A Low-Cost Airport Location Method Based on Mobile Terminals

Abstract

A low-cost positioning method for airports based on mobile terminals. In the positioning information acquisition stage, after using a mobile robotic cart based on laser SLAM to map the terminal building environment, a mobile terminal device is used to collect multi-source data for reference points in each divided area, obtaining the original fingerprint data fingerprint library Γ. In the positioning model establishment stage, first, the original data fingerprint library Γ is preprocessed, then multi-feature adaptive clustering is carried out, and finally, a prediction model is constructed. In the positioning model update stage, a small amount of data is collected for the terminal building area with large environmental changes, and an error correction model is constructed to maintain the high accuracy of the positioning result. The present invention utilizes the existing WIFI devices in the airport and the magnetic field information existing in nature, greatly reducing the hardware cost. At the same time, under the condition of meeting the positioning accuracy, this method omits the cumbersome process of constructing and updating the fingerprint library through manual collection in the traditional wireless positioning scheme, reducing a large amount of time and manpower consumption.

Description

Technical Field

[0001] The present invention belongs to the field of indoor positioning and navigation, and particularly relates to a low-cost positioning method for airports based on mobile terminals. Background Art

[0002] As an important place for transporting passengers, the construction of a smart airport must be centered around passengers, which is an inevitable trend in the contemporary civil aviation development of "people-oriented". Through the refined, collaborative, visualized, and intelligent operation and management of operations such as airport operations, flight guarantee, and commercial promotion, the safety, efficiency, convenience, and comfort of passengers during the travel process are ensured. Building a smart airport using new-generation information technologies such as the Internet of Things, big data, cloud computing, and mobile Internet is necessary for building China's intelligent comprehensive transportation management, for building the intelligent operation management of China's civil aviation, and is also the only way for China to transform from a major civil aviation country to a strong civil aviation country. Therefore, airports need to make full use of positioning and tracking technologies to perform real-time positioning and tracking on personnel and non-motorized ground service equipment, so as to facilitate operators to efficiently manage airport infrastructure, realize the analysis and early warning of civil aviation service quality, airport traffic, and project bidding, and provide decision-making support for civil aviation policies and industrial development.

[0003] Currently, with the rapid development of WIFI technology, WIFI devices have been installed in many large public buildings including airports. Making full use of WIFI positioning devices can not only meet the positioning requirements of airports but also reduce the cost of building the positioning system. At the same time, environmental magnetic field signals have also received extensive attention due to their ubiquitous and non-extra-setup characteristics. Currently, most WIFI positioning technologies and magnetic field positioning technologies adopt fingerprint positioning methods. The fingerprint positioning technology mainly relies on the fact that different spatial points in the surrounding environment have completely different multi-source information, and each spatial position has its own unique characteristics similar to the "fingerprint" of that point. By collecting and storing the multi-source signals of each of these position spatial points, a data fingerprint library for the area to be positioned is constructed. During online positioning, the collected multi-source information is matched with the data in the fingerprint library to obtain the current location, achieving the purpose of positioning.

[0004] However, problems such as large WIFI signal fading, severe multipath interference, and complex non-line-of-sight scenarios in the terminal building seriously affect the accuracy of fingerprint positioning. At the same time, due to problems such as the movement of the crowd and the appearance and disappearance of obstacles, the timeliness of the fingerprint library in the airport environment is low. The conventional method is to re-collect data for the fingerprint library, but this will consume a large amount of manpower and time and the effect is not significant enough. Therefore, how to maximize the accuracy of fingerprint positioning and reduce the problem of fingerprint library failure caused by environmental factors has become a research hotspot in airport positioning systems in recent years.

[0005] To solve the above problems, the prior art is as follows:

[0006] Application No.: CN 201810520528.X, Patent Name: Positioning Method, Positioning Device and Data Processing System Based on Collocation. This patent discloses a positioning method, a positioning device and a data processing system based on collocation. The method includes: collecting several different types of signals through an intelligent mobile terminal to form a random field signal composed of multiple signals; dividing the random field signal into a trend part and a random part; fitting the trend part with a preset function and fitting the random part with a covariance function; and restoring the accurate position of the intelligent terminal according to the fitting result. This method can provide users with sufficient and accurate personal positioning information, which can be applied to various industries and products based on location services, such as elderly care services and the location safety of children. It effectively improves the position restoration accuracy and can reduce the resource consumption of the server. Finally, the multi-dimensional information provided is applicable to the industrial applications of various positioning services.

[0007] It divides the random field signal into a trend part and a random part, and respectively uses a preset function and a covariance function for fitting, and finally restores the accurate position of the intelligent terminal, solving the problems of large resource consumption and low positioning accuracy of the positioning server. However, this method simply uses a preset function and least squares for fitting, and the accuracy needs to be further improved.

[0008] This application uses the currently widely popularized smart phone as the mobile terminal, comprehensively utilizes the existing WIFI devices and the magnetic field information existing in nature, and has a low cost. A deep learning network is used to establish different positioning models for fitting according to the clustering results, and the positioning prediction result has a high accuracy. In addition, this paper constructs a map of the positioning environment to realize automatic data collection, greatly reducing the problem of large workload in database construction.

[0009] Application No.: CN201810102422.8, Patent Title: A Method, Storage Medium and Intelligent Terminal for Indoor Positioning Based on Multi-source Data Fusion. This patent discloses a method, storage medium and intelligent terminal for indoor positioning based on multi-source data fusion. The method includes the steps of: constructing a conditional random field model that fuses multi-source data according to the feature function of the single-mode perception positioning result; implementing adaptive learning of model parameters by using the method assisted by indoor map to obtain real-time online model parameters; and inferring and positioning the indoor position of the user through the conditional random field model and the real-time online model parameters. By constructing a conditional random field model that can fuse multi-source heterogeneous data, the dependence on single-mode data is reduced, and the accuracy of indoor positioning based on smart phones is improved. At the same time, the real-time model parameters are obtained by using the indoor map matching method, and the adaptive model parameter adjustment is realized, which enhances the robustness and generality of the positioning algorithm. In addition, the method of behavior map matching is used in the indoor map matching, which improves the accuracy and efficiency of the matching.

[0010] It constructs a conditional random field model that fuses multi-source data by combining single-mode feature functions, and uses the method assisted by map to realize the adaptive learning of parameters, which enhances the robustness and generality of the positioning algorithm. However, the positioning result of this invention is greatly affected by pedestrians, and its running stability needs to be verified in complex environments.

[0011] The low-cost positioning method of this application comprehensively utilizes the existing WIFI devices and the magnetic field information existing in nature to construct a multi-feature adaptive clustering model to obtain a rough classification model of pedestrians, and then constructs a fine position prediction network for mobile devices. The adopted "coarse-fine" double-layer positioning model enhances the prediction accuracy on the basis of improving the prediction speed. At the same time, the error correction module can maintain the high accuracy of the positioning result for a long time. This invention can effectively solve the problems of dynamic airport environment, low positioning accuracy, low efficiency and high cost in complex environments. Summary of the Invention

[0012] The technical problem to be solved by this invention is to propose a low-cost airport positioning method based on mobile terminals to solve a series of problems in traditional airport positioning, such as high cost of building a positioning system, low positioning accuracy, easy invalidation of fingerprint database with environmental changes, and high cost caused by repeated collection consuming a large amount of human and material resources. This invention uses the currently widely popular smart phone as the mobile terminal, and utilizes the existing WIFI devices in the airport and the magnetic field information existing in nature, which greatly reduces the hardware cost. At the same time, under the condition of meeting the positioning accuracy, this method omits the cumbersome process of constructing and updating the fingerprint database through manual collection in the traditional wireless positioning scheme, and reduces the consumption of a large amount of time and manpower.

[0013] To achieve the above object, the technical solution adopted by this invention is:

[0014] A low-cost positioning method for airports based on mobile terminals, the specific steps are as follows:

[0015] Step 1, positioning information acquisition: The positioning information acquisition process includes using a mobile robotic cart based on laser SLAM to map the terminal building environment; dividing the indoor area of the terminal building, and using the mobile robotic cart to traverse the environmental reference points; collecting multi-source positioning information at each reference point to construct an original fingerprint database;

[0016] Step 2, positioning model establishment: The positioning model establishment process includes preprocessing the original data fingerprint library Γ, and screening out the fingerprint library that can be used for model establishment; performing multi-feature adaptive clustering on the processed fingerprint library to reduce the workload of data matching during online positioning and increase the accuracy of the positioning result; training the data of each clustering subset to obtain the final coordinate prediction model;

[0017] Step 3, positioning model update: The positioning model update process refers to when a major change occurs in the environment of a certain area of the terminal building due to layout or other reasons, using a mobile terminal to collect a small amount of data points to obtain a signal set Υ, inputting it into the coordinate prediction model of this area and comparing it with the real coordinate result to obtain an error data set Δ; then using the collected signal set Υ and error data set Δ together as a calibration data set and training it with an integrated evaluator to obtain an error correction model for correcting the positioning result.

[0018] As a further improvement of the present invention, the specific steps of step 1 are as follows;

[0019] Step 1.1, using a mobile robotic cart based on laser SLAM to map the entire terminal building environment to obtain an indoor map of the entire terminal building;

[0020] Step 1.2, dividing the indoor area of the terminal building into Ω k , k ∈ {1, 2,..., K}, dividing the area Ω k into N k meshes, and taking the geometric center of each mesh as a reference point where n ∈ {1, 2,..., N k}; autonomously planning the path of the mobile robotic cart to traverse each divided area, and simultaneously collecting multi-source positioning data at each reference point during the movement;

[0021] Step 1.3, aligning the position information and the collected multi-source positioning information through the positioning timestamp to obtain the original multi-source data fingerprint library Γ in the target environment.

[0022] As a further improvement of the present invention, the specific steps of step 2 are as follows;

[0023] Step 2.1: Preprocess the original multi-source data fingerprint database Γ to filter out the data fingerprint database that can be used for model establishment;

[0024] Step 2.2: Extract multi-dimensional features from the processed data fingerprint database for multi-feature adaptive k-means clustering;

[0025] Step 2.3: Train the model for each partitioned regional data to obtain the final coordinate prediction model.

[0026] As a further improvement of the present invention, the specific steps of Step 2.1 are as follows. First, at each reference point, calculate the mean μ and standard deviation σ for each dimension of WIFI data and magnetic field data respectively, and use the 3σ criterion to filter out gross errors. Secondly, perform Gaussian smoothing filtering on the processed fingerprint database data. Finally, in order to reduce the data fluctuation caused by periodic interference during measurement, use sliding window filtering for each dimension of data and replace the data within the window with the mean value.

[0027] As a further improvement of the present invention, the specific steps of Step 2.2 are as follows. First, select the initial points according to the regions divided by the terminal building, determine the initial clustering centers according to the relative coordinate features in the fingerprint database data, and select the center point position of each region as the initial clustering center. Then, perform multi-feature adaptive clustering according to the WIFI signal strength feature and magnetic field strength feature of the collected fingerprint database: set the initial dynamic adaptive weight μ to 1 / 2, calculate the Euclidean distance from each sample to each initial clustering center, and classify all samples according to the principle of assigning labels according to the minimum distance. Then, for each class c i , recalculate the centroid value of this cluster. The objective function of multi-feature adaptive clustering is:

[0028]

[0029] In the formula: μ is the adaptive weight, and the weight value of the centers calculated from the two perspectives is adaptively adjusted according to this weight, is the vector of sample x j under the WIFI perspective; is the vector of sample x j under the magnetic force information perspective; is the clustering center vector of the data under the WIFI perspective, is the clustering center vector of the data under the magnetic force information perspective;

[0030] Among them, the clustering center is updated by the following formula:

[0031]

[0032] Its dynamic adaptive weight takes μ as a parameter in a clustering process and determines its optimal value u through cross-validation. opt .

[0033] As a further improvement of the present invention, step 3 is specifically as follows;

[0034] Step 3.1, when a major change occurs in the environment of a certain area of the terminal due to layout or the like, use a mobile terminal to collect a small number of data points in this area to obtain a signal set Υ;

[0035] Step 3.2, input the collected signal data set Υ into the coordinate prediction model of this area to obtain a prediction result, and compare the prediction result with the true coordinate result of the collection point to obtain an error data set Δ;

[0036] Step 3.3, use the collected signal set Υ and the error data set Δ together as a calibration data set and train them using an integrated evaluator to obtain their non-linear mapping relationship as an error correction model for correcting the positioning result.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] After adopting the above technical means, the present invention can obtain the following advantages: Due to adopting this technical solution, the present invention uses a mobile robot control module to control the robot to move along a specified path, and at the same time collects multi-source information in the target environment, with high efficiency and good effect, without manual participation, greatly saving human resources; use the data processing module to remove gross errors, smooth filter, and segment compress the multi-source information, and at the same time perform multi-feature extraction on each multi-source information, use the multi-feature adaptive clustering method to classify the original data fingerprint library, and then model each classified sub-fingerprint library, greatly reducing the computational workload in the online positioning stage, saving computing resources while improving the positioning accuracy; when a major change occurs in the environment of a certain area, collect multi-source information of a small number of points for this environment, process it and establish an error correction model to correct the error of the original prediction model, which can maximize the positioning accuracy of the prediction model to maintain at an excellent level, without rebuilding the library, greatly saving labor and material costs. Facing a dynamic and complex environment, compared with the background technology method, this technology has high positioning accuracy, high efficiency, and low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flowchart of an airport low-cost positioning method based on a mobile terminal;

[0040] Figure 2 is a flowchart of a multi-feature adaptive clustering method;

[0041] Figure 3 is a flowchart of a positioning model update method. Specific Embodiments

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts belong to the scope of protection of the present invention. Figure 1 is a flowchart of an airport low-cost positioning method based on a mobile terminal, Figure 2 is a flowchart of a multi-feature adaptive clustering method; Figure 3 is a flowchart of a positioning model update method.

[0043] An airport low-cost positioning method based on a mobile terminal includes three stages: positioning information acquisition, positioning model establishment, and positioning model update; among them, in the positioning information acquisition stage, after using a mobile robotic cart based on laser SLAM to map the terminal building environment, a mobile terminal device is used to collect multi-source data of reference points in each divided area to obtain an original fingerprint data fingerprint library Γ; in the positioning model establishment stage, the original data fingerprint library Γ is first preprocessed, then multi-feature adaptive clustering is performed, and finally a prediction model is constructed; in the positioning model update stage, a small amount of data is collected for the terminal building area with large environmental changes, and an error correction model is constructed to maintain the high accuracy of the positioning result.

[0044] As Figure 1 shown, the specific steps are as follows:

[0045] (1) Positioning information acquisition: The positioning information acquisition process includes using a mobile robotic cart based on laser SLAM to map the terminal building environment; dividing the indoor area of the terminal building, and using the mobile robotic cart to traverse the environmental reference points; collecting multi-source positioning information at each reference point to construct the original fingerprint database Γ;

[0046] (2) Positioning model establishment: The positioning model establishment process includes preprocessing the original data fingerprint library Γ to screen out the fingerprint library that can be used for model establishment; performing multi-feature adaptive clustering on the processed fingerprint library to reduce the workload of data matching during online positioning and increase the accuracy of the positioning result; training the model for each cluster data to obtain the final coordinate prediction model;

[0047] (3) Location model update: The location model update process refers to the situation where when a major change occurs in the environment of a certain area of the terminal building due to layout or other factors, a small number of data points are collected using a mobile terminal to obtain a signal set Υ. After inputting it into the coordinate prediction model of this area and comparing it with the real coordinate results, an error data set Δ is obtained. Then, the collected signal set Υ and the error data set Δ are jointly used as a calibration data set and trained using an integrated evaluator to obtain an error correction model for correcting the location results.

[0048] The steps for obtaining the location information are as follows:

[0049] a. Use a mobile robotic cart based on laser SLAM to map the entire terminal building environment to obtain an indoor map of the entire terminal building. When the mobile robot moves in each area of the terminal building, use a lidar to collect the point cloud information of the entire area. By matching and comparing two pieces of point cloud at different times, calculate the distance and attitude changes of the relative movement of the mobile robot to complete the positioning of the robot itself. By using a mobile robotic cart, an indoor map of the airport terminal building can be obtained conveniently and quickly, providing good prior information for subsequent area division and data collection.

[0050] b. Divide the indoor area of the terminal building into Ω k , k ∈ {1, 2,..., K}, divide the area Ω k into N k meshes, and take the geometric center of each mesh as a reference point where n ∈ {1, 2,..., N k}; Autonomously plan the path of the mobile robotic cart to traverse each divided area, and at the same time, collect multi-source location data at each reference point in real time during the movement: At collect the information data from M k APs, including the BSSID name of the AP and the WIFI signal strength, and the RSS sample value is where represents the RSS sample value from the m-th AP collected at , m ∈ {1, 2,..., M k}; Collect the geomagnetic information at each reference point in the form of , corresponding to the north, east, and vertical intensities of the geomagnetic field respectively; Collect the position coordinates of the mobile cart output in real time at each reference point

[0051]

[0052] c. Align the location information and the collected multi-source location information through the positioning timestamp to obtain the original multi-source data fingerprint database Γ in the target environment. Using a mobile robotic cart to automatically collect the fingerprint database can greatly reduce the cost of manual collection and improve scalability.

[0053] The steps for establishing the positioning model are as follows:

[0054] a. Preprocess the original multi-source data fingerprint database Γ to screen out the data fingerprint database available for model establishment. First, at each reference point, calculate the mean μ and standard deviation σ for each dimension of WIFI data and magnetic field data respectively, and use the 3σ criterion to filter out gross errors. Second, perform Gaussian smoothing filtering on the processed fingerprint database data. Finally, in order to reduce the data fluctuations caused by periodic interference during measurement, sliding window filtering is used for each dimension of data, and the mean value is used to replace the data within the window. For example, when collecting data, the occlusion caused by pedestrians walking may increase the data fluctuations. The occlusion time is generally no more than 10s, and the collection frequency is 1HZ, so the window can be set to 10. Sliding window filtering can retain the information fluctuations obtained to the greatest extent and compress the size of the database to the greatest extent, reducing the complexity of subsequent learning. Obtain the representative value of the signal strength of each reference point from each AP location point, and save it together with the location coordinates of this reference point as the fingerprint database for online matching;

[0055] Extract multi-dimensional features from the processed data fingerprint database for multi-feature adaptive k-means clustering. Since the collected data contains multi-source information, we perform feature classification, such as Figure 2 As shown, three key feature information are obtained: WIIF signal strength, magnetic field signal strength, relative position coordinates; perform multi-feature adaptive k-means clustering on the fingerprint database according to these key features to reduce the workload of data matching during online positioning. Specifically, first, select the initial points according to the areas divided by the terminal building, determine the initial clustering centers according to the relative coordinate features in the fingerprint database data, and select the center point position of each area as the initial clustering center; then, perform multi-feature adaptive clustering on the collected fingerprint database according to the WIFI signal strength feature and the magnetic field strength feature: set the initial dynamic adaptive weight μ to 1 / 2, calculate the Euclidean distance from each sample to each initial clustering center, and classify all samples according to the principle of assigning labels according to the minimum distance. Then, for each class c i , recalculate the centroid value of this cluster. The objective function of multi-feature adaptive clustering is:

[0056]

[0057] In the formula: μ is the adaptive weight, and the weight values of the centers calculated from the two perspectives are adaptively adjusted according to this weight. is the vector of sample x j from the perspective of WIFI; is the vector of sample x j from the perspective of magnetic information; is the cluster center vector of the data from the perspective of WIFI, is the cluster center vector of the data from the perspective of magnetic information;

[0058] Among them, the cluster center is updated by the following formula:

[0059]

[0060] Its dynamic adaptive weight regards μ as a parameter in a clustering process, and determines its optimal value u through cross-validation opt .

[0061] b. Perform model training on the data of each divided area. In this example, a one-dimensional CNN is used for model training to obtain the final coordinate prediction model; during online positioning, the multi-source data collected is put into the classifier. First, rough positioning is performed and assigned to the general airport area; then, the trained CNN model in this area is called for prediction to obtain the specific positioning coordinates of the passenger.

[0062] The update steps of the positioning model are as follows:

[0063] a. When a major change occurs in the environment in a certain area of the terminal building due to layout and other factors, such as Figure 3 shown, use the mobile terminal to collect a small number of data points in this area to obtain the signal set Υ;

[0064] b. Input the collected signal data set Υ into the coordinate prediction model of this area to obtain the prediction result, and compare the prediction result with the real coordinate result of the collection point to obtain the error data set Δ;

[0065] c. Use the collected signal set Υ and the error data set Δ together as the calibration data set and train it with an integrated evaluator such as Xgboost to obtain an error correction model for correcting the positioning result. When positioning again, the signal value collected by the passenger in real time is first input into the classification model for area classification and classified into the current area. Then, the signal data is input into the coordinate prediction model of this area to obtain its initial solution result. Then, the signal data is input into the integrated evaluator to obtain its error correction result. Finally, the error correction value is used to correct the initial solution result, and the real position of the airport passenger is obtained by combining the results of the two.

[0066] The above are only the preferred embodiments of the present invention, and are not any other form of limitation to the present invention. Any modification or equivalent change made according to the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A low-cost positioning method for airports based on mobile terminals, the specific steps are as follows, characterized in that: Step 1, positioning information acquisition: The positioning information acquisition process includes using a mobile robotic cart based on laser SLAM to map the terminal environment; dividing the indoor area of the terminal, and using the mobile robotic cart to traverse the environmental reference points; collecting multi-source positioning information at each reference point to construct an original fingerprint database; Step 2, positioning model establishment: The process of establishing the positioning model includes preprocessing the original data fingerprint library to screen out the fingerprint library that can be used for model establishment; performing multi-feature adaptive clustering on the processed fingerprint library to reduce the workload of data matching during online positioning and increase the accuracy of the positioning result; training models for each clustering subset data to obtain the final coordinate prediction model; The specific content of step 2 is as follows; Step 2.1, preprocess the original multi-source data fingerprint database to screen out the data fingerprint database that can be used for model establishment; Step 2.2, extract multi-dimensional features from the processed data fingerprint database for multi-feature adaptive k-means clustering; The specific steps of step 2.2 are as follows. First, initial points are selected according to the areas divided by the terminal building, and initial clustering centers are determined according to the relative coordinate features in the fingerprint database. The center point position of each area is selected as the initial clustering center. Then, multi-feature adaptive clustering is performed on the collected fingerprint database according to the WIFI signal strength feature and the magnetic field strength feature: set the initial dynamic adaptive weight to be 1 / 2, calculate the Euclidean distance from each sample to each initial clustering center, and classify all samples according to the principle of assigning labels based on the minimum distance. Then, for each class , recalculate the value of the cluster centroid. The objective function of multi-feature adaptive clustering is: ； In the formula: is the adaptive weight, and the weight value of the center calculated from the two perspectives is adaptively adjusted according to this weight. is the sample vector under the WIFI perspective of is the sample vector under the magnetic information perspective of is the clustering center vector of the data under the WIFI perspective, is the clustering center vector of the data under the magnetic information perspective; Among them, the cluster center is updated by the following formula: ； Its dynamic adaptive weight will be regarded as a parameter in a clustering process and its optimal value will be determined through cross-validation ; Step 2.3, perform model training on the data of each divided area to obtain a final coordinate prediction model; Step 3, positioning model update: The positioning model update process refers to the situation where when there are significant changes in the environment of a certain area of the terminal building due to layout effects, a small number of data points are collected using a mobile terminal to obtain a signal set , which is input into the coordinate prediction model of this area and compared with the true coordinate results to obtain an error data set ; then the collected signal set and the error data set are jointly used as a calibration data set and trained using an integrated evaluator to obtain an error correction model for correcting the positioning results.

2. A low-cost positioning method for airports based on mobile terminals according to claim 1, characterized in that: The specific content of step 1 is as follows; Step 1.1, use a mobile robotic cart based on laser SLAM to map the entire terminal environment to obtain an indoor map of the entire terminal; Step 1.2, divide the indoor area of the terminal building , and divide the area into meshes, and take the geometric center of each mesh as a reference point , where ; autonomously plan the path of the mobile robot to traverse each divided area, and simultaneously collect multi-source positioning data at each reference point in real time during the movement; Step 1.

3. Align the position information and the collected multi-source positioning information through the positioning timestamp to obtain the original multi-source data fingerprint library in the target environment .

3. A low-cost positioning method for airports based on mobile terminals according to claim 2, characterized in that: The specific steps of step 2.1 are as follows. First, at each reference point, calculate the mean value of each dimension of WIFI data and magnetic field data respectively and the standard deviation , and use the 3σ criterion to filter out gross errors; secondly, perform Gaussian smoothing filtering on the processed fingerprint database data; finally, in order to reduce the data fluctuation caused by periodic interference during measurement, use sliding window filtering for each dimension of data, and replace the data within the window with the mean value.

4. A low-cost positioning method for airports based on mobile terminals according to claim 1, characterized in that: The specific content of step 3 is as follows; Step 3.1, when a significant environmental change occurs in a certain area of the terminal building due to layout influence, use a mobile terminal to collect a small number of data points in this area to obtain a signal set ; Step 3.2, input the collected signal data set into the coordinate prediction model of this area to obtain a prediction result, and compare the prediction result with the true coordinate result of the collection point to obtain an error data set ; Step 3.3, using the collected signal set and the error data set together as the calibration data set, training with an integrated evaluator to obtain the non-linear mapping relationship between the two as the error correction model for correcting the positioning result.

Images