Coding result optimization method of coding platform based on address distribution density

Abstract

The invention provides a coding result optimization method of a coding platform based on address distribution density, which belongs to the technical field of geocoding result data processing, and comprises the following steps: S1, carrying out word segmentation on a text address; s2, searching a large house number coordinate and a small house number coordinate (A and B) which are closest to the word segmentation result and have the same parity with the word segmentation result in the interest point address base; s3, importing A and B into a building vector layer, and determining respective closest building surface units; s4, connecting geometric centers of the building vector layers in the building vector layers to form a line segment L; obtaining the number N of building surface units intersected with the line segment L and calculating the address distribution density r; s5, calculating a surface unit index P; and searching and obtaining an optimal building surface unit along the line segment L, and taking the geometric center coordinate of the optimal building surface unit as a final coding result. Compared with a uniform interpolation method, the method is higher in coding result accuracy.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of geographic coding result data processing, and particularly relates to a coding result optimization method of a coding platform based on address distribution density. BACKGROUND

[0002] Geocoding is a technology of converting the text description of an address into geographic coordinates. The current online geocoding platform (such as Baidu, Tencent, Gaode, Sogou) all rely on this technology to provide geographic coordinate services.

[0003] When the address text is input into the coding platform, if the platform database does not have a coding result corresponding to the address text, two coding results closest to the address text, having the same parity and one large and one small house number are selected, and the corresponding coding result is generated by average interpolation. For example, assuming that the address text is "No. 18, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province", the platform does not match the coding result of the address after querying the database, then "No. 14, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province" and "No. 20, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province" are selected as the two closest house numbers with even parity; then the interval between the two is divided into six equal parts (since No. 14 and No. 20 are six house numbers apart, each part corresponds to one house number interval); finally, No. 18, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province is estimated to be at the end of the fourth equal part by interpolation, and the coding result of this position is output.

[0004] However, this way of dividing the interval between the two closest coding results and then interpolating is not accurate. Because the building distribution in the city is uneven, for example, No. 18, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province may be closer to No. 14 and farther from No. 20, and relying on equal interpolation will cause the output coding result to deviate from the true position. SUMMARY

[0005] The application provides a coding result optimization method of a coding platform based on address distribution density, which is used to improve the accuracy of the coding result.

[0006] In order to achieve the above purpose, the application provides the following technical content: A coding result optimization method of a coding platform based on address distribution density, comprising the following steps: S1: adopting CRF, performing word segmentation processing on the text address input into the coding platform to obtain a word segmentation result; S2: according to the word segmentation result, finding two house numbers closest to the address text, having the same parity and one large and one small in the point of interest address database, and obtaining the coordinates of the two house numbers, respectively denoted as A ( X A ,Y A )and B ( X B , Y B ); S3: A ( X A , Y A )and B ( X B , Y B ) into the building vector layer, and judge A ( X A , Y A )and B ( X B , Y B ) are respectively recorded as the first building surface unit and the second building surface unit; S4: In the building vector layer, connect the geometric centers of the first building surface unit and the second building surface unit to form a line segment L ; Get the line segment L Number of intersecting building polygon units N ; Use the word segmentation results in S1 and the two house numbers obtained in S2 to calculate the address distribution density r ; S5: Based on address distribution density r Sum and line segment L Number of intersecting building polygon units N , calculate the surface unit index P ; According to the index P , along the line segment L Find and obtain the best building surface unit, and output the geometric center coordinates of the best building surface unit as the final encoding result.

[0007] Furthermore, step S3 includes the following steps: S3.1: Respectively A ( X A , Y A )and B ( X B , Y B ) as the center and R as the radius, and mark them as circle O1 and circle O2 respectively; S3.2: If the circle O1 encompasses any part or the whole of any one building surface unit, the building surface unit is included in set A; if the circle O2 encompasses any part or the whole of any one building surface unit, the building surface unit is included in set B; S3.3: Calculate the Euclidean distance of each edge midpoint of all building surface units in set A to A ( X A , Y A ) respectively, and obtain the building surface unit corresponding to the edge with the shortest distance, which is recorded as the first building surface unit; calculate the Euclidean distance of each edge midpoint of all building surface units in set B to B ( X B , Y B ) respectively, and obtain the building surface unit corresponding to the edge with the shortest distance, which is recorded as the second building surface unit.

[0008] Further, the calculation formula of the address distribution density in step S4 is: r

[0009] In the formula, Vx is the house number to be matched, i.e. the house number in the text address in step S1; V 1 is the house number with smaller value found in step S2; V 2 is the house number with larger value found in step S2.

[0010] Further, step S5 includes the following steps: S5.1: The calculation formula of the index P is:

[0011] In the formula, N represents the number of building surface units intersected by the line segment L , and represents the floor function; S5.2: Taking the first building surface unit as the starting point, traverse the building surface units in the direction of the line segment L to the second building surface unit, and when the P th building surface unit is reached, the geometric center coordinates of the building surface unit are taken as the final encoding result.

[0012] Further, in step S3.1, the value of R is 50 m.

[0013] The above technical solutions can achieve the following beneficial effects:​ The scheme is based on the unevenness of building distribution, first finds two house numbers closest to the segmentation result, same parity and one large and one small in the point of interest address library, and obtains the corresponding coordinates A X A Y A and B X B Y B , based on the building vector layer, A X A Y A and B X B Y B , to obtain the first building surface unit and the second building surface unit, and further obtain the line segment L ; the address distribution density r is proposed, and the line segment L and the number of building surface units are obtained in combination with the building vector layer, so as to calculate the index P , and finally through the index P , the final encoding result is obtained, the accuracy of the encoding result is improved compared with the original uniform interpolation. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is the flow chart of the whole method; Figure 2 is the building surface unit graph of the method. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme of the present application will be described clearly and completely below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0016] As shown in Figure 1 , a coding result optimization method of a coding platform based on address distribution density includes the following steps: S1: using CRF (Conditional Random Field), the text address of the input coding platform is segmented to obtain the segmentation result.

[0017] ​​​​​​​​Specifically, for example, the text address is: No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province, and the segmentation result is "Jiangsu Province", "Nanjing City", "Jiangning District", "Focheng West Road", and "No. 17".

[0018] S2: According to the segmentation result, find two house numbers that are closest to the address, have the same parity, and are one large and one small in the point of interest address library (POI database), and obtain the coordinates of the two house numbers, respectively denoted as A ( X A , Y A ) and B ( X B , Y B ).

[0019] Specifically, since "No. 17, Focheng West Road, Jiangning District, Nanjing City, Jiangsu Province" does not have a corresponding address in the POI database, two house numbers that are closest to the address, have the same parity, and are one large and one small are found in the POI database, which are assumed to be "No. 13, Jiangning District, Nanjing City, Jiangsu Province" and "No. 19, Jiangning District, Nanjing City, Jiangsu Province". Based on the two addresses, the POI database outputs the coordinates corresponding to the two addresses, that is, A ( X A , Y A ) and B ( X B , Y B ).

[0020] S3: Import A ( X A , Y A ) and B ( X B , Y B ) into the building vector layer, and respectively judge the building surface units closest to A ( X A , Y A ) and B ( X B , Y B ), as shown in Figure 2 .

[0021] In the building vector layer, each building face unit is a polygon, which can be a three-side polygon or a five-side polygon. The building face unit refers to an independent spatial unit formed by the actual contour of a building as a boundary in a geographic information system.

[0022] S3 specifically comprises the following steps: S3.1: respectively taking A ( X A , Y A ) and B ( X B , Y B ) as the centers and 50 m as the radius to make circles, and respectively recording the circles as circle O1 and circle O2.

[0023] S3.2: if circle O1 encompasses a part or the whole of any building face unit, the building face unit is included in set A; if circle O2 encompasses a part or the whole of any building face unit, the building face unit is included in set B.

[0024] Specifically, assuming that circle O1 encompasses the whole of building face unit B1, a part of building face unit B2, …, and a part of building face unit Bn, building face unit B1, building face unit B2, …, and building face unit Bn are included in set A.

[0025] Assuming that circle O2 encompasses the whole of building face unit C1, a part of building face unit C2, …, and a part of building face unit Cm, building face unit C1, building face unit C2, …, and building face unit Cn are included in set B.

[0026] The building face units in set A and set B have the same building face units, which does not affect the scheme.

[0027] S3.3: calculating the Euclidean distances of A ( X A , Y A ) to the midpoints of all sides of the building face units in set A, respectively, and obtaining the building face unit corresponding to the shortest side, which is recorded as the first building face unit; calculating the Euclidean distances of B ( X B , Y B) to the midpoints of each edge of all building surface units in set B, and obtain the building surface unit corresponding to the edge with the smallest distance, which is recorded as the second building surface unit; specifically comprising the following steps: S3.3.1: Assume that all building surface elements in set A have S1 edges; set j Edge( j ∈[1,S1]) is the midpoint coordinate of ( X j,mid ,Y j,mid ),but A ( X A , Y A ) to j Midpoint of edge( X j,mid ,Y j,mid ) is:

[0028] In formula (1), D j express A ( X A , Y A ) to the first j Midpoint of edge( X j,mid , Y j,mid )’s Euclidean distance.

[0029] S3.3.2: In set A, filter out A ( X A , Y A ) and obtain the corresponding building surface unit, which is recorded as the first building surface unit.

[0030] The formula is:

[0031] In formula (2), min(*) means taking the minimum value.

[0032] Assume that after calculations in steps S3.3.1 and S3.3.2, A ( X A , Y A ) to j The Euclidean distance between the midpoints of the edges is the shortest, and jThe building surface unit corresponding to the edge is building surface unit B1, and the building surface unit B1 is the first building surface unit.

[0033] S3.3.3: Assuming that all building surface units in set B have S2 edges in total; set the midpoint coordinates of the first edge ( k k ∈[1, S2]) as ( X k,mid ,Y k,mid ), then the Euclidean distance from the first edge midpoint ( B X B , Y B ) to the midpoint of the S2th edge ( k X k,mid ,Y k,mid ) is:

[0034] In formula (3), D k represents the Euclidean distance from the first edge midpoint ( B X B , Y B ) to the midpoint of the S2th edge ( k X k,mid , Y k,mid ).

[0035] S3.3.4: In set B, filter out the edge with the shortest distance to ( B X B , Y B ), and obtain the corresponding building surface unit, denoted as the second building surface unit.

[0036] The formula is:

[0037] In formula (4), min(*) represents the minimum value.

[0038] Assuming that after steps S3.3.3 and S3.3.4 are performed, B X B , Y B ) to the midpoint of the S2th edge is the shortest, and the S2th edge is the edge with the shortest distance to ( k k ​​​​​​​​The building surface unit corresponding to the edge is the building surface unit C1, and the building surface unit C1 is the second building surface unit.

[0039] when A ( X A , Y A )and B ( X B , Y B ) have the shortest distance to the same building surface unit, then the building surface unit is both the first building surface unit and the second building surface unit.

[0040] S4: In the building vector layer, connect the geometric centers of the first building surface unit and the second building surface unit to form a line segment L , get the line segment L Number of intersecting building polygon units N ; Use the word segmentation results in S1 and the two house numbers obtained in S2 to calculate the address distribution density r ; Specifically, suppose the smaller house number is set as V 1. Set the larger house number as V 2, then the address distribution density r The calculation formula is:

[0041] In formula (5), r Indicates address distribution density; Vx is the house number to be matched, that is, the house number "No. 17" in the text address "No. 17, Jiangning District, Nanjing City, Jiangsu Province" in step S1. V 1 is the house number with the smaller value found in S2; V 2 is the house number with the larger value found in S2, that is, V 1 is the "No. 13" in step S2; V 2 is "No. 19" as an example in step S2.

[0042] when A ( X A , Y A )and B ( X B , Y B ) are all at the shortest distance to the same building surface unit, then the line segment L Does not exist, the number of building surface units intersecting with line segment L NDefault is 0.

[0043] S5: According to the address distribution density r And the number of building surface units intersected by the line segment L , calculate the surface unit index N ; according to the index P , find and get the best building surface unit along the line segment P , and output the geometric center coordinates of the best building surface unit as the final encoding result. Specifically, it includes the following steps: L S5.1: Calculate the index . P

[0044] The formula is:

[0045] In formula (6), N represents the number of building surface units intersected by the line segment L . represents the floor.

[0046] S5.2: Take the first building surface unit as the starting point, and traverse the building surface units along the line segment L to the second building surface unit. When the first P building surface unit is traversed, the geometric center coordinates of the building surface unit are taken as the final encoding result.

[0047] For example, calculated by formula (6): ; it can be seen from Figure 2 that the number of intersection points of the line segment L and the building surface unit is 2, so , and thus the first building surface unit is taken as the starting point, and the first traversed building surface unit is Figure 2 the building surface unit shaded by the diagonal line, and the geometric center of the building surface unit is taken as the final encoding result.

[0048] Example: The existing encoding method mainly includes point matching interpolation method and line matching interpolation method.

[0049] Point matching interpolation method example: assuming the text address is "No. 17", then the two door numbers closest to it, both odd numbers and one large and one small are "No. 13" and "No. 19", then the line segment formed by the two endpoints "No. 13" and "No. 19" is divided into 6 equal parts, then "No. 17" is located at the 4th equal point traversed from "No. 13", and the coordinates of the position are obtained as the final encoding result output.

[0050] ​Line matching interpolation method example: assuming the text address is "No. 17", "17" corresponds to the road Z, the door number of the end point at both ends of the road Z is assumed to be "No. 1" and "No. 100", then the road is divided into 100 parts, "No. 17" is located at the 17th equal point from "No. 1", and the coordinates of the position are obtained as the final encoding result output.

[0051] Compare the encoding results of the existing encoding method and the encoding method of the present scheme. The results are shown in Table 1.

[0052] Table 1 Comparison of encoding result accuracy of various algorithms

[0053] 7250 address information data containing accurate geographic coordinates were tested, and the results are shown in Table 1. From the data in Table 1, the standard deviation and mean of the present scheme are less than those of the other two methods, which indicates that the present scheme is superior to the other two methods in accuracy. Moreover, the present scheme has a good inhibitory effect on extreme errors, such as errors greater than 500m, with a data volume ratio of only 2.50%, which is much smaller than that of the other two methods. For general errors, i.e. errors greater than 100m and 150m, the present scheme also has a larger data volume ratio compared to the other two methods, indicating that the accuracy for general errors is still relatively accurate.

[0054] Based on the above ideal embodiments according to the present application, through the above description, relevant personnel can make various changes and modifications without deviating from the technical idea of the present application. The technical scope of the present application is not limited to the contents of the specification, and must be determined according to the scope of the claims.

Claims

1. An encoding result optimization method of an encoding platform based on address distribution density, characterized in that, The method comprises the following steps: S1: adopting CRF, performing word segmentation on the text address input into the coding platform to obtain a word segmentation result; S2: According to the word segmentation result, find two house numbers in the interest point address library that are closest to the house number, have the same parity, and are one large and one small, and obtain the coordinates corresponding to the two house number pairs, respectively denoted as A ( X A , Y A ) and B ( X B , Y B ); S3: the A ( X A , Y A ) and B ( X B , Y B ) are imported into the building vector layer, and the building surface units closest to A ( X A , Y A ) and B ( X B , Y B ) are determined respectively, and are recorded as the first building surface unit and the second building surface unit respectively; S4: In the building vector layer, connect the geometric centers of the first building surface unit and the second building surface unit to form a line segment L ; Get the line segment L Number of intersecting building polygon units N ; Use the word segmentation results in S1 and the two house numbers obtained in S2 to calculate the address distribution density r ; S5: according to the address distribution density r and the number of building surface units intersecting with the line segment L N , calculate the surface unit index P ; according to the index P , find and obtain the best building surface unit along the line segment L , and output the geometric center coordinates of the best building surface unit as the final encoding result.​ 2. The method of claim 1, wherein, Step S3 comprises the following steps: S3.1: circles are made with (S3.1.1) and (S3.1.2) as the center and R as the radius, respectively denoted as circle O1 and circle O2. A X A , Y A ) and B X B , Y B ) as the center and R as the radius, respectively denoted as circle O1 and circle O2.​​ S3.2: if the circle O1 encompasses a part or the whole of any one building surface unit, the building surface unit is included in set A; if the circle O2 encompasses a part or the whole of any one building surface unit, the building surface unit is included in set B; S3.3: Calculate A ( X A , Y A ) the Euclidean distance from each edge midpoint of all building surface units in set A to the center of set B, and obtain the building surface unit corresponding to the shortest edge, denoted as the first building surface unit; calculate B ( X B , Y B ) the Euclidean distance from each edge midpoint of all building surface units in set B to the center of set A, and obtain the building surface unit corresponding to the shortest edge, denoted as the second building surface unit.

3. The method of claim 1, wherein the method is characterized by: The address distribution density in step S4 r The calculation formula is: ； wherein Vx 1 is the house number to be matched, i.e. the house number in the text address in step S1 ; V 1 is the house number to be matched, i.e. the house number in the text address in step S1 ; V 1 is the house number to be matched, i.e. the house number in the text address in step S1 ; 4. The method of claim 3, wherein the method further comprises: Step S5 comprises the following steps: S5.1: Index P The calculation formula is: ； wherein N represents a line segment L the number of intersecting building face units, represents the floor of; S5.2: Starting from the first building surface unit, traverse along the line segment L to the second building surface unit. When the traversal reaches the P building surface unit, the coordinate of the geometric center of the building surface unit is the final encoding result.

5. The method of claim 2, wherein the method further comprises: In step S3.1, the value of R is 50 m.

Images