CN114705200A

CN114705200A - High-sampling-rate track map matching method based on path increment

Info

Publication number: CN114705200A
Application number: CN202210362773.9A
Authority: CN
Inventors: 刘远刚; 王浩岩; 李少华; 龙颖波; 蔡永香; 马潇雅
Original assignee: Yangtze University
Current assignee: Yangtze University
Priority date: 2022-04-07
Filing date: 2022-04-07
Publication date: 2022-07-05

Abstract

The invention discloses a high sampling rate track map matching method based on path increment, which formally defines a GNSS track point map matching process based on the path increment; extracting a candidate road subset from a road network by adopting a combined filtering method based on constraint conditions such as distance, road connectivity, direction consistency and the like; and taking the path between two adjacent intersection points in the candidate sub-network as an increment, and performing track point matching calculation based on the path increment. In the matching process, each incremental calculation process is a process of determining a subsequent matching road section by taking the intersection point as a starting point. In order to weaken the influence of parallel equidirectional road sections on the matching result, the similarity evaluation index combining the Hausdorff distance and the curvature is comprehensively considered at each intersection point. The method has higher accuracy and efficiency, can better process various complex road conditions, and is very suitable for high-sampling-rate GNSS track matching application scenes in the road network of the complex city.

Description

High-sampling-rate track map matching method based on path increment

Technical Field

The invention relates to a track map matching method, in particular to a high sampling rate track map matching method based on path increment.

Background

The progress of position acquisition and mobile computing technology generates a large amount of spatial trajectory data, which represents the mobility of various moving objects and plays a crucial role in various fields such as real-time path planning, road network updating, travel rule discovery and the like. However, different mobile devices have positioning errors, and the road is converted into a linear element under an abstract plane from a planar element in the real world through a map synthesis process, so that a deviation exists between the track and the road where the track is actually located. Therefore, before processing and analyzing the trajectory data, a map matching should be performed so that the trajectory is correctly positioned over the road. In urban application scenarios, the advance of Global positioning System (GNSS) technology and the development of big data processing technology enable the positioning data of urban areas to be acquired at very short time intervals, for example, 1 second or 5 seconds. How to satisfy the matching between the high sampling rate track and the complex urban road network becomes a challenge under the condition of considering both the efficiency and the accuracy.

Existing map matching algorithms can be divided into three categories: local algorithms, incremental algorithms, and global algorithms. The local map matching algorithm only considers a single GNSS point on the track at a time, and does not consider the relationship between the current track point and the front and rear points. Such methods have the advantages of high efficiency and easy implementation, but are very prone to matching errors in practical applications. The incremental map matching algorithm considers adjacent track points, introduces the spatial distribution and the movement rule of the track points such as topology, motion state, transition probability and the like into the matching process, and accordingly improves the performance of the algorithm. However, in the incremental matching algorithm, incorrect matching results of previous points may accumulate and affect matching of subsequent points. The global map matching algorithm maps the whole trajectory to paths in the road network based on similarity measures, and since the GNSS points on the trajectory are considered in batches, the error propagation problem of the incremental method is not generated. Therefore, the algorithm is not sensitive to the sampling rate, and ensures that the matching precision is higher under different sampling rates, but also causes the problem of low efficiency under the condition of high sampling rate.

According to the characteristics of the algorithms, when the application scene of the high sampling rate and the complex urban road network is oriented, the matching idea based on the increment is more widely applied, but the existing related algorithm still has the following two problems when processing the high sampling rate track: (1) the algorithm matching time is not only related to the number of track points, but also closely related to the complexity of a vehicle driving road section, and a more time-consuming calculation process is brought by a higher sampling rate and road complexity; (2) at complex road sections such as intersection points, since moving objects generally travel at a low speed, the distance between the trace points is made smaller relative to other regions, and thus mismatching is also more likely to result.

Disclosure of Invention

In order to overcome the defects of the technology, the invention provides a high-sampling-rate track map matching method based on path increment, and the calculation efficiency and the matching accuracy at a complex road section are further improved.

In order to solve the technical problems, the invention adopts the technical scheme that: a high sampling rate track map matching method based on path increment comprises the following steps:

step S1, formally defining a GNSS track point map matching process based on path increment;

step S2, extracting candidate road subnets from the road network by adopting a multi-constraint combination filtering method;

and step S3, taking the path between two adjacent intersection points in the candidate road sub-network as an increment, and performing track point matching calculation based on the path increment.

Preferably, in step S1, the GNSS track point map matching process based on the path increment is defined as: and screening out candidate road subnetworks G ' (V ', E ') according to the track T and the road network G (V, E), and then performing track matching calculation by taking the paths between two adjacent road junction points in the candidate road subnetworks as increments, thereby gradually determining the process of the actual motion path of the object.

Preferably, the specific process of step S2 is: and combining the spatial distribution and movement rules of two adjacent track points with the characteristics of the roads nearby the two track points, and eliminating a large number of irrelevant road sections by adopting a series of filtering conditions of distance constraint, road connectivity constraint, direction consistency constraint and single-point communication road section pruning to obtain a candidate road sub-network G ' (V ', E ').

Preferably, in step S2, the distance constraint is implemented by dynamically adjusting the buffer area, i.e. taking the track point as the center of circle and r as the center of circle_p0Constructing a buffer area for the initial radius, and if the candidate road section set is empty, enlarging the radius by a fixed step length r_p，stepAnd circulating the steps until the candidate road section set at least comprises one road section.

Preferably, in step S2, the road connectivity constraint includes a topological connectivity constraint and a direction connectivity constraint, where the topological connectivity constraint: for any point p in the trajectory_iIts candidate road segment set C (p)_i) In which there is a section e_iAnd candidate segment set C (p)_i+1) In a certain path section e_jThe endpoints are connected or overlapped;

directional connectivity constraints: for any point p in the trajectory_iSet of candidate links C (p)_i) In which there is a section e_iAnd candidate segment set C (p)_i+1) In a certain section e_jAre connected or overlapped, and when e_iAnd e_jWhen connecting, the two sections should meet the requirement of connecting end to end at the connecting position.

Preferably, in step S2, the directional consistency constraint includes a length ratio constraint and an azimuth angle constraint, where the length ratio constraint: if the length ratio of the distance between two continuous track points to the distance between the projection points of the two continuous track points on the candidate road section is less than 0.8 of a threshold value, rejecting the candidate road section;

and (4) azimuth angle constraint: and calculating the included angle between the track line segment and the projection line segment on the candidate path, and rejecting the candidate line segment when the included angle is more than 90 degrees.

Preferably, in the azimuth constraint, the path with the opposite direction is eliminated by calculating the difference between the vector azimuth of the track and the vector azimuth of the projection point of the track on the candidate path, wherein the vector azimuth is the true north direction and rotates clockwise to true northThe horizontal angle passed by the front vector is calculated by the following formula:

where α is the azimuth of the current vector, (x)₁，y₁)，(x₂，y₂) Respectively, the head and tail point coordinates of the vector.

Preferably, in step S2, the "pruning" of the single point connection segment means: and traversing each road section in sequence, respectively calculating the degrees of the starting point and the ending point of each road section, and deleting other single-point communication road sections except the head road section and the tail road section.

Preferably, the specific process of step S3 is:

firstly, determining an initial road section, namely selecting the road section with the highest similarity with a track line from candidate road sections of an initial track point as the initial road section;

after the initial road section is determined, entering an incremental calculation process, and determining subsequent matching road sections step by taking a road junction point as a starting point in each incremental calculation process;

when the vertex connected with the matched road section is a transition point, recording the current road section, and continuously determining the next road section;

when the top point connected with the matched road section is a road junction point, entering the next incremental calculation process; and when the single connected point appears, taking the single connected point as a track terminal point, and ending the matching process.

Preferably, in step S3, when the incremental advancing direction is determined at the intersection point, a similarity evaluation index combining the Hausdorff distance and the curvature is used to reduce the influence of the parallel equidirectional road segment on the matching result.

The invention improves the matching efficiency and simultaneously considers the processing of the complex road sections, and better solves the contradiction between the accuracy and the efficiency of the high-sampling-rate data in the complex road network matching. The method eliminates the influence of the non-communication road sections and the reverse road sections on the matching process through the combination filtering of the road network, and adopts the path increment to replace the track increment in the traditional increment matching method, so that the matching result is more consistent with the real vehicle driving route. And a combination of curvature integral and Hausdorff distance is adopted at the intersection point to replace the geometric measurement in the traditional matching algorithm so as to improve the matching accuracy of the starting road section and the intersection point.

Drawings

FIG. 1 is a flow chart of the present invention.

Fig. 2 is a schematic diagram of the filtering result of the road network in step S2.

Fig. 3 is a flowchart of the road network filtering in step S2.

Fig. 4 is a schematic diagram of the direction connectivity constraint in step S2.

Fig. 5 is a schematic diagram of the filtering of the road network by the directional consistency constraint in step S2.

Fig. 6 is a flowchart of the trajectory matching in step S3.

Fig. 7 is a pseudo code of the algorithm for determining the start road segment in step S3.

Fig. 8 is a pseudo code of the incremental matching algorithm performed in step S3.

Fig. 9 is a pseudo code of the link increment matching algorithm at the intersection point in step S3.

Fig. 10 is a schematic diagram of the matching process at the intersection point in step S3.

Fig. 11 is a schematic diagram illustrating the determination of the candidate point position in step S3.

FIG. 12 is a diagram of test trajectory data in a map according to an embodiment.

FIG. 13 is a graph showing the evaluation of the accuracy of the map matching experiment in the example and a comparison thereof.

Fig. 14 is a graph of the matching result of the test trace 1 in the example.

Fig. 15 is a graph of the matching result of the test trace 2 in the example.

Fig. 16 is a graph showing the matching result of the test trace 3 in the example.

Fig. 17 is a graph showing the matching result of the test trace 4 in the example.

FIG. 18 is a graph showing the comparison of the efficiency of each algorithm in the examples.

FIG. 19 is a diagram illustrating the number of times each algorithm matching unit is called in the embodiment.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

A high sampling rate track map matching method based on path increment as shown in fig. 1 includes the following steps:

step S1, formally defining the GNSS track point map matching process based on the path increment as: preliminarily screening out candidate road subnets G ' (V ', E ') from the road networks G (V, E) by using the track T, and then performing track matching calculation by taking the paths between two adjacent road junctions in the subnets as increments, thereby gradually determining the actual motion path of the object.

Formalized definition of a GNSS track point map matching process based on path increment specifically comprises the following steps:

the trajectory T is a set of location points acquired by the mobile device and may be denoted as T ═ p_i1., N }, where p is_iIs a track point. Road network is a digital representation of a real road system, and can be described by a directed graph G (V, E), where V is a set of vertices { V }_j1.. m., E is a set of edges { E |_k1., S }, vertex v_jThe number of associated edges is called vertex v_jThe number of degrees of (d) is denoted as d (v)_j). For arbitrary vertex v_jIf d (v)_j) 1, v_jIs a single connection point; if d (v)_j) When it is 2, call v_jIs a transition point; if d (v)_j) V is not less than 3_jIs a road junction point. The edge in G (V, E) is called a link, and path P is a set of a series of connected links, denoted as P ═ E_o，...，e_j}. The candidate track point is located at the vertex v_jSet of trajectory points in the neighborhood, denoted I (v)_j) Referred to as candidate points for short. Set of candidate road segments as and track points p_iIs represented as C (p)_i). The road sub-network G ' (V ', E ') is a road network for map matching after being filtered, i.e. the road sub-network is a road network

Satisfy the requirement of

Defining the path between two adjacent intersection points as an increment, in the invention, the track point map matching process based on the path increment is defined as: and (4) screening out road sub-networks G ' (V ', E ') according to the track T and the road networks G (V, E), then performing incremental calculation and finally determining the actual motion path of the object.

And step S2, extracting candidate road subnets from the road network by adopting a multi-constraint combination filtering method.

The map matching is carried out in a complex urban road network, and the difficulty is the matching of intersections and parallel roads. The track of the area often has a plurality of candidate road sections with similar distances and similar shapes, which easily causes mismatching. Therefore, the space distribution and the movement rule of two adjacent track points are combined with the characteristics of the connectivity, the direction and the like of the nearby roads, the candidate road sections are filtered, and a large number of irrelevant road sections are removed. As shown in fig. 2, the schematic diagram of the filtering result of the road network is shown, wherein the original road network is filtered by a series of conditions including distance constraint, road connectivity constraint, direction consistency constraint and "single-point communication road segment" pruning "to obtain candidate road subnets which are compared and matched with the track points.

Specifically, as shown in fig. 3, the multi-constraint combination filtering method for a road network includes the following steps:

s21, distance constraint is realized by a buffer area which is dynamically adjusted, namely, a track point is taken as a circle center, and r is taken as_p0Constructing a buffer area for the initial radius, and if the candidate road section set is empty, enlarging the radius by a fixed step length r_p.stepAnd circulating the steps until the candidate road section set at least comprises one road section.

S22, the road connectivity constraint comprises a topological connectivity constraint and a direction connectivity constraint.

Topological connectivity constraint: for any point p in the trajectory_iIts candidate road segment set C (p)_i) In which there is a section e_iAnd candidate segment set C (p)_i+1) In a certain section e_jThe endpoints are connected or overlapped;

directional connectivity constraints: for any point p in the trajectory_iSet of candidate road segments C (p)_i) In which there is a section e_iAnd candidate segment set C (p)_i+1) In a certain path section e_jAre connected or overlapped, and when e_iAnd e_jWhen connecting, the two sections should meet the requirement of connecting end to end at the connecting position. As shown in fig. 4(a), for a one-way traffic path, the road segments therein can be abstracted into a set of directed line segments, and if the directions of the set of line segments are consistent, the direction connectivity constraint is satisfied. Two cases of satisfying topological connectivity but not directional connectivity are reflected in fig. 4(b), and such a segment should be filtered. Since bidirectional road segments have no directional restrictions, only directional connectivity constraints when unidirectional road segments are connected to each other need to be considered.

And S23, the direction consistency constraint comprises a length ratio constraint and an azimuth angle constraint.

Length ratio constraint: and if the length ratio of the distance between two continuous track points to the distance between the projected points of the two continuous track points on the candidate road section is less than 0.8 of the threshold value, rejecting the candidate road section.

As shown in FIG. 5(a), p₁，p₂For two successive trace points, P₁，P₂And P₃Three paths containing candidate road sections are provided, and the three paths can be used as p₁，p₂Matching path of p₁，p₂The projection on the corresponding candidate road section is the actual position of the track point, and for the track with high sampling rate, the distribution of the projection and the actual road section is generally in approximate parallel relation. Based on this knowledge, before comparing the track direction with the road segment direction, the transverse road segments are firstly rejected through the length ratio of the distance between the continuous track points and the distance between the projected points of the continuous track points on the candidate road segments, and meanwhile, the condition that the track sequence is inconsistent with the sequence of the corresponding candidate road segments is avoided. When the ratio is set to be less than 0.8 according to experiments, a better removing effect can be achieved. In FIG. 4(a), the track points are oriented toward P₃Length Len after projection₃And the distance between the trace points is far smaller, so that the path can be eliminated.

As shown in fig. 5(b), after the length ratio constraint filtering is completed, the remaining paths are two paths with opposite directions, and the path with opposite directions can be eliminated by calculating the difference between the vector azimuth angles between the trajectory and the projection points of the trajectory on the path, where the vector azimuth angle is the horizontal angle through which the due north rotates clockwise to the current vector, and the calculation formula is:

The specific implementation flow of the directional filtering is as follows: firstly, respectively obtaining track points p_i，p_i+1On the path P_jA projected point thereon; respectively calculating based on the directions of the track and the path

Is in azimuth of₀And azimuth of its projection point

③ when

When, path P_jIs removed. As shown in FIG. 5(b), respectively calculate

Can reject the reverse path P₁Fig. 5(c) shows the result after the culling, and only the path in the same direction as the track is reserved.

S24, the "pruning" of the single-point communication link means: and traversing each road section in sequence, respectively calculating the degrees of the starting point and the ending point of each road section, and deleting other single-point communication road sections except the head road section and the tail road section. In the sub-network formed by the rest of the links, if the vertex degrees of other links are 1 besides the head-tail links, the links can be 'pruned' because the single-point communication links cannot be used as the intermediate links on the final matching path. In fig. 2, the road network after the constraint filtering of the connectivity and the direction consistency of the road has single-point communication road sections, that is, the road sections are "pruned", which may increase the computation load of the subsequent map matching and need to be eliminated.

And step S3, taking the path between two adjacent intersection points in the candidate road sub-network as an increment, and performing track point matching calculation based on the path increment. And determining a subsequent matching road section by taking the intersection point as a starting point in each incremental calculation process. When the incremental advancing direction is judged at the intersection point, two similarity evaluation indexes of the Hausdorff distance and the curvature are comprehensively considered, so that the influence of the parallel equidirectional road sections on the matching result is weakened.

Taking a path between two adjacent intersection points in the candidate subnet as an increment, performing track point matching calculation based on the path increment, as shown in fig. 6, which specifically includes the following steps:

s31, determining an initial road section: and determining an initial road section by adopting a 'look ahead' thought, namely selecting a road section with the highest similarity with the track line from candidate road sections of the initial track point as the initial road section. As shown in fig. 7, is an algorithmic pseudo code to determine an initial road segment.

S32, determining the next road section: after the initial road segment is determined, an incremental calculation process is entered. As shown in fig. 8, each incremental calculation process in the algorithm takes the intersection point as a starting point to determine a subsequent matching road segment. When the connection vertex of the matched road section is a transition point, recording the current road section, and continuously determining the next road section; when the top point connected with the matched road section is a road junction point, entering the next incremental calculation process; and ending the matching process until the single connected point appears. In the matching process, the similarity evaluation of the road sections and the track lines at each road junction point adopts a similarity evaluation index combining the Hausdorff distance and the curvature.

S33, outputting a matching result: the matching at the intersection point is used to determine the direction of the incremental matching, and the pseudo code of the algorithm is shown in fig. 9. Firstly, constructing a candidate point set I based on intersection points and search radiuses, then obtaining candidate points and projection points of the candidate points on corresponding road sections, calculating the similarity of tracks and the road sections, and finally determining the incremental matching direction through comparing the similarity.

FIG. 10 depicts an intersection point v₁The matching process of (1). As shown in FIG. 10(a), e₁For the section already matched by the last increment, v₁For the start of the next increment match, the matching direction of the subsequent increment needs to be determined, i.e. from path e₁→e₂And e₁→e₃In determining the path matching with the trajectory, fig. 10(b) and 10(c) respectively depict the matching process of two different paths, and e can be easily determined by calculating the similarity between the trajectory and two sets of points of the projection points thereof₃Are subsequent matching road segments. Wherein, the candidate point set I should include the track points of the passed and failed intersection points at the same time, otherwise, the step length r is used_c.stepStepwise increasing search radius r_cUntil the condition is satisfied.

And judging the position relation between the trace points and the intersection points in the candidate point set I by using the projection of the intersection points on the connecting lines of the adjacent trace points. As shown in FIG. 11(a), p_i，p_i+1Is two continuous tracing points, v_kFor crossing points, set v_kIs' v_kAt p_i，p_i+1The projected point of (a). The determination method comprises the following steps: from v_kTo line segment p_i，p_i+1Drawing a perpendicular line, if the foot falls on the line segment p_i，p_i+1In the above, the foot is regarded as v_kOtherwise, take p_i，p_i+1Middle distance v_kThe nearer end point is the projection point. Further, when v is judged_k' at p_i，p_i+1In time between, p_iPoint v of not passing through the road_k，p_i+1Has passed v_k(ii) a When v is_k' with p_iWhen overlapping, p_i，p_i+1All have passed v_kAs shown in FIG. 11 (b); when v is_k' and p_i+1When overlapping, p_i，p_i+1All fail to pass v_kAs shown in fig. 11 (c).

The similarity at the intersection point is the synthesis of two indexes of Hausdorff distance and bending, and the calculation formula is as follows:

wherein H_difIs the absolute value of the difference in Hausdorff distance between the track and the corresponding path, K_difThe value range of the similarity is (0, 1) for the absolute value of the curvature integral difference between the track and the corresponding road]The larger the value, the more similar the current path and trajectory.

The Hausdorff distance is a measure describing the degree of similarity between two sets of points, assuming that there are two sets of sets a ═ a₁，a₂，...，a_p}，B＝{b₁，b₂，...，b_q-the Hausdorff distance between these two sets of points is: h (a, B) ═ max (H (a, B), H (B, a)) where H (a, B) ═ maxmin | | a-B |, (a ∈ a, B ∈ B), H (B, a) ═ maxmin | | B-a |, (a ∈ a, B ∈ B), | | | | | | | | | | B-a |, (a ∈ a, B | | | | is the distance between point set a and point set B.

And the curvature is used for describing the geometric shape characteristics of the candidate point at the intersection point and the road section to be matched. In the map matching problem, both the trajectory and the road are discretized into a sequence of points, and therefore a tangent vector representing the position is approximated by a vector formed by two points adjacent to each other in front and back in the discretized point sequence. Let the smoothing curve gamma be discretized into a sequence of points [ a ]₁，a₂，...，a_n]Then, the curvature integral of the discretized curve is calculated by the following formula:

when calculating the curvature of each point, the true north direction is used as a reference.

The present invention will be further described with reference to the following examples.

Examples

In order to verify the effectiveness of the algorithm, partial taxi tracks are selected from a GeoLife 1.3 public data set provided by Microsoft Asia institute for experiment, and the track sampling interval is 1 s. The road network data is a road traffic network in Beijing City, the selected road section of the track includes a large number of parallel road sections, complex intersections and other special road sections, and the test track is shown in figure 12.

Before matching, track data are preprocessed, abnormal track points which possibly exist in a track are eliminated based on moving speed, and the abnormal track points can be divided into two types: firstly, the moving speed of the track is far lower than the normal running speed or the stagnation point, and secondly, the point deviating from the normal track range is caused by the positioning error. The two types of abnormal points are shown as abnormal points on the speed, and can be removed through a speed threshold value, and the abnormal points with the speed less than 1m/s or the speed more than 35m/s are removed in the experiment.

The environment of the experimental host is AMD Ryzen 536006-Core Processor CPU @3.60GHz, 16.0GB RAM and WIN10 x 64; the programming environment used is Python 3.6.

In order to verify the effect of the algorithm, the invention provides a curvature integral constrained map matching algorithm (hereinafter referred to as curvature integral algorithm for short, and relevant reference documents are Zhe Zeng, Tong Zhang, Qingquan Li, Zhongheng Wu, Haixiang Zou)&Chunxian Gao.Curvedness feature constrained map matching for low-frequency probe vehicle data[J]International Journal of geographic Information Science, 2016, 30: 4,660-: newson, p., Krumm, j.: hidden markov map matching through noise and spark [ J]In: ACM SIGSPATIAL International conference on Advances in Geographic Information Systems, 2009, 336-. Quantitative evaluation is mainly performed from the perspective of accuracy and efficiency of matching results. The evaluation indexes of the algorithm matching precision are as follows: path Mismatch score (RMF), accuracy (Precision), Recall (Recall). Wherein, the formula of RMF is:

in the formula, L₊For the length of the error increase in the algorithm calculation result compared to the true path, L_-Calculating a length of error reduction in the result compared to the true path for the algorithm; the calculation formulas of the accuracy rate and the recall rate are respectively as follows:

and

in the formula L_matchedFor the total length of the correct matching path in the algorithm calculation result, L_inferredFor the total length of the algorithm calculation result, L_realThe total length of the real path.

Considering the GNSS positioning error and the road width in the experimental data set, and determining the radius of the buffer area of the candidate road section _prAnd candidate point buffer radius _crIs set to 30m and the corresponding step size is set to 5 m. Meanwhile, in order to compare the matching performance of the correlation algorithm under different sampling frequencies, the sampling interval is gradually increased to 10 seconds in the experimental process, the corresponding algorithm is realized according to the description in the literature, and the indexes are utilized to carry out a comparison experiment.

The matching result index statistics are shown in fig. 13, when the sampling interval is 1s in the test data set, the accuracy of the algorithm provided by the invention reaches 97.70%, the recall rate reaches 97.10%, and the RMF is 5.23%, which are greatly improved compared with other two types of algorithms, indicating that the algorithm of the invention can ensure higher matching accuracy under high sampling rate. The reason mainly includes two aspects: (1) the algorithm filter filters out most wrong road sections, and the possibility of wrong matching is greatly reduced. Fig. 14, 15, 16, and 17 show the matching results of the data sets at a sampling interval of 1 s. In the graph, the HMM algorithm has a large number of conditions that tracks are matched to road sections in opposite directions in matching, and the road sections in the opposite directions are filtered out before the algorithm is matched; (2) the matching process based on the incremental algorithm provided by the invention ensures the uniqueness of paths among intersection points and weakens the adverse effect of parallel road sections on the matching result. As shown in fig. 15, the curvature integration algorithm has a situation that both parallel links are selected, because the method selects a matching link based on a point pair, and when the track sampling rate is high, two links parallel in the same direction may be selected in different point pairs, thereby causing a wrong matching, while the algorithm of the present invention avoids a wrong selection of a parallel link. Meanwhile, the algorithm also shows a good effect in matching a complex scene in which the driving direction changes by using multi-layer interworking, and the result is shown in fig. 17. However, as the sampling interval is increased, the number of candidate points in intersection point matching is gradually reduced, so that the accuracy and recall rate of the algorithm are gradually reduced, and no obvious advantage exists when the sampling interval is about 10s, as shown in fig. 13(a) and 13(b), which indicates that the algorithm of the present invention is sensitive to the sampling rate change. However, due to the uniqueness of the algorithm to acquire the path, the RMF is still significantly better than other algorithms.

The statistics of the efficiency of the above algorithms are shown in fig. 18. The shorter the sampling interval, the more significant the efficiency advantage of the present algorithm. The reason for this is that the number of times of calling the matching unit in the two comparison algorithms is closely related to the number of trace points, and frequent calling of the matching unit undoubtedly results in higher time cost. The algorithm only makes the calling times of the calculating unit relevant to the number of the road junction points in the road network by taking the path as the increment, and the complexity of the road network is greatly reduced after the road network is combined and filtered, thereby further improving the efficiency.

Fig. 19 is the number of times of calling the matching unit in each algorithm when the sampling interval is 1s, and when the track sampling interval is short, the number of track points is far greater than the number of intersection points to be matched, so the number of times of calling the matching unit of the algorithm is far lower than those of other two algorithms, and the efficiency of the algorithm is obviously improved. With the increase of the sampling interval, the number of the track points is gradually reduced, but the number of the intersection points is unchanged, so the efficiency advantage of the algorithm gradually disappears, and the efficiency of the algorithm is close to that of the global matching algorithm.

Compared with the existing curvature integral constraint map matching algorithm and the map matching algorithm based on the hidden Markov model, the algorithm provided by the invention has higher accuracy and efficiency when solving the map matching problem of the high-sampling-rate GNSS track points, and can better process various complex road section conditions. Therefore, the method is very suitable for the high-sampling-rate GNSS track matching application scene in the complex urban road network.

The above embodiments are not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make variations, modifications, additions or substitutions within the technical scope of the present invention.

Claims

1. A high sampling rate track map matching method based on path increment is characterized in that: the method comprises the following steps:

and step S3, taking the path between two adjacent intersection points in the candidate road sub-network as increment, and performing track point matching calculation based on the path increment.

2. The path increment-based high-sampling-rate track map matching method according to claim 1, characterized in that: in step S1, the GNSS track point map matching process based on the path increment is defined as: and screening out candidate road subnetworks G ' (V ', E ') according to the track T and the road network G (V, E), and then performing track matching calculation by taking the paths between two adjacent road junction points in the candidate road subnetworks as increments, thereby gradually determining the process of the actual motion path of the object.

3. The path increment-based high-sampling-rate track map matching method according to claim 1, characterized in that: the specific process of step S2 is as follows: and (3) combining the spatial distribution and movement rules of two adjacent track points with the characteristics of the nearby roads, and eliminating a large number of irrelevant road sections by adopting a series of filtering conditions of distance constraint, road connectivity constraint, direction consistency constraint and single-point communication road section pruning to obtain candidate road subnets G ' (V ', E ').

4. The path increment-based high sample rate track map matching of claim 3The method is characterized in that: in step S2, the distance constraint is implemented by dynamically adjusting the buffer area, i.e., using the track point as the center of circle and r as the center of circle_p0Constructing a buffer area for the initial radius, and if the candidate road section set is empty, enlarging the radius by a fixed step length r_p.stepAnd circulating the steps until the candidate road section set at least comprises one road section.

5. The path increment-based high-sampling-rate track map matching method according to claim 3, characterized in that: in step S2, the road connectivity constraint includes a topology connectivity constraint and a direction connectivity constraint, where the topology connectivity constraint: for any point p in the trajectory_iIts candidate road segment set C (p)_i) In which there is a section e_iAnd candidate segment set C (p)_i+1) In a certain section e_jThe endpoints are connected or overlapped;

directional connectivity constraints: for any point p in the trajectory_iSet of candidate links C (p)_i) In which there is a section e_iAnd candidate segment set C (p)_i+1) In a certain path section e_jAre connected or overlapped, and when e_iAnd e_jWhen connecting, the two sections should meet the requirement of connecting end to end at the connecting position.

6. The path increment-based high-sampling-rate track map matching method according to claim 3, characterized in that: in step S2, the directional consistency constraint includes a length ratio constraint and an azimuth angle constraint, where the length ratio constraint: if the length ratio of the distance between two continuous track points to the distance between the projection points of the two continuous track points on the candidate road section is less than 0.8 of a threshold value, rejecting the candidate road section;

and (4) azimuth angle constraint: and calculating the included angle between the track line segment and the projection line segment on the candidate path, and rejecting the candidate line segment when the included angle is larger than 90 degrees.

7. The path increment-based high-sampling-rate track map matching method according to claim 6, wherein: in the azimuth constraint, the track is calculatedAnd eliminating a path opposite to the vector azimuth angle difference between the projection points of the track on the candidate path, wherein the vector azimuth angle is a horizontal angle passing through from the due north direction to the current vector along the clockwise rotation direction, and the calculation formula is as follows:

8. The path increment-based high-sampling-rate track map matching method according to claim 3, characterized in that: in step S2, the "pruning" of the single point connection link means: and traversing each road section in sequence, respectively calculating the degrees of the starting point and the ending point of each road section, and deleting other single-point communication road sections except the head road section and the tail road section.

9. The path increment-based high-sampling-rate track map matching method according to claim 1, characterized in that: the specific process of step S3 is as follows:

after the initial road section is determined, entering an incremental calculation process, and gradually determining a subsequent matching road section by taking a road junction point as a starting point in each incremental calculation process;

when the top point connected with the matched road section is a road junction point, entering the next incremental calculation process; and when the single connected point appears, taking the single connected point as a track end point, and ending the matching process.

10. The path increment-based high-sampling-rate track map matching method of claim 9, wherein: in step S3, when the incremental advancing direction is determined at the intersection, a similarity evaluation index combining the Hausdorff distance and the curvature is used to reduce the influence of the parallel equidirectional road segment on the matching result.