CN109145171A

CN109145171A - A kind of multiple dimensioned map data updating method

Info

Publication number: CN109145171A
Application number: CN201810810684.XA
Authority: CN
Inventors: 陈利燕; 何华贵; 林鸿; 肖炜枝; 张鹏程; 陈飞; 张百灵
Original assignee: Guangzhou Urban Planning Survey and Design Institute
Current assignee: Guangzhou Urban Planning Survey and Design Institute
Priority date: 2018-07-23
Filing date: 2018-07-23
Publication date: 2019-01-04
Anticipated expiration: 2038-07-23
Also published as: CN109145171B

Abstract

The invention discloses a method for updating multi-scale map data, which includes the following steps: performing multi-scale target matching on new large-scale data and small-scale data to be updated; constructing element-level association relationships; detecting element change information; Comprehensive redefinition; object-oriented incremental updates; spatial inconsistency detection and processing. The multi-scale map data update method disclosed by the invention can effectively solve the problems that the prior art cannot accurately identify the change information and predict the update and transfer scale of the change elements, the matching range of the elements is small, and the efficiency of updating the map elements is low.

Description

A kind of multiple dimensioned map data updating method

Technical field

The present invention relates to multiple dimensioned map rejuvenation technical field more particularly to a kind of multiple dimensioned map data updating methods.

Background technique

For national Major Strategic project and wisdom cities such as the monitoring of geographical national conditions, National land space optimization, ecological red line protections City's construction, it is reliable, applicable and timely geographical spatial data has great importance.The Up-to-date state of geo-spatial data is to measure One of important logo of its application value directly restricts its use value and use scope.Currently, all ground level in the whole nation with It lists and a county-level city has all carried out the construction of digital city more than 400, wherein digital city construction more than ground level is complete It completes in face.The base surveying focus of work is produced by data and turns to data maintenance, especially to existing multi-scale map vector The Up-to-date state of database updates and consistency maintenance.

Scale and Up-to-date state are the most basic features of map datum, and the map datum of different scale can express earth sky Between phenomenon or entity different levels form, structure and details；The data of different tenses can then express terrestrial space phenomenon Or entity is with time-varying process, trend and rule.Multiple dimensioned expression of the various scale maps as real world, holds Different social applications is carried on a shoulder pole, is gradually being promoted and is being goed deep into every profession and trade application.Therefore, one with real world is kept Cause property, realizing that multiple dimensioned electronic map quickly updates is the core work content of modern base surveying.

It is existing that matched method is carried out to multiple dimensioned map are as follows:

(1) map trunk road network to be matched is chosen, map space to be matched is divided into several network meshes, and match not Network meshes between ruler in proportion；(2) white space skeleton line is constructed under the network meshes after matching, by network meshes subdivision At group and white space skeleton line mesh；(3) sky of each network meshes of small scale map is completed by same steps Between subdivision；(4) the white space skeleton line mesh of large-scale map is constructed according to the above method, and is extracted in each network meshes Settlement place group；(5) settlement place group, white space skeleton line mesh and the settlement place in different scale map are matched step by step Entity obtains the map architecture option using white space skeleton line mesh between the different scale map of unit.

The present inventor has found in the practice of the invention, and following technical problem exists in the prior art:

It can only be matched by image information to map element；If map elements generate variation, to small scale map When being updated, needs all elements in map to be matched again, lead to matching efficiency when having map elements update It is lower；It is unable to get forecasting model database, the matching of variation element when to map rejuvenation next time is predicted；Using trunk roads Net building skeleton line reduces element matching range since the area coverage of road is smaller to a certain extent.

Summary of the invention

The embodiment of the present invention provides a kind of multiple dimensioned map data updating method, and it is unpredictable to can effectively solve the prior art Change element, element matching range is smaller, the problem for having matching efficiency when map elements update lower.

One embodiment of the invention provides a kind of multiple dimensioned map data updating method, includes the following steps:

Multiscale target matching is carried out to new large scale data and small scale data to be updated；

Establishing element grade incidence relation；

Carry out factor change infomation detection；

Cartographic generaliztion is carried out to redefine；

Carry out the incremental update of object-oriented；

Carry out the detection of space inconsistency and processing.

As an improvement of the above scheme, described more to new large scale data and small scale data progress to be updated The matched method of scaled target is as follows:

When matching target is dotted entity, matched using semantic similarity and Euclidean distance similarity；

When matching target is Linear Entity, matched by calculating Hausdorff distance and node；

When matching target is planar entity, matched by position adjacent degree or overlapping similarity.

As an improvement of the above scheme, carry out that matched the specific method is as follows using semantic similarity and Distance conformability degree:

Proper noun is extracted from reference data place name, address and POI data library, obtains proprietary vocabulary and near synonym group；

Proprietary vocabulary and near synonym group are added to dictionary；

Dictionary is imported into segmenter, obtains word segmentation result；

By word segmentation result creation inverted entry index, and save to index document；

Location information is extracted from reference data place name, address, POI data library and other data containing location information, is obtained To position character information；

Position character information is imported into segmenter, obtains word segmentation result；

Trie tree is established according to word segmentation result；

Search result is obtained in index document by the search index in trie tree；

According to the judgment formula (1) of following semantic similarity, is sorted, matched to search result by semantic similarity As a result:

In formula, sim (w_k, S_i) | i ∈ (1,2,3 ... n) indicates query term search term S_iWith the W of reference data document d_k's Semantic similarity, max { sim (w_k, S_i) | i ∈ (1,2,3 ... n) } indicate and W_kThe maximum value of semantic similarity.

As an improvement of the above scheme, the specific method of the establishing element grade incidence relation is using alternative manner, tool Body construction step is as follows:

The target collection for enabling new large scale data include is OL={ OL1, OL2 ..., OLm }, small scale to be updated Footage according to comprising target collection be OS={ OS1, OS2 ..., OSn }, set TL and set TS are defined as storing respectively From the interim set of set OL and the target of set OS, and initialize TL=Φ, TS=Φ；

Step 1: if meeting OS ≠ Φ, taking out target in OS and store to TS, and execute step 4；If being unsatisfactory for OS ≠ Φ, hold Row step 2；

Step 2: if meeting OL ≠ Φ, executing step 3；If being unsatisfactory for OL ≠ Φ, terminate matching process；

Step 3: traversing each target Oi (i=S1, S2 ..., Sn) in interim set TS, pass through object matching query set The target that matches with Oi in OL is closed, by the target to match from taking out and be saved in array A in set OL；If meeting array A is sky, executes step 5；If being unsatisfactory for array A as sky, interim set TL is emptied, and by the goal displacement in array A to temporarily In set TL；

Step 4: traversing each target Oj (j=L1, L2 ..., Lm) in interim set TL, pass through object matching query set The target to match in OS with Oj is closed, the target to match is taken out from set OS and is saved in array A；If meeting array A For sky, step 5 is executed；If being unsatisfactory for array A as sky, interim set TS is emptied, and the goal displacement in array A is collected to interim It closes in TS, executes step 3；

Step 5: taking out the target in interim set TL and interim set TS respectively, be recorded as matching pair, establishing element closes Connection relationship；TL and TS are emptied simultaneously, execute step 1.

As an improvement of the above scheme, poor using variation relation type, overlapping when carrying out the factor change infomation detection Different degree, shape similarity, size similitude and geometric area judge that factor change is believed as characteristic index, by decision-tree model Breath；

Wherein, the variation relation type is judged by spatial object overlay analysis, including 1:0,0:1,1:1, m: 1 (m > 1), m:0 (m > 1), m:n (m >=1, n > 1) six seed types, specific mode classification are as follows:

Enabling new large scale data is D1, and small scale data to be updated are D2；

If meet some target in D1 does not have matching object in D2, shows as single element target and increase newly, point Class is Class1: 0；

If meet some target in D2 does not have matching object in D1, shows as single element target and disappear, point Class is type 0:1；

If meeting in D1 and D2 has object matching to correspond to, but expansion, shrinkage phenomenon is being locally present in the target, is classified as Class1: 1；

If meeting in D1, multiple adjacent targets are corresponding with single target matching in D2, show as the merging of adjacent target, point Class is type m:1 (m > 1)；

If meeting the type variation to treat adjacent targets multiple in D1 as a whole, there is no matching correspondence in D2 Target or target complex, show as the newly-increased of element group, be classified as type m:0 (m > 1).

If meeting, single or multiple adjacent targets in D1 are corresponding with object matchings multiple in D2, and the structure between element target is closed System changes, and is classified as type m:n (m >=1, n > 1)；

The following formula of calculation formula (2) of the overlapping diversity factor:

In formula, φ is overlapping diversity factor, N_i(i=1,2 ..., I) is include I in match group from new large scale The target of footage evidence, P_j(j=1,2 ..., J) it is the J mesh from small scale data to be updated for including in match group Calculation is shipped in mark, ∩ expression, and ∪ indicates union；

The following formula of single target shape index calculation formula (3) of the shape similarity:

In formula, O_aFor single target, ShapeIndex (O_a) it is single target shape index, Perimeter (O_a) it is mesh Target perimeter, Area (O_a) be target area；

The size similitude is target and small scale footage to be updated in large scale data new in match group The area ratio of target in；

The geometric area includes target area (new_area) and small ratio to be updated in new large scale data The area (old_area) of target of the example footage in.

As an improvement of the above scheme, the process of the judgement factor change of the decision-tree model is as follows:

The variation relation of element is classified；

If change type is Class1: 0, the target area (new_area) in new large scale data is calculated, if meeting New_area > area change standard value determines that the element belongs to changed new element, if being unsatisfactory for new_area > area Change standard value, determines that the element is not belonging to changed new element；

If change type is type 0:1, determine that the element belongs to changed new element；

If change type is Class1: 1, overlapping diversity factor φ is calculated, if being unsatisfactory for φ > overlapping diversity factor standard value, is determined The element is not belonging to changed new element, if meeting φ > overlapping diversity factor standard value, calculates shape index ShapeIndex (O_a), if being unsatisfactory for ShapeIndex (O_a) > shape similarity standard value determines that the element belongs to changed new element, if Meet ShapeIndex (O_a) > shape similarity standard value calculates size similitude, if it is similar to meet size similitude > size Property standard value, determines that the element is not belonging to changed new element, if being unsatisfactory for size similitude > size similarity standard Value, determines that the element belongs to changed new element；

If change type is type m:1 (m > 1), overlapping diversity factor φ is calculated, if meeting φ > overlapping diversity factor standard value, Determine that the element belongs to changed new element, if being unsatisfactory for φ > overlapping diversity factor standard value, determines that the element is not belonging to send out The new element for changing；

If change type is type m:0 (m > 1), the target area (new_area) in new large scale data is calculated, If meeting new_area > area change standard value, determine that the element belongs to changed new element, if being unsatisfactory for new_area > area change standard value determines that the element is not belonging to changed new element；

If change type is type m:n (m >=1, n > 1), determine that the element belongs to changed new element.

As an improvement of the above scheme, the progress cartographic generaliztion is redefined is described in the form of hexa-atomic group, is specifically retouched It states are as follows: (< layer identification code >, < operation operator >, < attribute code >, < index item >, < lower limit >, < upper limit >)；

Wherein, < layer identification code > determines that the property layer that this rule is applicable in, < operation operator > determine the synthetic operation of this rule, The operation includes deletion, merging and abbreviation；< attribute code > determines the objectives that this rule is applicable under certain layer；< index item > is true The characteristic item that set pattern is then directed to；< upper limit > and < lower limit > determines the value range of index item；

Hexa-atomic group of the general meaning can be expressed as: when the target in < layer identification code > has < attribute code >, and its < index When item > is less than < upper limit > and is greater than < lower limit >, < operation operator > is executed.

As an improvement of the above scheme, in the increment updating method for carrying out object-oriented, the update operation of object can It is divided into creation, deletion, geometric modification and attribute modification；

For newly-increased object, operated using creation；

For the object of disappearance, delete operation is used；

For the object of geometry or attribute change occurs, geometric modification or attribute modification are carried out；

For the object of merging, decomposition and polymerization, carries out deleting former object, create matching target object.

As an improvement of the above scheme, the specific method is as follows for the progress space inconsistency detection and processing:

Area target is shared into the inconsistent type in side and is divided into intersection type, mutually release, intertexture type；

The neutrodyne that unification processing mode is divided into biting connecions processing and precision equality that positioning accuracy is dominant is handled；

Establish six kinds of modes of spatial relationship consistency maintenance: intersection type+biting connecions,<2>intersection type+neutrodyne,<3>phase Release+biting connecions,<4>are mutually release+neutrodyne,<5>intertexture type+biting connecions and<6>intertexture type+neutrodyne；

The inconsistent regional area in boundary expressed based on the neighbouring analysis detection of Delaunay triangulation network model by triangle collection；

It is inconsistent to boundary to correct by triangulation network skeleton line drawing；

Boundary is made to correct loss of significance by equal part feature of the triangulation network skeleton line on uniformly subdivision minimum；

Mesh data topological relation connectivity correcting method is divided into point-wire connectivity conflict correction, line-wire connectivity punching Prominent correction and line-face connectivity conflict correction；

The linear river endpoint that conflict relationship will be present by extending segmental arc extends to the position of point target, makes linear river The connectivity conflict correction between point-line is realized in the connection of stream and dotted well；

By moving back a displacement connection method and mobile endpoint location method, the connectivity conflict correction between line-line is realized；

Using mobile endpoint location method and newly-increased segmental arc method, line-face connectivity conflict correction is realized.

As an improvement of the above scheme, it if element to be updated is road element, can also be carried out by BP neural network It updates, specific update method is as follows:

Obtain the object composition of large scale data and the training sample in small scale data；

Sample training is carried out, the update being passed and the update not being passed；

The training of BP nerve is carried out, weight matrix and bias vector are obtained, constructs forecasting model database；

The change information in new large scale data and small scale data to be updated is obtained, variation characteristic is calculated and refers to Mark；

In conjunction with the forecasting model database obtained through the training of BP nerve, the scale transmitting of change information is differentiated；

Small scale data to be updated are updated；

Wherein, it is as follows to carry out the housebroken process of BP mind:

Using training sample as input value and target value；

Initialization weight is combined to obtain predicted value by prediction process target value；

By predicted value and target value entrance loss function, penalty values are obtained；

Penalty values are calculated into gradient by backpropagation, obtain new initialization weight.

A kind of multiple dimensioned map data updating method provided in an embodiment of the present invention has as follows compared with prior art The utility model has the advantages that

Matched using multiscale target, including semantic similarity matching, Euclidean distance similarity mode, Hausdorff away from From matching, Knot Searching and position adjacent degree or overlapping similarity mode, improve the matched range of target component, accuracy and Accuracy, wherein the comparison based on semantic similarity, can effectively solve because data source it is inconsistent caused by with put it is not of the same name Phenomenon greatly improves matched accuracy；Factor change is judged by establishing element grade incidence relation and decision-tree model Information carries out part update when map elements change, and existing element is avoided to repeat to update, and improves map match effect Rate, the standard value of decision-tree model can be obtained or are manually set by the training in not same area, to adapt to different zones environment Under the conditions of variation identification；Cartographic generaliztion is described by way of hexa-atomic group to redefine, and keeps the matching operation of map more succinct；It is logical The detection of space inconsistency and processing are crossed, cartographic accuracy, stability and reliability are improved；When element to be updated is wanted for road It when plain, is updated by BP neural network, the demand of multi-Scale Intelligentization update can be substantially met, operation week is greatly reduced Phase reduces operating cost, improves operating efficiency, saves the activity duration.

Detailed description of the invention

Fig. 1 is a kind of flow diagram of multiple dimensioned map data updating method provided in an embodiment of the present invention.

Fig. 2 is a kind of signal of the decision-tree model of multiple dimensioned map data updating method provided in an embodiment of the present invention Figure.

Fig. 3 is that a kind of cartographic generaliztion of multiple dimensioned map data updating method provided in an embodiment of the present invention redefines front and back Comparison diagram.

Fig. 4 is a kind of BP neural network change information of multiple dimensioned map data updating method provided in an embodiment of the present invention Matched flow diagram.

Specific embodiment

Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other Embodiment shall fall within the protection scope of the present invention.

It is a kind of flow diagram of multiple dimensioned map data updating method provided in an embodiment of the present invention referring to Fig. 1, tool Steps are as follows for body update:

S1, multiscale target matching is carried out to new large scale data and small scale data to be updated；

Wherein, it when matching target is dotted entity, is matched using semantic similarity and Euclidean distance similarity: from Proper noun is extracted in reference data place name, address and POI data library, obtains proprietary vocabulary and near synonym group；By proprietary vocabulary Dictionary is added to near synonym group；Dictionary is imported into segmenter, obtains word segmentation result；Word segmentation result is created into inverted entry rope Draw, and saves to index document；It is mentioned from reference data place name, address, POI database and other data containing location information Location information is taken, position character information is obtained；Position character information is imported into segmenter, obtains word segmentation result；It is tied according to participle Fruit establishes trie tree；Search result is obtained in index document by the search index in trie tree；It is similar according to following semanteme The judgment formula (1) of degree sorts to search result by semantic similarity, obtains matching result:

In formula, sim (w_k, S_i) | i ∈ (1,2,3 ... n) indicates query term search term S_iWith the W of reference data document d_k's Semantic similarity.max{sim(w_k, S_i) | i ∈ (1,2,3 ... n) } indicate and W_kThe maximum value of semantic similarity.

S2, establishing element grade incidence relation；

Wherein, the specific method of establishing element grade incidence relation is using alternative manner, and specific construction step is as follows:

S3, factor change infomation detection is carried out；

It referring to fig. 2, is a kind of decision-tree model of multiple dimensioned map data updating method provided in an embodiment of the present invention Schematic diagram；

Wherein, using variation relation type, overlapping diversity factor, shape similarity, size similitude and geometric area conduct Characteristic index judges factor change information by decision-tree model；

Judge variation relation type by spatial object overlay analysis, including 1:0,0:1,1:1, m:1 (m > 1), m:0 (m > 1), m:n (m >=1, n > 1) six seed types, specific mode classification are as follows:

In formula, O_aFor single target, ShapeIndex (Oa) is single target shape index, Perimeter (O_a) it is mesh Target perimeter, Area (O_a) be target area；

Referred to using variation relation type, overlapping diversity factor, shape similarity, size similitude and geometric area as characteristic The process of the judgement factor change of target decision-tree model is as follows:

The variation relation of element is classified；

S4, progress cartographic generaliztion redefine；

Wherein, it carries out cartographic generaliztion and redefines the description in the form of hexa-atomic group, specifically describe are as follows: (< layer identification code >, < behaviour Make operator >, < attribute code >, < index item >, < lower limit >, < upper limit >)；

< layer identification code > determines the property layer that this rule is applicable in, and < operation operator > determines the synthetic operation of this rule, described Operation includes deletion, merging and abbreviation；< attribute code > determines the objectives that this rule is applicable under certain layer；< index item > determines rule The characteristic item being then directed to；< upper limit > and < lower limit > determines the value range of index item；

Hexa-atomic group of the general meaning can be expressed as: when the target in < layer identification code > has < attribute code >, and its < index When item > is less than < upper limit > and is greater than < lower limit >, executing < operation operator > referring to Fig. 3 is that one kind provided in an embodiment of the present invention is more The cartographic generaliztion of scale map data updating method redefines the comparison diagram of front and back.

S5, the incremental update for carrying out object-oriented；

Wherein, the update operation of object can be divided into creation, deletion, geometric modification and attribute modification；For newly-increased object, It is operated using creation；For the object of disappearance, delete operation is used；For the object of geometry or attribute change occurs, into Row geometric modification or attribute modification；For the object of merging, decomposition and polymerization, carries out deleting former object, create matching Target object.

S6, the detection of space inconsistency and processing are carried out；

Wherein, the specific method is as follows:

It further, referring to fig. 4, is a kind of BP mind of multiple dimensioned map data updating method provided in an embodiment of the present invention Flow diagram through network change information matches can also pass through BP neural network if element to be updated is road element It is updated, specific update method is as follows:

The training of BP nerve is carried out as follows: using training sample as input value and target value；Target value is combined It initializes weight and predicted value is obtained by prediction process；By predicted value and target value entrance loss function, penalty values are obtained；It will damage Mistake value calculates gradient by backpropagation, obtains new initialization weight.

Terminate the training of BP nerve, obtain weight matrix and bias vector, constructs forecasting model database；

The change information in new large scale data and small scale data to be updated is obtained, variation characteristic is calculated and refers to Mark, variation characteristic index include length, width, grade, degree of communication, reticular density, roading density and center Jie of road etc.；

Small scale data to be updated are updated.

The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as Protection scope of the present invention.

Claims

1. a multi-scale map data update method, is characterized in that, comprises the steps:

Perform multi-scale target matching on new large-scale data and small-scale data to be updated;

Build feature-level associations;

Perform element change information detection;

Carry out cartographic synthesis redefinition;

Perform object-oriented incremental updates;

Perform spatial inconsistency detection and processing.

2. a kind of multi-scale map data update method as claimed in claim 1 is characterized in that, the described method that multi-scale target matching is carried out to new large-scale data and small-scale data to be updated is as follows:

When the matching target is a point entity, semantic similarity and Euclidean distance similarity are used for matching;

When the matching target is a linear entity, the matching is performed by calculating the Hausdorff distance and the node;

When the matching target is a planar entity, it is matched by position adjacency or overlapping similarity.

3. a kind of multi-scale map data update method as claimed in claim 2, is characterized in that, adopts semantic similarity and distance similarity to carry out the concrete method of matching as follows:

Extract proper nouns from reference place names, addresses and POI databases to obtain proper words and synonyms;

Add proper words and synonyms to the thesaurus;

Import the thesaurus into the tokenizer to get the word segmentation result;

Create an inverted document index from the word segmentation result and save it to the index document;

Extract location information from reference place names, addresses, POI databases and other data containing location information to obtain location character information;

Import the position character information into the tokenizer to get the word segmentation result;

Build a search tree according to the word segmentation results;

Obtain the retrieval result in the index document through the retrieval index in the retrieval tree;

According to the judgment formula (1) of the following semantic similarity, the retrieval results are sorted according to the semantic similarity, and the matching results are obtained:

In the formula, sim(w _k , S _i )|i∈(1,2,3…n) represents the semantic similarity between the search term Si of the query item and the _W _k of the reference document d, max{sim(w _k , S _i )|i∈(1,2,3…n)} represents the maximum value of semantic similarity with W _k .

4. A kind of multi-scale map data update method as claimed in claim 1, is characterized in that, the concrete method of described construction element level association is to adopt iterative method, and concrete construction steps are as follows:

Let the target set contained in the new large-scale data be OL={OL1,OL2,...,OLm}, and the target set contained in the small-scale data to be updated is OS={OS1,OS2,...,OSn}, and the set TL and Set TS is defined as a temporary set that stores objects from set OL and set OS respectively, and initializes TL=Φ, TS=Φ;

Step 1: If OS≠Φ is satisfied, take out the target in OS and store it in TS, and go to Step 4; if OS≠Φ is not satisfied, go to Step 2;

Step 2: If OL≠Φ is satisfied, go to step 3; if OL≠Φ is not satisfied, end the matching process;

Step 3: Traverse each target Oi (i=S1, S2, . Save it in array A; if it is satisfied that array A is empty, go to step 5; if it is not satisfied that array A is empty, empty the temporary collection TL, and transfer the target in array A to the temporary collection TL;

Step 4: Traverse each target Oj (j=L1, L2, . into the array A; if it is satisfied that the array A is empty, go to step 5; if it is not satisfied that the array A is empty, clear the temporary set TS, and transfer the target in the array A to the temporary set TS, and go to step 3;

Step 5: Respectively take out the targets in the temporary set TL and the temporary set TS, record them as matching pairs, and construct an element association relationship; at the same time clear the TL and TS, and perform step 1.

5. A method for updating multi-scale map data as claimed in claim 1, characterized in that, when the element change information is detected, the change relationship type, overlap difference degree, shape similarity, size similarity and geometric area are adopted. As a characteristic index, the change information of elements is judged through a decision tree model;

Wherein, the type of change relationship is judged by spatial object superposition analysis, including 1:0, 0:1, 1:1, m:1 (m>1), m:0 (m>1), m:n ( m≥1, n>1) six types, the specific classification methods are as follows:

Let the new large-scale data be D1, and the small-scale data to be updated be D2;

If a target in D1 is satisfied and there is no matching object in D2, it means that a single element target is newly added, and it is classified as type 1:0;

If a target in D2 does not have a matching object in D1, it means that a single element target disappears, and it is classified as type 0:1;

If it is satisfied that D1 and D2 have target matching correspondence, but the target has local expansion and contraction phenomenon, it is classified as type 1:1;

If it is satisfied that multiple adjacent targets in D1 correspond to a single target in D2, it is manifested as a merger of adjacent targets, and is classified as type m:1 (m>1);

If this type of change is satisfied, the multiple adjacent targets in D1 are regarded as a whole, and there is no matching target or target group in D2, which is manifested as a new element group, and is classified as type m:0 (m>1) ;

If it is satisfied that a single or multiple adjacent targets in D1 match with multiple targets in D2, the structural relationship between the element targets changes, and it is classified as type m:n (m≥1, n>1);

The calculation formula of the overlap difference degree is the following formula (2):

In the formula, φ is the degree of overlap difference, Ni ( _i =1,2,…,I) is the I target from the new large-scale data included in the matching group, _Pj (j=1,2,…, J) is J included in the matching group from the targets of the small scale data to be updated, ∩ represents the intersection operation, and ∪ represents the parallel operation;

The calculation formula of the single target shape index of the shape similarity is the following formula (3):

In the formula, O _a is a single target, ShapeIndex(O _a ) is a single target shape index, Perimeter(O _a ) is the perimeter of the target, and Area(O _a ) is the area of the target;

The size similarity is the area ratio of the target in the new large-scale data in the matching group to the target in the small-scale data to be updated;

The geometric area includes the target area (new_area) in the new large-scale data and the target area (old_area) in the small-scale data to be updated.

6. a kind of multi-scale map data update method as claimed in claim 5 is characterized in that, the flow process that the judgment element of described decision tree model changes is as follows:

Classify the changing relationship of elements;

If the change type is type 1:0, calculate the target area (new_area) in the new large-scale data. If it satisfies the new_area> area change standard value, it is determined that the element is a new element that has changed. If it does not meet the new_area> area change standard value, determine that the element is not a new element that has changed;

If the change type is type 0:1, it is determined that the element is a new element that has changed;

If the change type is type 1:1, calculate the overlap difference degree φ, if it does not satisfy the standard value of φ > overlap difference degree, determine that the element does not belong to the new element that has changed, if it meets the standard value of φ > overlap difference degree, calculate the shape index ShapeIndex(O _a ), if it does not satisfy the ShapeIndex(O _a )> standard value of shape similarity, it is determined that the element belongs to a new element that has changed, and if it meets the standard value of ShapeIndex (O _a ) > shape similarity, the calculated size is similar If it satisfies the standard value of size similarity > size similarity, it is determined that the element does not belong to the new element that has changed; if it does not satisfy the standard value of size similarity > size similarity, it is determined that the element belongs to the new element that has changed;

If the change type is type m:1 (m>1), calculate the overlap difference degree φ, if it satisfies the standard value of the overlap difference degree, it is determined that the element belongs to the new element that has changed, if it does not satisfy the standard value of the overlap difference degree , determine that the element is not a new element that has changed;

If the change type is type m:0 (m>1), the target area (new_area) in the new large-scale data is calculated. If it satisfies the standard value of new_area> area change, it is determined that the element belongs to the new element that has changed. new_area> standard value of area change, determine that the element does not belong to the new element that has changed;

If the change type is type m:n (m≥1, n>1), it is determined that the element belongs to a new element that has changed.

7. a kind of multi-scale map data update method as claimed in claim 1 is characterized in that, described carrying out cartographic synthesis redefinition adopts the form description of six-tuple, and is specifically described as: (<layer code>, <operation calculation. sub>, <attribute code>, <indicator item>, <lower limit>, <upper limit>);

Among them, <layer code> determines the property layer to which this rule applies, <operation operator> determines the comprehensive operation of this rule, and the operations include deletion, merging and simplification; Specific goals; <indicator item> determines the characteristic item targeted by the rule; <upper limit> and <lower limit> determine the value range of the indicator item;

The general meaning of the six-tuple can be expressed as: when the target in <layer code> has <attribute code>, and its <index item> is less than <upper limit> and greater than <lower limit>, execute <operation operator>.

8. a kind of multi-scale map data update method as claimed in claim 1 is characterized in that, in described carrying out object-oriented incremental update method, the update operation of object can be divided into creation, deletion, geometry modification and attribute modification ;

For newly added objects, use the create operation;

For objects that disappear, use the delete operation;

For objects whose geometry or attributes have changed, perform geometry modification or attribute modification;

For merged, decomposed and aggregated objects, delete the original object and create a matching target object.

9. A kind of multi-scale map data updating method as claimed in claim 1, is characterized in that, the concrete method of described carrying out spatial inconsistency detection and processing is as follows:

Divide the inconsistent types of shared edges of planar objects into intersection type, phase separation type, and interweaving type;

The unification processing methods are divided into occlusal processing with superior positioning accuracy and adjustment processing with equal precision;

Six ways to establish the consistency maintenance of spatial relationship: intersection type + bite type, <2> intersection type + adjustment type, <3> separation type + bite type, <4> separation type + adjustment type , <5> interweaving type + bite type and <6> interweaving type + adjustment type;

Proximity analysis based on Delaunay triangulation model to detect the local area of inconsistent boundary expressed by triangle set;

Correct the boundary inconsistency by extracting the skeleton line of the triangulation;

The loss of boundary correction accuracy is minimized by the bisection feature of the triangulation skeleton line in the spatial division;

The connectivity correction methods of mesh data topology relationship are divided into point-line connectivity conflict correction, line-line connectivity conflict correction and line-surface connectivity conflict correction;

By extending the arc segment, the end point of the linear river with conflicting relationship is extended to the position of the point-shaped target, so that the linear river and the point-shaped well are connected, and the connection conflict between the point and the line can be corrected;

The connection conflict correction between line-line is realized by the method of back-off point shift connection and the method of moving end point position;

The line-surface connectivity conflict correction is realized by the method of moving the end point position and the new arc segment method.

10. a kind of multi-scale map data update method as claimed in claim 1 is characterized in that, if the element to be updated is road element, can also be updated through BP neural network, and the concrete update method is as follows:

Get object combinations of training samples in large-scale data and small-scale data;

Perform sample training to get passed updates and undelivered updates;

Perform BP neural training, obtain weight matrix and bias vector, and build a prediction model library;

Obtain the change information in the new large-scale data and the small-scale data to be updated, and calculate the change characteristic index;

Combined with the prediction model library obtained by BP neural training, the scale transmission of the change information is judged;

Update the small scale data to be updated;

Among them, the process of BP neural training is as follows:

Use training samples as input and target values;

Combine the target value with the initialization weight to obtain the predicted value through the prediction process;

Input the predicted value and target value into the loss function to get the loss value;

The loss value is back-propagated to calculate the gradient, and the new initialization weight is obtained.