A digital topographic map linkage update method for element matching

Abstract

The invention relates to a method for linking and updating a digital topographic map of element matching. The method comprises the following steps: abstracting various types of topographic elements in a digital topographic map into element entities according to their shapes; establishing buffer zones of various element entities respectively, and obtaining candidate matching pairs corresponding to the types of the element entities in the buffer zones of the element entities respectively; using similarity features of the candidate matching pairs of various element entities as training sample sets for machine learning, and training two matching relationship recognition models for each type of element entity, namely, model one and model two; using model one of various element entities to perform initial recognition on point, line and surface entity data respectively, and using model two of various element entities to perform linking and matching on various unmatched element entities; defining data fusion rules, and incrementally updating heterogeneous topographic map data based on linking and matching results; and realizing dynamic updating of graphics and attribute information, thereby providing technical support for establishing a topographic integrated spatiotemporal database.

Description

Technical Field

[0001] The present invention relates to the field of digital topographic map updating, and particularly to a method for linked updating of digital topographic maps with element matching. Background Art

[0002] Topographic maps visualize geographical elements such as landforms, houses, and vegetation in the form of points, lines, and surfaces, enabling people to better understand the morphological changes of the Earth's surface. They have a wide range of applications in geographical cognitive exploration, urban planning and decision-making support, and geographical information system applications. In particular, road planning, drainage system design, traffic network optimization, land use planning, etc. in municipal construction all rely on accurate and up-to-date topographic data. There is also a need to actively explore a new basic surveying and mapping production model of "unified planning, hierarchical implementation, and collaborative updating" to promote the transformation of traditional single-scale databases into entity-based and integrated spatio-temporal databases. These current situations mean that there is an urgent need to integrate, fuse, and update multi-source topographic map data with multiple scales, multiple time phases, etc. to promote the intelligent growth of cities and assist in the informatization construction of cities.

[0003] Based on the existing topographic map data, quickly and effectively dynamically updating the current terrain is one of the important issues that need to be solved for data integration. Currently, the general process of related research is to overlay the original topographic map with the latest image data, visually interpret the changed areas of the images, and then set the rules for dynamic updating of the topographic map, such as currency, accuracy, consistency, etc. Finally, digitalization or field re-surveying or re-measuring is carried out accordingly to update the topographic map in the changed areas. This method focuses on updating graphic data by combining image data with human-computer interaction, without taking into account the update requirements of attribute data. Moreover, visually identifying changed elements is obvious for local areas, but for large-scale topographic map dynamic updating, the efficiency is low, which does not meet the actual requirements of quickly and efficiently updating topographic element data. In addition, technologies for extracting vector data based on images are developing rapidly, and the sources of vector data acquisition are diverse, promoting the incremental updating and fusion requirements of multi-source vector data. Summary of the Invention

[0004] In view of the technical problems existing in the prior art, the present invention provides a method for linked updating of digital topographic maps with element matching, which realizes the dynamic updating of graphic and attribute information and provides technical support for establishing a topographic integrated spatio-temporal database.

[0005] According to a first aspect of the present invention, there is provided a method for linked updating of digital topographic maps with element matching, including: Step 1, abstracting each type of topographic element in the digital topographic map into an element entity according to its shape, where the element entity includes: point entity, line entity, and surface entity;

[0006] Step 2: Establish buffers for each of the element entities, and obtain respective candidate matching pairs corresponding to the type of each element entity in the buffer of the element entity.

[0007] Step 3: Define the separated, inclusion, connection, and intersection relationships between each of the element entities in terms of internal, boundary, and external relationships; divide the association relationships of each of the element entities into strong associations, medium associations, and weak associations.

[0008] Step 4: Use the similarity features of the candidate matching pairs of various element entities as the training sample set for machine learning. For each type of element entity, train two matching relationship recognition models, namely Model One and Model Two; among them, point entity data is used to train the point entity matching relationship recognition model, line entity data is used to train the line entity matching relationship recognition model, and surface entity data is used to train the surface entity matching relationship recognition model; the similarity features include: topological similarity, which is calculated based on the association relationships divided in Step 3.

[0009] Step 5: Use Model One of each type of element entity to perform primary recognition on point, line, and surface entity data respectively, extract the matched elements in the primary recognition results as the landmark sets in various element entities, establish the information transfer models of various types of elements, and use Model Two of various types of element entities to perform linkage matching on the remaining unmatched various element entities based on the information transfer model and with the landmark sets as constraints.

[0010] Step 6: Define data fusion rules, and perform incremental update on the heterogeneous topographic map data based on the linkage matching results.

[0011] Based on the above technical solutions, the present invention can also be improved as follows.

[0012] Optionally, Step 1 further includes: unifying the formats of the source data and target data in the digital topographic map dataset, correcting the systematic errors of the spatial information, and constructing the category, name, and attribute table information of each type of element; the attribute table information includes: common attributes and special attributes.

[0013] Optionally, the candidate matching pairs include: 1:1 matching pairs and 1:N, M:1, and M:N matching pairs caused by actual changes, map compilation, and production errors; N and M are natural numbers greater than 1.

[0014] Optionally, in step 4, the matching relationship recognition model selects OCSVM as the classifier. After creating samples and extracting similarity features, a Gaussian kernel function is selected, and the maximum F1 score of the test set is used to obtain the Gaussian kernel bandwidth g ∈ (0, 1) and the relaxation factor v ∈ (0, 1]. For the input sample, OCSVM outputs label 1 or label -1 to represent the matching relationship and non-matching relationship respectively, and outputs the decision distance value of the sample. The decision distance value is a positive number, and the closer it is to 0, the greater the matching probability.

[0015] Optionally, in step 4, the training sample set of the point entity is divided into a strict training sample set and a general training sample set according to the Euclidean distance of the candidate matching pairs of the point entity; the training sample set of the line entity is divided into a strict training sample set and a general training sample set according to the closest distance of the candidate matching pairs of the line entity; the training sample set of the surface entity is divided into a strict training sample set and a general training sample set according to the area overlap degree of the candidate matching pairs of the surface entity; the model one is obtained by training with the strict sample set, and the model two is obtained by training with the general sample set.

[0016] Optionally, in step 4,

[0017] When the feature entity is a point entity, the similarity features as the training sample data set include: category similarity, name similarity, spatial distance, and topological similarity;

[0018] When the feature entity is a line entity, the similarity features as the training sample data set include: Hausdorff distance, name similarity, category relationship similarity, and topological similarity;

[0019] When the feature entity is a surface entity, the similarity features as the training sample data set include: position similarity, direction similarity, area similarity, shape similarity, and topological similarity.

[0020] Optionally, the calculation formula of the topological similarity is:

[0021] TS = a×TP q +b×TP z +c×TP r

[0022] where a, b, and c are the weights corresponding to strong association, medium association, and weak association of the association relationship respectively; TP represents the topological strength, including: the topological strength TP of strong association q 、the topological strength TP of medium association z and the topological strength TP of weak association r ;

[0023]

[0024] n represents the number of neighboring entities, and x h represents the tag value indicating whether the neighbor matches. If the neighbor matches, then x h = 1; if the neighbor does not match, then x h = -1.

[0025] Optionally, in step 5, the landmark set is a set of landmark elements with prominent features or higher accuracy than a set threshold in the feature set.

[0026] Optionally, in step 5, the construction process of the information transfer model includes:

[0027] Step 501: Construct a constrained Delaunay triangulation of the vertices of the surface entity and extract the skeleton of the blank area based on the triangulation. The line entity coincides with or is a subset of the skeleton line, and each mesh of the skeleton line corresponds to a surface entity, representing the influence range of the surface entity, to obtain the connection between the surface entity and the line entity;

[0028] Step 502: Overlap the skeleton line meshes with the point entity set, and determine whether the point entity is inside the mesh or on the skeleton line of the mesh, to obtain the point entities within the influence range of the corresponding surface entity, to obtain the connection between the surface entity and the point entity;

[0029] Step 503: Obtain the information transfer model of the point entity, line entity, and surface entity from the topological relationships obtained in steps 501 and 502.

[0030] Optionally, the data fusion rules defined in step 6 include:

[0031] For multi-temporal data, compare the currency of heterogeneous data, and retain the elements with a more recent data production time and better currency;

[0032] For multi-scale data, retain the data with a smaller expression granularity and a larger scale;

[0033] For attribute information, with the aim of increasing the richness and integrity of the attribute information, expand the unique attributes and attribute values of each data source into the integrated data, and uniformly store the attribute values of different time versions to provide a data basis for change analysis;

[0034] The rule for retaining entities is timeliness > accuracy > integrity.

[0035] A digital topographic map linkage update method based on feature matching provided by the present invention has the following beneficial effects: (1) The topographic map features are divided into three types of entities: points, lines, and surfaces. A similarity calculation method based on self-features and adjacent features is proposed, and the association relationships and weights of different types of entities are defined. This is beneficial for mutual verification between different types of entities in the full-feature update, and improves the matching accuracy of multi-source data in complex situations; (2) The Delaunay triangulation network and the skeleton line are introduced to construct the topological relationships of point, line, and surface entities. One-to-one association relationships are established by defining rules, and information transmission is carried out to further improve the matching accuracy; (3) Based on the matching results of entities, the linkage update of different types of entities is realized by defining information fusion rules. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 FIG. is a flowchart of the implementation steps of a digital topographic map linkage update method based on feature matching provided by an embodiment of the present invention;

[0037] Figure 2 FIG. is a schematic diagram of an example for obtaining candidate matching pairs of heterogeneous topographic feature data provided by an embodiment of the present invention;

[0038] Figure 3 FIG. is a schematic diagram of the topological relationship of point, line, and surface entities provided by an embodiment of the present invention;

[0039] Figure 4 FIG. is a schematic diagram of the calculation of topological similarity of point, line, and surface features provided by an embodiment of the present invention;

[0040] Figure 5 FIG. is a schematic diagram of the linkage matching of line and surface vector features provided by an embodiment of the present invention;

[0041] Figure 6 FIG. is a schematic diagram of the linkage matching of point and surface vector features provided by an embodiment of the present invention;

[0042] Figure 7 FIG. is a schematic diagram of an example of the information transmission model of point, line, and surface entities provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The principles and features of the present invention will be described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0044] Figure 1 FIG. is a flowchart of a digital topographic map linkage update method based on feature matching provided by the present invention. As Figure 1 shown, the update method includes:

[0045] Step 1: Abstract the terrain elements of various types in the digital topographic map into element entities according to their shapes. The element entities include: point entities, line entities, and surface entities.

[0046] Step 2: For heterogeneous terrain element data, establish buffers for each element entity respectively, and obtain each candidate matching pair corresponding to the type of the element entity in the buffer of the element entity.

[0047] Step 3: Define the disjoint, inclusion, connection, and intersection relationships between each element entity in terms of internal, boundary, and external relationships; divide the association relationships of each element entity into strong associations, medium associations, and weak associations.

[0048] Step 4: Use the similarity features of the candidate matching pairs of various element entities as the training sample set for machine learning. Two matching relationship recognition models, namely Model 1 and Model 2, are trained for each type of element entity; among them, point entity data is used to train the point entity matching relationship recognition model, line entity data is used to train the line entity matching relationship recognition model, and surface entity data is used to train the surface entity matching relationship recognition model; the similarity features include: topological similarity, and the topological similarity is calculated based on the association relationships divided in Step 3.

[0049] Step 5: Use Model 1 of each type of element entity to perform initial recognition on point, line, and surface entity data respectively, extract the matched elements in the initial recognition results as the landmark sets in various element entities, establish the information transfer models of various types of elements, and use the landmark sets as constraints according to the information transfer models, and use Model 2 of each type of element entity to perform linkage matching on the remaining unmatched various element entities.

[0050] Step 6: Define the data fusion rules, and perform incremental update on the heterogeneous topographic map data based on the linkage matching results.

[0051] A method for linkage update of a digital topographic map with element matching provided by the present invention realizes the dynamic update of graphic and attribute information, and provides technical support for establishing a terrain integrated spatio-temporal database.

[0052] Embodiment 1

[0053] Embodiment 1 provided by the present invention is an embodiment of a method for linkage update of a digital topographic map with element matching provided by the present invention. As can be seen from Figure 1 it, the embodiment of this update method includes:

[0054] Step 1: Abstract the terrain elements of various types in the digital topographic map into element entities according to their shapes. The element entities include: point entities, line entities, and surface entities.

[0055] In a possible embodiment, step 1 further includes: unifying the formats of the source data and target data in the digital topographic map dataset, correcting the systematic errors of the spatial information, and constructing the category, name, and attribute table information of various types of elements; the attribute table information includes: common attributes and special attributes.

[0056] In specific implementation, according to the shapes of various types of elements, point elements such as electric poles and lighthouses in the topographic map can be abstracted as point entities P = {p1, p2, … p m}, linear elements such as roads and subways can be abstracted as line entities L = {l1, l2, … l n}, and planar elements such as houses and lakes can be abstracted as surface entities A = {a1, a, … a r}.

[0057] Preprocess the digital topographic map dataset, including unifying the formats of the source data and target data. In addition to correcting the systematic errors of the spatial information, it is also necessary to construct its category, name, and attribute table information. The common attribute table structure of point entities is composed of {code, name, coordinate point, category, datum, source, update date}, the common attribute table structure of line entities is composed of {code, name, coordinate string, category, datum, source, update date}, and the common attribute table structure of surface entities is composed of {code, name, coordinate string, category, datum, source, update date}. In addition to the common attributes, some entities also have special attributes, such as roads having "number of parking spaces" and buildings having "material", etc.

[0058] Step 2, for heterogeneous topographic element data, establish buffers for each element entity respectively, and obtain each candidate matching pair corresponding to the type of the element entity in the buffer of the element entity.

[0059] In a possible embodiment, the candidate matching pairs include: 1:1 matching pairs and 1:N, M:1, and M:N matching pairs caused by actual changes, map generalization, production errors, etc.; N and M are natural numbers greater than 1.

[0060] In specific implementation, the buffer method in step 2 is used to obtain candidate matching pairs, and only entities of the same type are obtained as candidate matching pairs. For example, for point entities, other line and surface entity types covered by the buffer are not within the candidate matching range, but other types of entities are used as neighboring information to support the matching process later. The buffer distance of point entities is 40m, the buffer distance of line entities is 20m, and the buffer distance of surface entities is 40m. As Figure 2 (b) shows, if the elements in the point buffer area range are (p2, l2, a2), then the candidate element is only p2; if the elements in the line buffer area range are (p3, p4, l2), then the candidate element is only l2; if the elements in the surface buffer area range are (p2, l2, a2, a3), then the candidate element is only p2.

[0061] Step 3: Define the disjoint, inclusion, connection, and intersection relationships between each element entity in terms of internal, boundary, and external relationships; classify the association relationships of each element entity into strong association, medium association, and weak association.

[0062] Since point, line, and surface entities do not exist in isolation in the real world and there is more or less a certain association between them, in the embodiments of the present invention, the nine-intersection model of plane geometry is used to describe the topological relationship of geometric figures, and the topological relationship as shown in Figure 3 is defined in terms of internal, boundary, and external relationships. In addition to the disjoint relationship between point, line, and surface entities, it also includes the inclusion, connection, and intersection relationships in the figure; classify the association relationships between entities into strong association, medium association, and weak association as the weight factors for topological similarity calculation in Step 4.

[0063] In specific implementation, the basis for the definition of the association relationship is the specification of the "Dictionary of Fundamental Geographic Information Elements". Composite elements, such as "seasonal river" composed of a center line, a directed line, and a range line, "karez" composed of a positioning point and a directed line, "island" composed of a positioning point and a range surface, etc., the relationship between point, line, and surface entities with an association relationship is defined as a strong association, and a weight of 0.5 is assigned; the relationship between entities with overlapping, connecting, and inclusion relationships is defined as a medium association. For example, Figure 3 the examples shown of a point on a line, a point within a surface, etc. In the specification, it is stipulated that there is a connection relationship between underground pipelines and underground pipeline entrances and exits, between rivers and reservoirs, an overlapping relationship between roads and boundaries, an inclusion relationship between underground canals and underground canal outlets, between buildings and pillars, etc., and a weight of 0.3 is assigned; other adjacent relationships are assigned a weight of 0.2.

[0064] Step 4: Use the similarity features of candidate matching pairs of various element entities as the training sample set for machine learning. For each type of element entity, two matching relationship recognition models are trained, namely Model 1 and Model 2; among them, point entity data is used to train the point entity matching relationship recognition model, line entity data is used to train the line entity matching relationship recognition model, and surface entity data is used to train the surface entity matching relationship recognition model; the similarity features include: topological similarity, and the calculation of topological similarity is obtained based on the association relationship divided in Step 3.

[0065] In a possible embodiment, in step 4, the matching relationship recognition model selects OCSVM as the classifier. After creating samples and extracting similarity features, a Gaussian kernel function is selected, and the maximum F1 score of the test set is used to obtain the Gaussian kernel bandwidth g ∈ (0, 1) and the relaxation factor v ∈ (0, 1]. The classifier parameters g and v are set, that is, an OCSVM classification model is constructed to obtain the matching relationship recognition model. For the input sample, the OCSVM outputs label 1 or label -1, representing the matching relationship and the non-matching relationship respectively, and outputs the decision distance value of the sample. The decision distance value is a positive number, and the closer it is to 0, the greater the matching probability.

[0066] In specific implementation, when constructing the training sample data, candidate matching pairs with matching relationships can be subjectively selected according to expert experience.

[0067] In a possible embodiment, in step 4, according to the Euclidean distance of the candidate matching pairs of point entities, the training sample set of point entities is divided into a strict training sample set and a general training sample set; according to the nearest distance of the candidate matching pairs of line entities, the training sample set of line entities is divided into a strict training sample set and a general training sample set; according to the area overlap degree of the candidate matching pairs of surface entities, the training sample set of surface entities is divided into a strict training sample set and a general training sample set; the strict sample set is used for training to obtain model one, and the general sample set is used for training to obtain model two.

[0068] In specific implementation, the matching relationship recognition model is obtained by training with the training sample set. Specifically, a strict training sample set (Euclidean distance < 15m) and a general training sample set (no distance limit) are selected from point entities, a strict training sample set (nearest distance < 20m) and a general training sample set (no distance limit) are selected from line entities, and a strict training sample set (area overlap degree > 0.5) and a general training sample set (no overlap degree limit) are selected from surface entities. Model one is obtained by training the model with the strict sample set, and model two is obtained by training the model with the general sample set. A total of 6 matching relationship recognition models are obtained for point, line, and surface entities.

[0069] In a possible embodiment, when the feature entity is a point entity, the similarity features as the training sample data set include: category similarity, name similarity, spatial distance, and topological similarity.

[0070] When the feature entity is a line entity, the similarity features as the training sample data set include: Hausdorff distance, name similarity, category relationship similarity, and topological similarity.

[0071] When the feature entity is a surface entity, the similarity features as the training sample data set include: position similarity, direction similarity, area similarity, shape similarity, and topological similarity.

[0072] In specific implementation, the similar features extracted in step 4 are used as training samples. According to the principles of effectiveness, reliability, integrity, and sufficiency of the features, category similarity, name similarity, spatial distance, and topological similarity are selected as the training samples for point entities, Hausdorff distance, name similarity, category relationship similarity, and topological similarity are selected as the training samples for line entities, and position similarity, direction similarity, area similarity, shape similarity, and topological similarity are selected as the training samples for surface entities.

[0073] In the specific implementation process:

[0074] (1) For point entities, the buffer is used to obtain the candidate matching pairs of point entities, and the category similarity, name similarity, spatial distance, and topological similarity are calculated.

[0075] Suppose the point entities to be matched in two heterogeneous point entity datasets P1 and P2 are p i and p j respectively. Then the calculation formulas for the four similarity features of point entities are as follows:

[0076] (1) Category similarity

[0077] According to the classification and coding of basic geographic information elements, the categories are divided into three levels: major category C1, middle category C2, and minor category C3. Suppose p i and p j belong to categories t1 and t2 respectively. If t1 ∈ C3 & t2 ∈ C3, the category distance is 1; if t1 ∈ C2 & t2 ∈ C2 but then the category distance is 2; if t1 ∈ C1 & t2 ∈ C1 but then the category distance is 3. The calculation formula for the category similarity index can be obtained as follows:

[0078]

[0079] In the formula, D ij represents the distance between point entity categories; S typ (i, j) represents the category similarity; σ is the maximum distance between point entity categories. In the embodiments of the present invention, the maximum distance between categories is 3.

[0080] (2) Name similarity

[0081] The Levenshtein distance algorithm is used to calculate the number of edits (insertions, deletions, or replacements) between strings. Its core idea is to calculate the cost of single-character edits (insertions, deletions, replacements) required to change one string into another. The smaller the conversion cost, the higher the similarity between the two names. The calculation formula is as follows:

[0082]

[0083] Among them, S and T respectively represent p i and p j 's name strings. Lev S，T (i, j) refers to the distance between the first i characters in S and the first j characters in T. After obtaining the minimum edit distance required to convert from one point entity name to another, we need to normalize this value so that the final result is a value in the range [0, 1]. The smaller the result, the lower the string-based name similarity between the two point entities. The formula for calculating the name similarity index is as follows:

[0084]

[0085] Among them, max(S i , L j ) represents the maximum number of characters in strings S and T, and S Lev (i, j) is the name similarity of the point entity pair (p i , p j ).

[0086] (3) Spatial distance

[0087] The longitude and latitude information is one of the important characteristic attributes of point entity data and is a basic characteristic for calculating the similarity of point positions. In the embodiments of the present invention, the geometric distance threshold method is selected to calculate the spatial similarity of spatial point pairs: First, calculate the Euclidean distance d(i, j) between two points p i (x i , y i ) and p j (x j , y j ) in the projected coordinate system, and then, considering the data distribution state of the data set, set a suitable geometric distance threshold d, and compare the obtained distance value with the set geometric distance threshold d to calculate the spatial similarity S dis , whose value range is [0, 1], and the larger the value, the more similar. The calculation formula is as follows:

[0088]

[0089] (4) Topological similarity

[0090] Let the two point entities to be matched be p1 and q1 respectively. The candidate matching set of p1 is C1 = {C P1 = (q1, q2,... q i ); L P1 = (m1, m2,... m j ); A P1 = (b1, b2,... b k)}, the candidate matching set of q1 is C2 = {C q1 =(p1, p2,... p p ) ; L q1 =(l1, l2,... l q ) ; A q1 =(a1, a,... a s )}.

[0091] The topological feature can be described as the support degree of the topological proximity between the entity p1 to be matched and q1 to the already matched entity pairs. It is manifested that the number of the topological proximity of the point entity p1 to be matched to the already matched neighbors is the same as that of the corresponding point entity q1. And different types and different topological relationships also have different support degrees for the entity pairs to be matched. Figure 4 The ways of obtaining topological relationships are listed, showing some examples of points, lines, and planes supporting the entities to be matched.

[0092] Here, taking the matching of the point entity p1 in Figure 4 (a) as an example, the adjacency relationship between the point p1 and other points is obtained through the Delaunay triangulation. The neighboring point entities in the buffer area are used as neighboring supports, such as p7; the first line entity intersecting with it in each direction in the buffer area is found as a neighboring support, such as l1; the surface entity is abstractly expressed as the centroid, and a Delaunay triangulation is jointly constructed with the point entity. The neighboring surface entities in the buffer area are used as neighboring supports, such as a1. The specific calculation process of the topological similarity between p1 and q1 is as follows:

[0093] First, according to the definition of the association relationship in step 3, calculate the topological strengths of strong association, medium association, and weak association of p1 and q1 respectively. The topological strength is used to describe the relationship between the entity and its topologically adjacent entities, and is calculated by the proportion of the neighboring matches in the total number of neighbors. The calculation formula is as follows:

[0094]

[0095] where n represents the number of neighboring entities, and x h represents the label value indicating whether it is a match. If the neighboring entity is a match, then x h = 1; if the neighboring entity is not a match, then x h = -1. The topological strength of strong association is expressed as TP q , the topological strength of medium association is expressed as TP z , and the topological strength of weak association is expressed as TP r .

[0096] Then, the topological strength is weighted to obtain the topological similarity. The calculation formula is as follows:

[0097] TS = 0.5×TP q + 0.3×TPz +0.2×T P r

[0098] (2) For line entities, candidate matching pairs of line entities are obtained using buffers, and the Hausdorff distance, name similarity, category similarity, and topological similarity are calculated. Among them, the name similarity, category similarity, and topological similarity are calculated in the same way as for point entities, and the calculation process of the Hausdorff distance is as follows:

[0099] Let the line entities to be matched in two heterogeneous line entity datasets L1 and L2 be l i and l j , the sets of break points of l i and l j are P = {p1, p2,... p i} and Q = {q1, q2,... q i} respectively, then the Hausdorff distance of the line entity can be defined as:

[0100] d H (P, Q) = max(d h (P, Q), d h (P, Q))

[0101] In the formula, d h (p, q) = max p∈P min q∈Q ||p - q||, represents the maximum value of the set of minimum distances from any point p in point set P to any point q in point set Q, |||| represents the Euclidean distance between two points; d h (q, p) = max q∈Q min p∈P ||q - p||, similarly. For line entity matching, the arc segments can be first decomposed into discrete point sets composed of break points, and then the Hausdorff distance is used to calculate their similarity.

[0102] In order to be consistent with other similarity value ranges, the Hausdorff distance is normalized here, and the formula is as follows:

[0103]

[0104] In the formula, d(i, j) is a threshold set artificially according to the distribution characteristics of line entities.

[0105] (3) For face entities, candidate matching pairs of face entities are obtained using buffers, and position similarity, direction similarity, area similarity, shape similarity, and topological similarity are calculated. The topological similarity is calculated in the same way as for point entities. The calculation of the remaining similarities is as follows:

[0106] (1) The calculation formula for the location similarity index is as follows:

[0107]

[0108] Among them, (x1, y1) and (x2, y2) are the centroid coordinates of two surface elements a and b to be matched, and U is the maximum value of the distances of the boundary point sets of the polygons a and b to be matched; the measurement value range is [0, 1], and the larger the value, the more similar it means.

[0109] (2) The calculation formula for the direction similarity index is as follows:

[0110] sim dir (a, b) = |cos(|θ b -θ g |)|

[0111] Among them, θ (·) is the long side direction of the minimum circumscribed rectangle of the surface element, and the measurement value range is [0, 1]. The larger the value, the more similar it means.

[0112] (3) The calculation formula for the area similarity index is as follows:

[0113]

[0114] Among them, Area(·) is the area of the surface element, a ∩ b is the intersection area of the surface elements a and b, and min is the minimum value function; the measurement value range is [0, 1], and the larger the value, the more similar it means.

[0115] (4) The calculation formula for the shape similarity index is as follows:

[0116]

[0117] Among them, e (·) (·) is the cumulative value of the turning angles of the surface element, max is the maximum value function, and the measurement value range is [0, 1]. The larger the value, the more similar it means.

[0118] Step 5: Use Model 1 of various element entities to initially identify point, line, and surface entity data respectively, extract the matched elements in the initial identification results as the landmark sets in various element entities, establish the information transfer models of various elements, and based on the information transfer models and using the landmark sets as constraints, use Model 2 of various element entities to perform linkage matching on the remaining unmatched various element entities.

[0119] In a possible implementation manner, in Step 5,

[0120] The landmark set is a set of landmark elements with prominent features or higher accuracy than the set threshold in the element set.

[0121] In specific implementation, the landmark set refers to a set of landmark elements with prominent features or high precision in the terrain element set. It accounts for a relatively small proportion in the entire terrain element set but has obvious features and is easily visually interpreted. The result of the initial recognition is used as known adjacent information to calculate topological similarity, providing information support for subsequent matching.

[0122] In a possible embodiment, the information transfer model enhances the richness and accuracy of entity information by establishing connections between various elements and mutually transferring and enhancing spatial information, attribute information, etc. The specific implementation idea is as follows:

[0123] Step 501: Construct a constrained Delaunay triangulation of the vertices of the face entity and extract the skeleton of the blank area based on the triangulation. The line entity coincides with or is a subset of the skeleton line. Each mesh of the skeleton line corresponds to a face entity, representing the influence range of the face entity, and the connection between the face entity and the line entity is obtained.

[0124] Step 502: Overlap the mesh of the skeleton line with the point entity set, and determine whether the point entity is inside the mesh or on the skeleton line of the mesh, obtaining the point entity within the influence range of the corresponding face entity, and obtaining the connection between the face entity and the point entity.

[0125] Step 503: Obtain the information transfer model of the point entity, line entity, and face entity from the topological relationships obtained in Step 501 and Step 502.

[0126] As Figure 5 shown in the line-face transfer model, the light gray triangulation is the constrained Delaunay triangulation constructed by encrypting the vertex set of the face entity set A, and the triangle skeleton line L is extracted according to the following rules: It is divided into three categories according to the number of valid adjacent triangles: When there is only one side with an adjacent triangle, the connection line between the midpoint of this side and its opposite vertex is the skeleton line (Type I triangle); when there are two sides with adjacent triangles, the connection line between the midpoints of these two sides is the skeleton line (Type II triangle); when all three sides have adjacent triangles, the connection lines between the center and the midpoints of the three sides are the skeleton lines (Type III triangle); the skeleton line mesh G surrounding the face entity is obtained. Since the line entity coincides with or is a subset of the skeleton line, the information transfer model between the line entity and the face entity is established.

[0127] The topological relationship between the line entity and the face entity established based on the skeleton grid establishes an association relationship between the originally independent line entity and face entity. For example Figure 7 (a) shown First, determine the matching pair with the highest matching probability, and then determine the matching situation of related entities with low matching probability according to the information transfer model. For example, if there is an association between a face entity pair (a1, b1) and a line entity pair (l1, m1), after being recognized by the OCSVM discriminator, (1, m) is matched while (a1, b1) is not. The matching result of the skeleton grid associated with it can be obtained from the topological relationship transfer model, and then it can be obtained that the face entity pair (a1, b1) uniquely contained in the skeleton grid is also matched. The principle of judging the matching relationship of line entities based on face entities is the same as above.

[0128] The topological relationship between point entities and face entities established based on the Delaunay triangulation network establishes an association relationship between originally independent point entities and face entities. As shown in the point-face transfer models in Figure 7 (b)(c), p i is a point entity, l i is a skeleton line, a i is a face entity abstracted as a point. The closed surface formed by the skeleton lines is a mesh. There is a unique face entity and several point entities inside the mesh. The face entity and the point entity have a unique proximity relationship constructed by the Delaunay triangulation network. Taking (a1, b1) and (p1, q1) as examples, a1 and p1 are first-order proximities, and b1 and q1 are also first-order proximities and are included in the same mesh. If after being recognized by the OCSVM discriminator, (p1, q1) is matched while (a1, b1) is not, then select the face entity with the highest matching probability with b1 from the set of first-order proximity face entities {a2, a5, a6, a7} of p1 or select the face entity with the highest matching probability with a1 from the set of first-order proximity face entities {b1, b5, b6, b7} of q1 as the matching result after the point-face entity information transfer. The principle of judging the matching relationship of point entities based on face entities is the same as above.

[0129] Step 6, define the data fusion rule, and perform incremental update on the heterogeneous topographic map data based on the linked matching result.

[0130] In a possible implementation manner, the data fusion rule in step 6 is to analyze the information redundancy, complementarity, and currency of different data sets, fully consider the position accuracy, attribute accuracy, integrity, logical consistency, and time accuracy characteristics of vector elements, and update the graphic information and attribute information with higher feature accuracy and usage value into the new data. The defined data fusion rules include:

[0131] For multi-temporal data, compare the currency of heterogeneous data and retain the elements with a recent data production time and better currency.

[0132] For multi-scale data, retain the data with a small expression granularity and a large scale, and the data has high precision.

[0133] For attribute information, in order to increase the richness and completeness of attribute information, the attributes and attribute values ​​unique to each data source are expanded into the integrated data, and the attribute values ​​of different time versions are stored uniformly to provide a data basis for change analysis.

[0134] For the above fusion strategies, the rule for retaining entity determination is timeliness > accuracy > completeness.

[0135] Specifically, the information is divided into updateable information and non-updateable information. For example, the number of floors and materials of a building are updateable information, and the ID number that has no practical meaning is non-updateable information. Subsequent fusion operations are only for updateable information. For the zero-to-one matching relationship of point, line and surface entities, the newly added data is retained. For the one-to-zero matching relationship, the new data is saved to the final data. For the one-to-one matching relationship of point entities, the qualitative and quantitative attributes are directly assigned.

[0136] There are four matching situations for line entities, namely one-to-one, one-to-many, many-to-one, and many-to-many. For one-to-one, values ​​are assigned directly; for one-to-many, qualitative type attributes are assigned directly, such as the "number of parking spaces" and other quantitative type attributes of the road line entity, which are allocated according to the length ratio; for many-to-one, qualitative type attributes are assigned directly, such as the "number of parking spaces" and other quantitative type attributes of the road line entity, which are assigned cumulatively; for many-to-many, the source data are first merged into a single entity, the attributes are merged, and then the target data is assigned according to the "one-to-many" rule.

[0137] There are four matching situations for face entities, namely one-to-one, one-to-many, many-to-one, and many-to-many. For one-to-one, values ​​are assigned directly; for one-to-many, qualitative type attributes (building material, category, etc.) and non-accumulative quantitative type attributes (number of floors, height, etc. of buildings) are assigned directly, and accumulative quantitative type attributes (building area, population, etc.) are allocated according to area ratio; for many-to-one, the qualitative type attributes with the largest area ratio are assigned to the new data, non-accumulative type attributes are assigned according to the maximum value, and accumulative quantitative type attributes are assigned cumulatively; for many-to-many, the source data are first merged into a single entity, the attributes are merged, and then the target data is assigned according to the "one-to-many" rule.

[0138] After the above steps, the 1:1, 1:N, M:1, M:N and empty matching relationships of point, line and surface entities in heterogeneous topographic map data can be identified, and the linkage update of graphic data and attribute data can be realized based on the matching results and information fusion rules.

[0139] A digital topographic map linkage update method with element matching provided by the present invention has the following beneficial effects: (1) The topographic map elements are divided into three types of entities: points, lines, and surfaces. A similarity calculation method based on self - characteristics and adjacent characteristics is proposed, and the association relationships and weights of different types of entities are defined. This is beneficial for mutual verification between different types of entities in the full - element update, and improves the matching accuracy of multi - source data in complex situations; (2) The Delaunay triangulation network and the skeleton line are introduced to construct the topological relationships of point, line, and surface entities. One - to - one association relationships are established by defining rules, and information is transmitted to further improve the matching accuracy; (3) Based on the matching results of the entities, the linkage update of different types of entities is realized by defining information fusion rules.

[0140] It should be noted that in the above - mentioned embodiments, the descriptions of each embodiment have their own focuses. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0141] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer - usable storage media (including but not limited to disk storage, CD - ROM, optical storage, etc.) containing computer - usable program code.

[0142] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general - purpose computer, a special - purpose computer, an embedded computer, or other programmable data - processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data - processing devices generate means for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0143] These computer program instructions can also be stored in a computer - readable memory that can direct a computer or other programmable data - processing device to work in a specific manner, so that the instructions stored in the computer - readable memory generate a manufactured article including instruction means, and the instruction means realizes the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0144] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, causing a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 or steps for implementing the functions specified in one block or a plurality of blocks.

[0145] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept. Therefore, the appended claims are intended to be construed to cover the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0146] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A digital topographic map linkage update method for element matching, characterized in that, The update method includes: Step 1: Abstract the terrain elements of various types in the digital topographic map into element entities according to their shapes. The element entities include: point entities, line entities, and surface entities. Step 2: Establish buffers for each of the element entities respectively, and obtain each candidate matching pair corresponding to the type of the element entity in the buffer of the element entity. Step 3: Define the separated, included, connected, and intersecting relationships between each of the element entities in terms of internal, boundary, and external relationships; divide the association relationships of each of the element entities into strong associations, medium associations, and weak associations. Step 4: Use the similarity features of the candidate matching pairs of various element entities as the training sample set for machine learning. Two matching relationship recognition models, namely Model One and Model Two, are trained for each type of element entity. Among them, point entity data is used to train the point entity matching relationship recognition model, line entity data is used to train the line entity matching relationship recognition model, and surface entity data is used to train the surface entity matching relationship recognition model; the similarity features include: topological similarity, which is calculated based on the association relationships divided in Step 3. Step 5: Use Model One of each type of element entity to perform primary recognition on point, line, and surface entity data respectively, extract the matched elements in the primary recognition results as the landmark sets in various element entities, establish the information transfer models of various types of elements, and use Model Two of various types of element entities to perform linkage matching on the remaining unmatched various element entities based on the information transfer model and with the landmark sets as constraints. Step 6: Define the data fusion rule, and perform incremental update on the heterologous topographic map data based on the linkage matching results. In the matching relationship recognition model described in step 4, the OCSVM is selected as the classifier. After creating samples and extracting similarity features, the Gaussian kernel function is selected, and the Gaussian kernel bandwidth is obtained using the maximized F1 score of the test set. and the relaxation factor ; for the input sample, the OCSVM outputs label 1 or label -1, indicating a matching relationship and a non-matching relationship respectively, and outputs the decision distance value of the sample. The decision distance value is a positive number, and the closer it is to 0, the greater the matching probability. In Step 4, divide the training sample set of the point entity into a strict training sample set and a general training sample set according to the Euclidean distance of the candidate matching pairs of the point entity; divide the training sample set of the line entity into a strict training sample set and a general training sample set according to the nearest distance of the candidate matching pairs of the line entity; divide the training sample set of the surface entity into a strict training sample set and a general training sample set according to the area overlap degree of the candidate matching pairs of the surface entity; use the strict training sample set for training to obtain Model One, and use the general training sample set for training to obtain Model Two. The calculation formula for the topological similarity is: Among them, a, b, and c are the corresponding weights for strong association, medium association, and weak association of the association relationship, respectively. Indicates the topological strength, including: the topological strength of strong association , the topological strength of medium association and the topological strength of weak association ; n represents the number of neighboring entities, represents the tag value indicating whether the neighbor matches. If the neighbor matches, if the neighbor does not match, .

2. The update method according to claim 1, characterized in that, Step 1 further includes: unifying the formats of the source data and target data in the digital topographic map dataset, correcting the systematic errors of the spatial information, and constructing the category, name, and attribute table information of each type of element; the attribute table information includes: common attributes and special attributes.

3. The update method according to claim 1, characterized in that, The candidate matching pairs include: 1:1 matching pairs and 1:N, M:1, and M:N matching pairs caused by actual changes, map compilation, and production errors; N and M are natural numbers greater than 1.

4. The update method according to claim 1, characterized in that, In Step 4, When the element entity is a point entity, the similarity features as the training sample data set include: category similarity, name similarity, spatial distance, and topological similarity. When the feature entity is a line entity, the similarity features as the training sample data set include: Hausdorff distance, name similarity, category relationship similarity, and topological similarity; When the feature entity is a surface entity, the similarity features as the training sample data set include: position similarity, direction similarity, area similarity, shape similarity, and topological similarity.

5. The update method according to claim 1, characterized in that, In step 5, the landmark set is a set of landmark features in the feature set that are prominent or have a precision higher than a set threshold.

6. The update method according to claim 1, characterized in that, In step 5, the construction process of the information transfer model includes: Step 501: Construct a constrained Delaunay triangulation of the vertices of the surface entity and extract the skeleton of the blank area based on the triangulation. The line entity coincides with or is a subset of the skeleton line, and each mesh of the skeleton line corresponds to a surface entity, representing the influence range of the surface entity, to obtain the connection between the surface entity and the line entity; Step 502: Overlap the mesh of the skeleton line with the point entity set, and determine whether the point entity is inside the mesh or on the skeleton line of the mesh, to obtain the point entity within the influence range of the corresponding surface entity, to obtain the connection between the surface entity and the point entity; Step 503: Obtain the information transfer model of the point entity, line entity, and surface entity from the topological relationships obtained in steps 501 and 502.

7. The update method according to claim 1, characterized in that, The data fusion rules defined in step 6 include: For multi-temporal data, compare the currency of the heterogeneous data and retain the features with a more recent data production time and better currency; For multi-scale data, retain the data with a smaller expression granularity and a larger scale; For attribute information, with the aim of increasing the richness and integrity of the attribute information, expand the unique attributes and attribute values of each data source into the integrated data, and uniformly store the attribute values of different time versions to provide a data basis for change analysis; The retention entity determination rule is timeliness > accuracy > integrity.

Images