A map updating method based on wireless signals and adversarial learning

Through adversarial learning technology, the problem of reduced positioning accuracy and high update cost in wireless signal indoor positioning technology due to environmental changes is solved, and high-precision and low-cost signal map updates are achieved to adapt to changes in complex indoor environments.

CN114882140BActive Publication Date: 2024-11-26SUN YAT SEN UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202210464466.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-29
Publication Date
2024-11-26
Estimated Expiration
2042-04-29

AI Technical Summary

Technical Problem

Existing indoor positioning technology based on wireless signals in complex indoor environments is caused by equipment changes and scene changes, which affects positioning accuracy and is expensive to manually collect and update.

Method used

Adversarial learning-based method is adopted to convert the signal fingerprint into a high-resolution thermal map, and train data is generated through scene constraint downsampling and random slice replacement, and an adversarial learning network is built, signal map is reconstructed, and update costs are reduced and accuracy is improved.

Benefits of technology

It realizes high-precision and low-cost signal map updates, adapts to scene changes, and reduces the time and cost of manual acquisition and update.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114882140B_ABST
    Figure CN114882140B_ABST
Patent Text Reader

Abstract

The present invention discloses a map updating method based on wireless signals and adversarial learning, the method comprising: converting the collected signal fingerprint into the form of a high-resolution heat map; downsampling the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map; randomly selecting a square area in the low-resolution heat map, using the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map, and completing the construction of training data; inputting the training data into an adversarial learning network for training to obtain a corresponding reconstruction model; and using the reconstruction model to reconstruct the signal map. The present invention has high precision and low cost, and can be widely used in the field of target positioning technology.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of target positioning, and in particular to a map updating method based on wireless signals and adversarial learning. Background Art

[0002] With the continuous development of mobile terminal technology and the huge application needs of industrial end users, indoor positioning technology has been widely used in commercial marketing, industrial production, cultural and educational entertainment, medical rescue and other fields. The current popular indoor positioning technologies include: Wi-Fi, geomagnetism, Bluetooth, ultra-wideband, image, radio frequency identification, etc. Among them, Wi-Fi-based positioning technology benefits from the extensive infrastructure deployment, which reduces the implementation cost of its positioning solution and is easier to implement than other positioning technologies. Its main technical solutions can be divided into three methods: based on arrival time, based on arrival angle and based on signal fingerprint. Among them, the signal fingerprint-based method has been widely studied and applied because it has low requirements on equipment accuracy and does not require additional deployment.

[0003] However, since wireless signals are easily affected by factors such as device changes and scene changes in complex indoor dynamic environments, the fingerprints collected manually in the early stage are quite different from those in the later stage of positioning. Outdated fingerprint maps seriously affect the accuracy of positioning, and the manual collection of fingerprint maps is time-consuming and labor-intensive, and the update cost is high. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a map updating method based on wireless signals and adversarial learning with high accuracy and low cost.

[0005] One aspect of the present invention provides a map updating method based on wireless signals and adversarial learning, comprising:

[0006] Convert the collected signal fingerprints into high-resolution heat maps;

[0007] Downsampling the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map;

[0008] Randomly select a square area in the low-resolution heat map, and use the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map to complete the construction of training data;

[0009] Inputting the training data into an adversarial learning network for training to obtain a corresponding reconstruction model;

[0010] The signal map is reconstructed using the reconstruction model.

[0011] Optionally, converting the collected signal fingerprint into a high-resolution heat map includes:

[0012] Acquire a wireless access point set and a reference point set of a target site in an offline phase;

[0013] Generate a signal fingerprint according to the signal strength value of each reference point in the reference point set received from different wireless access points;

[0014] constructing an initial signal map according to the signal fingerprint;

[0015] The initial signal map is converted into a high-resolution heat map of wireless access points.

[0016] Optionally, the downsampling operation on the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map includes:

[0017] Extracting obstacle areas where signals cannot be collected from the high-resolution thermal map, and using a binary mask to describe scene information of the obstacle areas;

[0018] Selecting down-sampling elements of the high-resolution heat map using a target window;

[0019] In the target window of a preset size, the entire operation window is filled with the value of the first non-zero element, and a small window selection operation is used to traverse the high-resolution heat map to complete the down-sampled heat map;

[0020] The binary mask is applied to the high-resolution thermal map to obtain a low-resolution thermal map under scene constraints.

[0021] Optionally, the method further includes the step of performing a data augmentation operation on the training data by random slice replacement, which step is specifically:

[0022] A square area is randomly selected in the low-resolution heat map, and the square area is used to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map.

[0023] Optionally, inputting the training data into an adversarial learning network for training to obtain a corresponding reconstruction model includes:

[0024] Constructing the adversarial learning network through a deep residual network generator and a generated heat map quality discriminator;

[0025] During training of the adversarial learning network, reducing a generated heat map by the deep residual network generator;

[0026] While training the deep residual network generator, the generated heat map quality discriminator is trained so that the generated heat map quality discriminator can determine whether the input image is a real heat map or a heat map output by the generator;

[0027] Determine the loss function of the adversarial learning network according to the loss function of the deep residual network generator and the loss function of the generated heat map quality discriminator;

[0028] According to the loss function of the adversarial learning network, the training of the adversarial learning network is completed to obtain a reconstruction model.

[0029] Optionally, reconstructing the signal map using the reconstruction model includes:

[0030] Collect sparse signal fingerprints in actual scenarios and calculate low-resolution heat maps corresponding to each wireless access point;

[0031] Inputting the low-resolution thermal map into a reconstruction model to obtain a predicted high-resolution thermal map;

[0032] A new signal map is calculated based on the predicted high-resolution heat map.

[0033] Optionally, the new signal map is calculated as follows:

[0034]

[0035]

[0036]

[0037] in, Represents the nth reference point RP n The received mth wireless access point AP m The signal strength value of Represents the reconstructed AP m The pixel value of the corresponding position of the reference point n in the high-resolution thermal map; Represents the maximum signal value in the signal map; Represents the minimum signal value in the signal map; S represents the number of signal acquisitions; Represents the signal fingerprint of the nth reference point; M t Represents the new signal map at time t.

[0038] An embodiment of the present invention further provides a map updating device based on wireless signals and adversarial learning, comprising:

[0039] The first module is used to convert the collected signal fingerprint into the form of a high-resolution heat map;

[0040] The second module is used to downsample the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map;

[0041] The third module is used to randomly select a square area in the low-resolution heat map, and use the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map to complete the construction of training data;

[0042] The fourth module is used to input the training data into the adversarial learning network for training to obtain a corresponding reconstruction model;

[0043] The fifth module is used to reconstruct the signal map using the reconstruction model.

[0044] Another aspect of an embodiment of the present invention further provides an electronic device, including a processor and a memory;

[0045] The memory is used to store programs;

[0046] The processor executes the program to implement the method described above.

[0047] Another aspect of the embodiments of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the method described above.

[0048] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0049] The embodiment of the present invention converts the collected signal fingerprint into the form of a high-resolution heat map; downsamples the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map; randomly selects a square area in the low-resolution heat map, and uses the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map to complete the construction of training data; inputs the training data into an adversarial learning network for training to obtain a corresponding reconstruction model; and uses the reconstruction model to reconstruct the signal map. The present invention has high precision and low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0051] Figure 1 An overall step flow chart provided for an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of the structure of an adversarial training reconstruction network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] In view of the problems existing in the prior art, the present invention proposes a method for indoor positioning map reconstruction based on signal fingerprint adversarial learning. The method effectively solves the problem of a significant decrease in positioning accuracy caused by outdated signal maps in indoor positioning methods based on wireless signal fingerprints. At the same time, the method can simultaneously handle practical problems such as AP and site changes. The method mainly includes three parts: (1) a signal heat map construction module, (2) a signal heat map data enhancement module, and (3) a reconstruction model based on adversarial learning.

[0055] First, the signal heat map construction module converts the collected signal fingerprints into the form of high-resolution heat maps, and uses the scene-constrained downsampling method to calculate the corresponding low-resolution heat maps. The scene-constrained downsampling method better retains the scene features during the downsampling process. Secondly, the signal heat map data enhancement module randomly selects square areas in the low-resolution heat map and uses the area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map. This method effectively enhances the richness of training data and improves the reconstruction accuracy. Finally, the reconstruction model based on adversarial learning adds a discriminator to the residual network generator to make the generator's reconstructed heat map closer to the real heat map, thereby improving the texture features of the generated heat map and improving the reconstruction and positioning accuracy. In the update stage, only a small number of reference point fingerprints in the scene need to be collected, and the sparse heat map is input into the reconstruction model to obtain the corresponding high-resolution heat map, which greatly reduces the update cost of the signal map.

[0056] Specifically, one aspect of the present invention provides a map updating method based on wireless signals and adversarial learning, comprising:

[0057] Convert the collected signal fingerprints into high-resolution heat maps;

[0058] Downsampling the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map;

[0059] Randomly select a square area in the low-resolution heat map, and use the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map to complete the construction of training data;

[0060] Inputting the training data into an adversarial learning network for training to obtain a corresponding reconstruction model;

[0061] The signal map is reconstructed using the reconstruction model.

[0062] Optionally, converting the collected signal fingerprint into a high-resolution heat map includes:

[0063] Acquire a wireless access point set and a reference point set of a target site in an offline phase;

[0064] Generate a signal fingerprint according to the signal strength value of each reference point in the reference point set received from different wireless access points;

[0065] constructing an initial signal map according to the signal fingerprint;

[0066] The initial signal map is converted into a high-resolution heat map of wireless access points.

[0067] Optionally, the downsampling operation on the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map includes:

[0068] Extracting obstacle areas where signals cannot be collected from the high-resolution thermal map, and using a binary mask to describe scene information of the obstacle areas;

[0069] Selecting down-sampling elements of the high-resolution heat map using a target window;

[0070] In the target window of a preset size, the entire operation window is filled with the value of the first non-zero element, and a small window selection operation is used to traverse the high-resolution heat map to complete the down-sampled heat map;

[0071] The binary mask is applied to the high-resolution thermal map to obtain a low-resolution thermal map under scene constraints.

[0072] Optionally, the method further includes the step of performing a data augmentation operation on the training data by random slice replacement, which step is specifically:

[0073] A square area is randomly selected in the low-resolution heat map, and the square area is used to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map.

[0074] Optionally, inputting the training data into an adversarial learning network for training to obtain a corresponding reconstruction model includes:

[0075] Constructing the adversarial learning network through a deep residual network generator and a generated heat map quality discriminator;

[0076] During training of the adversarial learning network, reducing a generated heat map by the deep residual network generator;

[0077] While training the deep residual network generator, the generated heat map quality discriminator is trained so that the generated heat map quality discriminator can determine whether the input image is a real heat map or a heat map output by the generator;

[0078] Determine the loss function of the adversarial learning network according to the loss function of the deep residual network generator and the loss function of the generated heat map quality discriminator;

[0079] According to the loss function of the adversarial learning network, the training of the adversarial learning network is completed to obtain a reconstruction model.

[0080] Optionally, reconstructing the signal map using the reconstruction model includes:

[0081] Collect sparse signal fingerprints in actual scenarios and calculate low-resolution heat maps corresponding to each wireless access point;

[0082] Inputting the low-resolution thermal map into a reconstruction model to obtain a predicted high-resolution thermal map;

[0083] A new signal map is calculated based on the predicted high-resolution heat map.

[0084] Optionally, the new signal map is calculated as follows:

[0085]

[0086]

[0087]

[0088] in, Represents the nth reference point RP n The received mth wireless access point AP m The signal strength value of Represents the reconstructed AP m The pixel value of the corresponding position of the reference point n in the high-resolution thermal map; Represents the maximum signal value in the signal map; Represents the minimum signal value in the signal map; S represents the number of signal acquisitions; Represents the signal fingerprint of the nth reference point; M t Represents the new signal map at time t.

[0089] An embodiment of the present invention further provides a map updating device based on wireless signals and adversarial learning, comprising:

[0090] The first module is used to convert the collected signal fingerprint into the form of a high-resolution heat map;

[0091] The second module is used to downsample the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map;

[0092] The third module is used to randomly select a square area in the low-resolution heat map, and use the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map to complete the construction of training data;

[0093] The fourth module is used to input the training data into the adversarial learning network for training to obtain a corresponding reconstruction model;

[0094] The fifth module is used to reconstruct the signal map using the reconstruction model.

[0095] Another aspect of an embodiment of the present invention further provides an electronic device, including a processor and a memory;

[0096] The memory is used to store programs;

[0097] The processor executes the program to implement the method described above.

[0098] Another aspect of the embodiments of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the method described above.

[0099] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0100] The specific implementation principle of the present invention is described in detail below in conjunction with the accompanying drawings:

[0101] like Figure 1 As shown, the present invention discloses a method for indoor positioning map reconstruction based on signal fingerprint adversarial learning. First, the signal acquisition and construction of the initial signal map are performed in the offline stage to generate paired high- and low-resolution AP heat maps; secondly, the CutPaste data enhancement method is used to enhance the diversity of training data; finally, the adversarial learning reconstruction network is trained to obtain a signal map reconstruction model. Specifically, it mainly includes the following steps:

[0102] Step 1: In the offline phase, there are M wireless access points and N reference points in the site, and the initial signal map Μ0={f1,f2,...,f N}, where fingerprint It is in RP n Received AP m When collecting signal fingerprints, each reference point needs to collect signals S times, so

[0103] Step 2: Initial signal map Μ0={f1,f2,...,f N} is converted into the form of AP heat map. For the pixel value in the W×H signal heat map in, is the minimum signal value in the signal map, and the corresponding is the maximum value of the signal in the signal map. Therefore, the initial signal map Μ0 is converted into M×S AP high-resolution heat maps. The present invention represents the high-resolution heat map as AP m High-resolution heat map of the sth sampling.

[0104] Step 3: Use the scene-constrained downsampling algorithm to downsample the high-resolution heat map to obtain a low-resolution heat map

[0105] Step 3-1: From the heat map In the example above, we extract the obstacle area where the signal cannot be collected and use a binary mask Mask. s Describes scene information.

[0106] Step 3-2: In a W×H heat map, use w s ×h s In the small window, the value of the first non-zero element is used to fill the entire operation window, and the small window selection operation is used to traverse the high-resolution heat map to complete the down-sampling heat map. The pixel value calculation method in the low-resolution heat map is:

[0107]

[0108] Step 3-3: Use binary mask Mask s Acting on heatmaps Above, we get a low-resolution heat map under scene constraints. (In the subsequent steps, it is simplified as ), the calculation process is:

[0109]

[0110] Step 4: Apply the CutPaste data augmentation method to increase the diversity of training data. Extract random slices from Paste it into The corresponding area in . The specific enhanced training data is defined as:

[0111]

[0112]

[0113] Step 5: The enhanced paired training data I HR and I LR Input into the adversarial learning network for training to obtain the corresponding reconstruction model.

[0114] Step 5-1: The adversarial learning network mainly includes a deep residual network generator G and a generated heat map quality discriminator D. The specific network structure is as follows Figure 2 shown.

[0115] Step 5-2: During the training process, the training goal of the generator G is to reduce the generated heat map Try to get as close to the actual heat map as possible I HR . Its MSE loss function is defined as:

[0116]

[0117] Step 5-3: While training the generator, the discriminator D is trained. The continuous iterative training enables the discriminator to determine whether the input image is a real heat map or a heat map output by the generator. Thus, in the adversarial training, the output of the generator is constantly approaching the real heat map. In the training of the generator, its adversarial loss function is defined as:

[0118] L ADV =log[1-D(G(I LR ))].

[0119] Step 5-4: Combining MSE loss and adversarial loss, the loss function of generator training is:

[0120] L G =L MSE +αL ADV .

[0121] Step 6: When the signal map changes significantly in the scene or needs to be updated regularly, the signal map is reconstructed using the reconstruction model.

[0122] Step 6-1: The fingerprint collector collects the sparse signal fingerprint in the scene at time t and calculates AP m Corresponding low-resolution heat map

[0123] Step 6-2: Input into the trained reconstruction model G, and output the corresponding predicted high-resolution thermal map

[0124] Step 6-3: Obtain a new signal map from the reconstructed high-resolution thermal map. The signal map is calculated as follows:

[0125]

[0126]

[0127]

[0128] In summary, the present invention discloses a method for indoor positioning signal map reconstruction based on signal fingerprint adversarial learning. This method can be used to perform sparse sampling in the positioning scene to complete the reconstruction of the dense signal map, so as to solve the problem of serious reduction in positioning accuracy caused by the expiration of the signal map. The advantages of the technical solution of the present invention are mainly: (1) constructing a sparse to dense mapping model to greatly reduce the manpower and time costs required for signal map updates; (2) using an adversarial learning model to enhance the texture details of the generated signal heat map, thereby improving the location distinction ability of the reconstructed signal map; (3) the scene-constrained downsampling method performs scene recognition in the heat map, which effectively improves the model's adaptability to scene changes; (4) the CutPaste data enhancement method improves the diversity of the training data set and further optimizes the reconstruction accuracy of the reconstruction model.

[0129] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0130] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0131] If the 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, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0132] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0133] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0134] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0135] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0136] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0137] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A map updating method based on wireless signals and adversarial learning, characterized in that: include: Convert the collected signal fingerprints into high-resolution heat maps; Downsampling the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map; Randomly select a square area in the low-resolution heat map, and use the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map to complete the construction of training data; Inputting the training data into an adversarial learning network for training to obtain a corresponding reconstruction model; reconstructing a signal map using the reconstruction model; The downsampling operation of the high-resolution heat map using the scene-constrained downsampling method to obtain a corresponding low-resolution heat map includes: Extracting obstacle areas where signals cannot be collected from the high-resolution thermal map, and using a binary mask to describe scene information of the obstacle areas; Selecting down-sampling elements of the high-resolution heat map using a target window; In the target window of a preset size, the entire operation window is filled with the value of the first non-zero element, and a small window selection operation is used to traverse the high-resolution heat map to complete the down-sampled heat map; Applying the binary mask to the high-resolution thermal map to obtain a low-resolution thermal map under scene constraints; The using the reconstruction model to reconstruct the signal map comprises: Collect sparse signal fingerprints in actual scenarios and calculate low-resolution heat maps corresponding to each wireless access point; Inputting the low-resolution thermal map into a reconstruction model to obtain a predicted high-resolution thermal map; A new signal map is calculated based on the predicted high-resolution heat map.

2. A map updating method based on wireless signals and adversarial learning according to claim 1, characterized in that: The method of converting the collected signal fingerprint into a high-resolution heat map includes: Acquire a wireless access point set and a reference point set of a target site in an offline phase; Generate a signal fingerprint according to the signal strength value of each reference point in the reference point set received from different wireless access points; constructing an initial signal map according to the signal fingerprint; The initial signal map is converted into a high-resolution heat map of wireless access points.

3. A map updating method based on wireless signals and adversarial learning according to claim 1, characterized in that: The step of inputting the training data into an adversarial learning network for training to obtain a corresponding reconstruction model includes: Constructing the adversarial learning network through a deep residual network generator and a generated heat map quality discriminator; During training of the adversarial learning network, reducing a generated heat map by the deep residual network generator; While training the deep residual network generator, the generated heat map quality discriminator is trained so that the generated heat map quality discriminator can determine whether the input image is a real heat map or a heat map output by the generator; Determine the loss function of the adversarial learning network according to the loss function of the deep residual network generator and the loss function of the generated heat map quality discriminator; According to the loss function of the adversarial learning network, the training of the adversarial learning network is completed to obtain a reconstruction model.

4. A map updating method based on wireless signals and adversarial learning according to claim 1, characterized in that: The new signal map is calculated as follows: in, Represents the nth reference point RP n The received mth wireless access point AP m The signal strength value of Represents the reconstructed AP m The pixel value of the corresponding position of the reference point n in the high-resolution thermal map; Represents the maximum signal value in the signal map; Represents the minimum signal value in the signal map; S represents the number of signal acquisitions; Represents the signal fingerprint of the nth reference point; M t Represents the new signal map at time t.

5. A device for implementing the map updating method based on wireless signals and adversarial learning as described in any one of claims 1 to 4, characterized in that: include: The first module is used to convert the collected signal fingerprint into the form of a high-resolution heat map; The second module is used to downsample the high-resolution heat map using a scene-constrained downsampling method to obtain a corresponding low-resolution heat map; The third module is used to randomly select a square area in the low-resolution heat map, and use the square area to replace the corresponding area in the high-resolution heat map as a new low-resolution heat map to complete the construction of training data; The fourth module is used to input the training data into the adversarial learning network for training to obtain a corresponding reconstruction model; The fifth module is used to reconstruct the signal map using the reconstruction model.

6. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement the method according to any one of claims 1 to 4.

Citation Information

Patent Citations

  • Indoor position fingerprint map generation method and system based on adversarial migration network

    CN110234085A

  • Indoor positioning method for reconstructing wifi fingerprint map

    CN110300370A

  • Wi-Fi fingerprint map reconstruction method and device, terminal device and medium

    CN113108792A

  • Global event-based avatar

    US10656797B1