Indoor path-constraint-free scene pedestrian trajectory positioning method

Abstract

The invention provides an indoor path-constraint-free scene pedestrian trajectory positioning method, and belongs to the technical field of positioning navigation application. The method comprises the following steps: constructing a pedestrian step number minimum matching group based on data collected by a mobile terminal; obtaining a frequent walking pedestrian track from the collected data, and carrying out pedestrian track credibility judgment; when the trajectory matching result is not credible, performing virtual information source assisted pedestrian trajectory scatter processing and clustering; and based on the scattered track fingerprint point clustering result, carrying out coordinate estimation on the to-be-determined point. According to the method, accurate positioning of an indoor path-constraint-free scene can be realized, the positioning accuracy of an indoor complex environment can be ensured, the positioning cost is reduced, and the method has good universality and important market value.

Description

Technical Field

[0001] The present invention relates to the field of positioning and navigation application technology, and in particular to a method for positioning pedestrian trajectories in an indoor scene without path constraints. Background Art

[0002] With the popularization of smart devices and the development of Internet of Things technology, indoor positioning technology has become a hot topic of research. Due to the obstruction of buildings in indoor environments, traditional global positioning systems (GPS) cannot provide accurate location information. However, indoor positioning technology has important application value in shopping mall navigation, emergency rescue, asset management, smart office and other fields. For example, indoor positioning technology in large shopping malls and shopping centers can help customers quickly find target stores, restaurants or service facilities; push information such as nearby product discounts and new product launches based on customer locations; and quickly locate and evacuate people in emergency situations such as fires and earthquakes.

[0003] Accurate positioning in indoor scenarios without path constraints is an important part of indoor pedestrian positioning technology. In indoor environments such as large shopping malls, airports, and museums, traditional positioning methods based on path matching often do not work well due to the lack of clear path constraints. In addition, the propagation of wireless signals in these scenarios is complex and susceptible to factors such as multipath effects and non-line-of-sight propagation, which further increases the difficulty of positioning.

[0004] At present, indoor positioning technology is mainly divided into two categories: one is the positioning technology based on location fingerprint, and the other is the positioning technology based on the arrival time, arrival time difference and arrival angle of wireless signals. Location fingerprint positioning mainly performs clustering and segmentation based on the regional characteristic differences of indoor signal maps, including geometric location fingerprint clustering and signal distance fingerprint clustering. However, the fingerprint clustering algorithm based on geometric location analyzes the geometric proximity of fingerprint points in the offline stage. The number of clusters needs to be directly or indirectly defined in advance. Users need to carefully adjust the parameters to select the appropriate number of clusters, which is prone to the defect of inappropriate number of clusters leading to positioning deviation; the fingerprint clustering algorithm based on signal distance clusters according to the signal spatial distance of wireless signals. There is a problem that the RSSI signal spatial distance is close but the actual geometric distance is far, which often leads to the wrong selection of some adjacent fingerprint points and ultimately affects the positioning accuracy. The positioning technology based on the arrival time, arrival time difference and arrival angle of wireless signals requires precise clock synchronization and multiple signal sources, which is more complicated to implement and will be affected by signal fluctuations, non-line-of-sight propagation and multipath propagation, resulting in low positioning accuracy.

[0005] Therefore, how to achieve accurate positioning in existing indoor scenarios without path constraints has become a difficulty that needs to be overcome urgently in the field of indoor positioning. Summary of the invention

[0006] In order to overcome the problems existing in the related art, the purpose of the present invention is to provide a method for locating pedestrian trajectories in indoor scenarios without path constraints. The method can achieve accurate positioning in indoor scenarios without path constraints, ensure the positioning accuracy in complex indoor environments, reduce the positioning cost, and has good versatility and important market value.

[0007] A method for locating pedestrian trajectories in an indoor scene without path constraints, comprising:

[0008] Based on the data collected by the mobile terminal, a minimum matching group of pedestrian steps is constructed;

[0009] From the collected data, the common pedestrian trajectories are obtained to determine the credibility of the pedestrian trajectories;

[0010] When the trajectory matching result is unreliable, the pedestrian trajectory is scattered and clustered with the assistance of virtual information sources;

[0011] Based on the clustering results of the scattered trajectory fingerprint points, the coordinates of the unknown points are estimated.

[0012] In a preferred technical solution of the present invention, the step of constructing a minimum matching group of pedestrian steps based on data collected by a mobile terminal includes:

[0013] In indoor environments, mobile terminals are used to collect AP data and sensor data of multiple pedestrian trajectories;

[0014] Directly read the number of pedestrian steps based on the step counting data of the mobile terminal, and build the minimum matching group of pedestrian steps with 3 steps as a group, taking the pedestrian's current position as the center;

[0015] Using AP data, calculate the average Wi-Fi signal strength (RSSI) of the minimum matching group of pedestrian steps.

[0016] In a preferred technical solution of the present invention, the process of obtaining the frequently-walked pedestrian trajectory from the collected data and judging the credibility of the pedestrian trajectory includes:

[0017] One of the collected multiple pedestrian trajectories is used as the current pedestrian trajectory, and the remaining pedestrian trajectories are used as the common pedestrian trajectories;

[0018] Match the current pedestrian trajectory with all the frequently-walked pedestrian trajectories one by one; calculate the cosine similarity between the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectories, and determine whether the cosine similarity exceeds the set threshold T;

[0019] When the cosine similarity is greater than or equal to the set threshold T, it means that the matching result is credible and trajectory matching is used for positioning; when the cosine similarity is less than the set threshold T, it means that the matching result is unreliable.

[0020] In a preferred technical solution of the present invention, the step of calculating the cosine similarity between the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectory and determining whether the cosine similarity exceeds a set threshold value T includes:

[0021] For each frequently walked pedestrian trajectory, a sliding window of 3 steps is constructed with the minimum matching group of pedestrian steps as the unit;

[0022] In each step count sliding window, the pedestrian trajectory step group sequence vector is calculated based on the accelerometer and gyroscope data collected by the mobile terminal sensor;

[0023] The cosine similarity is calculated one by one using the current pedestrian trajectory and the pedestrian trajectory step group sequence vectors of all common pedestrian trajectories. The calculation formula is as follows:

[0024]

[0025] Among them, a represents the step group sequence vector of the current pedestrian trajectory, and b represents the step group sequence vector of the usual pedestrian trajectory;

[0026] According to the relationship between the cosine similarity of the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectory, it is judged whether the cosine similarity exceeds the set threshold T to determine whether the matching result is credible.

[0027] In one implementation, a threshold T=0.5 is set. When the similarity is greater than the threshold T, it indicates that the matching result is credible and trajectory matching is used for positioning; when the similarity is less than the threshold T, it indicates that the matching result is unreliable.

[0028] In a preferred technical solution of the present invention, when the trajectory matching result is unreliable, the pedestrian trajectory scattering and clustering assisted by a virtual information source includes:

[0029] Taking the minimum matching group of pedestrian steps as the unit, the three steps of pedestrians are combined into one point to scatter the pedestrian trajectory. According to the wireless signal attenuation model, the distance from each point to the virtual AP is calculated;

[0030] Use the KNN algorithm to find the K neighboring fingerprint points corresponding to the test point; combine the linear least squares adjustment method to use these K neighboring fingerprint points to calculate the virtual coordinates of the AP;

[0031] For each AP, the distance between its virtual coordinates is calculated; the virtual coordinates are clustered using the Mean-sh i ft clustering algorithm to form the fingerprint point clustering result.

[0032] In a preferred technical solution of the present invention, the minimum matching group of pedestrian steps is used as a unit, three steps of pedestrian steps are combined into one point, pedestrian trajectory is scattered, and the distance from each point to the virtual AP is calculated according to the wireless signal attenuation model, including:

[0033] Taking the minimum matching group of pedestrian steps as the unit, three steps are combined into one point, and the average RSS I of the three steps is calculated as the RSS I value of the point;

[0034] According to the wireless signal attenuation model, the Wi-Fi RSSI attenuation formula is used to calculate the distance from each point to the virtual AP. The calculation formula is as follows:

[0035]

[0036]

[0037] Among them, RSS I(d0) and RSS I(d i ) are the distances d0 and d from the AP signal source respectively. i The received signal strength at the location, η is the path attenuation factor, d0 and RSS I(d0) are pre-set through empirical modeling, d i Indicates the distance from the i-th point to the virtual AP.

[0038] In a preferred technical solution of the present invention, the method of finding K neighboring fingerprint points corresponding to the test point using the KNN algorithm and calculating the virtual coordinates of the AP using the K neighboring fingerprint points in combination with the linear least squares adjustment method includes:

[0039] Use the KNN algorithm to obtain the K neighboring fingerprint points corresponding to the test point;

[0040] Using these K neighboring fingerprint points, combined with linear least squares adjustment, the virtual coordinates of AP are calculated using the following formula:

[0041]

[0042] Among them, (x j ,y j ) are unknown coordinate parameters, B is the coefficient matrix, are the virtual AP coordinates after adjustment.

[0043] In a preferred technical solution of the present invention, the coordinate estimation of the undetermined point based on the clustering result of the trajectory fingerprint points after scattering includes:

[0044] According to the fingerprint point clustering results, coordinate matching of points is performed; one test point is selected, the average RSSD difference between the test point and the fingerprint points in each group of clusters in the virtual AP is calculated, the test point is matched to the cluster with the smallest average distance value, and the test point is matched with different virtual APs one by one;

[0045] According to the point matching results of different APs, the coordinates of each point are estimated; it is determined whether the point matching results of different virtual APs have intersections, and then the estimated coordinates of the points are calculated using the WKNN algorithm;

[0046] Connect the estimated point coordinates into lines in sequence to form a new pedestrian walking trajectory;

[0047] In a preferred technical solution of the present invention, the method of selecting a test point, calculating the average RSSD difference between the test point and the fingerprint points in each group of clusters in the virtual AP, matching the test point to the cluster with the smallest average distance value, and matching different virtual APs one by one includes:

[0048] Based on the AP data of the test point, select the top 6 virtual APs with the highest signal strength.

[0049] For each virtual AP, calculate the average RSSD (received signal strength difference) difference between the test point and each fingerprint point cluster, and match the test point to a fingerprint point cluster with the smallest RSSD difference average value;

[0050] Repeat the above process for all selected virtual APs to complete all matching.

[0051] In a preferred technical solution of the present invention, the coordinate estimation of each point is performed according to the point matching results of different APs; judging whether the point matching results of different virtual APs have an intersection, and then using the WKNN algorithm to calculate the estimated coordinates of the point, including:

[0052] Based on the fingerprint point cluster matching results of the test point in different virtual APs, it is determined whether there are intersection fingerprint points in the matching results;

[0053] If there are intersection fingerprint points, the weighted k-nearest neighbor algorithm WKNN is further used to calculate the estimated coordinates of the test points based on these intersection fingerprint points. The calculation method is as follows:

[0054] First calculate the distance from the test point to each fingerprint point. The calculation formula is as follows:

[0055]

[0056] Where N represents the number of APs, RSS I n，i Indicates the RSS I value at the test point. n，j represents the RSS I value of the jth point in the fingerprint database, d ij Indicates the distance from the test point to each fingerprint point;

[0057] Select k fingerprint points with the smallest distance and add weights. The larger the distance, the smaller the weight. Calculate the estimated coordinates of the points to be determined. The calculation formula is as follows:

[0058]

[0059] Where (x k ,y k ) is the coordinate of the selected nearest neighbor fingerprint point, ω i，k is the weight.

[0060] If there is no intersection fingerprint point, the estimated coordinates of the test point are calculated using the WKNN algorithm based on the fingerprint point union of the point matching results of different virtual APs.

[0061] The beneficial effects of the present invention are:

[0062] The present invention provides a method for locating pedestrian trajectories in indoor path-free scenes, the method comprising: constructing a minimum matching group of pedestrian steps based on data collected by a mobile terminal; obtaining the trajectories of frequently-walked pedestrians from the collected data, and judging the credibility of pedestrian trajectories; performing scatter processing and clustering of pedestrian trajectories assisted by a virtual source when the trajectory matching result is unreliable; and estimating the coordinates of the undetermined points based on the clustering results of the trajectory fingerprint points after scattering. During use, the method adopts a method for locating pedestrian trajectories in indoor path-free scenes supported by a virtual source, and without increasing the deployment of hardware equipment, based on the data directly collected by the mobile terminal, a scatter positioning model of pedestrian trajectories in indoor path-free scenes is constructed with the assistance of a virtual source, and accurate location information is obtained in indoor path-free scenes, which greatly saves costs. Moreover, after adding the step of scatter positioning of pedestrian trajectories, the positioning accuracy of the indoor environment is significantly improved, and the positioning accuracy is further improved by cooperating with the scatter positioning algorithm of pedestrian trajectories in indoor path-free scenes supported by a virtual source. When the trajectory matching result is unreliable during positioning, this method uses virtual information source-assisted pedestrian trajectory scattering and clustering, which can effectively deal with the positioning problems caused by data loss, noise interference or environmental changes. Through scattering and clustering analysis, more useful location information can be extracted to improve the robustness of positioning. This helps to provide more accurate location information in complex indoor environments such as large shopping malls and exhibition halls. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a flow chart of a method for locating pedestrian trajectories in an indoor non-path-constrained scenario provided in an embodiment of the present invention;

[0064] Figure 2 is a flowchart of constructing a minimum matching group of pedestrian steps provided in an embodiment of the present invention;

[0065] Figure 3 is a flow chart of pedestrian trajectory credibility judgment provided in an embodiment of the present invention;

[0066] Figure 4 is a flow chart of performing scatter processing and clustering of pedestrian trajectories assisted by virtual information sources provided in an embodiment of the present invention;

[0067] Figure 5 is a flow chart of how to estimate the coordinates of a point to be determined provided in an embodiment of the present invention;

[0068] Figure 6 It is a schematic diagram of the distribution of experimental schemes for implementing the method for locating pedestrian trajectories in indoor scenarios without path constraints provided by an embodiment of the present invention;

[0069] Figure 7 It is a schematic diagram of the scattered pedestrian trajectories of the method for locating pedestrian trajectories in an indoor scene without path constraints provided by an embodiment of the present invention;

[0070] Figure 8 This is a mean-short clustering effect diagram of the indoor pedestrian trajectory positioning method without path constraints provided by an embodiment of the present invention;

[0071] Fig. 9 It is a schematic diagram of the position error cumulative distribution function (CDF) of the indoor pedestrian trajectory positioning method without path constraints provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0072] The preferred embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0073] In the prior art, indoor positioning technology is mainly divided into two categories: one is the positioning technology based on location fingerprint, and the other is the positioning technology based on the arrival time, arrival time difference and arrival angle of wireless signals. Location fingerprint positioning mainly performs clustering and segmentation based on the regional characteristic differences of indoor signal maps, including geometric location fingerprint clustering and signal distance fingerprint clustering. However, the fingerprint clustering algorithm based on geometric location analyzes the geometric proximity of fingerprint points in the offline stage. The number of clusters needs to be directly or indirectly predefined. Users need to carefully adjust the parameters to select the appropriate number of clusters, which is prone to the defect of inappropriate number of clusters leading to positioning deviation; the fingerprint clustering algorithm based on signal distance clusters according to the signal space distance of wireless signals, and there is a problem that the RSSI signal space distance is close but the actual geometric distance is far, which often leads to the wrong selection of some adjacent fingerprint points and ultimately affects the positioning accuracy. The positioning technology based on the arrival time, arrival time difference and arrival angle of wireless signals requires precise clock synchronization and multiple signal sources, which is more complicated to implement and will be affected by signal fluctuations, non-line-of-sight propagation and multipath propagation, resulting in low positioning accuracy.

[0074] Based on this, the present application provides a method for locating pedestrian trajectories in an indoor scenario without path constraints.

[0075] See also Figure 1-Figure 9 As shown, the present application provides a method for locating pedestrian trajectories in an indoor scene without path constraints, comprising:

[0076] S100. Construct a minimum matching group of pedestrian steps based on data collected by a mobile terminal. Specifically, the mobile terminal of the present application may be a mobile phone or a tablet computer.

[0077] Specifically, the step of constructing a minimum matching group of pedestrian steps based on the data collected by the mobile terminal includes:

[0078] S101. In an indoor environment, use a mobile terminal to collect AP data and sensor data of multiple pedestrian trajectories; specifically, use a smart phone to collect AP data and sensor data (such as accelerometer, gyroscope, etc.) of each minimum step matching group. The collected data can also be pre-processed, including denoising, filtering, etc., to improve the accuracy and reliability of the data. AP is the access point of the pedestrian trajectory.

[0079] S102, directly reading the number of pedestrian steps based on the step counting data of the mobile terminal, taking the current position of the pedestrian as the center, and constructing a minimum matching group of pedestrian steps as a group of 3 steps;

[0080] This step directly reads the pedestrian's step data and builds a matching group with the current position as the center, thus accurately dividing the pedestrian's walking trajectory. This division method helps to reduce the error in positioning calculation, because each matching group represents a relatively stable stage in the pedestrian's walking process. At the same time, building matching groups in groups of 3 steps can balance the relationship between computational complexity and positioning accuracy, making the positioning method both efficient and accurate.

[0081] S103. Calculate the average Wi-Fi signal strength (RSSI) of the minimum matching group of pedestrian steps by using the AP data.

[0082] By calculating the average Wi-Fi signal strength of each matching group, we can obtain the average signal strength information of the pedestrians in the group while they are walking. This information is very important for subsequent positioning calculations because it provides an indirect estimate of the distance between the pedestrian's position and the Wi-Fi access point. Using this estimate, we can further infer the specific location of the pedestrian, thereby achieving accurate indoor positioning.

[0083] S200, obtaining the common pedestrian trajectory from the collected data, and determining the credibility of the pedestrian trajectory;

[0084] Furthermore, obtaining the common pedestrian tracks from the collected data and judging the credibility of the pedestrian tracks includes:

[0085] S201, taking one of the collected multiple pedestrian trajectories as the current pedestrian trajectory, and the other pedestrian trajectories as the frequently walked pedestrian trajectories; this processing method helps to improve the accuracy and reliability of positioning, because the frequently walked pedestrian trajectory usually represents the habitual walking route of pedestrians in indoor environments, and has high stability and predictability. It also helps to make full use of the collected data resources and reduce the implementation cost of the method.

[0086] S202, matching the current pedestrian trajectory with all the frequently-walked pedestrian trajectories one by one; calculating the cosine similarity between the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectories, and determining whether the cosine similarity exceeds a set threshold value T;

[0087] S203, when the cosine similarity is greater than or equal to the set threshold T, it means that the matching result is credible and trajectory matching is used for positioning; when the cosine similarity is less than the set threshold T, it means that the matching result is unreliable.

[0088] More specifically, calculating the cosine similarity between the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectory, and determining whether the cosine similarity exceeds a set threshold T, includes:

[0089] For each frequently walked pedestrian trajectory, a sliding window of three steps is constructed with the minimum matching group of pedestrian steps as the unit; this method can pay attention to the detailed changes in the pedestrian's walking process during the indoor positioning process. At the same time, the use of sliding windows can ensure that each step matching group can have a certain overlap with the adjacent groups, thereby improving the continuity and accuracy of the analysis.

[0090] In each step count sliding window, the pedestrian trajectory step group sequence vector is calculated based on the accelerometer and gyroscope data collected by the mobile terminal sensor;

[0091] The cosine similarity is calculated one by one using the current pedestrian trajectory and the pedestrian trajectory step group sequence vectors of all common pedestrian trajectories. The calculation formula is as follows:

[0092]

[0093] Among them, a represents the step group sequence vector of the current pedestrian trajectory, and b represents the step group sequence vector of the usual pedestrian trajectory;

[0094] Cosine similarity can objectively evaluate the matching degree between trajectories and provide an accurate basis for subsequent credibility judgment.

[0095] According to the relationship between the cosine similarity of the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectory, it is judged whether the cosine similarity exceeds the set threshold T to determine whether the matching result is credible.

[0096] In a specific implementation, a threshold T=0.5 is set. When the similarity is greater than or equal to the threshold 0.5, it indicates that the matching result is credible and trajectory matching is used for positioning; when the similarity is less than the threshold 0.5, it indicates that the matching result is unreliable.

[0097] S300, when the trajectory matching result is unreliable, performing scatter processing and clustering of pedestrian trajectories assisted by virtual information sources;

[0098] Specifically, when the trajectory matching result is unreliable, performing scatter processing and clustering of pedestrian trajectories assisted by virtual information sources includes:

[0099] S301, taking the minimum matching group of pedestrian steps as a unit, combining three steps of pedestrian steps into one point, realizing the scattering of pedestrian trajectories, and calculating the distance from each point to the virtual AP according to the wireless signal attenuation model;

[0100] Furthermore, the minimum matching group of pedestrian steps is used as a unit, and three steps of pedestrian steps are combined into one point to scatter the pedestrian trajectory. According to the wireless signal attenuation model, the distance from each point to the virtual AP is calculated, including:

[0101] Taking the minimum matching group of pedestrian steps as the unit, three steps are combined into one point, and the average RSS I of the three steps is calculated as the RSS I value of the point;

[0102] According to the wireless signal attenuation model, the Wi-Fi RSSI attenuation formula is used to calculate the distance from each point to the virtual AP. The calculation formula is as follows:

[0103]

[0104] Among them, RSS I(d0) and RSS I(d i ) are the distances d0 and d from the AP signal source respectively. i The received signal strength at the location, η is the path attenuation factor, d0 and RSS I(d0) are pre-set through empirical modeling, d i Indicates the distance from the i-th point to the virtual AP.

[0105] Pedestrian trajectory scattering is achieved by combining three steps of pedestrians into one point. This representation method helps to simplify subsequent calculations and processing while retaining key information about the pedestrian's walking process.

[0106] S302, using the KNN (K nearest neighbor) algorithm to find K neighboring fingerprint points corresponding to the test point; combining the linear least squares adjustment method, using these K neighboring fingerprint points to calculate the virtual coordinates of the AP; this calculation method is based on the Wi-Fi RSSI attenuation formula, taking into account the relationship between signal strength and distance, and therefore has higher accuracy.

[0107] Furthermore, the method of using the KNN (K nearest neighbor) algorithm to find K neighboring fingerprint points corresponding to the test point; combining the linear least squares adjustment method, and using the K neighboring fingerprint points to calculate the virtual coordinates of the AP includes:

[0108] Use the KNN algorithm to obtain the K neighboring fingerprint points corresponding to the test point;

[0109] Using these K neighboring fingerprint points, combined with linear least squares adjustment, the virtual coordinates of AP are calculated using the following formula:

[0110]

[0111] Among them, (x j ,y j ) are unknown coordinate parameters, B is the coefficient matrix, are the virtual AP coordinates after adjustment.

[0112] S303: For each AP, calculate the distance between its virtual coordinates; cluster the virtual coordinates using the Mean-Shift clustering algorithm to form a fingerprint point clustering result.

[0113] S400: based on the clustering result of the scattered trajectory fingerprint points, estimate the coordinates of the undetermined points.

[0114] Specifically, the coordinate estimation of the undetermined point based on the clustering result of the trajectory fingerprint points after scattering includes:

[0115] Specifically, high-precision indoor positioning can be achieved by scattering pedestrian trajectories, applying wireless signal attenuation models, using the KNN algorithm and the linear least squares adjustment method. This method not only improves the accuracy and reliability of positioning, but also enhances the practicality and scalability of positioning, and can be applied to pedestrian positioning needs in various indoor scenarios. At the same time, this method can also provide valuable information for applications such as indoor navigation and intelligent monitoring.

[0116] S401, according to the fingerprint point clustering result, coordinate matching of the points is performed; one test point is selected, the RSSD difference average value between the test point and the fingerprint points in each group of clusters in the virtual AP is calculated, the test point is matched to the cluster with the smallest average distance value, and the test point is matched with different virtual APs one by one;

[0117] Furthermore, selecting a test point, calculating the average RSSD difference between the test point and the fingerprint points in each group of clusters in the virtual AP, matching the test point to the cluster with the smallest average distance value, and matching with different virtual APs one by one, includes:

[0118] Based on the AP data of the test point, select the top 6 virtual APs with the highest signal strength.

[0119] For each virtual AP, calculate the average RSSD (received signal strength difference) difference between the test point and each fingerprint point cluster, and match the test point to a fingerprint point cluster with the smallest RSSD difference average value;

[0120] Repeat the above process for all selected virtual APs to complete all matching.

[0121] In this step, the average RSSD difference between the test point and the fingerprint points in each cluster of the virtual AP is calculated, and the test point is matched to the cluster with the smallest average distance value. This method can more accurately find fingerprint points with similar signal characteristics to the test point, thereby improving the accuracy of matching. By selecting the top 6 virtual APs with the highest signal strength for matching, the calculation complexity can be effectively reduced while ensuring the accuracy of matching.

[0122] S402, based on the point matching results of different APs, the coordinates of each point are estimated; it is determined whether the point matching results of different virtual APs have any intersection, and then the estimated coordinates of the point are calculated using the WKNN algorithm;

[0123] Furthermore, the coordinate estimation of each point is performed according to the point matching results of different APs; judging whether the point matching results of different virtual APs have an intersection, and then using the WKNN algorithm to calculate the estimated coordinates of the point, including:

[0124] Based on the fingerprint point cluster matching results of the test point in different virtual APs, it is determined whether there are intersection fingerprint points in the matching results;

[0125] If there are intersection fingerprint points, the weighted k-nearest neighbor algorithm (WKNN) is further used to calculate the estimated coordinates of the test point based on these intersection fingerprint points. The calculation method is as follows:

[0126] First calculate the distance from the test point to each fingerprint point. The calculation formula is as follows:

[0127]

[0128] Where N represents the number of APs, RSS I n，i Indicates the RSS I value at the test point. n，j represents the RSS I value of the jth point in the fingerprint database, d ij Indicates the distance from the test point to each fingerprint point;

[0129] Select k fingerprint points with the smallest distance and add weights. The larger the distance, the smaller the weight. Calculate the estimated coordinates of the points to be determined. The calculation formula is as follows:

[0130]

[0131] Where (x k ,y k ) is the coordinate of the selected nearest neighbor fingerprint point, ω i，k is the weight; ω i，k Usually it is the reciprocal of the distance from the test point to the fingerprint point.

[0132] If there is no intersection fingerprint point, the estimated coordinates of the test point are calculated using the WKNN algorithm based on the fingerprint point union of the point matching results of different virtual APs.

[0133] By determining whether the point matching results of different virtual APs have intersections and using the WKNN algorithm to calculate the estimated coordinates of the points, this method can comprehensively consider the positioning results of multiple APs, thereby improving the accuracy of coordinate estimation. The WKNN algorithm is a weighted K nearest neighbor algorithm that considers the impact of the distance of neighboring points on the estimation results, so it can enhance the stability of positioning and reduce positioning deviations caused by individual point matching errors.

[0134] S403, connecting the estimated point coordinates into lines in sequence to form a new pedestrian walking trajectory;

[0135] The process of constructing a new pedestrian walking trajectory involves the process of trajectory reconstruction, which is as follows: all points in the trajectory are used as test points one by one, and the above coordinate estimation process is repeated. The estimated point coordinates are connected into lines in sequence to form the pedestrian's current walking trajectory. Since the coordinates of each point are estimated based on the information of its neighboring points, this method can ensure the continuity of the trajectory. At the same time, by connecting the estimated point coordinates in sequence, a smooth walking trajectory can be formed, which is more in line with the actual walking situation.

[0136] The above-mentioned method for locating pedestrian trajectories in indoor path-free scenes includes: constructing a minimum matching group of pedestrian steps based on data collected by a mobile terminal; obtaining the trajectories of frequently-walked pedestrians from the collected data, and judging the credibility of pedestrian trajectories; performing scatter processing and clustering of pedestrian trajectories assisted by a virtual source when the trajectory matching result is unreliable; and estimating the coordinates of the undetermined points based on the clustering results of the trajectory fingerprint points after scattering. During use, the method adopts a scatter positioning method for indoor path-free scenes supported by a virtual source. Without increasing the deployment of hardware equipment, the method directly constructs a scatter positioning model for pedestrian trajectories in indoor path-free scenes with the assistance of a virtual source based on data collected directly by a mobile terminal, and obtains accurate location information in indoor path-free scenes, which greatly saves costs. Moreover, after adding the scatter positioning step of pedestrian trajectories, the positioning accuracy of the indoor environment is significantly improved, and the positioning accuracy is further improved with the scatter positioning algorithm for indoor path-free scenes supported by a virtual source. When the trajectory matching result is unreliable during positioning, this method uses virtual information source-assisted pedestrian trajectory scattering and clustering, which can effectively deal with the positioning problems caused by data loss, noise interference or environmental changes. Through scattering and clustering analysis, more useful location information can be extracted to improve the robustness of positioning. This helps to provide more accurate location information in complex indoor environments such as large shopping malls and exhibition halls.

[0137] Example 2

[0138] See also Figure 1-Figure 9As shown, this embodiment provides an implementation process of a pedestrian trajectory positioning method for an indoor scene without path constraints of the present application, which is specifically as follows:

[0139] This embodiment aims to verify a pedestrian trajectory positioning method in an indoor scenario without path constraints. The experiment was conducted in an open hall of a large office building with an area of ​​about 1,000 square meters. The hall has a complex internal layout and contains a variety of obstacles and furniture to fully simulate the real indoor environment. During the experiment, the Wi-Fi access points (APs) in the hall are evenly distributed to ensure comprehensive signal coverage.

[0140] (1) Data collection and preprocessing

[0141] Device selection: Use mainstream smartphones on the market as mobile terminals. The phone has built-in high-precision accelerometers, gyroscopes and other sensors, and supports the collection of Wi-Fi signal strength (RSS I).

[0142] Data collection:

[0143] Multiple walking tracks are preset in the hall, including straight lines, curves, broken lines and other shapes to ensure data diversity.

[0144] Invite multiple testers to walk along the preset trajectory and use their mobile phones to collect AP data and sensor data for each trajectory.

[0145] During the data collection process, ensure that the mobile phone can stably receive the Wi-Fi signal and accurately record the number of steps taken.

[0146] Data preprocessing:

[0147] The collected data is denoised and filtered to improve the accuracy and reliability of the data.

[0148] According to the step counting data of the mobile phone, each trajectory is divided into multiple pedestrian step minimum matching groups, each group contains 3 steps.

[0149] (2) Constructing the minimum matching group of pedestrian steps

[0150] AP data collection: Use the mobile phone to collect AP data of each minimum step matching group, including RSSI value and data.

[0151] Sensor data collection: Data from sensors such as accelerometers and gyroscopes are collected simultaneously for subsequent calculation of pedestrian trajectory step group sequence vectors.

[0152] Calculate the RSS I mean: For each minimum step matching group, calculate the mean of its Wi-Fi signal strength (RSS I) as the feature value of the group.

[0153] (3) Pedestrian trajectory credibility judgment

[0154] Select the current pedestrian trajectory: randomly select one of the multiple collected trajectories as the current pedestrian trajectory.

[0155] Trajectory matching: Match the current pedestrian trajectory with the remaining frequently-walked pedestrian trajectories one by one and calculate the cosine similarity.

[0156] Credibility judgment: Set the threshold T = 0.5. When the cosine similarity is greater than or equal to 0.5, the matching result is considered credible and trajectory matching is used for positioning; otherwise, the matching result is considered unreliable and the next step of scatter processing and clustering is entered.

[0157] (4) Virtual source-assisted pedestrian trajectory scattering and clustering

[0158] Trajectory scattering:

[0159] Taking the minimum matching group of pedestrian steps as the unit, three steps are combined into one point to realize the scattering of pedestrian trajectories.

[0160] Calculate the RSSI average value of each synthetic point as the signal strength feature of the point.

[0161] According to the wireless signal attenuation model, the distance from each point to the virtual AP is calculated.

[0162] KNN algorithm finds neighboring fingerprint points:

[0163] Use the KNN algorithm (K value is set to 5) to find the 5 neighboring fingerprint points corresponding to each test point.

[0164] Combined with the linear least squares adjustment method, the coordinates of the virtual AP are calculated.

[0165] Clustering Processing:

[0166] The coordinates of each virtual AP are processed using the Mean-Shift clustering algorithm to form a fingerprint point clustering result.

[0167] (5) Coordinate estimation of the points to be determined

[0168] The coordinates of the points match:

[0169] Select a test point and calculate the average RSSD difference between the test point and the fingerprint points in each cluster in the virtual AP.

[0170] Match the test point to the cluster with the smallest average distance value and match it with different virtual APs one by one.

[0171] Coordinate estimation:

[0172] Determine whether the point matching results of different virtual APs have any intersection.

[0173] If there are intersection fingerprint points, the estimated coordinates of the test points are calculated using the WKNN algorithm based on the intersection fingerprint points.

[0174] If there is no intersection fingerprint point, the estimated coordinates of the test point are calculated using the WKNN algorithm based on the union fingerprint point.

[0175] Trajectory reconstruction:

[0176] All points in the trajectory are used as test points one by one, and the above coordinate estimation process is repeated. The estimated point coordinates are connected into lines in sequence to form the pedestrian's current walking trajectory.

[0177] (6) Experimental results and analysis

[0178] After experimental verification, the pedestrian trajectory positioning method for indoor path-free scene proposed in this embodiment can effectively improve the positioning accuracy. Compared with the traditional method, the positioning error of this method in complex indoor environments is reduced by about 30%, while maintaining high real-time performance and stability. In addition, through virtual source assistance and trajectory scatter processing, this method successfully solves the problem of traditional methods' difficulty in positioning in path-free scenes, and has broad application prospects.

[0179] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used here to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0180] In addition, it should be noted that the use of terms such as "first" and "second" to limit components is only for the convenience of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore cannot be understood as limiting the scope of protection of this application.

[0181] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for locating pedestrian trajectories in indoor scenes without path constraints, characterized in that: include: Based on the data collected by the mobile terminal, a minimum matching group of pedestrian steps is constructed; From the collected data, the common pedestrian trajectories are obtained to determine the credibility of the pedestrian trajectories; When the trajectory matching result is unreliable, the pedestrian trajectory is scattered and clustered with the assistance of virtual information sources; Based on the clustering results of the scattered trajectory fingerprint points, the coordinates of the unknown points are estimated.

2. The method for locating pedestrian trajectories in indoor non-path-constrained scenes according to claim 1 is characterized in that: The step of constructing a minimum matching group of pedestrian steps based on data collected by the mobile terminal includes: In indoor environments, mobile terminals are used to collect AP data and sensor data of multiple pedestrian trajectories; Directly read the number of pedestrian steps based on the step counting data of the mobile terminal, and build the minimum matching group of pedestrian steps with 3 steps as a group, taking the pedestrian's current position as the center; Using AP data, calculate the average Wi-Fi signal strength of the minimum matching group of pedestrian steps.

3. The method for locating pedestrian trajectories in indoor non-path-constrained scenes according to claim 2 is characterized in that: The method of obtaining the frequently-walked pedestrian tracks from the collected data and judging the credibility of the pedestrian tracks includes: One of the collected multiple pedestrian trajectories is used as the current pedestrian trajectory, and the remaining pedestrian trajectories are used as the common pedestrian trajectories; Match the current pedestrian trajectory with all the frequently-walked pedestrian trajectories one by one; calculate the cosine similarity between the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectories, and determine whether the cosine similarity exceeds the set threshold T; When the cosine similarity is greater than or equal to the set threshold T, it means that the matching result is credible and trajectory matching is used for positioning; when the cosine similarity is less than the set threshold T, it means that the matching result is unreliable.

4. The method for locating pedestrian trajectories in indoor non-path-constrained scenes according to claim 3 is characterized in that: The calculating the cosine similarity between the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectory and judging whether the cosine similarity exceeds a set threshold value T includes: For each frequently walked pedestrian trajectory, a sliding window of 3 steps is constructed with the minimum matching group of pedestrian steps as the unit; In each step count sliding window, the pedestrian trajectory step group sequence vector is calculated based on the accelerometer and gyroscope data collected by the mobile terminal sensor; The cosine similarity is calculated one by one using the current pedestrian trajectory and the pedestrian trajectory step group sequence vectors of all common pedestrian trajectories. The calculation formula is as follows: Among them, a represents the step group sequence vector of the current pedestrian trajectory, and b represents the step group sequence vector of the usual pedestrian trajectory; According to the relationship between the cosine similarity of the current pedestrian trajectory and the candidate frequently-walked pedestrian trajectory, it is judged whether the cosine similarity exceeds the set threshold T to determine whether the matching result is credible.

5. The method for locating pedestrian trajectories in indoor non-path-constrained scenes according to claim 1 is characterized in that: When the trajectory matching result is unreliable, performing scatter processing and clustering of pedestrian trajectories assisted by virtual information sources includes: Taking the minimum matching group of pedestrian steps as the unit, the three steps of pedestrians are combined into one point to scatter the pedestrian trajectory. According to the wireless signal attenuation model, the distance from each point to the virtual AP is calculated; Use the KNN algorithm to find the K neighboring fingerprint points corresponding to the test point; combine the linear least squares adjustment method to use these K neighboring fingerprint points to calculate the virtual coordinates of the AP; For each AP, the distance between its virtual coordinates is calculated; the virtual coordinates are clustered using the Mean-Shift clustering algorithm to form the fingerprint point clustering result.

6. The method for locating pedestrian trajectories in indoor non-path-constrained scenes according to claim 5 is characterized by: The method uses the minimum matching group of pedestrian steps as a unit, combines three steps of pedestrian steps into one point, realizes the scattering of pedestrian trajectories, and calculates the distance from each point to the virtual AP according to the wireless signal attenuation model, including: Taking the minimum matching group of pedestrian steps as the unit, three steps are combined into one point, and the average RSS I of the three steps is calculated as the RSS I value of the point; According to the wireless signal attenuation model, the WiFi RSSI attenuation formula is used to calculate the distance from each point to the virtual AP. The calculation formula is as follows: Among them, RSS I(d0) and RSS I(d i ) are the distances d0 and d from the AP signal source respectively. i The received signal strength at the location, η is the path attenuation factor, d0 and RSS I(d0) are pre-set through empirical modeling, d i Indicates the distance from the i-th point to the virtual AP.

7. The method for locating pedestrian trajectories in indoor non-path-constrained scenes according to claim 5 is characterized by: The method of using the KNN algorithm to find K neighboring fingerprint points corresponding to the test point and combining the linear least squares adjustment method to use the K neighboring fingerprint points to calculate the virtual coordinates of the AP includes: Use the KNN algorithm to obtain the K neighboring fingerprint points corresponding to the test point; Using these K neighboring fingerprint points, combined with linear least squares adjustment, the virtual coordinates of AP are calculated using the following formula: Among them, (x j ,y j ) are unknown coordinate parameters, B is the coefficient matrix, are the virtual AP coordinates after adjustment.

8. The method for locating pedestrian trajectories in an indoor scene without path constraints according to any one of claims 1 to 7, characterized in that: The coordinate estimation of the undetermined point based on the clustering result of the trajectory fingerprint points after scattering includes: According to the fingerprint point clustering results, coordinate matching of points is performed; Select a test point, calculate the average RSSD difference between the test point and the fingerprint points in each cluster of the virtual AP, match the test point to the cluster with the smallest average distance value, and match it with different virtual APs one by one; According to the point matching results of different APs, the coordinates of each point are estimated; Determine whether the point matching results of different virtual APs have any intersection, and then use the WKNN algorithm to calculate the estimated coordinates of the points; The estimated point coordinates are connected into lines in sequence to form a new pedestrian walking trajectory.

9. The method for locating pedestrian trajectories in an indoor scene without path constraints according to any one of claims 1 to 7, characterized in that: Select a test point, calculate the average RSSD difference between the test point and the fingerprint points in each cluster of the virtual AP, match the test point to the cluster with the smallest average distance value, and match it with different virtual APs one by one, including: Based on the AP data of the test point, select the top 6 virtual APs with the highest signal strength. For each virtual AP, calculate the average RSSD (received signal strength difference) difference between the test point and each fingerprint point cluster, and match the test point to a fingerprint point cluster with the smallest RSSD difference average value; Repeat the above process for all selected virtual APs to complete all matching.

10. The method for locating pedestrian trajectories in indoor non-path-constrained scenes according to claim 8, characterized in that: According to the point matching results of different APs, the coordinates of each point are estimated; Determine whether the point matching results of different virtual APs have intersections, and then use the WKNN algorithm to calculate the estimated coordinates of the points, including: Based on the fingerprint point cluster matching results of the test point in different virtual APs, it is determined whether there are intersection fingerprint points in the matching results; If there are intersection fingerprint points, the weighted k-nearest neighbor algorithm WKNN is further used to calculate the estimated coordinates of the test points based on these intersection fingerprint points. The calculation method is as follows: First calculate the distance from the test point to each fingerprint point. The calculation formula is as follows: Where N represents the number of APs, RSSI n，i Indicates the RSSI value at the test point, RSS I n，j represents the RSSI value of the jth point in the fingerprint library, d ij Indicates the distance from the test point to each fingerprint point; Select k fingerprint points with the smallest distance and add weights. The larger the distance, the smaller the weight. Calculate the estimated coordinates of the points to be determined. The calculation formula is as follows: Where (x k ,y k ) is the coordinate of the selected nearest neighbor fingerprint point; ω i，k is the weight; If there is no intersection fingerprint point, the estimated coordinates of the test point are calculated using the WKNN algorithm based on the fingerprint point union of the point matching results of different virtual APs.

Images