Method for obtaining uncertainty region of pedestrian continuous trajectory and related device

Abstract

The present invention discloses a method for obtaining an uncertainty region for a continuous trajectory of a pedestrian and related devices. The method includes: collecting and obtaining motion information corresponding to the continuous trajectory of a target pedestrian and position information of a reference point therein; according to the motion information and the position information, obtaining a corresponding reconstructed trajectory through a position recursion model, and obtaining a reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian after optimizing the reconstructed trajectory according to the position information of the reference point; obtaining target features corresponding to each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory; according to the target features, obtaining uncertainty error values corresponding to each reconstructed and optimized coordinate point through a trained uncertainty error prediction model; and obtaining an uncertainty region corresponding to the reconstructed and optimized trajectory according to the reconstructed and optimized coordinate points and the uncertainty error values and using it as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian. The present invention is beneficial to improving the accuracy of obtaining the region where the pedestrian may be located.

Description

Technical Field

[0001] The present invention relates to the technical field of positioning, and particularly relates to a method for obtaining an uncertainty area of a pedestrian's continuous trajectory and related devices. Background Art

[0002] With the development of science and technology, various intelligent services provided by intelligent devices are more and more widely applied in users' work and life. Among them, location-based intelligent services have also been widely used. When using location-based intelligent services, it is necessary to obtain the positions that a user may pass through during continuous movement.

[0003] In the prior art, the movement trajectory of a pedestrian is usually directly collected (for example, collected through positioning technologies such as GPS), and each point on the movement trajectory is regarded as the position passed by the pedestrian during continuous movement. The problem of the prior art is that the above-mentioned movement trajectory directly collected contains uncertainty errors and does not represent the true trajectory of the pedestrian. Therefore, directly regarding each point on the above-mentioned movement trajectory as the position passed by the pedestrian during continuous movement is not conducive to improving the accuracy of obtaining the area where the pedestrian may be located.

[0004] Therefore, the prior art still needs to be improved and developed. Summary of the Invention

[0005] The main purpose of the present invention is to provide a method for obtaining an uncertainty area of a pedestrian's continuous trajectory and related devices, aiming to solve the problem in the prior art that directly regarding each point on the movement trajectory directly collected as the position passed by the pedestrian during continuous movement is not conducive to improving the accuracy of obtaining the area where the pedestrian may be located.

[0006] To achieve the above purpose, in the first aspect of the present invention, a method for obtaining an uncertainty area of a pedestrian's continuous trajectory is provided. The method for obtaining an uncertainty area of a pedestrian's continuous trajectory includes:

[0007] Collect the motion information corresponding to the continuous trajectory of a target pedestrian and the position information of reference points in the continuous trajectory of the target pedestrian. The motion information includes a plurality of pedestrian step length values and a plurality of heading values, and one pedestrian step length value corresponds to one heading value. The reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the reference points includes the coordinate values of the reference points;

[0008] According to the above motion information and the position information of the above reference point, obtain the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model, and optimize the reconstructed trajectory according to the position information of the above reference point to obtain the reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian;

[0009] Obtain the target features corresponding to each reconstructed and optimized coordinate point in the above reconstructed and optimized trajectory. Among them, the target features corresponding to a reconstructed and optimized coordinate point include the reconstructed and optimized step value, reconstructed and optimized heading value, position update time difference, current cumulative step value, and speed value corresponding to this reconstructed and optimized coordinate point;

[0010] According to the above target features, obtain the uncertainty error values corresponding to each of the above reconstructed and optimized coordinate points through a trained uncertainty error prediction model;

[0011] Obtain the uncertainty region corresponding to the above reconstructed and optimized trajectory based on the above reconstructed and optimized coordinate points and the above uncertainty error values, and use it as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian. Among them, the above target uncertainty region is the region where the real trajectory corresponding to the continuous trajectory of the target pedestrian is located.

[0012] Optionally, the above acquisition of the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference point in the continuous trajectory of the target pedestrian includes:

[0013] Real-time acquisition of the above heading value through the magnetometer, accelerometer, and gyroscope in the movable intelligent device corresponding to the target pedestrian, and acquisition of the above pedestrian step value through the above accelerometer;

[0014] Through the above movable intelligent device, obtain the coordinate value of the above reference point according to at least one of a variety of preset positioning methods. Among them, the above variety of preset positioning methods include positioning methods based on wireless communication networks and positioning methods based on code scanning.

[0015] Optionally, the above obtaining the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model according to the above motion information and the position information of the above reference point includes:

[0016] Input the above motion information and the position information of the above reference point into the above preset position recursion model, and obtain the above reconstructed trajectory according to the preset position recursion formula in the above preset position recursion model. Among them, the above reconstructed trajectory includes a plurality of successively connected reconstructed coordinate points, and one of the above reconstructed coordinate points is connected to the next reconstructed coordinate point according to the reconstructed step value and reconstructed heading value corresponding to this reconstructed coordinate point.

[0017] Optionally, after optimizing the above-mentioned reconstructed trajectory according to the position information of the above-mentioned reference points, the reconstructed optimized trajectory corresponding to the continuous trajectory of the target pedestrian is obtained, including:

[0018] Obtain each reconstructed coordinate point corresponding to each of the above-mentioned reference points and use it as a target point;

[0019] According to the position information of the above-mentioned reference points, the position information of the above-mentioned target points, and a preset loss function, with the goal of minimizing the position information loss value between the above-mentioned reference points and their corresponding target points, optimize the reconstruction step value and reconstruction heading value of each of the above-mentioned target points to obtain the corresponding reconstructed optimized step value and reconstructed optimized heading value. Use the optimized above-mentioned target points as reconstructed optimized coordinate points, and obtain the above-mentioned reconstructed optimized trajectory. Among them, the above-mentioned reconstructed optimized trajectory includes multiple successively connected reconstructed optimized coordinate points. One above-mentioned reconstructed optimized coordinate point is connected to the next reconstructed optimized coordinate point according to the reconstructed optimized step value and reconstructed optimized heading value corresponding to this reconstructed optimized coordinate point.

[0020] Optionally, the target features corresponding to one reconstructed optimized coordinate point further include the current cumulative walking distance, current distance ratio, current duration ratio, current step ratio, current heading change amount, and current cumulative heading change amount corresponding to this reconstructed optimized coordinate point;

[0021] Among them, the above-mentioned current cumulative walking distance is the sum of the reconstructed optimized step lengths of all current passing points. The above-mentioned current passing points include all reconstructed optimized coordinate points from the starting point to this reconstructed optimized coordinate point of the above-mentioned reconstructed optimized trajectory;

[0022] The above-mentioned current distance ratio is the ratio of the above-mentioned current cumulative walking distance to the total distance of the trajectory. The total distance of the trajectory is the sum of the reconstructed optimized step values of all trajectory passing points. The trajectory passing points include all reconstructed optimized coordinate points from the starting point to the ending point of the above-mentioned reconstructed optimized trajectory;

[0023] The above-mentioned current duration ratio is the ratio of the current duration to the total duration of the trajectory. The current duration is the cumulative duration corresponding to all the above-mentioned current passing points. The total duration of the trajectory is the cumulative duration from the starting point to the ending point of the above-mentioned reconstructed optimized trajectory;

[0024] The above-mentioned current step ratio is the ratio of the above-mentioned current cumulative step value to the total number of steps of the trajectory. The total number of steps of the trajectory is the cumulative number of steps from the starting point to the ending point of the above-mentioned reconstructed optimized trajectory;

[0025] The above-mentioned current heading change amount is the difference between the reconstructed optimized heading value corresponding to this reconstructed optimized coordinate point and the reconstructed optimized heading value corresponding to the previous reconstructed optimized coordinate point;

[0026] The above-mentioned current cumulative heading change amount is the sum of the heading start change amounts corresponding to all the above-mentioned current passing points, and the heading start change amount is the difference between the reconstructed and optimized heading value between the above-mentioned current passing point and the start point of the above-mentioned reconstructed and optimized trajectory.

[0027] Optionally, for any one reconstructed and optimized coordinate point, the uncertainty error value corresponding to the reconstructed and optimized coordinate point is calculated and obtained according to the following steps:

[0028] The target features corresponding to the above-mentioned reconstructed and optimized coordinate point are spliced to obtain a multi-dimensional feature vector;

[0029] The above-mentioned feature vector is input into the above-mentioned trained uncertainty error prediction model, and the uncertainty error value corresponding to the above-mentioned reconstructed and optimized coordinate point output by the above-mentioned uncertainty error prediction model is obtained;

[0030] Wherein, the above-mentioned uncertainty error prediction model is a hybrid deep learning model, and the above-mentioned uncertainty error prediction model includes a one-dimensional convolutional neural network layer, a long short-term memory neural network layer and a fully connected layer.

[0031] Optionally, the above-mentioned obtaining the uncertainty region corresponding to the above-mentioned reconstructed and optimized trajectory according to the above-mentioned reconstructed and optimized coordinate point and the above-mentioned uncertainty error value and using it as the target uncertainty region corresponding to the continuous trajectory of the above-mentioned target pedestrian includes:

[0032] The error circles corresponding to each of the above-mentioned reconstructed and optimized coordinate points are obtained respectively, wherein the center of an error circle corresponding to one of the above-mentioned reconstructed and optimized coordinate points is the reconstructed and optimized coordinate point, and the radius is the uncertainty error value corresponding to the reconstructed and optimized coordinate point;

[0033] The union regions of the error circles corresponding to each of the above-mentioned reconstructed and optimized coordinate points are obtained in sequence, wherein the union region of the error circles corresponding to one of the above-mentioned reconstructed and optimized coordinate points is the union region corresponding to the error circles of the reconstructed and optimized coordinate point and the next adjacent reconstructed and optimized coordinate point;

[0034] The union of the union regions of the error circles corresponding to all the above-mentioned reconstructed and optimized coordinate points is used as the uncertainty region corresponding to the above-mentioned reconstructed and optimized trajectory, and the uncertainty region corresponding to the above-mentioned reconstructed and optimized trajectory is used as the target uncertainty region corresponding to the continuous trajectory of the above-mentioned target pedestrian.

[0035] The second aspect of the present invention provides a system for obtaining the uncertainty region of a continuous pedestrian trajectory, wherein the system for obtaining the uncertainty region of a continuous pedestrian trajectory includes:

[0036] A data acquisition module, configured to collect and obtain the motion information corresponding to the continuous trajectory of a target pedestrian and the position information of reference points in the continuous trajectory of the target pedestrian. The motion information includes a plurality of pedestrian step values and a plurality of heading values, and one of the pedestrian step values corresponds to one of the heading values. The reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the reference points includes the coordinate values of the reference points;

[0037] A trajectory reconstruction module, configured to obtain a reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model according to the motion information and the position information of the reference points, and obtain a reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian after optimizing the reconstructed trajectory according to the position information of the reference points;

[0038] A target feature extraction module, configured to obtain the target features corresponding to each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory. The target features corresponding to one reconstructed and optimized coordinate point include the reconstructed and optimized step value, the reconstructed and optimized heading value, the position update time difference, the current cumulative step value, and the speed value corresponding to the reconstructed and optimized coordinate point;

[0039] An error value calculation module, configured to obtain the uncertainty error value corresponding to each reconstructed and optimized coordinate point through a trained uncertainty error prediction model according to the target features;

[0040] An uncertainty region generation module, configured to obtain an uncertainty region corresponding to the reconstructed and optimized trajectory based on the reconstructed and optimized coordinate points and the uncertainty error values, and use it as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian. The target uncertainty region is the region where the true trajectory corresponding to the continuous trajectory of the target pedestrian is located.

[0041] A third aspect of the present invention provides an intelligent terminal. The intelligent terminal includes a memory, a processor, and an uncertainty region acquisition program for pedestrian continuous trajectories stored on the memory and executable on the processor. When the uncertainty region acquisition program for pedestrian continuous trajectories is executed by the processor, it implements the steps of any one of the above-mentioned uncertainty region acquisition methods for pedestrian continuous trajectories.

[0042] A fourth aspect of the present invention provides a computer-readable storage medium. An uncertainty region acquisition program for pedestrian continuous trajectories is stored on the computer-readable storage medium. When the uncertainty region acquisition program for pedestrian continuous trajectories is executed by a processor, it implements the steps of any one of the above-mentioned uncertainty region acquisition methods for pedestrian continuous trajectories.

[0043] As can be seen from the above, in the solution of the present invention, motion information corresponding to the continuous trajectory of the target pedestrian and position information of reference points in the continuous trajectory of the target pedestrian are collected and obtained. Among them, the motion information includes a plurality of pedestrian step values and a plurality of heading values, and one of the pedestrian step values corresponds to one of the heading values. The reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the reference points includes the coordinate values of the reference points. According to the motion information and the position information of the reference points, a reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian is obtained through a preset position recursion model, and a reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian is obtained after optimizing the reconstructed trajectory according to the position information of the reference points. Target features corresponding to each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory are obtained. Among them, the target features corresponding to one reconstructed and optimized coordinate point include the reconstructed and optimized step value, the reconstructed and optimized heading value, the position update time difference, the current cumulative step number, and the speed value corresponding to the reconstructed and optimized coordinate point. According to the target features, an uncertainty error value corresponding to each reconstructed and optimized coordinate point is obtained through a trained uncertainty error prediction model. An uncertainty region corresponding to the reconstructed and optimized trajectory is obtained according to the reconstructed and optimized coordinate points and the uncertainty error values and used as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian, where the target uncertainty region is the region where the true trajectory corresponding to the continuous trajectory of the target pedestrian is located.

[0044] Compared with the prior art, in the solution of the present invention, each point on the continuous trajectory of the target pedestrian that can be directly collected and obtained is not regarded as the position passed by the pedestrian during continuous movement. Instead, according to the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference point therein, the reconstructed trajectory corresponding to the continuous trajectory of the above target pedestrian is obtained through a preset position recursion model, and further optimized to obtain a reconstructed and optimized trajectory. By reconstructing and optimizing the trajectory, the uncertainty included in the trajectory is eliminated (or minimized as much as possible), which is beneficial to improving the accuracy of obtaining the position of the user (i.e., the pedestrian). Then, the target features of each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory are extracted, and the uncertainty error value of each reconstructed and optimized coordinate point is calculated through a trained uncertainty error prediction model. Finally, the corresponding target uncertainty region is generated according to each reconstructed and optimized coordinate point and its corresponding uncertainty error value, which is beneficial to obtaining more accurately the area that the user may pass through, that is, improving the accuracy of obtaining the user's position. The target uncertainty region represents the region where the true trajectory corresponding to the continuous trajectory of the above target pedestrian is located, that is, the region where the points that the pedestrian may pass through during continuous movement are determined after eliminating the uncertainty error of the trajectory. In this way, the solution of the present invention can reduce the influence of uncertainty in the continuous trajectory of the pedestrian and reduce the influence of errors in the process of collecting the original continuous trajectory of the target pedestrian, and generate a target uncertainty region corresponding to the true trajectory of the pedestrian. The above target uncertainty region can cover the true trajectory of the pedestrian, reflect the possible position of the pedestrian, and is smaller than the area range limited by the trajectory obtained through direct original collection, and can define a more accurate range, which is beneficial to improving the accuracy of obtaining the area where the pedestrian may be located, that is, the range where the true trajectory is located can be determined more accurately and the above true trajectory can be better covered. Furthermore, it is beneficial to improve the accuracy and effect of using location-based intelligent services. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0046] Figure 1 It is a schematic flowchart of a method for obtaining an uncertainty region for a continuous trajectory of a pedestrian provided by an embodiment of the present invention;

[0047] Figure 2 It is a schematic diagram of a reconstructed trajectory obtained by reconstruction provided by an embodiment of the present invention;

[0048] Figure 3It is a schematic diagram of the acquisition route of the continuous trajectory of the target pedestrian provided by the embodiment of the present invention;

[0049] Figure 4 It is a schematic comparison diagram of the reconstructed and optimized trajectory and the real trajectory provided by the embodiment of the present invention;

[0050] Figure 5 It is a schematic structural diagram of the uncertainty error prediction model provided by the embodiment of the present invention;

[0051] Figure 6 It is a schematic diagram of the obtained target uncertainty region provided by the embodiment of the present invention;

[0052] Figure 7 It is another schematic diagram of the obtained target uncertainty region provided by the embodiment of the present invention;

[0053] Figure 8 It is a schematic diagram of the composition module of the uncertainty region acquisition system for the continuous trajectory of pedestrians provided by the embodiment of the present invention;

[0054] Figure 9 It is a schematic diagram of the internal structure principle of the intelligent terminal provided by the embodiment of the present invention. Detailed implementation manners

[0055] In the following description, specific details such as specific system structures and technologies are presented for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present invention. However, those skilled in the art should clearly understand that the present invention can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present invention.

[0056] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0057] It should also be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0058] It should be further understood that the term "and / or" used in the specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the related listed items, and includes these combinations.

[0059] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Persons skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0060] With the development of science and technology, various intelligent services provided by intelligent devices are increasingly widely applied in users' work and life. Among them, location-based intelligent services have also been widely used. When using location-based intelligent services, it is necessary to obtain the locations that a user may pass through during continuous movement.

[0061] In the prior art, usually the movement trajectory of a pedestrian is directly collected (for example, collected through positioning technologies such as GPS), and each point on this movement trajectory is regarded as the location passed by the pedestrian during continuous movement. The problem with the prior art is that the above-mentioned movement trajectory directly collected contains uncertainty errors and does not represent the true trajectory of the pedestrian. Therefore, directly regarding each point on the above-mentioned movement trajectory as the location passed by the pedestrian during continuous movement is not conducive to improving the accuracy of obtaining the area where the pedestrian may be located, and thus is not conducive to enhancing the usage accuracy and effect of location-based intelligent services.

[0062] Specifically, location-based intelligent services play an indispensable role in the application of intelligent terminals, and how to accurately obtain continuous and reliable location information of users in a large-scale scenario is the key to location-based intelligent services. Modeling the uncertainty of pedestrian movement in the actual application process (especially in complex and diverse urban environments) is an important task to promote geospatial data mining and analysis.

[0063] In an application scenario, an uncertainty prediction model can be used for modeling the uncertainty of pedestrian movement. However, usually, predictions are directly made based on the collected pedestrian trajectory, which is easily affected by the uncertainty contained in the pedestrian trajectory. Specifically, due to the complexity and diversity of urban indoor scenarios, there are the following problems with the positioning technology and uncertainty error estimation technology in large-scale urban indoor scenarios based on intelligent terminals: the current uncertainty prediction model only considers the movement distance or speed and cannot well adapt to time-varying measurement errors; there is a lack of continuous trajectory derivation and optimization methods in indoor scenarios, making it difficult to further reduce the influence of uncertainty errors; there is a lack of an overall calculation and evaluation scheme for the uncertainty area of large-scale indoor trajectories.

[0064] To solve at least one of the above-mentioned multiple problems, in the solution of this embodiment, motion information corresponding to the continuous trajectory of the target pedestrian and position information of reference points in the continuous trajectory of the target pedestrian are collected and obtained. Among them, the motion information includes a plurality of pedestrian step values and a plurality of heading values, and one of the pedestrian step values corresponds to one of the heading values. The reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the reference points includes the coordinate values of the reference points; according to the motion information and the position information of the reference points, a reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian is obtained through a preset position recursion model, and a reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian is obtained after optimizing the reconstructed trajectory according to the position information of the reference points; target features corresponding to each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory are obtained. Among them, the target features corresponding to one reconstructed and optimized coordinate point include the reconstructed and optimized step value, the reconstructed and optimized heading value, the position update time difference value, the current cumulative step number value, and the speed value corresponding to this reconstructed and optimized coordinate point; according to the target features, an uncertainty error prediction model that has been trained is used to obtain the uncertainty error values corresponding to each of the above-mentioned reconstructed and optimized coordinate points; an uncertainty region corresponding to the reconstructed and optimized trajectory is obtained according to the reconstructed and optimized coordinate points and the uncertainty error values and used as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian, where the target uncertainty region is the region where the real trajectory corresponding to the continuous trajectory of the target pedestrian is located.

[0065] Compared with the prior art, in the solution of this embodiment, each point on the continuous trajectory of the target pedestrian that can be directly acquired is not regarded as the position passed by the pedestrian during continuous movement. Instead, according to the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference point therein, the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian is obtained through a preset position recursion model, and further optimized to obtain a reconstructed and optimized trajectory. By reconstructing and optimizing the trajectory, the uncertainty included in the trajectory is eliminated (or minimized as much as possible), which is beneficial to improving the accuracy of obtaining the position of the user (i.e., the pedestrian). Then, the target features of each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory are extracted, and the uncertainty error value of each reconstructed and optimized coordinate point is calculated through a trained uncertainty error prediction model. Finally, the corresponding target uncertainty region is generated according to each reconstructed and optimized coordinate point and its corresponding uncertainty error value, which is beneficial to obtaining more accurately the area that the user may pass through, that is, improving the accuracy of obtaining the user's position. The target uncertainty region represents the region where the true trajectory corresponding to the continuous trajectory of the target pedestrian is located, that is, the region where the points that the pedestrian may pass through during continuous movement are determined after eliminating the uncertainty error of the trajectory. In this way, the solution of this embodiment can reduce the influence of uncertainty in the continuous trajectory of the pedestrian, reduce the influence of errors in the acquisition process of the original continuous trajectory of the target pedestrian, and generate a target uncertainty region corresponding to the true trajectory of the pedestrian, which is beneficial to improving the accuracy of obtaining the area where the pedestrian may be located. Furthermore, it is beneficial to improve the accuracy and effect of using location-based intelligent services.

[0066] Furthermore, the solution of this embodiment can be specifically applied to the application scenario of estimating the uncertainty region of the pedestrian trajectory in a large-scale indoor space for intelligent terminals, so as to improve the positioning and trajectory reconstruction accuracy of intelligent terminals in a large-scale urban indoor space. At the same time, a hybrid deep learning model is used to model and analyze the uncertainty error of the reconstructed and optimized trajectory, and finally a target uncertainty region that can reflect the region where the true trajectory is located is generated.

[0067] Specifically, in this embodiment, important factors (i.e., multiple target features) affecting the positioning trajectory accuracy in a large-scale space are comprehensively considered, and the error sources (i.e., the continuous trajectory of the target pedestrian containing uncertainty) are modeled and evaluated, and the original continuous trajectory of the target pedestrian obtained by positioning is reconstructed and optimized; at the same time, a hybrid deep learning model integrating a one-dimensional convolutional neural network layer, a long short-term memory neural network layer, and a fully connected layer is designed as an uncertainty error prediction model, and this uncertainty error prediction model is trained by extracting factors (i.e., target features) affecting trajectory uncertainty; finally, the uncertainty of the pedestrian trajectory (i.e., the continuous trajectory of the target pedestrian containing uncertainty) in a large-scale indoor space is effectively estimated, and its corresponding target uncertainty region is generated, which is beneficial to improving the acquisition accuracy of the area where the pedestrian may be located, and further beneficial to improving the usage accuracy and effect of location-based intelligent services.

[0068] Exemplary method

[0069] As Figure 1 shown, an embodiment of the present invention provides a method for obtaining an uncertainty region of a continuous pedestrian trajectory. Specifically, the above method includes the following steps:

[0070] Step S100, collect and obtain the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference points in the continuous trajectory of the target pedestrian. Among them, the above motion information includes a plurality of pedestrian step values and a plurality of heading values, and one of the above pedestrian step values corresponds to one of the above heading values. The above reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the above reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the above reference points includes the coordinate values of the above reference points.

[0071] Among them, the above continuous trajectory of the target pedestrian is a continuous trajectory of the target pedestrian that can be collected by a mobile device (or based on positioning technologies such as GPS). It should be noted that the continuous trajectory of the target pedestrian obtained by collection contains uncertainty errors, so the continuous trajectory of the target pedestrian obtained by collection does not represent the true trajectory during the pedestrian's movement.

[0072] Specifically, in this embodiment, it is not necessary to obtain the complete continuous trajectory of the target pedestrian for processing, and only the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of each reference point in the continuous trajectory of the target pedestrian need to be obtained. Among them, the above motion information includes a plurality of pedestrian step values and a plurality of heading values collected during the movement of the target pedestrian (i.e., the user), and one pedestrian step value and one heading value are associated to form the operation data corresponding to a position. Among them, the above pedestrian step value represents the step length corresponding to one step of the target pedestrian, and the above heading value represents the direction of the target pedestrian's movement in one step corresponding to the step length value.

[0073] In this embodiment, the position information of multiple reference points included in the continuous trajectory of the target pedestrian is also collected and obtained. The above position information is used to specifically indicate the spatial position of the corresponding reference point. Among them, the above reference point is a point whose real position can be determined in the continuous trajectory of the target pedestrian (for example, the target pedestrian can use a smart device to scan a QR code to obtain the position of the reference point). The above position information may include specific longitude and latitude information, position grid codes or coordinate values for indicating positions, etc. In this embodiment, coordinate values are taken as an example for specific description, but it is not a specific limitation.

[0074] It should be noted that in this embodiment, the number of pedestrian step values (and their corresponding heading values) collected and obtained can be greater than the number of reference points, that is, it is not necessary to determine the positions of too many reference points with clear positions during the movement. For example, only the starting point and the ending point of the trajectory can be set as reference points, or an appropriate number of reference points can be set during the movement (for example, a QR code corresponding to a reference point is set every 50 meters to determine the reference point position), which can avoid the user (i.e., the target pedestrian) from repeatedly scanning the code for positioning, and is beneficial to improving the data collection efficiency and the user experience.

[0075] Specifically, in this embodiment, the above collection and acquisition of the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference points in the continuous trajectory of the target pedestrian include: the heading value is collected and obtained in real time through the magnetometer, accelerometer and gyroscope in the movable smart device corresponding to the target pedestrian, and the pedestrian step value is obtained through the accelerometer; through the above movable smart device, the coordinate value of the above reference point is obtained according to at least one of a variety of preset positioning methods, where the above variety of preset positioning methods include a positioning method based on a wireless communication network and a positioning method based on code scanning.

[0076] Among them, the movable smart device corresponding to the target pedestrian is a movable smart device carried by the target pedestrian, such as a mobile phone, a tablet computer, a smart watch, etc., which is not specifically limited here. Devices such as a magnetometer, an accelerometer and a gyroscope are provided in the above movable smart device for collecting information of the target pedestrian during the movement in real time (at a certain frequency). The above positioning method based on a wireless communication network may include positioning through Wi-Fi and / or positioning through Bluetooth, and the above positioning method based on code scanning may include a method of scanning a QR code for positioning.

[0077] It should be noted that the method for obtaining the uncertainty region of the continuous trajectory of pedestrians in this embodiment can be applied to devices such as computers, smartphones, and Internet of Things terminals, and is used for estimating the uncertainty of the pedestrian trajectory and generating the uncertainty region in a large-scale indoor space. By collecting, processing, and fusing multi-source sensor data (i.e., motion information collected by magnetometers, accelerometers, gyroscopes, etc.) and position information obtained from wireless signal data, the trajectory of pedestrians indoors is reconstructed. At the same time, a hybrid deep learning model is built to calculate (or estimate) the uncertainty error of the reconstructed trajectory, and finally, high-precision measurement of the trajectory uncertainty region in a large-scale urban indoor space is achieved.

[0078] Step S200, according to the above-mentioned motion information and the position information of the above-mentioned reference point, obtain the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model, and optimize the reconstructed trajectory according to the position information of the above-mentioned reference point to obtain the reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian.

[0079] Among them, the above-mentioned preset position recursion model is a model that is preset and trained to recursively reconstruct the trajectory position during the movement of the target pedestrian. In an application scenario, the graph optimization method is used to globally optimize the step length and heading value involved in the pedestrian trajectory in combination with the landmark information to reconstruct the original indoor trajectory. It should be noted that during the training of the uncertainty error prediction model, the corresponding training data of the uncertainty error prediction model can also be collected and processed through the above-mentioned position recursion model.

[0080] Specifically, the above-mentioned obtaining the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model according to the above-mentioned motion information and the position information of the above-mentioned reference point includes: inputting the above-mentioned motion information and the position information of the above-mentioned reference point into the above-mentioned preset position recursion model, and obtaining the above-mentioned reconstructed trajectory according to the preset position recursion formula in the above-mentioned preset position recursion model. Among them, the above-mentioned reconstructed trajectory includes a plurality of successively connected reconstructed coordinate points, and one above-mentioned reconstructed coordinate point is connected to the next reconstructed coordinate point according to the reconstructed step length value and reconstructed heading value corresponding to the reconstructed coordinate point.

[0081] After optimizing the above-mentioned reconstructed trajectory according to the position information of the above-mentioned reference points, the reconstructed optimized trajectory corresponding to the continuous trajectory of the target pedestrian is obtained, including: obtaining each reconstructed coordinate point corresponding to each of the above-mentioned reference points as a target point; according to the position information of the above-mentioned reference points, the position information of the above-mentioned target points, and a preset loss function, with the goal of minimizing the position information loss value between the above-mentioned reference point and its corresponding target point, optimizing the reconstruction step value and the reconstruction heading value of each of the above-mentioned target points to obtain corresponding reconstructed optimized step values and reconstructed optimized heading values, taking the optimized above-mentioned target points as reconstructed optimized coordinate points, and obtaining the above-mentioned reconstructed optimized trajectory, where the above-mentioned reconstructed optimized trajectory includes multiple sequentially connected reconstructed optimized coordinate points, and one above-mentioned reconstructed optimized coordinate point is connected to the next reconstructed optimized coordinate point according to the reconstructed optimized step value and the reconstructed optimized heading value corresponding to the reconstructed optimized coordinate point.

[0082] Among them, the above-mentioned position information loss value is used to indicate the position difference between the target point and the corresponding reference point, and the above-mentioned preset loss function is a function preset for calculating the loss value between positions, or the corresponding position information loss value can be calculated according to the preset loss function model.

[0083] It should be noted that when the number of generated reconstructed coordinate points is less than or equal to the number of reference points, only the reconstructed coordinate points corresponding to the reference points are obtained for optimization and the reconstructed optimized coordinate points are obtained. In this embodiment, when the number of generated reconstructed coordinate points is greater than the number of reference points, after taking the optimized target points as the reconstructed optimized coordinate points, all the remaining reconstructed coordinate points that do not need to be optimized (that is, the reconstructed coordinate points without corresponding reference points) are taken as the reconstructed optimized coordinate points. That is, in this embodiment, the number of reconstructed coordinate points is greater than the number of reference points, and the finally obtained reconstructed optimized coordinate points include the points obtained by optimizing the target points and the reconstructed coordinate points that do not need to be optimized. And the number of finally obtained reconstructed optimized coordinate points is equal to the number of reconstructed coordinate points.

[0084] Specifically, in this embodiment, a position recurrence model of the pedestrian's trajectory in the indoor environment is pre-constructed, and the above-mentioned position recurrence model can be constructed and trained in advance according to the real-time collected heading value and pedestrian step value, as well as the corresponding real position information. In this embodiment, the preset position recurrence formula in the above-mentioned position recurrence model is shown in the following formula (1):

[0085]

[0086] Among them, STP i (L i ,θ i ) represents the coordinate value of the i-th reconstructed coordinate point calculated, R 01Represents the starting point (coordinates) of the continuous trajectory of the target pedestrian, which can be obtained through Wi-Fi, Bluetooth, QR codes, etc. n represents the number of points for which recursion needs to be performed. In this embodiment, n is the same as the number of pedestrian step length values (and heading values) corresponding to the continuous trajectory of the target pedestrian. L i Represents the i-th pedestrian step length value of the motion information of the continuous trajectory of the above-mentioned target pedestrian, and can also be regarded as the step length value corresponding to the i-th point or the i-th step of the target pedestrian's movement in the continuous trajectory of the above-mentioned target pedestrian. Similarly, θ i Represents the i-th heading value of the motion information of the continuous trajectory of the above-mentioned target pedestrian.

[0087] Figure 2 Is a schematic diagram of a reconstructed trajectory obtained by reconstruction provided by an embodiment of the present invention. Among them, t1 represents the acquisition time corresponding to the first pedestrian step length value, t i Represents the acquisition time corresponding to the i-th pedestrian step length value, and so on. STP i Is equivalent to the calculated STP i (L i , θ i ), and so on, which will not be elaborated here. It should be noted that, as Figure 2 shown, the first point STP1 represents the starting point of the continuous trajectory of the target pedestrian, and its coordinates are the same as R in formula (1). 01 In this embodiment, the reconstructed coordinate points are sequentially obtained according to the order of the acquisition time, and the reconstructed coordinate points are sequentially connected to obtain Figure 2 the reconstructed trajectory shown. Figure 2 In STP i represents a reference point, and its coordinates are also obtained through Wi-Fi, Bluetooth, QR codes, etc.

[0088] It should be noted that in an application scenario, trajectory reconstruction can also be performed based on other pre-trained position recursion models. The above-mentioned pre-trained position recursion model can be a neural network model pre-trained according to position recursion training data. The above-mentioned position recursion training data includes pre-collected position recursion training step length values, position recursion training heading values, and multiple marked point coordinates. The above-mentioned marked point coordinates are real coordinates obtained by actual acquisition, and the number of marked point coordinates is the same as the number of position recursion training step length values and the number of position recursion training heading values. During training, the position recursion training step length values and position recursion training heading values in the position recursion training data are input into the position recursion model to obtain the predicted reconstructed point coordinates generated by the model. The model parameters of the position recursion model are adjusted according to the loss value between the predicted reconstructed point coordinates and the corresponding marked point coordinates, and the training is repeatedly iterated until the preset recursion training iteration threshold is reached or the loss value is less than the preset recursion training loss threshold to obtain the pre-trained position recursion model.

[0089] In this embodiment, a loss function for graph optimization (i.e., a preset loss function) is constructed in advance according to the above position recursion model. After obtaining more than two target points (i.e., the reconstructed coordinate points corresponding to the reference points), the position information loss value between the reference point and its corresponding target point is calculated according to this loss function. It should be noted that the position information corresponding to the above reference point can be directly read after the mobile phone corresponding to the target pedestrian scans the QR code corresponding to the reference point. After each new target point is obtained, a loss value calculation is performed. In this embodiment, the calculation for one of the target points is taken as an example for illustration. The above loss function is shown in the following formula (2):

[0090] ξ(L i ,θ i )=(z - STP i (L i ,θ i )) T ρ -1 (z - STP i (L i ,θ i )) (2)

[0091] Among them, ξ(L i ,θ i ) represents the position information loss value between the i-th reconstructed coordinate point and its corresponding reference point, and its specific value is obtained by the calculation method on the right side of formula (2). It should be noted that in this embodiment, when the i-th reconstructed coordinate point is the target point corresponding to a reference point, the i-th reconstructed coordinate point is regarded as the i-th target point. Therefore, the numbers between adjacent target points are not necessarily consecutive. z represents the coordinate value of the reference point corresponding to the i-th reconstructed coordinate point (obtained through Wi-Fi, Bluetooth, QR code, etc.), STP i (L i ,θ i ) represents the coordinate value of the i-th reconstructed coordinate point, and the superscript T represents matrix transpose. ρ -1 represents a preset skew-symmetric matrix, and its specific dimensions, etc., are set in advance according to actual requirements.

[0092] It should be noted that although the step size and heading of the reconstructed coordinate points in the reconstructed trajectory reconstructed according to the above formula (1) are known, there are errors, so optimization is required. In this embodiment, the reconstructed trajectory is optimized with the goal of minimizing the position information loss value obtained by the above formula (2). In one application scenario, an optimization model can be constructed in advance, and this optimization model optimizes the reconstructed trajectory according to the following formula (3):

[0093]

[0094] Among them, (L i , θ i ) * represents the combination of the reconstructed optimization step value and the reconstructed optimization heading value obtained after optimization corresponding to the i-th reconstructed coordinate point. In this embodiment, it can be represented by and respectively representing the reconstructed optimization step value and the reconstructed optimization heading value obtained after optimization corresponding to the i-th target point. represents solving the minimum value of the loss function. When the loss function reaches the minimum value, it is recorded as obtaining the optimal trajectory optimization result.

[0095] The above formula (3) shows how to perform trajectory optimization to obtain the optimal step value and heading value. It should be noted that for the reconstructed trajectory that needs to be optimized, only the position corresponding to the reconstructed coordinate point (i.e., the target point) of the reference point can be regarded as real. All the points between two target points are derived according to the above formula (1) based on the collected step value and heading value. Therefore, it is necessary to continue to optimize according to formulas (2) and (3). Optimize the reconstructed trajectory according to the above formulas (2) and (3), and the reconstructed optimization step values corresponding to the optimized reconstructed trajectory coordinate points are not necessarily the same, and the corresponding reconstructed optimization heading values are not necessarily the same.

[0096] It should be noted that a method for determining whether the loss function reaches the minimum value can be preset. For example, when the loss value calculated by the loss function reaches the preset minimum loss threshold, it is considered to reach the minimum value, or when the difference between the currently calculated loss value and the loss value obtained in the previous calculation is less than the preset difference threshold (i.e., convergence), it is considered to reach the minimum value. There can also be other methods, which are not specifically limited here.

[0097] In this embodiment, the input data of the above optimization model is the set of the step value and the heading value before the optimization of the current trajectory (i.e., the reconstructed trajectory), and the output data is the set of the optimized step value and the heading value. This model is used to optimize the process of obtaining the optimal solution.

[0098] Figure 3 is a schematic diagram of the acquisition route of the continuous trajectory of the target pedestrian provided by the embodiment of the present invention. Specifically, the data set in this embodiment is collected from a large indoor shopping mall. Different acquisition route schematic diagrams are as Figure 3 shown. Figure 4 is a schematic diagram for comparing the reconstructed optimized trajectory and the real trajectory provided by the embodiment of the present invention. Figure 4 In

[0099] Step S300: Obtain the target features corresponding to each reconstruction and optimization coordinate point in the above-mentioned reconstructed and optimized trajectory. Among them, the target features corresponding to one reconstruction and optimization coordinate point include the reconstruction and optimization step value, reconstruction and optimization heading value, position update time difference, current cumulative step value, and speed value corresponding to this reconstruction and optimization coordinate point.

[0100] Among them, the above-mentioned target features are features extracted from the reconstructed and optimized trajectory, which are used to reflect the motion state of each step in the reconstructed and optimized trajectory. In this embodiment, important factors affecting the accuracy of the large-scale indoor pedestrian trajectory after reconstruction and optimization are analyzed and extracted, and modeling is carried out. The factors for modeling (i.e., target features) are further used to train a deep learning model (i.e., uncertainty error prediction model) to improve the model estimation accuracy. Therefore, based on the target features, the uncertainty error value can be calculated through this uncertainty error prediction model.

[0101] In this embodiment, the target features corresponding to the u-th reconstruction and optimization coordinate point include the reconstruction and optimization step value obtained from the above-mentioned reconstructed and optimized trajectory Reconstruction and optimization heading value The position update time difference Δt required for each two position updates, the current cumulative step value step corresponding to the current moment u And the speed value between the current moment and the previous moment Among them, the position update time difference required for each two position updates is the time interval for the mobile phone to collect data such as step values. For example, in this embodiment, the mobile phone can be set to collect a pedestrian step value every 1 second, then the corresponding position update time difference Δt is 1 second. The above-mentioned current moment is the data collection moment corresponding to the u-th reconstruction and optimization coordinate point, that is, the moment corresponding to the u-th pedestrian step value in the motion information of the target pedestrian's continuous trajectory collected by the mobile phone, and it can be directly obtained through the mobile phone collecting data. The above-mentioned current cumulative step value represents how many steps the target pedestrian has walked in total at the current moment, and it can also be directly obtained through the mobile phone collecting data. In an application scenario, step u That is, it represents walking u steps in total.

[0102] Furthermore, to improve the prediction accuracy of the uncertainty error, more features can be extracted and combined as the above-mentioned target features. In this embodiment, the target features corresponding to one (the u-th) reconstruction and optimization coordinate point also include the current cumulative walking distance, current distance ratio, current duration ratio, current step ratio, current heading change amount, and current cumulative heading change amount corresponding to this reconstruction and optimization coordinate point.

[0103] Among them, the above-mentioned current cumulative walking distance is the sum of the reconstruction optimization step lengths of all currently passed points, where u represents the serial number of the last currently passed point, and the above currently passed points include all the reconstruction optimization coordinate points from the starting point of the above reconstruction optimization trajectory to this reconstruction optimization coordinate point.

[0104] The above current distance ratio is the ratio of the above current cumulative walking distance to the total distance of the trajectory. n represents the total number of reconstruction optimization coordinate points. The above total distance of the trajectory is the sum of the reconstruction optimization step length values of all the trajectory passed points, and the above trajectory passed points include all the reconstruction optimization coordinate points from the starting point of the above reconstruction optimization trajectory to the end point of the above reconstruction optimization trajectory.

[0105] The above current duration ratio PI t (u)=T(u) / T total , which is the ratio of the current duration to the total duration of the trajectory. The above current duration is the cumulative duration corresponding to all the above currently passed points, and the above total duration of the trajectory is the cumulative duration from the starting point of the above reconstruction optimization trajectory to the end point of the above reconstruction optimization trajectory. Specifically, the current duration and the total duration corresponding to the current moment can be obtained by counting the data collected in the mobile phone. For example, the total duration of the trajectory can be obtained by counting the time spent by the pedestrian walking through the entire trajectory.

[0106] The above current step ratio PI s (u)=step u / step total , which is the ratio of the above current cumulative step value to the total number of steps of the trajectory. The above total number of steps of the trajectory is the cumulative number of steps from the starting point of the above reconstruction optimization trajectory to the end point of the above reconstruction optimization trajectory, and can be obtained by counting the number of steps spent by the pedestrian walking through the entire trajectory.

[0107] The above current heading change is the difference between the reconstruction optimization heading value corresponding to this reconstruction optimization coordinate point and the reconstruction optimization heading value corresponding to the previous reconstruction optimization coordinate point.

[0108] The above current cumulative heading change is the sum of the heading starting change amounts corresponding to all the above currently passed points. The above heading starting change amount is the difference between the reconstruction optimization heading value between the above currently passed point and the starting point of the above reconstruction optimization trajectory.

[0109] It should be noted that in this embodiment, the data representing the ratio above is expressed in the form of a percentage for unified calculation.

[0110] Step S400, according to the above target features, obtain the uncertainty error values corresponding to each of the above reconstruction optimization coordinate points through the trained uncertainty error prediction model.

[0111] Among them, the above-trained uncertainty error prediction model is a pre-trained model for calculating and obtaining the uncertainty error value of a corresponding point according to the input target features.

[0112] In an application scenario, the above uncertainty error prediction model is trained according to the following steps: input the training features in the training data into the uncertainty error prediction model, and generate the predicted uncertainty error value corresponding to the training features through the above uncertainty error prediction model. Among them, the above training data includes training information groups corresponding to multiple consecutive training coordinate points (a group of consecutive training coordinate points are training coordinate points on the same training trajectory), and each group of training information groups includes the training features and the true uncertainty error value corresponding to the training coordinate point; according to the above predicted uncertainty error value and its corresponding true uncertainty error value, adjust the model parameters of the above uncertainty error prediction model, and continue to execute the above step of inputting the training features in the training data into the uncertainty error prediction model until the preset training conditions are met (for example, the number of iterations reaches the preset training iteration number threshold, or the loss between the obtained uncertainty error value and its corresponding true uncertainty error value is less than the preset error loss threshold), so as to obtain the trained uncertainty error prediction model.

[0113] It should be noted that the above training features correspond to the above target features, that is, each specific feature in the training features corresponds to each specific feature in the target features. The acquisition, processing, and extraction processes of the above training features can refer to the processing process of the target features above and will not be elaborated here.

[0114] In this embodiment, for any one of the reconstructed and optimized coordinate points, the uncertainty error value corresponding to the reconstructed and optimized coordinate point is calculated and obtained according to the following steps: splice the target features corresponding to the above reconstructed and optimized coordinate points to obtain a multi-dimensional feature vector; input the above feature vector into the above trained uncertainty error prediction model, and obtain the uncertainty error value corresponding to the above reconstructed and optimized coordinate point output by the above uncertainty error prediction model; among them, the above uncertainty error prediction model is a hybrid deep learning model, and the above uncertainty error prediction model includes a one-dimensional convolutional neural network layer, a long short-term memory neural network layer, and a fully connected layer.

[0115] Specifically, in this embodiment, a hybrid deep learning model (i.e., the uncertainty error prediction model) is used for uncertainty error estimation, and the advantages of a one-dimensional convolutional neural network and a long short-term memory network are combined to achieve trajectory uncertainty error estimation that takes into account both time continuity and multi-feature characteristics. It should be noted that the model parameters of the model are pre-trained through the above training process.

[0116] Specifically, in this embodiment, a one-dimensional convolutional neural network layer is designed as the first layer of the uncertainty error prediction model for learning and extracting trajectory features, including determining the format, dimension, specific parameters of the input and output vectors, and the neural network connection mode. The operation function corresponding to the above one-dimensional convolutional neural network layer is shown in the following formula (4):

[0117]

[0118] Among them, y j represents the output value of the one-dimensional convolutional neural network layer, x i represents the input value of the one-dimensional convolutional neural network layer, ζ represents a preset proportional value (a preset constant), k ij and b j represent the connection weight value and the bias value respectively. k ij and b j are model parameters determined during the model training process. It should be noted that the input value of the above one-dimensional convolutional neural network layer is a multi-dimensional feature vector (such as an 11-dimensional feature vector) obtained by splicing the above target features. The output value of the above one-dimensional convolutional neural network layer is a feature quantity obtained by further understanding.

[0119] Furthermore, the constructed long short-term memory (LSTM) neural network layer is used as the second layer of the uncertainty error prediction model for learning and training continuous feature vectors, and its corresponding operation function is shown in the following formula (5):

[0120]

[0121] Among them, X t is the feature input quantity corresponding to the above long short-term memory neural network layer, that is, the output value y j of the one-dimensional convolutional neural network layer. The above long short-term memory neural network layer has multiple nodes, h t-1 represents the input value received by the previous node, and h t represents the output value transmitted to the next node. W f , W i , W C and W o are the corresponding weight values, b f , b i , b C and b o are the corresponding bias compensation amounts, and their specific values can be determined by pre-training. f t represents the forgetting gate, which is used to indicate which features in C t-1 are used to calculate C t , C t-1Represents the state of the previous unit, C t Represents the state of the current unit. i t Represents the input gate, which is calculated from the feature input quantity X t and the input value h of the previous node received t-1 σ is a preset constant parameter, is the unit state update amount, o t represents the state value of the output gate, h t is the final output quantity, which is obtained from the output gate and the current unit state C t value.

[0122] Furthermore, the constructed fully connected layer is used as the third layer of the uncertainty error prediction model to connect the output result of the LSTM network layer and calculate the uncertainty error value of the final output MLP(y i )。Among them, y i is the output value h of the LSTM network layer t , and it is also used as the input of the fully connected layer. The fully connected layer actually performs the final summary output on the output vector of the LSTM network, and finally obtains the uncertainty estimation result (i.e., the uncertainty error value), and this uncertainty error value represents how much the position error corresponding to the current gait is.

[0123] Figure 5 is a schematic structural diagram of an uncertainty error prediction model provided by an embodiment of the present invention. As Figure 5 shown, in a specific application scenario, the above uncertainty error prediction model specifically includes a convolutional layer, a pooling layer, a normalization layer, a long short-term memory network layer, and a fully connected layer. Among them, Figure 5 the part shown in the dotted box in is the specific structural schematic diagram of the long short-term memory network layer. As Figure 5 shown, in this embodiment, by inputting a feature vector into the above uncertainty error prediction model, the output vector of the model (i.e., the uncertainty error value corresponding to the reconstructed and optimized coordinate point) can be finally obtained.

[0124] Step S500, obtain the uncertainty region corresponding to the above reconstructed and optimized trajectory according to the above reconstructed and optimized coordinate point and the above uncertainty error value, and use it as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian, where the above target uncertainty region is the region where the real trajectory corresponding to the continuous trajectory of the target pedestrian is located.

[0125] Specifically, the above-mentioned target uncertainty region is the region where the true trajectory (which cannot be directly obtained) corresponding to the continuous trajectory of the above-mentioned target pedestrian (the trajectory directly obtained through the positioning technology and including errors) is located. Based on the solution of this embodiment, a more accurate uncertainty region can be calculated and generated, and the possible positions that the target pedestrian (i.e., the user) may pass through can be determined more accurately, which is beneficial to improving the accuracy of obtaining the region where the pedestrian may be located, and further beneficial to improving the accuracy and effect of using location-based intelligent services.

[0126] In this embodiment, obtaining the uncertainty region corresponding to the reconstructed and optimized trajectory based on the above-mentioned reconstructed and optimized coordinate points and the above-mentioned uncertainty error value and using it as the target uncertainty region corresponding to the continuous trajectory of the above-mentioned target pedestrian includes: respectively obtaining the error circles corresponding to each of the above-mentioned reconstructed and optimized coordinate points, where the center of an error circle corresponding to a reconstructed and optimized coordinate point is the reconstructed and optimized coordinate point, and the radius is the uncertainty error value corresponding to the reconstructed and optimized coordinate point; sequentially obtaining the union regions of the error circles corresponding to each of the above-mentioned reconstructed and optimized coordinate points, where the union region of the error circles corresponding to a reconstructed and optimized coordinate point is the union region corresponding to the error circles of this reconstructed and optimized coordinate point and the next adjacent reconstructed and optimized coordinate point; taking the union of the union regions of the error circles corresponding to all the above-mentioned reconstructed and optimized coordinate points as the uncertainty region corresponding to the reconstructed and optimized trajectory, and using the uncertainty region corresponding to the reconstructed and optimized trajectory as the target uncertainty region corresponding to the continuous trajectory of the above-mentioned target pedestrian.

[0127] Specifically, a construction of the overall uncertainty region of the trajectory is built by expanding the size of the single-step uncertainty error in the trajectory to the global, and the region edge is extracted to effectively cover the true activity trajectory of the pedestrian. In one application scenario, what is finally obtained is the region edge, indicating that this uncertainty region is the region where the user's trajectory may be located. In the actual scenario, since the obtained pedestrian trajectories all contain uncertainty errors, the obtained trajectories do not represent the true trajectories of the pedestrians, and it is necessary to estimate the region where the true trajectories of the pedestrians may exist.

[0128] Specifically, taking the position corresponding to each step (i.e., each reconstructed and optimized coordinate point) in the above-mentioned reconstructed and optimized trajectory as a center (or node), and using the solved corresponding uncertainty error value as the radius to construct the corresponding error circle (x - x0) 2 +(y - y0) 2 =r 2 . Where (x, y) is the coordinate to be solved, (x0, y0) is the coordinate of the center, and r represents the radius. Sequentially solve the intersection points of two adjacent error circles (i.e., the intersection points of two adjacent uncertain regions) to determine the union region of the error circles. Specifically, the intersection points of two error circles are Among them, (x, y) are the coordinates to be solved, a represents the distance between the centers of the two error circles, and h represents the distance between the intersection points of the two error circles. (x0, y0) represents the center coordinates of the first error circle, (x1, y1) represents the coordinates of a known point on the first error circle, and (x2, y2) represents the coordinates of a known point on the second error circle.

[0129] Further, solve the area of the union of two adjacent uncertainty regions Among them, S represents the union of the uncertainty regions corresponding to two adjacent error circles, q1 and q2 represent the angles of the sectors in the intersection region of the two circles, r1 and r2 respectively represent the radii of the two circles, and d is the distance value between the centers of the two circles. Further, by comprehensively integrating the union of the union regions of the error circles corresponding to all the above-mentioned reconstructed and optimized coordinate points, the target uncertainty region corresponding to the above-mentioned reconstructed and optimized trajectory can be obtained.

[0130] Figure 6 is a schematic diagram of a target uncertainty region obtained in an embodiment of the present invention. Figure 6 The range enclosed by the dashed line in is the range of the above-mentioned target uncertainty region. For example, Figure 6 As shown, the target uncertainty region obtained according to the method of this embodiment basically covers the region where the true trajectory of the pedestrian is located, which is beneficial to improving the accuracy of obtaining the region where the pedestrian may be located.

[0131] In an application scenario, the above-mentioned reconstructed and optimized trajectory can also be downsampled, that is, only select some of the reconstructed and optimized coordinate points for subsequent calculations to reduce the data volume, which can reduce the data calculation volume on the premise of ensuring a certain accuracy and is beneficial to improving the processing efficiency.

[0132] In this embodiment, the solution in this application is also tested, trained, and compared with the solutions corresponding to other 4 different models (including the AUB model, the BAEE model, the UB model, and the LSTM model) on the dataset. Table 1 shows the corresponding results, and the comparison can be made according to the results shown in Table 1. By comparing the indoor trajectory uncertainty error estimation results of several different models, it can be found that the uncertainty region estimation model of the pedestrian indoor trajectory estimated by using the hybrid deep learning model provided in the embodiment of the present invention has achieved better estimation results, and the constructed target uncertainty region basically covers the true trajectory and has a higher coverage density. It should be noted that in Table 1, the AUB model, the BAEE model, the UB model, and the LSTM model respectively specifically represent the adaptive maximum distance model, the adaptive error ellipse model, the maximum distance model, and the long short-term memory model, ED represents the pre-set enhanced dataset, and HDL represents the uncertainty error prediction model in this embodiment.

[0133] Table 1

[0134]

[0135] Figure 7 It is a schematic diagram of another obtained target uncertainty region provided by an embodiment of the present invention. Figure 7 Specifically shown is a schematic diagram of the target uncertainty regions obtained after downsampling the same reconstructed and optimized trajectory using the AUB model, the BAEE model, the UB model, and the HDL model in this embodiment. As Figure 7 shown, the target uncertainty region obtained after downsampling by the HDL model in this embodiment can still cover the region where the true trajectory of the pedestrian is located, and a relatively smaller and more accurate (i.e., limited within a smaller range) region can be obtained, showing better effects compared with the other three models.

[0136] As can be seen from the above, in the method for obtaining the uncertainty region of the pedestrian's continuous trajectory provided by the embodiment of the present invention, each point on the directly collectable target pedestrian's continuous trajectory is not regarded as the position passed by the pedestrian during continuous movement. Instead, according to the motion information corresponding to the target pedestrian's continuous trajectory and the position information of the reference point therein, the reconstructed trajectory corresponding to the above target pedestrian's continuous trajectory is obtained through a preset position recursion model, and further optimized to obtain a reconstructed and optimized trajectory. By reconstructing and optimizing the trajectory, the uncertainty included in the trajectory is eliminated (or minimized as much as possible), which is beneficial to improving the accuracy of obtaining the position of the user (i.e., the pedestrian). Then, the target features of each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory are extracted, and the uncertainty error value of each reconstructed and optimized coordinate point is calculated through a trained uncertainty error prediction model. Finally, the corresponding target uncertainty region is generated according to each reconstructed and optimized coordinate point and its corresponding uncertainty error value, which is beneficial to more accurately obtaining the region that the user may pass through, that is, improving the accuracy of obtaining the user's position. This target uncertainty region represents the region where the true trajectory corresponding to the above target pedestrian's continuous trajectory is located, that is, the region where the points that the pedestrian may pass through during continuous movement are determined after eliminating the uncertainty error of the trajectory. In this way, the solution of this embodiment can reduce the influence of uncertainty in the pedestrian's continuous trajectory, reduce the influence of errors in the original acquisition process of the target pedestrian's continuous trajectory, and generate a target uncertainty region corresponding to the true trajectory of the pedestrian, which is beneficial to improving the accuracy of obtaining the region where the pedestrian may be located. Furthermore, it is beneficial to improving the accuracy and effect of using location-based intelligent services.

[0137] Specifically, in this embodiment, first, a graph optimization method is used to globally optimize the step lengths and heading values involved in the pedestrian trajectory in combination with the road punctuation information to reconstruct and optimize the original indoor trajectory. Then, the important factors affecting the accuracy of the reconstructed large-scale indoor pedestrian trajectory are analyzed and extracted, and a model is built. The factors in the model are further used to train a deep learning model to improve the model estimation accuracy. Next, a hybrid deep learning model is designed for estimating the uncertainty error of the reconstructed indoor trajectory, and the advantages of a one-dimensional convolutional neural network and a long short-term memory network are combined to achieve the estimation of the trajectory uncertainty error that takes into account both the time continuity characteristic and the multi-feature characteristic. Finally, a construction of the overall trajectory uncertainty region that expands from the magnitude of the single-step uncertainty error in the trajectory to the global is carried out, and the region edges are extracted to effectively cover the real activity trajectory of the pedestrian. This is beneficial to obtaining a better estimation result of the pedestrian trajectory uncertainty region in a large-scale indoor scenario.

[0138] Exemplary device

[0139] As Figure 8 shown in, corresponding to the above method for obtaining the uncertainty region of the pedestrian continuous trajectory, an embodiment of the present invention further provides a system for obtaining the uncertainty region of the pedestrian continuous trajectory. The above system for obtaining the uncertainty region of the pedestrian continuous trajectory includes:

[0140] A data acquisition module 610, configured to collect and obtain the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference points in the continuous trajectory of the target pedestrian. Among them, the above motion information includes a plurality of pedestrian step length values and a plurality of heading values, and one of the above pedestrian step length values corresponds to one of the above heading values. The above reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the above reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the above reference points includes the coordinate values of the above reference points;

[0141] A trajectory reconstruction module 620, configured to obtain the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model according to the above motion information and the position information of the reference points, and obtain the reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian after optimizing the reconstructed trajectory according to the position information of the reference points;

[0142] A target feature extraction module 630, configured to obtain the target features corresponding to each reconstructed and optimized coordinate point in the above reconstructed and optimized trajectory. Among them, the target feature corresponding to one reconstructed and optimized coordinate point includes the reconstructed and optimized step length value, the reconstructed and optimized heading value, the position update time difference value, the current cumulative step number value, and the speed value corresponding to the reconstructed and optimized coordinate point;

[0143] An error value calculation module 640 is configured to obtain the uncertainty error values corresponding to each of the above-mentioned reconstructed and optimized coordinate points according to the above-mentioned target features through a trained uncertainty error prediction model;

[0144] An uncertainty region generation module 650 is configured to obtain the uncertainty region corresponding to the above-mentioned reconstructed and optimized trajectory based on the above-mentioned reconstructed and optimized coordinate points and the above-mentioned uncertainty error values, and use it as the target uncertainty region corresponding to the continuous trajectory of the above-mentioned target pedestrian. The target uncertainty region is the region where the true trajectory corresponding to the continuous trajectory of the above-mentioned target pedestrian is located.

[0145] It should be noted that the specific structures and implementation manners of the above-mentioned uncertainty region acquisition system for pedestrian continuous trajectories and its various modules or units can refer to the corresponding descriptions in the above method embodiments, and will not be elaborated here.

[0146] It should be noted that the division methods of the various modules of the above-mentioned uncertainty region acquisition system for pedestrian continuous trajectories are not unique, and are not specifically limited here either.

[0147] Based on the above embodiments, the present invention also provides an intelligent terminal, and its principle block diagram can be as Figure 9 shown. The above intelligent terminal includes a processor, a memory, a network interface, and a display screen connected through a system bus. Among them, the processor of the intelligent terminal is used to provide computing and control capabilities. The memory of the intelligent terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a program for acquiring the uncertainty region of the pedestrian continuous trajectory. The internal memory provides an environment for the operation of the operating system and the program for acquiring the uncertainty region of the pedestrian continuous trajectory in the non-volatile storage medium. The network interface of the intelligent terminal is used to communicate with an external terminal through a network connection. When the program for acquiring the uncertainty region of the pedestrian continuous trajectory is executed by the processor, it implements the steps of any one of the above methods for acquiring the uncertainty region of the pedestrian continuous trajectory. The display screen of the intelligent terminal can be a liquid crystal display screen or an electronic ink display screen.

[0148] Those skilled in the art can understand that Figure 9 the principle block diagram shown is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the intelligent terminal to which the solution of the present invention is applied. The specific intelligent terminal may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0149] In one embodiment, an intelligent terminal is provided. The intelligent terminal includes a memory, a processor, and a program for obtaining an uncertainty region of a pedestrian's continuous trajectory stored on the memory and executable on the processor. When the program for obtaining an uncertainty region of a pedestrian's continuous trajectory is executed by the processor, the steps of any method for obtaining an uncertainty region of a pedestrian's continuous trajectory provided by the embodiments of the present invention are implemented.

[0150] The embodiments of the present invention also provide a computer-readable storage medium. A program for obtaining an uncertainty region of a pedestrian's continuous trajectory is stored on the computer-readable storage medium. When the program for obtaining an uncertainty region of a pedestrian's continuous trajectory is executed by a processor, the steps of any method for obtaining an uncertainty region of a pedestrian's continuous trajectory provided by the embodiments of the present invention are implemented.

[0151] It should be understood that the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0152] Those skilled in the art can clearly understand that, for the convenience and conciseness of description, only the above division of each functional unit and module is used as an example. In practical applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the above device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiments can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of the present invention. The specific working processes of the units and modules in the above device can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0153] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0154] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0155] In the embodiments provided by the present invention, it should be understood that the disclosed system / terminal device and method can be implemented in other ways. For example, the system / terminal device embodiments described above are only illustrative. For example, the above division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0156] If the above integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above method embodiments of the present invention, it can also be completed by a computer program instructing related hardware. The above computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above method embodiments can be implemented. Among them, the above computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The above computer-readable medium can include: any entity or device that can carry the above computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the above computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0157] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not deviate from the spirit and scope of the technical solutions of the present invention's various embodiments, and should all be included in the protection scope of the present invention.

Claims

1. A method for obtaining an uncertainty region for a pedestrian's continuous trajectory, characterized in that, The method for obtaining the uncertainty region of the continuous trajectory of a pedestrian includes the following steps: Collect the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference points in the continuous trajectory of the target pedestrian. Among them, the motion information includes a plurality of pedestrian step values and a plurality of heading values, and one pedestrian step value corresponds to one heading value. The reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the reference points includes the coordinate values of the reference points; According to the motion information and the position information of the reference points, obtain the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model, and obtain the reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian after optimizing the reconstructed trajectory according to the position information of the reference points; Obtain the target features corresponding to each reconstructed and optimized coordinate point in the reconstructed and optimized trajectory. Among them, the target features corresponding to one reconstructed and optimized coordinate point include the reconstructed and optimized step value, the reconstructed and optimized heading value, the position update time difference, the current cumulative step number, and the speed value corresponding to this reconstructed and optimized coordinate point; According to the target features, obtain the uncertainty error values corresponding to each reconstructed and optimized coordinate point through the trained uncertainty error prediction model; According to the reconstructed and optimized coordinate points and the uncertainty error values, obtain the uncertainty region corresponding to the reconstructed and optimized trajectory and use it as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian. Among them, the target uncertainty region is the region where the real trajectory corresponding to the continuous trajectory of the target pedestrian is located; The uncertainty error prediction model is a hybrid deep learning model. The uncertainty error prediction model includes a one-dimensional convolutional neural network layer, a long short-term memory neural network layer, and a fully connected layer. During the training process of the uncertainty error prediction model, the training data of the corresponding uncertainty error prediction model is collected and processed through the position recursion model; The target features corresponding to one reconstructed and optimized coordinate point further include the current cumulative walking distance, the current distance ratio, the current duration ratio, the current step ratio, the current heading change amount, and the current cumulative heading change amount corresponding to this reconstructed and optimized coordinate point; Among them, the current cumulative walking distance is the sum of the reconstructed and optimized step lengths of all current passing points. The current passing points include all reconstructed and optimized coordinate points from the starting point of the reconstructed and optimized trajectory to this reconstructed and optimized coordinate point; The current distance ratio is the ratio of the current cumulative walking distance to the total distance of the trajectory. The total distance of the trajectory is the sum of the reconstructed and optimized step values of all trajectory passing points. The trajectory passing points include all reconstructed and optimized coordinate points from the starting point of the reconstructed and optimized trajectory to the end point of the reconstructed and optimized trajectory; The current duration ratio is the ratio of the current duration to the total duration of the trajectory. The current duration is the cumulative duration corresponding to all the current passing points. The total duration of the trajectory is the cumulative duration from the starting point of the reconstructed and optimized trajectory to the end point of the reconstructed and optimized trajectory; The current step proportion is the ratio of the current cumulative step value to the total number of steps of the trajectory, and the total number of steps of the trajectory is the cumulative number of steps from the starting point to the ending point of the reconstructed and optimized trajectory; The current heading change amount is the difference between the reconstructed and optimized heading value corresponding to the reconstructed and optimized coordinate point and the reconstructed and optimized heading value corresponding to the previous reconstructed and optimized coordinate point; The current cumulative heading change amount is the sum of the heading starting change amounts corresponding to all the current passing points, and the heading starting change amount is the difference between the reconstructed and optimized heading value between the current passing point and the starting point of the reconstructed and optimized trajectory.

2. The method for obtaining the uncertainty region for the continuous trajectory of a pedestrian according to claim 1, wherein The acquisition of the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference points in the continuous trajectory of the target pedestrian includes: The heading value is acquired in real time through a magnetometer, an accelerometer and a gyroscope in the movable intelligent device corresponding to the target pedestrian, and the pedestrian step length value is acquired through the accelerometer; Through the movable intelligent device, the coordinate value of the reference point is acquired according to at least one of a plurality of preset positioning methods, wherein the plurality of preset positioning methods include a positioning method based on a wireless communication network and a positioning method based on code scanning.

3. The method for obtaining the uncertainty region for the continuous trajectory of pedestrians according to claim 1, wherein The acquisition of the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian according to the motion information and the position information of the reference point includes: The motion information and the position information of the reference point are input into the preset position recursion model, and the reconstructed trajectory is acquired according to a preset position recursion formula in the preset position recursion model, wherein the reconstructed trajectory includes a plurality of successively connected reconstructed coordinate points, and one reconstructed coordinate point is connected to the next reconstructed coordinate point according to the reconstructed step length value and the reconstructed heading value corresponding to the reconstructed coordinate point.

4. The method for obtaining the uncertainty region for the continuous trajectory of a pedestrian according to claim 3, characterized in that, The acquisition of the reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian after optimizing the reconstructed trajectory according to the position information of the reference point includes: Acquire each reconstructed coordinate point corresponding to each reference point and use it as a target point; According to the position information of the reference point, the position information of the target point and a preset loss function, with the goal of minimizing the position information loss value between the reference point and its corresponding target point, the reconstructed step length value and the reconstructed heading value of each target point are optimized to obtain the corresponding reconstructed and optimized step length value and the reconstructed and optimized heading value, the optimized target point is used as the reconstructed and optimized coordinate point, and the reconstructed and optimized trajectory is obtained, wherein the reconstructed and optimized trajectory includes a plurality of successively connected reconstructed and optimized coordinate points, and one reconstructed and optimized coordinate point is connected to the next reconstructed and optimized coordinate point according to the reconstructed and optimized step length value and the reconstructed and optimized heading value corresponding to the reconstructed and optimized coordinate point.

5. The method for obtaining the uncertainty region of the continuous trajectory of a pedestrian according to claim 1, wherein, For any reconstructed and optimized coordinate point, the uncertainty error value corresponding to the reconstructed and optimized coordinate point is calculated and acquired according to the following steps: The target features corresponding to the reconstructed and optimized coordinate point are spliced to obtain a multi-dimensional feature vector; Input the feature vector into the trained uncertainty error prediction model to obtain the uncertainty error value corresponding to the reconstructed and optimized coordinate point output by the uncertainty error prediction model.

6. The method for obtaining the uncertainty region for the continuous trajectory of a pedestrian according to claim 1, wherein, The obtaining of the uncertainty region corresponding to the reconstructed and optimized trajectory based on the reconstructed and optimized coordinate point and the uncertainty error value and using it as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian includes: Respectively obtain the error circles corresponding to each of the reconstructed and optimized coordinate points. Among them, the center of the error circle corresponding to one of the reconstructed and optimized coordinate points is the reconstructed and optimized coordinate point, and the radius is the uncertainty error value corresponding to the reconstructed and optimized coordinate point. Successively obtain the union regions of the error circles corresponding to each of the reconstructed and optimized coordinate points. Among them, the union region of the error circles corresponding to one of the reconstructed and optimized coordinate points is the union region corresponding to the error circles of this reconstructed and optimized coordinate point and the next adjacent reconstructed and optimized coordinate point. Take the union of the union regions of the error circles corresponding to all the reconstructed and optimized coordinate points as the uncertainty region corresponding to the reconstructed and optimized trajectory, and use the uncertainty region corresponding to the reconstructed and optimized trajectory as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian.

7. An uncertainty region acquisition system for continuous pedestrian trajectories, characterized in that, The system for obtaining the uncertainty region of the continuous trajectory of a pedestrian includes: A data acquisition module, configured to collect and obtain the motion information corresponding to the continuous trajectory of the target pedestrian and the position information of the reference points in the continuous trajectory of the target pedestrian. Among them, the motion information includes a plurality of pedestrian step values and a plurality of heading values, and one pedestrian step value corresponds to one heading value. The reference points include at least two points passed by the continuous trajectory of the target pedestrian, and the reference points include the starting point of the continuous trajectory of the target pedestrian. The position information of the reference points includes the coordinate values of the reference points. A trajectory reconstruction module, configured to obtain the reconstructed trajectory corresponding to the continuous trajectory of the target pedestrian through a preset position recursion model according to the motion information and the position information of the reference points, and obtain the reconstructed and optimized trajectory corresponding to the continuous trajectory of the target pedestrian after optimizing the reconstructed trajectory according to the position information of the reference points. A target feature extraction module, configured to obtain the target features corresponding to each of the reconstructed and optimized coordinate points in the reconstructed and optimized trajectory. Among them, the target features corresponding to one reconstructed and optimized coordinate point include the reconstructed and optimized step value, the reconstructed and optimized heading value, the position update time difference value, the current cumulative step number value, and the speed value corresponding to the reconstructed and optimized coordinate point. An error value calculation module, configured to obtain the uncertainty error values corresponding to each of the reconstructed and optimized coordinate points through the trained uncertainty error prediction model according to the target features. An uncertainty region generation module, configured to obtain the uncertainty region corresponding to the reconstructed and optimized trajectory based on the reconstructed and optimized coordinate point and the uncertainty error value and use it as the target uncertainty region corresponding to the continuous trajectory of the target pedestrian. Among them, the target uncertainty region is the region where the real trajectory corresponding to the continuous trajectory of the target pedestrian is located. The uncertainty error prediction model is a hybrid deep learning model, and the uncertainty error prediction model includes a one-dimensional convolutional neural network layer, a long short-term memory neural network layer, and a fully connected layer; during the training of the uncertainty error prediction model, training data of the corresponding uncertainty error prediction model is collected and processed through the position recursion model; The target features corresponding to a reconstructed and optimized coordinate point further include the current cumulative walking distance, the current distance ratio, the current duration ratio, the current step ratio, the current heading change amount, and the current cumulative heading change amount corresponding to the reconstructed and optimized coordinate point; Among them, the current cumulative walking distance is the sum of the reconstructed and optimized step lengths of all current passing points, and the current passing points include all reconstructed and optimized coordinate points from the starting point of the reconstructed and optimized trajectory to this reconstructed and optimized coordinate point; The current distance ratio is the ratio of the current cumulative walking distance to the total distance of the trajectory. The total distance of the trajectory is the sum of the reconstructed and optimized step length values of all trajectory passing points. The trajectory passing points include all reconstructed and optimized coordinate points from the starting point of the reconstructed and optimized trajectory to the ending point of the reconstructed and optimized trajectory; The current duration ratio is the ratio of the current duration to the total duration of the trajectory. The current duration is the cumulative duration corresponding to all the current passing points, and the total duration of the trajectory is the cumulative duration from the starting point of the reconstructed and optimized trajectory to the ending point of the reconstructed and optimized trajectory; The current step ratio is the ratio of the current cumulative step value to the total number of steps of the trajectory. The total number of steps of the trajectory is the cumulative number of steps from the starting point of the reconstructed and optimized trajectory to the ending point of the reconstructed and optimized trajectory; The current heading change amount is the difference between the reconstructed and optimized heading value corresponding to this reconstructed and optimized coordinate point and the reconstructed and optimized heading value corresponding to the previous reconstructed and optimized coordinate point; The current cumulative heading change amount is the sum of the starting heading change amounts corresponding to all the current passing points. The starting heading change amount is the difference between the reconstructed and optimized heading value between the current passing point and the starting point of the reconstructed and optimized trajectory; 8. An intelligent terminal, characterized in that The intelligent terminal includes a memory, a processor, and an uncertainty area acquisition program for pedestrian continuous trajectories stored on the memory and executable on the processor. When the uncertainty area acquisition program for pedestrian continuous trajectories is executed by the processor, the steps of the uncertainty area acquisition method for pedestrian continuous trajectories according to any one of claims 1-6 are implemented.

9. A computer-readable storage medium, characterized in that, An uncertainty area acquisition program for pedestrian continuous trajectories is stored on the computer-readable storage medium. When the uncertainty area acquisition program for pedestrian continuous trajectories is executed by the processor, the steps of the uncertainty area acquisition method for pedestrian continuous trajectories according to any one of claims 1-6 are implemented.

Images