CN112866900B - Fine-grained WiFi fingerprint map real-time construction method based on crowdsourcing data - Google Patents
Fine-grained WiFi fingerprint map real-time construction method based on crowdsourcing data Download PDFInfo
- Publication number
- CN112866900B CN112866900B CN202110019264.1A CN202110019264A CN112866900B CN 112866900 B CN112866900 B CN 112866900B CN 202110019264 A CN202110019264 A CN 202110019264A CN 112866900 B CN112866900 B CN 112866900B
- Authority
- CN
- China
- Prior art keywords
- data
- fingerprint map
- wifi
- fine
- grained
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The invention discloses a method for constructing a fine-grained WiFi fingerprint map in real time based on crowdsourcing data. According to the method, based on sparse and noisy position information collected by crowdsourcing mobile equipment and sample information of a WiFi list obtained through scanning, a fine-grained WiFi fingerprint map is finally obtained through data acquisition, data cleaning, wiFi fingerprint map pair generation and fine-grained WiFi fingerprint map generation network construction and training, and crowdsourcing data containing position noise can be fully utilized for further practical application work such as positioning. Compared with the traditional fingerprint map construction method, the method firstly reduces labor cost, can construct a real-time updated fine-grained WiFi fingerprint map after certain data are accumulated, and meanwhile, compared with the traditional crowdsourcing map construction method, the method makes full use of data containing noise, greatly improves data utilization efficiency, solves a difficulty in crowdsourcing data use, and can ensure enough precision for subsequent positioning and other applications.
Description
Technical Field
The invention relates to a WiFi positioning technology, in particular to a method for constructing a fine-grained WiFi fingerprint map in real time based on crowdsourcing data.
Background
Currently, the mainstream positioning system is a GPS positioning system, but there are many limitations in using GPS, for example, in an outdoor shade road or an indoor environment, signals are blocked, and positioning accuracy is affected. Therefore, a WiFi positioning system is introduced to supplement a GPS positioning system, and better positioning effect can be achieved outdoors and indoors.
However, the most important problem encountered in WiFi fingerprint positioning is the construction and iterative update of a fingerprint map, which usually requires great labor cost to maintain the map and continuously acquire new data. This makes this positioning method encounter many obstacles in practical use. Therefore, a method which saves cost, can update the map in real time and meets the precision requirement is needed to be found, so that the WiFi positioning can be better applied.
At present, crowdsourcing is a hot topic, but crowdsourced data is often accompanied by inaccuracy, and the update of a WiFi fingerprint map through crowdsourcing is also researched by people in the past, but the noise of the crowdsourcing is not well processed, so that most data is abandoned, and the effect is not good.
Disclosure of Invention
In order to solve the problems that a WiFi fingerprint map is calibrated through manual measurement and a crowdsourcing method comprises high noise, the invention provides a method for constructing a fine-grained WiFi fingerprint map in real time based on crowdsourcing data.
The purpose of the invention is realized by the following technical scheme: a crowdsourcing data-based real-time construction method for a fine-grained WiFi fingerprint map comprises the following steps:
(1) Data acquisition and cleaning: obtaining crowdsourced mobile device data in a target area, including location at time of data record generation, received signal strength to WiFi hotspotTime stamp and positioning accuracy; to the positioning accuracy not satisfying the set positioning accuracy threshold a o The data of the AP are filtered, the AP is cleaned, the mobile AP is deleted, and the N residual AP data after cleaning are stored.
(2) Generating a WiFi fingerprint map pair:
for an area of N APs, the WiFi fingerprint map of the area is analogized to an N-channel r × c pixel picture M. Setting a positioning accuracy threshold a according to the step (1) o Gridding the target area, segmenting all the data cleaned in the step (1) at certain time intervals, averaging the intensity values of the same AP in each grid in a time slice to obtain an original fingerprint map M of each time slice o 。
Meanwhile, high-precision data are selected from the data cleaned in the step (1) to obtain a fine-grained high-precision fingerprint map M in a corresponding time slice f Will M f As a label, with the original fingerprint map M o Together forming a WiFi fingerprint map pair. M o And M f Together as training data, a pair of maps, M, is generated in each time slice o For r x c pixel pictures, M f Typically 2r x 2c, 3r x 3c or 4r x 4c pixel pictures, i.e. high precision a f Is generally set as a o 1/4 to 1/2 times of the total weight of the composition.
(3) Constructing and training a fine-grained WiFi fingerprint map generation network:
constructing a fine-grained WiFi fingerprint map generation network, wherein the network comprises a three-dimensional convolution layer, a residual error feature extraction module and a pixel recombination module; original fingerprint map M sliced with T time o Inputting the three-dimensional convolution layer to obtain a low-layer characteristic diagram, inputting the low-layer characteristic diagram into a residual error characteristic extraction module, and inputting the low-layer characteristic diagram into a pixel recombination module after extracting space-time characteristics by the residual error characteristic extraction module; the Pixel recombination module comprises a pooling layer, a convolution layer, a BN layer and a Pixel buffer layer which are sequentially connected, and after the characteristics of T time slices are fused by the data through the pooling layer, the sum M is obtained through the convolution layer, the BN layer and the Pixel buffer layer f The feature map of the same pixel dimension (2r × 2c, 3r × 3c or 4r × 4c) is finally output through the convolution layer after the ReLU activation, and is recorded as
The original fingerprint map M which is obtained in the step (2) and contains the current time slice T and the near T-1 historical time slices o Fine-grained high-precision fingerprint map M of current time slice t f Training a fine-grained WiFi fingerprint map generation network as training data, and considering M in the training process f Introducing a mask matrix A when a large number of grids have data loss m For labelling M f Whether the original data of one AP exists in one grid or not is marked as 1 if the original data of one AP exists, and otherwise, the original data of one AP is marked as 0; the loss of the network is defined as:
(4) Generating a fine-grained high-precision WiFi fingerprint map: the original fingerprint map M which is obtained in the step (2) and contains the current time slice T and the near T-1 historical time slices o And (4) inputting the fine-grained WiFi fingerprint map generation network trained in the step (3) to obtain a complete high-precision fine WiFi fingerprint map.
Further, the data acquisition specifically includes: in a certain time period, acquiring crowdsourcing mobile equipment data in a target area, wherein the crowdsourcing mobile equipment data comprises position data and WiFi scanning data, distinguishing by using a unique Mac address of each WiFi hotspot, and constructing a WiFi hotspot data set W = { W } for N' total scanned WiFi hotspots 1 ,W 2 ,…,W N′ W, single WiFi hotspot data W i The corresponding WiFi hotspot is w i From all scans to w i A total of k mobile device records of i ＝{m i1 ,m i2 ,…,m ik In which each record m ij ＝(l ij ,a ij ,s ij ,t ij )，l ij ＝(x ij ,y ij ) For the position at the time of the generation of the piece of mobile device data record, a ij Positioning accuracy, s, for the strip of mobile device data records ij Pair w when generating for the strip of mobile device data records i Received signal strength of t ij To generate a time stamp of the time of recording.
Further, in the data acquisition step, if the scene is an outdoor scene, x ij Denotes longitude, y ij Represents latitude, a ij The GPS positioning accuracy of the data record of the mobile equipment; if it is an indoor scene, x ij 、y ij Position coordinates representing a floor of a building, a ij For the positioning accuracy set by the user.
Further, in the step (1), the AP is cleaned through a DBSCAN algorithm, the mobile AP is deleted, and the remaining N AP data after cleaning are stored in the database.
Further, the picture M in step (2) is represented as:
wherein p is ij N AP Signal Strength feature vectors [ rss ] representing the location 1 ,rss 2 ,…,rss N ]。
Further, in the step (2), the intensity values of the same AP in each grid are averaged in the time slice, and the grid with data missing is replaced with the minimum signal intensity value, so as to obtain the original fingerprint map M of each time slice o 。
Furthermore, the mobile equipment is positioned by combining a high-precision refined WiFi fingerprint map and a WiFi fingerprint positioning algorithm.
The invention has the technical effects and advantages that: the method utilizes crowdsourcing data containing larger noise to construct a real-time WiFi fingerprint map, solves the influence of overlarge noise, constructs a rough original WiFi fingerprint map through a large amount of original data, trains a deep learning network by combining a small amount of accurate data, learns the space-time correlation in the original data to obtain the high-resolution accurate WiFi fingerprint map, solves the problem that the accuracy of the WiFi fingerprint is reduced along with the time lapse, greatly saves the time and energy of off-line acquisition, does not need additional sensor assistance during map construction, and provides help for improving the WiFi positioning application scene.
Drawings
Fig. 1 is a schematic diagram of a method for constructing a crowdsourcing data-based fine-grained WiFi fingerprint map in real time according to an embodiment of the present invention;
fig. 2 is a block diagram of a fine-grained WiFi fingerprint map generation network structure provided by an embodiment of the present invention;
fig. 3 is a schematic diagram of a residual error feature extraction module in the fine-grained WiFi fingerprint map generation network according to the embodiment of the present invention;
FIG. 4 is a schematic diagram of a pixel recombination method according to an embodiment of the present invention;
fig. 5 shows the accumulated raw data, the raw WiFi fingerprint map, and the processed high-precision refined fingerprint map according to an embodiment of the present invention.
Detailed Description
The technical solution of the present invention is further described in detail with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, a method for constructing a fine-grained WiFi fingerprint map based on crowdsourcing data in real time provided by an embodiment of the present invention includes the following steps:
(1) Data acquisition and cleaning
In a certain time period, acquiring crowdsourcing mobile equipment data in a target area, wherein the crowdsourcing mobile equipment data comprises position data and WiFi scanning data, distinguishing by using a unique Mac address of each WiFi hotspot, and constructing a WiFi hotspot data set W = { W } for N' total scanned WiFi hotspots (APs) 1 ,W 2 ,…,W N′ W, single WiFi hotspot data W i The corresponding WiFi hotspot is w i From all scans to w i A total of k mobile device records of i ＝{m i1 ,m i2 ,…,m ik In which each record m ij ＝(l ij ,a ij ,s ij ,t ij )，l ij ＝(x ij ,y ij ) For the position at which the piece of mobile equipment data record was generated, a ij Location of data record for the strip of mobile devicePrecision, s ij Pair w when generating for the piece of mobile device data record i Received signal strength of t ij To generate a time stamp of the time of recording. If outdoor scene, x ij Denotes longitude, y ij Representing the latitude, a ij The GPS positioning accuracy of the data record for the mobile device; if it is an indoor scene, x ij 、y ij Position coordinates representing a floor of a building, a ij For the positioning accuracy set by the user.
To the positioning accuracy not meeting the set positioning accuracy threshold a o Filtering the data, cleaning the AP through algorithms such as DBSCAN and the like, deleting the mobile AP, and storing the N cleaned AP data into a database.
(2) WiFi fingerprint map pair generation
The WiFi signal strength of each position can be distinguished, all signal characteristics of the position are combined to be called position fingerprints, and the position fingerprint set of the whole area is a fingerprint map.
For a region of N APs, the WiFi fingerprint map of the region may be analogized to an N-channel r × c pixel picture M.
Wherein p is ij N AP Signal Strength feature vectors [ rss ] expressed as the location 1 ,rss 2 ,…,rss N ]。
Further processing the data obtained in the step (1), and setting a positioning accuracy threshold a according to the step (1) o Gridding the target area, wherein the side length of the grid is set to be 2a o And segmenting data according to a certain time interval, averaging the intensity value of the same AP in each grid in the time segment, replacing the grid with the minimum signal intensity value (usually-120 dB is adopted, and the actual situation is better met compared with 0) when the data is missing, and obtaining the original fingerprint map M of each time segment o This step uses all the data after the washing in step (1).
Meanwhile, high-precision data are selected from the data cleaned in the step (1) to obtain a fine-grained high-precision fingerprint map M in a corresponding time slice f (the positioning accuracy screening threshold is generally a o /2，a o A/3 or a o /4, corresponding trellis division side length of a o ，2a o A 3 and a o /2) mixing M f As a label, with the original fingerprint map M o Together forming a WiFi fingerprint map pair. Will M f The reason why the tag is not directly used as a positioning map is that under the condition of high-precision screening, most grids in the area have data loss, and the use requirement of a WiFi fingerprint positioning algorithm cannot be met. M o And M f Together as raw training data, a pair of maps, M, is generated within each time slice o For r x c pixel pictures, M f Typically 2r x 2c, 3r x 3c or 4r x 4c pixel pictures, i.e. high precision a f Is generally set as a o 1/4 to 1/2 times of the total weight of the composition.
(3) Constructing and training fine-grained WiFi fingerprint map generation network
Constructing a fine-grained WiFi fingerprint map generation network, as shown in FIG. 2, wherein the network comprises a three-dimensional convolution layer, a residual error feature extraction module and a pixel recombination module; original fingerprint map M sliced with T time o Inputting the three-dimensional convolution layer to obtain a low-layer characteristic diagram, inputting the low-layer characteristic diagram into a residual error characteristic extraction module, and inputting the low-layer characteristic diagram into a pixel recombination module after extracting space-time characteristics by the residual error characteristic extraction module; the Pixel recombination module comprises a pooling layer, a convolution layer, a BN layer and a Pixel buffer layer which are sequentially connected, and after the characteristics of T time slices are fused through the pooling layer, the data passes through the convolution layer, the BN layer and the Pixel buffer layer to obtain a sum M f The feature map of the same pixel dimension (2r × 2c, 3r × 3c or 4r × 4c) is finally output through the convolution layer after the ReLU activation, and is recorded as
The original fingerprint map M which is obtained in the step (2) and contains the current time slice T and the near T-1 historical time slices o Fine-grained high-precision fingerprint map M of current time slice t f As training data, training the fine-grained WiFi fingerprint map generation network, and considering M in the training process f Introducing a mask matrix A when a large number of grids have data loss m For labelling M f Whether the original data of one AP exists in one grid or not is marked as 1 if the original data of one AP exists, and otherwise, the original data of one AP is marked as 0; the loss of the network is defined as:
(4) Fine-grained high-precision WiFi fingerprint map generation
The original fingerprint map M which is obtained in the step (2) and contains the current time slice T and the near T-1 historical time slices o And (4) inputting the fine-grained WiFi fingerprint map generation network trained in the step (3) to obtain a complete high-precision fine WiFi fingerprint map.
(5) And positioning of the mobile equipment is realized by combining a high-precision refined WiFi fingerprint map and a WiFi fingerprint positioning algorithm (a KNN method can be adopted).
In one embodiment, the specific implementation steps for constructing and training the fine-grained WiFi fingerprint map generation network are as follows:
firstly, building a network model, and firstly converting an input T time-sliced N-channel original fingerprint map into a feature map X of 64N channels through a 3X 3 three-dimensional convolution layer and a ReLU activation function 1 。
This is followed by stacking 16 residual feature extraction modules as shown in fig. 3, which are mainly composed of two three-dimensional convolution layers and a ReLU activation function, which is described as: y is l ＝h(X l )+F(X l ,W l ) And X l+1 ＝f(Y l ) Wherein X is l And X l+1 Respectively representing the input and output of the ith residual feature extraction module, F is a residual function representing the learned residual, h represents identity mapping, and F is a ReLU activation function. Through this stacking of modules, learning features from the initial to the L-level can be obtained:
each residual error feature extraction module comprises a multilayer structure, specifically, two three-dimensional convolution layers are used in the module, compared with a common two-dimensional convolution, the residual error feature extraction module can learn the features of time dimension, and is suitable for a scene with a plurality of historical data accumulation.
After feature extraction, a fingerprint map higher than the original resolution is obtained through a pixel recombination module, specifically, T time slices of information are fused through an average pooling layer, then r × c feature maps of 4, 9 or 16 channels are obtained through a convolution layer, then values of multiple channels are arranged to various positions through a pixel shuffle operation as shown in fig. 4 to obtain high-resolution 2r × 2c, 3r × 3c or 4r × 4c feature maps, and finally, required 2r 2c, 3r × 3c or 4r × 4c high-precision fingerprint maps are obtained through transformation of the convolution layer.
After the network model is built, an original fingerprint map M containing a current time slice T and near T-1 historical time slices o Fine-grained high-precision fingerprint map M of current time slice t f And as training data, training the fine-grained WiFi fingerprint map generation network.
Fig. 5 is an accumulated raw data (multi-time accumulated data), an original WiFi fingerprint map, and a processed high-precision fine-grained fingerprint map provided by the embodiment of the present invention.
The embodiments described above are only a part of the embodiments of the present invention, and not all embodiments, and the specific embodiments described herein are only for explaining the present invention and do not limit the present invention, and all other embodiments obtained by those skilled in the art without making creative efforts based on the embodiments of the present invention belong to the protection scope of the present invention.
Claims (1)
1. A method for constructing a fine-grained WiFi fingerprint map in real time based on crowdsourcing data is characterized by comprising the following steps:
(1) Data acquisition and cleaning: obtaining crowdsourced mobile device data in a target area, including location at time of data record generation, received signal strength to WiFi hotspot, timestampAnd positioning accuracy; to the positioning accuracy not satisfying the set positioning accuracy threshold a o Filtering the data, cleaning the AP through a DBSCAN algorithm, deleting the mobile AP, and storing the N residual AP data after cleaning;
the data acquisition specifically comprises the following steps: in a certain time period, crowdsourcing mobile equipment data in a target area is obtained, the crowdsourcing mobile equipment data comprises position data and WiFi scanning data, unique Mac addresses of all WiFi hotspots are used as distinguishing, and a WiFi hotspot data set W = { W } is constructed for N' total scanned WiFi hotspots 1 ，W 2 ，...，W N′ W, single WiFi hotspot data W i The corresponding WiFi hotspot is w i From all scans to w i A total of k mobile device records of i ＝{m i1 ，m i2 ，...，m ik H, each record m ij ＝(l ij ，a ij ，s ij ，t ij )，l ij ＝(x ij ，y ij ) For the position at which the piece of mobile equipment data record was generated, a ij Positioning accuracy, s, for the strip of mobile device data records ij Pair w when generating for the strip of mobile device data records i Received signal strength of t ij To generate a time stamp of the record; if outdoor scene, x ij Denotes longitude, y ij Representing the latitude, a ij The GPS positioning accuracy of the data record of the mobile equipment; if it is an indoor scene, x ij 、y ij Position coordinates representing a floor of a building, a ij For the positioning accuracy set by the user;
(2) Generating a WiFi fingerprint map pair:
for the area of N APs, simulating the WiFi fingerprint map of the area into an N-channel r × c pixel picture M; picture M is represented as:
wherein p is ij N AP Signal Strength feature vectors [ rss ] representing the location 1 ，rss 2 ，...，rss]；
According to the positioning accuracy threshold a o Gridding the target area, segmenting all cleaned data at certain time intervals, averaging the intensity values of the same AP in each grid in the time slices, replacing the grid with the minimum signal intensity value when the data is missing, and obtaining the original fingerprint map M of each time slice o ；
Selecting high-precision data from the cleaned data to obtain a fine-grained high-precision fingerprint map M in a corresponding time slice f Will M f As a label, high precision a f Is set to a o 1/4 to 1/2 times of the total weight of the composition; and the original fingerprint map M o Jointly forming a WiFi fingerprint map pair;
(3) Constructing and training a fine-grained WiFi fingerprint map generation network:
constructing a fine-grained WiFi fingerprint map generation network, wherein the network comprises a three-dimensional convolution layer, a residual error feature extraction module and a pixel recombination module; original fingerprint map M sliced by T time o Inputting the three-dimensional convolution layer to obtain a low-layer characteristic diagram, inputting the low-layer characteristic diagram into a residual error characteristic extraction module, and inputting the low-layer characteristic diagram into a pixel recombination module after extracting space-time characteristics by the residual error characteristic extraction module; the Pixel recombination module comprises a pooling layer, a convolution layer, a BN layer and a Pixel buffer layer which are sequentially connected, and after the characteristics of T time slices are fused by the data through the pooling layer, the sum M is obtained through the convolution layer, the BN layer and the Pixel buffer layer f And (4) after the characteristic graph with the same pixel dimension is activated by the ReLU, finally outputting through the convolution layer, and recording as
The original fingerprint map M which is obtained in the step (2) and contains the current time slice T and the near T-1 historical time slices o Fine-grained high-precision fingerprint map M of current time slice t f As training data, training the fine-grained WiFi fingerprint map generation network, and introducing a masking matrix A in the training process m For labelling M f If there is raw data of an AP in a certain mesh, if so, it is determined that the AP exists in the meshLabeled 1, otherwise labeled 0; the loss of the network is defined as:
(4) Generating a fine-grained high-precision WiFi fingerprint map: the original fingerprint map M which is obtained in the step (2) and contains the current time slice T and the near T-1 historical time slices o Inputting the fine-grained WiFi fingerprint map generated network trained in the step (3) to obtain a complete high-precision fine WiFi fingerprint map; and positioning of the mobile equipment is realized by combining a high-precision refined WiFi fingerprint map and a WiFi fingerprint positioning algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110019264.1A CN112866900B (en) | 2021-01-07 | 2021-01-07 | Fine-grained WiFi fingerprint map real-time construction method based on crowdsourcing data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110019264.1A CN112866900B (en) | 2021-01-07 | 2021-01-07 | Fine-grained WiFi fingerprint map real-time construction method based on crowdsourcing data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112866900A CN112866900A (en) | 2021-05-28 |
| CN112866900B true CN112866900B (en) | 2023-03-31 |
Family
ID=76004956
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110019264.1A Active CN112866900B (en) | 2021-01-07 | 2021-01-07 | Fine-grained WiFi fingerprint map real-time construction method based on crowdsourcing data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112866900B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114095863B (en) * | 2021-10-18 | 2023-12-19 | 西安邮电大学 | WIFI fingerprint library updating method based on sparse automatic encoder |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106658422A (en) * | 2017-01-09 | 2017-05-10 | 深圳市智开科技有限公司 | Network side positioning method and network side positioning system for aiming at high-sparse WiFi data |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102638889B (en) * | 2012-03-21 | 2014-10-15 | 浙江大学 | Indoor wireless terminal positioning method based on Bayes compression sensing |
| CN103338509A (en) * | 2013-04-10 | 2013-10-02 | 南昌航空大学 | WSN (wireless sensor network) indoor positioning method based on hidden markov models |
| CN103916820B (en) * | 2014-03-31 | 2017-06-06 | 浙江大学 | Wireless indoor location method based on access point stability |
| CN105516929B (en) * | 2016-01-25 | 2019-08-27 | 图优化(北京)科技有限公司 | Establish method, indoor orientation method and the corresponding intrument of indoor map data |
| CN106714109B (en) * | 2017-01-12 | 2020-08-25 | 上海交通大学 | WiFi fingerprint database updating method based on crowdsourcing data |
| CN110012428B (en) * | 2019-05-22 | 2020-12-29 | 合肥工业大学 | Indoor positioning method based on WiFi |
-
2021
- 2021-01-07 CN CN202110019264.1A patent/CN112866900B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106658422A (en) * | 2017-01-09 | 2017-05-10 | 深圳市智开科技有限公司 | Network side positioning method and network side positioning system for aiming at high-sparse WiFi data |
Non-Patent Citations (1)
| Title |
|---|
| 结合CNN和WiFi指纹库的室内定位算法;曹建荣 等;《计算机系统应用》(第07期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112866900A (en) | 2021-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110245709B (en) | 3D point cloud data semantic segmentation method based on deep learning and self-attention | |
| CN109743683B (en) | Method for determining position of mobile phone user by adopting deep learning fusion network model | |
| EP2033140B1 (en) | Classifying image regions based on picture location | |
| CN105911518A (en) | Robot positioning method | |
| CN109598220B (en) | People counting method based on multi-input multi-scale convolution | |
| CN112949407B (en) | Remote sensing image building vectorization method based on deep learning and point set optimization | |
| CN109151750B (en) | LTE indoor positioning floor distinguishing method based on recurrent neural network model | |
| CN114973002A (en) | Improved YOLOv 5-based ear detection method | |
| CN111539453B (en) | Global ionized layer electron total content prediction method based on deep cycle neural network | |
| CN111797920B (en) | Remote sensing extraction method and system for depth network impervious surface with gate control feature fusion | |
| CN109145836A (en) | Ship target video detection method based on deep learning network and Kalman filtering | |
| CN112866900B (en) | Fine-grained WiFi fingerprint map real-time construction method based on crowdsourcing data | |
| CN111414954A (en) | Rock image retrieval method and system | |
| CN110598564A (en) | OpenStreetMap-based high-spatial-resolution remote sensing image transfer learning classification method | |
| CN111640116B (en) | Aerial photography graph building segmentation method and device based on deep convolutional residual error network | |
| CN115421158B (en) | Self-supervision learning solid-state laser radar three-dimensional semantic mapping method and device | |
| CN114912707A (en) | Air quality prediction system and method based on multi-mode fusion | |
| CN113469226A (en) | Street view image-based land utilization classification method and system | |
| CN111242028A (en) | Remote sensing image ground object segmentation method based on U-Net | |
| Andersson et al. | Combining street-level and aerial images for dengue incidence rate estimation | |
| CN113989296A (en) | Unmanned aerial vehicle wheat field remote sensing image segmentation method based on improved U-net network | |
| CN113034511A (en) | Rural building identification algorithm based on high-resolution remote sensing image and deep learning | |
| CN108462992B (en) | Indoor positioning method based on super-resolution reconstruction Wi-Fi fingerprint map | |
| CN115493596A (en) | Semantic map construction and navigation method for mobile robot | |
| CN112668615B (en) | Satellite cloud picture prediction method based on depth cross-scale extrapolation fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |