An underground space electromagnetic positioning method and device based on trajectory time series analysis

Abstract

The present invention provides an underground space electromagnetic positioning method and device based on trajectory time series analysis, which relates to the field of positioning technology. The method includes: obtaining the electromagnetic field phase difference and magnetic field signal strength of a target to be positioned, and constructing time series input data according to the electromagnetic field phase difference and magnetic field signal strength; inputting the time series input data into the constructed time series electromagnetic positioning model; wherein, the time series electromagnetic positioning model includes a screening gate, a sensing gate and a repositioning gate; and according to the time series input data and the time series electromagnetic positioning model, the target pose information is output in real time. The present invention realizes real-time continuous positioning of a high-precision moving target by autonomously sensing and capturing target positioning information through a time series electromagnetic positioning network model.

Description

Technical Field

[0001] The present invention relates to the technical field of positioning, and in particular to an underground space electromagnetic positioning method and device based on trajectory time series analysis. Background Art

[0002] With the development of information technology, people's demand for positioning in complex indoor and underground non-exposed spaces is increasing day by day. In environments such as underground parking lots, underground pipe galleries, subways, and mines, timely obtaining the position information of personnel and mobile devices is crucial for ensuring normal living and production order. Traditional indoor wireless positioning methods, such as RFID (Radio Frequency Identification), ZigBee, WIFI, or UWB (Ultra Wide Band) and other technologies, have reflection, diffraction, and scattering phenomena when the high-frequency wireless signals they use propagate indoors, resulting in multipath effects, affecting the accuracy of indoor positioning, and being unable to penetrate buildings for propagation. The near-field electromagnetic positioning system uses NFER (Near Field Electromagnetic Ranging) technology to obtain the distance between the target and the positioning base station, and then uses a multi-base station positioning algorithm to perform real-time positioning on the target. The near-field electromagnetic positioning system uses low-frequency electromagnetic signals, which have a small path attenuation when penetrating the medium and are not affected by multipath effects, and are suitable for underground space positioning.

[0003] However, the transmitting antennas of near-field electromagnetic positioning systems generally use orthogonal magnetic coil antennas, which have strong directivity. When using a near-field electromagnetic positioning system to position a moving target, the movement trajectory and attitude of the target will affect the convergence performance of the near-field ranging algorithm, resulting in a decrease in positioning accuracy and timeliness. Summary of the Invention

[0004] To solve the technical problems of the existing technology such as the decrease in positioning accuracy and timeliness, the embodiments of the present invention provide an underground space electromagnetic positioning method and device based on trajectory time series analysis. The technical solutions are as follows:

[0005] On the one hand, an underground space electromagnetic positioning method based on trajectory time series analysis is provided. This method is implemented by an underground space electromagnetic positioning device, and the method includes:

[0006] S1. Obtain the electromagnetic field phase difference and magnetic field signal intensity of the target to be positioned, and construct time series input data according to the electromagnetic field phase difference and magnetic field signal intensity.

[0007] S2. Input the time series input data into the constructed time series electromagnetic positioning model; wherein, the time series electromagnetic positioning model includes a screening gate, a perception gate, and a repositioning gate.

[0008] S3. Output the target pose information in real time according to the timing input data and the timing electromagnetic positioning model.

[0009] Optionally, the construction process of the timing electromagnetic positioning model in S2 includes:

[0010] S21. Construct an electromagnetic positioning data set; wherein, the electromagnetic positioning data set includes electromagnetic field phase difference samples, magnetic field signal strength samples, and corresponding target pose information samples.

[0011] S22. Construct a screening gate.

[0012] S23. Construct a perception gate.

[0013] S24. Construct a relocalization gate.

[0014] S25. Construct a timing electromagnetic positioning network according to the screening gate, the perception gate, and the relocalization gate.

[0015] S26. Train the timing electromagnetic positioning network according to the electromagnetic positioning data set to generate a timing electromagnetic positioning model.

[0016] Optionally, the construction of the screening gate in S22 includes:

[0017] S221. For the current moment t, obtain the target pose information sample at the previous moment t - 1, the electromagnetic field phase difference sample at the current moment t, and the magnetic field signal strength sample at the current moment t.

[0018] S222. Input the target pose information sample at the previous moment t - 1 into the first self - dimensionality - increasing layer to obtain a feature map MO1.

[0019] S223. Input the feature map MO1 into a convolutional module to obtain a feature map MO2.

[0020] S224. Input the electromagnetic field phase difference sample at the current moment t and the magnetic field signal strength sample at the current moment t into the second self - dimensionality - increasing layer to obtain a feature map ME1.

[0021] S225. Input the feature map ME1 into a global pooling layer and output a feature vector ME2.

[0022] S226. Input the feature vector ME2 into an SE layer to obtain a feature vector ME3.

[0023] S227. Multiply the feature map MO2 by the feature vector ME3 to obtain a screening gate output feature map ML0.

[0024] Optionally, the construction of the perception gate in S23 includes:

[0025] S231. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO3, and perform a 3×3 convolution operation on the deep feature matrix MO3 to obtain a deep feature matrix MO4.

[0026] S232. Add the feature map MO1 and the deep feature matrix MO4 element - by - element to obtain the first sum matrix of the perception gate.

[0027] S233. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME4, and perform a 3×3 convolution operation on the deep feature matrix ME4 to obtain a deep feature matrix ME5.

[0028] S234. Add the feature map ME1 and the deep feature matrix ME5 element - by - element to obtain the second sum matrix of the perception gate.

[0029] S235. Multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate matrix - by - matrix, and perform global element activation through the Sigmoid layer to obtain the voting matrix MO5 of the perception gate.

[0030] S236. Multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate matrix - by - matrix, and perform activation through the Tanh layer to obtain the perception matrix MO8.

[0031] S237. Multiply the voting matrix MO5 of the perception gate and the perception matrix MO8 matrix - by - matrix to obtain the output feature map ML1 of the perception gate.

[0032] Optionally, constructing the relocation gate in S24 includes:

[0033] S241. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO9, and perform a 3×3 convolution operation on the deep feature matrix MO9 to obtain a deep feature matrix MO 10 .

[0034] S242. Add the feature map MO1 and the deep feature matrix MO 10 element - by - element to obtain the first sum matrix of the relocation gate.

[0035] S243. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME8, and perform a 3×3 convolution operation on the deep feature matrix ME8 to obtain a deep feature matrix ME9.

[0036] S244. Add the feature map ME1 and the deep feature matrix ME9 element - by - element to obtain the second sum matrix of the relocation gate.

[0037] S245, perform matrix multiplication on the first sum matrix of the repositioning gate and the second sum matrix of the repositioning gate, and perform global element activation through the Sigmoid layer to obtain the repositioning gate voting matrix MO 11 .

[0038] S246, obtain electromagnetic positioning memory unit L t , for the electromagnetic positioning memory unit L t Perform two 3×3 convolution operations in sequence to obtain the convolution result L3, and combine the convolution result L3 with the electromagnetic positioning memory unit L t The elements are summed and activated through the Tanh layer to obtain the relocation matrix L4.

[0039] S247, relocate the gate voting matrix MO 11 Perform matrix multiplication with the relocation matrix L4 to obtain the relocation gate output feature map L5.

[0040] S248. Input the relocation gate output feature map L5 into the self-dimensionality reduction layer to obtain the relocation gate output feature Ot.

[0041] Optionally, in S25, a sequential electromagnetic positioning network is constructed according to the screening gate, the perception gate, and the relocation gate, including:

[0042] S251. For the current moment t, the target posture information sample at the previous moment t-1, the electromagnetic field phase difference sample at the current moment t, and the magnetic field signal strength sample at the current moment t are input into the screening gate to obtain the screening gate output feature map ML0.

[0043] S252 , subtract the screening gate output characteristic map ML0 from the electromagnetic positioning memory unit Lt-1 at the previous time t-1 to obtain the electromagnetic positioning memory unit MLt-1.

[0044] S253. Input the feature map MO1 and the feature map ME1 in the screening gate into the perception gate to obtain the perception gate output feature map ML1.

[0045] S254. Add the electromagnetic positioning memory unit MLt-1 and the perception gate output feature map ML1 to obtain the electromagnetic positioning memory unit Lt.

[0046] S255. Input the feature map MO1, feature map ME1 and electromagnetic positioning memory unit Lt into the repositioning gate to obtain the repositioning gate output feature Ot, which is the target posture information at the current moment.

[0047] On the other hand, an underground space electromagnetic positioning device based on trajectory time series analysis is provided, which is applied to the underground space electromagnetic positioning method based on trajectory time series analysis, and the device includes:

[0048] An acquisition module, configured to acquire the electromagnetic field phase difference and the magnetic field signal strength of the target to be located, and construct time-series input data according to the electromagnetic field phase difference and the magnetic field signal strength.

[0049] An input module, configured to input the time-series input data into the constructed time-series electromagnetic positioning model; wherein, the time-series electromagnetic positioning model includes a screening gate, a perception gate and a repositioning gate.

[0050] An output module, configured to output the target pose information in real time according to the time-series input data and the time-series electromagnetic positioning model.

[0051] Optionally, the input module is further configured to:

[0052] S21. Construct an electromagnetic positioning data set; wherein, the electromagnetic positioning data set includes electromagnetic field phase difference samples, magnetic field signal strength samples and corresponding target pose information samples.

[0053] S22. Construct a screening gate.

[0054] S23. Construct a perception gate.

[0055] S24. Construct a repositioning gate.

[0056] S25. Construct a time-series electromagnetic positioning network according to the screening gate, the perception gate and the repositioning gate.

[0057] S26. Train the time-series electromagnetic positioning network according to the electromagnetic positioning data set to generate a time-series electromagnetic positioning model.

[0058] Optionally, the input module is further configured to:

[0059] S221. For the current moment t, acquire the target pose information sample at the previous moment t-1, the electromagnetic field phase difference sample at the current moment t, and the magnetic field signal strength sample at the current moment t.

[0060] S222. Input the target pose information sample at the previous moment t-1 into the first self-dimension-raising layer to obtain a feature map MO1.

[0061] S223. Input the feature map MO1 into the convolutional module to obtain a feature map MO2.

[0062] S224. Input the electromagnetic field phase difference sample at the current moment t and the magnetic field signal strength sample at the current moment t into the second self-dimension-raising layer to obtain a feature map ME1.

[0063] S225. Input the feature map ME1 into the global pooling layer and output a feature vector ME2.

[0064] S226. Input the feature vector ME2 into the SE layer to obtain a feature vector ME3.

[0065] S227. Multiply the feature map MO2 by the feature vector ME3 to obtain the filtered gate output feature map ML0.

[0066] Optionally, the input module is further configured to:

[0067] S231. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO3, and perform a 3×3 convolution operation on the deep feature matrix MO3 to obtain a deep feature matrix MO4.

[0068] S232. Element-wise add the feature map MO1 and the deep feature matrix MO4 to obtain the first sum matrix of the perception gate.

[0069] S233. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME4, and perform a 3×3 convolution operation on the deep feature matrix ME4 to obtain a deep feature matrix ME5.

[0070] S234. Element-wise add the feature map ME1 and the deep feature matrix ME5 to obtain the second sum matrix of the perception gate.

[0071] S235. Matrix multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate, and perform global element activation through a Sigmoid layer to obtain the voting matrix MO5 of the perception gate.

[0072] S236. Matrix multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate, and perform activation through a Tanh layer to obtain the perception matrix MO8.

[0073] S237. Matrix multiply the voting matrix MO5 of the perception gate and the perception matrix MO8 to obtain the perception gate output feature map ML1.

[0074] Optionally, the input module is further configured to:

[0075] S241. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO9, and perform a 3×3 convolution operation on the deep feature matrix MO9 to obtain a deep feature matrix MO 10 .

[0076] S242. Element-wise add the feature map MO1 and the deep feature matrix MO 10 to obtain the first sum matrix of the relocation gate.

[0077] S243. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME8, and perform a 3×3 convolution operation on the deep feature matrix ME8 to obtain a deep feature matrix ME9.

[0078] S244. Add the elements of the feature map ME1 and the deep feature matrix ME9 to obtain the second sum matrix of the relocation gate.

[0079] S245. Multiply the first sum matrix of the relocation gate and the second sum matrix of the relocation gate, and perform global element activation through the Sigmoid layer to obtain the voting matrix MO of the relocation gate. 11 .

[0080] S246. Obtain the electromagnetic positioning memory unit L t , and perform two 3×3 convolution operations on the electromagnetic positioning memory unit L t in sequence to obtain the convolution result L3. Add the convolution result L3 and the electromagnetic positioning memory unit L t element-wise, and perform activation through the Tanh layer to obtain the relocation matrix L4.

[0081] S247. Multiply the voting matrix MO of the relocation gate 11 and the relocation matrix L4 to obtain the output feature map L5 of the relocation gate.

[0082] S248. Input the output feature map L5 of the relocation gate into the self-dimension reduction layer to obtain the output feature Ot of the relocation gate.

[0083] Optionally, the input module is further used for:

[0084] S251. For the current time t, input the target pose information sample at the previous time t - 1, the electromagnetic field phase difference sample at the current time t, and the magnetic field signal intensity sample at the current time t into the screening gate to obtain the output feature map ML0 of the screening gate.

[0085] S252. Subtract the output feature map ML0 of the screening gate from the electromagnetic positioning memory unit Lt - 1 at the previous time t - 1 to obtain the electromagnetic positioning memory unit MLt - 1.

[0086] S253. Input the feature map MO1 and the feature map ME1 in the screening gate into the perception gate to obtain the output feature map ML1 of the perception gate.

[0087] S254. Add the electromagnetic positioning memory unit MLt - 1 and the output feature map ML1 of the perception gate to obtain the electromagnetic positioning memory unit Lt.

[0088] S255. Input the feature map MO1, the feature map ME1, and the electromagnetic positioning memory unit Lt into the relocation gate together to obtain the output feature Ot of the relocation gate, that is, the target pose information at the current time.

[0089] On the other hand, a subterranean space electromagnetic positioning device is provided. The subterranean space electromagnetic positioning device includes: a processor; a memory storing computer-readable instructions thereon, and when the computer-readable instructions are executed by the processor, any one of the methods in the above-mentioned subterranean space electromagnetic positioning method based on trajectory time series analysis is implemented.

[0090] On the other hand, a computer-readable storage medium is provided, and at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement any one of the methods in the above-mentioned subterranean space electromagnetic positioning method based on trajectory time series analysis.

[0091] The beneficial effects brought by the technical solutions provided in the embodiments of the present invention at least include:

[0092] In the present invention, the subterranean space electromagnetic positioning method based on trajectory time series analysis outputs target positioning information in real time through a time series electromagnetic positioning network. Compared with the traditional method where the movement trajectory and attitude of the target will affect the positioning accuracy and timeliness of the near-field ranging algorithm, it has strong time series signal processing capabilities, and can learn more essential electromagnetic positioning characterization mechanisms as data and time accumulate, so as to give more accurate real-time target positioning information.

[0093] The time series electromagnetic positioning network is designed for a specific task, that is, subterranean space electromagnetic positioning. It includes three structures: a screening gate, a sensing gate, and a repositioning gate. The designs of the three gate structures are respectively used to filter redundant historical positioning information, sense and store the current electromagnetic positioning representation, and reposition the current pose of the target. The three gate structures are connected in series in sequence, which can enhance the anti-interference ability of the electromagnetic positioning method, and through in-depth mining and interaction of long-term positioning historical experience and current signals, achieve accurate target positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] 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 the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0095] Figure 1 is a flowchart of a subterranean space electromagnetic positioning method based on trajectory time series analysis provided by an embodiment of the present invention;

[0096] Figure 2 is a schematic structural diagram of a subterranean space electromagnetic positioning based on trajectory time series analysis provided by an embodiment of the present invention;

[0097] Figure 3 is a flowchart for constructing a time series electromagnetic positioning model provided by an embodiment of the present invention;

[0098] Figure 4 is the screening gate structure diagram provided by the embodiments of the present invention;

[0099] Figure 5 is the sensing gate structure diagram provided by the embodiments of the present invention;

[0100] Figure 6 is the relocation gate structure diagram provided by the embodiments of the present invention;

[0101] Figure 7 is the timing electromagnetic positioning network structure diagram provided by the embodiments of the present invention;

[0102] Figure 8 is the block diagram of an underground space electromagnetic positioning device based on trajectory timing analysis provided by the embodiments of the present invention;

[0103] Figure 9 is the structural schematic diagram of an underground space electromagnetic positioning device provided by the embodiments of the present invention. Specific embodiments

[0104] The following describes the technical solutions in the present invention with reference to the accompanying drawings.

[0105] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of the word "example" is intended to present concepts in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0106] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when not emphasizing their differences, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when not emphasizing their differences, the meanings they express are the same.

[0107] In the embodiments of the present invention, sometimes subscripts such as W1 may be written in a non-subscript form such as W1. When not emphasizing their differences, the meanings they express are the same.

[0108] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0109] An embodiment of the present invention provides an underground space electromagnetic positioning method based on trajectory time series analysis. This method can be implemented by an underground space electromagnetic positioning device, which can be a terminal or a server. As Figure 1 shown in the flowchart of the underground space electromagnetic positioning method based on trajectory time series analysis, the processing flow of this method can include the following steps:

[0110] S1. Obtain the electromagnetic field phase difference and magnetic field signal strength of the target to be positioned, and construct time series input data according to the electromagnetic field phase difference and magnetic field signal strength.

[0111] S2. Input the time series input data into the constructed time series electromagnetic positioning model; wherein, the time series electromagnetic positioning model includes a screening gate, a sensing gate, and a repositioning gate.

[0112] S3. According to the time series input data and the time series electromagnetic positioning model, output the target pose information in real time.

[0113] In a feasible implementation manner, the present invention constructs time series input data by collecting the phase difference between the field signal and the magnetic field signal within the sampling period and the magnetic field signal strength, and then outputs the target positioning information in real time through the time series electromagnetic positioning network.

[0114] Furthermore, the time series electromagnetic positioning network is designed for underground space electromagnetic positioning and includes three structures: a screening gate, a sensing gate, and a repositioning gate. The designs of the three gate structures are respectively used to filter redundant historical positioning information, sense and store the current electromagnetic positioning representation, and reposition the current pose of the target. The three gate structures are connected in series in sequence, which can enhance the anti-interference ability of the electromagnetic positioning method, and realize accurate target positioning through in-depth mining and interaction of long-term positioning historical experience and current signals.

[0115] Specifically, as Figure 2 shown, the present invention can emit the electromagnetic signal required for the positioning target through the signal transmitting part; the signal receiving part receives the electric field part and the magnetic field part of the positioning signal, and the magnetic field signal is received by an array antenna to receive the signal strength of the magnetic field signal in multiple directions; the preprocessing part processes the received electric field signal and magnetic field signal to obtain the phase difference between the electric field signal and the magnetic field signal and the field strength of the magnetic field signal; the positioning model part designs a time series electromagnetic positioning network and trains it based on a large amount of electromagnetic field phase difference and field strength data of different trajectories to obtain a positioning model. In actual use, the data to be processed is input into the model; the positioning result part outputs the position of the target.

[0116] Furthermore, in the present invention, the signal transmitting antenna uses an orthogonal magnetic antenna to transmit the target electromagnetic signal.

[0117] The electric field signal received by the electric field antenna at the receiving end is E, and the magnetic field signal received by the magnetic field array antenna is , where is the number of magnetic field antennas.

[0118] In the preprocessing stage, LAMBDA (Least square AMBiguity Decorrelation Adijustment) is used to obtain the phase difference between the electric field signal and the magnetic field signal . The sample mean square value is used as the magnetic field signal strength .

[0119] Optionally, as Figure 3 shown, in the positioning model stage, the time-series electromagnetic positioning model can be obtained through the following steps S21 - S26:

[0120] S21. Construct an electromagnetic positioning data set; among them, the electromagnetic positioning data set includes electromagnetic field phase difference samples, magnetic field signal strength samples, and corresponding target pose information samples.

[0121] In a feasible implementation, an underground target positioning test scenario is arranged, and a large number of electromagnetic field phase differences, field strength data, and corresponding target pose information under different trajectories are collected and made, and the collection period is T.

[0122] Furthermore, the phase difference ( ) and the field strength data ( ) are both time-series signals with dimension T n. Let the electromagnetic field phase difference and field strength data corresponding to time t be Et, then Et is composed of two pieces of data with dimension 1 n.

[0123] The target pose information is a time-series signal with dimension T 4 , where is the coordinate of the target in the space coordinate system, is the angle between the target orientation and the due north. Let the target pose data corresponding to time t be Ot, then the dimension of Ot is 1 4.

[0124] S22. Construct a screening gate.

[0125] Optionally, the above step S22 may include the following steps S221 - S227:

[0126] S221. For the current time t, obtain the target pose information sample at the previous time t - 1, the electromagnetic field phase difference sample at the current time t, and the magnetic field signal strength sample at the current time t.

[0127] S222. Input the target pose information sample at the previous moment t-1 into the first self-ascending dimension layer to obtain the feature map MO1.

[0128] S223. Input the feature map MO1 into the convolutional module to obtain the feature map MO2.

[0129] S224. Input the electromagnetic field phase difference sample at the current moment t and the magnetic field signal intensity sample at the current moment t into the second self-ascending dimension layer to obtain the feature map ME1.

[0130] S225. Input the feature map ME1 into the global pooling layer and output the feature vector ME2.

[0131] S226. Input the feature vector ME2 into the SE layer to obtain the feature vector ME3.

[0132] S227. Multiply the feature map MO2 by the feature vector ME3 to obtain the screening gate output feature map ML0.

[0133] In a feasible implementation manner, as Figure 4 shown, the inputs of the screening gate are the target pose information Ot-1 at the previous moment and the electromagnetic field phase difference and field strength data Et at the current moment. First, input Ot-1 into the self-ascending dimension layer. In the self-ascending dimension layer, Ot-1 first multiplies with its own transpose to obtain a 4 ×4 matrix; then perform three consecutive deconvolution operations with a stride of 2 and a number of channels of 32 on it to obtain a feature map MO1 with a size of 32 ×32 ×32. This step enriches the spatio-temporal feature representation of the target pose information from different dimensions and provides a diverse data basis for subsequent feature mining. Then input the feature map MO1 into the convolutional module, and perform feature mining through three consecutive convolutional operations. In this step, each convolutional operation consists of a 3 ×3 convolutional layer, a Relu activation layer, and a BN layer. After feature mining, output a feature map MO2 with a size of 32 ×32 ×32.

[0134] Subsequently, input Et into its self-ascending dimension layer. In the self-ascending dimension layer, the phase difference data with dimension 1 ×n and the field strength data with dimension 1 ×n respectively multiply with the transpose of each other to obtain two n ×n matrices, and perform channel splicing to obtain an n ×n ×2 data matrix; then perform three consecutive deconvolution operations with corresponding strides and a number of channels of 32 on this data matrix to obtain a size of 32 ×32 The feature map ME1 of 32. This step enriches the spatio-temporal feature representation of the target electromagnetic information from different dimensions, providing a diverse data basis for subsequent feature mining. Then, the feature map ME1 is input into the global pooling layer for spatial dimension average pooling to obtain a size of 1 1 The feature vector ME2 of 32, and perform SE attention activation on it to obtain an equal-sized feature vector ME3.

[0135] Multiply the feature map MO2 by the feature vector ME3 to obtain an output feature map ML0 with a final size of 32 32 32.

[0136] The above is the internal structure of the screening gate. The screening gate extracts the previous moment's positioning information and the current electromagnetic information to obtain redundant features irrelevant to the current moment's positioning task, so as to facilitate filtering them in the later stage.

[0137] S23. Construct a perception gate.

[0138] Optionally, the above step S23 may include the following steps S231 - S237:

[0139] S231. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO3, and perform a 3×3 convolution operation on the deep feature matrix MO3 to obtain a deep feature matrix MO4.

[0140] S232. Add the feature map MO1 and the deep feature matrix MO4 element by element to obtain the first sum matrix of the perception gate.

[0141] S233. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME4, and perform a 3×3 convolution operation on the deep feature matrix ME4 to obtain a deep feature matrix ME5.

[0142] S234. Add the feature map ME1 and the deep feature matrix ME5 element by element to obtain the second sum matrix of the perception gate.

[0143] S235. Multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate, and perform global element activation through the Sigmoid layer to obtain the voting matrix MO5 of the perception gate.

[0144] S236. Multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate, and perform activation through the Tanh layer to obtain the perception matrix MO8.

[0145] S237. Multiply the voting matrix MO5 of the perception gate and the perception matrix MO8 to obtain the output feature map ML1 of the perception gate.

[0146] In a feasible implementation, as Figure 5 shown, the input of the perception gate is the output feature maps MO1 and ME1 of the self-dimension-increasing layer in the screening gate, which respectively represent the previous moment's positioning information and the current electromagnetic positioning feature. First, a voting matrix is constructed. The output feature map MO1 is input into a convolutional layer with a convolution kernel of 1 ×1 and 3 ×3, and the deep feature matrices MO3 and MO4 are obtained in sequence. The elements of MO1 itself and its deep feature MO4 are added together to obtain a sum matrix, which retains the original electromagnetic positioning information while obtaining rich positioning voting information. The same operation as Figure 5 shown is performed on ME1 to obtain a sum matrix of ME1 itself and its deep feature ME5. Then, the two sum matrices are multiplied matrix-wise, and global element activation is performed through a Sigmoid layer to obtain the voting matrix MO5, that is, all output elements are between 0 and 1, which is used to vote on the importance of the information of this part.

[0147] Subsequently, a perception matrix is constructed. The construction path of the perception matrix is the same as that of the voting matrix. The difference is that after the two sum matrices are multiplied matrix-wise, activation is performed through a Tanh layer to obtain the perception matrix MO8, and all its output elements are between ±1. While effectively capturing the non-linear relationship of the positioning signal, it realizes feature scaling and accelerates the convergence of subsequent network training.

[0148] Finally, the voting matrix MO5 and the perception matrix MO8 are multiplied matrix-wise to obtain the output feature map ML1 with a final size of 32 ×32 ×32, which represents the refresh degree of the current moment's positioning information.

[0149] The above is the internal structure of the perception gate. The perception gate extracts the previous moment's positioning information to obtain useful historical positioning data, and captures the key information guiding positioning update by extracting the current electromagnetic signal features.

[0150] S24. Construct a repositioning gate.

[0151] Optionally, the above step S24 may include the following steps S241 - S248:

[0152] S241. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO9, and perform a 3×3 convolution operation on the deep feature matrix MO9 to obtain a deep feature matrix MO 10 .

[0153] S242. Add the elements of the feature map MO1 and the deep feature matrix MO 10 to obtain the first sum matrix of the repositioning gate.

[0154] S243. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME8, and perform a 3×3 convolution operation on the deep feature matrix ME8 to obtain a deep feature matrix ME9.

[0155] S244. Add the feature map ME1 and the deep feature matrix ME9 element - by - element to obtain the relocation gate second sum matrix.

[0156] S245. Multiply the relocation gate first sum matrix and the relocation gate second sum matrix, and perform global element activation through the Sigmoid layer to obtain the relocation gate voting matrix MO 11 。

[0157] S246. Obtain the electromagnetic positioning memory unit L t and perform two consecutive 3×3 convolution operations on the electromagnetic positioning memory unit L t to obtain a convolution result L3. Add the convolution result L3 and the electromagnetic positioning memory unit L t element - by - element, and perform activation through the Tanh layer to obtain the relocation matrix L4.

[0158] S247. Multiply the relocation gate voting matrix MO 11 and the relocation matrix L4 to obtain the relocation gate output feature map L5.

[0159] S248. Input the relocation gate output feature map L5 into the self - dimensionality reduction layer to obtain the relocation gate output feature Ot.

[0160] In a feasible implementation, as Figure 6 shown, the input of the relocation gate is the output feature maps MO1 and ME1 of the self - dimensionality increase layer in the screening gate, representing the previous moment's positioning information and the current electromagnetic positioning feature respectively. First, construct the voting matrix, and the construction process is the same as that of the perception gate voting matrix. After calculation, obtain the voting matrix MO 11 for voting on the importance of each element information in the feature map with a dimension of 32 32 ×32.

[0161] Subsequently, construct the relocation matrix. The input feature is L t . L t is the electromagnetic positioning memory unit. Perform two consecutive convolution operations with a convolution kernel of 3 t ×3 on L successively, add the convolution result L3 and the input feature L t element - by - element, and then input the sum result into the Tanh layer for activation to obtain the relocation matrix L4.

[0162] Finally, multiply the voting matrix MO 11Perform matrix multiplication with the relocation matrix L4 to obtain a final size of 32 32 The output feature map L5 of size 32, which represents the positioning information at the current moment. Input L5 into the self-dimension reduction layer. In the self-dimension reduction layer, first perform global average pooling to reduce the dimension of the output feature map of size 32 32 from 32 to 1 1 to obtain a one-dimensional vector of size 32. Then output this vector through a fully connected layer to obtain an output feature of size 1 4. This output feature is Ot, which is the pose information of the target at the current moment.

[0163] The above is the internal structure of the relocation gate. The relocation gate can extract and learn the rich electromagnetic positioning and target state representations in the electromagnetic positioning memory unit, and activate the key information therein through the voting matrix, thereby generating accurate current target pose information.

[0164] S25. Construct a temporal electromagnetic positioning network according to the screening gate, the sensing gate, and the relocation gate.

[0165] Optionally, the above step S25 may include the following steps S251 - S255:

[0166] S251. For the current moment t, input the target pose information sample at the previous moment t - 1, the electromagnetic field phase difference sample at the current moment t, and the magnetic field signal intensity sample at the current moment t into the screening gate to obtain the screening gate output feature map ML0.

[0167] S252. Subtract the screening gate output feature map ML0 from the electromagnetic positioning memory unit Lt - 1 at the previous moment t - 1 to obtain the electromagnetic positioning memory unit MLt - 1.

[0168] S253. Input the feature map MO1 and the feature map ME1 in the screening gate into the sensing gate to obtain the sensing gate output feature map ML1.

[0169] S254. Add the electromagnetic positioning memory unit MLt - 1 and the sensing gate output feature map ML1 to obtain the electromagnetic positioning memory unit Lt.

[0170] S255. Input the feature map MO1, the feature map ME1, and the electromagnetic positioning memory unit Lt into the relocation gate together to obtain the relocation gate output feature Ot, which is the target pose information at the current moment.

[0171] In a feasible implementation, the structure of the temporal electromagnetic positioning network is as Figure 7As shown in the figure. First, the target pose information \(O_{t - 1}\) at the previous moment and the electromagnetic field phase difference and field strength data \(E_{t}\) at the current moment are input into the screening gate, and the feature map \(M_{L0}\) is output. \(M_{L0}\) contains redundant features that are irrelevant to the current moment's positioning task. Subtract \(M_{L0}\) from the electromagnetic positioning memory unit \(L_{t - 1}\) at the previous moment to obtain the memory unit \(M_{Lt - 1}\). This step is used to delete the redundant features in the memory unit \(L_{t - 1}\) that are irrelevant to the current moment's positioning task.

[0172] Then, the feature maps \(M_{O1}\) and \(M_{E1}\) in the screening gate are input into the perception gate, and the feature map \(M_{L1}\) is output. Add the memory unit \(M_{Lt - 1}\) and \(M_{L1}\) to obtain \(L_{t}\), so that the memory unit obtains valuable electromagnetic positioning information at the current moment.

[0173] Input the feature maps \(M_{O1}\), \(M_{E1}\), and \(L_{t}\) into the repositioning gate together, and output \(O_{t}\), that is, the pose information of the target at the current moment.

[0174] The above is the structure of the sequential electromagnetic positioning network. Continuously input the electromagnetic signals at the current moment into this network, and the pose information of the target at the corresponding moment can be output.

[0175] S26. Train the sequential electromagnetic positioning network according to the electromagnetic positioning data set to generate a sequential electromagnetic positioning model.

[0176] In a feasible implementation, the sequential electromagnetic positioning model is generated as follows: Use the constructed electromagnetic positioning data set to train the above sequential electromagnetic positioning network until the algorithm converges to obtain the final sequential electromagnetic positioning model.

[0177] The above is the generation process of the sequential electromagnetic positioning model.

[0178] After obtaining the sequential electromagnetic positioning model, in actual applications, input the collected electromagnetic field phase difference and magnetic field signal strength into the sequential electromagnetic positioning model to obtain the position and orientation of the target , where is the coordinate of the target in the space coordinate system, is the angle between the target's orientation and the due north.

[0179] In the embodiment of the present invention, the underground space electromagnetic positioning method based on trajectory sequential analysis outputs the target positioning information in real time through the sequential electromagnetic positioning network. Compared with the traditional method where the target's motion trajectory and pose will affect the positioning accuracy and timeliness of the near-field ranging algorithm, it has strong sequential signal processing capabilities, and can learn more essential electromagnetic positioning characterization mechanisms as data and time accumulate, so as to give more accurate real-time target positioning information.

[0180] The timing electromagnetic positioning network is designed for a specific task, namely electromagnetic positioning in underground space. It includes three structures: a screening gate, a sensing gate, and a repositioning gate. The designs of the three gate structures are respectively used to filter redundant historical positioning information, sense and store the current electromagnetic positioning representation, and reposition the current pose of the target. The three gate structures are connected in series in sequence, which can enhance the anti-interference ability of the electromagnetic positioning method, and achieve accurate target positioning through in-depth mining and interaction of long-term positioning historical experience and current signals.

[0181] Figure 7 It is a block diagram of an underground space electromagnetic positioning device shown according to an exemplary embodiment. This device is used for the underground space electromagnetic positioning method based on trajectory timing analysis. Refer to Figure 7 This device includes an acquisition module 310, an input module 320, and an output module 330. Among them:

[0182] The acquisition module 310 is used to acquire the electromagnetic field phase difference and magnetic field signal strength of the target to be positioned, and construct timing input data according to the electromagnetic field phase difference and magnetic field signal strength.

[0183] The input module 320 is used to input the timing input data into the constructed timing electromagnetic positioning model; among them, the timing electromagnetic positioning model includes a screening gate, a sensing gate, and a repositioning gate.

[0184] The output module 330 is used to output the target pose information in real time according to the timing input data and the timing electromagnetic positioning model.

[0185] Optionally, the input module 320 is further used for:

[0186] S21. Construct an electromagnetic positioning data set; among them, the electromagnetic positioning data set includes electromagnetic field phase difference samples, magnetic field signal strength samples, and corresponding target pose information samples.

[0187] S22. Construct a screening gate.

[0188] S23. Construct a sensing gate.

[0189] S24. Construct a repositioning gate.

[0190] S25. Construct a timing electromagnetic positioning network according to the screening gate, the sensing gate, and the repositioning gate.

[0191] S26. Train the timing electromagnetic positioning network according to the electromagnetic positioning data set to generate a timing electromagnetic positioning model.

[0192] Optionally, the input module 320 is further used for:

[0193] S221. At the current moment t, obtain the target pose information sample at the previous moment t - 1, the electromagnetic field phase difference sample at the current moment t, and the magnetic field signal intensity sample at the current moment t.

[0194] S222. Input the target pose information sample at the previous moment t - 1 into the first self - dimensionality - increasing layer to obtain the feature map MO1.

[0195] S223. Input the feature map MO1 into the convolutional module to obtain the feature map MO2.

[0196] S224. Input the electromagnetic field phase difference sample at the current moment t and the magnetic field signal intensity sample at the current moment t into the second self - dimensionality - increasing layer to obtain the feature map ME1.

[0197] S225. Input the feature map ME1 into the global pooling layer and output the feature vector ME2.

[0198] S226. Input the feature vector ME2 into the SE layer to obtain the feature vector ME3.

[0199] S227. Multiply the feature map MO2 by the feature vector ME3 to obtain the screening gate output feature map ML0.

[0200] Optionally, the input module 320 is further used for:

[0201] S231. Perform a 1×1 convolution operation on the feature map MO1 to obtain the deep feature matrix MO3, and perform a 3×3 convolution operation on the deep feature matrix MO3 to obtain the deep feature matrix MO4.

[0202] S232. Element - wise add the feature map MO1 and the deep feature matrix MO4 to obtain the perception gate first sum matrix.

[0203] S233. Perform a 1×1 convolution operation on the feature map ME1 to obtain the deep feature matrix ME4, and perform a 3×3 convolution operation on the deep feature matrix ME4 to obtain the deep feature matrix ME5.

[0204] S234. Element - wise add the feature map ME1 and the deep feature matrix ME5 to obtain the perception gate second sum matrix.

[0205] S235. Matrix - multiply the perception gate first sum matrix and the perception gate second sum matrix, and perform global element activation through the Sigmoid layer to obtain the perception gate voting matrix MO5.

[0206] S236. Matrix - multiply the perception gate first sum matrix and the perception gate second sum matrix, and perform activation through the Tanh layer to obtain the perception matrix MO8.

[0207] S237. Multiply the sensing gate voting matrix MO5 with the sensing matrix MO8 to obtain the sensing gate output feature map ML1.

[0208] Optionally, the input module 320 is further configured to:

[0209] S241. Perform a 1×1 convolution operation on the feature map MO1 to obtain the deep feature matrix MO9, and perform a 3×3 convolution operation on the deep feature matrix MO9 to obtain the deep feature matrix MO 10 .

[0210] S242. Add the feature map MO1 and the deep feature matrix MO 10 element-wise to obtain the relocating gate first sum matrix.

[0211] S243. Perform a 1×1 convolution operation on the feature map ME1 to obtain the deep feature matrix ME8, and perform a 3×3 convolution operation on the deep feature matrix ME8 to obtain the deep feature matrix ME9.

[0212] S244. Add the feature map ME1 and the deep feature matrix ME9 element-wise to obtain the relocating gate second sum matrix.

[0213] S245. Multiply the relocating gate first sum matrix and the relocating gate second sum matrix, and perform global element activation through the Sigmoid layer to obtain the relocating gate voting matrix MO 11 .

[0214] S246. Obtain the electromagnetic positioning memory unit L t , perform two consecutive 3×3 convolution operations on the electromagnetic positioning memory unit L t to obtain the convolution result L3, add the convolution result L3 and the electromagnetic positioning memory unit L t element-wise, and perform activation through the Tanh layer to obtain the relocating matrix L4.

[0215] S247. Multiply the relocating gate voting matrix MO 11 and the relocating matrix L4 to obtain the relocating gate output feature map L5.

[0216] S248. Input the relocating gate output feature map L5 into the self-dimension reduction layer to obtain the relocating gate output feature Ot.

[0217] Optionally, the input module 320 is further configured to:

[0218] S251. For the current time t, input the target pose information sample at the previous time t - 1, the electromagnetic field phase difference sample at the current time t, and the magnetic field signal strength sample at the current time t into the screening gate to obtain the screening gate output feature map ML0.

[0219] S252. Subtract the electromagnetic positioning memory unit Lt-1 at the previous moment t-1 from the screening gate output feature map ML0 to obtain the electromagnetic positioning memory unit MLt-1.

[0220] S253. Input the feature map MO1 and the feature map ME1 in the screening gate into the perception gate to obtain the perception gate output feature map ML1.

[0221] S254. Add the electromagnetic positioning memory unit MLt-1 and the perception gate output feature map ML1 to obtain the electromagnetic positioning memory unit Lt.

[0222] S255. Input the feature map MO1, the feature map ME1, and the electromagnetic positioning memory unit Lt into the repositioning gate together to obtain the repositioning gate output feature Ot, which is the target pose information at the current moment.

[0223] In the embodiment of the present invention, the underground space electromagnetic positioning method based on trajectory time series analysis outputs the target positioning information in real time through the time series electromagnetic positioning network. Compared with the traditional method where the movement trajectory and posture of the target will affect the positioning accuracy and timeliness of the near-field ranging algorithm, it has strong time series signal processing capabilities, and can learn more essential electromagnetic positioning characterization mechanisms as data and time accumulate, so as to give more accurate real-time target positioning information.

[0224] The time series electromagnetic positioning network is designed for a specific task, that is, underground space electromagnetic positioning. It includes three structures: a screening gate, a perception gate, and a repositioning gate. The designs of the three gate structures are respectively used to filter redundant historical positioning information, perceive and store the current electromagnetic positioning representation, and reposition the current pose of the target. The three gate structures are connected in series in sequence, which can enhance the anti-interference ability of the electromagnetic positioning method, and realize accurate target positioning through in-depth mining and interaction of long-term positioning historical experience and current signals.

[0225] Figure 8 is a schematic structural diagram of an underground space electromagnetic positioning device provided by an embodiment of the present invention. As Figure 8 shown, the underground space electromagnetic positioning device may include the above-mentioned Figure 7 underground space electromagnetic positioning device based on trajectory time series analysis shown. Optionally, the underground space electromagnetic positioning device 410 may include a first processor 2001.

[0226] Optionally, the underground space electromagnetic positioning device 410 may further include a memory 2002 and a transceiver 2003.

[0227] Among them, the first processor 2001 is connected to the memory 2002 and the transceiver 2003, such as through a communication bus.

[0228] Next, in combination withFigure 8 Specifically introduce each component of the underground space electromagnetic positioning device 410:

[0229] Among them, the first processor 2001 is the control center of the underground space electromagnetic positioning device 410, which can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or can be an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. For example: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0230] Optionally, the first processor 2001 can execute various functions of the underground space electromagnetic positioning device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0231] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 8 CPU0 and CPU1 shown in

[0232] In a specific implementation, as an embodiment, the underground space electromagnetic positioning device 410 may also include multiple processors, such as Figure 8 the first processor 2001 and the second processor 2004 shown in

[0233] Among them, the memory 2002 is used to store software programs for executing the solution of the present invention and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiments and will not be elaborated here.

[0234] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or any other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently and be coupled to the first processor 2001 through an interface circuit ( Figure 8 not shown) of the underground space electromagnetic positioning device 410. The embodiments of the present invention do not make specific limitations thereto.

[0235] The transceiver 2003 is used to communicate with a network device or with a terminal device.

[0236] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 8 not separately shown). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0237] Optionally, the transceiver 2003 may be integrated with the first processor 2001 or may exist independently and be coupled to the first processor 2001 through an interface circuit ( Figure 8 not shown) of the underground space electromagnetic positioning device 410. The embodiments of the present invention do not make specific limitations thereto.

[0238] It should be noted that Figure 8 the structure of the underground space electromagnetic positioning device 410 shown does not constitute a limitation to the router. The actual knowledge structure recognition device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0239] In addition, the technical effects of the underground space electromagnetic positioning device 410 may refer to the technical effects of the underground space electromagnetic positioning method based on trajectory time series analysis described in the above method embodiments, and will not be elaborated here.

[0240] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0241] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM) or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM) and direct rambus RAM (DR RAM).

[0242] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, or a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0243] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context before and after.

[0244] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0245] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0246] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled professionals may 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.

[0247] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0248] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0249] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0250] In addition, the functional units in various embodiments of the present invention can be integrated in one processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0251] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0252] As described above, the above are only specific implementation manners of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An underground space electromagnetic positioning method based on trajectory time series analysis, characterized in that The method includes: S1. Obtain the electromagnetic field phase difference and magnetic field signal strength of the target to be located, and construct time-series input data according to the electromagnetic field phase difference and magnetic field signal strength; S2. Input the time-series input data into the constructed time-series electromagnetic positioning model; wherein, the time-series electromagnetic positioning model includes a screening gate, a perception gate, and a repositioning gate; S3. According to the time-series input data and the time-series electromagnetic positioning model, output the target pose information in real time; The construction process of the time-series electromagnetic positioning model in S2 includes: S21. Construct an electromagnetic positioning data set; wherein, the electromagnetic positioning data set includes electromagnetic field phase difference samples, magnetic field signal strength samples, and corresponding target pose information samples; S22. Construct a screening gate; S23. Construct a perception gate; S24. Construct a repositioning gate; S25. According to the screening gate, the perception gate, and the repositioning gate, construct a time-series electromagnetic positioning network; S26. Train the time-series electromagnetic positioning network according to the electromagnetic positioning data set to generate a time-series electromagnetic positioning model; The constructing of the time-series electromagnetic positioning network according to the screening gate, the perception gate, and the repositioning gate in S25 includes: S251. For the current moment t, input the target pose information sample at the previous moment t-1, the electromagnetic field phase difference sample at the current moment t, and the magnetic field signal strength sample at the current moment t into the screening gate to obtain the screening gate output feature map ML0; S252. Subtract the screening gate output feature map ML0 from the electromagnetic positioning memory unit Lt-1 at the previous moment t-1 to obtain the electromagnetic positioning memory unit MLt-1; S253. Input the feature maps MO1 and ME1 in the screening gate into the perception gate to obtain the perception gate output feature map ML1; S254. Add the electromagnetic positioning memory unit MLt-1 and the perception gate output feature map ML1 to obtain the electromagnetic positioning memory unit Lt; S255. Input the feature maps MO1, ME1, and the electromagnetic positioning memory unit Lt into the repositioning gate together to obtain the repositioning gate output feature Ot, which is the target pose information at the current moment.

2. The underground space electromagnetic positioning method based on trajectory time series analysis according to claim 1, wherein The constructing of the screening gate in S22 includes: S221. For the current moment t, obtain the target pose information sample at the previous moment t-1, the electromagnetic field phase difference sample at the current moment t, and the magnetic field signal strength sample at the current moment t; S222. Input the target pose information sample at the previous moment t-1 into the first self-dimension-raising layer to obtain the feature map MO1; S223. Input the feature map MO1 into the convolutional module to obtain the feature map MO2; S224. Input the electromagnetic field phase difference sample at the current moment t and the magnetic field signal strength sample at the current moment t into the second self-dimension-raising layer to obtain the feature map ME1; S225. Input the feature map ME1 into the global pooling layer to output the feature vector ME2; S226. Input the feature vector ME2 into the SE layer to obtain the feature vector ME3; S227. Multiply the feature map MO2 by the feature vector ME3 to obtain the screening gate output feature map ML0.

3. The underground space electromagnetic positioning method based on trajectory time series analysis according to claim 1, characterized in that The construction of the perception gate in S23 includes: S231. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO3, and perform a 3×3 convolution operation on the deep feature matrix MO3 to obtain a deep feature matrix MO4; S232. Element-wise add the feature map MO1 and the deep feature matrix MO4 to obtain the first sum matrix of the perception gate; S233. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME4, and perform a 3×3 convolution operation on the deep feature matrix ME4 to obtain a deep feature matrix ME5; S234. Element-wise add the feature map ME1 and the deep feature matrix ME5 to obtain the second sum matrix of the perception gate; S235. Matrix multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate, and perform global element activation through the Sigmoid layer to obtain the perception gate voting matrix MO5; S236. Matrix multiply the first sum matrix of the perception gate and the second sum matrix of the perception gate, and perform activation through the Tanh layer to obtain the perception matrix MO8; S237. Matrix multiply the perception gate voting matrix MO5 and the perception matrix MO8 to obtain the perception gate output feature map ML1.

4. The electromagnetic positioning method for underground space based on trajectory time series analysis according to claim 1, wherein The construction of the repositioning gate in S24 includes: S241. Perform a 1×1 convolution operation on the feature map MO1 to obtain a deep feature matrix MO9, and perform a 3×3 convolution operation on the deep feature matrix MO9 to obtain a deep feature matrix MO 10 ; S242. Add the feature map MO1 and the deep feature matrix MO 10 element-wise to obtain the first sum matrix of the relocation gate; S243. Perform a 1×1 convolution operation on the feature map ME1 to obtain a deep feature matrix ME8, and perform a 3×3 convolution operation on the deep feature matrix ME8 to obtain a deep feature matrix ME9; S244. Element-wise add the feature map ME1 and the deep feature matrix ME9 to obtain the second sum matrix of the repositioning gate; S245. Matrix-multiply the first relocation gate sum matrix and the second relocation gate sum matrix, and perform global element activation through the Sigmoid layer to obtain the relocation gate voting matrix MO 11 ; S246. Obtain the electromagnetic positioning memory unit L t , and perform two 3×3 convolution operations on the electromagnetic positioning memory unit L t in sequence to obtain a convolution result L3. Add the convolution result L3 element-wise to the electromagnetic positioning memory unit L t and activate it through the Tanh layer to obtain a repositioning matrix L4; S247. Multiply the relocation gate voting matrix MO 11 by the relocation matrix L4 to obtain the relocation gate output feature map L5; S248. Input the repositioning gate output feature map L5 into the self-dimension reduction layer to obtain the repositioning gate output feature Ot.

5. An underground space electromagnetic positioning device based on trajectory time series analysis, the underground space electromagnetic positioning device based on trajectory time series analysis is used to implement the underground space electromagnetic positioning method based on trajectory time series analysis according to any one of claims 1-4, characterized in that, The device includes: An acquisition module, configured to acquire the electromagnetic field phase difference and the magnetic field signal strength of the target to be located, and construct time-series input data according to the electromagnetic field phase difference and the magnetic field signal strength; An input module, configured to input the time-series input data into the constructed time-series electromagnetic positioning model; wherein, the time-series electromagnetic positioning model includes a screening gate, a perception gate, and a repositioning gate; An output module, configured to output the target pose information in real time according to the time-series input data and the time-series electromagnetic positioning model; The construction process of the time-series electromagnetic positioning model includes: S21. Construct an electromagnetic positioning data set; wherein, the electromagnetic positioning data set includes electromagnetic field phase difference samples, magnetic field signal strength samples, and corresponding target pose information samples; S22. Construct a screening gate; S23. Construct a perception gate; S24. Construct a repositioning gate; S25. Construct a time-series electromagnetic positioning network according to the screening gate, the perception gate, and the repositioning gate; S26. Train the time-series electromagnetic positioning network according to the electromagnetic positioning data set to generate a time-series electromagnetic positioning model; Construct a temporal electromagnetic positioning network according to the screening gate, perception gate, and repositioning gate, including: S251. For the current time t, input the target pose information sample at the previous time t-1, the electromagnetic field phase difference sample at the current time t, and the magnetic field signal intensity sample at the current time t into the screening gate to obtain the screening gate output feature map ML0; S252. Subtract the screening gate output feature map ML0 from the electromagnetic positioning memory unit Lt-1 at the previous time t-1 to obtain the electromagnetic positioning memory unit MLt-1; S253. Input the feature maps MO1 and ME1 in the screening gate into the perception gate to obtain the perception gate output feature map ML1; S254. Add the electromagnetic positioning memory unit MLt-1 and the perception gate output feature map ML1 to obtain the electromagnetic positioning memory unit Lt; S255. Input the feature maps MO1, ME1, and the electromagnetic positioning memory unit Lt into the repositioning gate together to obtain the repositioning gate output feature Ot, which is the target pose information at the current time.

6. An underground space electromagnetic positioning device, characterized in that, The underground space electromagnetic positioning device includes: A processor; A memory, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor, the method described in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that, Program code is stored in the computer-readable storage medium, and the program code is called by the processor to execute the method described in any one of claims 1 to 4.

Images