CN112580479A - Geomagnetic indoor positioning system based on cavity convolution neural network - Google Patents

Geomagnetic indoor positioning system based on cavity convolution neural network Download PDF

Info

Publication number
CN112580479A
CN112580479A CN202011466245.5A CN202011466245A CN112580479A CN 112580479 A CN112580479 A CN 112580479A CN 202011466245 A CN202011466245 A CN 202011466245A CN 112580479 A CN112580479 A CN 112580479A
Authority
CN
China
Prior art keywords
fingerprint
geomagnetic
positioning
sequence
neural network
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.)
Pending
Application number
CN202011466245.5A
Other languages
Chinese (zh)
Inventor
陈彦如
王伟
倪振心
陈良银
杨彦兵
刘诗佳
郭敏
张媛媛
胡顺仿
王浩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chengdu Yishuqiao Technology Co ltd
Original Assignee
Chengdu Yishuqiao Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chengdu Yishuqiao Technology Co ltd filed Critical Chengdu Yishuqiao Technology Co ltd
Priority to CN202011466245.5A priority Critical patent/CN112580479A/en
Publication of CN112580479A publication Critical patent/CN112580479A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/12—Fingerprints or palmprints
    • G06V40/13—Sensors therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/29—Geographical information databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/12—Fingerprints or palmprints
    • G06V40/1365—Matching; Classification
    • 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
    • H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
    • 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
    • H04W4/023—Services making use of location information using mutual or relative location information between multiple location based services [LBS] targets or of distance thresholds
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/33—Services specially adapted for particular environments, situations or purposes for indoor environments, e.g. buildings
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Databases & Information Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Multimedia (AREA)
  • Human Computer Interaction (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Remote Sensing (AREA)
  • Collating Specific Patterns (AREA)

Abstract

The invention discloses a geomagnetic indoor positioning system based on a cavity convolution neural network. According to the characteristic that the geomagnetic fingerprint is uniquely associated with the space position coordinate, the fingerprint segmentation length evaluation algorithm is designed, and the fingerprint segmentation length is calculated according to the abundance degree of the geomagnetic fingerprint in the indoor space. And an overlapped segmentation method is adopted to segment the fingerprint sequence, each segmented fingerprint is mapped to a position point, the density of the positioning points is increased, and the fingerprint characteristics are fully utilized. In order to match and identify the geomagnetic fingerprint more accurately, the invention designs a fingerprint classification model based on a cavity convolution neural network to distinguish each segmented fingerprint after converting the geomagnetic fingerprint sequence into a curve difference matrix. The void volume in the classification model can acquire more effective fingerprint information, and has higher fingerprint matching accuracy after a large amount of fingerprint data training. The geomagnetic indoor positioning system improves the positioning accuracy with less positioning cost.

Description

Geomagnetic indoor positioning system based on cavity convolution neural network
Technical Field
The invention belongs to the field of indoor positioning, and relates to a deep learning-based geomagnetic indoor positioning system and an implementation method thereof.
Background
In recent years, with the development of mobile internet technology, Location-Based Service (LBS) applications are emerging continuously, and the LBS applications penetrate into the aspects of people's clothes and people's residences, thereby greatly improving the convenience of people's daily life. A mature Global Positioning System (GPS) is used as a technical support of LBS, and can achieve a high Positioning accuracy in an outdoor environment, and a Positioning effect cannot meet a demand of people for an Indoor Location-Based Service (ILBS) due to signal blocking and multipath effects indoors. Therefore, indoor positioning technology has been rapidly developed, and various signals are used in the indoor positioning technology, such as Wi-Fi (Wireless-Fidelity), Bluetooth (Bluetooth), Radio Frequency Identification (RFID), infrared, Ultra Wide Band (UWB), terrestrial magnetism, visible light, ultrasonic waves, Frequency Modulation (FM), Inertial Navigation System (INS), images, and the like.
The indoor positioning scheme needs to consider the deployment and maintenance cost of the scheme while considering the positioning accuracy. The positioning precision of technologies such as RFID, ultrasonic wave and UWB can reach the meter level, but the deployment cost and the energy consumption of the equipment are relatively high, so that the positioning scheme is difficult to implement and is difficult to popularize and apply. The cost of the Wi-Fi and inertial navigation sensing technologies is low, but the positioning accuracy is not satisfactory. Such as Wi-Fi, have the effect of signal fluctuations and multipath effects and cannot be located effectively in a complex environment. In contrast, the geomagnetic field-based positioning technology not only has stable signals and high positioning accuracy, but also does not require additional equipment, and is gradually the focus of research.
The existing geomagnetic scheme also has a lot of defects, mainly including: 1) the fingerprint collection and the fingerprint database construction are time-consuming and labor-consuming: the existing research needs a worker to survey in a positioning place, manually acquire fingerprint data in a space, mark a position and establish a mapping relation between the fingerprint and the position. 2) The positioning accuracy is limited: most of the existing geomagnetic positioning schemes have low fingerprint matching accuracy, and the problem of sequence alignment during matching still exists, so that the ideal positioning accuracy can be achieved only by higher positioning and walking cost, and the positioning effect is not ideal under short distance. 3) The positioning time delay is longer: the traditional fingerprint similarity matching algorithm needs to traverse all fingerprints in a fingerprint database, and the algorithm is time-consuming; the particle filter algorithm has a large calculation amount, particles may be converged slowly, and a certain time delay exists during positioning.
Disclosure of Invention
The invention aims to solve the problems of higher fingerprint acquisition cost, limited positioning accuracy, higher positioning walking cost and higher positioning time delay of the existing geomagnetic indoor positioning algorithm, trains a fingerprint classification model by a deep learning method, establishes a mapping relation between a position coordinate point and a geomagnetic fingerprint and realizes a low-cost, efficient and accurate geomagnetic indoor positioning system.
The core technical idea of the invention is to realize positioning by matching and identifying geomagnetic fingerprints through a deep learning model, and the system comprises two stages of off-line training and on-line positioning. 1) An off-line training stage: collecting geomagnetic fingerprint information in space, training a geomagnetic fingerprint classification model after processing, and establishing a mapping relation between geomagnetic fingerprints and positions. 2) And (3) in an online positioning stage: after receiving a positioning request of a user, calculating a positioning result through a fingerprint classification model and a related algorithm and feeding back the positioning result to the user.
The core technology for solving the problems comprises a fingerprint sequence segmentation length evaluation algorithm, an overlapped fingerprint segmentation method, a curve interpolation matrix and a geomagnetic fingerprint classification model.
(1) And (4) a fingerprint sequence segmentation length evaluation algorithm. The invention realizes positioning by solving the problem of classification matching of the geomagnetic fingerprint sequence, wherein one section of fingerprint sequence corresponds to a section of main path segment, and the corresponding positioning point is the coordinate point of the main path segment end point. Although the longer the length of the geomagnetic fingerprint sequence is, the larger the amount of information expressed is, the higher the uniqueness is in a limited spatial range, the relationship between the length and the uniqueness is not a linear increase. In order to seek a balance point between the positioning accuracy (uniqueness) and the positioning walking cost (fingerprint sequence length), the invention provides a fingerprint sequence segment length evaluation algorithm.
(2) Overlapped fingerprint cutting method. Because the problem of misplacement of the head and the tail of the sequence exists during actual fingerprint sequence matching, and the larger fingerprint sequence segmentation length can cause higher alignment cost and larger error, the invention provides an overlapped segmented geomagnetic fingerprint acquisition algorithm based on the segmented geomagnetic fingerprint acquisition algorithm, and the subsequences are overlapped to a certain extent.
(3) And (5) curve interpolation matrix. The geomagnetic fingerprint sequence is a one-dimensional space-time sequence, equipment difference exists, absolute numerical values of the fingerprint sequence cannot be directly used for classification tasks, and only relative relation and existing high-dimensional space characteristics can be searched among sequence numerical values. In order to better describe the relative relation between sequence values, the invention provides a curve difference matrix, and absolute values are replaced by differences between sampling points to describe the shape of a curve formed by a fingerprint sequence, so that a neural network can better sense the high-dimensional spatial characteristics of data.
(4) Geomagnetic fingerprint classification model. The basic convolution neural network is composed of a convolution layer, a pooling layer and a full-connection layer, wherein the convolution layer utilizes convolution to check the image for feature extraction, the pooling layer plays a role in down-sampling, and finally, information obtained by combining different neurons through the full-connection layer is the feature extracted by the neural network to the whole image. The curve difference matrix of the geomagnetic fingerprint sequence is different from a general image, and each value of the matrix corresponds to the difference between two sampling points on the fingerprint sequence, and the difference is neither irrelevant nor redundant, but is effective information contributing to the output of the model. Therefore, in the feature extraction process of the curve difference matrix, the information is reduced or blurred regardless of the maximum pooling or the average pooling, and the method is not suitable for being used more frequently. On the other hand, pooling can increase the size of the receptive field, each convolution output contains larger range information, a convolution neural network without pooling or with little pooling can lose a certain receptive field and cannot sense the characteristics of a larger range of the fingerprint sequence. The feature extraction of the geomagnetic fingerprint sequence needs to avoid information loss and guarantee that information in different ranges and scales is obtained as much as possible, and the cavity convolution just meets the requirements.
Drawings
FIG. 1 is a structural diagram of a geomagnetic positioning system based on a hole convolution neural network;
FIG. 2 is a diagram of a floor plan partition according to the present invention;
FIG. 3 is a diagram illustrating an exemplary algorithm for estimating the segment length of a fingerprint sequence according to the present invention;
FIG. 4 is a diagram illustrating the segmentation of the overlapped geomagnetic fingerprint sequence according to the present invention;
FIG. 5 is a diagram of a fingerprint classification model according to the present invention;
FIG. 6 is a diagram of fingerprint matching misalignment correction according to the present invention;
Detailed Description
The present invention will be further described with reference to the accompanying drawings, wherein the system structure is shown in fig. 1, and mainly includes four parts, namely, geomagnetic fingerprint acquisition, geomagnetic fingerprint processing, geomagnetic fingerprint classification model training, and real-time position estimation. It should be noted that the description of the embodiments is only for the purpose of facilitating understanding of the present invention, and does not limit the present invention.
1. Geomagnetic fingerprint collection
Real-time magnetic field value (geo) acquired by geomagnetic field sensor built in mobile phonex,geoy,geoz) The mobile phone coordinate system is based on, the numerical value can change along with the change of the mobile phone posture, and the mobile phone coordinate system has no referential property. To avoid this problem, the geomagnetic field value based on the mobile phone coordinate system needs to be converted into the geomagnetic field value (g) based on the terrestrial coordinate system through the conversion matrixx,gy,gz). And the terrestrial magnetism value g of the terrestrial coordinate system in the east-west directionxAlmost 0, not being taken as the geomagnetic fingerprint, the final geomagnetic fingerprint of a single sample can be expressed as G ═ G (G)y,gz,gxyz) Wherein g isxyzModulo of the three-axis fingerprint vector. However, G cannot uniquely associate a position coordinate point in a certain space, and only multiple sampling values are collected into a geomagnetic fingerprint sequence S ═ G1,G2,…,GnThe unique mapping relationship between the geomagnetic fingerprint and the position point P can be guaranteed. In order to reduce the workload of acquiring geomagnetic fingerprint data, the invention uses an intelligent trolley carrying a mobile phone to acquire the geomagnetic fingerprint data. The main path of a floor is divided into four thickened lines as shown in fig. 2.
2. Geomagnetic fingerprint processing
An exemplary diagram of a fingerprint sequence segment length evaluation algorithm is shown in fig. 3. According to the algorithm, all geomagnetic fingerprint sequences are divided according to directions, the length (the number of segmented samples) of each segment is assumed to be SSN, and the SSN is traversed within a set maximum and minimum value range. N times fingerprint sampling sequence of total path in same direction { S1,S2,…,SnAnd taking the samples, and comparing the similarity between n segmentation fragments in the m groups of segmentation fragments transversely by adopting DTW (dynamic time warping) among the samples in each cycle. Through experimental statistics, the similarity between the same segments meets the chi-square distribution, the average value mu and the standard deviation sigma of the similarity values between all the segments are calculated, and the distinguishing threshold theta of the current segment is set to be mu + sigma. Then traversing each complete sequence SiEach sequence is cut according to the length of the current SSN and then labeled, and U is a label set of all segments of the sequence. And then, calculating similarity of m segments of the same sequence pairwise by adopting DTW (differential time warping), if the result is not greater than the previously calculated distinguishing threshold value, considering that the two segments are similar, and removing labels of the two segments from the label set. And finally, calculating and finding out the SSN value when the distinguishing rate converges and approaches to 1, namely the optimal length of the segments of the geomagnetic fingerprint sequence.
An overlapped type segmented geomagnetic fingerprint acquisition algorithm is adopted based on the segmented geomagnetic fingerprint acquisition algorithm, certain overlapping is carried out between the subsequences, and an overlapping proportion of 50% is selected for convenient calculation to carry out an experiment, as shown in fig. 4. The method can also more fully utilize the characteristics of the sequence on the basis of increasing the segmentation precision of the sequence.
The invention provides a curve difference matrix, which uses the difference between sampling points to replace absolute numerical values to describe the shape of a curve formed by a fingerprint sequence, so that a neural network can better sense the high-dimensional spatial characteristics of data. Compared with the normalization processing, the characteristic of the discrete degree between sequences is reserved. Suppose the segmented geomagnetic fingerprint sequence is Sxyz＝{g1,g2,…,gnThe curve difference matrix D is as follows:
Figure BDA0002832820670000041
after conversion, a two-dimensional image with richer contents is formed, different types of curves correspond to different curve difference value matrix diagrams, and the image category is clear. The horizontal and vertical coordinates of the matrix all represent the coordinates of sampling points of the sequence, and the difference value between the sampling points has high-dimensional characteristics in space, so that the condition of classifying by using an image processing mode is met, and the feature extraction can be performed by adopting a convolutional neural network to realize the classification of the geomagnetic fingerprint sequence.
3. Geomagnetic fingerprint classification model training
The invention provides a segmented fingerprint classification model based on a void convolution neural network, which has a model structure shown in fig. 5, wherein Conv represents a convolution layer, f × f × c represents the size of a convolution kernel f, the number of channels c, and d represents the expansion rate of void convolution. Pool denotes the pooling layer, here the largest pooling layer of size 2 x 2, with a step s of 2. FC is a fully connected neural network layer of different feature numbers. And the last but one FC has auxiliary data input and is trained together with the fingerprint sequence high-dimensional features extracted in the front after being spliced. And Dropout is added after the first FC to prevent overfitting, increasing the generalization performance of the model. All Conv are followed by a Batch Normalization (BN). Modified Linear units (relus) are used as activation functions except for the last layer. Since it is a multi-sort task, the activation function of the last FC employs the Softmax function. The expansion rate of the cavity convolution adopts 1,2,5, 1 and 2 sawtooth structures to meet the HDC design rule, so that the characteristics of different scales can be sensed, and the grid effect is avoided. The input of the model is a curve difference matrix respectively converted by geomagnetic fingerprint sequences of three dimensions, and the input direction parameter O is used as auxiliary data to train the model together after passing through a convolution layer and a full connection layer. A Keras deep learning framework is selected as a model framework, TensorFlow is used at the rear end, an Adam optimization algorithm is selected by an optimization method of a neural network, and a coordinated _ cross control loss function is selected.
4. Real-time position estimation
Because the nature of geomagnetic fingerprint sequence matching is the feature matching between a fingerprint sampling sequence and a fingerprint sequence in a certain segment on a path, although an overlapped segmentation mode is adopted, the problem of sequence dislocation still exists. Assuming that the path length corresponding to the fingerprint segment length SSN is FSL, when the walking start point is located between some two positioning points, the maximum dislocation length on the path is FSL/4, and the maximum dislocation length of the fingerprint sequence is SSN/4. In order to solve the problem, the subsection analyzes the problem and provides a position correction method.
The worst case is first analyzed: the maximum dislocation length occurs when the pedestrian real-time fingerprint Sample sequence is in the middle of two segmented fingerprint sequences divided on the path (as shown by Sample line segment in fig. 6). In this case, the fingerprint sample sequence only contains partial subsequences of two segmented fingerprint sequences, and if the sample sequence is fed into the classification model, the obtained result will have larger deviation. In order to ensure the validity of matching between sequences, it is first required to ensure that the fingerprint sample sequence at least contains a segmented fingerprint sequence with a complete anchor point of a path, so that a subsequence of the fingerprint sample sequence can be matched to the segmented fingerprint sequence, and thus the location is successful. Therefore, the problem of insufficient fingerprint information acquisition is solved by increasing the sampling length of the fingerprint sequence, wherein the increased sampling length is FSL/4 (shown as a dotted line segment in FIG. 6), and the sampling length L in actual positioning issampleThe (minimum walking distance) requirements are:
Figure BDA0002832820670000051
when the fingerprint sampling sequence is ensured to contain the segmented fingerprint sequence, the next work is to align the segmented fingerprint and carry out correct matching. The method adopts a sliding window algorithm to search, the window size is FSL, and meanwhile, the sliding step size slide is set to be FSL/10 in consideration of positioning accuracy and calculation cost.
And finally, the real-time positioning process comprises the steps of firstly carrying out moving average filtering pretreatment on the geomagnetic fingerprint sequence and the acceleration sequence acquired by the pedestrian, and then calculating the transfer threshold of the FSM algorithm. And dividing the acceleration sequence of each step, and calculating the step length of the single step and the walking distance. And judging whether the walking distance meets the shortest distance requirement obtained by a segment length evaluation algorithm or not after obtaining the walking distance, quitting the algorithm to prompt the walking to continue if the walking distance does not meet the shortest distance requirement, segmenting the fingerprint sequence by a sliding window algorithm if the walking distance meets the requirement, converting the segmented segments into a curve difference matrix, splicing the curve difference matrix with the direction parameters, sequentially entering a geomagnetic fingerprint classification model, and returning the pre-determined sites and the corresponding confidence degrees of the segmented fingerprint sequences. And finally, acquiring the positioning point and the index value with the maximum confidence coefficient, and calculating to obtain the final positioning point coordinate through position correction.
The user usage scenario of the invention is as follows:
in some large buildings there are more narrow and complex walking paths where pedestrians often get lost position and need to obtain location based services in real time, i.e. know where self or other people or objects are located in the indoor space. In the above scenario, the method has the advantages of obtaining a better effect, and being lower in equipment deployment cost and positioning walking cost and lower in positioning time delay compared with the existing method.

Claims (5)

1. The utility model provides a geomagnetic type indoor positioning system based on cavity convolution neural network which characterized in that: the accuracy and efficiency of real-time positioning are improved under the condition of lower positioning cost through the model constructed by the neural network. The system comprises the mobile terminal and the server terminal, and after a user initiates a positioning request through the mobile terminal, the server terminal can quickly return a positioning result, so that good positioning experience is provided for the user.
The mobile terminal is divided into a fingerprint acquisition module and a real-time positioning module, a geomagnetic fingerprint sequence of a main path in an indoor space is acquired through the fingerprint acquisition module of the mobile terminal before a positioning system is used, then a segmentation length of the fingerprint sequence is calculated through a geomagnetic fingerprint segmentation length evaluation algorithm, and each segmentation fingerprint is converted into a curve difference matrix after segmentation is carried out by adopting an overlapped segmentation method. Training a geomagnetic fingerprint classification model based on a cavity convolutional neural network at a server by taking the processed geomagnetic fingerprint as a data set, and establishing a mapping relation between the geomagnetic fingerprint and a position coordinate. After a user sends a positioning request through the mobile terminal real-time positioning module, the server terminal firstly processes various sensing data acquired in real time, the walking distance of the user is calculated, the geomagnetic fingerprint is converted into a curve difference matrix and then sent into a geomagnetic fingerprint classification model, the most possible positioning position point is predicted, and finally, the result is returned to the mobile terminal of the user after position correction.
2. The geomagnetic fingerprint sequence segment length evaluation algorithm according to claim 1, wherein: segmenting the geomagnetic fingerprint sequence in the space by different segment lengths, calculating the discrimination between the segments, and finally obtaining the shortest length which can ensure the mutual discrimination between the segmented fingerprints. And when the geomagnetic fingerprint data set is processed, segmenting according to the obtained length. The corresponding distance in space, namely the shortest walking length required by positioning, can also be calculated according to the segment length.
3. The overlappingly bisecting method of geomagnetic fingerprints as defined in claim 1, wherein: when the fingerprint sequence is segmented, overlapping segmentation is carried out by adopting 50% of overlapping proportion, the segmentation mode can more fully utilize geomagnetic fingerprint information in a limited space, and alignment cost during segmented fingerprint matching is reduced while locating points are segmented at a finer granularity.
4. The curve difference matrix according to claim 1, wherein: and the difference values among the sampling points are used for replacing absolute numerical values to describe the shape of a curve formed by the fingerprint sequence, and the relative relation among the numerical values of the sequence is better described, so that the neural network can better sense the high-dimensional spatial characteristics of the data. Compared with the normalization processing, the curve difference value matrix reserves the characteristic of the discrete degree between the sequences. The one-dimensional geomagnetic fingerprints form a two-dimensional image with richer contents after conversion, different types of curves correspond to different curve difference value matrix diagrams, and the image categories are clear. The horizontal and vertical coordinates of the matrix all represent the coordinates of sampling points of the sequence, and the difference value between the sampling points has high-dimensional characteristics in space.
5. The geomagnetic fingerprint classification model based on the hole convolutional neural network according to claim 1, wherein: in order to learn the characteristics of contribution of curve difference matrixes corresponding to more segmented geomagnetic fingerprints to model output during training, reduce information reduction and fuzziness caused by maximum pooling and average pooling, ensure a certain receptive field of a neural network and select hole convolutions with different expansion rates to replace a common pooling layer. The model consists of 6 hole convolution layers with expansion ratios cycling in {1,2,5} order, 1 maximum pooling layer, and 5 full-link layers. In a word, the hollow convolution in the model has the advantages that a large receptive field can be obtained under the condition of not losing information, and effective geomagnetic fingerprint information can be better learned. The classification model trained by a large number of geomagnetic fingerprint data sets improves the classification accuracy.
CN202011466245.5A 2020-12-13 2020-12-13 Geomagnetic indoor positioning system based on cavity convolution neural network Pending CN112580479A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011466245.5A CN112580479A (en) 2020-12-13 2020-12-13 Geomagnetic indoor positioning system based on cavity convolution neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011466245.5A CN112580479A (en) 2020-12-13 2020-12-13 Geomagnetic indoor positioning system based on cavity convolution neural network

Publications (1)

Publication Number Publication Date
CN112580479A true CN112580479A (en) 2021-03-30

Family

ID=75134789

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011466245.5A Pending CN112580479A (en) 2020-12-13 2020-12-13 Geomagnetic indoor positioning system based on cavity convolution neural network

Country Status (1)

Country Link
CN (1) CN112580479A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113469083A (en) * 2021-07-08 2021-10-01 西安电子科技大学 SAR image target classification method and system based on anti-sawtooth convolution neural network
CN114040347A (en) * 2021-10-29 2022-02-11 中国石油大学(华东) Signal fingerprint positioning method based on deep confidence network
CN114440888A (en) * 2022-01-14 2022-05-06 中山大学 Indoor positioning method and device based on sequence grouping sliding window
CN113469083B (en) * 2021-07-08 2024-05-31 西安电子科技大学 SAR image target classification method and system based on antialiasing convolutional neural network

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113469083A (en) * 2021-07-08 2021-10-01 西安电子科技大学 SAR image target classification method and system based on anti-sawtooth convolution neural network
CN113469083B (en) * 2021-07-08 2024-05-31 西安电子科技大学 SAR image target classification method and system based on antialiasing convolutional neural network
CN114040347A (en) * 2021-10-29 2022-02-11 中国石油大学(华东) Signal fingerprint positioning method based on deep confidence network
CN114440888A (en) * 2022-01-14 2022-05-06 中山大学 Indoor positioning method and device based on sequence grouping sliding window
CN114440888B (en) * 2022-01-14 2023-05-16 中山大学 Indoor positioning method and device based on sequence grouping sliding window

Similar Documents

Publication Publication Date Title
CN110012428B (en) Indoor positioning method based on WiFi
CN104064051B (en) A kind of passenger's portable mobile terminal and a locating information dynamic matching method that rides in a bus
CN110536245B (en) Deep learning-based indoor wireless positioning method and system
CN111479231A (en) Indoor fingerprint positioning method for millimeter wave large-scale MIMO system
CN109029429B (en) WiFi and geomagnetic fingerprint based multi-classifier global dynamic fusion positioning method
CN104807460A (en) Indoor positioning method and system for unmanned aerial vehicle
CN104517289A (en) Indoor scene positioning method based on hybrid camera
CN112580479A (en) Geomagnetic indoor positioning system based on cavity convolution neural network
CN110401977B (en) Multi-floor indoor positioning method based on Softmax regression multi-classification recognizer
CN110426037A (en) A kind of pedestrian movement track real time acquiring method under enclosed environment
CN111985389A (en) Basin similarity discrimination method based on basin attribute distance
CN114241464A (en) Cross-view image real-time matching geographic positioning method and system based on deep learning
CN110727714A (en) Resident travel feature extraction method integrating space-time clustering and support vector machine
CN111050282A (en) Multi-time fuzzy inference weighted KNN positioning method
CN111461251A (en) Indoor positioning method of WiFi fingerprint based on random forest and self-encoder
CN113239753A (en) Improved traffic sign detection and identification method based on YOLOv4
CN109665464A (en) A kind of method and system that movable type fork truck automatically tracks
Zhang et al. Regional Double-Layer, High-Precision Indoor Positioning System Based on iBeacon Network
CN109766969B (en) RFID indoor positioning algorithm based on asynchronous dominant motion evaluation
Wang et al. Indoor position algorithm based on the fusion of wifi and image
CN113947636B (en) Laser SLAM positioning system and method based on deep learning
CN114710831B (en) RFID label positioning system based on deep learning
CN110691336A (en) Double-scale positioning algorithm based on integrated learning and relative positioning
CN113316080B (en) Indoor positioning method based on Wi-Fi and image fusion fingerprint
CN114630266A (en) Multimode data fusion indoor positioning system based on neural network

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