Geographic coordinate optimization method based on interest point constraint multi-source network geocoding platform

Abstract

The invention provides a geographic coordinate optimization method based on a point-of-interest constraint multi-source network geocoding platform, which belongs to the field of geocoding result data processing, and comprises the following steps: S1, a user uses a CRF model to divide an input text address into a rough address feature and a precise address feature; s2, generating a minimum geographic encirclement circle according to the rough address features; s3, theoretical coordinates are calculated according to the accurate address features, and the theoretical coordinates are marked as (Xm, Ym); s4, coding results of all the geocoding platforms form a candidate set; s5, if the coding result in the candidate set is not in the minimum geographic encirclement, outputting the theoretical coordinates (Xm, Ym); and if at least one coding result in the candidate set is located in the minimum geographic encirclement, calculating optimized geographic coordinates (X < optimal > and Y < optimal >) and outputting the optimized geographic coordinates (X < optimal > and Y < optimal >). According to the method, coding result errors of all geocoding platforms are reduced, and the coding result accuracy is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of geocoding result data processing, and in particular relates to a geo-coordinate optimization method based on a multi-source network geocoding platform constrained by points of interest. Background Art

[0002] Geocoding is a technology that converts the text description of an address into geographic coordinates. Current online geocoding platforms (such as Baidu, Tencent, AutoNavi, and Sogou) all rely on this geocoding technology to provide geographic coordinate services.

[0003] When using the above-mentioned existing geocoding platforms, for example: typing "No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province" in the above-mentioned multiple geocoding platforms respectively, the above-mentioned geocoding platforms will query the corresponding encoding results in their corresponding databases based on the typed text address "No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province" through their respective semantic similarity algorithms, and then feedback the corresponding encoding results to the user. The encoding results are the geographic coordinates corresponding to the text address "No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province".

[0004] Since the sources of data in the databases of the above-mentioned multiple geocoding platforms are different, or the design of the semantic similarity algorithms in each geocoding platform is different, the encoding results output by each geocoding platform are also different. For example, for the same text address, the encoding result output by geocoding platform K1 is P1, and the encoding result output by geocoding platform K2 is P2. There are large errors in the geographic coordinates output by multiple geocoding platforms, with an average error of more than 500m. This makes it impossible to provide users with accurate geographic coordinate services, causing inconvenience to users. For example, takeaway personnel cannot accurately reach the location based on the geographic coordinates provided by a certain encoding platform.

[0005] In order to reduce the geographic coordinate errors of multiple geocoding platforms for the same text address and provide users with more accurate geographic coordinate services, it is necessary to propose a geographic coordinate optimization method for a multi-source network geocoding platform based on point of interest constraints. Summary of the invention

[0006] The present invention proposes a geographic coordinate optimization method for a multi-source network geocoding platform based on interest point constraints, which is used to reduce the coding result errors output by each geocoding platform.

[0007] In order to solve the above problems, the present invention proposes the following solutions: The geographic coordinate optimization method based on the multi-source network geocoding platform with interest point constraints includes the following steps: S1: The user inputs a text address into the CRF model, and the CRF model extracts address features from the input text address and divides the address features into rough address features and precise address features; S2: Based on the rough address features, POI data is filtered in the POI address database to generate the minimum geographic encirclement; S3: All POI data within the minimum geographic encirclement are used as a new POI database; query the new POI database, calculate theoretical coordinates based on precise address features, and mark the theoretical coordinates as ( X m , Y m ); S4: Enter the text address in S1 into each geocoding platform respectively, obtain the coding result of each geocoding platform, and form the coding result of each geocoding platform into a candidate set; S5: If none of the encoding results in the candidate set is within the minimum geographic encirclement, then the theoretical coordinates are output ( X m , Y m ); If at least one encoding result in the candidate set is within the minimum geographic encirclement, calculate the optimized geographic coordinates ( X 优 , Y 优 ), output optimized geographic coordinates ( X 优 , Y 优 ).

[0008] Further, step S2 includes the following steps: S2.1: The rough address features and all POI data in the POI address database are input into the BERT model; the BERT model vectorizes the rough address features and the text addresses of all POI data; S2.2: Calculate the cosine similarity between the rough address feature and the text address of each POI data; The formula is:

[0009] In formula (1), A is the vector representation of the rough address feature; B is a vector representation of a POI data text address in the POI address database; Represents the dot product of two vectors; and They are A Vector Sum B L2 norm of a vector; CSis the cosine similarity; S2.3: Set the cosine similarity threshold R and compare it with the vector A POI data with cosine similarity greater than or equal to R are marked as POI data for the encirclement to be constructed; using the minimum polygon method, all POI data for the encirclement to be constructed are constructed into a polygonal area, and the constructed polygonal area is the minimum geographic encirclement.

[0010] Further, step S3 includes two cases: Case 1: If there is at least one house number of the POI data text address in the new POI database that is the same as the house number corresponding to the precise address feature, the calculation formula for the theoretical coordinates is:

[0011] In formula (2), X m and Y m are the longitude and latitude of the theoretical coordinates; X i and Y i The longitude and latitude of the POI data with the same features as the precise address, , i is an integer; n is the number of POI data with the same exact address characteristics, n≥ 1; Case 2: If the new POI database does not contain the same house number as the precise address feature, then query the new POI database for two house numbers close to the house number with the precise address feature, and record them as C 1 and C 2, The calculation formula of the theoretical coordinates is:

[0012] In formula (3), X m and Y m are the longitude and latitude of the theoretical coordinates respectively; C x The house number corresponding to the precise address feature; and House numbers C 1's longitude and latitude; and House numbers C 2's longitude and latitude.

[0013] Furthermore, in step S4, the encoding results of each geocoding platform are recorded as: (X z1 , Y z1 )、( X z2 , Y z2 ),……、( X zu , Y zu ), the candidate set is: {( X z1 , Y z1 )、( X z2 , Y z2 ),……、( X zu , Y zu )}, u is the number of encoding results.

[0014] Further, step S5 includes the following steps: S5.1: If none of the encoding results in the candidate set is within the minimum geographic encirclement, then the theoretical coordinates are output ( X m , Y m ); S5.2: If at least one encoding result in the candidate set is within the minimum geographic encirclement, calculate the optimized geographic coordinates, and record the optimized geographic coordinates as ( X 优 , Y 优 ), output optimized geographic coordinates ( X 优 , Y 优 ); S5.2 specifically includes the following steps: S5.2.1: Calculate the Euclidean distance between each encoding result and the theoretical coordinates using the formula:

[0015] In formula (4), ; d k represents the theoretical coordinates ( X m , Y m ) and k Encoding results ( X zk , Yzk )’s Euclidean distance; S5.2.2: The coordinate weight of each encoding result is:

[0016] In formula (5), w k Indicates k The coordinate weight of each encoding result; u is the number of encoding results; d k represents the theoretical coordinates ( X m , Y m ) and k Encoding results ( X zk , Y zk )’s Euclidean distance; S5.2.3: Calculate optimized geographic coordinates:

[0017] In formula (6), w k Indicates k The coordinate weight of each encoding result; u is the number of encoding results; X zk and Y zk Respectively represent k The longitude and latitude of the encoded result; ( X 优 , Y 优 ) are the optimized geographic coordinates.

[0018] Furthermore, in step S2.3, the cosine similarity threshold R is 0.95.

[0019] By adopting the above technical solution, the beneficial effects that this solution can achieve are: This solution constructs a minimum geographic encirclement as a spatial constraint, calculates theoretical coordinates based on the POI data constraints in the minimum geographic encirclement, uses multiple geocoding platforms to generate coding results, and outputs theoretical coordinates or optimized geographic coordinates based on the relationship between multiple coding results and the minimum geographic encirclement, effectively considering the coding results of multiple geocoding platforms and improving the accuracy of coding results. It reduces the coding result errors caused by different databases of different geocoding platforms and different algorithms with similar semantics. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a logic flow chart of the optimization method; Figure 2 It is a schematic diagram of POI data in a POI address database; Figure 3 It is a schematic diagram of the POI data in the POI address database forming a minimum geographical encirclement; Figure 4 is the error histogram of the optimization method, other encoding platforms and the real address respectively; Figure 5 It is a process example diagram of this optimization method. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] See Figure 1 , a geographic coordinate optimization method based on a multi-source network geocoding platform with interest point constraints, comprising the following steps: S1: The user inputs a text address into the CRF model. The CRF model extracts address features from the input text address and divides the address features into coarse address features and precise address features.

[0023] The explanation of CRF model is: Conditional Random Field (CRF) model.

[0024] Suppose the text address input into the CRF model is: "No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province".

[0025] After the CRF model extracts the address features, the text address "No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province" is divided into rough address features and precise address features, where the rough address feature is: "Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province". The precise address feature is: "No. 17".

[0026] S2: Based on the rough address features, POI data is filtered in the POI address database to generate a minimum geographic encirclement.

[0027] The POI address database is downloaded from the open source POI address website.

[0028] Among them, POI is interpreted as: point of interest.

[0029] The POI address database contains several POI data. Figure 2 shown.

[0030] Step S2 specifically includes the following steps: S2.1: The rough address features and all POI data in the POI address database are input into the BERT model. The rough address features and the text addresses of all POI data are vectorized using the BERT model.

[0031] The POI data includes longitude, latitude and text address. Step S2.1 vectorizes only the text address of the POI data.

[0032] The BERT model is explained as follows: BERT (Bidirectional Encoder Representations from Transformers) is a pre-trained language model based on the Transformer architecture.

[0033] S2.2: Calculate the cosine similarity between the rough address feature and the text address of each POI data.

[0034] The formula is:

[0035] In formula (1), A is the vector representation of the rough address feature; B is a vector representation of a POI data text address in the POI address database; and They are A Vector Sum B L2 norm of the vector. CS is the cosine similarity. Represents the dot product of two vectors.

[0036] S2.3: Set the cosine similarity threshold R, filter the POI data according to the cosine similarity threshold R, and compare the vector A POI data with cosine similarity greater than or equal to R are marked as POI data of the encirclement to be constructed; the minimum geographic encirclement is constructed using the POI data of the encirclement to be constructed.

[0037] S2.3 specifically includes the following steps: S2.3.1: Combine with vector A POI data with cosine similarity greater than or equal to R are marked in the POI address database; the marked POI data are the POI data of the encirclement to be constructed; R is selected as 0.95 in this embodiment.

[0038] S2.3.2: Use the minimum polygon method to construct all the POI data of the encirclement to be constructed into a polygonal area. The constructed polygonal area is shown in Figure 3 The constructed polygonal area is the minimum geographic enclosing circle.

[0039] The cosine similarity between the text address of some POI data inside the constructed polygonal area and the rough address feature may be smaller than R, which has no impact on the overall solution.

[0040] S3: All POI data within the minimum geographic encirclement are used as a new POI database; query the new POI database, calculate theoretical coordinates based on precise address features, and mark the theoretical coordinates as ( X m , Y m ).

[0041] S3 specifically includes two situations: Case 1: If there is at least one house number of the POI data text address in the new POI database that is the same as the house number corresponding to the precise address feature, the calculation formula for the theoretical coordinates is:

[0042] In formula (2), X m and Y m are the longitude and latitude of the theoretical coordinates respectively; X i and Y i are the longitude and latitude of the POI data with the same features as the precise address, ; n is the number of POI data with the same exact address characteristics, n≥ 1, and it is an integer.

[0043] Explanation of the situation where multiple POI data text addresses and house numbers have the same address in the new POI database: Different stores on the east and west sides of a building may have the same house number, such as e The stores on the east and west sides are Store S1 and Store S2 respectively, but both Store S1 and Store S2 use the building e Therefore, different stores have the same house number. However, store S1 and store S2 have different geographical coordinates.

[0044] Case 2: If the new POI database does not contain the same house number as the precise address feature, then query the new POI database for two house numbers close to the house number with the precise address feature, and record them as C 1 andC 2. The calculation formula of theoretical coordinates is:

[0045] In formula (3), X m and Y m are the longitude and latitude of the theoretical coordinates respectively; C x The house number corresponding to the precise address feature; and House numbers C 1's longitude and latitude; and House numbers C 2's longitude and latitude.

[0046] Example: Assuming precise address characteristics C x The house number of the POI is "No. 17", and the house numbers "No. 16", "No. 18", "No. 19", and "No. 21" exist in the new POI database. Then the two house numbers close to "No. 17" are "No. 16" and "No. 18" respectively.

[0047] S4: Enter the text address in S1 into each geocoding platform respectively, obtain the encoding result of each geocoding platform, and form the encoding results of each geocoding platform into a candidate set.

[0048] Specifically, "No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province" is entered into various geocoding platforms, including but not limited to: Amap, Baidu, Sogou, Tencent, etc. Assume that there are u Geocoding platforms, each of which generates a coding result based on the text address, so a total of u encoding results, assuming that the encoding results are: ( X z1 , Y z1 )、( X z2 , Y z2 ),……、( X zu , Y zu ). The candidate set is: {( X z1 , Y z1 )、( X z2 , Y z2),……、( X zu , Y zu )}.

[0049] S5: If none of the encoding results in the candidate set is within the minimum geographic encirclement, then the theoretical coordinates are output ( X m , Y m ), theoretical coordinates ( X m , Y m ) is provided to the user as the final coding result; if at least one coding result in the candidate set is within the minimum geographic encirclement, the optimized geographic coordinates are calculated and recorded as ( X 优 , Y 优 ), output optimized geographic coordinates ( X 优 , Y 优 ), the optimized geographic coordinates ( X 优 , Y 优 ) is provided to users as the final encoding result.

[0050] S5 specifically includes the following steps: S5.1: If none of the encoding results in the candidate set is within the minimum geographic encirclement, then the theoretical coordinates are output ( X m , Y m ); Theoretical coordinates ( X m , Y m ) is provided to users as the final encoding result.

[0051] S5.2: If at least one encoding result in the candidate set is within the minimum geographic encirclement, calculate the optimized geographic coordinates, and record the optimized geographic coordinates as ( X 优 , Y 优 ), output optimized geographic coordinates ( X 优 , Y 优 ).

[0052] S5.2 specifically includes the following steps: S5.2.1: Calculate the Euclidean distance between each encoding result and the theoretical coordinates using the formula:

[0053] In formula (4), .

[0054] S5.2.2: Calculate the coordinate weight of each encoding result based on the Euclidean distance. The coordinate weight of each encoding result is:

[0055] In formula (5), w k Indicates k The coordinate weight of the encoding result. u is the number of encoding results. d k represents the theoretical coordinates ( X m , Y m ) and k Encoding results ( X zk , Y zk ) is the Euclidean distance.

[0056] S5.2.3: Calculate optimized geographic coordinates:

[0057] In formula (6), w k Indicates k The coordinate weight of the encoding result. u is the number of encoding results. X zk and Y zk Respectively represent k The longitude and latitude of the encoded result. ( X 优 , Y 优 ) is the optimized geographic coordinates, the optimized geographic coordinates ( X 优 , Y 优 ) is provided to users as the final encoding result.

[0058] Example analysis: In order to verify that this method can reduce the encoding result errors of multiple encoding platforms, this example selects 424 actual addresses and accurately obtains the geographic coordinates of the 424 actual addresses. The 424 actual addresses are all from Gulou District, Nanjing City, Jiangsu Province, and the address types are diverse. Due to the large amount of data, only part of the address data is provided. Some of the address data is shown in Table 1. Each actual address contains rough address features and precise address features.

[0059] Table 1 Actual address data

[0060] The real geographic coordinates corresponding to each actual address are obtained through manual coding and manual correction.

[0061] The coding results were obtained through the open API interfaces of the three existing coding platforms: Baidu, Amap and Tencent. Then the geographic coordinates are output through the optimization algorithm of this solution.

[0062] The statistical error data are shown in Table 2 and Figure 4 exhibit.

[0063] Table 2 Errors between the coding results of various types of geocoding platforms and the actual geographic coordinates

[0064] It can be seen from Table 2 that the optimization algorithm of this scheme has the smallest error with the actual true geographic coordinates, so the optimization algorithm of this scheme is conducive to reducing the geographic coordinate error.

[0065] from Figure 4 It can be seen that the error results of this optimization algorithm are mostly concentrated in the range of 0-50m. The number of errors decreases as the error range increases, and the number of errors greater than 500m is 0, which shows that this optimization algorithm can significantly reduce the large-scale errors greater than 500m.

[0066] Figure 5 It is a schematic diagram of the process of this optimization algorithm.

[0067] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A geographic coordinate optimization method based on a multi-source network geocoding platform with point of interest constraints, characterized in that: The following steps are involved: S1: The user inputs a text address into the CRF model, and the CRF model extracts address features from the input text address and divides the address features into rough address features and precise address features; S2: Based on the rough address features, POI data is filtered in the POI address database to generate the minimum geographic encirclement; S3: All POI data within the minimum geographic encirclement are used as a new POI database; query the new POI database, calculate theoretical coordinates based on precise address features, and mark the theoretical coordinates as ( X m , Y m ); S4: Enter the text address in S1 into each geocoding platform respectively, obtain the coding result of each geocoding platform, and form the coding result of each geocoding platform into a candidate set; S5: If none of the encoding results in the candidate set is within the minimum geographic encirclement, then the theoretical coordinates are output ( X m , Y m ); If at least one encoding result in the candidate set is within the minimum geographic encirclement, calculate the optimized geographic coordinates ( X 优 , Y 优 ), output optimized geographic coordinates ( X 优 , Y 优 ).

2. The geographic coordinate optimization method based on the multi-source network geocoding platform with point of interest constraints according to claim 1 is characterized in that: Step S2 includes the following steps: S2.1: The rough address features and all POI data in the POI address database are input into the BERT model; the BERT model vectorizes the rough address features and the text addresses of all POI data; S2.2: Calculate the cosine similarity between the rough address feature and the text address of each POI data; The formula is: ； In formula (1), A is the vector representation of the rough address feature; B is a vector representation of a POI data text address in the POI address database; Represents the dot product of two vectors; and They are A Vector Sum B L2 norm of a vector; CS is the cosine similarity; S2.3: Set the cosine similarity threshold R and compare it with the vector A POI data with cosine similarity greater than or equal to R are marked as POI data for the encirclement to be constructed; using the minimum polygon method, all POI data for the encirclement to be constructed are constructed into a polygonal area, and the constructed polygonal area is the minimum geographic encirclement.

3. The geographic coordinate optimization method based on the multi-source network geocoding platform with point of interest constraints according to claim 2 is characterized in that: Step S3 includes two cases: Case 1: If there is at least one house number of the POI data text address in the new POI database that is the same as the house number corresponding to the precise address feature, the calculation formula for the theoretical coordinates is: ； In formula (2), X m and Y m are the longitude and latitude of the theoretical coordinates; X i and Y i The longitude and latitude of the POI data with the same features as the precise address, , i is an integer; n is the number of POI data with the same exact address characteristics, n≥ 1; Case 2: If the new POI database does not contain the same house number as the precise address feature, then query the new POI database for two house numbers close to the house number with the precise address feature, and record them as C 1 and C 2, The calculation formula of the theoretical coordinates is: ； In formula (3), X m and Y m are the longitude and latitude of the theoretical coordinates respectively; C x The house number corresponding to the precise address feature; and House numbers C 1's longitude and latitude; and House numbers C 2's longitude and latitude.

4. The geographic coordinate optimization method based on the multi-source network geocoding platform with point of interest constraints according to claim 3 is characterized in that: In step S4, the coding results of each geocoding platform are recorded as: X z1 , Y z1 )、( X z2 , Y z2 ),……、( X zu , Y zu ), the candidate set is: {( X z1 , Y z1 )、( X z2 , Y z2 ),……、( X zu , Y zu )}, u is the number of encoding results.

5. The geographic coordinate optimization method based on the multi-source network geocoding platform with point of interest constraints according to claim 4 is characterized in that: Step S5 includes the following steps: S5.1: If none of the encoding results in the candidate set is within the minimum geographic encirclement, then the theoretical coordinates are output ( X m , Y m ); S5.2: If at least one encoding result in the candidate set is within the minimum geographic encirclement, calculate the optimized geographic coordinates, and record the optimized geographic coordinates as ( X 优 , Y 优 ), output optimized geographic coordinates ( X 优 , Y 优 ); S5.2 specifically includes the following steps: S5.2.1: Calculate the Euclidean distance between each encoding result and the theoretical coordinates using the formula: ; In formula (4), ;d k represents the theoretical coordinates ( X m , Y m ) and k Encoding results ( X zk , Y zk )’s Euclidean distance; S5.2.2: The coordinate weight of each encoding result is: ; In formula (5), w k Indicates k The coordinate weight of each encoding result; u is the number of encoding results; d k represents the theoretical coordinates ( X m , Y m ) and k Encoding results ( X zk , Y zk )’s Euclidean distance; S5.2.3: Calculate optimized geographic coordinates: ; In formula (6), w k Indicates k The coordinate weight of each encoding result; u is the number of encoding results; X zk and Y zk Respectively represent k The longitude and latitude of the encoded result; ( X 优 , Y 优 ) are the optimized geographic coordinates.

6. The geographic coordinate optimization method based on the multi-source network geocoding platform with point of interest constraints according to claim 2 is characterized in that: The cosine similarity threshold R in step S2.3 is 0.95.

Images