阅读说明：本技术 一种基于移动终端众包数据的室内导航路网提取方法 (Indoor navigation road network extraction method based on crowdsourcing data of mobile terminal ) 是由 阳媛 况余进 王庆 张小国 杨浩然 于 2021-08-16 设计创作，主要内容包括：本发明公开了一种基于移动终端众包数据的室内导航路网提取方法,属于室内定位导航领域。该算法首先通过对室内行人众包数据进行提取,解算得到行人轨迹；接着采用ST-DBSCAN算法去除冗余轨迹点,结合运动数据信息判定轨迹点所在区域,增添语义信息；最后基于Douglas-Peucker抽稀算法对解算出来的行人室内轨迹进行轨迹压缩,构建原始室内语义路网；该室内导航网络在后续获得新的轨迹时,会自动进行更新和优化,同时可以对后续轨迹进行约束和匹配,提高定位精度。该方法在面对未知环境时,能够快速、高效、低成本地获取室内导航路网,同时当环境动态变换后,可支持对路网进行动态修改。(The invention discloses an indoor navigation road network extraction method based on crowdsourcing data of a mobile terminal, and belongs to the field of indoor positioning navigation. The algorithm firstly extracts indoor pedestrian crowdsourcing data and resolves to obtain a pedestrian track; removing redundant track points by using an ST-DBSCAN algorithm, judging the area of the track points by combining motion data information, and adding semantic information; finally, performing track compression on the calculated indoor track of the pedestrian based on a Douglas-Peucker rarefaction algorithm to construct an original indoor semantic road network; when the indoor navigation network obtains a new track subsequently, the indoor navigation network can be automatically updated and optimized, and meanwhile, the subsequent track can be restrained and matched, so that the positioning precision is improved. The method can rapidly, efficiently and inexpensively acquire the indoor navigation road network when facing unknown environment, and can support dynamic modification of the road network after the environment is dynamically changed.)

1. An indoor navigation road network extraction method based on crowdsourcing data of a mobile terminal is characterized by specifically comprising the following steps:

step 1, acquiring pedestrian crowdsourcing indoor motion data, and combining a positioning algorithm to carry out track calculation: solving an indoor track by adopting a particle swarm positioning algorithm and combining multi-source data;

step 2, removing redundant track points based on an ST-DBSCAN algorithm, and adding semantic information to the track points in the indoor specific area;

removing redundant track points based on an ST-DBSCAN algorithm, and adding semantic information to the track points of indoor specific areas, wherein the method comprises the following specific steps:

step 2.1, setting a residence time threshold t according to the actual situation_thSpatial distance threshold s_thThe minimum point Minpts, and the speed threshold v is increased relative to the original ST-DBSCAN algorithm_thAs a constraint condition, judging wandering points and stagnation points of track points;

step 2.2, local density calculation is carried out on the redundant points formed by the wandering points and the stagnation points:

where ρ is_iIs the local density, d_ijThe space distance between the track point i and the track point j is set; d_cIs a radius threshold; sign (x) function is a decision function: when x is less than 0, sign (x) is-1, when x is 0, sign (x) is 0, when x is more than 0, sign (x) is 1, calculating local density of all redundant points in each area, finding out track core point m with maximum local density, using said core point as centre of circle, d_cCalculating the coordinate average value of all points in the neighborhood as the radius, and taking the coordinate average value as a clustering center point;

step 2.3, comprehensively judging the area range of the track point according to the acceleration, the angular velocity, the air pressure count value and the WIFI signal conversion condition in the obtained multi-source data;

step 2.4, adding indoor semantic information to the track points;

step 3, performing track compression on indoor tracks of pedestrians based on Douglas-Peucker thinning algorithm calculation to construct an original indoor semantic road network;

the method comprises the following steps of compressing the number of track points of the calculated indoor track of the pedestrian based on a Douglas-Peucker thinning algorithm, and constructing an original indoor semantic road network, wherein the method specifically comprises the following steps:

step 3.1, setting an expected upper limit M of the number of nodes of the compression track according to the motion condition, wherein the upper limit M of the number of nodes can be 5% of the number of the track, and ensuring that the track is compressed to a sufficient degree;

step 3.2, setting a self-adaptive distance threshold value to replace a uniform distance threshold value relative to the original Douglas-Peucker thinning algorithm, and taking the distance d of the farthest point as a temporary distance threshold value delta during first compression₁As the number of compression points increases, the threshold value Δ decreases by a certain step length, and the next compression point is away from the threshold value Δ₂The automatic attenuation is performed according to the following formula:

wherein r is the distance between the current farthest point and the connection point;

step 3.3, calculating and recording the similarity between the current compression track and the original track for each increase of the number of nodes, continuously reducing the threshold, performing comprehensive evaluation on the compression degree and the similarity of all track compression conditions in the range of the expected number of nodes when the number of nodes is greater than the upper limit of the expected number of nodes, selecting the optimal track, wherein the compressed optimal track set is { T }₁,T₂,...,T_n}；

And 4, carrying out track calculation on the new pedestrian indoor data, simultaneously carrying out matching processing on the new pedestrian indoor data and the original indoor semantic road network, and updating the original road network.

2. The indoor navigation road network extraction method based on the crowdsourcing data of the mobile terminal according to claim 1, wherein the indoor navigation road network extraction method comprises the following steps: and 2.4, adding indoor semantic information including galleries, rooms and stair openings to the track points.

3. The indoor navigation road network extraction method based on the crowdsourcing data of the mobile terminal according to claim 1, wherein the indoor navigation road network extraction method comprises the following steps:

in the step 4, the trajectory calculation is performed on the new indoor pedestrian data, and the similarity calculation and matching processing is performed on the new indoor pedestrian data and the original indoor semantic road network at the same time, so that the original road network is updated, and the specific steps are as follows:

step 4.1, the newly producedThe indoor track of the pedestrian is segmented according to turning points { L₁,L₂,...,L_iCalculating a subset of pedestrian tracks of each section and a road network track set { T }₁,T₂,...,T_nScore of track similaritySpatial distance scoringAnd comparing and scoring WIFI signal strength of each point in the track informationFinally, the scores are combined for weighting { omega }₁,ω₂,ω₃Voting to obtain voting result { Q }₁,Q₂,...,Q_i}；

Step 4.2, if the voting result shows that the new track is similar to the road network, setting a threshold value of the distance from the track node to the road network node, and replacing the track node smaller than the threshold value by the corresponding road network node; setting a threshold value of the distance from a track node to a road network line segment, wherein the track node smaller than the threshold value is replaced by a projection point of the node on a certain road network line segment;

and 4.3, if the voting result indicates that the new track is not similar to the road network or the node which does not meet the threshold condition is a new node, no constraint operation is performed. After which the database of the original road network is updated with the trajectory information of the new nodes.

Technical Field

The invention belongs to the field of indoor positioning navigation, and particularly relates to an indoor navigation road network extraction method based on crowdsourcing data of a mobile terminal.

Background

In recent years, with the progress of modernization becoming faster, large public buildings and places have increased like bamboo shoots in the spring after rain, and the range of indoor spaces has increased day by day. These have resulted in a large number of indoor services being required by people without leaving the location services supported by the indoor navigation system. The indoor navigation network is the basis of indoor navigation research. Therefore, how to automatically generate an indoor navigation network has become a research focus in recent years.

Indoor navigation networks have indispensable uses in indoor positioning and navigation, such as positioning services for shopping malls and navigation services for museums. At present, the generation of the indoor navigation network is usually realized by building a computer Aided Design (cad) drawing or a laser radar synchronous positioning and drawing (SLAM) method in a room, and the methods are tedious and expensive. Further, the navigation network thus generated is susceptible to changes in the external environment. Therefore, a practical and cheap method for dynamically constructing an indoor navigation network has important value and necessity.

In order to solve the above problems, patent publication (publication) No. CN111288999A discloses a method, an apparatus, and a device for detecting attributes of a pedestrian network based on a mobile terminal, which obtains road attributes by performing data processing and machine learning after obtaining sensor data of an outdoor mobile terminal, and gives the road attributes to an existing pedestrian network to obtain road network data with attribute information. The method focuses on judging and generating road network attributes for outdoor data in road network processing, and an effective method cannot be provided for how to construct a road network. Patent publication (notice) No. CN 109472416 a discloses an indoor path planning method, an indoor path planning apparatus, and a client terminal based on automatic road network data extraction, in which a path skeleton is extracted by performing image processing on an indoor distribution map, and a road network data structure is generated from path skeleton information. The method depends too much on static indoor information, neglects the conditions of complex indoor structure and variable real environment, and is difficult to effectively obtain the road network meeting the actual conditions.

Disclosure of Invention

In view of the above problems, the invention provides an indoor navigation road network extraction method based on crowdsourcing data of a mobile terminal, which solves the problems of large investment, complex implementation, inconvenience in dynamic modification and the like in the existing indoor navigation network construction technology.

The invention provides an indoor navigation road network extraction method based on crowdsourcing data of a mobile terminal, which specifically comprises the following steps:

step 1, acquiring pedestrian crowdsourcing indoor motion data, and combining a positioning algorithm to carry out track calculation: solving an indoor track by adopting a particle swarm positioning algorithm and combining multi-source data;

step 2, removing redundant track points based on an ST-DBSCAN algorithm, and adding semantic information to the track points in the indoor specific area;

removing redundant track points based on an ST-DBSCAN algorithm, and adding semantic information to the track points of indoor specific areas, wherein the method comprises the following specific steps:

step 2.1, setting a residence time threshold t according to the actual situation_thSpatial distance threshold s_thThe minimum point number Minpts, and the speed threshold v is increased relative to the original ST-DBSCAN algorithm_thAs a constraint condition, judging wandering points and stagnation points of track points;

step 2.2, local density calculation is carried out on the redundant points formed by the wandering points and the stagnation points:

where ρ is_iIs the local density, d_ijThe space distance between the track point i and the track point j is set; d_cIs the radius threshold; sign (x) function is a decision function: when x is less than 0, sign (x) is equal to-1, when x is equal to 0, sign (x) is equal to 0, when x is greater than 0, sign (x) is equal to 1, local densities of all redundant points in each area are calculated, and the local density is the largestA locus core point m, which is taken as a circle center, d_cCalculating the coordinate average value of all points in the neighborhood as the radius, and taking the coordinate average value as a clustering center point;

step 2.3, comprehensively judging the area range of the track point according to the acceleration, the angular velocity, the air pressure count value and the WIFI signal conversion condition in the obtained multi-source data;

step 2.4, adding indoor semantic information to the track points;

step 3, performing track compression on indoor tracks of pedestrians based on Douglas-Peucker thinning algorithm calculation to construct an original indoor semantic road network;

and (3) compressing the track points of the calculated indoor tracks of the pedestrians based on a Douglas-Peucker rarefaction algorithm to construct an original indoor semantic road network, and specifically comprising the following steps:

step 3.1, setting an expected upper limit M of the number of nodes of the compression track according to the motion condition, wherein the upper limit M of the number of nodes can be 5% of the number of the track, and ensuring that the track is compressed to a sufficient degree;

step 3.2, setting a self-adaptive distance threshold value to replace a uniform distance threshold value relative to the original Douglas-Peucker thinning algorithm, and taking the distance d of the farthest point as a temporary distance threshold value delta during first compression₁As the number of compression points increases, the threshold value Δ decreases by a certain step length, and the next compression point is away from the threshold value Δ₂The automatic attenuation is performed according to the following formula:

wherein r is the distance between the current farthest point and the connection point;

and 3.3, calculating and recording the similarity between the current compression track and the original track for each increase of the number of nodes, continuously reducing the threshold, performing comprehensive evaluation on the compression degree and the similarity of all track compression conditions in the range of the expected number of nodes when the number of nodes is greater than the upper limit of the expected number of nodes, selecting the optimal track, wherein the compressed optimal track set is { T }₁,T₂,...,T_n}；

And 4, carrying out track calculation on the new pedestrian indoor data, simultaneously carrying out matching processing on the new pedestrian indoor data and the original indoor semantic road network, and updating the original road network.

As a further improvement of the invention, the indoor semantic information added to the track points in step 2.4 comprises corridors, rooms and stair openings.

As a further improvement of the present invention, in step 4, the trajectory solution is performed on the new indoor pedestrian data, and the similarity calculation and matching processing is performed on the new indoor pedestrian data and the original indoor semantic road network, so as to update the original road network, and the specific steps are as follows:

step 4.1, segmenting the newly generated indoor track of the pedestrian according to turning points { L₁,L₂,...,L_iCalculating a pedestrian track subset and a road network track set { T) of each section₁,T₂,...,T_nScore of track similaritySpatial distance scoringAnd comparing and scoring WIFI signal strength of each point in the track informationFinally, the scores are combined for weighting { omega }₁,ω₂,ω₃Voting to obtain voting result { Q }₁,Q₂,...,Q_i}；

Step 4.2, if the voting result shows that the new track is similar to the road network, setting a threshold value of the distance from the track node to the road network node, and replacing the track node smaller than the threshold value by the corresponding road network node; setting a threshold value of the distance from a track node to a road network line segment, wherein the track node smaller than the threshold value is replaced by a projection point of the node on a certain road network line segment;

and 4.3, if the voting result indicates that the new track is not similar to the road network or the node which does not meet the threshold condition is a new node, no constraint operation is performed. After which the database of the original road network is updated with the trajectory information of the new node.

Compared with the prior art, the invention has the advantages that:

aiming at the technical problems of large investment, complex implementation, inconvenience in dynamic modification and the like of the current indoor road network construction technology, the invention provides an indoor navigation road network extraction method based on crowdsourcing data of a mobile terminal. Firstly, extracting indoor pedestrian crowdsourcing data, and resolving to obtain a pedestrian track; removing redundant track points by using an ST-DBSCAN algorithm, judging the area of the track points by combining motion data information, and adding semantic information; performing track compression on the calculated indoor track of the pedestrian based on a Douglas-Peucker rarefaction algorithm to construct an original indoor semantic road network; when the indoor navigation network obtains a new track subsequently, the indoor navigation network can be automatically updated and optimized, and meanwhile, the subsequent track can be restrained and matched, so that the positioning precision is improved. When facing unknown environment and dynamic environment, compared with the method in the background art, the method has the advantages of high acquisition speed, high efficiency and low cost.

Drawings

FIG. 1 is a flow chart of road network extraction according to the present invention;

FIG. 2 is a schematic diagram of crowdsourcing data acquisition and processing by a mobile terminal;

fig. 3 is a flow chart of a data processing procedure of road network construction.

Detailed Description

The invention is described in further detail below with reference to the following detailed description and accompanying drawings:

the invention provides an indoor navigation road network extraction method based on crowdsourcing data of a mobile terminal, which solves the problems of large investment, complex implementation, inconvenience in dynamic modification and the like in the existing indoor navigation network construction technology.

As a specific embodiment of the invention, the technical scheme adopted by the invention is as follows:

step S1: and acquiring pedestrian crowdsourcing indoor motion data, and performing track calculation by combining a positioning algorithm. The method specifically comprises the following steps:

s1.1: drawing an indoor map according to an indoor fire map or a plan, extracting line features of the drawn indoor map by adopting a self-study software algorithm, and converting the indoor map into a map vector matrix with wall features.

S1.2: and acquiring crowdsourcing data of the mobile terminal from the crowdsourcing database. Crowdsourcing data is subjected to preliminary effectiveness screening and contains three-axis gyroscope information, three-axis acceleration information, WIFI signal information and barometer information. As shown in particular in fig. 2.

S1.3: and carrying out multisource information fusion positioning on the obtained crowdsourcing data of the mobile terminal by adopting a particle swarm optimization, wherein the multisource information comprises PDR positioning information, map constraint information and WIFI signal information. Firstly, determining the initial position of a target according to a WIFI positioning result, and performing state transition updating on the particle swarm by using a PDR positioning output result as a state transition amount. Judging the map vector matrix and the track vector matrix to obtain intersection points of indoor tracks of pedestrians and walls, namely wall penetrating points; carrying out weight zeroing on the through-wall particles in the particle swarm to obtain the particle swarm with corrected map information;

s1.4: and after the particle filtering is finished, calculating an optimal value represented by the particle swarm to obtain an optimal coordinate, namely an optimal pedestrian indoor track.

Step S2: removing redundant track points based on an ST-DBSCAN algorithm, and adding semantic information to the track points in an indoor specific area, specifically comprising the following steps:

s2.1: setting a residence time threshold t according to actual conditions_thSpatial distance threshold s_thThe minimum point Minpts, and the speed threshold v is increased relative to the original ST-DBSCAN algorithm_thAnd as a constraint condition, judging wandering points and stagnation points of the track points.

S2.2: and local density calculation is carried out on the redundant points consisting of the wandering points and the stagnation points:

where ρ is_iIs the local density, d_ijThe space distance between the track point i and the track point j is set; d_cIs the radius threshold; sign (x) function is a decision function: sign (x) ═ 1 when x < 0, sign (x) ═ 0 when x > 0, and sign (x) ═ 1 when x > 0. And calculating the local density of all redundant points in each area, and finding out the track core point m with the maximum local density. Using the core point as the center of a circle, d_cCalculating the coordinate average value of all points in the neighborhood as the radius, and taking the coordinate average value as a clustering center point; calculating a cluster center point within the radius threshold range, wherein the coordinates are as follows:

s2.3: and comprehensively judging the area range of the track point according to the acceleration, the angular velocity, the air pressure count value and the WIFI signal conversion condition in the obtained multi-source data. Due to the fact that the difference of the motion postures of going upstairs and downstairs is obvious, the stair nodes can be judged according to the angular velocities { p, q, r }. When in a certain indoor area, the track point closes the angular velocityThe area is always stabilized between 0 and 4rad/s, the barometer is continuously changed, and the RSSI value detected by the WIFI signal fluctuates abnormally, so that the area is judged to be a stair area; the pedestrian stops more in the facility, so when the track point under the same track stays for a long time in a certain area and moves at a slow speed, the area is determined to be the area in the facility; and adding indoor semantic information to the track points.

Step S3: the method comprises the following steps of performing track compression on indoor tracks of a person to be solved based on lines calculated by a Douglas-Peucker thinning algorithm, and constructing an original indoor semantic road network, wherein the method specifically comprises the following steps:

s3.1: setting an upper limit M of the number of nodes of an expected compression track according to the motion condition, wherein the upper limit M of the number of nodes can be 5% of the number of the track points, and the track is compressed to a sufficient degree;

s3.2: and setting a self-adaptive distance threshold value to replace a uniform distance threshold value relative to the original Douglas-Peucker thinning algorithm. Taking the farthest point distance d as a temporary distance threshold value delta during the first compression₁. Following the point of compressionThe number increases and the threshold delta decreases in steps. Next compression point distance threshold delta₂The automatic attenuation is performed according to the following formula:

where r is the distance of the current farthest point from the connection point.

S3.3: and calculating the similarity between the current compressed track and the original track through a DTW algorithm and recording the similarity for the increase of the number of the nodes each time. And continuously reducing the threshold, and when the number of the nodes is larger than the upper limit of the expected number of the nodes, comprehensively evaluating the compression degree and the similarity of all track compression conditions in the range of the expected number of the nodes, and selecting the optimal track. The compressed optimal trajectory set is { T }₁,T₂,...,T_n}。

S3.4: near an indoor intersection, track T_iA turning point is generated; at the same time, the track T_a，T_bCross nodes are created between each other. When the turning points are enough, the area is an intersection node of an indoor road network. In order to keep the real turning points and the crossing nodes, a clustering algorithm is adopted to remove noise crossing points caused by track deviation. And after the aggregation of all the nodes is completed, connecting all the clustering centers to complete the reconstruction of the node road network. The node road network is stored as a small text file in a point form for convenient visualization.

And step S4, performing track calculation on the new indoor pedestrian data, and performing similarity calculation and matching processing on the new indoor pedestrian data and the original indoor semantic road network to update the original road network. The method comprises the following specific steps:

s4.1: segmenting the newly generated indoor trajectory of the pedestrian according to turning points { L₁,L₂,...,L_iCalculating a pedestrian track subset and a road network track set { T }of each section₁,T₂,...,T_nScore of track similaritySpatial distance scoringAnd comparing and scoring WIFI signal strength of each point in the track informationFinally, the scores are combined for weighting { omega }₁,ω₂,ω₃Voting to obtain voting result { Q }₁,Q₂,...,Q_iThe specific calculation formula is as follows: q ═ ω₁Sⁿ _i+ω₂Dⁿ _i+ω₃RSSIⁿ _i；

S4.2: if the voting result shows that the new track is similar to the road network, setting a threshold value of the distance from the track node to the road network node, and replacing the track node smaller than the threshold value by the corresponding road network node; setting a threshold value of the distance from a track node to a road network line segment, wherein the track node smaller than the threshold value is replaced by a projection point of the node on a certain road network line segment;

s4.3: and if the voting result indicates that the new track is not similar to the road network or the node which does not meet the threshold condition is a new node, no constraint operation is performed. After which the database of the original road network is updated with the trajectory information of the new nodes.

The above description is only one of the preferred embodiments of the present invention, and is not intended to limit the present invention in any way, but any modifications or equivalent variations made in accordance with the technical spirit of the present invention are also within the scope of the present invention as claimed.