CN113008226A - Geomagnetic indoor positioning method based on gated cyclic neural network and particle filtering - Google Patents
Geomagnetic indoor positioning method based on gated cyclic neural network and particle filtering Download PDFInfo
- Publication number
- CN113008226A CN113008226A CN202110180618.0A CN202110180618A CN113008226A CN 113008226 A CN113008226 A CN 113008226A CN 202110180618 A CN202110180618 A CN 202110180618A CN 113008226 A CN113008226 A CN 113008226A
- Authority
- CN
- China
- Prior art keywords
- geomagnetic
- particle
- indoor
- positioning
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/04—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by terrestrial means
- G01C21/08—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by terrestrial means involving use of the magnetic field of the earth
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
- G01C21/206—Instruments for performing navigational calculations specially adapted for indoor navigation
Landscapes
- Remote Sensing (AREA)
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Physics & Mathematics (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Life Sciences & Earth Sciences (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
- Navigation (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a geomagnetic indoor positioning method based on a gated recurrent neural network and particle filtering. The invention trains the gate control cyclic neural network through the built geomagnetic indoor database to match and position geomagnetic track signals. The trained gated recurrent neural network is used for matching and positioning, so that better positioning accuracy can be brought to the matching and positioning of the geomagnetic track signal, and compared with a commonly-used geomagnetic track signal matching algorithm based on dynamic time programming, the trained model reduces the real-time calculation amount in the matching and positioning process. The invention designs a system for carrying out real-time positioning by combining a particle filter algorithm on the basis of carrying out matching positioning on geomagnetic track signals by a neural network model. The system effectively utilizes the advantage of extracting the geomagnetic trajectory signal features by the gated recurrent neural network, and brings better real-time positioning accuracy for the particle filtering algorithm.
Description
Technical Field
The invention belongs to the field of indoor positioning, and particularly relates to a method for performing geomagnetic matching positioning based on a gated recurrent neural network and a method for performing real-time positioning by combining particle filtering.
Background
The indoor positioning technology has wide application value in daily life, such as positioning in a market, navigation in a parking lot and the like. The geomagnetism is a common indoor signal, and has the characteristics of no infrastructure and unique signal intensity at different positions. However, geomagnetism has a defect of low resolution, and similar situations can occur in different positions of geomagnetic signals in an indoor environment, so the effect of geomagnetism matching and positioning is often not good. The commonly used geomagnetic matching and positioning algorithm mainly acquires continuous geomagnetic track signals and a dynamic time planning algorithm to perform matching and positioning. The gate control recurrent neural network is an excellent recurrent neural network for extracting the characteristics of sequence data, and can better extract geomagnetic trajectory signals to perform matching positioning. Particle filtering is a real-time positioning method commonly used in the field of geomagnetic indoor positioning, and the positioning accuracy is effectively improved through continuous position estimation and real-time measurement update.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a geomagnetic indoor positioning method based on a gated recurrent neural network and particle filtering.
The invention mainly provides a building method of an indoor geomagnetic map database, and a gated recurrent neural network is used for extracting geomagnetic track signal characteristics to achieve a better matching positioning result.
The invention mainly utilizes a gate control cyclic neural network to carry out real-time positioning by combining a particle filter algorithm on the basis of matching and positioning geomagnetic track signals, and the whole invention is mainly divided into three stages: and (3) building an indoor geomagnetic map database, and training a model to perform matching positioning and real-time positioning of particle filtering. The construction stage of the indoor geomagnetic map database mainly divides an indoor map and collects geomagnetic track signals. In the stage of training the model for matching and positioning, the indoor geomagnetic map database is trained mainly through the gated recurrent neural network, so that the model has a more excellent matching and positioning effect. And in the real-time positioning stage of the particle filtering, the real-time positioning is mainly carried out by combining the particle filtering and the trained model. The method is implemented according to the following steps:
step 1, building an indoor geomagnetic map database:
1) the size of the indoor map is measured and divided into a plurality of grids.
2) And collecting a plurality of groups of geomagnetic track signals in the walking direction of the descending people at different times.
3) Assigning geomagnetic trajectory signals to the grids and assigning position tags to the geomagnetic trajectory signals.
Step 2, training the model to perform matching positioning:
1) and dividing the indoor geomagnetic map database into a training set and a test set.
2) And training a gated cyclic neural network model for matching and positioning.
Step 3, real-time positioning of particle filtering:
1) and constructing an indoor geomagnetic map according to the indoor geomagnetic map database.
2) And estimating the walking step length of the pedestrian according to a geomagnetic sensor real-time acquisition geomagnetic track signal and a pedestrian route presumption algorithm.
3) Based on an indoor geomagnetic chart and a trained gated cyclic neural network model, after dividing geomagnetic track signals collected in real time, carrying out real-time positioning by combining a particle filtering algorithm.
The method of the invention has the advantages and beneficial results that:
1. the geomagnetic signals are adopted, and basic equipment does not need to be additionally built and maintained.
2. The invention designs a construction method of a geomagnetic indoor database, and mainly performs matching positioning on geomagnetic track signals by training a gated recurrent neural network through the constructed geomagnetic indoor database. The trained gated recurrent neural network is used for matching and positioning, so that better positioning accuracy can be brought to the matching and positioning of the geomagnetic track signal, and compared with a commonly-used geomagnetic track signal matching algorithm based on dynamic time programming, the trained model reduces the real-time calculation amount in the matching and positioning process.
3. The invention designs a system for carrying out real-time positioning by combining a particle filter algorithm on the basis of carrying out matching positioning on geomagnetic track signals by a neural network model. The system effectively utilizes the advantage of extracting the geomagnetic trajectory signal features by the gated recurrent neural network, and brings better real-time positioning accuracy for the particle filtering algorithm.
Drawings
FIG. 1 is a diagram of a process framework of the present invention;
FIG. 2 is a schematic diagram of the indoor map partitioning of the present invention;
FIG. 3 is a block diagram of a gated recurrent neural network in an embodiment of the present invention;
fig. 4 is a structural diagram of a particle filter according to an embodiment of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific embodiments.
The invention mainly utilizes the gate control cyclic neural network to extract the characteristics of geomagnetic track signals and provides a real-time geomagnetic positioning method combining particle filtering on the basis of matching positioning of the gate control cyclic neural network. The frame is generally as shown in figure 1. Specifically, the method is carried out according to the following steps.
Step 1, building an indoor geomagnetic map database:
geomagnetic signals collected by geomagnetic sensor are represented by three-dimensional vectors<Mx,My,MZ>In this case, the three vector values represent geomagnetic signals measured on three directional axes of the sensor, respectively. The geomagnetic intensity signal is further expressed by a second-order norm of three vector values. The geomagnetic intensity signal M is as follows:
the geomagnetic trajectory signal is composed of successive geomagnetic intensity signals. As shown in fig. 2, the indoor geomagnetic map database is constructed by dividing a map into square grids with side length D (D is greater than or equal to 0.4m and less than or equal to 0.8m, the specific side length is determined according to the size of the indoor map, and the side length capable of equally dividing the indoor map is preferably selected), wherein the number of the grids is determined according to the size of the indoor map, and geomagnetic trajectory signals are collected for each grid in the walking direction of pedestrians. For the diversity of each grid data, geomagnetic trajectory signals are acquired at different times for each grid. And distributing a corresponding position label for the geomagnetic track signal acquired by each grid, wherein the position label is determined according to a relative coordinate system established by the indoor map and the last position of the geomagnetic track signal, the origin of the relative coordinate system is set to be a certain vertex of the indoor map, and the last position is the center point corresponding to the grid at the edge pointed by the geomagnetic track signal.
Step 2, training the model to perform matching positioning:
the indoor geomagnetic map database is divided, wherein 70% serves as a training set, and 30% serves as a test set for training the gated recurrent neural network. The adopted gated recurrent neural network comprises 3 GRU hidden layers as shown in fig. 3, and geomagnetic trajectory data in a training set is input according to time cycle. In the process of training the gated recurrent neural network, a square loss function is adopted as a loss function of the model, and the formula is as follows:
wherein xiAnd yiA position tag indicating a geomagnetic trajectory signal in the indoor geomagnetic map database,andand a position label representing the prediction of the geomagnetic trajectory signal by the gated recurrent neural network.
Step 3, real-time positioning based on particle filtering:
firstly, building an indoor geomagnetic map, and selecting the most representative data for each grid in the same direction for building, wherein the most representative geomagnetic track data is as follows: and calculating the average value of all geomagnetic track signals of the grids in the same direction, and screening out the data which is closest to the average value from all geomagnetic track signals as the most representative geomagnetic track data. The indoor geomagnetic map is used for the following particle filtering.
And the particle filter is positioned in real time by combining a trained gated recurrent neural network model. As shown in fig. 4, the particle filter is divided into five stages: particle initialization, particle propagation, particle weight update, position estimation and particle resampling. Gated recurrent neural networks are used in particle weight updates.
Particle initialization is firstly carried out, then particle weight updating is carried out, position estimation is carried out again, particle resampling is carried out after position estimation is finished, particle weight updating is carried out again after particle propagation, circulation is formed, and real-time indoor geomagnetic positioning is finished.
(1) Particle initialization:
in the real-time localization of geomagnetism, a certain number of particles are distributed at random positions in an indoor map. The set of particles broadcast in the indoor map is called a set of particles.
(2) Particle propagation: and estimating the position of the particle at the next moment. Estimating the particle position at the next moment according to the particle position at the previous moment, namely estimating the current walking step length and direction of the user, and assuming that the step length is l and the direction angle is theta, the formula of particle propagation is as follows:
where k represents a particle in the set of particles and t represents a state time. The step size and the direction angle are calculated by a pedestrian course presumption algorithm.
(3) Updating the weight of the particles: and updating the weight of the particle set in the indoor map according to the data acquired in real time. At an initial time, each particle in the set of particles is assigned a uniform weight, the sum of which is 1. The unified update of the particle weights is performed after particle initialization or particle propagation. The formula for particle weight update is as follows:
whereinIndicated is the weight of the kth particle in the set of particles at time t,expressed as the position weight of the particle at time t, ztThe geomagnetic trajectory signal is expressed as observation data at time t, that is, the geomagnetic trajectory signal acquired in real time.
For the calculation of the probability of the proposed distribution: firstly, calculating the difference value of the matching positioning position of the trained gate control cyclic neural network model on the geomagnetic track signal acquired in real time and the position of the particle, and finally obtaining the suggested distribution probability under Gaussian distribution of the difference value.
For the calculation of the measurement distribution probability: firstly, calculating a difference value between a geomagnetic track signal in an indoor geomagnetic map corresponding to the position of the particle and a geomagnetic track signal acquired in real time through a dynamic time programming algorithm, and finally obtaining a measurement distribution probability under Gaussian distribution of the difference value.
For the calculation of the transition distribution probability: firstly, the difference value of the predicted positioning position estimated by the position set obtained by the particle filter algorithm in the past and the position of the particle is obtained, and finally the transfer distribution probability is obtained by the difference value under the Gaussian distribution. Suppose with (x)t,yt) Representing the location of the particle filter algorithm at time t,indicating particles at time tThe filter algorithm estimates the predicted position location from the past derived set of locations. The formula for predicting the location is as follows:
where k represents the length of the past sequence trajectory that needs to be taken and is set to 6 in the positioning system. w is aiThe weights representing the step sizes at the past time, as can be seen from the second formula, the more the weight of the step size closest to the current time. In addition, after the particle weights are updated, the weights of all the particles are normalized so that the sum of the weights is 1.
(4) And (3) position estimation: estimating the position of the pedestrian at the current moment by the particle set, selecting the first 30% of particles in the particle set which are sorted from large to small according to the weight size, and estimating the position by an enhanced summation algorithm, wherein the formula is as follows:
where k represents the number of particles in the set of particles to take the higher weight fraction.
(5) And (3) resampling particles:
the aim of redistributing the particles in the particle set is to reduce the particles with low weight and increase the particles with high weight, thereby solving the problem of particle degradation.
Claims (4)
1. The geomagnetic indoor positioning method based on the gated recurrent neural network and the particle filtering is characterized by comprising the following steps of:
step 1, building an indoor geomagnetic map database:
1) measuring the size of the indoor map, and dividing the indoor map into a plurality of grids;
2) collecting a plurality of groups of geomagnetic track signals in the walking direction of the descending person at different times;
3) distributing geomagnetic track signals to grids and distributing position labels to the geomagnetic track signals;
step 2, training the model to perform matching positioning:
1) dividing an indoor geomagnetic map database into a training set and a test set;
2) training a gated cyclic neural network model for matching and positioning;
step 3, real-time positioning based on particle filtering:
1) constructing an indoor geomagnetic map according to an indoor geomagnetic map database;
2) the method comprises the steps of acquiring geomagnetic track signals in real time according to a geomagnetic sensor and estimating the walking step length of a pedestrian by a pedestrian route estimation algorithm;
3) based on an indoor geomagnetic chart and a trained gated cyclic neural network model, after dividing geomagnetic track signals collected in real time, carrying out real-time positioning by combining a particle filtering algorithm.
2. The indoor geomagnetic positioning method based on the gated recurrent neural network and the particle filtering according to claim 1, wherein step 1 is to construct an indoor geomagnetic map database, and specifically comprises the following operations:
geomagnetic signals collected by geomagnetic sensor are represented by three-dimensional vectors<Mx,My,Mz>Representing that the three vector values respectively represent geomagnetic signals measured on three direction axes of the sensor; the geomagnetic intensity signal is expressed by a second-order norm of three vector values; the geomagnetic intensity signal M is as follows:
the geomagnetic track signal is composed of continuous geomagnetic intensity signals; building an indoor geomagnetic map database, firstly, dividing a map into square grids with the side length of D, wherein D is more than or equal to 0.4m and less than or equal to 0.8m, the specific side length is determined according to the size of the indoor map, the number of the grids is determined by the size of the indoor map, and geomagnetic track signals are collected for each grid in the walking direction of pedestrians; acquiring geomagnetic track signals at different times for each grid for the diversity of each grid data; and distributing a corresponding position label for the geomagnetic track signal acquired by each grid, wherein the position label is determined according to a relative coordinate system established by the indoor map and the last position of the geomagnetic track signal, the origin of the relative coordinate system is set to be a certain vertex of the indoor map, and the last position is the center point corresponding to the grid at the edge pointed by the geomagnetic track signal.
3. The geomagnetic indoor positioning method based on the gated recurrent neural network and the particle filtering, according to claim 2, wherein the step 2 trains the training model to perform matching positioning, and the specific operations are as follows:
dividing an indoor geomagnetic map database, wherein 70% of indoor geomagnetic map databases are used as training sets, and 30% of indoor geomagnetic map databases are used as test sets for training the gated recurrent neural network; the adopted gate control cyclic neural network comprises 3 GRU hidden layers, and geomagnetic trajectory data in a training set are input according to time cycle; in the process of training the gated recurrent neural network, a square loss function is adopted as a loss function of the model, and the formula is as follows:
4. The geomagnetic indoor positioning method based on the gated recurrent neural network and the particle filtering, according to claim 3, wherein the step 3 is based on real-time positioning of the particle filtering, and specifically comprises the following operations:
firstly, building an indoor geomagnetic map, and selecting the most representative data for each grid in the same direction for building, wherein the most representative geomagnetic track data is as follows: calculating the average value of all geomagnetic track signals of the grids in the same direction, and screening out data which is closest to the average value in all geomagnetic track signals to serve as the most representative geomagnetic track data;
particle filtering is positioned in real time by combining a trained gated cyclic neural network model; particle filtering is divided into five stages: particle initialization, particle propagation, particle weight updating, position estimation and particle resampling; the gated recurrent neural network is used in particle weight updating;
particle initialization is firstly carried out, then particle weight updating is carried out, position estimation is carried out again, particle resampling is carried out after position estimation is finished, particle weight updating is carried out again after particle propagation, circulation is formed, and real-time indoor geomagnetic positioning is finished;
(1) particle initialization:
distributing a certain number of particles at random positions in an indoor map in the real-time positioning of the geomagnetism; the set of particles broadcast in the indoor map is called a set of particles;
(2) particle propagation: estimating the position of the particle at the next moment; estimating the particle position at the next moment according to the particle position at the previous moment, namely estimating the current walking step length and direction of the user, and assuming that the step length is l and the direction angle is theta, the formula of particle propagation is as follows:
wherein k represents a certain particle in the particle set, and t represents a certain state moment; step length and direction angle are calculated through a pedestrian route presumption algorithm;
(3) updating the weight of the particles: updating the weight of the particle set in the indoor map according to the data collected in real time; at the initial moment, each particle in the particle set is assigned with a weight with a uniform size, and the sum of the weights is 1; unified updating of particle weights is performed after particle initialization or particle propagation; the formula for particle weight update is as follows:
whereinIndicated is the weight of the kth particle in the set of particles at time t,expressed as the position weight of the particle at time t, ztThe geomagnetic trajectory signal is represented as observation data at the time t, namely geomagnetic trajectory signals collected in real time;
for the calculation of the probability of the proposed distribution: firstly, calculating a difference value between a matching positioning position of a trained gate control circulation neural network model on geomagnetic track signals acquired in real time and a position where a particle is located, and finally obtaining a suggested distribution probability under Gaussian distribution of the difference value;
for the calculation of the measurement distribution probability: firstly, calculating a difference value between a geomagnetic track signal in an indoor geomagnetic map corresponding to the position of the particle and a geomagnetic track signal acquired in real time through a dynamic time planning algorithm, and finally obtaining a measurement distribution probability under Gaussian distribution of the difference value;
for the calculation of the transition distribution probability: firstly, obtaining a difference value between a predicted positioning position estimated by a position set obtained by a particle filter algorithm in the past and a position where a particle is located, and finally obtaining a transition distribution probability under Gaussian distribution of the difference value; suppose with (x)t,yt) Representing the particle filter algorithm at time tThe position of the position to be determined,a predicted positioning position representing a position set estimate obtained in the past by the particle filter algorithm at time t; the formula for predicting the location is as follows:
where k represents the length of the trace that needs to take the past sequence, set to 6 in the positioning system; w is aiThe weight of the step length at the past moment is represented, and as can be seen from the second formula, the weight of the step length closest to the current moment is heavier; in addition, after the particle weights are updated, normalization processing is carried out on the weights of all the particles so that the sum of the weights is 1;
(4) and (3) position estimation: estimating the position of the pedestrian at the current moment by the particle set, selecting the first 30% of particles in the particle set which are sorted from large to small according to the weight size, and estimating the position by an enhanced summation algorithm, wherein the formula is as follows:
wherein k represents the number of particles in the set of particles to be weighted higher;
(5) and (3) resampling particles:
the aim of redistributing the particles in the particle set is to reduce the particles with low weight and increase the particles with high weight, thereby solving the problem of particle degradation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110180618.0A CN113008226B (en) | 2021-02-09 | 2021-02-09 | Geomagnetic indoor positioning method based on gated cyclic neural network and particle filtering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110180618.0A CN113008226B (en) | 2021-02-09 | 2021-02-09 | Geomagnetic indoor positioning method based on gated cyclic neural network and particle filtering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113008226A true CN113008226A (en) | 2021-06-22 |
CN113008226B CN113008226B (en) | 2022-04-01 |
Family
ID=76384061
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110180618.0A Active CN113008226B (en) | 2021-02-09 | 2021-02-09 | Geomagnetic indoor positioning method based on gated cyclic neural network and particle filtering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113008226B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114440888A (en) * | 2022-01-14 | 2022-05-06 | 中山大学 | Indoor positioning method and device based on sequence grouping sliding window |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018018994A1 (en) * | 2016-07-27 | 2018-02-01 | 无锡知谷网络科技有限公司 | Method and system for indoor positioning |
CN110388926A (en) * | 2019-07-12 | 2019-10-29 | 杭州电子科技大学 | A kind of indoor orientation method based on mobile phone earth magnetism and scene image |
CN110401978A (en) * | 2019-07-19 | 2019-11-01 | 中国电子科技集团公司第五十四研究所 | Indoor orientation method based on neural network and particle filter multi-source fusion |
CN110530371A (en) * | 2019-09-06 | 2019-12-03 | 电子科技大学 | A kind of indoor map matching process based on deeply study |
CN111256684A (en) * | 2020-01-18 | 2020-06-09 | 杭州电子科技大学 | Geomagnetic indoor positioning method based on multilayer gate control circulation unit network |
-
2021
- 2021-02-09 CN CN202110180618.0A patent/CN113008226B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018018994A1 (en) * | 2016-07-27 | 2018-02-01 | 无锡知谷网络科技有限公司 | Method and system for indoor positioning |
CN110388926A (en) * | 2019-07-12 | 2019-10-29 | 杭州电子科技大学 | A kind of indoor orientation method based on mobile phone earth magnetism and scene image |
CN110401978A (en) * | 2019-07-19 | 2019-11-01 | 中国电子科技集团公司第五十四研究所 | Indoor orientation method based on neural network and particle filter multi-source fusion |
CN110530371A (en) * | 2019-09-06 | 2019-12-03 | 电子科技大学 | A kind of indoor map matching process based on deeply study |
CN111256684A (en) * | 2020-01-18 | 2020-06-09 | 杭州电子科技大学 | Geomagnetic indoor positioning method based on multilayer gate control circulation unit network |
Non-Patent Citations (3)
Title |
---|
LEE NAMKYOUNG ET AL.: "AMID: Accurate Magnetic Indoor Localization Using Deep Learning", 《SENSORS》 * |
LIU ZHENGUANG ET AL.: "Fusion of Magnetic and Visual Sensors for Indoor Localization: Infrastructure-Free and More Effective", 《IEEE TRANSACTIONS ON MULTIMEDIA》 * |
刘宇等: "基于遗传算法优化神经网络的多源信息融合室内定位方法", 《中国惯性技术学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
Also Published As
Publication number | Publication date |
---|---|
CN113008226B (en) | 2022-04-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107818571B (en) | Ship automatic tracking method and system based on deep learning network and average drifting | |
CN110012428B (en) | Indoor positioning method based on WiFi | |
CN105263113B (en) | A kind of WiFi location fingerprints map constructing method and its system based on crowdsourcing | |
CN111368769B (en) | Ship multi-target detection method based on improved anchor point frame generation model | |
CN110135354B (en) | Change detection method based on live-action three-dimensional model | |
CN109115205A (en) | A kind of indoor fingerprint positioning method and system based on geomagnetic sensor array | |
CN107182036A (en) | The adaptive location fingerprint positioning method merged based on multidimensional characteristic | |
CN103049751A (en) | Improved weighting region matching high-altitude video pedestrian recognizing method | |
CN105869178A (en) | Method for unsupervised segmentation of complex targets from dynamic scene based on multi-scale combination feature convex optimization | |
CN109508585A (en) | A method of urban function region is extracted based on POI and high-resolution remote sensing image | |
CN109145836A (en) | Ship target video detection method based on deep learning network and Kalman filtering | |
CN110443830B (en) | Precipitation cloud cluster detection method based on Gaussian mixture background model | |
CN110213003A (en) | A kind of wireless channel large-scale fading modeling method and device | |
CN114627447A (en) | Road vehicle tracking method and system based on attention mechanism and multi-target tracking | |
CN111901749A (en) | High-precision three-dimensional indoor positioning method based on multi-source fusion | |
CN110388926A (en) | A kind of indoor orientation method based on mobile phone earth magnetism and scene image | |
CN111383273B (en) | High-speed rail contact net part positioning method based on improved structure reasoning network | |
CN110334584A (en) | A kind of gesture identification method based on the full convolutional network in region | |
CN106951829A (en) | A kind of notable method for checking object of video based on minimum spanning tree | |
CN113596989A (en) | Indoor positioning method and system for intelligent workshop | |
CN112100435A (en) | Automatic labeling method based on edge end traffic audio and video synchronization sample | |
CN113008226B (en) | Geomagnetic indoor positioning method based on gated cyclic neural network and particle filtering | |
WO2022242018A1 (en) | Indoor target positioning method based on improved cnn model | |
CN105654061A (en) | 3D face dynamic reconstruction method based on estimation compensation | |
CN113671599A (en) | Global climate mode-based login cyclone identification method |
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 |