Internet of Things smart phone indoor positioning method and system

Abstract

The invention provides an Internet of Things smart phone indoor positioning method and system, the method is applied to an Internet of Things-based indoor positioning system, and in the scheme, in response to a positioning operation of a user, a control terminal sends a positioning request to a positioning server; the positioning server respectively sends a data acquisition request to the plurality of access devices according to the positioning request, each access device acquires communication data of the mobile phone and returns the communication data to the positioning server, the communication data comprises signal intensities at a plurality of moments, and the positioning server sends the signal intensities to the plurality of access devices according to the communication data sent by the plurality of access devices. And predicting the position of the to-be-positioned user by using the position prediction model to obtain a target position of the to-be-positioned user, and returning the target position to the control terminal. According to the scheme, artificial intelligence is adopted to learn the signal change condition of the target in the moving process in the complex indoor environment, so that accurate positioning of the user or the target to be positioned is realized, and the reliability of the predicted position is improved.

Description

Technical Field

[0001] This application relates to the fields of Internet of Things and indoor positioning technology, and particularly relates to an indoor positioning method and system for Internet of Things smartphones. Background Art

[0002] With the increasingly rich lifestyle, the services and activities that indoor venues can provide are also increasing. More and more users choose to carry out various activities in large indoor venues. However, the internal structure and environment of large indoor venues are complex, resulting in users being prone to getting lost during activities, and the demand for indoor precise positioning and navigation is getting higher and higher.

[0003] To solve the above problems, the technical solutions adopted at the present stage mainly detect the communication signals between the user's mobile phone and the base stations, measure the time required for the signals to be sent and received, calculate the distance between the mobile phone and the base stations, and realize the positioning of the user's location in the indoor by measuring the distances from multiple base stations.

[0004] However, the complex building structure of indoor scenes, the large number of wall barriers, and the movement of the user's own position will all lead to unstable signals and inaccurate signal detection, and cannot achieve precise positioning of the user indoors. Summary of the Invention

[0005] This application provides an indoor positioning method and system for Internet of Things smartphones to solve the technical problems that the complex building structure of indoor scenes at the present stage, the large number of wall barriers, and the movement of the user's own position will all lead to unstable signals and inaccurate signal detection, and cannot achieve precise positioning of the user indoors.

[0006] The first aspect of this application provides an indoor positioning method for Internet of Things smartphones, which is applied to an indoor positioning system based on the Internet of Things. The indoor positioning system includes a plurality of access devices, a positioning server, and a control terminal distributed in a target venue. The method includes: In response to the user's positioning operation, the control terminal sends a positioning request to the positioning server, and the positioning request includes the identifier of the mobile phone of the user to be positioned; The positioning server receives the positioning request from the control terminal, and respectively sends a data acquisition request to the plurality of access devices according to the positioning request. The data acquisition request includes the identifier of the mobile phone; After each access device receives the data acquisition request, it acquires the communication data of the mobile phone and returns the communication data to the positioning server. The communication data includes the signal strength of the mobile phone detected at the current moment and the signal strengths at multiple moments before the current moment; The positioning server predicts the location of the user to be located according to the communication data sent by multiple access devices, and obtains the target location of the mobile phone. The location prediction model is a model obtained by training a recurrent neural network model based on a signal fingerprint database. The signal fingerprint database includes: the signal strength and acquisition time of the test terminal at each location in the indoor scene detected by multiple access devices set in the indoor scene during the historical time period; The positioning server returns the target location to the control terminal.

[0007] In a specific embodiment, the method further includes: The positioning server receives a positioning model configuration message sent by the cloud server, and the positioning model configuration message includes the location prediction model; The positioning server configures the location prediction model in the positioning program based on the positioning model configuration message.

[0008] In a specific embodiment, the method further includes: In the positioning server, in response to the operation of the user, an initial long short-term memory network (LSTM) model is constructed using Keras; The data in the signal fingerprint database is normalized to obtain a sample set. The sample set includes multiple samples, and each sample includes: an input signal matrix and a corresponding location. The input signal strength matrix includes signal strength sequences corresponding to multiple access devices. Each signal strength sequence corresponding to an access device includes the signal strength of the test terminal collected by the access device at any location in the test scene and the signal strength collected at at least two moments before the test terminal reaches the location; Based on the sample set, the LSTM model is trained to obtain the location prediction model.

[0009] In a specific embodiment, the step of training the LSTM model based on the sample set to obtain the location prediction model includes: Based on the samples in the sample set, the memory neural network of the LSTM model is optimized by an adaptive particle swarm optimization algorithm to obtain an LSTM model with optimized parameters; Based on the sample set, the LSTM model with optimized parameters is trained until a preset loss function converges to obtain the location prediction model.

[0010] In a specific embodiment, based on the samples in the sample set, the parameters of the memory neural network of the LSTM model are optimized by an adaptive particle swarm algorithm to obtain an LSTM model with optimized parameters, including: Initialize the particle swarm of the adaptive particle swarm algorithm to obtain an initial particle swarm, where the initial particle swarm includes multiple particles, and each particle is a parameter combination of the LSTM model; Determine the objective function of the LSTM model as the fitness function of the adaptive particle swarm algorithm, and the objective function is an accuracy evaluation function; Configure the parameter combination corresponding to each particle in the LSTM model, input the samples in the sample set into the LSTM model for prediction, and calculate the fitness of each particle using the fitness function according to the predicted value and the actual position of the sample to obtain the fitness value of each particle; Update the velocity and position of each particle according to the fitness value of each particle, and recalculate the fitness of each updated particle until the iteration number is reached to obtain a target particle swarm; Configure the parameter combination corresponding to the particle with the largest fitness value in the target particle swarm as the optimal parameter combination into the LSTM model to complete the parameter optimization.

[0011] In a specific embodiment, the optimal parameter combination includes the selected optimal number of units, the number of units in the Dense layer, the learning rate, and the dropout probability.

[0012] In a specific embodiment, before the positioning server returns the target position to the control terminal, the method further includes: The positioning server uses a filtering algorithm to optimize the target position to obtain a new optimized target position; Correspondingly, the positioning server returns the target position to the control terminal, including: The positioning server returns the new target position to the control terminal.

[0013] In a specific embodiment, the method further includes: The positioning server generates the trajectory route and the moving direction of the mobile phone in real time according to the obtained target positions of the mobile phone at multiple moments; The positioning server returns the trajectory route and the moving direction to the control terminal.

[0014] In a specific embodiment, the method further includes: After receiving the trajectory route and the moving direction, the control terminal displays the trajectory route and the moving direction in real time on the graphical user interface, and marks the current target position of the mobile phone on the trajectory route.

[0015] The second aspect of the present application provides an indoor positioning system, including: a plurality of access devices distributed in a target venue, a positioning server, and a control terminal; Among them, the control terminal is used to respond to the user's positioning operation, send a positioning request to the positioning server, and the positioning request includes the identifier of the mobile phone of the user to be located; The positioning server is used to receive the positioning request from the control terminal, and send a data acquisition request to each of the plurality of access devices according to the positioning request, and the data acquisition request includes the identifier of the mobile phone; Each access device is used to acquire the communication data of the mobile phone according to the received data acquisition request and return the communication data to the positioning server. The communication data includes the signal strength detected by the mobile phone at the current moment and the signal strengths at a plurality of moments before the current moment; The positioning server is further used to predict the position of the user to be located by using a position prediction model according to the communication data sent by the plurality of access devices, and obtain the target position of the mobile phone. The position prediction model is a model obtained by training a recurrent neural network model based on a signal fingerprint database. The signal fingerprint database includes: the signal strength and acquisition time of a test terminal at each position in the indoor scene detected by a plurality of access devices set in the indoor scene during a historical time period; The positioning server is further used to return the target position to the control terminal.

[0016] The third aspect of the present application further provides a computer-readable storage medium, in which computer execution instructions are stored, and when the computer execution instructions are executed by a processor, they are used to implement the Internet of Things smart phone indoor positioning method according to any one of the first aspect.

[0017] The fourth aspect of the present application further provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the Internet of Things smart phone indoor positioning method according to any one of the first aspect.

[0018] The indoor positioning method and system for Internet of Things smartphones provided by this application are applied to an indoor positioning system based on the Internet of Things. In this solution, in response to a user's positioning operation, the control terminal sends a positioning request to the positioning server. The positioning server sends data acquisition requests to multiple access devices respectively according to the positioning request. Each access device acquires the communication data of the mobile phone and returns the communication data to the positioning server. The communication data includes the signal strengths at multiple moments. The positioning server predicts the position of the user to be located by using a position prediction model based on the communication data sent by multiple access devices, obtains the target position of the user to be located, and returns the target position to the control terminal. In this solution, artificial intelligence is used to learn the signal change situation of the target during the movement in a complex indoor environment, so as to achieve precise positioning of the user or the target to be located and improve the reliability of the predicted position. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings here are incorporated into the specification and form a part of this specification, showing the embodiments consistent with this application, and are used together with the specification to explain the principles of this application.

[0020] Figure 1 It is a schematic diagram of an indoor positioning system based on the Internet of Things provided by this application; Figure 2 It is a schematic flowchart of Embodiment 1 of the indoor positioning method for Internet of Things smartphones provided by an embodiment of this application; Figure 3 It is a schematic flowchart of Embodiment 2 of the indoor positioning method for Internet of Things smartphones provided by an embodiment of this application; Figure 4 It is a schematic flowchart of Embodiment 3 of the indoor positioning method for Internet of Things smartphones provided by an embodiment of this application; Figure 5 It is a schematic flowchart of Embodiment 4 of the indoor positioning method for Internet of Things smartphones provided by an embodiment of this application; Figure 6 It is a schematic structural diagram of Embodiment 1 of the indoor positioning device for Internet of Things smartphones provided by an embodiment of this application; Figure 7 It is a schematic structural diagram of Embodiment 2 of the indoor positioning device for Internet of Things smartphones provided by an embodiment of this application; Figure 8 It is a schematic structural diagram of Embodiment 3 of the indoor positioning device for Internet of Things smartphones provided by an embodiment of this application; Figure 9 It is a schematic structural diagram of the electronic device provided by an embodiment of this application.

[0021] Through the above-mentioned accompanying drawings, specific embodiments of the present application have been shown, and there will be more detailed descriptions hereinafter. These drawings and textual descriptions are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Description of the Embodiments

[0022] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0023] In the era of the Internet of Things, location-based services have become one of the essential services in people's work and life. The Global Positioning System (GPS) and base station positioning technology basically meet the needs of users for location-based services in outdoor scenarios. However, 80% of a person's life is spent indoors, and a large number of positioning requirements for individual users, service robots, new Internet of Things devices, etc. also occur indoors; while in indoor scenarios, due to the obstruction of buildings, the signals of the global navigation satellite system rapidly attenuate or even completely reject, unable to meet the needs of navigation and positioning in indoor scenarios.

[0024] Currently, the commonly used indoor positioning methods are mainly divided into seven types in principle: proximity detection method, centroid positioning method, multilateration method, triangulation method, pole method, fingerprint positioning method, and dead reckoning method. Different indoor positioning methods select different observables and extract the information required by the algorithm through different observable extraction algorithms. According to the positioning principle and observables, a variety of indoor positioning technologies have been derived. The most commonly used technical solution for indoor positioning is still to detect the communication signal between the user's mobile phone and the base station, measure the time required for the signal to be sent and received, calculate the distance between the mobile phone and the base station, and realize the positioning of the user's location indoors by measuring the distances to multiple base stations. However, the complex building structure, many wall obstructions, and the user's own position movement in indoor scenarios will all lead to unstable signals and inaccurate signal detection, and cannot achieve precise positioning of the user indoors.

[0025] Based on this, when the inventor was researching the indoor positioning technology for large public places with complex structures such as shopping malls and airports, it was found that the problem of insufficient positioning accuracy in the prior art could be solved by developing an indoor positioning method and system for Internet of Things smartphones based on artificial intelligence technology, and it also had good effects in terms of scalability, accuracy, cost, complexity, and response time. In this solution, first, a terminal application program for data collection and real-time data transmission was developed for the indoor positioning system of Internet of Things smartphones to synchronously collect Received Signal Strength Indication (RSSI) data during daily processes, and then the RSSI data was sent to the positioning server. At the same time, real-time monitoring software was developed on the positioning server, which allowed the positioning server to communicate directly with smartphones to receive the detected RSSI data sets. Smartphones were used to collect the RSSI detected by each access device at each reference point (that is, each position divided in the indoor scene, which can be marked in advance). After collecting and sending all the data, the positioning server processed the data to form a signal fingerprint database. The next step was to train the recursive neural network algorithm based on the data in the signal fingerprint database, and finally obtain a position prediction model. Finally, in the online positioning stage, when real-time positioning was required, the positioning server collected the real-time RSSI of the mobile phone to be located through various access devices in the target location, used the previously obtained position prediction model to predict the position to obtain the target position, and fed it back to the control terminal or the user's mobile phone.

[0026] The technical solutions involved in this application will be introduced in detail through several specific embodiments below.

[0027] Figure 1 A schematic diagram of an indoor positioning system based on the Internet of Things provided in this application is as Figure 1 shown. In this indoor positioning system, at least a positioning server and multiple access devices are required. These access devices are respectively set at various positions in the target location. These access devices can be WIFI hotspots or wireless access points such as base stations (AccessPoint, abbreviated as: AP), devices that can detect the communication signal strength. This solution does not make specific limitations on this.

[0028] The positioning server can be a cloud server, a server dedicated to positioning, or a combination of multiple servers. After the training of the model is separately implemented on one of the servers, it can be configured on the server for positioning. This solution does not make limitations on this.

[0029] Further, during the actual application of the indoor positioning system, the system may further include a control terminal. It should be understood that the control terminal may be a public service terminal in a large target place such as a shopping mall, or a terminal such as a mobile phone or a computer of an associated user who needs to find the user to be located. Additionally, it may also be the terminal device of the user to be located, and this solution does not limit this.

[0030] Based on the above application scenarios, the implementation process of the indoor positioning method for Internet of Things smart phones provided in this application will be described in detail below.

[0031] Embodiment 1

[0032] Figure 2 is a schematic flowchart of Embodiment 1 of the indoor positioning method for Internet of Things smart phones provided in an embodiment of this application. As Figure 2 shown, the indoor positioning method for Internet of Things smart phones specifically includes the following steps: S101: In response to a positioning operation of a user, the control terminal sends a positioning request to the positioning server.

[0033] In this step, when in an indoor scenario such as a shopping mall, an airport, or a large indoor amusement park, when it is necessary to locate a certain user, a positioning operation can be initiated from the control terminal. The operating user on the control terminal operates the control terminal and inputs identification information such as a mobile phone number that can uniquely identify the mobile phone of the user to be located, and initiates the positioning server. In response to this operation, the control terminal can send a positioning request to the positioning server of the entire system, and the positioning request includes the identification of the mobile phone of the user to be located.

[0034] S102: The positioning server receives the positioning request from the control terminal and sends data acquisition requests to multiple access devices respectively according to the positioning request.

[0035] In this step, the data acquisition request includes the identification of the mobile phone.

[0036] For the positioning server, it receives the positioning request sent by the control terminal, which includes the identification of the mobile phone to be located. Based on this, the positioning server sends a data acquisition request carrying the identification of the mobile phone to multiple access devices in the current indoor positioning system, so that these access devices acquire communication data of the mobile phone based on this identification.

[0037] S103: After each access device receives the data acquisition request, it acquires the communication data of the mobile phone.

[0038] S104: Return the communication data to the positioning server. The communication data includes the signal strength of the mobile phone detected at the current moment and the signal strengths at multiple moments before the current moment.

[0039] In the above two steps, for each access device, after receiving a data acquisition request, communication data is acquired based on historical communication data and ongoing communication conditions. It should be understood that in the communication data in this solution, generally the communication signal strengths at multiple moments are required. For each access device, after receiving the data acquisition request, it can initiate communication with the mobile phone, detect the signal strengths at multiple moments to obtain the communication data, or it can also detect the signal strength at the current moment and obtain the signal strengths at several previous moments from the historical communication data to obtain the communication data. This solution does not limit this.

[0040] After each access device acquires the communication data, it returns the acquired communication data to the positioning server for analysis and processing respectively.

[0041] In this solution, it should be understood that due to the complex building structure of large indoor venues, although a relatively large number of access devices are arranged and the user to be located is indeed carrying the mobile phone and moving within the large venue, there are still cases where some access devices cannot detect the mobile phone at all. At this time, the communication data acquired by this access device can be 0 or empty. For example: If the configuration is based on data at three moments for location information, the communication data returned by the access device that fails to detect the mobile phone can include that the signal strengths corresponding to the three moments are all 0 or empty.

[0042] S105: The positioning server uses a location prediction model to predict the location of the user to be located based on the communication data sent by multiple access devices, and obtains the target location of the mobile phone.

[0043] In this step, after the positioning server acquires the communication data for the mobile phone to be located from each access device, it can use the pre-acquired location prediction model to analyze and process these data to obtain the target location of the mobile phone at the next moment. In this solution, the location prediction model is a model obtained by training a recursive neural network model based on a signal fingerprint database. The signal fingerprint database includes: the signal strengths and acquisition times of a test terminal at each location in an indoor scene detected by multiple access devices set in the indoor scene during a historical time period.

[0044] In specific implementation, since the LSTM deep learning model can effectively memorize and process the information of time series data, the LSTM model can be selected as the neural recursive network model for processing this location prediction. Other machine learning models can also be selected. This solution does not limit this.

[0045] S106: The positioning server returns the target location to the control terminal.

[0046] After the positioning server determines the target of the mobile phone to be positioned, the target location can be returned to the control terminal, so that the control terminal can display the target location, enabling the user with the need to know the specific location of the user to be positioned.

[0047] In this solution, a fingerprint database is established in advance by collecting signal fingerprints at various positions in a large indoor scene. A location prediction model is constructed based on a recurrent neural network model, and the location prediction model is obtained through model training based on the fingerprint database. Finally, this model is applied to some indoor positioning scenarios. When looking for the elderly, children, and other positioning needs, since the positioning prediction model has learned the signal instability factors caused by building structures and other factors at each position, that is, artificial intelligence is used to learn the signal changes during the movement of the target in a complex indoor environment to achieve precise positioning of the user or the target to be positioned, effectively improving the reliability of positioning prediction.

[0048] Based on the above embodiments, it can be known that in order to achieve precise positioning in a certain indoor scene, some data in this scene need to be collected in advance to build a signal fingerprint database for model training, and then the model is used in the actual positioning scene. The positioning server configuring the location prediction model includes at least the following two implementation schemes: The first implementation method: Perform model training on other servers, and after training, configure the obtained location prediction model into this positioning server for use.

[0049] For example, after standardizing the data in the signal fingerprint database, perform training of the location prediction model on the cloud server. After training, send the location prediction model to the positioning server. The positioning server receives the location model configuration message sent by the cloud server, and the location model configuration message includes the location prediction model. Then, based on the location model configuration message, configure the location prediction model in the positioning program for subsequent location prediction applications.

[0050] The second implementation method is that the positioning server providing positioning services for this indoor scene itself performs data collection and model training to obtain a location prediction model, and finally applies it in the positioning service.

[0051] Specifically, a test terminal can be set to perform tests at each position in the entire indoor scene. The specific test method can be that the test user takes the test device and performs multiple tests along different direction paths, detects the signal strength and collection time and stores them in the fingerprint database. After collecting complete data, perform standardization processing in the positioning server to obtain a sample set, and then perform model training to obtain the location prediction server.

[0052] Next, taking the selection of the LSTM model and the implementation of model training in the positioning server as an example (the process of model training on other servers is similar), the specific training process of the location prediction model provided in this application will be illustrated by way of example.

[0053] Before introducing the model training process of this embodiment, the construction of the signal fingerprint database will be introduced. First, location reference points are set in the test site (that is, the indoor scene where the final model will be applied). That is to say, all areas in the entire environment are simply divided by location. Especially for some environments with complex structures and layouts, the signal will be affected by the building structure and external devices, and location reference points need to be set at smaller intervals in these areas.

[0054] After the location reference points are set, the test user can wear a terminal for collecting signal strength (for example, a smart phone, a smartwatch, etc.), and the access device can collect signal strength at a preset frequency (this frequency can be set according to the actual situation, for example, collect once every second), taking into account both static and mobile scenarios. For the static scenario, the test user holds the test terminal to collect signal strength at each test location. For the mobile scenario, the user carries the test terminal and walks along a pre-designed route while recording the signal strength. For example, there are 6 access devices in the test scenario, and their location coordinates are (3,1), (10,1), (19,1), (3,6), (10,6), (19,6) respectively. For each location, the set of signal strengths collected (which can be WIFI signals, cellular network signals, or the signal strengths of other types of wireless communication technologies) can be defined as , where represents the signal strength of the th access device. If the th access device cannot be detected at the current location, then let be 0 or an empty number, or -100 (usually the default minimum value of RSSI). Suppose the set of signal strengths is recorded at a certain location. As the user moves, the test device moves to another location where the th access device cannot be detected. Then the signal strength detected by the th access device can be empty, 0, or the minimum value of the signal strength. However, there are other access devices j that detect the signal strength and record the signal strength, obtaining the set of signal strengths at this location.

[0055] Optionally, since the walking speeds of different users and the same user at different times are not consistent, in order to obtain the true position corresponding to the signal strength detected on the moving path, try to keep the user at a uniform speed within the same section of the path, then divide the path length by the total time to calculate the user's walking speed, and then interpolate according to the timestamps of the collected signal strengths between the start and end points of the preset route to obtain the true positions of the signal strengths of each sample on the moving path.

[0056] After collecting all the data through the above method, the various regions of the test scenario can be divided into grids of the same size (for example: 1m * 1m), and the size of the grid can be modified. The signal strengths collected by all access devices in the test scenario are collected at each grid vertex, and the collected signal strengths and position data are entered into the signal fingerprint database. For example, the received signal strength RSSI of the WIFI signal is determined by the position of the receiving device. As the communication distance increases, the strength of the signal will change due to the influence of the environment during the signal propagation process, and the magnitude of the signal strength attenuation is affected by the distance. However, through the above method of collecting signal strength, the change of the signal strength at each position is recorded in the database for the machine learning model to learn.

[0057] Embodiment 2

[0058] Figure 3 The following is a schematic flowchart of Embodiment 2 of the indoor positioning method for Internet of Things smartphones provided by the embodiments of the present application. As Figure 3 shown, the process of model training in the positioning server is as follows: S201: In the positioning server, in response to the user's operation, an initial LSTM model is constructed using keras.

[0059] In this step, the LSTM model constructed using keras can effectively remember and process the information of time series data. Its main structure is divided into two layers, namely the LSTM layer and the fully connected layer.

[0060] S202: Standardize the data in the signal fingerprint database to obtain a sample set. The sample set includes multiple samples, and each sample includes: an input signal matrix and the corresponding position. The input signal strength matrix includes signal strength sequences corresponding to multiple access devices. Each signal strength sequence corresponding to an access device includes the signal strength corresponding to any position of the test terminal collected by the access device and the signal strengths collected at at least two moments before the test terminal reaches the position.

[0061] Before training the model, it is necessary to process the data in the pre-collected fingerprint database to conform to the data structure of the input and output of the actual model. Since the discrete signal strengths and times collected through the test terminal in the test scenario before are all discrete, they need to be sorted into samples according to certain rules. Specifically, the final required position prediction model needs to use the signal strengths at several consecutive moments sent by multiple access devices as input to predict the position of the target at the next moment. Therefore, it is also necessary to standardize the data in the signal fingerprint database according to this rule to obtain training samples that conform to the model input and output structure.

[0062] In this embodiment, it should also be understood that since it is necessary to comprehensively consider the data collected by multiple access devices in actual implementation, the input of the LSTM model needs to be designed as different input units. Different input units are used to input the sequences collected by different access devices. The length of each sequence represents how many moments before are selected for the signal strength to calculate the current position prediction. For example, it can be set to 3 or 5 (specifically, it can be improved according to actual needs). It is necessary to recombine the data in the signal fingerprint database to obtain the input signal matrix for inputting into the LSTM model, and multiple samples composed of the positions corresponding to this signal matrix. For each input signal matrix, it includes the signal strength sequences corresponding to multiple access devices (such as the aforementioned 6 access devices). Each signal strength sequence corresponding to an access device includes the signal strength corresponding to any position of the test terminal collected by this access device and the signal strengths at several moments before the test terminal reaches the position. Through the processing in the above way, a sample set composed of multiple samples can be obtained.

[0063] S203: Based on the sample set, train the LSTM model to obtain a position prediction model.

[0064] In this step, after obtaining the sample set through processing, the LSTM model can be trained according to the samples in the sample set. In each training iteration, the data is fed into the model, and the loss value is calculated according to the preset loss function, and then the weights of the model are updated through the optimizer. And the results of the updated model are evaluated, and the model is tuned and optimized based on the evaluation results. The optimization also includes the optimization of the training method and the model architecture until the evaluation meets the requirements to obtain a position prediction model.

[0065] In a possible implementation, step 203 of the above embodiment may be specifically implemented as follows: Based on the samples in the sample set, the memory neural network of the LSTM model is optimized for parameters through an adaptive particle swarm optimization algorithm to obtain an LSTM model with optimized parameters. Then, the LSTM model with optimized parameters is trained based on the sample set until the preset loss function converges to obtain the position prediction model.

[0066] Specifically, the specific implementation steps for optimizing the parameters of the memory neural network of the LSTM model through the adaptive particle swarm optimization algorithm are as follows: Step 1: Initialize the particle swarm of the adaptive particle swarm optimization algorithm to obtain an initial particle swarm. The initial particle swarm includes multiple particles, and each particle is a parameter combination of the LSTM model.

[0067] Step 2: Determine the objective function of the LSTM model as the fitness function of the adaptive particle swarm optimization algorithm. The objective function is an accuracy evaluation function.

[0068] Step 3: Configure the parameter combination corresponding to each particle in the LSTM model, input the samples in the sample set into the LSTM model for prediction, and calculate the fitness of each particle using the fitness function according to the predicted value and the actual position of the sample to obtain the fitness value of each particle.

[0069] Step 4: Update the velocity and position of each particle according to the fitness value of each particle, and recalculate the fitness of each updated particle until the iteration number is reached to obtain the target particle swarm. Step 5: Configure the parameter combination corresponding to the particle with the largest fitness value in the target particle swarm as the optimal parameter combination into the LSTM model to complete the parameter optimization.

[0070] Optionally, the optimal parameter combination includes the selected optimal number of units, the number of units in the Dense layer, the learning rate, and the dropout probability.

[0071] Before introducing the technology of the solution of this embodiment in detail, first introduce the basic steps of the PSO algorithm. First, set the number of particles, initialize the positions and velocities of all particles, set the individual's optimal solution as the current position, and the optimal individual in the group as the current global optimal solution; in each iteration, calculate the fitness function value of each particle. If the current fitness function value is better than the individual's historical optimal value, update the individual's historical optimal value; if the current fitness function value is better than the global historical optimal value, update the global historical optimal value; when the iteration number reaches the maximum, the optimization process stops.

[0072] In this embodiment, the traditional PSO algorithm is improved by introducing chaotic sequence initialization to replace the random initialization of the particle population position in PSO. The pseudo-random initialization in the standard PSO may cause the population to gather in a certain local area, affecting the overall search performance. Chaotic sequences usually exhibit better randomness and distribute the initial population positions more widely in the search space, thus increasing the chance of finding the global optimal solution.

[0073] In the specific implementation, the Tent chaotic mapping is used to initialize the particle population position, and the formula is as follows: , where, ; In the above formula, is the chaotic sequence in the interval [0,1]; is the chaotic value, set to 0.7; is the population number; is the number of iterations; is the chaotic sequence that meets the particle range; and are the upper and lower boundaries of the particle range.

[0074] In addition, in the process of adjusting the particle weight in this scheme, a non-linear weight adjustment strategy is adopted. Compared with the linear weight adjustment, the non-linear decreasing weight adjustment is used, so that is larger in the early stage of the search, and the global search ability of the particle is stronger; so that is smaller in the later stage of the search, and the local optimization ability of the particle is stronger, improving the chance of finding the optimal solution. The improved inertia weight expression is:

[0075] In the formula, and are the maximum inertia weight and the minimum inertia weight, set to 0.7 and 0.4 respectively; is the maximum number of iterations.

[0076] Furthermore, the learning factors , of the particle are optimized simultaneously, so that decreases while increases. In the initial stage of the search, the self-learning ability of the particle is emphasized, and in the later stage, the social learning ability of the particle is emphasized, increasing the chance of obtaining the global optimal solution. The improved learning factor expression is: ; ; In the above formula, is the generation and are the maximum values of the individual and group learning factors, both set to 2.5; and are the minimum values of the individual and group learning factors, both set to 1.

[0077] Based on the above several improvements to the PSO algorithm, the adaptive particle swarm algorithm adopted in this solution is obtained.

[0078] After completing the algorithm improvement, based on the above improved adaptive particle swarm algorithm and the samples in the sample set, the parameters such as the number of hidden layer units and the learning rate in the LSTM model are optimized to obtain the final selected optimal parameter combination composed of the optimal number of units, the number of Dense layer units, the learning rate, and the dropout probability. The specific process is as follows: Use chaotic sequence initialization to initialize the particle swarm, generate a certain number of particles, and form an initial population of particles. Each particle represents a parameter combination in the LSTM model. This parameter combination can include at least one parameter to be optimized. In the preferred case, the above four parameters of the number of units, the number of Dense layer units, the learning rate, and the dropout probability can be selected.

[0079] The root mean square error between the prediction result of the LSTM model and the actual value can be used as the loss function and the fitness of the adaptive particle swarm algorithm. Configure the parameter combination corresponding to each particle in the LSTM model, input the samples in the sample set into the LSTM model, calculate the fitness of each particle according to the fitness function, and obtain the fitness value of each particle; update the speed and position of each particle according to the fitness value of each particle, and recalculate the fitness of each updated particle until the iteration times are reached to obtain the globally optimal target particle. The parameter combination corresponding to the position where the target particle is located can be configured in the LSTM model.

[0080] Then, configure other parameters and activation functions of the optimized LSTM model, set the calculation function for model training, divide the sample set into a training set and a validation set, train the optimized LSTM model with the samples in the training set, and test and verify the trained model based on the samples in the validation set for error analysis until the position prediction model in this application is obtained.

[0081] In the technical solution provided in this embodiment, the improved adaptive particle swarm algorithm optimizes the parameters of the LSTM model, which can improve the prediction ability of the LSTM model, that is, improve the accuracy of position prediction.

[0082] In a specific implementation of this solution, when selecting the LSTM model for tracking during training, the problem of gradient disappearance may lead to difficulties in modeling long-term dependencies of the model. Therefore, to improve the convergence speed, the ReLU function can be selected as the activation function. The derivative of this ReLU function is always 1 when the input is greater than 0, and there is no problem of gradient disappearance, which can effectively solve the problem of gradient disappearance. During the training process of the LSTM model, it is easier to learn and understand the characteristics of the data, obtaining better convergence effects and higher prediction accuracies.

[0083] Embodiment 3

[0084] Figure 4 is a schematic flowchart of Embodiment 3 of the indoor positioning method for Internet of Things smartphones provided by the embodiments of the present application. As Figure 4 shown, on the basis of any of the above embodiments, in another specific implementation manner of this indoor positioning method for Internet of Things smartphones, before the positioning server returns the target location to the control terminal, it further includes: S107: The positioning server optimizes the target location using a filtering algorithm to obtain a new optimized target location.

[0085] Correspondingly, S106 can be specifically implemented as: S1061: The positioning server returns the new target location to the control terminal.

[0086] In the technical solution of this embodiment, in order to further improve the accuracy of the predicted location, filtering can also be integrated with machine learning in the aforementioned solution during the prediction process in the positioning server. The filtering algorithm is used to correct the predicted target location, which can reduce the errors caused by sudden fluctuations in signal strength due to the complex indoor environment. Especially for people who are constantly moving, in this application, a position prediction model is first used for prediction and positioning, and the predicted target location is used as the input of the filter for optimization.

[0087] In this solution, because the motion states of the user to be located are continuous, correlated, and have a certain regularity, the method of filtering can provide better positioning accuracy and positioning stability. The particle filter algorithm is an approximate estimation method that uses particle sampling to simulate the state space at any time. The particle filter updates the weights and states of the particles based on information and the current observation values (keep the particles with high weights, discard the particles with low weights, and then weighted accumulate the predicted positions of the remaining particles, which is the current moment). Finally, it can output the position trajectory under the premise of high accuracy. Therefore, this solution preferably uses the particle filter method to optimize the target location.

[0088] Specifically, the algorithm process of optimizing the target location using the particle filter algorithm is as follows: (1) In the entire area of the preset range around the predicted target position, if the total number of position reference points in the preset range is N, then these position reference points are randomly distributed as particles with a total number of , and the state of the particle swarm is initialized, including the position of each particle, the movement direction angle and the movement speed . And the weight value of each particle is initialized to . Among them, and follow a uniform distribution, is a constant. However, when the user to be located is moving, the speed will not be constant, and there will be speed noise expressed as , and it follows a standard normal distribution.

[0089] (2) The particles move according to the particle state equation of the following formula to predict the next position of the particle swarm (( , ) represents the position of the particle at moment):

[0090] Among them, is the movement direction angle of the particle at moment.

[0091] (3) Update the particle weights according to the probability information that the Euclidean distance between the predicted value and the true position of each particle satisfies the Gaussian distribution function; (4) Calculate the number of effective particles. If the number of effective particles is less than a certain threshold, particle resampling is performed, and the weights of the resampled particles are normalized; (5) The positions represented by the particles resampled in (4) are weighted and accumulated, and the weighted sum obtained is the optimized new target position.

[0092] As the positioning target moves, in the subsequent position prediction process, every time the target is predicted to be a child, the above-mentioned particle filter optimization process is repeated, which can not only improve the positioning accuracy, but also achieve real-time position tracking.

[0093] In the IoT smartphone positioning method provided in this embodiment, a filtering algorithm is introduced to further optimize the target position predicted by the machine learning method, so as to improve the accuracy of the positioned position. In this preferred solution, the particle filter algorithm is selected for fusion. During the optimization process, the Euclidean distance between the real position and the estimated position is used as the position error and the time complexity (running time), as the index to evaluate the performance of the positioning algorithm, effectively reducing the error caused by the sudden change in signal strength caused by indoor obstacles and multipath phenomena, improving the positioning accuracy, and having a fast processing speed and good timeliness performance.

[0094] Embodiment 4

[0095] Figure 5 It is a schematic flowchart of Embodiment 4 of the IoT smartphone indoor positioning method provided by an embodiment of this application. As Figure 5 shown, on the basis of any of the above embodiments, the IoT smartphone indoor positioning method further includes the following steps: S301: The positioning server generates the trajectory route and moving direction of the mobile phone in real time according to the obtained target positions of the mobile phone at multiple moments.

[0096] S302: The positioning server returns the trajectory route and the moving direction to the control terminal.

[0097] S303: After receiving the trajectory route and the moving direction, the control terminal displays the trajectory route and the moving direction on the graphical user interface in real time, and marks the current target position of the mobile phone on the trajectory route.

[0098] In this solution, since the positioning server can obtain the position of the mobile phone in real time, after accumulating the position points at multiple moments, it can generate the movement trajectory of the mobile phone according to these trajectory points, that is, the trajectory route of the user to be located. Based on the target positions at several moments, the moving direction of the user to be located can also be determined. After obtaining the trajectory route and the moving direction, the trajectory route and the moving direction are returned to the control terminal. The control terminal can display them on the graphical user interface and at the same time mark the current target position of the mobile phone of the user to be located on the route.

[0099] Through this method, this solution can intuitively provide the control terminal with the moving direction, route and position of the user to be located, facilitating the user of the demand side to find the target and improving the efficiency and accuracy of finding the target.

[0100] Furthermore, this implementation method can interface with institutional platforms such as large shopping malls, security, medical, and scenic spots, visually see one's own or the other party's location and target orientation in three dimensions, form precise positioning, and react in real time to the tourist's terminal device, which has important applications in scenarios such as robot autonomous positioning and finding lost users. At the same time, during the implementation process of this solution, a large amount of effective information can be collected before, during, and after the event, and various possible risk events can be promptly warned.

[0101] Furthermore, the display in the control terminal can also select a three-dimensional space display. Especially for a scene space with a complex structure, the entire indoor scene can be three-dimensionally modeled in advance through image acquisition methods, combined with simultaneous localization and mapping technology, to obtain a three-dimensional space model. After obtaining the target position, trajectory route, and moving direction of the user to be located through the foregoing technical solution, a virtual object can be created in the three-dimensional space model, and the virtual object, trajectory route, and moving direction can be superimposed on the corresponding positions in the three-dimensional space model in the form of virtual images and displayed on the interface. The specific display of the trajectory and direction can include dynamic arrows, path lines, etc., and these elements can be updated in real time as the target moves. In this way, the intuitiveness of the interaction can be improved, and the efficiency can be increased when looking for lost people.

[0102] In summary, this application has developed an indoor positioning method and system for Internet of Things smartphones. This method combines signal strength positioning technology, signal strength fingerprint feature extraction, and intelligent algorithms, thus significantly improving the accuracy and efficiency of personnel positioning prediction in large shopping malls. By automatically collecting key signal fingerprint information, preferably, the IPSO-LSTM and filtering fusion positioning algorithm combination method is used to achieve mobile phone positioning, effectively improving the accurate prediction of indoor mobile phone positioning.

[0103] In the technical solution of this application, it should be understood that the foregoing method embodiments are only examples with smartphones as the target to be located. In actual applications, any terminal capable of wireless communication can use the above technical solution to achieve precise positioning prediction, such as a smartwatch, a tablet computer, a smart bracelet anti-loss wearable device, etc. This solution is not limited thereto.

[0104] Figure 6 FIG. 13 is a schematic structural diagram of Embodiment 1 of an indoor positioning device for Internet of Things smartphones provided by an embodiment of this application. As Figure 6 shown, the indoor positioning device for Internet of Things smartphones is used to implement the method on the positioning server side in the foregoing method embodiments. Specifically, the indoor positioning device 10 for Internet of Things smartphones includes: A receiving module 11, configured to receive a positioning request from the control terminal; The sending module 12 is configured to send data acquisition requests to the multiple access devices respectively according to the positioning request, and the data acquisition request includes the identifier of the mobile phone; The receiving module 11 is further configured to receive communication data returned by each access device, and the communication data includes the signal strength of the mobile phone detected at the current moment and the signal strengths at multiple moments before the current moment.

[0105] The processing module 13 is configured to predict the position of the user to be located by using a position prediction model according to the communication data sent by multiple access devices, and obtain the target position of the mobile phone. The position prediction model is a model obtained by training a recurrent neural network model based on a signal fingerprint database. The signal fingerprint database includes: the signal strength and acquisition time of a test terminal at each position in an indoor scene detected by multiple access devices set in the indoor scene during a historical time period; The sending module 12 is further configured to return the target position to the control terminal.

[0106] Optionally, the receiving module 11 is further configured to receive a positioning model configuration message sent by a cloud server, and the positioning model configuration message includes the position prediction model; The processing module 13 is further configured to configure the position prediction model in a positioning program based on the positioning model configuration message.

[0107] Optionally, the processing module 13 is further configured to: In the positioning server, in response to a user's operation, construct an initial long short-term memory network (LSTM) model by using Keras; Normalize the data in the signal fingerprint database to obtain a sample set. The sample set includes multiple samples, and each sample includes: an input signal matrix and a corresponding position. The input signal strength matrix includes signal strength sequences corresponding to multiple access devices. Each signal strength sequence corresponding to an access device includes the signal strength of the test terminal detected at any position in the test scene by the access device and the signal strengths collected at at least two moments before the test terminal reaches the position; Based on the sample set, train the LSTM model to obtain the position prediction model.

[0108] Optionally, the processing module is specifically configured to: Based on the samples in the sample set, optimize the parameters of the memory neural network of the LSTM model by using an adaptive particle swarm algorithm to obtain an LSTM model with optimized parameters; Based on the sample set, train the LSTM model with optimized parameters until the preset loss function converges to obtain the position prediction model.

[0109] Optionally, the processing module 13 is further specifically configured to: Initialize the particle swarm of the adaptive particle swarm algorithm to obtain an initial particle swarm, where the initial particle swarm includes multiple particles, and each particle is a parameter combination of the LSTM model; Determine the objective function of the LSTM model as the fitness function of the adaptive particle swarm algorithm, and the objective function is an accuracy evaluation function; Configure the parameter combination corresponding to each particle in the LSTM model, input the samples in the sample set into the LSTM model for prediction, and calculate the fitness of each particle using the fitness function based on the predicted value and the actual position of the sample to obtain the fitness value of each particle; Update the speed and position of each particle according to the fitness value of each particle, and recalculate the fitness of each updated particle until the iteration number is reached to obtain a target particle swarm; Configure the parameter combination corresponding to the particle with the largest fitness value in the target particle swarm as the optimal parameter combination into the LSTM model to complete parameter optimization.

[0110] Optionally, the optimal parameter combination includes the selected optimal number of units, the number of units in the Dense layer, the learning rate, and the dropout probability.

[0111] Optionally, the processing module 13 is further configured to: optimize the target position using a filtering algorithm to obtain a new optimized target position.

[0112] Correspondingly, the sending module 12 is specifically configured to return the new target position to the control terminal.

[0113] Optionally, the processing module 13 is further configured to generate the trajectory route and the moving direction of the mobile phone in real time according to the obtained target positions of the mobile phone at multiple moments; The sending module 12 is further specifically configured to return the trajectory route and the moving direction to the control terminal.

[0114] Figure 7 This is the structural schematic diagram of the second embodiment of the Internet of Things smart phone indoor positioning device provided by the embodiment of the present application. As Figure 7 shown, this Internet of Things smart phone indoor positioning device is used to implement the method on any access device side in the foregoing method embodiments. Specifically, this Internet of Things smart phone indoor positioning device 20 includes: A receiving module 21, configured to receive a data acquisition request sent by a positioning server, where the data acquisition request includes an identifier of the mobile phone. A processing module 22, configured to acquire communication data of the mobile phone according to the data acquisition request, where the communication data includes signal strengths detected at the current moment of the mobile phone and signal strengths at multiple moments before the current moment. A sending module 23, configured to return the communication data to the positioning server.

[0115] Figure 8 FIG. is a schematic structural diagram of Embodiment 3 of the indoor positioning device for an Internet of Things smart phone provided by an embodiment of the present application. As Figure 8 shown, the indoor positioning device for an Internet of Things smart phone is used to implement any method on the control terminal side in the foregoing method embodiments. Specifically, the indoor positioning device 30 for an Internet of Things smart phone includes: A sending module 33, configured to, in response to a positioning operation of a user, send a positioning request to the positioning server by the control terminal, where the positioning request includes an identifier of a mobile phone of a user to be positioned. A receiving module 31, configured to receive a target location returned by the positioning server. A display module 32, configured to display the target location on a graphical user interface.

[0116] Optionally, the receiving module 31 is further configured to receive a trajectory route and a moving direction sent by the positioning server. The display module 32 is further configured to display the trajectory route and the moving direction on the graphical user interface in real time, and mark the current target location of the mobile phone on the trajectory route.

[0117] The indoor positioning device for an Internet of Things smart phone provided in any of the foregoing embodiments is used to execute the technical solutions in the foregoing method embodiments, and the implementation principles and technical effects are similar, which will not be elaborated herein.

[0118] Figure 9 FIG. is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 9 shown, the electronic device 400 includes: a processor 401, a memory 402, and a communication interface 403. The memory 402 is used to store a computer program, and the processor 401 is used to execute the computer program to implement the technical solutions in any of the foregoing method embodiments.

[0119] In actual implementation, the electronic device 400 may be implemented as any one of a positioning server, a control terminal, and an access device, and this solution is not limited thereto.

[0120] This embodiment provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the microscopic measurement data processing method in the above embodiment.

[0121] This embodiment also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the microscopic measurement data processing method provided in any one of the above embodiments.

[0122] Those skilled in the art will readily conceive of other implementations of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present application are pointed out by the following claims.

[0123] It should be understood that the present application is not limited to the exact structures already described and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A method for indoor positioning of an Internet of Things smartphone, characterized in that: Applied to an indoor positioning system based on the Internet of Things, the indoor positioning system includes a plurality of access devices distributed in a target location, a positioning server and a control terminal, and the method includes: In response to the user's positioning operation, the control terminal sends a positioning request to the positioning server, wherein the positioning request includes an identifier of the mobile phone of the user to be positioned; The positioning server receives a positioning request from the control terminal, and sends data acquisition requests to the multiple access devices respectively according to the positioning request, wherein the data acquisition request includes an identifier of the mobile phone; After receiving the data acquisition request, each access device acquires the communication data of the mobile phone and returns the communication data to the positioning server, wherein the communication data includes the signal strength of the mobile phone detected at the current moment and the signal strength at multiple moments before the current moment; The positioning server predicts the position of the user to be positioned using a position prediction model according to the communication data sent by the multiple access devices, and obtains the target position of the mobile phone, wherein the position prediction model is a model obtained by training a recursive neural network model based on a signal fingerprint database, wherein the signal fingerprint database includes: the signal strength and acquisition time of each position of the test terminal in the indoor scene detected by multiple access devices set in the indoor scene within a historical time period; The positioning server returns the target location to the control terminal.

2. The method for indoor positioning of an IoT smartphone according to claim 1, characterized in that: The method further comprises: The positioning server receives a positioning model configuration message sent by the cloud server, wherein the positioning model configuration message includes the position prediction model; The positioning server configures the location prediction model in a positioning program based on the positioning model configuration message.

3. The method for indoor positioning of an IoT smartphone according to claim 1, characterized in that: The method further comprises: In the positioning server, in response to the user's operation, an initial long short-term memory network LSTM model is constructed using keras; Standardize the data in the signal fingerprint database to obtain a sample set, wherein the sample set includes multiple samples, each sample includes: an input signal matrix and a corresponding position, the input signal strength matrix includes multiple access device corresponding signal strength sequences, and the signal strength sequence corresponding to each access device includes the signal strength corresponding to any position of the test terminal in the test scene collected by the access device and the signal strength collected at least two times before the test terminal reaches the position; Based on the sample set, the LSTM model is trained to obtain the location prediction model.

4. The method for indoor positioning of an IoT smartphone according to claim 3, characterized in that: The step of training the LSTM model based on the sample set to obtain the location prediction model includes: Based on the samples in the sample set, the memory neural network of the LSTM model is optimized by using an adaptive particle swarm algorithm to obtain an LSTM model with optimized parameters; The LSTM model after parameter optimization is trained based on the sample set until the preset loss function converges to obtain the position prediction model.

5. The method for indoor positioning of an IoT smartphone according to claim 4, characterized in that: The method of optimizing the parameters of the memory neural network of the LSTM model based on the samples in the sample set by using an adaptive particle swarm algorithm to obtain the LSTM model after parameter optimization includes: Initializing a particle swarm of the adaptive particle swarm algorithm to obtain an initial particle swarm, wherein the initial particle swarm includes a plurality of particles, each particle being a parameter combination of the LSTM model; Determine the objective function of the LSTM model as the fitness function of the adaptive particle swarm algorithm, and the objective function is an accuracy evaluation function; The parameter combination corresponding to each particle is configured in the LSTM model, and the samples in the sample set are input into the LSTM model for prediction, and the fitness of each particle is calculated according to the predicted value and actual position of the sample using the fitness function to obtain the fitness value of each particle; The speed and position of each particle are updated according to the fitness value of each particle, and the fitness of each updated particle is recalculated until the preset number of iterations is reached to obtain the target particle group; The parameter combination corresponding to the particle with the largest fitness value in the target particle group is configured as the optimal parameter combination into the LSTM model to complete parameter optimization.

6. The method for indoor positioning of an IoT smartphone according to claim 5, characterized in that: The optimal parameter combination includes the selected optimal number of units, number of Dense layer units, learning rate and dropout probability.

7. The method for indoor positioning of an IoT smartphone according to any one of claims 1 to 6, characterized in that: Before the positioning server returns the target location to the control terminal, the method further includes: The positioning server optimizes the target position using a filtering algorithm to obtain an optimized new target position; Correspondingly, the positioning server returns the target location to the control terminal, including: The positioning server returns the new target location to the control terminal.

8. The method for indoor positioning of an IoT smartphone according to claim 7, characterized in that: The method further comprises: The positioning server generates the trajectory route and movement direction of the mobile phone in real time according to the acquired target positions of the mobile phone at multiple times; The positioning server returns the trajectory route and the moving direction to the control terminal.

9. The method for indoor positioning of an IoT smartphone according to claim 8, characterized in that: The method further comprises: After receiving the trajectory route and the moving direction, the control terminal displays the trajectory route and the moving direction in real time on a graphical user interface, and marks the current target position of the mobile phone on the trajectory route.

10. An indoor positioning system, characterized in that: include: Multiple access devices, positioning servers and control terminals distributed in the target location; The control terminal is used to send a positioning request to the positioning server in response to the user's positioning operation, and the positioning request includes the identifier of the mobile phone of the user to be positioned; The positioning server is used to receive a positioning request from the control terminal, and send data acquisition requests to the multiple access devices respectively according to the positioning request, wherein the data acquisition request includes the identification of the mobile phone; Each access device is used to obtain the communication data of the mobile phone according to the received data acquisition request and return the communication data to the positioning server, wherein the communication data includes the detected signal strength of the mobile phone at the current moment and the signal strength at multiple moments before the current moment; The positioning server is further used to predict the position of the user to be positioned using a position prediction model according to the communication data sent by the multiple access devices, and obtain the target position of the mobile phone, wherein the position prediction model is a model obtained by training a recursive neural network model based on a signal fingerprint database, wherein the signal fingerprint database includes: the signal strength and acquisition time of each position of the test terminal in the indoor scene detected by multiple access devices set in the indoor scene within a historical time period; The positioning server is also used to return the target location to the control terminal.

Images