A method and system for automatically constructing indoor corridor maps based on crowdsourcing

Collect sensor data through smartphones and use pedestrian track calculation and grid method to build indoor corridor maps, solving the problems of high cost of building traditional indoor maps and difficulty in updating, realizing automatic updates and universal map construction, and improving the indoor map Practicality.

CN114111784BActive Publication Date: 2024-11-22JIANGXI WANGKAI CONSTR CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111251948.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-26
Publication Date
2024-11-22
Estimated Expiration
2041-10-26

AI Technical Summary

Technical Problem

The existing indoor map construction methods have problems such as large expenses in equipment and personnel and difficulty in updating, especially when indoor environment changes, it is difficult to quickly update the map.

Method used

A crowdsourcing-based method is adopted to collect pedestrian sensor data through smartphones, and indoor corridor maps are constructed using pedestrian track calculation, path simplification, path merging and grid methods to realize automatic updates and low-cost map construction.

Benefits of technology

It realizes automatic construction and dynamic update of indoor corridor maps, reduces equipment and personnel costs, is suitable for general map construction of different buildings, and improves the practicality and application prospects of maps.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114111784B_ABST
    Figure CN114111784B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for building an indoor corridor map based on crowdsourcing, and the method comprises the following steps: step S1, data collection and preprocessing based on built-in sensors of a smartphone; step S2, using a pedestrian dead reckoning method to calculate and obtain pedestrian paths; step S3, path simplification; step S4, path merging; step S5, building a corridor map. The present invention collects sensor data of pedestrians during their daily indoor corridor walking through their smartphones, and then automatically builds an indoor corridor map, which solves the problem of large equipment and personnel expenses in traditional indoor map construction, and has the advantage of automatically updating the map when the indoor environment changes, and is easy to promote and use.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of indoor positioning, and in particular relates to a crowdsourcing-based indoor corridor map construction method and system. Background Art

[0002] Location-based services (LBS) are services related to user location that obtain the location information of mobile terminal users through the operator's wireless network or global positioning system (GPS). With the rise of LBS services, a large number of mobile applications based on LBS services have emerged. For example, positioning services, navigation services, e-commerce and other applications, these applications and services have brought convenience to people's lives.

[0003] Applications based on LBS services have one thing in common, that is, they all need a carrier of location information, which is the common map we see. The creation of maps has a long history, and the production of indoor maps today can be said to be independent of each other. Some continue the traditional outdoor method, collecting data offline through special personnel and equipment to draw maps; some draw maps by obtaining architectural design structure drawings; and some draw maps by encouraging users to actively upload data. These methods have their own advantages and disadvantages, and can solve the problem of indoor map drawing, but they also have their drawbacks. For example, the quality of the map drawn by professionals and equipment is undoubtedly the best, but there are defects in high manpower and financial costs. In addition, due to limited personnel and equipment, the coverage is not wide, and the variability of the indoor environment also brings great trouble to the update of indoor maps.

[0004] Currently, crowdsourcing is a very popular technology, which refers to the practice of a company or organization outsourcing work tasks (usually large tasks) that were previously performed by employees to non-specific mass volunteers on a voluntary basis. Indoor map construction can also be achieved by crowdsourcing. Specifically, crowdsourcing-based indoor map construction refers to collecting trajectory data of people walking indoors through the smartphones of mass volunteers, and then constructing pedestrians' walking trajectory maps and approximate indoor floor plans through a series of algorithms. Indoor maps constructed by crowdsourcing can save the expenses of mapmakers and equipment; maps can be dynamically updated when the indoor environment changes; after the data collection program and map construction program are written, they can be used for different buildings, and there is no need to formulate a separate map construction plan for each building, which is universal. Because the trajectory of indoor rooms may involve privacy, the present invention mainly solves the problem of indoor corridor map construction based on crowdsourcing, that is, the construction of maps of indoor public areas (corridors). Summary of the invention

[0005] In view of the above situation, the present invention proposes a method and system for automatically constructing an indoor corridor map based on crowdsourced data. The method collects sensor data of pedestrians during their daily walking in indoor corridors through their smartphones, and then automatically constructs an indoor corridor map, which solves the problem of high equipment and personnel expenses in traditional indoor map construction. The method also has the advantage of automatically updating the map when the indoor environment changes, and is easy to promote and use.

[0006] The present invention adopts the following technical scheme:

[0007] The crowdsourcing-based indoor corridor map construction method specifically includes the following steps:

[0008] Step S1, data collection and preprocessing based on the built-in sensors of the smartphone.

[0009] Step S2, using a pedestrian dead reckoning method to calculate and obtain a pedestrian path;

[0010] Step S3, path simplification;

[0011] Step S4, path merging;

[0012] Step S5, constructing a corridor map.

[0013] Further, step S1 specifically includes:

[0014] S1.1 Pedestrians hold their smartphones horizontally and walk normally along the corridor area without forming a closed trajectory, and collect three-axis acceleration, gyroscope and magnetometer data at a sampling frequency of 50Hz. The starting point of all paths is consistent, and each corridor path is collected at least N times. Pedestrians press the "Start Collection" button to start collecting data, and press the "End Collection" button to end the collection when they reach the end of the corridor or enter an indoor room.

[0015] S1.2 reads the collected acceleration, gyroscope and magnetometer data, uses a 3rd-order low-pass filter with a cutoff frequency of 0.2π to remove noise from the collected acceleration data, and calculates the three-axis combined acceleration;

[0016]

[0017] where a x (t), a y (t) and a z (t) is the accelerometer three-axis filtered data at time t.

[0018] Furthermore, in step S2, a pedestrian dead reckoning (PDR) method is used to calculate and obtain the pedestrian path, as follows:

[0019] S2.1 PDR consists of three parts: gait detection, step length estimation and direction angle calculation.

[0020] Gait detection is to determine whether a pedestrian is walking based on the characteristic that the acceleration in one step is a continuous positive and negative peak. This step uses the peak detection method to detect gait. Step length estimation is used to estimate the distance between pedestrians' steps. This step uses the following formula to estimate the step length L of the kth step: k .

[0021]

[0022] α, β, γ are parameters, a max,k and a min,k is the peak and valley value of the combined acceleration at step k, f k is the walking frequency, v k is the acceleration variance, satisfying:

[0023]

[0024]

[0025] t k s and t k e Respectively represent the starting time and ending time of the kth step, a t represents the acceleration value in the kth step, represents the mean acceleration of the kth step, N k Indicates the number of acceleration sampling points in the kth step.

[0026] In terms of azimuth calculation, since the gyroscope is not affected by magnetic field and has the characteristics of high accuracy over short distances, this step uses the gyroscope to obtain the change in azimuth within one step:

[0027]

[0028] Among them, θ k g is the change in the direction angle at the kth step, M is the number of gyroscope sampling points at the kth step, ρ i and Δt are the value and sampling interval of the gyroscope Z-axis sampling point, respectively.

[0029] S2.2 obtains the pedestrian path based on the PDR algorithm. The position of the pedestrian at the kth step can be given by the following formula:

[0030]

[0031] Where (x0, y0) is the initial position, θ0 is the initial direction angle, (x k ,y k ) is the k-th step position, and the set of all step coordinates is the pedestrian’s walking path.

[0032] Furthermore, in step S3, the path simplification is specifically as follows:

[0033] S3.1 Path simplification refers to representing the pedestrian path obtained by PDR as an ordered set of key points (starting point, turning point, and end point). The present invention stipulates that the starting point of all paths is the same, and the last step of the path is defined as the end point. The turning point must meet two conditions: 1) The absolute value of the heading angle change at the turning point is the local maximum and is greater than the threshold 2) The absolute value of the sum of the changes in the direction angle within the window with the turning point as the center and the window length as len is greater than the threshold Right now:

[0034]

[0035] Then the kth step is considered to be the step where the turning point is located, so the path can be simplified to:

[0036] Tran i =(i1, i2, ...i n )

[0037] Tran i is the i-th simplified path, i n is the nth key point of the current path, satisfying:

[0038]

[0039] Type(·) represents the type of key point. The starting point, turning point, and end point are assigned values ​​of 0, 1, and 2 respectively.

[0040] S3.2 This step uses the path information matrix Mes i To record the key point number, type, direction angle, position and other information of the simplified path numbered i, thereby obtaining the path information matrix set Mes corresponding to all simplified paths.

[0041] Furthermore, in step S4, path merging is specifically as follows:

[0042] This step uses the square matrix M op To record the connectivity between key points in all paths, the dimension of the matrix is ​​equal to the number of different key points identified. Specifically, when the jth and kth key points are connected (i.e., there is a straight path), M j,k op The value is 1, otherwise it is 0. The specific steps of path merging are as follows:

[0043] S4.1 Initialize the matrix M using the information of the first trajectory Tran1 op .

[0044] S4.2 Take a path information matrix Mes from the set Mes ot , M op With Mes ot To be specific, first detect M op and Mes ot The similarity of the key points in M, if similar, update M op The information of the corresponding key points in (direction angle, spatial position and similar point set LS); if they are not similar, expand M op Square to record Mes ot dissimilar key points in (i.e. updating path connectivity information).

[0045] Furthermore, key point similarity detection is achieved by judging the type, spatial position and direction angle of two key points, that is, it must meet the following conditions:

[0046] P op ∈M op , P ot ∈Mes ot ,satisfy

[0047]

[0048] P op , P ot M op With Mes ot The key points in Mes are respectively used to obtain the type, direction angle, X, and Y coordinate information of the key points in Mes. and is the judgment threshold.

[0049] S4.3 Repeat step S4.2 until Mes is traversed and finally the path matrix M that records the connectivity of all key points is obtained. op (Record it as M opa ), and M opa The similarity point set LS of each key point in , and the updated path information matrix set Mes.

[0050] Furthermore, in step S5, the corridor map is constructed as follows:

[0051] This step divides the corridor map construction into sections. The construction of the corridor map is mainly divided into the following four steps:

[0052] S5.1, M opa The starting point of the path is set to the current key point Cur.

[0053] S5.2, obtain the key point set ConS connected to the current key point Cur Cur , iterate over Cons cur Each key point Con Cur (Con Cur ∈Cons Cur ), based on Cur and Con respectively Cur The similarity point set is extracted between Cur and Con Cur PDR point cloud data within.

[0054] S5.3, for the current point Cur and Con Cur The point cloud data between the corridors is optimized using the key point information in Mes, and the sub-corridors are constructed using the spatial grid method.

[0055] S5.4, update the current key point Cur, and repeat steps S5.2 and S5.3 until the construction of all sub-corridor maps is completed.

[0056] Further, in step S5.2, the point cloud data refers to a set of coordinates of all steps between two key points in the corresponding PDR path. The key point set Cons connected to the current key point Cur Cur By M opa get.

[0057] Furthermore, the specific steps of extracting the point cloud data in step S5.2 are as follows:

[0058] S5.2.1, according to M opa Get the key point set connected to Cur as Cons Cur ;

[0059] S5.2.2, from Cons Cur Take a point Con Cur (Con Cur ∈Cons Cur );

[0060] S5.2.3, according to the points Cur and Con Cur Get the similarity point set of Cur and Con Cur The PDR path number between

[0061] S5.2.4, extract the coordinates of each step between these two points from the corresponding numbered PDR path;

[0062] S5.2.5, repeat steps S5.2.2 to S5.2.4 until all Cons are traversed. Cur .

[0063] Further, in step S5.3, optimizing the point cloud data means scaling and translating the extracted point cloud data using the coordinates of the key points in the updated Mes, so that the subsequently constructed map is more accurate. First, the coordinates of the key points in the PDR path and Mes are obtained, and the distance between the two key points is calculated respectively. The scaling factor α is the inverse of the ratio of the two distances. The point cloud data is scaled using α, and the scaled point cloud data is translated on the coordinate axis by β1 and β2 distances (β1 and β2 are related to the coordinates of the key points in the PDR path and Mes), and the optimized PDR point cloud data is obtained.

[0064] Further, in step S5.3, using the grid method to construct a sub-corridor means covering the corridor with a grid, then counting the valid grids, and finally regularizing the obtained valid grids to obtain a sub-corridor map. A valid grid refers to a grid in which the number of point cloud data falling into the grid exceeds a threshold. Regularization is to fill the constructed valid grid area into a regular area, so that the constructed corridor map is closer to the actual corridor map. The steps are as follows:

[0065] Cur and Con Cur The optimized point cloud data is represented as

[0066] P′={(x k p′ ,y k p′ )|k=1,2,3...m}

[0067] m is the number of point clouds in the corridor. In this step, a square with a side length of l is used as the basic grid to cover the corridor (the grid area is represented by Sq). Assume that the coordinates of the center point of the grid in the i-th row and j-th column are

[0068] Sq i,j =(x i,j ,y i,j )

[0069] Then the coordinates of the four vertices of the grid (Q, R, S, T) are:

[0070]

[0071] If there exists P′ v satisfy:

[0072]

[0073] Then we think that P′ v The point cloud falls into the current grid. If the number of point clouds falling into the current grid exceeds the threshold Thr num , then the grid is considered valid, and all valid grids can be obtained.

[0074] Since there are "holes" in the effective grid area of ​​Sq, in order to make the constructed corridor close to the actual corridor, the present invention regularizes the effective grid in Sq. First, the effective row range in Sq is calculated. min , Row max ] and the effective column range [Col min , Col max ]. Define Ave row is the average number of valid grids in the valid rows of Sq (valid rows refer to rows whose valid grid numbers are not 0), Ave col The average effective grid number of Sq effective columns is:

[0075]

[0076]

[0077] Among them, Count(Sq) represents the number of valid grids in Sq, NumR and NumC are the number of valid rows and columns in Sq respectively. min 、Row max They are respectively satisfied that the number of valid grids exceeds Ave row The minimum and maximum row subscripts of the valid rows, Col min 、Col max They are respectively satisfied that the number of valid grids exceeds Ave col After obtaining the valid row and column range, all grids within the valid range are regarded as valid grids, thus completing the regularization of the valid grid. The regularized valid grid area is Cur and Con Cur The map of the sub-corridor in the room.

[0078] When iterating over Cons Cur Each key point Con Cur After constructing the sub-corridor maps respectively, Cur is continuously updated based on the breadth-first idea, and the above steps are repeated to finally obtain a complete corridor map.

[0079] The present invention also discloses a crowdsourcing-based indoor corridor map construction system, which includes the following modules:

[0080] Data collection and preprocessing module, based on data collection and preprocessing of built-in sensors in smartphones;

[0081] Pedestrian dead reckoning module, which uses pedestrian dead reckoning method to calculate and obtain pedestrian paths;

[0082] A path simplification module, used to simplify the path;

[0083] Path merging module, used to merge paths;

[0084] Map building module, used to build corridor maps.

[0085] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0086] The present invention proposes a new method and system for automatically constructing indoor corridor maps. The indoor corridor map can be constructed by using only sensor data collected by a smart phone. This not only solves the problem of high labor and tool costs in traditional indoor map construction, but also the constructed indoor map can be automatically updated after the indoor environment changes. In general, the technology of the present invention has good practicality and application prospects in reality. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0088] Figure 2 is a simplified schematic diagram of the path;

[0089] Figure 3 It is a schematic diagram of path merging;

[0090] Figure 4 This is a schematic diagram of the corridor construction;

[0091] Figure 5 is a schematic diagram of grid regularization;

[0092] Figure 6 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0093] The present invention will be further described below in conjunction with the embodiments and drawings.

[0094] like Figure 1 As shown, the crowdsourcing-based indoor corridor map construction method proposed in this embodiment includes the following five steps:

[0095] Step S1, data collection and preprocessing based on the built-in sensors of the smartphone;

[0096] Step S2, using pedestrian dead reckoning (PDR) method to calculate and obtain the pedestrian path;

[0097] Step S3, path simplification;

[0098] Step S4, path merging;

[0099] Step S5, constructing a corridor map.

[0100] The specific process of step S1:

[0101] S1.1 Pedestrians hold their smartphones horizontally and walk normally along the corridor area without forming a closed trajectory, and collect three-axis acceleration, gyroscope and magnetometer data at a sampling frequency of 50Hz. The starting point of all paths is consistent, and each corridor path is collected at least N times. Pedestrians press the "Start Collection" button to start data collection, and press the "End Collection" button to end data collection when they reach the end of the corridor or enter an indoor room.

[0102] S1.2 reads the collected acceleration, gyroscope and magnetometer data, uses a 3rd-order low-pass filter with a cutoff frequency of 0.2π to remove noise from the collected acceleration data, and calculates the three-axis combined acceleration;

[0103]

[0104] where a x (t), a y (t) and a z (t) is the accelerometer three-axis filtered data at time t.

[0105] The specific process of step S2:

[0106] S2.1 PDR includes three parts: gait detection, step length estimation and direction angle calculation. Gait detection refers to judging whether a pedestrian is walking based on the characteristic that the acceleration in one step is a continuous positive and negative peak. The present invention detects gait based on the peak detection method. Step length estimation is used to estimate the distance between pedestrians' steps. The following formula is used to dynamically estimate the step length L of the kth step: k .

[0107]

[0108] α, β and γ are parameters, which are set to 0.4, 0.02, and 0.02 respectively. max,k and a min,k is the peak and valley value of the combined acceleration in the kth step, f k is the walking frequency, v k is the acceleration variance, satisfying:

[0109]

[0110]

[0111] t k s and t k e Respectively represent the starting time and ending time of the kth step, a t represents the acceleration value in the kth step, represents the mean acceleration of the kth step, N kIn terms of azimuth angle calculation, since the gyroscope is not disturbed by the magnetic field and has the characteristics of high accuracy over short distances, the present invention uses the gyroscope to obtain the change in azimuth angle within one step:

[0112]

[0113] Among them, θ k g is the change in the direction angle at the kth step, M is the number of gyroscope sampling points at the kth step, ρ i and Δt are the value and sampling interval of the gyroscope Z-axis sampling point, respectively.

[0114] S2.2 obtains the pedestrian path based on the PDR algorithm. The initial direction angle θ can be obtained from the magnetometer s , after knowing the initial position of the pedestrian, the position of each step of the pedestrian can be calculated by the step length and direction angle. The formula is as follows:

[0115]

[0116] Where (x0, y0) is the initial position, θ0 is the initial direction angle, (x k ,y k ) is the k-th step position, and the set of all step coordinates is the pedestrian’s walking path.

[0117] The specific process of step S3:

[0118] S3.1 Path simplification refers to representing the pedestrian path obtained by PDR as an ordered set of key points (starting point, turning point, and end point). The present invention assumes that the starting point of all paths is the same and defines the last step of the path as the end point. The turning point must meet two conditions: 1) The absolute value of the direction angle change at the turning point is the local maximum and is greater than the threshold 2) The absolute value of the sum of the changes in the direction angle within the window with the turning point as the center and the window length as len is greater than the threshold Right now:

[0119]

[0120] Then the kth step is considered to be the step where the turning point is located. len is set to 23, 70 and 4 respectively. Thus a path can be simplified to:

[0121] Tran i =(i1, i2, ...i n )

[0122] Tran i is the i-th simplified path, i nis the nth key point of the current path, satisfying:

[0123]

[0124] Type(·) represents the type of key point. The starting point, turning point, and end point are assigned values ​​of 0, 1, and 2 respectively.

[0125] S3.2 The present invention utilizes the path information matrix Mes i To record the key point number, type, direction angle (turning point records turning angle), position and other information of the simplified path numbered i, the path information matrix set Mes corresponding to all simplified paths is obtained. See the attached schematic diagram Figure 2 .

[0126] The specific process of step S4 is as follows:

[0127] The present invention uses square matrix M op To record the connectivity between key points in all paths, the dimension of the matrix is ​​equal to the number of different key points identified. Specifically, when the jth and kth key points are connected (i.e., there is a straight path), M j,k op The value is 1, otherwise it is 0. The specific steps of path merging are as follows:

[0128] S4.1 Initialize the matrix M using the information of the first trajectory Tran1 op .

[0129] S4.2 Take a path information matrix Mes from the set Mes ot , M op With Mes ot Merge (see attached Figure 3 ). Specifically, first detect the similarity of key points, and if they are similar, update M op Corresponding key point information (direction angle, spatial position and similar point set LS); if not similar, expand M op Square to record Mes ot The pseudo code of path merging is as follows:

[0130]

[0131]

[0132] The above pseudo code shows the process of path merging. Line 1 of the code determines the number of loops to be Mes ot The number of key points included, the code line 3 initializes M op Current processing point set P opl , line 4 of the code traverses Mes otThe key point in the code line 6 is SD(·) to determine P opl Is there a and P ot Similar points, SD(P ot , P opl ) will traverse P in turn opl The key point in P opl If there are key points in , the value is returned as True if the following conditions are met:

[0133]

[0134]

[0135] P ot For Mes ot The key points in Mes are Type(·), Angle(·), PosX(·), and Posy(·), respectively, which are used to obtain the type, direction angle, X, and Y coordinate information of the key points in Mes. The key points are similar only if the key points have the same type and the absolute value of the direction angle difference is less than or greater than (different turns at the same turning point), spatial distance is less than The present invention sets is 20, is 150, is 2.5.

[0136] The RP(·) function in line 7 of the code is used to update the path information matrix Mes op The values ​​of the corresponding similar points in the update similar point set LS, that is:

[0137]

[0138] Where M1 is the number of updates of the corresponding key point information.

[0139] The similar point set refers to the union of each key point and its similar points, which is used for subsequent point cloud data extraction. Figure 3 A1 and B1 are judged to be similar, and A2 and B2 are judged to be similar, so the similarity point set of A1 and A2 is updated. and Line 9 of the code FN(·) is used to update P opl , get the next connected region. Line 11 of the code EM(·) expands M op To record connectivity information, the EM(·) pseudo code is as follows:

[0140]

[0141]

[0142] Attached Figure 3 For schematic diagram, Figure 3 The SD(·) function in the figure determines that A3 and B3 are not similar, so the matrix is ​​expanded. First, the rows are expanded (corresponding to the third line of the code), then the columns are expanded (corresponding to the fifth and twelfth lines of the code), and finally the connectivity information is recorded (corresponding to the sixth and tenth lines of the code). op Dimension records Mes ot The connectivity information in .

[0143] S4.3 Repeat step S4.2 until Mes is traversed and finally the path matrix M that records the connectivity of all key points is obtained. op (Record it as M opa ), and M opa The similarity point set LS of each key point in and the updated path information set Mes.

[0144] The specific process of step S5:

[0145] The present invention divides the construction of the corridor map into the construction of a section of sub-corridor maps, which is mainly divided into the following four steps:

[0146] S5.1, M opa The starting point of the path is set to the current key point Cur.

[0147] S5.2, get the key point set Cons connected to the current key point Cur Cur , iterate over Cons Cur Each key point Con Cur (Con Cur ∈Cons Cur ), based on Cur and Con respectively Cur The similarity point set is extracted between Cur and Con Cur PDR point cloud data within.

[0148] S5.3, for the current point Cur and Con Cur The point cloud data between the corridors is optimized using the key point information in Mes, and the sub-corridors are constructed using the spatial grid method.

[0149] S5.4, update the current key point Cur, and repeat steps S5.2 and S5.3 until the construction of all sub-corridor maps is completed.

[0150] Furthermore, the point cloud data in step S5.2 refers to a set of coordinates of all steps between two key points in the corresponding PDR path. The key point set Cons connected to the current key point Cur Cur By M opaget.

[0151] S5.1 The pseudo code of the corridor construction algorithm is as follows:

[0152]

[0153]

[0154] M opa The starting point of the path is set to the current point Cur, and the algorithm will build a corridor map based on the breadth-first idea. Figure 4 Assume that A1-A2-A3-A4, A1-A2-B3-B4, A1-A2-C3-C4 and A1-A2-D3-D4 are M opa Recorded indoor connectivity information. Based on the breadth-first idea, the algorithm will first construct the corridor {A1, A2} at A1, then construct the corridors {A2, A3}, {A2, B3}, {A2, C3}, {A2, D3} at A2, and construct the corridors {A3, A4}, {B3, B4}, {C3, C4} at A3, B3, and C3 respectively, until S1 is empty, the algorithm ends, and the corridor map is constructed.

[0155] S5.2 Point cloud data extraction. The specific steps are:

[0156] S5.2.1, according to M opa Get the key point set connected to Cur as Cons Cur ;

[0157] S5.2.2, from Cons Cur Take a point Con Cur (Con Cur ∈Cons Cur );

[0158] S5.2.3, according to the points Cur and Con Cur The similarity set of Cur and Con Cur The PDR path number between

[0159] S5.2.4, extract the coordinates of each step between these two points from the corresponding numbered PDR path;

[0160] S5.2.5, repeat steps S5.2.2 to S5.2.4 until all Cons are traversed. Cur .

[0161] The connected set Cons of the above Cur Cur M opa In step S5.2.3, we first obtain the similarity point set LS of Cur. Cur,Con Cur The similarity set of Since the key point number records the corresponding PDR path number and the sequence number of the key point in the path, the corresponding point cloud data can be extracted based on this. Cur The path number contained in is represented as Un Cur , The path number contained in is represented as Then the set of path numbers passing through this corridor is Un Cur and The intersection of Using LS cur and The step number information of the corresponding key point in the PDR path can be used to extract the point cloud data between the two key points in the corresponding PDR path.

[0162] Attach Figure 4 For example, when constructing the corridor {A1, A2} segment, the key point connected by A1 is Cons Cur For A2, assume that the similarity points of A1 and A2 are:

[0163]

[0164] Depend on It can be seen that there are four paths (Un Cur ={A, B, C, D}) through A1, by It can be seen that there are numbers If there are three paths passing through A2, then the three paths numbered A, B, and C contain the point cloud data required for the corridor {A1, A2}. The set of path numbers passing through this corridor is:

[0165]

[0166] After obtaining the PDR path number passing through the corridor, the point cloud data in the corresponding PDR path can be extracted using the step number information of the key points. The point cloud data obtained from the above corridor {A1, A2} segment can be expressed as:

[0167]

[0168] in:

[0169]

[0170]

[0171]

[0172] Represents the set of coordinates of each step of the PDR path numbered A within the key points A1 and A2. is the step number of key point A1, Info A Represents the set of coordinates for each step in the PDR path numbered A. Similarly, the point cloud data set P of the corridor {A1, A2} segment is obtained.

[0173] S5.3 Optimize point cloud data and construct sub-corridor maps.

[0174] S5.3.1 Optimize point cloud data. Optimizing point cloud data means scaling and translating the extracted point cloud data using the coordinates of the key points in the updated Mes, so that the subsequently constructed map is more accurate. First, the coordinates of the key points in the PDR path and Mes are obtained, and the distance between the two key points is calculated respectively. The scaling factor α is the inverse of the ratio of the two distances. α is used to scale the point cloud data, and the scaled point cloud data is translated on the coordinate axis by β1 and β2 distances (β1 and β2 are related to the coordinates of the key points in the PDR path and Mes), and the optimized PDR point cloud data is obtained.

[0175] To optimize the obtained For example, first get the coordinates of the key points in Mes and calculate the distance between the two points:

[0176]

[0177]

[0178]

[0179] Then get the coordinates of the key points in the PDR path and calculate the distance between the two points:

[0180]

[0181]

[0182]

[0183] The optimized It is expressed as:

[0184]

[0185]

[0186] Where (x A ,y A )for The original point cloud data can be optimized in the same way. The point cloud data is finally obtained, and the optimized point cloud set P′ is obtained.

[0187] S5.3.2 Sub-corridor map construction. Sub-corridor map construction refers to covering the corridor with grids, then counting the valid grids, and finally regularizing the obtained valid grids to finally obtain the sub-corridor map. Valid grids refer to grids in which the number of point cloud data falling into the grid exceeds the threshold. Regularization is to fill the constructed valid grid area into a regular area, so that the constructed corridor map is closer to the actual corridor map. See below for detailed steps:

[0188] The optimized point cloud data in the same sub-corridor is represented as

[0189] P′={(x k p′ ,y k p′ )|k=1,2,3...m}

[0190] m is the number of point clouds in the corridor. The present invention uses a square with a side length of l as the basic grid to cover the corridor (the grid area is represented by Sq). Assume that the coordinates of the center point of the grid in the i-th row and j-th column are

[0191]

[0192] Then the coordinates of the four vertices (Q, R, S, T) of the mesh are:

[0193]

[0194] If there exists P′ v satisfy:

[0195]

[0196] Then we think that P′ v The point cloud falls into the current grid. If the number of point clouds falling into the current grid exceeds the threshold Thr num , then the grid is considered to be valid, and all valid grids can be obtained. Since there are "holes" in the valid grid area of ​​Sq, in order to make the constructed corridor close to the actual corridor, the present invention regularizes it. Regularization requires calculating the valid row range of Sq [Row min , Row max ] and the effective column range [Col min , Col max ], the pseudo code is as follows:

[0197]

[0198]

[0199] Ave row is the average number of valid grids in the valid rows of Sq (valid rows refer to rows whose valid grid numbers are not 0), Ave col The average effective grid number of Sq effective columns is:

[0200]

[0201]

[0202] Among them, Count(Sq) represents the number of valid grids in Sq, NumR and NumC are the number of valid rows and columns in Sq respectively. min 、Row max They are respectively satisfied that the number of valid grids exceeds Ave row The minimum and maximum row subscripts of the valid rows, Col min 、Col max They are respectively satisfied that the number of valid grids exceeds Ave col After obtaining the valid row and column range, all grids within the valid range are regarded as valid grids, thereby completing the regularization of the effective grid. The regularized effective grid area is Cur and Con Cur See the attached map of the sub-corridor Figure 5 .

[0203] S5.4 Traversing Cons Cur Each key point Con Cur After constructing the sub-corridor maps respectively, Cur is continuously updated based on the breadth-first idea, and the above steps are repeated to finally obtain a complete corridor map.

[0204] like Figure 6 As shown, this embodiment discloses a crowdsourcing-based indoor corridor map construction system, which specifically includes the following modules:

[0205] Data collection and preprocessing module, based on data collection and preprocessing of built-in sensors in smartphones.

[0206] Pedestrian dead reckoning module, which uses pedestrian dead reckoning method to calculate and obtain pedestrian paths;

[0207] Path simplification module, path simplification;

[0208] Path merging module, path merging;

[0209] Map construction module, builds corridor map.

[0210] In this embodiment, the data collection and preprocessing module is as follows:

[0211] S1.1 Pedestrians hold their smartphones horizontally and walk normally along the corridor area without forming a closed trajectory, and collect three-axis acceleration, gyroscope and magnetometer data at a sampling frequency of 50Hz. The starting point of all paths is consistent, and each corridor path is collected at least N times. Pedestrians press the "Start Collection" button to start collecting data, and press the "End Collection" button to end the collection when they reach the end of the corridor or enter an indoor room.

[0212] S1.2 reads the collected acceleration, gyroscope and magnetometer data, uses a 3rd-order low-pass filter with a cutoff frequency of 0.2π to remove noise from the collected acceleration data, and calculates the three-axis combined acceleration;

[0213]

[0214] where a x (t), a y (t) and a z (t) is the accelerometer three-axis filtered data at time t.

[0215] In this embodiment, the pedestrian dead reckoning module uses the pedestrian dead reckoning (PDR) method to calculate and obtain the pedestrian path, as follows:

[0216] S2.1PDR includes three parts: gait detection, step length estimation and direction angle calculation. Gait detection refers to judging whether a pedestrian is walking based on the characteristic that the acceleration in one step is a continuous positive and negative peak. The present invention uses the peak detection method to detect gait. Step length estimation is used to estimate the distance between pedestrians' steps. The present invention uses the following formula to estimate the step length L of the kth step: k .

[0217]

[0218] α, β, γ are parameters, a max,k and a min,k is the peak and valley value of the combined acceleration at step k, f k is the walking frequency, v k is the acceleration variance, satisfying:

[0219]

[0220]

[0221] t k s and t k e Respectively represent the starting time and ending time of the kth step, a t represents the acceleration value in the kth step, represents the mean acceleration of the kth step, N kIndicates the number of acceleration sampling points in the kth step

[0222] In terms of azimuth calculation, since the gyroscope is not affected by magnetic field interference and has the characteristics of high accuracy over short distances, the present invention uses the gyroscope to obtain the change in azimuth within one step:

[0223]

[0224] Among them, θ k g is the change in the direction angle at the kth step, M is the number of gyroscope sampling points at the kth step, ρ i and Δt are the value and sampling interval of the gyroscope Z-axis sampling point, respectively.

[0225] S2.2 obtains the pedestrian path based on the PDR algorithm. The position of the pedestrian at the kth step can be given by the following formula:

[0226]

[0227] Where (x0, y0) is the initial position, θ0 is the initial direction angle, (x k ,y k ) is the k-th step position, and the set of all step coordinates is the pedestrian’s walking path.

[0228] In this embodiment, the path simplification module is specifically as follows:

[0229] S3.1 Path simplification refers to representing the pedestrian path obtained by PDR as an ordered set of key points (starting point, turning point, and end point). The present invention stipulates that the starting point of all paths is the same, and the last step of the path is defined as the end point. The turning point must meet two conditions: 1) The absolute value of the heading angle change at the turning point is the local maximum and is greater than the threshold 2) The absolute value of the sum of the changes in the direction angle within the window with the turning point as the center and the window length as len is greater than the threshold Right now:

[0230]

[0231] Then the kth step is considered to be the step where the turning point is located, so the path can be simplified to:

[0232] Tran i =(i1, i2, ...i n )

[0233] Tran i is the i-th simplified path, i n is the nth key point of the current path, satisfying:

[0234]

[0235] Type(·) represents the type of key point. The starting point, turning point, and end point are assigned values ​​of 0, 1, and 2 respectively.

[0236] S3.2 The present invention utilizes the path information matrix Mes i To record the key point number, type, direction angle, position and other information of the simplified path numbered i, thereby obtaining the path information matrix set Mes corresponding to all simplified paths.

[0237] In this embodiment, the path merging module is specifically as follows:

[0238] The present invention uses square matrix M op To record the connectivity between key points in all paths, the dimension of the matrix is ​​equal to the number of different key points identified. Specifically, when the jth and kth key points are connected (i.e., there is a straight path), M j,k op The value is 1, otherwise it is 0. The specific steps of path merging are as follows:

[0239] S4.1 Initialize the matrix M using the information of the first trajectory Tran1 op .

[0240] S4.2 Take a path information matrix Mes from the set Mes ot , M op With Mes ot Specifically, first detect M op and Mes ot The similarity of the key points in M, if similar, update M op The information of the corresponding key points in (direction angle, spatial position and similar point set LS); if they are not similar, expand M op Square to record Mes ot dissimilar key points in (i.e. updating path connectivity information).

[0241] Furthermore, key point similarity detection is achieved by judging the type, spatial position and direction angle of two key points, that is, it must meet the following conditions:

[0242] P op ∈M op , P ot ∈Mes ot ,satisfy

[0243]

[0244] P op , P ot M op With Mesot The key points in Mes are respectively used to obtain the type, direction angle, X, and Y coordinate information of the key points in Mes. and T is the judgment threshold.

[0245] S4.3 Repeat step S4.2 until Mes is traversed and finally the path matrix M that records the connectivity of all key points is obtained. op (Record it as M opa ), and M opa The similarity point set LS of each key point in , and the updated path information matrix set Mes.

[0246] In this embodiment, the map construction module is specifically as follows:

[0247] The present invention divides the construction of the corridor map into the construction of a section of sub-corridor maps, which is mainly divided into the following four steps:

[0248] S5.1, M opa The starting point of the path is set to the current key point Cur.

[0249] S5.2, get the key point set Cons connected to the current key point Cur Cur , iterate over Cons Cur Each key point Con Cur (Con Cur ∈Cons Cur ), based on Cur and Con respectively cur The similarity point set is extracted between Cur and Con cur PDR point cloud data within.

[0250] S5.3, for the current point Cur and Con Cur The point cloud data between the corridors is optimized using the key point information in Mes, and the sub-corridors are constructed using the spatial grid method.

[0251] S5.4, update the current key point Cur, and repeat steps S5.2 and S5.3 until the construction of all sub-corridor maps is completed.

[0252] Furthermore, the point cloud data in step S5.2 refers to a set of coordinates of all steps between two key points in the corresponding PDR path. The key point set Cons connected to the current key point Cur Cur By M opa get.

[0253] Furthermore, the specific steps of extracting the point cloud data in step S5.2 are as follows:

[0254] S5.2.1, according to M opa Get the key point set connected to Cur as Cons Cur ; S5.2.2, from ConS Cur Take a point Con Cur (Con Cur ∈Cons Cur );

[0255] S5.2.3, according to the points Cur and Con Cur Get the similarity point set of Cur and Con Cur The PDR path number between

[0256] S5.2.4, extract the coordinates of each step between these two points from the corresponding numbered PDR path;

[0257] S5.2.5, repeat steps S5.2.2 to S5.2.4 until all Cons are traversed. Cur .

[0258] Furthermore, the optimized point cloud data in step S5.3 refers to scaling and translating the extracted point cloud data using the coordinates of the key points in the updated Mes, so that the subsequently constructed map is more accurate. First, the coordinates of the key points in the PDR path and Mes are obtained, and the distance between the two key points is calculated respectively. The scaling coefficient α is the inverse of the ratio of the two distances. The point cloud data is scaled using α, and the scaled point cloud data is translated on the coordinate axis by β1 and β2 distances (β1 and β2 are related to the coordinates of the key points in the PDR path and Mes), and the optimized PDR point cloud data is obtained.

[0259] Furthermore, the construction of the sub-corridor using the grid method in step S5.3 refers to covering the corridor with a grid, then counting the valid grids, and finally regularizing the obtained valid grids to obtain a sub-corridor map. A valid grid refers to a grid in which the number of point cloud data falling into the grid exceeds a threshold. Regularization is to fill the constructed valid grid area into a regular area, so that the constructed corridor map is closer to the actual corridor map. The steps are as follows:

[0260] Cur and Con Cur The optimized point cloud data is represented as

[0261] P′={(x k p′ ,y k p′ )|k=1,2,3...m}

[0262] m is the number of point clouds in the corridor. The present invention uses a square with a side length of l as the basic grid to cover the corridor (the grid area is represented by Sq). Assume that the coordinates of the center point of the grid in the i-th row and j-th column are

[0263]

[0264] Then the coordinates of the four vertices of the grid (Q, R, S, T) are:

[0265]

[0266] If there exists P′ v satisfy:

[0267]

[0268] Then we think that P′ v The point cloud falls into the current grid. If the number of point clouds falling into the current grid exceeds the threshold Thr num , then the grid is considered valid, and all valid grids can be obtained.

[0269] Since there are "holes" in the effective grid area of ​​Sq, in order to make the constructed corridor close to the actual corridor, the present invention regularizes the effective grid in Sq. First, the effective row range in Sq is calculated. min , Row max ] and the effective column range [Col min , CoI max ]. Define Ave row is the average number of valid grids in the valid rows of Sq (valid rows refer to rows whose valid grid numbers are not 0), Ave col The average effective grid number of Sq effective columns is:

[0270]

[0271]

[0272] Among them, Count(Sq) represents the number of valid grids in Sq, NumR and NumC are the number of valid rows and columns in Sq respectively. min 、Row max They are respectively satisfied that the number of valid grids exceeds Ave row The minimum and maximum row subscripts of the valid rows, Col min 、Col max They are respectively satisfied that the number of valid grids exceeds Ave colAfter obtaining the valid row and column range, all grids within the valid range are regarded as valid grids, thus completing the regularization of the valid grid. The regularized valid grid area is Cur and Con Cur The map of the sub-corridor in the room.

[0273] When iterating over Cons Cur Each key point Con Cur After constructing the sub-corridor maps respectively, Cur is continuously updated based on the breadth-first idea, and the above steps are repeated to finally obtain a complete corridor map.

[0274] The above embodiments are preferred implementations of the present invention, but the implementations of the present invention are not limited to the above embodiments. Components and specific algorithms not specified in the present invention can be implemented using existing technologies.

[0275] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. A crowdsourcing-based indoor corridor map construction method, characterized by The steps include: Step S1, data collection and preprocessing based on the built-in sensors of the smartphone; Step S2, using a pedestrian dead reckoning method to calculate and obtain a pedestrian path; Step S3, path simplification; Step S4, path merging; Step S5, constructing a corridor map; In step S4, use the square matrix M op Record the connectivity between key points in all paths. The dimension of the matrix is ​​equal to the number of different key points identified. Specifically, when the jth and kth key points are connected, there is a straight path. j,k op The value is 1, otherwise it is 0; the specific steps of path merging are as follows: S4.1, initialize the matrix M using the information of the first trajectory Tran1 op ; S4.2, take out a path information matrix Mes from the set Mes ot , M op With Mes ot To merge; specifically, first detect M op and Mes ot The similarity of the key points in M, if similar, update M op The information of the corresponding key points in , including the direction angle, spatial position and similar point set LS; If not similar, expand M op Square to record Mes ot The dissimilar key points in the update path connectivity information; Key point similarity detection is achieved by judging the type, spatial position and direction angle of two key points, that is, it must meet the following requirements: P op ∈M op ,P ot ∈Mes ot P op , P ot M op With Mes ot The key points in Mes, Type(·), Angle(·), PosX(·), and PosY(·) are used to obtain the type, direction angle, X, and Y coordinate information of the key points in Mes, respectively. and is the judgment threshold; S4.3, repeat step S4.2 until Mes is traversed, and finally the path matrix M that records the connectivity of all key points is obtained op , record it as M opa , and M opa The similarity point set LS of each key point in the , and the updated path information matrix set Mes; Step S5 is specifically as follows: S5.1, M opa The starting point of the middle path is set to the current key point Cur; S5.2, get the key point set Cons connected to the current key point Cur Cur , iterate over Cons Cur Each key point Con Cur ,Con Cur ∈Cons Cur , based on Cur and Con respectively Cur The similarity point set is extracted between Cur and Con Cur PDR point cloud data within; S5.3, for the current point Cur and Con Cur The point cloud data between the corridors are optimized by using the key point information in Mes, and the sub-corridors are constructed respectively by using the spatial grid method; S5.4, update the current key point Cur, and repeat steps S5.2 and S5.3 until the construction of all sub-corridor maps is completed.

2. The crowdsourcing-based indoor corridor map construction method according to claim 1, characterized in that: Step S1 specifically includes: S1.1, pedestrians walk horizontally along the corridor area without forming a closed trajectory while holding a smartphone, and collect three-axis acceleration, gyroscope and magnetometer data; the starting point of all paths is consistent, and each corridor path is collected at least N times, N ≥ 8; S1.2, read the collected acceleration, gyroscope and magnetometer data, use a low-pass filter to remove noise in the collected acceleration data, and calculate the three-axis combined acceleration; Among them, a x (t), a y (t) and a z (t) is the accelerometer three-axis filtered data at time t.

3. The crowdsourcing-based indoor corridor map construction method according to claim 2, characterized in that: Step S2 is specifically as follows: S2.1, pedestrian dead reckoning methods include gait detection, step length estimation, and direction angle calculation; Gait was detected using the peak detection method; Estimate the step length L of the kth step using the following formula: k , α, β, γ are parameters, a max,k and a min,k is the peak and valley value of the combined acceleration at step k, f k is the walking frequency, v k is the acceleration variance, satisfying: t k s and t k e Respectively represent the starting time and ending time of the kth step, a t represents the acceleration value in the kth step, represents the mean acceleration of the kth step, N k Indicates the number of acceleration sampling points in the kth step; Use the gyroscope to get the change in direction angle within one step: Among them, θ k g is the change in the direction angle at the kth step, M is the number of gyroscope sampling points at the kth step, ρ i and Δt are the value and sampling interval of the gyroscope Z-axis sampling point, respectively; S2.2, based on the pedestrian dead reckoning method, the pedestrian path is obtained, and the position of the pedestrian at the kth step is given by the following formula: Among them, (x0, y0) is the initial position, θ0 is the initial direction angle, (x k ,y k ) is the k-th step position, and the set of all step coordinates is the pedestrian’s walking path.

4. The crowdsourcing-based indoor corridor map construction method according to claim 3, characterized in that: Step S3 is as follows: S3.1, the pedestrian path obtained by the pedestrian dead reckoning method is represented as an ordered set consisting of only the starting point, turning point, and end point; the starting point of all paths is defined as the same, and the last step of the path is defined as the end point; the turning point must meet two conditions: 1) the absolute value of the heading angle change at the step where the turning point is located is the local maximum and is greater than the threshold 2) The absolute value of the sum of the changes in the direction angle within the window with the turning point as the center and the window length as len is greater than the threshold Right now: Then the kth step is considered to be the step where the turning point is located, and thus the path is simplified to: Tran i =(i1,i2,…i n ) Tran i is the i-th simplified path, i n is the nth key point of the current path, satisfying: Type(·) represents the type of key point, and the starting point, turning point, and end point are assigned values ​​of 0, 1, and 2 respectively; S3.2, using the path information matrix Mes i The key point number, type, direction angle, and position information of the simplified path numbered i are recorded, thereby obtaining the path information matrix set Mes corresponding to all simplified paths.

5. The crowdsourcing-based indoor corridor map construction method according to claim 1, characterized in that: The specific steps of extracting point cloud data in step S5.2 are as follows: S5.2.1, according to M opa Get the key point set connected to Cur as Cons Cur ; S5.2.2, from Cons Cur Take a point Con Cur ,Con Cur ∈Cons Cur ; S5.2.3, according to the points Cur and Con Cur Get the similarity point set of Cur and Con Cur The PDR path number between S5.2.4, extract the coordinates of each step between these two points from the corresponding numbered PDR path; S5.2.5, repeat steps S5.2.2 to S5.2.4 until all Cons are traversed. Cur .

6. The crowdsourcing-based indoor corridor map construction method according to claim 1, characterized in that: In step S5.3, first obtain the coordinates of the key points in the PDR path and Mes, and calculate the distance between the two key points respectively. The scaling factor α is the inverse of the ratio of the two distances. α is used to scale the point cloud data, and the scaled point cloud data is translated on the coordinate axis by distances β1 and β2 respectively to obtain the optimized PDR point cloud data.

7. The crowdsourcing-based indoor corridor map construction method according to claim 1, 5 or 6, characterized in that the step S5.3 The specific steps are as follows: Cur and Con Cur The optimized point cloud data is represented as P′={(x k p′ ,y k p′ )|k=1,2,3…m} m is the number of point clouds in the corridor. A square with a side length of l is used as the basic grid to cover the corridor. The grid area is represented by Sq. Assume that the coordinates of the center point of the grid in the i-th row and j-th column are Sq i,j (x) i,j ,and i,j ) Then the coordinates of the four vertices of the grid (Q, R, S, T) are: If there is P v 'satisfy: It is believed that P v ′The point cloud falls into the current grid; if the number of point clouds falling into the current grid exceeds the threshold Thr num , then the grid is considered valid, and all valid grids are obtained; Regularize the valid grid in Sq. First, calculate the valid row range in Sq. min ,Row max ] and the effective column range [Col min ,Col max ]; define Ave row is the average number of effective grids in the effective row of Sq, Ave col The average effective grid number of Sq effective columns is: Among them, Count(Sq) represents the number of valid grids in Sq, NumR and NumC represent the number of valid rows and columns in Sq respectively; then Row min 、Row max They are respectively satisfied that the number of valid grids exceeds Ave row The minimum and maximum row subscripts of the valid rows, Col min 、Col max They are respectively satisfied that the number of valid grids exceeds Ave col The minimum and maximum column subscripts of the valid columns; after obtaining the valid row and column range, all grids within the valid range are regarded as valid grids, thus completing the regularization of the effective grid. The regularized effective grid area is Cur and Con Cur The sub-corridor map of the room; When iterating over Cons Cur Each key point Con Cur After constructing the sub-corridor maps respectively, Cur is continuously updated based on the breadth-first idea, and the above steps are repeated to finally obtain a complete corridor map.

8. A crowdsourcing-based indoor corridor map construction system, characterized by Includes the following modules: Data collection and preprocessing module, based on data collection and preprocessing of built-in sensors in smartphones; Pedestrian dead reckoning module, which uses pedestrian dead reckoning method to calculate and obtain pedestrian paths; A path simplification module, used to simplify the path; Path merging module, used to merge paths; Map building module, used to build corridor maps; In path merging, use the square matrix M op Record the connectivity between key points in all paths. The dimension of the matrix is ​​equal to the number of different key points identified. Specifically, when the jth and kth key points are connected, there is a straight path. j,k op The value is 1, otherwise it is 0; the specific steps of path merging are as follows: S4.1, initialize the matrix M using the information of the first trajectory Tran1 op ; S4.2, take out a path information matrix Mes from the set Mes ot , M op With Mes ot To merge; specifically, first detect M op and Mes ot If the similarity is similar, update M op The information of the corresponding key points in , including the direction angle, spatial position and similar point set LS; If not similar, expand M op Square to record Mes ot The dissimilar key points in the update path connectivity information; Key point similarity detection is achieved by judging the type, spatial position and direction angle of two key points, that is, it must meet the following requirements: P op ∈M op ,P ot ∈Mes ot P op , P ot M op With Mes ot The key points in Mes, Type(·), Angle(·), PosX(·), and PosY(·) are used to obtain the type, direction angle, X, and Y coordinate information of the key points in Mes, respectively. and is the judgment threshold; S4.3, repeat step S4.2 until Mes is traversed, and finally the path matrix M that records the connectivity of all key points is obtained op , record it as M opa , and M opa The similarity point set LS of each key point in the , and the updated path information matrix set Mes; The construction of the corridor map is as follows: S5.1, M opa The starting point of the middle path is set to the current key point Cur; S5.2, get the key point set Cons connected to the current key point Cur Cur , iterate over Cons Cur Each key point Con Cur ,Con Cur ∈Cons Cur , based on Cur and Con respectively Cur The similarity point set is extracted between Cur and Con Cur PDR point cloud data within; S5.3, for the current point Cur and Con Cur The point cloud data between the corridors are optimized by using the key point information in Mes, and the sub-corridors are constructed respectively by using the spatial grid method; S5.4, update the current key point Cur, and repeat steps S5.2 and S5.3 until the construction of all sub-corridor maps is completed.