Projection coordinate direct conversion error correction method based on neural network and related equipment

Abstract

The invention provides a projection coordinate direct conversion error correction method based on a neural network and related equipment, and relates to the technical field of coordinate conversion. The method comprises the following steps: acquiring projection coordinates under a first reference system to be converted; for the projection coordinates to be converted under the first reference system, obtaining the projection coordinates of the projection coordinates under the second reference system through polynomial fitting; inputting the projection coordinates in the first reference system and the projection coordinates, obtained by fitting, in the second reference system into a trained neural network model to judge whether a conversion error is within a preset error range; and if the conversion error is not in the preset error range, correcting the projection coordinate under the second reference system obtained by fitting according to the error statistic corresponding to the control point set. According to the method, the projection coordinate conversion precision under different reference systems can be improved on the premise of consuming less time resources and computing resources.

Description

Technical Field

[0001] The present invention relates to the technical field of coordinate transformation, and particularly relates to a method and related device for directly correcting the projection coordinate transformation error based on a neural network. Background Art

[0002] A Spatial Reference System or Coordinate Reference System is a coordinate system used to accurately measure the positions on the Earth's surface. In the field of geographic information science, the spatial reference system is divided into a geographic coordinate system and a projection coordinate system according to longitude and latitude coordinates and plane rectangular coordinates. For the transformation between two projection coordinate systems with known ellipsoid and projection parameters, an indirect transformation method can be adopted. As shown in Figure 1 , the projection coordinates in the A reference system are inversely calculated and transformed into the geographic coordinate system in the A reference system, and then transformed into the geographic coordinate system in the B reference system through a three-parameter or seven-parameter transformation method, and then transformed into the projection coordinates in the B reference system through a forward projection calculation. For two projection coordinate systems with unknown ellipsoid parameters or projection parameters, a direct transformation method is adopted, generally based on a known set of control points, and the direct transformation between the two coordinate systems is realized by fitting a polynomial.

[0003] One of the greatest advantages of a neural network is that it can determine the relationship between two coordinate systems without a mathematical model. Therefore, the neural network method has also been introduced into coordinate transformation, and there are mainly three ideas: (1) Using the known control point coordinates as a training set to let the neural network learn the transformation relationship between the coordinates, such as "Wu Zhaofu, Gao Fei, Gong Peng. GPS Coordinate Transformation Based on BP Neural Network [J]. Engineering Investigation, 2004, (04): 58-59, 63.". (2) Improving the accuracy by comparing the algorithms of various neural networks or constructing a more complex neural network model, such as "Cui Fang, Zhao Shuxu. Convolutional Neural Network GPS Coordinate Transformation Method [J]. Bulletin of Surveying and Mapping, 2019, (03): 1-5.". (3) First, realizing coordinate transformation through traditional methods, and then training an error model through a neural network to better improve the accuracy, such as "Li Dajun, Zou Shilin, Liu Yingzi, etc. Projection Transformation Method Based on Neural Network [J]. Bulletin of Surveying and Mapping, 2003, (03): 24-26, 30.".

[0004] The above ideas essentially hope that the neural network model can achieve coordinate transformation by learning to establish the connection between two coordinate systems, which can be regarded as a regression problem. However, in actual engineering practice, limited by the number of control points, in small-sample regression problems, the neural network model is prone to overfitting problems, reducing the generalization ability of the model. In some cases, the conversion accuracy of the trained neural model is even lower than that of traditional polynomial methods. If the number of control points is increased on a large scale, in fact, high accuracy can be achieved by establishing a grid model and other methods. The requirements of complex neural networks for computing resources and time make neural networks fall into the dilemma of "using a sledgehammer to crack a nut" in coordinate transformation problems. At the same time, since the neural network is a black-box model, it is difficult to estimate the error of the converted coordinate accuracy, and it is even more difficult to control the error accuracy after conversion. Summary of the Invention

[0005] In order to further improve the accuracy of direct projection coordinate conversion, considering the limited number of control points, the present invention converts the coordinate conversion problem (essentially a regression problem) into a classification problem, and provides a method and related device for direct projection coordinate conversion error correction based on a neural network.

[0006] In a first aspect, the present invention provides a method for direct projection coordinate conversion error correction based on a neural network, including:

[0007] Obtain the projection coordinates in the first reference system to be converted;

[0008] For the projection coordinates in the first reference system to be converted, obtain their projection coordinates in the second reference system through polynomial fitting;

[0009] Input the projection coordinates in the first reference system and the projection coordinates in the second reference system obtained by fitting into the trained neural network model to determine whether the conversion error is within a preset error range;

[0010] If the conversion error is not within the preset error range, correct the projection coordinates in the second reference system obtained by fitting according to the error statistic corresponding to the control point set.

[0011] Further, the training process of the neural network model includes:

[0012] Obtain a control point set and divide it into a training set and a test set. For any point in the training set, use the projection coordinates of the point in the first reference system and the projection coordinates of the point in the second reference system obtained by polynomial fitting as the input of the preset neural network, and use whether the conversion error is within the preset error range as the output of the preset neural network to optimize the parameters of the neural network.

[0013] Further, taking whether the conversion error is within a preset error range as the output of a preset neural network specifically includes:

[0014] Using a first parameter, a second parameter, and a third parameter to comprehensively represent whether the conversion error in the horizontal direction is within a preset horizontal error range, and using a fourth parameter, a fifth parameter, and a sixth parameter to comprehensively represent whether the conversion error in the vertical direction is within a preset vertical error range.

[0015] Further, taking whether the conversion error is within a preset error range as the output of a preset neural network specifically includes:

[0016] Using a first parameter and a second parameter to comprehensively represent whether the distance conversion error is within a preset distance error range, using a third parameter and a fourth parameter to comprehensively represent whether the conversion error in the horizontal direction is negative or positive, and using a fifth parameter and a sixth parameter to comprehensively represent whether the conversion error in the vertical direction is negative or positive.

[0017] Further, if the conversion error is not within the preset error range, then correcting the projected coordinates in the second reference system obtained by fitting according to the error statistic corresponding to the control point set, specifically including:

[0018] If the distance conversion error is not within the preset distance error range, then determining the coordinate direction corresponding to the parameter with a higher confidence according to the output of the neural network model, so as to correct the coordinates in the corresponding direction of the projected coordinates in the second reference system obtained by fitting, including: if the conversion error in the coordinate direction corresponding to the parameter with a higher confidence is negative, then adding N times the standard deviation to the coordinate value in the corresponding direction of the projected coordinates, if the conversion error in the coordinate direction corresponding to the parameter with a higher confidence is positive, then subtracting N times the standard deviation from the coordinate value in the corresponding direction of the projected coordinates; wherein, the standard deviation is the standard deviation of the distance conversion error corresponding to the control point set or the standard deviation of the conversion error in the coordinate direction corresponding to the parameter with a higher confidence, and N is a positive number less than or equal to 1.

[0019] Further, the neural network uses an error backpropagation neural network.

[0020] In a second aspect, the present invention provides a device for directly correcting the conversion error of projected coordinates based on a neural network, including:

[0021] An acquisition module, configured to acquire the projected coordinates in a first reference system to be converted;

[0022] A coordinate conversion module, configured to, for the projected coordinates in the first reference system to be converted, obtain the projected coordinates in a second reference system by polynomial fitting;

[0023] An error judgment module, configured to input the projection coordinates in the first reference system and the projection coordinates in the second reference system obtained by fitting into a trained neural network model, so as to judge whether the conversion error is within a preset error range;

[0024] A coordinate correction module, configured to correct the projection coordinates in the second reference system obtained by fitting according to the error statistic corresponding to the control point set when the conversion error is not within the preset error range.

[0025] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the program, the method described in the first aspect is implemented.

[0026] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0027] The beneficial effects of the present invention are as follows:

[0028] (1) The method and related device for directly correcting the projection coordinate conversion error based on a neural network according to the present invention convert the coordinate conversion problem (essentially a regression problem) into a classification problem under the condition of limited number of control points, and fully utilize the determination advantage of the neural network under the condition of small samples; and combined with the determination of the neural network and the error model statistic, a relatively robust correction strategy is made to ensure that the error is continuously optimized, thereby improving the accuracy of coordinate conversion.

[0029] (2) In the method for directly correcting the projection coordinate conversion error based on a neural network according to the present invention, the training of the neural network model is very simple and fast, and does not require a large amount of time and computing resources. Description of the Drawings

[0030] Figure 1 For the conversion between projection coordinates in different reference systems;

[0031] Figure 2 It is a schematic flowchart of a method for directly correcting the projection coordinate conversion error based on a neural network provided by an embodiment of the present invention;

[0032] Figure 3 It is a schematic diagram of coordinate conversion;

[0033] Figure 4 It is a schematic flowchart of a method for directly correcting the projection coordinate conversion error based on a neural network provided by an embodiment of the present invention;

[0034] Figure 5Schematic diagram of the training process of a neural network model provided by an embodiment of the present invention;

[0035] Figure 6 Error correction strategy combining neural network model determination and error model statistic provided by an embodiment of the present invention;

[0036] Figure 7 Accuracy rate of determination of a neural network model provided by an embodiment of the present invention in "whether within the error range" and "positive or negative of the error direction";

[0037] Figure 8 Percentage of type II errors and type III errors in all records that require error correction (i.e., the number of records determined to be outside the error range);

[0038] Figure 9 Schematic diagram of the structure of a device for directly converting projection coordinates based on a neural network and correcting errors provided by an embodiment of the present invention;

[0039] Figure 10 Structure block diagram of an electronic device provided by an embodiment of the present invention. Specific implementation manners

[0040] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, rather than all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] The basic assumption of the present invention is that on the premise of admitting the establishment of the second law of geography (spatial heterogeneity), it is considered that there is a certain connection between the errors caused by any coordinate conversion method and the spatial distribution. On the basis of this assumption, instead of letting the neural network learn the direct connection between the two coordinate systems, the neural network is allowed to learn the connection between "whether the error is within a certain range" and "spatial distribution", so as to convert the problem of coordinate conversion calculation into a classification problem.

[0042] To better understand the technical solution of the present invention, the technical problems to be solved by the present invention are defined as follows:

[0043] Assume that a set C containing n control points is known, that is, the projection coordinates of n control points in two reference systems are known, which are respectively expressed as follows:

[0044] (1) n projection coordinates in the A reference system: (x A1 , y A1 ), (xA2 , y A2 ), ……(x An , y An );

[0045] (2) The n projection coordinates in the B reference system: (x B1 , y B1 ), (x B2 , y B2 ), ……(x Bn , y Bn );

[0046] For any unknown point k, the known projection coordinates in the A reference system are Solve for the projection coordinates of k in the B reference system to make it as close as possible to the true value.

[0047] To solve the above problems, as Figure 2 shown, an error correction method for direct conversion of projection coordinates based on a neural network provided by an embodiment of the present invention includes the following steps:

[0048] S101: Obtain the projection coordinates in the first reference system to be converted;

[0049] S102: For the projection coordinates in the first reference system to be converted, obtain their projection coordinates in the second reference system through polynomial fitting;

[0050] Specifically, due to different selected reference ellipsoids and different projection methods, different coordinate systems will be formed. The first reference system and the second reference system in the present invention represent coordinate systems under different reference ellipsoids, such as the 1954 Beijing coordinate system, the 1980 Xi'an coordinate system, the 2000 National coordinate system, and the WGS-84 coordinate system. If the geographic coordinates in the first reference system are known, the geographic coordinates can be converted into the projection coordinates in the first reference system by using GIS software or an online coordinate system conversion tool. It should be noted that the projection coordinates in the two reference systems in the embodiments of the present invention adopt the same projection method.

[0051] In practical applications, the direct conversion method of projection coordinates between two reference systems through polynomial fitting includes, but is not limited to, 2D Helmert transformation (four parameters), affine transformation six-parameter method, and quadratic polynomial, etc.

[0052] (1) 2D Helmert transformation. It mainly includes: Coordinate system A is transformed into coordinate system B through translation, scaling, and rotation, as Figure 3 (a) shown. The 2D Helmert transformation formula is as follows, where T x , T yis a translation variable, s is a scale variable, and θ is the angle of rotation.

[0053]

[0054] (2) Six-parameter method of affine transformation. Coordinate system A and coordinate system B are not orthogonal, and the scale factors in the X and Y directions may also be different. As shown in Figure 3 (b). The conversion formula of the six-parameter method of affine transformation is as follows, where T x , T y is a translation variable, S x , S y are the scale variables in the X and Y directions respectively, θ is the angle of rotation, and β is the non-orthogonal intersection angle between the two coordinate systems.

[0055]

[0056] (3) Quadratic polynomial transformation. Affine transformation can be regarded as a first-degree polynomial transformation. The general formula of polynomial transformation is as follows.

[0057]

[0058] In practical applications, the power degree of the polynomial should not be too high. Generally, it is considered that the power degree of the polynomial is preferably not higher than 3.

[0059] The parameters of the above three conversion methods are all obtained by solving the parameters of the known control points through the least squares method.

[0060] S103: Input the projection coordinates in the first reference system and the projection coordinates in the second reference system obtained by fitting into the trained neural network model to determine whether the conversion error is within the preset error range;

[0061] Specifically, the neural network can be a traditional machine learning network, such as a BP neural network, or a neural network based on deep learning, such as a convolutional neural network, a recurrent neural network, etc. It should be noted that compared with other network types, in practical applications, the BP neural network is the preferred neural network model of the present invention.

[0062] S104: If the conversion error is not within the preset error range, correct the projection coordinates in the second reference system obtained by fitting according to the error statistic corresponding to the control point set.

[0063] Specifically, the error statistic includes, but is not limited to, the statistic of the conversion error in the horizontal direction, vertical direction, or distance aspect; at the same time, the statistic can include the average value, maximum value, minimum value, and standard deviation, etc. For example, the standard deviation of the distance conversion error corresponding to the control point set can be used to correct the projection coordinates in the second reference system obtained by fitting.

[0064] For example, as Figure 4 shown, for the projection coordinates to be solved and converted in the A reference system first, the projection coordinates in the B reference system are obtained by means of polynomial fitting Then, the distribution range of the conversion error is determined through the trained neural network model, and then the error correction is carried out by combining the statistics of the error model of the known control points, so as to improve the accuracy of the model.

[0065] The method for directly correcting the conversion error of projection coordinates based on a neural network provided by the embodiments of the present invention converts the coordinate conversion problem (essentially a regression problem) into a classification problem under the condition of limited number of control points, and fully utilizes the determination advantage of the neural network under the condition of small samples; the correction strategy combines the determination of the neural network and the statistics of the error model, and makes a relatively robust correction strategy to ensure that the error is continuously optimized, thereby improving the accuracy of coordinate conversion.

[0066] In one embodiment, as Figure 5 shown, for the above-mentioned neural network model, the embodiments of the present invention also provide a training method for the neural network model, which specifically includes the following steps:

[0067] Obtain a control point set and divide it into a training set and a test set. For any point in the training set, the projection coordinates of the point in the first reference system and the projection coordinates obtained by polynomial fitting in the second reference system are used as the input of the preset neural network, and whether the conversion error is within the preset error range is used as the output of the preset neural network to optimize the parameters of the neural network. Figure 5 Among them, (x′ B1 , y′ B1 ), (x′ B2 , y′ B2 ),... (x′ Bn , y′ Bn ) represent the projection coordinates obtained by calculating the projection coordinates in the A reference system through polynomial fitting and converting them to the B reference system.

[0068] Specifically, for the training of the neural network, the input is defined as the projection coordinates of any point i in the control point set in the A reference system and the projection coordinates in the B reference system calculated by polynomial fitting, that is, {x Ai , y Ai , x′ Bi , y′ Bi}, i ∈ C. The purpose is to let the neural network know the spatial distribution of the control points in each reference system. It should be noted that in order to adapt to the requirements of the neural network (such as the BP neural network) for the input, it is necessary to normalize the input coordinates.

[0069] In the embodiments of the present invention, for the output of the neural network during the training process, according to different types of conversion errors selected, the determination of whether the conversion error is within the preset error range will also be different, mainly including the following two determination strategies.

[0070] The first error determination strategy: Determine the output of the neural network according to whether the conversion errors in the horizontal and vertical directions are within the preset error range. This output is a six-parameter output, assumed to be R1 - R6;

[0071] (1) The first three parameters represent the distribution of the error in the horizontal direction; theoretically, when the confidence of one of the parameters is 1, the confidence of the other two parameters should be 0, that is

[0072]

[0073] (2) The last three parameters represent the distribution of the error in the vertical direction; theoretically, when the confidence of one of the parameters is 1, the confidence of the other two parameters should be 0, that is

[0074]

[0075] Among them, σ x represents the standard deviation of the conversion error in the horizontal direction corresponding to the control point set, σ y represents the standard deviation of the conversion error in the vertical direction corresponding to the control point set, Δx i represents the conversion error in the horizontal direction corresponding to the control point i, and Δy i represents the conversion error in the vertical direction corresponding to the control point i.

[0076] The second error determination strategy: Determine the output of the neural network according to whether the distance conversion error is within the preset error range. This output is also a six-parameter output, assumed to be R1 - R6;

[0077] (1) The first two parameters indicate whether the distance conversion error is within the error range; theoretically, when the confidence of one of the parameters is 1, the confidence of the other parameter should be 0, that is

[0078]

[0079] (2) The middle two parameters are used to determine whether the conversion error in the horizontal direction is negative or positive; theoretically, when the confidence of one of the parameters is 1, the confidence of the other parameter should be 0, that is

[0080]

[0081] (3) The last two parameters are used to determine whether the conversion error in the vertical direction is negative or positive; theoretically, when the confidence level of one parameter is 1, the confidence level of the other parameter should be 0, that is

[0082]

[0083] where, Δdist i represents the Euclidean distance between the conversion error in the horizontal direction and the conversion error in the vertical direction corresponding to the control point i, and σ dist represents the standard deviation of the distance conversion error corresponding to the set of control points, and Δx i represents the conversion error in the horizontal direction corresponding to the control point i, and Δy i represents the conversion error in the vertical direction corresponding to the control point i.

[0084] It should be noted that in the above embodiments, only the corresponding standard deviation is used to determine the error range as an example. According to needs, users can also select other error statistics to determine the error range.

[0085] Theoretically, the values of the above six parameters are 0 or 1; however, in the actual training process, the neural network generally uses the sigmoid function, and the output target value cannot be less than or equal to 0 or greater than or equal to 1. Therefore, in fact, the values of the above six parameters are 0.01 or 0.99.

[0086] For the method for directly correcting the projection coordinate conversion error based on a neural network provided by the embodiments of the present invention, the training of the neural network model is very simple and fast, and does not require a large amount of time and computing resources.

[0087] It can be understood that in the case of a relatively small number of samples, the second determination method is actually better than the first determination method because it is only a binary classification (whether the error exceeds the threshold, and whether the errors in the X direction and Y direction are negative or positive). Therefore, in one embodiment, taking the second determination method as an example, an error correction strategy is provided, which specifically includes the following steps.

[0088] For the second determination method, the determination of the neural network model needs to go through two stages. The first stage is to determine whether the distance conversion error is within the error range. If the distance conversion error is within the error range, then there is no need to correct the projection coordinates in the second reference system obtained by fitting. Only when it is determined that the distance conversion error is outside the error range, does it enter the second stage. The second stage is to assist in error correction by determining whether the conversion errors in the X direction and Y direction are positive or negative. Taking the standard deviation as the error range as an example, the actual situation of the projection coordinate conversion, the determination of the neural network model, and the error correction are shown in Table 1.

[0089] Table 1 Actual situation of projection coordinate conversion, determination of neural network model, and error correction

[0090]

[0091]

[0092] It is not difficult to find from Table 1 that the model prediction in the first stage avoids the situation of further correcting errors for coordinates that are already within the standard deviation range. However, the most important impact on error optimization is the determination in the second stage. Therefore, the accuracy of the neural network model determination in the second stage is crucial. If the error in a certain direction is small, the neural network model is very likely to make a wrong determination. Therefore, the thinking can be changed. It only needs to determine the positive and negative of the direction where the maximum error value is located, rather than determining the positive and negative of the error in the X or Y direction. As shown in Table 2, since the error in the X direction is too small, the indexes of the negative and positive values in the X direction are very close, and all judgments are wrong. However, since the error in the Y direction is large, all judgments are correct. Based on the above assumptions, similarly, in the test stage, the present invention determines the error direction corresponding to the parameter with a larger confidence level output by the neural network as the maximum error direction, so as to correct the coordinates in the maximum error direction.

[0093] Table 2 Example of positive and negative judgments of the neural network model for the X direction, Y direction, and maximum error direction

[0094]

[0095] The error correction strategy is as Figure 6 shown. In one embodiment, the adopted error correction strategy specifically includes the following steps: If the distance conversion error is not within the preset distance error range, determine the coordinate direction corresponding to the parameter with a larger confidence level according to the output of the neural network model, so as to correct the error of the coordinate in the corresponding direction in the projection coordinate under the second reference system obtained by fitting, including: If the conversion error in the coordinate direction corresponding to the parameter with a larger confidence level is negative, add N times the standard deviation to the coordinate value in the corresponding direction of the projection coordinate; if the conversion error in the coordinate direction corresponding to the parameter with a larger confidence level is positive, subtract N times the standard deviation from the coordinate value in the corresponding direction of the projection coordinate; where the standard deviation is the standard deviation of the distance conversion error corresponding to the control point set or the standard deviation of the conversion error in the coordinate direction corresponding to the parameter with a larger confidence level, and N is a positive number less than or equal to 1. For example, if the standard deviation is the standard deviation of the distance conversion error, according to experience, N can be If the standard deviation is the standard deviation of the conversion error in the horizontal direction or the vertical direction, according to experience, N can be 1.

[0096] Similarly, for the first determination method, if it is determined according to the output of the neural network model that the conversion error in the horizontal direction (or vertical direction) is not within the preset horizontal error range, then error correction is performed on the coordinates in the horizontal direction (or vertical direction) of the projected coordinates in the second reference system obtained by fitting, including: if the conversion error in the horizontal direction (or vertical direction) is less than the left endpoint value of the preset horizontal error range, then add N times the standard deviation to the coordinates in the horizontal direction (or vertical direction); if the conversion error in the horizontal direction (or vertical direction) is greater than the right endpoint value of the preset horizontal error range, then subtract N times the standard deviation from the coordinates in the horizontal direction (or vertical direction); where the standard deviation is the standard deviation of the conversion error in the horizontal direction (or vertical direction) corresponding to the control point set, and N is a positive number less than or equal to 1.

[0097] To verify the effectiveness of the solution of the present invention, the present invention also provides the following experimental data.

[0098] (I) Experimental process

[0099] (1) Projection coordinate conversion of known control points based on polynomials.

[0100] In this experiment, 438 control points in a certain area were selected. In the selection of polynomial fitting, the method with the smallest solution error for the control points in this area - the six-parameter affine transformation method - was selected. The average value of the solved error was 0.630 meters, the maximum value was 2.485 meters, the minimum value was 0.012 meters, and the standard deviation was 0.494 meters.

[0101] (2) Training of a neural network model for projection coordinate conversion error for small samples.

[0102] Converting the calculation problem of coordinate system conversion into a classification problem of error determination is an important key to solving the problem of a small number of control point samples. The Back Propagation Neural Network (BP Neural Network for short) is a commonly used and very effective neural network. At the same time, using the same data set to repeatedly train the BP neural network for small sample data is also helpful for updating the weights during the gradient descent process, so as to better use the BP neural network for classification.

[0103] Since the amount of data of 438 control points is still small, two measures are taken to enhance the samples and make the results more credible: (1) randomly generate 6 groups of training sets and test sets according to a ratio of 19:1; (2) in addition, each training set is trained repeatedly for multiple times. After practical optimization, finally, the input layer of the BP neural network has 4 input nodes, 20 nodes in the middle hidden layer, 6 nodes in the output layer, and is trained repeatedly 200 times.

[0104] (3) Combine the determination of the neural network model and the error correction of the error model statistic.

[0105] The error correction is carried out according to Figure 6 the given correction strategy.

[0106] (II) Experimental Results

[0107] (1) The results of 6 test sets show that the neural network model is superior to the determination of "whether within the error range" in the "positive and negative determination of the error direction", as Figure 7 shown.

[0108] (2) In the determination of "whether within the error range", two types of errors will occur: The first type is that it is determined to be within the error range outside the error range. Since it will not be revised anymore, this type of error has no impact on the error correction; the second type is that it is determined to be outside the error range within the error range, and this type of error will affect the correction result. In addition, under the condition of being determined to be outside the error range, the incorrect determination of the positive and negative of the maximum error direction will ultimately affect the error correction, which is called the third type of error in this paper. Figure 8 is the percentage of the second type of error and the third type of error in all records that require error correction (that is, the number of records determined to be outside the error range). It can be seen that although the percentage of the second type of error is still relatively high (more than 20%), due to the very low percentage of the third type of error, the reliability of the error correction is also guaranteed.

[0109] (3) According to the error correction strategy given in this paper, the corrected results of 6 test sets are shown in Table 3:

[0110] Table 3 Statistics of the cumulative sum of errors and the average value of errors before and after modification of the test set

[0111]

[0112] All the above test sets can be significantly optimized. In fact, it benefits from the first two experimental conclusions. Even if some records make the second type of error, due to the very low error rate of the third type of error, the error correction can also be improved, as shown in the example in Table 4.

[0113] Table 4 Example of error correction combining the determination of the neural network model and the error model statistic

[0114]

[0115]

[0116] Such as Figure 9As shown in the figure, an embodiment of the present invention further provides a device for correcting the direct conversion error of projection coordinates based on a neural network, including an acquisition module, a coordinate conversion module, an error judgment module, and a coordinate correction module.

[0117] Among them, the acquisition module is used to acquire the projection coordinates in the first reference system to be converted; the coordinate conversion module is used to obtain the projection coordinates in the second reference system by polynomial fitting for the projection coordinates in the first reference system to be converted; the error judgment module is used to input the projection coordinates in the first reference system and the fitted projection coordinates in the second reference system into the trained neural network model to judge whether the conversion error is within the preset error range; the coordinate correction module is used to correct the fitted projection coordinates in the second reference system according to the error statistic corresponding to the control point set when the conversion error is not within the preset error range.

[0118] It should be noted that the device for correcting the direct conversion error of projection coordinates based on a neural network provided by the embodiment of the present invention is to implement the method, and its functions can be specifically referred to the above method embodiments, which will not be elaborated here.

[0119] Figure 10 An example of the physical structure diagram of an electronic device is shown in Figure 10 As shown in the figure, the electronic device may include: a processor 1001, a communication interface 1002, a memory 1003, and a communication bus 1004. Among them, the processor 1001, the communication interface 1002, and the memory 1003 complete mutual communication through the communication bus 1004. The processor 1001 can call the logical instructions in the memory 1003 to execute a method for correcting the direct conversion error of projection coordinates based on a neural network. The method includes: acquiring the projection coordinates in the first reference system to be converted; obtaining the projection coordinates in the second reference system by polynomial fitting for the projection coordinates in the first reference system to be converted; inputting the projection coordinates in the first reference system and the fitted projection coordinates in the second reference system into the trained neural network model to judge whether the conversion error is within the preset error range; if the conversion error is not within the preset error range, then correcting the fitted projection coordinates in the second reference system according to the error statistic corresponding to the control point set.

[0120] In addition, when the logical instructions in the above-mentioned memory 1003 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0121] An embodiment of the present invention further provides a computer program product. The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the method for correcting the direct conversion error of projection coordinates based on a neural network provided in each of the above method embodiments.

[0122] An embodiment of the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for correcting the direct conversion error of projection coordinates based on a neural network provided in each of the above method embodiments.

[0123] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and 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 invention.

Claims

1. A method for directly correcting the error of projection coordinate conversion based on a neural network, characterized in that Including: Obtain the projected coordinates in the first reference system to be converted; For the projected coordinates in the first reference system to be converted, obtain their projected coordinates in the second reference system through polynomial fitting; Input the projected coordinates in the first reference system and the projected coordinates in the second reference system obtained by fitting into the trained neural network model to determine whether the conversion error is within the preset error range; If the conversion error is not within the preset error range, correct the projected coordinates in the second reference system obtained by fitting according to the error statistic corresponding to the control point set.

2. The method for correcting the direct conversion error of projection coordinates based on a neural network according to claim 1, wherein The training process of the neural network model includes: Obtain the control point set and divide it into a training set and a test set. For any point in the training set, use the projected coordinates of the point in the first reference system and the projected coordinates of the point in the second reference system obtained by polynomial fitting as the input of the preset neural network, and use whether the conversion error is within the preset error range as the output of the preset neural network to optimize the parameters of the neural network.

3. A method for directly converting and error correcting projection coordinates based on a neural network according to claim 2, characterized in that, Using whether the conversion error is within the preset error range as the output of the preset neural network specifically includes: Use the first parameter, the second parameter, and the third parameter to comprehensively represent whether the conversion error in the horizontal direction is within the preset horizontal error range, and use the fourth parameter, the fifth parameter, and the sixth parameter to comprehensively represent whether the conversion error in the vertical direction is within the preset vertical error range.

4. A method for directly converting and error correcting projection coordinates based on a neural network according to claim 2, characterized in that, Using whether the conversion error is within the preset error range as the output of the preset neural network specifically includes: Use the first parameter and the second parameter to comprehensively represent whether the distance conversion error is within the preset distance error range, use the third parameter and the fourth parameter to comprehensively represent whether the conversion error in the horizontal direction is negative or positive, and use the fifth parameter and the sixth parameter to comprehensively represent whether the conversion error in the vertical direction is negative or positive.

5. A method for directly converting and error correcting projection coordinates based on a neural network according to claim 4, characterized in that, If the conversion error is not within the preset error range, correcting the projected coordinates in the second reference system obtained by fitting according to the error statistic corresponding to the control point set specifically includes: If the distance conversion error is not within the preset distance error range, determine the coordinate direction corresponding to the parameter with a higher confidence according to the output of the neural network model, and thus correct the coordinates in the corresponding direction of the projected coordinates in the second reference system obtained by fitting, including: if the conversion error in the coordinate direction corresponding to the parameter with a higher confidence is negative, add N times the standard deviation to the coordinate value in the corresponding direction of the projected coordinates; if the conversion error in the coordinate direction corresponding to the parameter with a higher confidence is positive, subtract N times the standard deviation from the coordinate value in the corresponding direction of the projected coordinates; where the standard deviation is the standard deviation of the distance conversion error corresponding to the control point set or the standard deviation of the conversion error in the coordinate direction corresponding to the parameter with a higher confidence, and N is a positive number less than or equal to 1.

6. A method for correcting the direct conversion error of projection coordinates based on a neural network according to any one of claims 1 to 5, characterized in that, The neural network uses a backpropagation neural network.

7. An apparatus for correcting the direct conversion error of projection coordinates based on a neural network, characterized in that, Including: An acquisition module for obtaining the projected coordinates in the first reference system to be converted; A coordinate conversion module for obtaining the projected coordinates of the first reference system to be converted in the second reference system through polynomial fitting; An error judgment module, configured to input the projection coordinates in the first reference system and the projection coordinates in the second reference system obtained by fitting into a trained neural network model to determine whether the conversion error is within a preset error range; A coordinate correction module, configured to correct the projection coordinates in the second reference system obtained by fitting according to the error statistic corresponding to the control point set when the conversion error is not within the preset error range.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, the method described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the method described in any one of claims 1 to 6 is implemented.

Images