set to obtain a trained data conversion neural network model; inputting the coordinates of the points to be measured within the data conversion partition into the trained data conversion neural network model to obtain the coordinate conversion values ​​of the points to be measured.

[0063] In this embodiment, by calculating the difference between the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted, the corresponding relationship between the coordinate value of the control point in the current coordinate system and the coordinate difference of the control point is determined, and according to the corresponding relationship between the coordinate value of the control point in the current coordinate system and the coordinate difference of the control point, a data conversion neural network model is established and trained to obtain a trained data conversion neural network model.

[0064] In some embodiments of the present application, a data conversion neural network model is trained according to a data set, including: calculating a first loss value according to the difference between the coordinate value of the current coordinate system of the control point in the data set and the coordinate value of the coordinate system to be converted, and calculating a second loss value according to a preset allowable distance value; setting weight values ​​of the first loss value and the second loss value, and determining a final loss value based on the first loss value and the second loss value; determining whether the data conversion neural network model converges according to the final loss value, and if the data conversion neural network model converges, outputting the trained data conversion neural network model.

[0065] In some embodiments of the present application, determining a final loss value based on the first loss value and the second loss value includes: calculating a final loss value of the data conversion neural network model according to a final loss value calculation formula, wherein the final loss value calculation formula is specifically:

[0066] L=αL p +βL q

[0067] Among them, L is the final loss value, p i is the preset optimal coordinate difference, r i is the difference between the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted, Q is the preset allowable distance value, α and β are the preset first weight coefficient and the preset second weight coefficient respectively.

[0068] In this embodiment, L p is the first loss value, L q The second loss value is the second loss value, and the data conversion neural network model is trained based on the first loss value and the second loss value to optimize the prediction value of the data conversion neural network model.

[0069] In this embodiment, the preset first weight coefficient and the preset second weight coefficient are both set by historical operating experience. The loss value is controlled by changing the size of the weight coefficient to improve the accuracy of the model prediction result.

[0070] In some embodiments of the present application, determining whether the data conversion neural network model has converged is based on the final loss value, including: obtaining the current number of iterations of the data conversion neural network model, and determining whether the current number of iterations has reached a preset maximum number of iterations. If so, determining that the model has converged; if not, calculating the difference between the final loss value corresponding to the current number of iterations and the final loss value corresponding to the previous iteration. If the difference between the final loss value corresponding to the current number of iterations and the final loss value corresponding to the previous iteration is less than a first preset threshold, determining that the model has converged.

[0071] In this embodiment, when the preset maximum number of iterations is not reached, the difference between the final loss value corresponding to the current number of iterations and the final loss value corresponding to the previous iteration is used to determine whether the model has converged. This ensures the calculation accuracy of the model while minimizing the model training time, which is beneficial to improving the model calculation speed and efficiency.

[0072] S103, determining a model error value according to the coordinate conversion value of the point to be measured, and determining a final conversion coordinate data sequence according to the model error value.

[0073] In this embodiment, the coordinate conversion value determined according to the data conversion neural network model is corrected based on the model error value, and the final coordinate conversion value is determined according to the correction result to form a final conversion coordinate data sequence.

[0074] In some embodiments of the present application, a final conversion coordinate data sequence is determined based on a model error value, including: obtaining the measured coordinate value of the center control point, and calculating the measured perimeter value of the data conversion partition based on the measured coordinate value of the center control point; inputting the coordinate value of the center control point to be converted into a data conversion neural network model to obtain a predicted coordinate value of the center control point, and calculating the predicted perimeter value of the data conversion partition based on the predicted coordinate value of the center control point; calculating the difference between the measured perimeter value of the data conversion partition and the predicted perimeter value of the data conversion partition to obtain a model error value, and determining a data conversion correction value based on the model error value; and correcting the coordinate conversion value of the point to be measured based on the data conversion correction value to obtain a final conversion coordinate data sequence.

[0075] In this embodiment, since the data conversion partition is a polygon surrounded by a central control point, the shape variable of the polygon is calculated by calculating the measured value and predicted value of the perimeter of the polygon surrounded by the data conversion partition, thereby obtaining the model error value.

[0076] In some embodiments of the present application, determining a data conversion correction value according to a model error value includes:

[0077] Obtain a preset model error tolerance value, and calculate the difference between the model error value and the preset model error tolerance value; determine whether the difference between the model error value and the preset model error tolerance value is greater than a second preset threshold value; if the difference between the model error value and the preset model error tolerance value is greater than the second preset threshold value, set the first correction value as the data conversion correction value; if the difference between the model error value and the preset model error tolerance value is less than or equal to the second preset threshold value, determine whether the difference between the model error value and the preset model error tolerance value is greater than a third preset threshold value; if the difference between the model error value and the preset model error tolerance value is greater than the third preset threshold value, set the second correction value as the data conversion correction value; if the difference between the model error value and the preset model error tolerance value is less than or equal to the third preset threshold value, set the third correction value as the data conversion correction value.

[0078] In this embodiment, the preset model error tolerance value is set according to historical experience, and the first correction value, the second correction value and the third correction value decrease in sequence. The smaller the difference between the model error value and the preset model error tolerance value, the lower the data conversion correction value. By correcting the coordinate transformation value of the measured point, a coordinate transformation value with higher accuracy is obtained.

[0079] S104, splicing the final conversion coordinate data sequence, and obtaining the final data conversion railway map according to the data splicing result.

[0080] In this embodiment, the final conversion coordinate data sequences of each data conversion partition are spliced, and the splicing result is displayed on the railway electronic map to obtain the final data conversion railway map.

[0081] In some embodiments of the present application, a final data conversion railway map is obtained based on the data splicing results, including: obtaining the coordinate conversion value of each point to be measured in the corrected data conversion partition, and drawing an electronic map of the data conversion partition according to the coordinate conversion value of each point to be measured; and performing edge processing on the electronic map of each data conversion partition based on the averaging method to obtain the final data conversion map.

[0082] In this embodiment, the electronic maps of the data conversion zones are spliced ​​to obtain a final data conversion map. Since the edges of the data conversion zones in the final data conversion map may be deformed and misaligned, the electronic maps of the data conversion zones are spliced ​​using the averaging method, and the coordinates of the points to be joined on both sides of the misaligned edges are calculated as the coordinates after the splicing, so as to realize the splicing of the electronic maps of the data conversion zones and form a unified and complete railway map.

[0083] By applying the above technical scheme, the present invention obtains the coordinate data set of the control points in the projection zone of the current railway map, divides the data conversion partition according to the coordinate data set of the control points; obtains the coordinate data set of the control points in the projection zone of the data conversion partition, establishes a data conversion neural network model according to the coordinate data set of the control points in the data conversion partition, performs coordinate data conversion on each point to be measured based on the data conversion neural network model, and obtains the coordinate conversion value of the point to be measured; determines the model error value according to the coordinate conversion value of the point to be measured, and determines the final conversion coordinate data sequence according to the model error value; splices the final conversion coordinate data sequence, and obtains the final data conversion railway map according to the data splicing result. Finally, a high-accuracy, unified and complete railway map is formed.

[0084] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by hardware, or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present invention.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A railway multi-projection band data conversion method, characterized in that: The method comprises: Obtain the coordinate data set of the control points in the projection zone of the current railway map, and divide the data conversion partitions according to the coordinate data set of the control points; Acquire a control point coordinate data set of the projection zone in the data conversion partition, establish a data conversion neural network model according to the control point coordinate data set in the data conversion partition, perform coordinate data conversion on each point to be measured based on the data conversion neural network model, and obtain a coordinate conversion value of the point to be measured; Determine the model error value according to the coordinate conversion value of the point to be measured, and determine the final conversion coordinate data sequence according to the model error value; The final conversion coordinate data sequence is spliced, and the final data conversion railway map is obtained according to the data splicing result; The data conversion partitions are divided according to the coordinate data set of the control points, including: According to the coordinate data set of the control points, the control points whose distances to the edge of the projection zone are less than a preset distance are selected, and the control points whose distances to the edge of the projection zone are less than the preset distance are formed into a plurality of edge coordinate data sets; Set the number of cluster centers k, cluster the edge coordinate data set based on the k-means clustering algorithm, and obtain k cluster centers; Set the control point closest to the cluster center as the central control point, and connect the k central control points end to end in sequence to form a data conversion partition; The final transformation coordinate data sequence is determined according to the model error value, including: Obtain the measured value of the coordinates of the central control point, and calculate the measured value of the perimeter of the data conversion partition according to the measured value of the coordinates of the central control point; The coordinate value to be converted of the central control point is input into the data conversion neural network model to obtain the coordinate prediction value of the central control point, and the perimeter prediction value of the data conversion partition is calculated according to the coordinate prediction value of the central control point; Calculate the difference between the measured value of the perimeter of the data conversion partition and the predicted value of the perimeter of the data conversion partition to obtain a model error value, and determine the data conversion correction value according to the model error value; The coordinate transformation value of the measured point is corrected according to the data transformation correction value to obtain the final transformation coordinate data sequence.

2. The railway multi-projection band data conversion method according to claim 1 is characterized in that: Clustering edge coordinate data sets based on the k-means clustering algorithm includes: Obtain the coordinate values ​​of each control point in the edge coordinate data set, and randomly select k initial cluster centers according to the coordinate values ​​of the control points; The distance between each control point and the initial cluster center is calculated according to the coordinate value of the control point, and each control point is divided into each cluster according to the distance between each control point and the initial cluster center; The average value of the control points is calculated according to the coordinate values ​​of the control points in each cluster after clustering, and the cluster center is updated according to the average value of the control points; Repeat the iteration and set new cluster centers until the cluster centers no longer change, and finally obtain the k cluster centers.

3. The railway multi-projection band data conversion method according to claim 1, characterized in that: A data conversion neural network model is established according to the control point coordinate data set in the data conversion partition, and coordinate data conversion is performed on each test point based on the data conversion neural network model, including: Obtain the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted in the data conversion partition, and calculate the difference between the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted; A data set is established according to the difference between the coordinate value of the current coordinate system of the control point and the coordinate value of the coordinate system to be converted, and a data conversion neural network model is established according to the data set; The data conversion neural network model is trained according to the data set to obtain a trained data conversion neural network model; The coordinates of the points to be measured in the data conversion partition are input into the trained data conversion neural network model to obtain the coordinate conversion values ​​of the points to be measured.

4. The railway multi-projection band data conversion method according to claim 3 is characterized in that: The data conversion neural network model is trained based on the data set, including: Calculate a first loss value according to the difference between the coordinate value of the control point in the data set in the current coordinate system and the coordinate value of the coordinate system to be converted, and calculate a second loss value according to a preset allowable distance value; Setting weight values ​​of the first loss value and the second loss value, and determining a final loss value based on the first loss value and the second loss value; Whether the data conversion neural network model converges is determined based on the final loss value. If the data conversion neural network model converges, the trained data conversion neural network model is output.

5. The railway multi-projection band data conversion method according to claim 4 is characterized in that: Determining a final loss value based on the first loss value and the second loss value includes: The final loss value of the data conversion neural network model is calculated according to the final loss value calculation formula. The final loss value calculation formula is specifically as follows: L=αL p +βL q Among them, L is the final loss value, p i is the preset optimal coordinate difference, r i is the difference between the coordinate value of the control point in the current coordinate system and the coordinate value of the coordinate system to be converted, Q is the preset allowable distance value, α and β are the preset first weight coefficient and the preset second weight coefficient respectively.

6. The railway multi-projection band data conversion method according to claim 5, characterized in that: Determine whether the data conversion neural network model converges based on the final loss value, including: Obtain the number of iterations of the current data conversion neural network model, and determine whether the current number of iterations reaches the preset maximum number of iterations. If so, determine that the model has converged; If not reached, the difference between the final loss value corresponding to the current iteration number and the final loss value corresponding to the previous iteration is calculated. If the difference between the final loss value corresponding to the current iteration number and the final loss value corresponding to the previous iteration is less than the first preset threshold, the model is determined to have converged.

7. The railway multi-projection band data conversion method according to claim 1, characterized in that: Determine the data conversion correction value based on the model error value, including: Obtaining a preset model error tolerance value, and calculating a difference between the model error value and the preset model error tolerance value; Determine whether the difference between the model error value and the preset model error tolerance value is greater than a second preset threshold value, and if the difference between the model error value and the preset model error tolerance value is greater than the second preset threshold value, set the first correction value as the data conversion correction value; If the difference between the model error value and the preset model error tolerance value is less than or equal to the second preset threshold, determining whether the difference between the model error value and the preset model error tolerance value is greater than a third preset threshold; If the difference between the model error value and the preset model error tolerance value is greater than a third preset threshold, setting the second correction value as the data conversion correction value; If the difference between the model error value and the preset model error tolerance value is less than or equal to the third preset threshold, the third correction value is set as the data conversion correction value.

8. The railway multi-projection band data conversion method according to claim 1, characterized in that: The final data conversion railway map is obtained based on the data splicing results, including: Obtaining the coordinate conversion value of each point to be measured in the corrected data conversion partition, and drawing an electronic map of the data conversion partition according to the coordinate conversion value of each point to be measured; Based on the averaging method, the electronic maps of each data conversion zone are processed to obtain the final data conversion map.

Images