CN112566009B - Participation type indoor positioning system based on geomagnetism - Google Patents

Abstract

The invention discloses a participating type indoor positioning system based on geomagnetism and an implementation method thereof. The system mainly comprises a fingerprint library construction scheme for reducing manpower and a matching scheme for improving precision and efficiency; the fingerprint database construction scheme is used for mining map topology from data contributed by a user and constructing a corresponding geomagnetic fingerprint database, does not need plane drawings, does not limit user sampling behaviors and can be widely used; the geomagnetic matching scheme divides a geomagnetic sequence into monotonous intervals, dynamically sets the size of a matching window, and meanwhile, the pre-judging mechanism reduces the complexity of calculation and can quickly return a correct positioning result after a user sends a positioning request.

Description

Participation type indoor positioning system based on geomagnetism

Technical Field

The invention belongs to the field of indoor positioning, and relates to an indoor positioning system based on a geomagnetic fingerprint database and an implementation method thereof.

Background

In the era of mobile internet, location-Based Service (LBS) applications create great commercial value and have great influence on society. In the field of outdoor Positioning, the Global Positioning System (GPS) has been highly successful. However, GPS signals are blocked by walls, and accurate positioning cannot be performed in indoor scenes. Therefore, indoor positioning technology has been rapidly developed, and various signals are used in the indoor positioning technology, such as Wi-Fi (Wireless Fidelity), ultra Wide Band (UWB), RFID (Radio Frequency Identification), radio signal FM (Radio Modulation), inertial sensor INS (Inertial Navigation System), image, geomagnetism, sound wave, and visible light.

The indoor positioning scheme needs to consider the deployment and maintenance costs of the scheme while considering the improvement of the positioning accuracy, and the high deployment cost can cause the difficulty in implementing the positioning scheme and is difficult to popularize and apply. The main problems with existing research are 1) the need to deploy additional infrastructure: deployment and maintenance costs of the infrastructure can limit the expansion of the technology. Technologies such as those based on visible light, acoustic, RFID, UWB, etc., all require the support of the underlying device. 2) The complex indoor environment affects the stability of the positioning signal: the indoor environment is not as spacious as the outdoor environment, and the radio frequency signal can have the problems of signal attenuation, multipath effect and the like under the indoor environment. For example, ultrasonic signals are easily interfered by personnel flow and multipath effects; wi-Fi signals are prone to signal degradation problems.

Geomagnetism does not require deployment of additional infrastructure, and thus positioning solutions using geomagnetism are superior to other solutions from a cost perspective. Meanwhile, the geomagnetic signals are more stable, larger fluctuation caused by the change of indoor environment can be avoided, and the geomagnetic sensor is more suitable for indoor positioning. The existing geomagnetic scheme also has a lot of defects, mainly including: 1) The time-consuming and labor-consuming fingerprint database construction process comprises the following steps: the existing research needs an engineer to perform reconnaissance in a positioning place, manually acquire actual fingerprint data, and mark a position, thereby establishing the mapping of the fingerprint and the position. 2) The crowd-sourced library building scheme in which users participate relies heavily on user interaction: the crowdsourcing technology can contract the work collected in the database building process to the smart phone user, and the scheme can effectively reduce the cost of customizing the fingerprint database. However, the prior art solutions rely heavily on the interaction of the mobile phone user, and the user can only upload data with a fixed format, and is required to actively mark the data. Frequent interaction can reduce the enthusiasm of the user, and if the data uploaded by the user is wrongly marked or is missing, the precision of the fingerprint database can be greatly reduced. 3) It is necessary to provide a plan view of the location site: in practical application, a plan view of a positioning field is difficult to obtain. Floor plans of venues such as airports, teaching buildings, etc. may involve confidential problems. 4) The fingerprint matching scheme has low fixed precision and efficiency: the current sequence matching solution has the problem that the physical end points of the sequences cannot be aligned, so that the positioning accuracy is not ideal. Meanwhile, due to the limitation of sampling frequency, the magnitude of data of the fingerprint database to be matched is large. The existing scheme has high calculation complexity and poor convergence, so that the positioning efficiency is not high.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: the indoor positioning system is based on geomagnetism, is constructed by few workers, and guarantees positioning accuracy and positioning efficiency, and the implementation method thereof are provided.

The participating indoor positioning system based on the geomagnetism comprises a training stage and a positioning stage, wherein the training stage is used for completing the construction of a map and a fingerprint database, and the positioning stage is used for responding to a positioning request of a user by matching the geomagnetism fingerprint. 1) A training stage: mining spatial structure information (map topology) from data provided by participating users; the association maps the geomagnetic fingerprint with the location. 2) A positioning stage: and receiving a request sent by a user, and feeding back a positioning result in time.

The participating geomagnetic library construction scheme adopted for solving the technical problems comprises three parts, namely trajectory data acquisition, user trajectory processing and geomagnetic fingerprint library construction.

A. When a user walks, the smart phone terminal APP automatically records track data and uploads the track data to the server side. The trajectory data includes not only data of the magnetometer but also data of the acceleration sensor and data of the attitude sensor.

B. After receiving the user data, the server needs to perform data preprocessing, which includes geomagnetic value processing, angle data processing and step number detection.

And C, combining the tracks on the same path by the BM-MAP algorithm, and connecting the tracks after combination to form a complete MAP topology. First, the participating users will pass through the turn angle when sampling, and the BM-MAP algorithm identifies the turn part in the trajectory and cuts the long trajectory into trajectory segments according to this part. BM-MAP defines buckets in order to distinguish between different physical corner location trajectories. A bucket represents a corner position in physical space and track segments from the same direction and the same corner position should be sorted into the same bucket. Then, the BM-MAP algorithm divides the track segments into different buckets, and further merges the track segments of the same path. The algorithm communicates map topology according to the information of the bucket, and spatial structure information is obtained. The BM-MAP algorithm provides a step displacement model, which divides a plane space into intervals with step as a unit, wherein each step interval corresponds to a position in an actual scene. After the position intervals are divided, the algorithm associates corresponding geomagnetic fingerprints for the intervals. Finally, BM-MAP algorithm proposes a correction mechanism based on geomagnetic fingerprint feature sequence to reduce the error of fingerprint mapping.

The geomagnetic matching scheme adopted by the invention for solving the technical problems comprises two stages of off-line sub-library and on-line matching. 1) Performing off-line warehouse separation: some operations may be performed on the fingerprint repository during the offline phase without user interaction requests to reduce the matching pressure during the online phase. Therefore, the matching scheme introduces angle information and performs sub-library on the geomagnetic fingerprint library. In the online stage, matching is not needed in a global fingerprint library, and matching is only needed in a corresponding angle sub-library. 2) And (3) online matching: in the online stage, an MI-DTW algorithm based on the DTW algorithm is provided, the fingerprint sequence is divided into monotone intervals, and the length of a matching sequence is dynamically set; meanwhile, a pre-judging mechanism based on the interval is provided, and the calculation times are reduced.

Drawings

FIG. 1 is a block diagram of a geomagnetic based participant positioning system;

FIG. 2 is a partial track segmentation T, L type of map construction according to the present invention;

FIG. 3 is a step displacement model according to the present invention;

FIG. 4 is an exemplary illustration of an L-shaped split barrel in a barrel according to the present invention;

FIG. 5 is a same group trajectory merge involved in the present invention;

FIG. 6 is a different set of track merges involved in the present invention;

FIG. 7 is a diagram of the same edge merging involved in the present invention;

FIG. 8 is a bucket consolidation process related to the present invention;

FIG. 9 is a flow diagram of an offline binning process in accordance with the present invention;

FIG. 10 is a sliding window fingerprint matching graph in accordance with the present invention;

FIG. 11 is a diagram of a user fingerprint to template fingerprint alignment relationship in accordance with the present invention;

FIG. 12 is a monotone interval division diagram according to the present invention;

FIG. 13 is a schematic diagram illustrating an interval anticipation situation according to the present invention;

Detailed Description

The invention will be further described with reference to the accompanying drawings. It should be noted that the description of the embodiments is only for assisting understanding of the present invention, and the present invention is not limited thereto.

The system as shown in fig. 1 is mainly divided into two important implementation parts: the method comprises a geomagnetic fingerprint database construction scheme and a geomagnetic matching scheme.

The participating geomagnetic fingerprint database construction scheme comprises the following stages:

1. user information extraction

The types of track data uploaded by users are mainly classified into three types: 1) A geomagnetic value; 2) An attitude sensor value; 3) An acceleration sensor value. Due to the difference of mobile phone devices and the diversification of use scenes, the acquired values are different. Thus, for each geomagnetic sample value f _i ＝(m _x ，m _y ，m _z ) Calculating f _i Of (2)

Extracting processed geomagnetic fingerprint sequence { f ₁ ，f ₂ ，f ₃ ，...，f _n }. The value of the attitude sensor is used for extracting angle information, and the value range of the reading of the attitude sensor in some mobile phones is [ -180 degrees, 180 degrees °]While the value range in other mobile phones is [0 °,360 ° ]]. In order to unify data specification, the algorithm is uniformly converted into [0 degrees and 360 degrees ] when recording the data of the database]The interval of (2). The value of the acceleration sensor is used to extract the step information in the trajectory. The value of the acceleration sensor is also a three-dimensional data

Wherein a is _i Is the value obtained by taking the modulus of the ith acceleration sampling value. The acceleration values are first preprocessed once with a smoothing filter. And after processing, low-pass filtering is applied to filter out high-frequency components of the acceleration value. After filtering, calculating the number of steps by adopting a sliding window peak value detection scheme if a _i In the window range [ t _i -t _w /2，t _i +t _w / ₂ ]Maximum internal value, then a _i Is judged as a peak value and is recorded as one step. Wherein t is _i Is the current point in time, t _w Is the size of the sliding window.

2. Trajectory segmentation

According to the scheme, the corner position is selected as a key point, and the track is segmented according to the key point. Influenced by building structures, corners being commonExhibiting a 90 deg. variation and a partial annular corner exhibiting a 45 deg. variation. The position of the corner in the track can be easily identified through the angle information acquired in the track information extraction stage. And forming a track segment after the track segmentation, and giving a type definition of the track segment for the convenience of subsequent description: 1) Type T: turn type, which refers to corner type track segment. 2) And (3) type L: line type, refers to straight type track segments. In fig. 2, the red portion is a T-shaped track segment and the blue portion is an L-shaped track segment. In order to facilitate the construction of the map topology, an additional segment label is added for the L shape when the track is segmented,

and

indicating that the L segment is cut from the right side of the T-shape with the number i. According to the sampling sequence and the segmentation sequence, the segment label gives direction information of the L-type, and indicates the direction of the segmented L-type relative to the T-type on the original track.

BM-MAP Algorithm

The BM-MAP comprises two parts of a bucket dividing rule and a combining rule. Classifying the T-shaped track sections by a bucket dividing rule, and determining the number of buckets; while it also distributes the L-shaped track segments into the respective buckets. Combining the tracks according to the barrel on the basis of the step number displacement model by the combining rule, and excavating a map topology; and meanwhile, the position area of the map is divided, and corresponding geomagnetic fingerprints are related.

A. The barrel dividing rule is as follows: two parts are involved: t type divides the bucket and L type divides the bucket. The T-shape is a track segment of a corner type and represents the joint position of the corridor and the corridor in the building structure. The algorithm needs to classify the segmented T-shaped track segment set first and then segment the same T-shaped track segments at the physical positions together. If the T-type set is divided into n classes, the algorithm creates n buckets. Having established the number of buckets, the L-type needs to be assigned to the associated bucket.

(1) T-shaped sub-barrel

T-shaped fingerprint sequences from the same position have higher similarity, onlyAll T-type geomagnetic fingerprints need to be classified by calculating their DTW similarity. Two T-shaped geomagnetic sequences F are given ₁ ＝(f ₁ ，f ₂ ，...，f _m ) And F ₂ ＝(f ₁ ′，f ₂ ′，...，f _n '), the DTW algorithm maintains a set of costs ω = ω (1), ω (1),..,. Omega (k), where ω (k) = [ i (k), j (k)]，[i(k)，j(k)]Show F ₁ And F ₂ I (k) and j (k) are respectively F ₁ And F ₂ An index set of sample points. Cost function omega ^* Is defined as follows:

where δ (ω (k)) is a distance value obtained by the inverse operation of cosine similarity:

the similarity of two T-shaped geomagnetic sequences can be obtained by calculating omega ^* It is found that the lower the calculated value, the higher the similarity of the two trajectories. Dividing similar T types into the same barrel, and generating n barrels B after T types are divided into barrels _n ＝{T _i ，T _j ，...，T _k In which B is _n Is a barrel of reference number n, each B _n There is at least one T-shaped section.

(2) L-shaped sub-barrel

Most of the L-shaped tracks are obtained by track segmentation, and a small part of the L-shaped tracks are scattered track segments uploaded by users. Only the L-type with the associated information (L-type with segment labels) is considered when extracting the topology in this section. And traversing and inquiring all L-shaped segment tags, and dividing the L-shaped segments into different buckets according to tag values. If query L _i Segment tag of T _i And knows T _i The corresponding barrel is B _i Then L is _i Should be classified into bucket B _i In (1). There may be 2 pieces of tag information for an L-type, and this L-type is divided into 2 buckets and also divided into 2 bucketsOne bucket mapping table information mapList (B) needs to be updated _i ，B _j )。(B _i ，B _j ) Indicating that the left side of the L-shape is barrel B _i The right side is a barrel B _j . The mapList information is used for bucket-to-bucket connectivity, physically representing corner-to-corner connectivity. The entire map topology can be connected through the mapList.

B. Merging rules

The step number displacement model arranges the track time sequence into a sequence divided by step numbers. Each step in the divided track is an interval, and the interval corresponds to a mark point at an actual position, which is called a position mark point and is marked as l. l. the _k Representing one actual position in the localization space, each actual position uniquely corresponding to one/. Figure 3 is an exemplary diagram of a step shift model,

is an L-shaped track segment without division, and the length of the number of steps is 9. According to the division mode of the step number displacement model, the method is to

Dividing the space into 9 step intervals, each step interval corresponding to a position in the actual space, and the ith interval being expressed as l _i 。

The map topology is composed of an edge point graph, and the map topology can be stored in a data structure by using an adjacency matrix. The rows and columns in the adjacency matrix represent the points in the graph, and the values of the matrix represent the distances between the points. The BM-MAP defines a matrix M as a MAP topology matrix, and the number of rows and columns in the M is determined by the number of the position marking points l. M [ l ] _i ][l _j ]And the distance between the ith position mark point and the jth position mark point is represented, and the value of the distance is the difference value calculated by the step number and the step length between the two position mark points. The distance between the position points can be calculated through a step number displacement model, and the map topology of the space can be extracted.

After the map topology matrix M is created, the geomagnetic fingerprint needs to be associated with position information. Defining a mapping table associating geomagnetic fingerprints and positions as a hashMap, wherein the hashMap maintains step digitsAnd (4) moving the mapping relation between the position mark points in the model and the geomagnetic fingerprint. For a geomagnetic fingerprint sequence F = { F in an arbitrary trajectory ₁ ，f ₂ ，f ₃ ，...，f _n Is f thereof _i Geomagnetic value associated position mark point l _i 。

The merge rule is divided into two parts: the intra-bucket trajectory merge and the inter-bucket trajectory merge.

(1) In-barrel track merging

Fig. 4 has 4L-shaped traces divided into B1 buckets. L is a radical of an alcohol ₁ = {1,2,3,4} represents L ₁ Is a track comprising 4 steps in length. Same principle L ₂ ＝{5，6，7}、L ₃ {8,9, 10, 11, 12, 13} and L ₄ = {14, 15, 16} each represents L ₂ 、L ₃ And L ₄ The step number lengths of (a) are 3, 6 and 3, respectively. When the 4 track sections are combined together in the bucket, the position marking points l of the 4 track sections are established, meanwhile, the distance of the corresponding map topological matrix M is calculated, and finally, the position information of the corresponding geomagnetic fingerprint is associated. The in-bucket merge rule includes three phases:

a) L-shaped edge in barrel

The L-shaped segment labels have directional information, and there should be 3L-shaped segments to the left of B1 in FIG. 4: l is ₁ 、L ₂ And L ₃ On the right, there should be an L-shape: l is ₄ 。

b) Merging at the same side

The BM-MAP proposes a grouping rule for the same-edge construction. The grouping rules divide the L-shapes of the same physical path into the same group (L) ₁ And L ₂ Are in a group, L ₃ One group), the tracks of the same group are merged first, and the tracks of different groups are merged later. Grouping is carried out by calculating the similarity of L types, and the similarity calculation mainly comprises two steps: and calculating the difference value of the track steps. Extracting the trajectory L _i And L _j Step number of Step _i And Step _j Difference δ = | Step _i -Step _j L; and calculating the similarity of the geomagnetic fingerprints. The calculation scheme refers to equations (X) and (X).

1) Same set of trajectory rules

When combined, not aligned with the direction of the barrel, butThe distance is aligned according to the other direction, and after the alignment, the distance calculation error caused by the difference of track steps is reduced by calculating the average value of the distances. As shown in FIG. 5, the long step interval is selected to establish the position mark point during merging, so the merged position mark point is { l } ₁ ，l ₂ ，l ₃ ，l ₄ And adding corresponding row and column values to the map topology matrix M. L is ₁ And L ₂ Instead of aligning in the B1 direction, the distances of the positions in M are calculated by aligning {1} and {5} instead of {4} and {7}, averaging the distances of {1,2,3} and {5,6,7} after alignment, and allowing d to be d ₃₄ ＝d ₃₄ /2. And after the M is updated, the label of the mark point of the geomagnetic fingerprint associated position is updated, and the mapping in the hashMap is updated.

2) Merging of diverse group trajectories

In FIG. 6, L ₃ L to be combined with ₁₂ Merging, selecting long step interval to establish position mark points during merging, wherein the position mark points after merging in figure 6 are { l } ₅ ，l ₆ ，l ₁ ，l ₂ ，l ₃ ，l ₄ }. Aligned in the direction of the bucket for different combinations and calculations of distance, L in FIG. 6 ₃ And L ₁₂ Will {6} and { l in the direction of B1 ₄ Align and calculate {3,4,5,6} and { l }simultaneously ₁ ，l ₂ ，l ₃ ，l ₄ And updating the distance matrix M for the mean value of the distances, associating the upper position mark points for the geomagnetic fingerprint and updating the hashMap.

c) Different edge merge rules

After the same-side combination is completed, the position marking points { l of the two sides of the barrel are determined _i ，l _m ，...，l _k And { l } and _j ，...，l _n and communicating the position mark points on the two sides according to the connectivity of the barrel. As shown in FIG. 7, B1 determines the left position mark point as { l after merging on the same side ₁ ，l ₂ ，...，l ₆ Denoted by { l } on the right ₇ ，...，l ₉ }. The distance between the location points can be calculated based on the connectivity of B1, as shown by l in FIG. 7 ₄ To l ₇ Is equal to

And similarly, updating the distance in the map topology matrix M.

(2) Inter-bucket merge rules

For each bucket, the merging is performed according to the rule of merging in the bucket. After the merging in the buckets is completed, the topology of the map needs to be merged and communicated between the buckets. The buckets are related according to a mapList mapping table, and the mapList records the bucket B _i And B _j The communication relationship between them. Take one of the records mapList (B) _i ，B _j )＝L _n ，(B _i ，B _j ) With directional information, indicating L _n n left side is B _i And the right side is B _j When the barrels are combined, B needs to be updated _i And B _j If B _i The set of location points on the right is { l } _i1 ，l _i2 ，...，l _in }，B _j The set of position points on the left side is { l } _j1 ，l _j2 ，...，l _jn The merging position point is { l } ₁ ，l ₂ ，...，l _n Updating the distance of the relevant position points in M according to the connectivity, and increasing B _i To B _j The distance of (c).

FIG. 8 shows a complete bucket merge process, two buckets B in the figure _i And B _j They are first combined separately in the barrel. Because B _i And B _j There is a connected L-type, and this connection is recorded in the mapList. According to the rule of merging between buckets, B _i And B _j The inter-bucket distance of (a) is updated.

The geomagnetic matching scheme is divided into the following stages:

1. offline sub-library based on angle information

Fig. 9 is a flow chart of offline binning, in which a fingerprint bin is divided into n sub-bins according to angle information, and each sub-bin corresponds to information in one direction. When a server receives a positioning request of a user, firstly extracting angle information in data; then, the angle index in the fingerprint library is inquired and positioned into the corresponding sub-library. The number of fingerprints to be matched can be greatly reduced after the fingerprints are sorted according to the angles.

MI-DTW algorithm

The geomagnetic fingerprint to be matched uploaded by the user is a relatively short time sequence, and needs to be matched with a longer template fingerprint sequence in a fingerprint database. The short uploaded fingerprint is defined as a user fingerprint U, and the long geomagnetic fingerprint in the fingerprint database is defined as a template fingerprint T. MI-DTW is a dynamic fingerprint matching algorithm based on a sliding window, and the working process of the geomagnetic matching algorithm based on the sliding window is shown in FIG. 10. When the DTW algorithm is used, two fingerprint sequences are required to be physically aligned in end points, otherwise, the cost value calculated by the cumulative function of the DTW algorithm has errors. In fig. 11, AB is a template fingerprint, and CD is a user fingerprint to be matched. CD can match with C 'D' on AB, but CD and C 'D' cannot be aligned in time, and CD is shorter than C 'D'. In order to match the CD and C 'D' better, the size of the sliding window cannot be simply set to the length of the CD.

And the MI-DTW divides the sequence into monotone intervals, and dynamically determines the length of a sliding window according to the length of the monotone interval, so that the matching is more flexible.

MI-DTW monotonic interval division

Fig. 12 is a monotone interval division diagram. The template sequence AB is labeled as 0 or 1 according to the monotone increase and decrease of the waveform. The rising interval is marked as 1 and the falling interval is marked as 0. Each template geomagnetic fingerprint sequence can be divided according to the rule. The MI-DTW first partitions the user sequence by monotonic intervals, such as CD to [101010]. The template sequence is then divided by monotonic intervals, as AB processed to [10101010101010101010]. At this time, the size of the sliding window w is set to be an interval of [101010], and the fingerprint and the user fingerprint are selected on the template sequence according to the sliding window to calculate the DTW similarity. The next sliding window still slides according to the interval, and the sliding distance is just the distance of one interval. If starting from the interval 1, the next time the interval of 1 is aligned, the interval of 0 is skipped by each sliding. After the DTW similarity of the whole template sequence is calculated, a similarity sequence can be obtained, and finally the best matched template is confirmed according to a threshold value.

MI-DTW (Muting Interval based) step number prejudging mechanism

Fig. 13 illustrates a phenomenon encountered in matching. The fingerprint sequence interval for the current user is [101], which is a relatively short sequence corresponding to only a physically short distance. The S interval [101] on the template fingerprint sequence is a very long sequence, corresponding to a physically long distance. According to the rule, when the window slides, the DTW similarity between the user interval and the S interval [101] is calculated. However, in practical cases, the S interval is a long interval, and the user sequence interval is a short interval, which will not be similar, and calculating the DTW value between them is a waste of computing resources. If the S interval [101] is not similar to the user interval in advance under the actual condition, the DTW similarity does not need to be calculated again, and the matching speed is further accelerated. First, it is necessary to ensure that there is a difference in physical length between the S interval [101] and the user interval [101] to directly skip the similarity calculation. Directly calculating the time difference between intervals or the difference between sample points is not reliable because fingerprints at physically the same end point also have a temporal difference.

MI-DTW introduces the number of steps to calculate the difference in interval length. Since the S interval [101] is physically very different from the user interval [101], the difference in the number of steps between the two intervals is also very large. If the step number difference is detected when DTW matching is carried out each time, and coarse-grained screening is carried out once, the matching times of DTW fine measurement can be reduced.

Claims (4)

1. A participating indoor positioning system based on geomagnetism is characterized in that a fingerprint library construction scheme and a geomagnetism matching scheme which do not consume manpower are adopted, wherein the fingerprint library construction scheme (1) which reduces construction cost is adopted to mine map topology from data contributed by a user and construct a corresponding geomagnetism fingerprint library, the geomagnetism matching scheme (2) can return a correct positioning result after the user sends a positioning request, and the positioning system comprises a mobile terminal (3) and a server terminal (4) and provides good positioning experience;

the mobile terminal (3) is divided into a participating user terminal (5) and a positioning user terminal (6), and the server terminal (4) receives data contributed by the participating user terminal to construct a geomagnetic fingerprint database (7); constructing a partial barrel combination algorithm, namely a Bucket Merging Map Construction algorithm, namely a BM-MAP algorithm for short, and the algorithm utilizes the thought of segmentation combination, firstly identifying key points, namely corner positions, in user data, then segmenting the user walking data according to the key points, then carrying out barrel division on fingerprints according to the key points, finally carrying out combination according to the barrel, calculating the distance according to a step displacement model during combination, simultaneously associating corresponding geomagnetic fingerprints, extracting geomagnetic, angle and step information, then combining the user walking data overlapped on a physical space position through a Map topology mining mechanism (8), simultaneously connecting non-overlapping data, extracting Map topology information from the processed data, needing not to obtain a plane Map of a positioning place in advance, and finally mapping the geomagnetic fingerprints with the positions through the region positions to generate a geomagnetic fingerprint division database (7); when a user carries out positioning, the system divides the fingerprint database into databases according to angle information (9), a positioning user end (6) sends a positioning request, a server receives the positioning request, then the positioning request is positioned into a corresponding sub-database, a geomagnetic matching algorithm based on a monotone Interval, namely a Monotonic Interval-Dynamic Time Warping algorithm (10), is operated to match the geomagnetic fingerprint database, and finally a user positioning result is returned.

2. A geomagnetic-based participant indoor positioning system, according to claim 1, wherein: the fingerprint database construction scheme (1) provides a step number displacement model, a plane space is divided into intervals with step numbers as units, and a map is divided into positions according to the step numbers.

3. A geomagnetic based participant indoor positioning system according to claim 1, wherein: the geomagnetic matching scheme (2) provides an angle-based off-line database partitioning mechanism (9), geomagnetic fingerprints at the same angle are partitioned into the same sub-database, and the geomagnetic fingerprints are located in the corresponding sub-database by judging angle information during location.

4. A geomagnetic based participant indoor positioning system according to claim 1, wherein: the geomagnetic matching scheme (2) provides an MI-DTW algorithm (10), which divides a geomagnetic sequence according to a monotone interval, dynamically sets the size of a sliding window through the monotone interval, and simultaneously provides a prejudgment mechanism based on step numbers.

Images