WO2011047728A1

WO2011047728A1 - Matching method for a digital vector map

Info

Publication number: WO2011047728A1
Application number: PCT/EP2009/063937
Authority: WO
Inventors: Heiko Mund
Original assignee: Tele Atlas B.V.
Priority date: 2009-10-22
Filing date: 2009-10-22
Publication date: 2011-04-28

Abstract

Sequentially ordered trace points (P) are matched to a line segment (w) in a generated network when: 1) their offset is less than a predetermined offset limit; 2) there are at least a minimum number (n) of consecutive trace points (P_k... P_k+n-1); 3) for trace points (P_j, P_l), the measure of m(P_j,w) is less than or equal to m(P_l,w), if k ≤ j < I < k+n; and 4) the heading angle for each trace point (P_i) is less than a predetermined maximal angle value. Alternate large (D) and small (d) offset limits are established for Condition 1), and alternate large (Φ) and small (φ) maximal angle values are established for Condition 4). Initially, the small offset limit (d) is applied for Condition 1) and the small maximal angle value (φ) for Condition 4). Thereafter, the large offset limit (D) is substituted for Condition 1) and the large (Φ) maximal angle value for Condition 4) when a trace point (P_k) fulfills Condition 1). Following this, a reversion to the small offset limit (d) is made for Condition 1) and the small maximal angle value (φ) for Condition 4) when a trace point (P_k+n) has an offset greater than the large offset limit (D).

Description

MATCHING METHOD FOR A DIGITAL VECTOR MAP

CROSS REFERENCE TO RELATED APPLICATIONS

[0001] None.

BACKGROUND OF THE INVENTION

Field of the Invention

[0002] This invention relates generally to a method for updating and extending digital vector maps using probe data in cases where references between GPS traces and an existent road network are needed, and more particularly toward a method for matching one road network to another road map for the computation of speed profiles, the computation of additional road attributes, the detection of changes in the road network, the comparison of two road networks, and the like.

Related Art

[0003] Navigation systems, electronic maps (also known as digital maps), and geographical positioning devices are increasingly used by travelers to assist with various navigation functions, such as to determine the overall position and orientation of the traveler and/or vehicle, find destinations and addresses, calculate optimal routes, and provide real-time driving guidance. Typically, the navigation system includes a small display screen or graphic user interface that portrays a network of streets as a series of line segments, including a centre line running approximately along the centre of each street or path. The traveler can then be generally located on the digital map close to or with regard to that centre line.

[0004] Digital maps are expensive to produce and update, since exhibiting and processing road information is very costly. Surveying methods or digitizing satellite images are commonly employed techniques for creating a digital map. Furthermore, digital maps are likely to contain inaccuracies or systematic errors due to faulty or inaccurate input sources or flawed inference procedures. Once a digital map has been created, it is costly to keep map information up to date, since road geometry changes over time. In some regions of the world, digital maps are not available at all.

[0005] Figures 1A-1C depict a digital vector map in the form of roads. Figure 1A represents major motorways or driving routes. Figure IB depicts the major motorways of Figure 1A plus an interconnecting network of secondary roads. Figure 1C illustrates all of the information of Figure IB together with an extended network of tertiary streets and alleys. As will be appreciated by reference to these figures, in combination with the expense and effort required to produce digital maps, it may be the case that an existing roadway map or network is incomplete in its depiction of all roadways or paths within a given region. Furthermore, due to the evolving nature of networks which may include but are not limited to roadways and paths, changes may occur over time such that an existing digital map may no longer accurately portray current conditions.

[0006] Figure 2 illustrates a fractional section of a digital vector map, in this case a bidirectional roadway supporting two-way traffic, containing junctions J and line segments W1 . . .W9. Together, they constitute a graph with several additional properties. A main trunk of the roadway is indicated at 10 and a branch road extending generally perpendicularly from the main trunk 10 is indicated at 12. It is known, for example, to take probe data inputs from low- cost positioning systems and handheld devices and mobile phones with integrated GPS functionality for the purpose of incrementally learning a map using certain clustering technologies. The input to be processed consists of recorded GPS traces in the form of a standard ASCII stream, which is supported by almost all existing GPS devices. The output is a road map in the form of a directed graph with nodes and edges annotated with travel time information. Travelers appropriately fitted with navigation devices and traversing the main trunk 10 and branch 12 junction may thus create a trace map like that shown in Figure 2, with trace points or nodes created at regular distances. The nodes and edges are stored in a digital vector map table or database.

[0007] Road geometry is inferred through this technique, and the collected data points are then refined by filtering and partitioning algorithms. Of course, there exists other cases where references between GPS traces and an existing road network are needed, such as to compute speed profiles, compute additional road attributes, detect changes in the road network, and to compare two road networks, to name a few.

[0008] For such applications, there exist already different map matching algorithms. These algorithms can be categorized as on-line or off-line. For on-line algorithms, only the current and the previous GPS points are available. On the other hand, off-line algorithms can use additionally some or even all future GPS points. It may also be beneficial to categorize the prior art map matching algorithms as complete or incomplete. A complete map matching allocates each trace point to any line segment. With this approach it is possible that a trace point is far away from the matched line segment. Therefore one has to assure that the digital vector map is complete. If this is not the case one has to allow that a trace point is not allocated to a line segment in any case. Algorithms which allow unmatched points can be categorized as incomplete map matching. For different classes of algorithms exists different map matching methods.

[0009] For the purpose of the generation, improving or extension of digital maps using GPS traces or other probe data only incomplete map matchings are useful. This is because the considered digital map is not just yet complete.

[0010] The «-points matching algorithm (whereby n is a natural number > 2) is one example of an incomplete, off-line algorithm. This particular algorithm considers not only the position of a line point related to the network, but also the match status of the neighboring previous and future points. A description of the «-points matching algorithm follows a necessary description of the algorithm terms, which may be best understood by reference to Figures 3 and 4. In Figure 3, a digital vector map contains junctions 14 and line segments W1. . .W5. Together, they constitute a graph with several additional properties. The junctions 14 are the nodes and the line segments W1 . . .W5 are the edges of the graph. For a unidirectional map the graph is directed and for a bidirectional map it is undirected. Every line segment Wi connects two junctions 14. On the contrary, in each junction 14 meets just one or least three line segments Wi. (Only in exceptional cases will just two line segments meet in a junction.) The junctions 14 and the line segments W\ are usually associated with several attributes, including for example weight value, measure and heading.

[0011] The geometry of a line segment w is often described as a polygonal chain (also called polygonal curve, polygonal path, or piecewise linear curve). Alternatively one can also use other curves like splines, circle segments or clothoids. However because each curve can be sufficiently accurately approximated through a polygonal chain, usually polygonal chains are used. The vertices or nodes of a polygonal chain are called shape points 18 because they define the shape of the curve. Of course, it is possible to change a shape point to a junction under appropriate circumstances, for example if an attribute changes.

[0012] Figure 4 depicts the simple digital vector map in which a generated network is represented by a single line segment w defined by a polygonal chain having nodes N1 -N8, some or all of which may be classified as shape points. In this example, the points Nl and N8 would represent the junctions 14 of the line segment w. Adjacent the line segment w is a trace line collected by trace data. The trace line is plotted from a sequence of trace points P1 -P8. Directionality is represented by the number progression 1 , 2, 3, ... 8 associated with the points N and trace points P.

[0013] Two values are allocated to each point P1 -P8 of the trace relative to the line segment w — an offset and a measure m. (The offset can be signed +/- where the sign depends on which side of the line segment w the trace point falls. However for most purposes an unsigned offset value is acceptable.) The offset is the shortest distance of the trace point to the line segment w. The measure m is the length from the first junction or point (e.g., Nl) of the line segment w to the orthogonal projection of the trace point to the line segment w. Respectively, the measures ml , ml, mi, mA, and mA for trace points PI , P2, P3, P4, and P5 are illustrated in Figure 4. Because the orthogonally projected point for any given trace point P is in general not unique, the measure m is also in general not unique. Consider, for example, that the measures for trace points P5 and P6 may be substantially equal. For this reason, it is necessary to define a measure m by reference to both a trace point P and particular line segment w. Therefore, an accurate description of any given measure may be noted like this: m(?i,w).

[0014] According to the «-points matching algorithm, to match a point P_; of a trace line to a network element w, the following conditions are required:

(1) The offset of P_; to w has to be smaller than a given maximal offset value.

(2) There are at least n consecutive points (P^ ... P_/t+n-/; where k < i < k+n) which have also an offset to w smaller than the given maximal offset value, n is a fixed number greater or equal to 2.

(3) For the measures m(P/,w) and m(P_/,w), of the points P, and P_/ to the network element w, m(?_j,w) < m(?i,w), if k <j < l < k+n.

[0015] It is possible that more than one line segment w fulfills these conditions. In this case the "best" line segment is chosen. One simple way to determine which line segment is best is to use a greedy strategy, in which the network element is selected having the smallest offset of the first point in the sequence of the consecutive points which fulfill the conditions. If a bidirectional map matching is desired, Condition (3) above can be modified as follows: (3') For the measures m(P/,w) and m(P_/,w), of the points P, and P_/ to the network element w, m P_j,w)≤ m(Pi,w), for all j, I with k <j < l < k+n, or m(P/,w) > m(P_/,w), for all j, I with A: <j < l < k+n.

[0016] Returning to the example shown in Figure 4, trace points PI ,... , P5 of the trace line are matched to the network element or line segment w. However the trace points P6, P7 and P8 are not matched because the offset of these points to the network element w is larger than a given maximal offset value. The predetermined maximum offset value is also referred to below as an offset limit.

[0017] Sometimes a trace runs nearly parallel to a line segment w. If the offset of some trace points is lower and of some other points greater than the maximum offset value, the match status will change often. An example of this can be observed by reference to Figure 5. Only the trace points P3, P4, P5 and P8, P9, P10 have an offset value lower than a small offset limit d. If the maximum offset limit is set at d only the trace points P3, P4, P5 and P8, P9, P10 are matched to the line segment w (assuming n < 3). So one would get two portions of matched trace points which are separated by the unmatched points P6 and P7. As can be appreciated, this is an undesirable situation and results in less reliable, more difficult matching of the trace data to line segments w in a generated network.

SUMMARY OF THE INVENTION

[0018] According to a first aspect of this invention, a method is provided for map matching in any cases where references between collected trace data and an existent generated network are needed by matching a sequence of trace points (P^ ... P_/t+n-/) to a sufficiently near line segment (w) in the generated network. A generated network is provided containing at least one line segment (w) spatially associated within a geographic coordinate system, the line segment (w) having a first point (Ν;) at one end thereof. A plurality of sequentially ordered trace points (P) are collected, and then an offset is calculated for each trace point (P_;) as the shortest distance to the line segment (w). Also, a measure (m) is calculated for each trace point (P_;). The measure (m) is the length along the line segment (w) from the first point (Ν;) to an orthogonal projection of the trace point (P_;) onto the line segment (w). A plurality of sequential trace points (P^ ... Pk+n-i) are matched to the line segment (w) if the following Conditions are met: 1 ) The offset for each trace point (P_;) is less than a predetermined offset limit defined by a generally consistent orthogonal spacing from the line segment (w); and

2) The plurality of sequential trace points (P^ . . . ¥k+n-i) comprise at least a minimum number («) of consecutive trace points (P^ . . . Vk+η-ϊ, where k < i < k+n) each having an offset to the line segment (w) less than the predetermined offset limit; and

3) For any two trace points (P/, P/) in the sequence, the measure of m(P/,w) is less than or equal to m(P_/,w), if k≤j < I < k+n;

[0019] Alternate large (D) and small (d) offset limits are established for Condition 1 ). The orthogonal spacing from the line segment (w) of the large offset limit (D) is greater than the orthogonal spacing of the small offset limit (d). Initially, the small offset limit (d) is applied for Condition 1 ). Thereafter, the large offset limit (D) is substituted for Condition 1 ) as soon as a trace point (P^) is found to fulfill Condition 1 ). A reversion to the small offset limit (d) is made for Condition 1 ) as soon as a trace point (P_/t+„) is encountered having an offset greater than the large offset limit (D).

[0020] For the sake of clarity, it should be understood that point sequences P_k , . . . , P_k+n-_i (k < i < k + n) do not have to contain exactly n points. The sequence has to contain only least n points. Therefore it may be more effective in some contexts to replace the previous statement by the folio wing P_k , . . . , P_k, (k≤ i≤ k' , k'≥ k + n - 1) . It should be understood that these are essentially equivalent expressions referring to the same underlying concept.

[0021] According to another aspect of this invention, a method is provided for map matching in any cases where references between collected trace data and an existent generated network are needed by matching a sequence of trace points (P^ . . . P_/t+n-/) to a sufficiently near line segment (w) in the network. A plurality of sequential trace points (P^ . . . Pk+n-i) are matched to the line segment (w) if the following Conditions are met:

2) The plurality of sequential trace points (P^ . . . ¥k+n-i) comprise at least a minimum number («) of consecutive trace points (P^ . . . Vk+n- , where k < i < k+n) each having an offset to the line segment (w) less than the predetermined offset limit; and

4) The angle between the headings of Ρ_ί and the projected point of Ρ_ί on w is less than a predetermined maximal angle value. [0022] Alternate large (Φ) and small (φ) maximal angle values are established for Condition 4). Initially, the small maximal angle value (φ) is used for Condition 4). Thereafter, the large (Φ) maximal angle value is substituted for Condition 4) once a trace point (P^) has fulfilled Condition 4). Following this, a reversion to the small maximal angle value (φ) is made for Condition 4) after a trace point (P_/t₊„) is encountered having a heading angle (30) greater than the large maximal angle value (Φ).

[0023] The subject invention proposes several possibilities to improve the «-points matching algorithm described above. One is to implement a kind of hysteresis effect as stated above. Another is to use the heading to the next trace point and from respectively previous trace point as stated above. Of course it is possible to use both improvements together.

BRIEF DESCRIPTION OF THE DRAWINGS

[0024] These and other features and advantages of the present invention will become more readily appreciated when considered in connection with the following detailed description and appended drawings, wherein:

[0025] Figures 1A-C depict a roadway network as representative of one form of digital vector map wherein Figure 1A shows the major roadways, Figure IB shows the interconnecting network and Figure 1C depicts a street network;

[0026] Figure 2 is a fragmentary view of a road map having a main trunk intersected at a junction by a branch road;

[0027] Figure 3 is simplified depiction of a digital vector map showing its junctions, line segments, and shape points;

[0028] Figure 4 is an illustration of the manner in which a trace line resulting from probe data is initially set in a coordinate system alongside the preexisting line segment in a digital vector map;

[0029] Figure 5 is an illustration of a trace line set alongside a line segment in a digital vector map and depicting the offset distances and measures for several trace points relative to the line segment, showing further that with a maximal offset value d only the trace points P3, P4, P5 and P8, P9, P10 will match to the line segment w and that a gap at trace points P6 and P7 can be avoided by using two different maximal offset values or limits d and D; and

[0030] Figure 6 is a simplified view of a trace line providing a definition of the heading angle. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

[0031] The following paragraphs describe the algorithm of this invention with hysteresis by means of Figure 5. In order to avoid frequent changes of the match status, two different maximal offset values or limits are used, a lower value d and a greater value D. At the start the first point of the trace is considered. Successively, the conditions for the next trace points P2, P3... are considered. At first we use the lower value d as the offset limit or maximal offset value. As soon as a trace point is encountered having an offset smaller than d, the offset limit is reset to the greater value D. This situation continues then until a trace point P12 is encountered which has an offset greater than D, at which time the offset limit reverts to the lower value d.

[0032] In the example of Figure 5, the trace points PI and P2 are unmatched because the offset is greater than the initial limit d. The offset of trace point P3 is smaller than d, therefore it fulfills the Condition (1) and for the following trace point P2 the maximal offset limit is reset to D. The offset of the points P4 and P5 is smaller than D and therefore they fulfill Condition (1) too. If n = 3 then trace points P3, P4, P5 fulfill also the Conditions (1) and (3) and therefore they are matched to the line segment w. The trace points P6,..., Pl l also have an offset smaller than D. They also fulfill the Conditions (1) - (3) and therefore they are matched to the line segment w.

[0033] The offset of trace point P12 is greater than D and it is not matched to the line segment w because it does not satisfy Condition (1). For the following points P13,... , P15 the maximum offset limit reverts to d. These points are likewise unmatched because they do not fulfill Condition (1). So the trace points P3,... , PI 1 are the only matched points.

[0034] A lack of symmetry can be observed in this example. Point P2 and point PI 1 are both greater than d and smaller than D. However trace point P2 is unmatched because the offset limit for this point was d whereas point PI 1 is matched because the offset limit for this point was D. In order to improve the symmetry of these results, therefore, we modify the hysteresis.

[0035] This modification is accomplished by identifying transition pairs 20, 22 within the plurality of sequentially ordered trace points P. Transition pairs 20, 22 are defined as any two sequential trace points P^;, P^ and Vk_+n- P_/t+n in which one trace point P^ or Vk_+n-i fulfills Condition 1) and the other trace point

respectively, does not fulfill Condition 1). Furthermore, it is helpful to designate between leading 20 and trailing 22 transition pairs. A leading transition pair 20 comprises two sequential trace points Y_k-i, in which the offset limit for the first trace point P^/, is the small offset limit d and the predetermined offset limit for the second trace point P^ is the large offset limit D. A trailing transition pair 22, by contrast, comprises two sequential trace points V_k+n- P_/t_+n in which the offset limit for the first trace point V_k+n-i, is the large offset limit D and the offset limit for the second trace point P^_+n is the small offset limit d.

[0036] Upon reaching a trailing transition pair 22, i.e., a trace point where the current offset limit changes from D to d, we go back step by step. For each trace point we test the conditions again with the alternate maximal offset limit d until reaching a trace point with an offset smaller than d.

[0037] Referring still to the example of Figure 5, at trace point P12 the maximal offset value or limit changes from D to d. Upon reaching the trace point P12 attention is drawn one step back to trace point PI 1. There again the condition (1) is re -tested with the alternate offset limit d, and it is discovered that the offset of Pl l is greater than the reset maximal offset limit d. Hence the Condition (1) is not fulfilled and trace point Pl l is removed from the sequence of the matched points. With trace point PI 1 removed, consideration is then given to the trace point P10 because the trailing transition pair 22 has now changed to points P10 and Pl l . Because the offset of P10 is smaller than d, the Condition (1) holds. Of course the Conditions (1) - (3) are still fulfilled. As a consequence, only trace points P3,..., P10 are matched to the line segment w. Notice that if the trace line w were to end with point Pl l , then a trailing transition pair 22 is not declared because a point is not identified where the maximal offset value changes from D to d. In such a case, one would not go back and remove the point Pl l , and the trace point Pl l would remain in the list of the matched points.

[0038] It is meaningful to have the same effect in the beginning of the trace, at any instance of a leading transition pair 20. So if the trace point PI would not be in the trace it would be meaningful to add the trace point P2 to the list of the matched points. Therefore only a small modification is necessary. In the example of Figure 5, P2 and P3 represent the leading transition pair 20. After the first trace point PI , the offset limit is changed to d because its offset is greater than D. After this the algorithm works like above. So the trace point P2 is signed as "unmatched". If however the trace point PI is not contained in the trace, for trace point P2 we use the initial maximal offset value D. Hence, if PI did not exist, then trace point P2 would fulfill the Condition (1). Together with the following points it fulfils also the conditions (1) and (3). Thus point P2 would be matched to the trace line under those circumstances where PI does not exist.

[0039] However, using this modified technique, if all points of a trace have an offset between d and D, then the whole trace would be matched to the line segment w although no point has an offset lower than d. If this is an undesirable situation, it would be necessary to test for it after the fact.

[0040] This modification of the hysteresis can be described in generic terms as follows. In a trailing transition pair 22, the first trace point Vk+n-i is excluded if the trace point Vk+n-₂ immediately preceding the first trace point Vk+n-i has an offset greater than the small offset limit d. This step of excluding the first trace point in the trailing transition pair 22 with a new trailing transition pair is then repeated with the new trailing transition pair over and over again as needed. The application is mirrored for leading transition pairs 20. In that case, the first trace point Vk-i in a leading transition pair is added to the matched plurality of sequential trace points if the trace point Yk-₂ immediately preceding the first trace point P^/ has an offset smaller than the large offset limit D. As with the trailing transition pair, if this procedure results in a change in the leading transition pair, the steps are repeated with the new points until such time as one of the Conditions fails.

[0041] An example is provided to describe the subject map matching algorithm with respect to Figure 6. According to this example, the «-points map matching algorithm comprises four steps: A) Determination of a candidate list which have to fulfill the condition (1), B) Verifying of the conditions (2) and (3), C) Selection of the matched road lines using a greedy strategy, and D) Backtracking (optional). Figure 6 shows a digital vector map which contains four line segments LI ,..., L4. Remark that L4 is not connected with the other three line segments. There is also a trace line which contains 8 trace points PI ,..., P8. In order to match this trace line to the digital map a list of matching candidates is determined. For each trace point PI ,..., P8 we detect all lines with an offset smaller than d respectively D. That means that they fulfill the condition (1).

Table 1 : List of the matching candidates.

[0042] Note that for P4-L4, first the higher value D is used. However the next point P5 has an offset to line L4 greater than D. Therefore we go back to point P4 and test the condition with the lower maximal offset value d. For P5-L1 , first the higher value D is used. However the next point P6 has an offset to line LI greater than D. Therefore we go back to point P5 and test the condition with the lower maximal offset value d. For P5-L2, First the higher value D is used. However the next point P6 has an offset to line L2 greater than D. Therefore we go back to point P5 and test the condition with the lower maximal offset value d.

[0043] Of course if one does not want to use the hysteresis, one has only to set d=D. In the same manner one can use the heading of the trace points. A short overview over the matching candidates is shown in Table 2.

Table 2: Overview over the matching candidates

[0044] In the next step, conditions (2) and (3) are verified. If n=3 then only LI , L4 and L3 fulfill the condition (2). We assume that all lines fulfill the condition (3). So we have to remove only L2 from Table 2, as shown in Table 3.

Table 3: Overview over the matching candidates after removing line L2

[0045] How one can see in Table 3 there are trace points which have still more than one matching candidates which fulfill all the conditions ( 1) - (3). In order to select one of the matching candidates we use a greedy strategy. We start with P 1. There is only one matching candidate. So we choose the line segment LI . If there is more than one candidate we choose the candidate with the smallest offset value. After the choosing of LI we remain on line LI so long as possible. So we choose LI also for the trace points P2, P3, P4 and P5. Because P6 don't match to line LI we have to choose a new matching line. Therefore we choose for P6 the line segment L3. With the greedy strategy we remain also for the following trace points P7 and P8 on the line segment L3. So we get the following allocation to the line segments presented in Table 4.

Table 4: Matching result with the greedy strategy.

[0046] The greedy strategy is very suitable to avoid frequently changing of the matched line. In this example, the greedy strategy is useful to avoid matchings like L1 -L1 -L4-L1. Sometimes, however, the greedy strategy is too conservative. That means it remains too long on the same line than is appropriate under the circumstances. As can be seen in the example in Figure 6, the trace point P5 should better match to line L3 than to line LI . To avoid this occasional error an additional backtracking step can be used. According to this backtracking step, if point P6 is reached it is discovered that one must change the line segment from LI to L3. At this point, it is appropriate to go back to the previous trace points in order to check the matched lines again. [0047] In the example at point P6, we go back to trace point P5 and compare the offset d₅ of

P5 to L3 with the offset d\ to LI . Because d₅ < d\ the matched line of P5 is changed from LI to L3. Then one goes back to point P4. However the offset value d of P4 to LI is smaller than the offset value d of P4 to L3. Therefore the matched line is not changed and one goes forward to point P7. After the backtracking the following matching result is obtained, as shown in Table 5.

Table 5: Matching result after the backtracking step.

[0048] Figures 7A-C and 8 describe a variant of the subject algorithm using of the heading. The heading (also called course) is the direction in which a vehicle drives. It is important in this context to distinguish between headings, angles and angle measures. A heading gives just a direction. Usually it is measured clockwise from the north direction (see Figures 7A and 7B). So it is also an angle measure, however always referring to a reference direction. The values are between 0° and 360° (although some use values between -180° and 180°).

[0049] The heading from one trace point to the next can be expressed as the angle measured from a specific reference datum clockwise from 0° through 360° to the line. As reference usually the north direction is selected as the datum. There are different possibilities to define the heading of a GPS trace respective to a polygonal chain. Often the direction from the current trace point P_; to the next point P_;+; is used. However this is not defined for the last trace point P_;_;. Correspondingly the direction can be used from the previous point P_;_; to the current point P_;. (Of course this is not defined for the first trace point in a trace line.) In order that the heading is independent of the sense of direction of the trace, a mean value of both current and preceding headings can be used.

[0050] In the same way, the headings of the line segment of the digital vector map are computed. For the algorithm also the headings of points of the line segment w are needed which are not shape points. For such a point the heading is simply computed from the previous shape point P_; to the next shape point P_;+;. Alternatively one can use the weighted mean of the headings of the previous shape point and the next shape point. This will avoid discontinuities of the headings.

[0051] Additionally to the measure m and the offset we compute for every trace point the angle 30 between the headings of the trace point and it's projection to the line segment w. If the digital vector map is a unidirectional network, the (unsigned) angle between these two headings in the range [0°, 180°] can be used. However, for a bidirectional map matching, the angle should be in the range [0°, 90°]. Therefore it would be necessary to choose the supplementary angle if the angle is greater than 90°.

[0052] As additional Condition for the matching of a point P_; of a trace P to a line segment w the requirement is established:

(4) The angle between the headings of P_; and the projected point of P_; on w has to be smaller than a given maximal angle value.

[0053] Said another way, the condition 4) is the angle between the headings of P_t and the projected point of P_l on w is less than a predetermined maximal angle value. Or even more exactly, it is the absolute value of the angle difference between the headings of P_t and the projected point of P_t on w is less than a predetermined maximal angle value.

[0054] On principle one has to distinguish between an angle and the measure of the angle although usually both are termed angle. Also usually one uses the same symbol for the angle and the angle measure. The angle itself is a geometric object consisting of two rays with the same origin. Each angle has a unique angle measure. On the other hand there are infinite many angles which have the same angle measure. Usually angle measures are between 0° and 180°. Correctly the above angles one has to term undirected angles because the order of the two rays is not of interest. The angle measures of such undirected angles are unsigned (between 0° and 180°).

[0055] There exist also directed angles where the order of the two rays is important. Such directed angles have signed measures between -180° and 180°. If one changes the order of the two rays then the sign of the angle measure changes. In the case of this invention, the headings are such directed angles. All other angles are always undirected angles. Because the first ray of the heading angle is always the north direction the second ray is unique determined (except for translations) through the measure of the heading angle. Therefore usually one uses the measure of the heading angle. This measure of the heading angle is called shortly heading. On the other hand also the second ray itself is called heading.

[0056] In condition (4), the angle between the heading of P_l and the heading of the projected point of Ρ_ί is of concern. Because Ρ_ί and the projected point of Ρ_ί are not identical, both headings have different origins. Therefore no angle exists between these two headings (rays). Nevertheless it is possible to compute the angle measure between these two headings. Therefore one may translate the headings so that both have the same origin (see Figure 7C). Then it is convenient to measure the angle. The angle measure is independent to which point one translates the angles. One can easily compute this angle measure as the absolute value of the difference of both headings (measures). In this way it is possible to compute the angle measure between two headings if a unidirectional map matching is used. However for a bidirectional map matching it is generally considered undesirable to distinguish between the forward and backward direction of the trace. Therefore the minimum of the angle measure itself is chosen and the measure of the adjacent angle (see Figure 8). In this case the angle measure is always between 0° and 90°. This is the only difference between uni-directional and bi-directional map matching apart from the condition (3) respectively (3').

[0057] To use the headings together with the hysteresis effect the two different maximal angle values are set as small φ and large Φ. In other words, the angle measured from an angle bisector to the line segment w is greater for the large Φ maximal angle value than for the small φ maximal angle value. Between these small φ and large Φ values it is possible to switch in the same manner described above for the offset limits d and D. If the maximal values are D and Φ, at the appropriate transition zones 20, 22 the values will change to d and φ if both Conditions (1) and (4) failed. On the contrary if the maximal values are d and φ the values change only to D and Φ if both Conditions (1) and (4) are fulfilled. Advantageously, in cases where headings are used, it is possible to set n=\ . In this case the Conditions (1) and (3) are always fulfilled. So it is possible to disregard them altogether.

[0058] In situations where it is desirable to apply both the heading and hysteresis effect components of this invention, Condition 1) and Condition 4) parameters are applied in tandem. Thus, when a switch is made from the initial large offset limit D and large maximal angle value Φ , both the offset limit and maximal angle values are changed simultaneously to d and φ respectively if one of the conditions (1) or (4) failed. On the contrary if the maximal values are d and φ the values change only to D and Φ if both conditions (1) and (4) are fulfilled.

[0059] The foregoing invention has been described in accordance with the relevant legal standards, thus the description is exemplary rather than limiting in nature. Variations and modifications to the disclosed embodiment may become apparent to those skilled in the art and fall within the scope of the invention. Accordingly the scope of legal protection afforded this invention can only be determined by studying the following claims.

Claims

claimed is:

A method for matching a sequence of collected trace data points (P_l , ... , P_r ) to a digital vector map, said method comprising the steps of:

(i) providing a digital vector map {M) containing a certain number of line segments w_l,...,w_s e M spatially associated within a geographic coordinate system;

(ii) collecting a plurality of sequentially ordered trace points (P_l , ... , P_r ) ;

(iii) determination for each trace point P_t, i e {l,...,r} a subset . c M of line segments which have a distance to the trace point P_t smaller than or equal a predetermined offset limit d;

(iv) calculating an offset d for each trace point P_t, i e {l,...,r} to each line segment w e . as the shortest distance between the trace point P_i and the line segment w;

(v) determination of subsets W_{ c ., i e {l,...,r}, where PFJ contains all w e . for which exists a plurality of least a minimum number n of consecutive trace points P_k,...,P_k, (k<i<k',k'≥k + n-l) so that we _y. for each k< j<k;

(vi) selection of one line segment w W_j for each z^'e{l,...,r} as matched line segment for P_{; if W_{ is empty for any i e {l,...,r} the trace point is classified as "unmatched".

The method of Claim 1 further including the step of

(iv*) calculating a measure /;/ :^* for each trace point P_t, i e {l,...,r} and each line segment w e . as the length along the line segment w from the first point of w to an orthogonal projection of the trace point P_! onto the line segment w.

The method of Claims 1 and 2 further including the step of

(v*) determination of subsets V_i c W_{, i e {l,...,r}, where V_i contains all w e W_{ for which for all consecutive trace points P_k,...,P_k, (k<i<k') with weW_h for each k<h<k' the following condition holds: m <m for all j, I with k < j <l < k'; after this step V_x,...,V_r will renamed into W_x,...,W_r.

The method of Claims 1 and 2 further including the step of

(v**) determination of subsets V_i czW^i e {l,...,r}, where V_i contains all w e W_{ for which for all consecutive trace points P_k,...,P_k, (k<i<k') with weW_h for each k<h<k' the following condition holds: m <m for all j, I with k≤ j < I≤ k' or m ≥ η for all j, I with k < j <l<k; after this step V_x,...,V_r will renamed into W_x,...,W_r. The method of any of Claims 1-4 further including the steps of

(iv') calculating an absolute angle for each trace point P_{, i e {l,...,r} to each line segment w e . as the absolute value of the angle difference between the heading of the trace point P_! and the heading of the line segment w at the position of the orthogonal projection of the trace point P_! onto the line segment w;

(ν') removing all line segments w e . , i e {l, ... , r) with φ™ > φ from . , whereas φ is a predetermined maximal angle value.

The method of any of Claims 1-4 further including the steps of

(iv') calculating an absolute angle φ™ for each trace point P_t, i e {l,...,r} to each line segment w e . as the absolute value of the angle difference between the heading of the trace point P_! and the heading of the line segment w at the position of the orthogonal projection of the trace point P_! onto the line segment w;

(ν') removing all line segments w e M i e {l,...,r} with f(d ,(p )> f(d,(p) from . , whereas / is a certain function of the offset and the absolute angle, d is a predetermined offset limit and φ is a predetermined maximal angle value.

The method of any of Claims 1 -6 whereas the step of the selection of one line segment w W_j for each i e {l, ... , r) as matched line segment for P_! is based on a greedy strategy which contains the following steps of:

(vii) starting with the first trace point P_l all trace points (P_l P_r ) will sequentially processed;

(viii) if W_l is empty the trace point P_! is classified as "unmatched";

(ix) if P_j = P_l the first trace point or P^ is classified as "unmatched" then for P_! the matched line segment is the line segment w e PFJ with the lowest offset value d ;

(x) if P_i__l,i>\ is matched to a line segment weW^ and wiW_t then for P_i the matched line segment is the line segment w e PFJ with the lowest offset value d ;

(xi) if P_i__l,i>\ is matched to a line segment weW^ and weW_j then for P_i the matched line segment is also the line segment w eW

The method of Claim 7 further including the step of backtracking:

(xii) starting with the last trace point P_r all trace points (Ρ^.,.,Ρ^ will sequentially processed in reverse order;

(xiii) if \<i <r and w e W_M is matched to P_M and w is matched to Ρ_ί and w eW_j and d < d then w is matched to Ρ_ί instead w eW_r

9. The method of any of Claims 1 -6 whereas the step of the selection of one line segment w W_j for each i e {l, ... , r) as matched line segment for P_{ is based on a greedy strategy which contains the following steps of:

(vii*) starting with the first trace point P_l all trace points (P_l P_r ) will sequentially processed;

(viii*) if W_j is empty the trace point P_i is classified as "unmatched";

(ix*) if P_t = P_l the first trace point or Ρ_ί is classified as "unmatched" then for P_t the matched line segment is the line segment w eW_t with the lowest absolute angle

(x*) if P_l__l,i>l is matched to a line segment w W_l__l and wiW_t then for P_l the matched line segment is the line segment w W_j with the lowest absolute angle

(xi*) if P₁__l,i>\ is matched to a line segment weW^ and w W_j then for P_! the matched line segment is also the line segment w W_j.

10. The method of Claim 9 further including the step of backtracking:

(xii*) starting with the last trace point P_r all trace points (Ρ^.,.,Ρ^ will sequentially processed in reverse order;

(xiii*) if \<i <r and w e W_M is matched to P_M and w is matched to Ρ_ί and w eW_j and φ™ < φ™ then w is matched to Ρ_ί instead w eW_r

11. The method of any of Claims 1 -6 whereas the step of the selection of one line segment w W_j for each i e {l, ... , r) as matched line segment for P_{ is based on a greedy strategy and contains the following steps of:

(vii**) starting with the first trace point P_l all trace points (P_l P_r ) will sequentially processed;

(viii**) if W_j is empty the trace point P_i is classified as "unmatched";

(ix**) if P_t = P_l the first trace point or Ρ_ί is classified as "unmatched" then for P_t the matched line segment is the line segment weW_j with the lowest value f(d ,φ ), whereas /is a certain function of the offset d and the absolute angle

(x**) if P_{i X},i>\ is matched to a line segment weW^ and wiW_t then for P_i the matched line segment is the line segment weW_j with the lowest value f(d?,<P?);

(xi**) if P_{i X},i>\ is matched to a line segment weW^ and weW_j then for P_i the matched line segment is also the line segment w eW

12. The method of Claim 11 further including the step of backtracking:

(xii**) starting with the last trace point P_r all trace points (Ρ^.,.,Ρ^ will sequentially processed in reverse order;

(xiii**) if \ <i <r and w e W_M is matched to P_M and w is matched to Ρ_ί and w e W_t and f(d ,q> )< f(d? ,q>?) then w e W_t is matched to P_t instead w≡W_t.

13. A method for matching a sequence of collected trace data points (P_l,...,P_r) to a digital vector map, said method comprising the steps of:

(iii) predetermination of a smaller offset limit d and a greater offset limit D;

(iv) determination for each trace point P_t, i e {l,...,r} a subset . c of line segments which have a distance to the trace point P_t smaller than or equal the greater offset limit D;

(v) calculating an offset d™ for each trace point P_t, i e {l,...,r} to each line segment w e . as the shortest distance between the trace point P_! and the line segment w;

(vi) determination of subsets W_{ c ., / e {l,...,r}, where Ρίζ. contains all w e . for which exists a plurality of least a minimum number « of consecutive trace points P_k,...,P_k. {k<i<k k^,≥k + n-\) so that weM_j for each k < j < k' and

(vii) selection of one line segment w W_j for each z^'e{l,...,r} as matched line segment for P_{; if Ρίζ. is empty for any i e {l,...,r} the trace point P_{ is classified as "unmatched".

14. The method of Claim 13 further including the step of

(v*) calculating a measure m™ for each trace point P_t, i e {l,...,r} and each line segment w e M_l as the length along the line segment w from the first point of w to an orthogonal projection of the trace point P_t onto the line segment w.

15. The method of Claims 13 and 14 further including the step of

(vi*) determination of subsets V_i c W_{, i e {l,...,r}, where V_i contains all w e W_{ for which for all consecutive trace points P_k,...,P_k, (k≤i<k') with weW_h for each k<h<k' the following condition holds: m <m for all j, I with k≤ j <l < k'; after this step V_l,...,V_r will renamed into W_x,...,W_r.

16. The method of Claims 13 and 14 further including the step of

(vi**) determination of subsets V_i czW^i e {l,...,r}, where V_i contains all w e W_{ for which for all consecutive trace points P_k,...,P_k, (k<i<k') with weW_h for each k<h<k the following condition holds: mj≤m™ for all j, I with k≤ j < I≤ k' or m ≥ η for all j, I with k < j <l<k; after this step V_x,...,V_r will renamed into W_x,...,W_r.

17. The method of any of Claims 13-16 further including the steps of

(iii') predetermination of a smaller angle limit φ and a greater angle limit Φ ;

(ν') calculating an absolute angle φ™ for each trace point P_t, i e {l,...,r} to each line segment w e . as the absolute value of the angle difference between the heading of the trace point P_t and the heading of the line segment w at the position of the orthogonal projection of the trace point P_t onto the line segment w;

(vi') determination of subsets £/. c ., i e {l,...,r}, where U_{ contains all w e . for which exists a plurality of least a minimum number n of consecutive trace points P_k,...,P_k. (k<i<k',k'>k + n-\) so that < for each k < j < k' and φ - Ψ ^and φ - Ψ ; ^aft^er this ^sleP U_l,...,U_r will renamed into M_x,..., M_r .

18. The method of any of Claims 13-16 further including the steps of

(iii") predetermination of a smaller angle limit φ and a greater angle limit Φ ;

(v") calculating an absolute angle for each trace point P_{, i e {l,...,r} to each line segment w e . as the absolute value of the angle difference between the heading of the trace point P_! and the heading of the line segment w at the position of the orthogonal projection of the trace point P_! onto the line segment w; (vi") determination of subsets U_{ c ., i e {l,...,r}, where U_{ contains all w e . for which exists a plurality of least a minimum number n of consecutive trace points P_k,...,P_k, (k<i<k',k'≥k + n-l) so th for each k<j<K and f(d_k ,<¾^w)< f{d,(p) and

reas / is a certain function of the offset and the absolute angle; after this step U_l,...,U_r will renamed into M_x,..., M_r .

19. The method of any of Claims 13 - 18 whereas the step of the selection of one line segment w W_t for each i e {l,... , r} as matched line segment for P_l is based on a greedy strategy which contains the following steps of:

(viii) starting with the first trace point P_l all trace points (P_l P_r ) will sequentially processed;

(ix) if W_j is empty the trace point P_i is classified as "unmatched";

(x) if P_t = P_l the first trace point or Ρ_ί is classified as "unmatched" then for P_t the matched line segment is the line segment w e W_l with the lowest offset value d™ ;

(xi) if Ρ_ί , z > 1 is matched to a line segment w e W_l__l and w i W_t then for P_l the matched line segment is the line segment w e W_l with the lowest offset value d™ ;

(xii) if Ρ_ί , z > 1 is matched to a line segment w e W_t__x and w e W_l then for P_l the matched line segment is also the line segment w W_j .

20. The method of Claim 19 further including the step of backtracking:

(xiii) starting with the last trace point P_r all trace points (Ρ^. , . ,Ρ^ will sequentially processed in reverse order;

(xiv) if 1 < i < r and w e W_M is matched to P_M and w is matched to Ρ_ί and w e W_j and d < d then w e PFJ is matched to P_t instead w e W_r

21. The method of any of Claims 13 - 18 whereas the step of the selection of one line segment w W_t for each i e {l,... , r} as matched line segment for P_{ is based on a greedy strategy which contains the following steps of:

(viii*) starting with the first trace point P_l all trace points (P_l P_r ) will sequentially processed;

(ix*) if W_j is empty the trace point P_i is classified as "unmatched";

(x*) if P_t = P_l the first trace point or Ρ_ί is classified as "unmatched" then for P_t the matched line segment is the line segment w e PFJ with the lowest absolute angle

(xi*) if P_{i X} , i > \ is matched to a line segment w e W_t__x and w i W_t then for P_i the matched line segment is the line segment w with the lowest absolute angle

(xii*) if P_j_i , i > \ is matched to a line segment w e W_t__x and w <≡W_t then for P_i the matched line segment is also the line segment w e W

22. The method of Claim 21 further including the step of backtracking:

(xiii*) starting with the last trace point P_r all trace points (Ρ^.,.,Ρ^ will sequentially processed in reverse order;

(xiv*) if 1 < i < r and w e W_M is matched to P_M and w is matched to Ρ_ί and w eW_j and φ™ < φ™ then w eW_t is matched to P_t instead w eW_r

23. The method of any of Claims 13 - 18 whereas the step of the selection of one line segment w W_t for each i e {l,...,r} as matched line segment for P_{ is based on a greedy strategy and contains the following steps of:

(viii**) starting with the first trace point P_l all trace points (P_l,...,P_r) will sequentially processed;

(ix**) if W_j is empty the trace point P_i is classified as "unmatched";

(x**) if P_t = P_l the first trace point or Ρ_ί is classified as "unmatched" then for P_t the matched line segment is the line segment weW_j with the lowest value f{d ,φ ), whereas /is a certain function of the offset d and the absolute angle

(xi**) if P_{i X}, i> \ is matched to a line segment w e W_t__x and w <£ W_t then for P_i the matched line segment is the line segment weW_j with the lowest value f(d?,<P?);

(xii**) if P_{i X}, i> \ is matched to a line segment w e Ρίζ.^ and w then for ^. the matched line segment is also the line segment w eW

24. The method of Claim 23 further including the step of backtracking:

(xiii**) starting with the last trace point P_r all trace points (Ρ^.,.,Ρ^ will sequentially processed in reverse order;

(xiv**) if \ <i<r and weW_M is matched to and w&W_t is matched to ^. and w e W_t and f(d ,q> )< f(d? ,q>?) then w e ^ is matched to P_t instead w≡W_t.