CN105608073A - Modeling method for GIS data - Google Patents

Abstract

The invention relates to a modeling method for GIS (Geographic Information System) data. The modeling method comprises the following steps: establishing a road network set containing stationary points with semantic information, calculating public interval lengths of edges in a semantic trajectory a and a semantic trajectory b, similarity between two edges based on the stationary points and normalized similarity based on distance, calculating similarity between the two edges in the semantic trajectory a and the semantic trajectory b, calculating first similarity and second similarity of the semantic trajectory a and the semantic trajectory b according to the similarity between the two edges and finally obtaining maximum weighted similarity of the semantic trajectory a and the semantic trajectory b. Therefore, according to the modeling method provided by the invention, when the similarity between the trajectories is calculated, semantic information is used, so that calculation results are more accurate.

Description

A kind of modeling method for GIS data

Technical field

The application relates to the application of track data, especially a kind of modeling method for GIS data.

Background technology

In recent years, along with in daily life universal of the smart machine that carries GIS sensor, to the research of track dataAlso more and more burning hoter. Wherein, track similarity is as the basic index of the base application such as geography recommendation, stroke prediction, its calculatingMethod has been the basic fundamental of location-based service.

But in technology, the computational methods of track similarity are mostly just calculated the shape between track in coarseness ground nowSimilarity, seldom takes the semantic information of adhering on track into account.

Summary of the invention

The application provides a kind of modeling method for GIS data, solves in prior art, calculates track similarity and does not examineConsider the problem of semantic information.

For a modeling method for GIS data, comprise the following steps:

Obtain GIS data, from GIS data, detect the stationary point with semantic information;

Foundation comprises the road network set G with semantic information stationary point_ST(VV_ST，EE_ST, wherein, the set that V is node,E is the set on limit, V_STFor the node set of stationary point composition, E_STBe the limit that is linked to be of two stationary points and stationary point withThe set on the limit that other nodes are linked to be, road network set G_STIn comprise semantic track a and semantic track b;

The public region length on limit in computing semantic track a and semantic track b | sim (e_a，i，e_b，j) |, wherein, e_a，iRepresentI article of limit in semantic track a, e, represents j article of limit in semantic track b;

The length of the public region of each opposite side in semantic track a and semantic track b is added, obtains:

$Sim({\mathrm{STN}}_a,{\mathrm{STN}}_b)=\frac{{\Σ}_{i=1}^n{\Σ}_{j=1}^m\left|sim(e_{a,i},e_{b,j})\right|}{\max \left\{\right|S_a|,|S_b\left|\right\}};$

Wherein, STN_aRepresent semantic track a, STN represents semantic track b, | S_a| be the overall length of stationary point in semantic track aDegree, | S_b| be the total length of stationary point in semantic track b;

The similarity Sim on two limits based on stationary point in computing semantic track a and semantic track b_s

Based on the distance between different edge in semantic track a and semantic track b, calculate the regularization similarity based on distanceSim

Similarity in computing semantic track a and semantic track b between two limits:

sim_w(e_a，i，e_b，j)＝βsim_d(e_a，i，e_b，j)+(1-β)sim_s(e_a，i，e_b，j)；

Wherein, 0≤β≤1st, weight coefficient;

The first similarity of computing semantic track a and semantic track b is again:

By the timeline of semantic track b translation Δ t in each step, then calculate semantic track a and semantic track b after translationThe second similarity:

${\mathrm{Sim}}_{W,r}({\mathrm{STN}}_a,{\mathrm{STN}}_b)=\frac{\mathrm{\Δ}T-r\×\mathrm{\Δ}t}{\mathrm{\Δ}T}{\mathrm{Sim}}_W({\mathrm{STN}}_a,{\mathrm{STN}}_b^r);$

Wherein, Δ T=max (| e_a，1.tin-e_b，m.tout|，|e_b，1.tin-e_a，n.tout|), n is limit in semantic track aSum, m is the sum on limit in semantic track b, e_a，1.tin be through limit e_a，The timestamp of start node, e_b，1.tin be processLimit e, the timestamp of start node, e_b，m.tout be through limit e_b，mThe timestamp of end node, e_a，n.tout be through limit e_a，nThe timestamp of end node, in the translation of r step the timeline of r × Δ t be labeled as

According to the first similarity and the second similarity, obtain the maximum weighted similarity of track a and track b:

The application's beneficial effect is, because the application sets up the road network set comprising with semantic information stationary point, to countCalculate the similarity on the public region length on limit in semantic track a and semantic track b, two limits based on stationary point and based on distanceRegularization similarity, then the similarity between two limits in computing semantic track a and semantic track b, according to the phase between limitLike degree, the first similarity of computing semantic track a and semantic track b and the second similarity, finally obtain track a and track bMaximum weighted similarity. Therefore, the application, in the time of the similarity of calculating between track, has used semantic information, makes result of calculation morePrecisely.

Brief description of the drawings

Fig. 1 is the schematic diagram of certain road network in embodiment 1;

Fig. 2 is the shape schematic diagram of certain road network track in embodiment 1;

Fig. 3 be in embodiment 1 certain mark the road network schematic diagram of stationary point;

Fig. 4 is the corresponding timeline schematic diagram of semantic track in Fig. 3;

Fig. 5 is the schematic diagram that in embodiment 1, semantic track a and semantic track b compare.

Detailed description of the invention

By reference to the accompanying drawings the present invention is described in further detail below by detailed description of the invention.

Embodiment 1:

For a modeling method for GIS data, comprise the following steps:

S101: obtain GIS data, detect the stationary point with semantic information from GIS data.

GIS is GIS-Geographic Information System, can obtain by GIS location technology user's GIS data, the position clothes of for example micro-letterBusiness etc., available k-means clustering method detects the stationary point with semantic information from GIS data, and a stationary point refers toThe entity of a movement has stopped the place of certain hour in certain geographic range, and semantic information comprises time and position coordinatesEtc. information.

S102: set up the road network set G comprising with semantic information stationary point_ST(VV_ST，EE_ST, wherein, V is nodeSet, the set that E is limit, V_STFor the node set of stationary point composition, E_STBe that the limit that is linked to be of two stationary points and one stayPut the set on the limit being linked to be with other nodes, road network set G_STIn comprise semantic track a and semantic track b.

As shown in Figure 1, road network has comprised node v1, v2, v3, v4, v5 and v6, and each node represents intersection,Line between node is limit, and a limit represents a section, and semantic track is by a series of stationary points with semantic informationComposition. Fig. 2 shows the schematic diagram of track in road network, and arrow wherein represents moving direction, as can be seen from Figure, and this semanteme railMark is v1-> v3-> v5-> v4-> v2.

For semantic information is attached on trace information, we add route using the stationary point detecting as new nodeIn net, form final road network set G_ST(VV_ST，EE_ST. Each stationary point needs to mark two timestamps: the time of adventAnd time departure, therefore, need to define final semantic track: P_ST Wherein, v is stationary point or node, nFor the number of stationary point and node, tin is the time of advent, and tout is time departure. In Fig. 3, υ₁，υ₃，υ₅，υ₄，υ₂∈ V is jointPoint, υ₇，υ₈，υ₉，υ₁₀∈V_STFor stationary point.

By P_STBe expressed as a timeline being formed by stationary point and intersection. Fig. 4 has represented semantic track in Fig. 3Corresponding timeline, timeline is made up of a series of continuous panes, and each pane has represented stretchThe running time of journey. The length of pane has just represented from an intersection long to the running time of next crossingDegree. In Fig. 4, have five panes (v1, v3), (v3, v5), (v5, v4), (v4, v2), and t1, t4, t5, t10, t13 is respectivelyRepresent that mobile entity is through v1, v3, v5, v4, timestamp when v2. Meanwhile, the stationary point on the semantic track of having gone back mark, in figureOblique line part represent the residence time of mobile entity at stationary point, the starting point of oblique line part represents to arrive the time of stationary pointTin, the terminal of oblique line part represents to leave the time tout of stationary point. Language in four oblique line parts difference presentation graphs 3 in Fig. 4Four stationary point v7 of justice track, v8, v9, v10, be respectively (t2, t3) its residence time, (t6, t7), (t8, t9), (t11,t12)。

S103: the public region length on limit in computing semantic track a and semantic track b | sim (e_a，i，e_b，j) |, wherein,e_a，iRepresent i article of limit in semantic track a, e, represents j article of limit in semantic track b.

Semantic track a and semantic track b are object to be studied, hereinafter, use STN_aRepresent semantic track a, STN representsSemantic track b. Public region refers to different the mobile entities identical or close stationary point of institute's co-occurrence in interval at one time,Wherein, close stationary point refers to that the distance between stationary point is less than certain penalties. Limit in semantic track a and semantic track bPublic region be semantic track a arbitrarily on one side and semantic track b identical or close stationary point in one side arbitrarily.

S104: the length of the public region of each opposite side in semantic track a and semantic track b is added, obtains:

$Sim({\mathrm{STN}}_a,{\mathrm{STN}}_b)=\frac{{\Σ}_{i=1}^n{\Σ}_{j=1}^m\left|sim(e_{a,i},e_{b,j})\right|}{\max \left\{\right|S_a|,|S_b\left|\right\}};$

Wherein, | S_a| be the total length of stationary point in semantic track a, | S_b| be the total length of stationary point in semantic track b.In formula, molecule is the total length of the public region of all stationary points, and denominator is the maximum of stationary point total length in two tracks.If the length of public region equals the length of the maximum, similarity is 1; If there is no public region, its similarity is0。

Here it should be noted that, before calculating public region, must first find out semantic track a and semantic track bBetween public region. Cross and progressively filter out not identical public region by following steps: (1) first checks two semantic railsWhether mark there is overlapping running time; (2) overlapping if running time exists, reexamine them and whether be under the jurisdiction of sameLimit; (3) if two tracks be under the jurisdiction of same limit and running time exist overlapping, reexamine on same limit, whether exist funnyStay overlapping stationary point of time; (4) finally check that whether residence time overlapping stationary point is identical or close.

As shown in Figure 5, four above-mentioned steps of specific explanations. In figure, be semantic track a and semantic track b. Obviously, semantemeTrack a and semantic track b have shared the running time of a very long time. Then, check whether they are under the jurisdiction of same limit.In figure, (v1, the v2) of semantic track a and semantic track b (v10, v3) do not belong to same limit, do not have public affairs between themInterval altogether. The Part II of semantic track a and semantic track b is all (v3, v4), and they have respectively three and two to stayPoint. Because the distance between v7 and v12 is less than θ, between them, there is public region; , between v13, also there is public area in like manner v7Between. Although it is overlapping that the running time between v8 and v14 exists, the distance between them is greater than threshold values θ, so there are not public affairsInterval altogether.

S105: the similarity Sim on two limits based on stationary point in computing semantic track a and semantic track b_s

Because strict constraint when modeling on time dimension, the similarity having between a lot of tracks is zero. Such as, even ifArticle two, the shape of semantic track is identical with stationary point, but two tracks on time dimension without any overlapping,The similarity of two tracks is still 0. But this tolerance can not characterize the similarity between track definitely.

Therefore timeline that, can translation track. Article two, may not there is not public region in semantic track before translation, but flatAfter moving, can share one section of public region. But translation meeting causes similarity distortion between track. So we can be with the timeThe distance that line moves is weight, and the calculating of public region is applied to intervention. The public region of stationary point and stationary point semantemeLabel is all included in and is considered.

S105: based on the distance between different edge in semantic track a and semantic track b, calculate the regularization based on distanceSimilarity Sim

In the time weighing the similarity on two limits, for fear of by two very close limits in road network, because do not have publicInterval and show that its similarity is 0 conclusion, need to consider the distance between different edge in road network. Therefore, calculate based on distanceRegularization similarity.

S106: the similarity in computing semantic track a and semantic track b between two limits:

sim_w(e_a，i，e_b，j)＝βsim_d(e_a，i，e_b，j)+(1-β)sim_s(e_a，i，e_b，j)；

Wherein, 0≤β≤1st, weight coefficient.

Result of calculation in comprehensive step S104 and step S105, finally obtains the similarity on two limits.

S106: the first similarity of computing semantic track a and semantic track b is again:

S107: by the timeline of semantic track b translation Δ t in each step, then calculate semantic track a and semanteme after translationThe second similarity of track b:

${\mathrm{Sim}}_{W,r}({\mathrm{STN}}_a,{\mathrm{STN}}_b)=\frac{\mathrm{\Δ}T-r\×\mathrm{\Δ}t}{\mathrm{\Δ}T}{\mathrm{Sim}}_W({\mathrm{STN}}_a,{\mathrm{STN}}_b^r);$

Wherein, Δ T=max (| e_a，1.tin-e_b，m.tout|，|e_b，1.tin-e_a，n.tout|), n is limit in semantic track aSum, m is the sum on limit in semantic track b, e_a，1.tin be through limit e_a, the timestamp of start node, e_b，1.tin be warpCross limit e, the timestamp of start node, e_b，m.tout be through limit e_b，mThe timestamp of end node, e_a，n.tout be through limite_a，nThe timestamp of end node, in the translation of r step the timeline of r × Δ t be labeled as

Step S105 above mentions, in the time of the timeline of relatively different tracks, can translation timeline, and suppose that we are oftenThe distance of the timeline of one step translation STN is all Δ t, in the translation of r step the timeline of r × Δ t be labeled asOn this basis, calculate the second similarity of semantic track a and semantic track b.

S108: according to the first similarity and the second similarity, obtain the maximum weighted similarity of track a and track b:

Embodiment 2:

On the basis of embodiment 1, the public region length on limit in semantic track a and semantic track b | sim (e_a，i，e_b，j)| calculating formula be:

$\left|sim(e_{a,i},e_{b,j})\right|={\Σ}_{k=1}^{count\left(S_{a,i}\right)}{\Σ}_{l=1}^{count\left(S_{b,j}\right)}|s_{a,k}\∩s_{b,j}|;$ Semantic track a and semantemeThe similarity Sim on two limits based on stationary point in track b_s Calculating formula be:

The public region that this formula has comprised stationary point and stationary point semantic label, its Part I has calculated public region and has accounted forThe ratio of stationary point maximum length, Part II has calculated the similarity of two semantic label set.

Regularization similarity Sim based on distance in semantic track a and semantic track bCalculating formula be:

${\mathrm{sim}}_d(e_{a,i},e_{b,j})=\frac{\max \left(\right|\left|{\mathrm{STN}}_a\right||,|\left|{\mathrm{STN}}_b\right|\left|\right)-Dist(e_{a,i},e_{b,j})}{\max \left\{\right|\left|{\mathrm{STN}}_a\right||,|\left|{\mathrm{STN}}_b\right|\left|\right\}};$

Need to say, calculate the regularization similarity based on distance, need e_a,iAnd e, between distance B ist (e_a，i，e_b，j) meet following condition:

|Dist(e_a，i，e_b，j)≤max{||STN_a||，||STN_b||}。

In above-mentioned several formula, s_a，kFor k stationary point in semantic track a, s_b，lFor the l in semantic track bStationary point, countS_a,i|S_a,i|,countS,|S,|，LS_a,iAnd LS, be respectively S_a,iAnd S, the semanteme of upper all stationary pointsTag set, 0≤α≤1st, weight parameter.

Embodiment 3:

On the basis of embodiment 1 or embodiment 2, in computing semantic track a and semantic track b, the public region on limit is longDegree | sim (e_a，i，e_b，j) |, also comprise before:

Judge whether to meet following condition:

Wherein, 1≤k≤count (S_a，i)，1≤l≤count(S_b，j)；

If meet, have public region; If do not meet, the length of public region is 0.

Judge whether the limit in semantic track a and semantic track b exists public region, in the time that it meets above-mentioned condition,Illustrate and have public region; If do not meet, the length of public region is 0. Only have in the time that public region exists, can realize follow-upCalculating.

Further, by contrast summit, check between two limits whether have public region. Suppose STN_aComprise n barLimit, e_a,iIts i article of limit; Suppose that STN comprises m bar limit, e, its j article of limit, can obtain:

Wherein 1≤i≤n, 1≤j≤m, e_a，i≠e_b，jRepresent e_a,iAnd e, not same limit. | e_a，i.tin be through e_a,iTimestamp when start node, e_a，i.tout be through e_a,iTimestamp when end node.

Further, suppose STN_aComprise g stationary point, | s_a，k∈V_STK stationary point wherein; STN comprises hStationary point, s_b，l∈V_STL stationary point wherein. can obtain:

,, in the time meeting above formula, illustrate between two stationary points and have public region.

Wherein 1≤k≤g, 1≤l≤h, s_a，k.tin be s_a, time of advent stamp, s_a，k.tout be s_a, time departureStamp, dist (s_a，k，s_b，l) be s_a, ands. between Euclidean distance.

Above content is in conjunction with concrete embodiment further description made for the present invention, can not assert thisBright concrete enforcement is confined to these explanations. For general technical staff of the technical field of the invention, not de-Under the prerequisite of the present invention's design, can also make some simple deduction or replace.

Claims (3)

1. for a modeling method for GIS data, it is characterized in that: comprise the following steps:

Obtain GIS data, from GIS data, detect the stationary point with semantic information;

Foundation comprises the road network set with semantic information stationary pointWherein, V is nodeSet, the set that E is limit, V_STFor the node set of stationary point composition, E_STBe that the limit that is linked to be of two stationary points and one stayPut the set on the limit being linked to be with other nodes, road network set G_STIn comprise semantic track a and semantic track b;

The public region length on limit in computing semantic track a and semantic track b | sim (e_a，i，e_b，j) |, wherein, e_a，iRepresent semanticI article of limit in track a,Represent j article of limit in semantic track b;

The length of the public region of each opposite side in semantic track a and semantic track b is added, obtains:

$Sim({\mathrm{STN}}_a,{\mathrm{STN}}_b)=\frac{{\mathrm{\Σ}}_{i=1}^n{\mathrm{\Σ}}_{j=1}^m\left|sim(e_{a,i},e_{b,j})\right|}{max\left\{\right|S_a|,|S_b\left|\right\}};$

Wherein, STN_aRepresent semantic track a,Represent semantic track b, | S_a| be the total length of stationary point in semantic track a,|S_b| be the total length of stationary point in semantic track b;

The similarity on two limits based on stationary point in computing semantic track a and semantic track b

Based on the distance between different edge in semantic track a and semantic track b, calculate the regularization similarity based on distance

Similarity in computing semantic track a and semantic track b between two limits:

sim_w(e_a，i，e_b，j)＝βsim_d(e_a，i，e_b，j)+(1-β)sim_s(e_a，i，e_b，j)；

Wherein, 0≤β≤1st, weight coefficient;

The first similarity of computing semantic track a and semantic track b is again:

By the timeline of semantic track b translation Δ t in each step, then calculate the of semantic track a and semantic track b after translationTwo similarities:

${\mathrm{Sim}}_{W,r}({\mathrm{STN}}_a,{\mathrm{STN}}_b)=\frac{\mathrm{\Δ}T-r\×\mathrm{\Δ}t}{\mathrm{\Δ}T}{\mathrm{Sim}}_W({\mathrm{STN}}_a,{\mathrm{STN}}_b^r);$

Wherein, Δ T=max (| e_a，1.tin-e_b，m.tout|，|e_b，1.tin-e_a，n.tout|), n is the total of limit in semantic track aNumber, m is the sum on limit in semantic track b, e_a，1.tin be through limitThe timestamp of start node, e_b，1.tin be through limitThe timestamp of start node, e_b，m.tout be through limit e_b，mThe timestamp of end node, e_a，n.tout be through limit e_a，nThe timestamp of end node, in the translation of r step the timeline of r × Δ t be labeled as

According to the first similarity and the second similarity, obtain the maximum weighted similarity of track a and track b:

2. method according to claim 1, is characterized in that:

The public region length on limit in semantic track a and semantic track b | sim (e_a，i，e_b，j) | calculating formula be:

$\left|sim(e_{a,i},e_{b,j})\right|={\Σ}_{k=1}^{count\left(S_{a,i}\right)}{\Σ}_{l=1}^{count\left(S_{b,j}\right)}|s_{a,k}\∩s_{b,j}|;$

The similarity on two limits based on stationary point in semantic track a and semantic track bCalculating formula be:

${\mathrm{sim}}_s\left(e_{a,i},e_{b,j}\right)=\mathrm{\α}\frac{\left|sim\left(e_{a,i},e_{b,j}\right)\right|}{\max \left({\Σ}_{s_{a,k}\∈S_{a,i}}\left|s_{a,k}\right|,{\Σ}_{s_{b,l}\∈S_{b,j}}\left|s_{b,l}\right|\right)}+\left(1-\mathrm{\α}\right)\left|\frac{L\left(S_{a,i}\right)\∩L\left(S_{b,j}\right)}{L\left(S_{a,i}\right)\∪L\left(S_{b,j}\right)}\right|;$

Regularization similarity based on distance in semantic track a and semantic track bCalculating formula be:

${\mathrm{sim}}_d\left(e_{a,i},e_{b,j}\right)=\frac{\max \left(\left|\right|{\mathrm{STN}}_a\left|\right|,\left|\right|{\mathrm{STN}}_b\left|\right|\right)-Dist\left(e_{a,t},e_{b,j}\right)}{\max \left\{\left|\right|{\mathrm{STN}}_a\left|\right|,\left|\right|{\mathrm{STN}}_b\left|\right|\right\}};$

Wherein, s_a，kFor k stationary point in semantic track a, s_b，lFor l stationary point in semantic track b,WithRespectively S_a，iWithUpper all funnyThe semantic label set at stationary point, 1≤α≤1st, weight parameter.

3. method according to claim 1 and 2, is characterized in that:

The public region length on limit in computing semantic track a and semantic track b | sim (e_a，i，e_b，j) |, also comprise before:

Judge whether to meet following condition:

Wherein, 1≤k≤count (S_a，i)，1≤l≤count(S_b，j)；

If meet, have public region; If do not meet, the length of public region is 0.