Indoor positioning error calibration method, device, system and storage medium

Abstract

The present invention discloses an indoor positioning error calibration method and device, system, and storage medium, including: step S1, obtaining human behavior information according to IMU data; step S2, collecting initial path information through PDR; step S3, obtaining a reconstructed indoor path according to the association between human behavior data and key nodes of the initial path information; step S4, correcting the heading estimation and step length estimation according to the reconstructed indoor path. The technical solution of the present invention is adopted to improve the accuracy of PDR positioning.

Description

Technical Field

[0001] The present invention belongs to the field of indoor positioning technology, and in particular relates to an indoor positioning error calibration method and device, a system, and a storage medium. Background Art

[0002] PDR (Pedestrian Dead Reckoning) faces severe cumulative errors (CE) that arise from the estimation of three parameters: heading, stride, and step length. Many efforts have been made to reduce this error. One approach is to improve algorithms such as Kalman filtering, particle filtering, and deep learning. These methods correct the heading estimate and step length. However, the measurement errors are persistent, and subsequent corrections are based on biased positions, causing CE to occur continuously.

[0003] Invention content

[0004] The technical problem to be solved by the present invention is to provide a PDR indoor positioning error calibration method and device, system, and storage medium based on behavior matching and a few sample markers.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A method for calibrating indoor positioning errors, comprising:

[0007] Step S1, obtaining human behavior information according to IMU data;

[0008] Step S2: Collect initial path information through PDR;

[0009] Step S3, obtaining a reconstructed indoor path according to the association between the human behavior information behavior data and the key nodes of the initial path information;

[0010] Step S4: Correct the heading estimation and step length estimation according to the reconstructed indoor path.

[0011] Preferably, in step S1, the IMU data is input into the HAR attention network to extract human behavior information.

[0012] Preferably, in step S3, based on the key nodes of the initial path information of the human behavior information, the reconstructed indoor path is obtained by spatial matching of reference point coordinates with behavior semantics.

[0013] The present invention also provides an indoor positioning error calibration device, comprising:

[0014] A first acquisition module is used to obtain human behavior information based on IMU data;

[0015] The second acquisition module is used to collect initial path information through the PDR;

[0016] A reconstruction module, used to obtain a reconstructed indoor path according to the association between the human behavior information behavior data and the key nodes of the initial path information;

[0017] The correction module is used to correct the heading estimation and the step length estimation according to the reconstructed indoor path.

[0018] Preferably, the first acquisition module inputs the IMU data into the HAR attention network to extract human behavior information.

[0019] Preferably, the reconstruction module obtains the reconstructed indoor path through spatial matching of reference point coordinates and behavior semantics based on key nodes of the initial path information of the human behavior information.

[0020] An embodiment of the present invention further provides an indoor positioning error calibration system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an indoor positioning error calibration method when executed by the processor.

[0021] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes an indoor positioning error calibration method when running.

[0022] The present invention adopts a lightweight self-attention model to classify the behavior sequences in the training data and match the behaviors with landmarks to reconstruct indoor paths. By sequentially connecting the time-continuous behaviors with the spatially discrete landmarks, a spatial reconstruction path inside the building is constructed to assist in PDR (Pedestrian Dead Reckoning) positioning, and the heading is corrected according to the similarity between the new predicted path and the existing path. When an activity matches a landmark, the positioning estimate is recalibrated to align it with the identified landmark, thereby correcting the accumulated error. When performing heading estimation, deep learning technology is applied to alleviate the problem of sensor yaw misalignment in IMU data. The technical solution of the present invention is adopted to improve the accuracy of PDR positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0024] Figure 1This is a flow chart of the indoor positioning error calibration method according to an embodiment of the present invention;

[0025] Figure 2 Schematic diagram of HAR self-attention network;

[0026] Figure 3 Schematic diagram of a multilayer perceptron for heading estimation. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] Embodiment 1:

[0030] like Figure 1 As shown, an embodiment of the present invention provides an indoor positioning error calibration method.

[0031] Step S1, obtaining human behavior information according to IMU data;

[0032] Step S2: Collect initial path information through PDR;

[0033] Step S3, obtaining a reconstructed indoor path according to the association between the human behavior information behavior data and the key nodes of the initial path information;

[0034] Step S4: Correct the heading estimation and step length estimation according to the reconstructed indoor path.

[0035] As an implementation method of an embodiment of the present invention, in step S1, IMU data is input into the HAR attention network to extract human behavior information.

[0036] Further, if Figure 2 As shown, the IMU data is taken as input and initially linearly embedded. Using the attention mechanism, the data is re-quantized, which involves calculating the dot product of Q and K and then multiplying it by the matrix V, thereby focusing on areas with greater weight importance, namely:

[0037]

[0038] in, are the three inputs to the self-attention layer: queries, keys, and values, tq, tk, and tv are the number of elements in the different inputs, and dq, dk, and dv represent the corresponding element dimensions. Prevent the Softmax function from entering a region with very small gradients. The output of a query is calculated as a weighted sum of values, where the weight of each value is calculated by the specified function of the query and the corresponding key.

[0039]

[0040] Among them, O represents the output of the 1×1 convolutional layer, represented by W o And the attention layer Attention weighted. The final output of this network is:

[0041]

[0042] Among them, HAR result is the recognition result, Represents a fully connected layer.

[0043] Furthermore, the input data of the HAR attention network is: the data collected by the IMU, the size of each sample is 3*200, 3 represents the accelerometer x, y, z axes, and 200 is the data length.

[0044] Embedding: Perform embedding operations on the input data, mainly using a fully connected method to increase the dimension of the data from 200 to 256 dimensions.

[0045] Attention layer: The input vector will generate query, key, and value through three different linear transformations. Their weight dimensions are all 256*3, and the final dimension is 3*3. Then the attention score is calculated, score = QK T ,Here, is the transpose of the key matrix, and the shape of the calculated score matrix is ​​3*3, which represents the similarity between each input. In order to prevent the score from being too large when calculating the dot product, the score will be scaled: Then, the softmax function is used to calculate the attention weights. The final output matrix shape of the attention layer is 3*256.

[0046] Fully connected layer: The output of the attention layer is expanded into 1 dimension, and compressed to 512 using a fully connected layer, and then another fully connected layer compresses the 512 elements to the output dimension.

[0047] HAR attention network output: The number of output categories is 6, indicating the number of categories.

[0048] The function of the HAR attention network is to perform behavior recognition. After the recognition is completed, the heading is corrected using the matched feature points to reduce the cumulative error.

[0049] As an implementation method of the embodiment of the present invention, in step S3, based on the key nodes of the initial path information of the human behavior information, the reconstructed indoor path is obtained by spatial matching of the reference point coordinates and the behavior semantics.

[0050] Furthermore, according to the timestamp of behavior recognition, the path nodes are matched in the time domain, and the heading estimation of the traditional PDR is transformed into coordinates in different time periods according to the known reference points. The angle of deviation of each behavior landmark from the reference point is defined as:

[0051]

[0052] Among them, the angle θ represents the angle between the original estimated point and the reference point, (x pre ,y pre ) represents the estimated coordinates of PDR, (x ref ,y ref ) represents the coordinates of the reference position.

[0053] The deviation angle can be used to correct the predicted path of the PDR. The formula for correcting each predicted point is as follows:

[0054]

[0055] Among them, x′ pre ,y′ pre Indicates the coordinates after heading correction.

[0056] After the heading transformation of all coordinates at each time interval, the reconstruction of the indoor path is completed.

[0057] As an implementation of an embodiment of the present invention, in step S4, the step length estimation method is the same as the heading estimation method; wherein, according to the reconstructed indoor path, the heading is accurately estimated by MLP (multi-layer perceptron), and the heading estimation is corrected, such as Figure 3 As shown, specifically including:

[0058] Multilayer perceptron network input: accelerometer data, the data shape is 3*50, representing 0.5s data.

[0059] Fully connected layer: There are 4 fully connected layers. The first layer converts from 150 dimensions to 40 dimensions; the second layer converts from 40 dimensions to 120 dimensions; the third layer converts from 120 dimensions to 60 dimensions; and the fourth layer converts from 60 dimensions to 30 dimensions.

[0060] Multilayer perceptron network output: estimated value of heading, dimension is 1.

[0061] Embodiment 2:

[0062] The embodiment of the present invention further provides an indoor positioning error calibration device, comprising:

[0063] A first acquisition module is used to obtain human behavior information based on IMU data;

[0064] The second acquisition module is used to collect initial path information through the PDR;

[0065] A reconstruction module, used to obtain a reconstructed indoor path according to the association between the human behavior information behavior data and the key nodes of the initial path information;

[0066] The correction module is used to correct the heading estimation and the step length estimation according to the reconstructed indoor path.

[0067] As an implementation method of an embodiment of the present invention, the first acquisition module inputs IMU data into the HAR attention network to extract human behavior information.

[0068] As an implementation method of the embodiment of the present invention, the reconstruction module obtains the reconstructed indoor path through spatial matching of reference point coordinates and behavior semantics based on key nodes of the initial path information of the human behavior information.

[0069] Embodiment 3:

[0070] An embodiment of the present invention further provides an indoor positioning error calibration system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes an indoor positioning error calibration method when executed by the processor.

[0071] Embodiment 4:

[0072] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes an indoor positioning error calibration method when running.

[0073] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for indoor positioning error calibration, characterized in that: include: Step S1, obtaining human behavior information according to IMU data; Step S2: Collect initial path information through PDR; Step S3, obtaining a reconstructed indoor path according to the association between the human behavior information behavior data and the key nodes of the initial path information; Step S4, correcting the heading estimation and step length estimation according to the reconstructed indoor path; In step S1, the IMU data is input into the HAR attention network to extract human behavior information; In step S3, based on the key nodes of the initial path information of the human behavior information, the reconstructed indoor path is obtained by spatial matching of the reference point coordinates and the behavior semantics.

2. An indoor positioning error calibration device, characterized in that: include: A first acquisition module is used to obtain human behavior information based on IMU data; The second acquisition module is used to collect initial path information through the PDR; A reconstruction module, used to obtain a reconstructed indoor path according to the association between the human behavior information behavior data and the key nodes of the initial path information; A correction module, used for correcting the heading estimation and the step length estimation according to the reconstructed indoor path; The first acquisition module inputs IMU data into the HAR attention network to extract human behavior information; The reconstruction module obtains the reconstructed indoor path based on the key nodes of the initial path information of the human behavior information through spatial matching of the reference point coordinates and the behavior semantics.

3. An indoor positioning error calibration system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the indoor positioning error calibration method as claimed in claim 1 is executed.

4. A storage medium, characterized in that: The storage medium stores a computer program, and the computer program executes the indoor positioning error calibration method as claimed in claim 1 when running.

Images