Address and coordinate based geocode error identification method

Abstract

The application provides a geographic coding error identification method based on addresses and coordinates, belongs to the technical field of geographic coding, and comprises the following steps. n S1: performing geographic coding on each text address through a geographic coding platform and converting each text address into corresponding coded coordinates in a WGS-84 coordinate system; each text address is in a standard format; S2: for any coded coordinate, error identification is performed on the coded coordinate to obtain an identification type of the coded coordinate; the identification type comprises a potential error type, a no-error type and an unidentifiable type; if the coded coordinate is a no-error type coded coordinate, an error value 0 of the no-error type coded coordinate is directly output; and S3: for the coded coordinate belonging to the potential error type, an error value of the coded coordinate is calculated. The method can determine the error value of the coded coordinate output by the geographic coding platform, thereby realizing error identification and precision evaluation of the geographic coding result and improving accurate understanding of the geographic coding.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of geocoding, and particularly relates to a geocoding error identification method based on an address and coordinates. BACKGROUND

[0002] Geocoding is a core technology for converting a text address into geographic coordinates, and has now been deeply integrated into daily life - from ordering takeout, calling a ride-hailing car to receiving and sending express, all of which need to rely on this technology to realize the association between an address and a spatial position. It is also the underlying support for map navigation and various location services, ensuring that the textual address information can be accurately mapped to a specific point on an electronic map. However, whether using Baidu, Gaode or other commercial platforms, the encoded coordinates often have large or small errors. Such errors will affect the efficiency of delivery personnel and the use of address data by researchers. For industries that rely on accurate locations, how to evaluate and identify these errors has become a difficult problem to be solved.

[0003] Currently, there are mainly two ideas for evaluating the errors of these coordinates.

[0004] The first one is to "check with standard answers". That is, using the high-precision address library measured by the government or the "true coordinates" collected by manually carrying a GPS device in the field as a standard to verify the errors of commercial coding. Although this method is direct, for most ordinary users and enterprises, the cost of obtaining "standard answers" is extremely high, the threshold is extremely high, and even involves confidential data, which is not realistic to popularize and popularize.

[0005] The second one is to "calculate by theory". This method does not pursue true coordinates, but tries to analyze the fuzziness of the input address and the stability of the coding algorithm itself to calculate the possible range of errors. However, the problem is that the fuzziness of the address and the error characteristics of the algorithm are not constant, and will dynamically change due to differences in regions, platforms and even algorithm versions, resulting in a large deviation between the calculation results and the actual situation.

[0006] For users using geocoding data, there are only input addresses and coordinates returned by the platform, which are not known to be accurate or not. Therefore, there is an urgent need for a method that can use these ready-made data to identify coding errors, so as to perform error identification and precision evaluation on the results of geocoding, and improve the accurate understanding of geocoding. SUMMARY

[0007] The application proposes a geocoding error identification method based on an address and coordinates, which can evaluate the errors of the coordinates returned by the geocoding platform.

[0008] In order to achieve the above purpose, the application proposes the following technical content:

[0009] A method for identifying address and coordinate-based geocoding errors, comprising the following steps:

[0010] S1: Perform geocoding on n textual addresses through a certain geocoding platform, and convert them into corresponding encoded coordinates in the WGS-84 coordinate system; each textual address is in standard format, i.e., in the form of: xx City xx District xx Road xx building name / facility name;

[0011] S2: For any encoded coordinate, perform error identification to obtain the identification type of the encoded coordinate; the identification type includes: potential error type, no error type, and unidentifiable type; if the encoded coordinate is a no error type encoded coordinate, directly output the error value 0 of the no error type encoded coordinate;

[0012] S3: For encoded coordinates belonging to the potential error type, calculate the error value; specifically comprising the following steps:

[0013] S3.1: Set the j encoded coordinate x j , y j ) belongs to the potential error type, j ∈[1, n ], and is an integer, set the search radius k R k established in the

[0014] S3.2: Take the j encoded coordinate x j , y j ) as the center, and the current search radius R k to search, when at least 3 encoded coordinates of no error type are found, first stop expanding the search radius, and calculate the Z-score score j j of the x j encoded coordinate y R k under the current search radius Z j ;

[0015] S3.3: If Z j is greater than the set threshold value Z m , then the j encoded coordinate x ​​j , y j The semantic similarity between the text address corresponding to the error-free type of encoding coordinates and the text address corresponding to the error-free type of encoding coordinates differs greatly; conversely, the semantic similarity difference is small. When the semantic similarity difference is small, S3.1-S3.2 are repeated, and the search radius is expanded in each iteration. R k And recalculate Z j This continues until the semantic similarity differs significantly; when the semantic similarity differs significantly, the first step is to obtain the [missing information]. j Each encoded coordinate ( x j , y j The corresponding error value E j ;

[0016] S3.4: For the first j Each encoded coordinate ( x j , y j ), within the corresponding search radius R k If its Z-score is consistently less than the set threshold Z, then... m If the number of error-free type encoded coordinates is always less than 3, then first calculate the Euclidean distance-semantic similarity correspondence using all error-free type encoded coordinates, and then use the first... j Each encoded coordinate ( x j , y j The semantic similarity between the coordinates of the first and second errors and the average semantic similarity of the coordinates of the first error type within a set range E is used to determine the first error type. j Each encoded coordinate ( x j , y j The corresponding error value.

[0017] Furthermore, step S2 specifically includes the following steps:

[0018] S2.1: For the first i Each encoded coordinate ( x i , y i ), i ∈[1, n ], and is an integer, with its coordinates ( x i , y i) a spatial query window is constructed with the center and radius R;

[0019] S2.2: Traverse all the encoded coordinates in the spatial query window, and divide them into two cases:

[0020] The first case: there is no other encoded coordinate in the spatial query window except the first encoded coordinate, then the first encoded coordinate belongs to the unrecognizable type; i x i , y i The second case: there is at least one other encoded coordinate in the spatial query window except the first encoded coordinate, then the first encoded coordinate does not belong to the unrecognizable type; i x i , y i

[0021] The second case: there is at least one other encoded coordinate in the spatial query window except the first encoded coordinate, then the first encoded coordinate does not belong to the unrecognizable type; i x i , y i i x i , y i

[0022] S2.3: When the first encoded coordinate does not belong to the unrecognizable type, judge whether the second encoded coordinate belongs to the potential error type or the no error type; specifically including the following steps: i x i , y i i x i , y i

[0023] S2.3.1: Obtain the text addresses corresponding to the other encoded coordinates in the spatial query window except the first encoded coordinate, and include them in the neighborhood address set i x i , y i N

[0024] S2.3.2: Calculate the distance between each text address in the neighborhood address set N and the first encoded coordinate i x i , y ​​​​​​​​​​​​​​​i semantic similarity between corresponding text addresses, and compared with a set threshold T s comparison; according to the comparison result, judging the first i encoded coordinate x i , y i belongs to a potential error type or a no error type;

[0025] The formula is:

[0026]

[0027] In the formula, S i,x indicates the first i encoded coordinate x i , y i semantic similarity between corresponding text addresses, and compared with a set threshold x U a indicates the first i encoded coordinate x i , y i matching degree of "city word segment" after text segmentation of the corresponding text address of the first x text address in the neighborhood address set, if matching, then U a =1; if not matching, then U a =0; U b indicates the first i encoded coordinate x i , y i matching degree of "district word segment" after text segmentation of the corresponding text address of the first x text address in the neighborhood address set, if matching, then U b =1; if not matching, then U b =0; U c indicates the first i encoded coordinate x i , y i ​The corresponding text address, after word segmentation, is the "street segment" and its corresponding neighborhood address in the set of neighboring addresses. x After segmenting the text address into words, the matching degree of the "street segment" is calculated. If a match is found, U c =1; if not matched, then U c =0; U d Indicates the first i Each encoded coordinate ( x i , y i The corresponding text address, after word segmentation, is "building name / facility name", which is related to the first word in the neighborhood address set. x The formula for the matching degree of "building name / facility name" after word segmentation of each text address is:

[0028]

[0029] In the formula, LCS (*,*) represents calculating the length of the longest common subsequence between two strings; len (*) indicates that the length of the character is calculated; L i Indicates the first i Each encoded coordinate ( x i , y i The corresponding text address, after word segmentation, is "building name / facility name". L x No. x After word segmentation of the text address, "Building Name / Facility Name" is used. If it is established, then the first i Each encoded coordinate ( x i , y i ) belongs to the error-free type; if Not valid, number i Each encoded coordinate ( x i , y i This belongs to the type with potential errors; A x Represents the set of neighborhood addresses. x A text address.

[0030] Further, in steps S3.2-S3.3, the calculation of the first... j Each encoded coordinate ( x j , y j)th encoding coordinate (x R k th encoding coordinate (x

[0031]

[0032] Z j th encoding coordinate (x j x j y j th encoding coordinate (x R k th encoding coordinate (x j x j y j th encoding coordinate (x

[0033] th encoding coordinate (x j x j y j th encoding coordinate (x R k Z j th encoding coordinate (x

[0034] th encoding coordinate (x j x j y j th encoding coordinate (x

[0035]

[0036] E j th encoding coordinate (x​​​​​​​​​​​​​​​j the corresponding error value of the x j , y j the corresponding error value of the R k the value of the current search radius.

[0037] Further, step S3.4 comprises the following steps:

[0038] S3.4.1: Calculate all error-free type of encoding coordinates Euclidean distance-semantic similarity corresponding relationship;

[0039] Statistical step S2 all error-free type of encoding coordinates, calculate the following data:

[0040] I. The semantic similarity between the text addresses corresponding to any two error-free type of encoding coordinates;

[0041] II. The Euclidean distance between any two error-free type of encoding coordinates;

[0042] S3.4.2: Set error interval; Specifically: [0, 50), [50, 150), [150, 250), [250, 350), [350, 450), [450, 550), [550, 650), [650, 750), [750, 850), [850, 950), [950, + ∞).

[0043] S3.4.3: According to the Euclidean distance calculated in S3.4.1, obtain the average semantic similarity of the text addresses located in the same error interval;

[0044] S3.4.4: Calculate the j the corresponding text address of the x j , y j the average semantic similarity of the text addresses corresponding to the multiple error-free type of encoding coordinates within the range with the j the corresponding text address of the x j , y j the average semantic similarity of the text addresses corresponding to the multiple error-free type of encoding coordinates within the range with the E

[0045]

[0046]

[0047] In the formula, S j,d the corresponding text address of the j the corresponding text address of the​​x j , y j The text address corresponding to ) and the first d Semantic similarity between text addresses corresponding to error-free encoded coordinates. d ∈[1, P ], and is an integer; P Indicates the first j Each encoded coordinate ( x j , y j With ) as the center, within the set range E The number of coordinates with no error type encoding; Indicates the first j Each encoded coordinate ( x j , y j The corresponding text address, within the set range E The average semantic similarity between the text addresses corresponding to the encoded coordinates within the range;

[0048] S3.4.5: Will Compare with all index values ​​in S3.4.3 to obtain the information related to... The closest index value is used to determine the index. j Each encoded coordinate ( x j , y j The corresponding error range;

[0049] When the j Each encoded coordinate ( x j , y j When the corresponding error interval does not belong to [0, 50) and [950, +∞), the median value of the error interval is taken as the first... j Each encoded coordinate ( x j , y j The corresponding error value; when the first j Each encoded coordinate ( x j , y j If the corresponding error interval is [0, 50), then the first... j Each encoded coordinate ( x j , y j This belongs to the error-free type, and the output error value is 0; when the first... jEach encoded coordinate ( x j , y j When the corresponding error range is [950, +∞), the output error value is 1000m.

[0050] Furthermore, the radius R = 50m; Z m =2.

[0051] The beneficial effects that can be achieved by adopting the above technologies are:

[0052] 1. This solution can obtain the error value of coded coordinates with potential error types, thereby realizing the error identification and accuracy assessment of geocoding results and improving the accurate understanding of geocoding.

[0053] 2. It can identify the magnitude of address errors without relying on any high-precision external data, and even without real coordinates, administrative divisions, road data, etc. It is applicable to various commercial geocoding platforms, avoids black-box algorithms, and does not depend on the confidence level provided by the platform. It can roughly determine the spatial distribution pattern and specific error quantification level simply by using the platform's API to obtain the encoded coordinates. Attached Figure Description

[0054] Figure 1 This is a flowchart of the method;

[0055] Figure 2 This is the logic diagram of this solution;

[0056] Figure 3 It is the first j Each encoded coordinate ( x j , y j A diagram illustrating the semantic difference between this and other text addresses.

[0057] Figure 4 It is the first j Each encoded coordinate ( x j , y j This is a diagram illustrating a situation where the semantics of the address differ significantly from those of other text addresses.

[0058] Figure 5 This is a distribution diagram of the number of errors in the example;

[0059] Figure 6 It is a distribution chart showing the number of each difference after subtracting the recognition error from the actual error. Detailed Implementation

[0060] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0061] As shown in Figure 1 and Figure 2 , the address and coordinate-based geographic coding error identification method comprises the following steps:

[0062] S1: performing geographic coding on n text addresses by a certain geographic coding platform. For any text address, the geographic coding platform first converts it into an encoded coordinate in its own coordinate system, and then converts it into an encoded coordinate in the WGS-84 coordinate system, and finally outputs the corresponding encoded coordinate in the WGS-84 coordinate system. Each text address is a standard format text address, i.e., the form conforms to: xx city xx district xx road xx building name / facility name;

[0063] For example, using the API interface provided by Gaode Map, performing geographic coding on n text addresses, obtaining n encoded coordinates in the GCJ-02 coordinate system, and converting them into encoded coordinates in the WGS-84 coordinate system.

[0064] S2: For any encoded coordinate, performing error identification to obtain the identification type of the encoded coordinate; the identification type includes: potential error type, no error type and unidentifiable type; if the encoded coordinate is a no error type encoded coordinate, the corresponding error value is 0. Specifically, it comprises the following steps:

[0065] S2.1: for the i encoded coordinate( x i , y i ), i ∈[1, n ], and is an integer, constructing a spatial query window with its coordinate( x i , y i ) as the center and radius R; R is a set value, generally 50m.

[0066] S2.2: traversing all encoded coordinates in the spatial query window, and dividing them into two cases:

[0067] The first case: excluding the ithe first encoded coordinate x i , y i ) is not identifiable, then the first encoded coordinate i x i , y i ) is of an unidentifiable type. The unidentifiable type means that the error of the first encoded coordinate i x i , y i ) cannot be obtained.

[0068] The second case: except for the first encoded coordinate i x i , y i ), there is at least one other encoded coordinate in the spatial query window, then the first encoded coordinate i x i , y i ) is not of the unidentifiable type.

[0069] S2.3: When the first encoded coordinate i x i , y i ) is not of the unidentifiable type, it is determined whether the first encoded coordinate i x i , y i ) is of a potential error type or a no error type. Specifically, the following steps are included:

[0070] S2.3.1: Obtain the text addresses corresponding to the other encoded coordinates in the spatial query window except for the first encoded coordinate i x i , y i ), and include them in a neighborhood address set N , and the formula is:

[0071]

[0072] In the formula, N represents the number of text addresses in the neighborhood address set; m A x represents the first encoded coordinate x ​​​​​​​​A text address.

[0073] S2.3.2: Calculate the neighborhood address set N Each text address is compared with the semantic similarity between the text address corresponding to the first i coding coordinate x i , y i and the set threshold value T s ; according to the comparison result, it is judged that the first i coding coordinate x i , y i belongs to the potential error type or the error-free type.

[0074] The formula is:

[0075]

[0076] In the formula, S i,x , represents the semantic similarity between the text address corresponding to the first i coding coordinate x i , y i and the first x text address in the neighborhood address set; 0.4, 0.3, 0.2 and 0.1 represent the weight;

[0077] U a , represents the matching degree of the "city word segment" after the text address corresponding to the first i coding coordinate x i , y i is segmented, and the "city word segment" after the first x text address in the neighborhood address set is segmented, if it matches, U a =1; if it does not match, U a =0; for example: the text address corresponding to the first i coding coordinate x i , y i is: GG City KK District PP Road DD Mansion, which is segmented into: "GG City", "KK District", "PP Road" and "DD Mansion"; the first x text address in the neighborhood address set is: "HH City JJ District", which is segmented into: "HH City" and "JJ District".The text address is: GG City KK District PP Road HH Building, which is segmented as: "GG City", "KK District", "PP Road", and "HH Building". Since the "city word segment" of both is "GG City", they match, so U a = 1.

[0078] U b represents the matching degree of the "district word segment" after segmenting the text address corresponding to the first i x i , y i x U b = 1 if matched; otherwise, U b = 0; as mentioned above, since the "district word segment" of both is "KK District", they match, so U b = 1.

[0079] U c represents the matching degree of the "street word segment" after segmenting the text address corresponding to the first i x i , y i x U c = 1 if matched; otherwise, U c = 0; as mentioned above, since the "street word segment" of both is "PP Road", they match, so U c = 1.

[0080] U d represents the matching degree of the "building name / facility name" after segmenting the text address corresponding to the first i x i , y i x

[0081]

[0082] In the formula,​​​​​​​​​LCS (*) means calculating the length of the longest common subsequence between two strings; len (*) means calculating the length of the character; L i means the i th encoding coordinate x i , y i corresponding to the text address after segmentation "building name / facility name"; L x means the x th text address after segmentation "building name / facility name".

[0083] If is true, the i th encoding coordinate x i , y i belongs to the error-free type; directly output the error value 0 of the error-free type encoding coordinate.

[0084] If is not true, the i th encoding coordinate x i , y i belongs to the potential error type. T s means the threshold value set as 0.9.

[0085] S3: For the encoding coordinates belonging to the potential error type, calculate their error values. Specifically, the following steps are included:

[0086] S3.1: Set the j th encoding coordinate x j , y j belongs to the potential error type, j ∈[1, n ], and is an integer; Set the search radius established at the k th expansion as:

[0087]

[0088] In the formula, R k means the search radius established at the k th expansion, unit m;

[0089] S3.2: Take the j th encoding coordinate xj , y j The current search radius is centered at ). R k When a search is performed and at least three error-free coded coordinates are found, the search radius is no longer expanded. Then, the following parameters are calculated respectively:

[0090] No. j Each encoded coordinate ( x j , y j The average semantic similarity between the text address corresponding to the error-free type of encoded coordinates and the text addresses corresponding to the error-free type of encoded coordinates. ;

[0091] The average semantic similarity between the text addresses corresponding to the encoded coordinates of multiple error-free types found in the search. ;

[0092] The standard deviation of semantic similarity among the text addresses corresponding to multiple error-free encoded coordinates found in the search. ;

[0093] Calculate the first j Each encoded coordinate ( x j , y j In the current search radius R k The Z-score is calculated based on the set threshold Z. m , judge the first j Each encoded coordinate ( x j , y j The semantic similarity between the text address corresponding to the search result and the text addresses corresponding to the multiple error-free type encoded coordinates is calculated. If the semantic similarity difference is small, steps S3.1-S3.2 are repeated, expanding the search radius in each iteration until the semantic similarity difference becomes large. If the semantic similarity difference is large, the result is obtained. j Each encoded coordinate ( x j , y j The corresponding error value.

[0094] Specifically, calculate the first j Each encoded coordinate ( x j , y j In the current search radius R kThe Z-score is calculated using the following formula:

[0095]

[0096] In the formula, Z j Indicates the first j Each encoded coordinate ( x j , y j In the current search radius R k The Z-score score is below;

[0097] S3.3: If If it is established, then the first j Each encoded coordinate ( x j , y j The semantic similarity between the text address corresponding to the error-free type of encoding coordinates and the text address corresponding to the error-free type of encoding coordinates differs greatly; conversely, the semantic similarity difference is small. Z m The threshold value is set.

[0098] Figure 3 The first j Each encoded coordinate ( x j , y j A diagram showing the semantic difference between this and other text addresses. Figure 4 The first j Each encoded coordinate ( x j , y j This is a diagram illustrating a situation where the semantics of the address differ significantly from those of other text addresses.

[0099] When the semantic similarity difference is small, repeat steps S3.1-S3.2, expanding the search radius in each iteration. R k And recalculate Z j until the semantic similarity differs greatly;

[0100] When the semantic similarity differs greatly, the first... j Each encoded coordinate ( x j , y j The corresponding error value is:

[0101]

[0102] In the formula, E j Indicates the first j Each encoded coordinate ( x j , y j The corresponding error value; R k This is the current search radius.

[0103] S3.4: For the first j Each encoded coordinate ( x j , y j ), within the corresponding search radius R k If its Z-score is consistently less than the set threshold Z, then... m If the number of error-free type encoded coordinates is always less than 3, then first calculate the Euclidean distance-semantic similarity correspondence using all error-free type encoded coordinates, and then use the first... j Each encoded coordinate ( x j , y j ) and the set range E The average semantic similarity of the encoded coordinates of multiple error-free types within the range is used to determine the semantic similarity of the first-order coordinate. j Each encoded coordinate ( x j , y j The corresponding error value. Specifically, it includes the following steps:

[0104] S3.4.1: Calculate the Euclidean distance-semantic similarity correspondence between the encoded coordinates of all error-free types.

[0105] Statistically analyze the encoded coordinates of all error-free types in step S2, and calculate the following data:

[0106] 1. Semantic similarity between the text addresses corresponding to any two error-free encoded coordinate types;

[0107] 2. The Euclidean distance between any two error-free type coded coordinates;

[0108] For example: Suppose that there are three error-free type encoded coordinates in step S2: T1, T2 and T3; and the text addresses corresponding to the three encoded coordinates T1, T2 and T3: A1, A2 and A3.

[0109] First, calculate the Euclidean distance between the pairwise error-free encoded coordinates: that is, the Euclidean distance D12 between T1 and T2; the Euclidean distance D13 between T1 and T3; and the Euclidean distance D23 between T2 and T3.

[0110] Next, calculate the semantic similarity between the text addresses corresponding to the encoded coordinates of each pair of error-free types: that is, the semantic similarity between A1 and A2 is S12; the semantic similarity between A1 and A3 is S13; and the semantic similarity between A2 and A3 is S23.

[0111] S3.4.2: Set the error range; specifically: [0,50), [50,150), [150,250), [250,350), [350,450), [450,550), [550,650), [650,750), [750,850), [850,950), [950, +∞).

[0112] S3.4.3: Based on the Euclidean distance calculated in S3.4.1, obtain the average semantic similarity of text addresses located in the same error interval;

[0113] Based on the Euclidean distance calculated in S3.4.1, the semantic similarity of elements belonging to the same error interval is statistically analyzed, the average semantic similarity in that interval is calculated, and the average semantic similarity is used as the index value of that error interval.

[0114] Continuing with the example in S3.4.1, assuming D12 and D13 are 80 and 90 respectively, it can be seen that they both belong to the interval [50, 150). Then, the average semantic similarity corresponding to this interval is calculated, i.e.: Q1 = 1 / 2 × (S12 + S13).

[0115] Assuming D23 is 180, it can be known that it belongs to the interval [150, 250). Next, the average semantic similarity corresponding to this interval is calculated. Since there is only one, Q2 = S23.

[0116] S3.4.4: Calculate the first... j Each encoded coordinate ( x j , y j The corresponding text address, and the first j Each encoded coordinate ( x j , y j (Set the range within the center of the circle) E The average semantic similarity of text addresses corresponding to multiple error-free encoded coordinate types within the range; E =1000m;

[0117] The formula is:

[0118]

[0119] In the formula, S j,d Indicates the first j Each encoded coordinate ( x j , y j The text address corresponding to ) and the first d Semantic similarity between text addresses corresponding to error-free encoded coordinates. d ∈[1, P ], and is an integer; P Indicates the first j Each encoded coordinate ( x j , y j With ) as the center, within the set range E The number of coordinates with no error type encoding; Indicates the first j Each encoded coordinate ( x j , y j The corresponding text address, within the set range E The average semantic similarity between text addresses corresponding to multiple error-free type encoded coordinates within the range.

[0120] S3.4.5: Will Compare with all index values ​​in S3.4.3 to obtain the information related to... The closest index value is used to determine the index. j Each encoded coordinate ( x j , y j The corresponding error range is determined, and the error value is obtained based on the corresponding error range.

[0121] Continuing with the example in S3.4.3, let's assume... After calculation, it equals Q3. Compare Q3 with the previously mentioned Q1 and Q2. If Q3 is smaller than Q1, then the... j Each encoded coordinate ( x j , y j If the error range is [50, 150), then the first... j Each encoded coordinate ( x j , y jIf Q3 is within the error range of [150, 250), and Q1 and Q2 are the same distance apart, then the larger error range, i.e., [150, 250), is taken.

[0122] When the j Each encoded coordinate ( x j , y j If the error interval corresponding to ) does not belong to either [0, 50) or [950, +∞), then the 1st error interval is taken. j Each encoded coordinate ( x j , y j The median of the corresponding error interval is used as the first... j Each encoded coordinate ( x j , y j The error value of ). For example: the first j Each encoded coordinate ( x j , y j If the error range is [150, 250), then the first... j The error value corresponding to the coded coordinate is 200, output the coded coordinate. j Error value of each coded coordinate.

[0123] When the j Each encoded coordinate ( x j , y j If the error interval corresponding to ) belongs to [0, 50), then the first ) is considered to be 0. j Each encoded coordinate ( x j , y j This belongs to the error-free type, and the output is the first... j Each encoded coordinate ( x j , y j The corresponding error value is 0.

[0124] When the j Each encoded coordinate ( x j , y j When the corresponding error interval is [950, +∞), the first... j Each encoded coordinate ( x j , y jThe corresponding error value is 1000 m .

[0125] Example:

[0126] In order to verify the feasibility of the scheme, a data set of 9500 POI addresses is collected in the GG city area, each address has a complete city-district-street-building name / facility name hierarchy, and the address also has real WGS84 coordinates and coded coordinates obtained by using a network coding platform.

[0127] The real error of each POI address is calculated, wherein the real error is the distance between the WGS84 coordinates of each POI address and the coded coordinates obtained by the network coding platform;

[0128] The identification error is: the error obtained after the scheme is processed;

[0129] For 9500 POI addresses, the relevant data of the real error and the identification error are shown in Table 1.

[0130] Table 1 Real error and coding error result statistics

[0131]

[0132] From the data in Table 1, it can be seen that:

[0133] In terms of quartiles: the real error and the identification error error distribution of the scheme are basically the same, which shows that the scheme can be more objective to meet the error distribution rule.

[0134] In terms of error mean: the error mean of the identification error of the scheme is smaller than the error mean of the real error, which shows that the identification error of the scheme is more accurate. Although in terms of standard deviation, the identification error of the scheme is not as good as the real error, but it does not affect the overall effect of the scheme.

[0135] Figure 5 The distribution graph of each error number is shown, for 9500 POI addresses, a total of 7030 POI addresses are applied, of which 2470 POI addresses are not identified.

[0136] From Figure 5 It can be seen that the number of 0 errors of the scheme is the most, and is greater than the number of real errors; this shows that the scheme has high credibility for the error-free result.

[0137] Figure 6The difference between the identification error and the real error is subtracted and the absolute value is taken. For 7030 POI addresses, the number of difference values of 0 is the largest, and the data with a difference value within 500m reaches 5007, accounting for 71.2% of the total data, which shows that the recognition effect of the scheme on the overall data coding coordinate error is remarkable and the accuracy is high.

[0138] The above ideal embodiments according to the application are used as inspiration, and through the above description, relevant personnel can make various changes and modifications without deviating from the technical idea of the application. The technical scope of the application is not limited to the content in the specification, and the technical scope must be determined according to the scope of claims.

Claims

1. A geocoding error identification method based on address and coordinates, characterized in that, Includes the following steps: S1: Through a certain geocoding platform n Each text address is geocoded and converted to the corresponding coded coordinates in the WGS-84 coordinate system; each text address is in a standard format, i.e., the format conforms to: xx city xx district xx road xx building name / facility name; S2: For any coded coordinate, perform error identification to obtain the identification type of the coded coordinate; the identification type includes: type with potential error, type without error, and unidentifiable type; if the coded coordinate is a type without error, directly output the error value of the type without error coded coordinate as 0; S3: For encoded coordinates that belong to the type with potential errors, calculate their error values; specifically, this includes the following steps: S3.1: Setting the first j Each encoded coordinate ( x j , y j This belongs to the type with potential errors. j ∈[1, n ], and is an integer, set the number of... k The search radius established during the subsequent expansion R k ; S3.2: with the first j Each encoded coordinate ( x j , y j The current search radius is centered at (). R k When a search is performed and at least three error-free coded coordinates are found, the search radius is no longer expanded, and the calculation of the... j Each encoded coordinate ( x j , y j In the current search radius R k Z-score Z j ; S3.3: If Z j Greater than the set threshold Z m If it is established, then the first j Each encoded coordinate ( x j , y j The semantic similarity between the text address corresponding to the error-free type of encoding coordinates and the text address corresponding to the error-free type of encoding coordinates differs greatly; conversely, the semantic similarity difference is small. When the semantic similarity difference is small, S3.1-S3.2 are repeated, and the search radius is expanded in each iteration. R k And recalculate Z j This continues until the semantic similarity differs significantly; when the semantic similarity differs significantly, the first step is to obtain the [missing information]. j Each encoded coordinate ( x j , y j The corresponding error value E j ; S3.4: For the first j Each encoded coordinate ( x j , y j ), within the corresponding search radius R k If its Z-score is consistently less than the set threshold Z, then... m If the number of error-free type encoded coordinates is always less than 3, then first calculate the Euclidean distance-semantic similarity correspondence using all error-free type encoded coordinates, and then use the first... j Each encoded coordinate ( x j , y j The semantic similarity between the coordinates of the first and second errors and the average semantic similarity of the coordinates of the first error type within a set range E is used to determine the first error type. j Each encoded coordinate ( x j , y j The corresponding error value.

2. The geocoding error identification method based on address and coordinates according to claim 1, characterized in that, Step S2 specifically includes the following steps: S2.1: For the first i Each encoded coordinate ( x i , y i ), i ∈[1, n ], and is an integer, with its coordinates ( x i , y i Construct a spatial query window centered on a radius R; S2.2: Traverse all the encoded coordinates in the spatial query window, dividing them into two cases: First case: Except for the first i Each encoded coordinate ( x i , y i If, apart from ), there are no other coded coordinates in the spatial query window, then the 1st... i Each encoded coordinate ( x i , y i This belongs to an unrecognizable type; The second case: except for the first i Each encoded coordinate ( x i , y i Besides, if at least one other coded coordinate exists in the spatial query window, then the first... i Each encoded coordinate ( x i , y i This does not belong to an unrecognizable type; S2.3: When the first i Each encoded coordinate ( x i , y i If it does not belong to an unrecognizable type, determine the first... i Each encoded coordinate ( x i , y i This belongs to either a type with potential errors or a type without errors; specifically, it includes the following steps: S2.3.1: Obtain the remainder of the first... i Each encoded coordinate ( x i , y i In addition to the text addresses corresponding to other coded coordinates in the spatial query window, the address set of neighboring addresses is also included. N middle; S2.3.2: Calculate the neighborhood address set N Each text address and the first i Each encoded coordinate ( x i , y i The semantic similarity between the corresponding text addresses is compared with a set threshold. T s Compare; based on the comparison results, determine the first... i Each encoded coordinate ( x i , y i It belongs to either a type with potential errors or a type without errors; The formula is: ; In the formula, S i,x Indicates the first i Each encoded coordinate ( x i , y i The text address corresponding to ) and the in the neighborhood address set x Semantic similarity between text addresses; 0.4, 0.3, 0.2, and 0.1 represent weights; U a Indicates the first i Each encoded coordinate ( x i , y i After segmenting the text address corresponding to the "city segment", the "city segment" is compared with the first segment in the neighborhood address set. x After segmenting the text address into words, the matching degree of the "city segment" is calculated. If a match is found, then... U a =1; if not matched, then U a =0; U b Indicates the first i Each encoded coordinate ( x i , y i The corresponding text address, after word segmentation, is the "region segment" and its relation to the neighbor address set. x After segmenting a text address into words, the matching score of the "segment" is calculated. If a match is found, then... U b =1; if not matched, then U b =0; U c Indicates the first i Each encoded coordinate ( x i , y i The corresponding text address, after word segmentation, is the "street segment" and its corresponding neighbor address in the set of neighbor addresses. x After segmenting the text address into words, the matching degree of the "street segment" is calculated. If a match is found, U c =1; if not matched, then U c =0; U d Indicates the first i Each encoded coordinate ( x i , y i After word segmentation of the corresponding text address, "building name / facility name" is compared with the first word in the neighborhood address set. x The formula for the matching degree of "building name / facility name" after word segmentation of each text address is: ; In the formula, LCS (*,*) represents calculating the length of the longest common subsequence between two strings; len (*) indicates that the length of the character is calculated; L i Indicates the first i Each encoded coordinate ( x i , y i The corresponding text address, after word segmentation, is "building name / facility name". L x No. x After word segmentation of the text address, "building name / facility name", if If it is established, then the first i Each encoded coordinate ( x i , y i ) belongs to the error-free type; if Not valid, number i Each encoded coordinate ( x i , y i This belongs to the type with potential errors; A x Represents the set of neighborhood addresses. x A text address.

3. The geocoding error identification method based on address and coordinates according to claim 2, characterized in that, In steps S3.2-S3.3, calculate the first... j Each encoded coordinate ( x j , y j In the current search radius R k The Z-score is calculated using the following formula: ; In the formula, Z j Indicates the first j Each encoded coordinate ( x j , y j In the current search radius R k The Z-score score is below; Indicates the first j Each encoded coordinate ( x j , y j The average semantic similarity between the text address corresponding to the searched text address and the text address corresponding to the encoded coordinates of multiple error-free types; This represents the average semantic similarity between the text addresses corresponding to multiple error-free encoded coordinates found during the search; This represents the standard deviation of semantic similarity among the text addresses corresponding to multiple error-free encoded coordinates found during the search. like If it is established, then the first j Each encoded coordinate ( x j , y j The semantic similarity between the text address corresponding to the error-free type of encoding coordinates and the text address corresponding to the error-free type of encoding coordinates differs greatly; conversely, the semantic similarity difference is small. When the semantic similarity difference is small, repeat steps S3.1-S3.2, expanding the search radius in each iteration. R k And recalculate Z j until the semantic similarity differs greatly; When the semantic similarity differs greatly, the first... j Each encoded coordinate ( x j , y j The corresponding error value is: In the formula, E j Indicates the first j Each encoded coordinate ( x j , y j The corresponding error value; R k This is the current search radius.

4. The geocoding error identification method based on address and coordinates according to claim 1, characterized in that, Step S3.4 includes the following steps: S3.4.1: Calculate the Euclidean distance-semantic similarity correspondence between the encoded coordinates of all error-free types; Statistically analyze the encoded coordinates of all error-free types in step S2, and calculate the following data:

1. Semantic similarity between the text addresses corresponding to any two error-free encoded coordinate types; 2. The Euclidean distance between any two error-free type coded coordinates; S3.4.2: Set the error range; specifically: [0,50), [50,150), [150,250), [250,350), [350,450), [450,550), [550,650), [650,750), [750,850), [850,950), [950, +∞); S3.4.3: Based on the Euclidean distance calculated in S3.4.1, obtain the average semantic similarity of text addresses located in the same error interval; S3.4.4: Calculate the first... j Each encoded coordinate ( x j , y j The corresponding text address, and the first j Each encoded coordinate ( x j , y j With ) as the center, within the set range E The average semantic similarity of text addresses corresponding to multiple error-free encoded coordinate types within the range; The formula is: ; In the formula, S j,d Indicates the first j Each encoded coordinate ( x j , y j The text address corresponding to ) and the first d Semantic similarity between text addresses corresponding to error-free encoded coordinates. d ∈[1, P ], and is an integer; P Indicates the first j Each encoded coordinate ( x j , y j With ) as the center, within the set range E The number of coordinates with no error type encoding; Indicates the first j Each encoded coordinate ( x j , y j The corresponding text address, within the set range E The average semantic similarity between text addresses corresponding to multiple error-free encoded coordinates; S3.4.5: Will Compare with all index values ​​in S3.4.3 to obtain the information related to... The closest index value is used to determine the index. j Each encoded coordinate ( x j , y j The corresponding error range; When the j Each encoded coordinate ( x j , y j When the corresponding error interval does not belong to [0, 50) and [950, +∞), the median value of the error interval is taken as the first... j Each encoded coordinate ( x j , y j The corresponding error value; when the first j Each encoded coordinate ( x j , y j If the corresponding error interval is [0, 50), then the first... j Each encoded coordinate ( x j , y j This belongs to the error-free type, and the output error value is 0; when the first... j Each encoded coordinate ( x j , y j When the corresponding error range is [950, +∞), the output error value is 1000m.

5. The geocoding error identification method based on address and coordinates according to claim 3, characterized in that, Radius R = 50m; Z m =2.

Images