CN110458887A - A kind of Weighted Fusion indoor orientation method based on PCA - Google Patents
A kind of Weighted Fusion indoor orientation method based on PCA Download PDFInfo
- Publication number
- CN110458887A CN110458887A CN201910636664.XA CN201910636664A CN110458887A CN 110458887 A CN110458887 A CN 110458887A CN 201910636664 A CN201910636664 A CN 201910636664A CN 110458887 A CN110458887 A CN 110458887A
- Authority
- CN
- China
- Prior art keywords
- frame image
- pca
- elm
- image
- result
- 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
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
- G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
- G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
- G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
- G01C21/165—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation combined with non-inertial navigation instruments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Automation & Control Theory (AREA)
- Image Analysis (AREA)
- Navigation (AREA)
Abstract
The invention discloses a kind of Weighted Fusion indoor orientation method based on PCA, comprising: training sample set is input in ELM neural network, establishes neural network model to training sample set using ELM regression algorithm by initialization ELM model;Test sample collection is input in trained neural network model, the relative displacement between consecutive frame image is obtained, relative displacement result is integrated, the position of frame image is obtained;Image fuzzy Judgment is introduced, calculates the position of current frame image as frame framing result using the position of Inertia information and previous frame image;Drift correction is carried out to original acceleration measuring signal using visual information, the acceleration after drift correction is subjected to quadratic integral, obtains inertial positioning result;Weight distribution is carried out to frame image, inertial positioning result using PCA, obtains final positioning result.The accumulated error of INS can be effectively controlled in this method, effectively solves the problems, such as VNS vulnerable to external disturbance.
Description
Technical field
The present invention relates to indoor positioning, information fusion and field of signal processing, more particularly to it is a kind of based on PCA (it is main at
Point analysis) Weighted Fusion indoor orientation method.
Background technique
In recent years, the demand rapid growth of indoor positioning service, indoor locating system become more and more popular.The whole world is fixed
Position system (Global Positioning System, GPS) is most popular positioning and navigation system, and GPS is in outdoor environment
Positioning accuracy can reach several meters when middle use, have good accuracy and higher confidence level.But due to wall barrier and more
Diameter effect, GPS can not provide reliable service in environment indoors.It is well known that indoor positioning item tracking indoors, in shop
It is played an important role in a variety of applications such as shopping guide and indoor navigation.Therefore, it is necessary to find more effective way to mention
For indoor positioning service.
Be currently suggested multiple indoor location technology: the location technology based on single piece of information source, the method realized extensively are
Received signal strength (RSS).Signal source can be Wi-Fi[1], FM, bluetooth etc..Wi-Fi is considered as most promising method,
It can be by Positioning Precision Control in the range of several meters.But Wi-Fi system, there are some disadvantages, Wi-Fi access point is usual
About 90 meters of covering radius of region, and it is easy the interference by other signals.Bluetooth indoor positioning has at low cost, power consumption
Low, the advantages that equipment volume is small, but for complicated space environment, the stability of bluetooth positioning system is slightly worse, is easy by outer
The influence of portion's noise signal.It can be seen that the location technology in existing single piece of information source is restricted by positioning accuracy and reliability, all
It can not popularization and application in daily life.
Recently, inertial navigation system (INS)[2]Have become the focus of indoor positioning research.Because it can be not outer
Position is provided in the case where portion's equipment, and there is quick data renewal speed, has small in size, at low cost, portability is strong
The features such as.But as time goes by, the error of gyroscope and accelerometer will increase rapidly, INS is merely able to realize
In short term, short distance positioning[3].With continuously improving for computer vision technique, domestic and international expert is to vision navigation system (VNS)
It is more and more interested.VNS[4]A kind of good method understood using vision data and perceive indoor environment is provided, it can be
High position precision is obtained in scene with characteristic matching abundant and identification.C.Piciarelli[5]It proposes a kind of by image
It is compared to realize the vision indoor positioning technologies of positioning (at this with the reference model of the visual signature with position mark
In referred to as VL algorithm), compared with non-vision navigation system, it, which has, contains much information, and positioning accuracy is high, it is noiseless the advantages that,
However it is ineffective in some cases, it such as blocks, light variation and personnel access interference etc..
Summary of the invention
The present invention provides a kind of Weighted Fusion indoor orientation method based on PCA, using ELM, (limit learns the present invention
Machine) regression algorithm obtain view-based access control model data location information, utilize visual information static feedback carry out drift correction, improve
Traditional inertial positioning finally utilizes principal component analytical method (PCA)[6]Improved inertia and vision positioning result are weighed
It reassigns, this method can effectively control the accumulated error of inertial navigation system (INS), effectively solution vision navigation system
(VNS) vulnerable to external disturbance the problem of, described below:
A kind of Weighted Fusion indoor orientation method based on PCA, the described method comprises the following steps:
ELM model is initialized, training sample set is input in ELM neural network, using ELM regression algorithm to training sample
This collection establishes neural network model;
Test sample collection is input in trained neural network model, the relative displacement between consecutive frame image is obtained,
Relative displacement result is integrated, the position of frame image is obtained;
Image fuzzy Judgment is introduced, is made using the position that the position of Inertia information and previous frame image calculates current frame image
For frame framing result;
Drift correction is carried out to original acceleration measuring signal using visual information, the acceleration after drift correction is carried out
Quadratic integral obtains inertial positioning result;
Weight distribution is carried out to frame image, inertial positioning result using PCA, obtains final positioning result.
Wherein, the training sample set specifically:
The SURF descriptor for extracting image, to the N number of SURF characteristic point of every frame image zooming-out, and carries out Feature Points Matching;It adopts
Matching result is handled with RANSAC algorithm, Mismatching point is removed, obtains affine transformation matrix;Calculate each frame image with
Relative displacement between next frame image true coordinate;
It is inputted affine matrix as training, relative displacement establishes training sample set as training output.
Further, the introducing image fuzzy Judgment is calculated current using the position of Inertia information and previous frame image
The position of frame image is as frame framing result specifically:
This index of introduced feature matching rate NS, by can be used for measuring whether current image obscures to NS given threshold;
When NS is less than threshold value, determine that the frame image is fuzzy, using the positioning result of the previous frame obtained by ELM, in conjunction with
Acceleration, the Inertia informations such as gyroscope obtain the position of present frame.
Wherein, described that drift correction is carried out to original acceleration measuring signal using visual information, after drift correction
Acceleration carries out quadratic integral specifically:
By the threshold value of translational movement between setting consecutive frame image, to distinguish motion state and stationary state.Work as translation vector
When less than threshold value, judgement is currently at stationary state, to carry out drift correction, otherwise regards as motion state and its original is kept to add
Velocity amplitude is constant, carries out quadratic integral to the acceleration after drift correction and obtains final position.
The beneficial effect of the technical scheme provided by the present invention is that:
(1) present invention merges inertia and visual information by PCA, so that inertia and visual information are rung in accuracy and frequency
It answers aspect to obtain complementation, improves positioning performance;
(2) present invention carries out drift correction to inertial data by carrying out static feedback using visual information, can be effective
Ground controls the error accumulation of inertial navigation system (INS);
(3) present invention carries out weight distribution to improved inertia and vision positioning result using PCA, and it is fused fixed to obtain
Position is as a result, the positioning result made is more accurate.
Detailed description of the invention
Fig. 1 is a kind of flow chart of Weighted Fusion indoor orientation method based on PCA;
The error accumulation distribution map contrast schematic diagram for the location algorithm that Fig. 2 is proposed by this method and document [5];
Fig. 3 is the concealed nodes number of vision positioning precision and ELM and the relation schematic diagram of activation primitive.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, embodiment of the present invention is made below further
Ground detailed description.
The embodiment of the invention provides a kind of Weighted Fusion indoor orientation method based on PCA, is mainly made of 3 parts:
The positioning of view-based access control model, the positioning based on inertia, and the positioning of the fusion based on PCA.
One, the positioning of view-based access control model:
Training sample set of the pretreatment creation comprising training input X and training output T is carried out to training vision data, will be instructed
Practice sample set to be input in ELM neural network, using ELM regression algorithm learning training sample set, establishes ELM neural network mould
Type.
Testing vision data are pre-processed, test sample collection is created, test sample collection is input to trained ELM
In neural network model, the relative displacement between consecutive frame image is obtained, then by integral calculation, obtains the position of each frame picture
Set coordinate.
Introduce image fuzzy Judgment simultaneously, when the image in video sequence is judged as fuzzy, using Inertia information and
The position of previous frame image calculates the position of present frame, the positioning result as vision.
Wherein, above-mentioned ELM neural network, ELM regression algorithm, establish the step of ELM neural network model and image
The step of fuzzy Judgment, is known to those skilled in the art, and the embodiment of the present invention does not repeat them here this.
Two, based on the positioning of inertia:
The static exercise state that video camera detects is fed back into inertial sensor, carries out drift correction.If at one section
When characteristic point pixel in time between successive frame hardly happens variation, just determine that object is in stationary motion state.However
Most of inertial data is not zero, thus using vision system by this static state-feedback to inertial sensor be used for into
Row drift correction removes inertial data cumulative errors.
By the threshold value of translational movement between setting consecutive frame image, to distinguish motion state and stationary state.Work as translation vector
When less than threshold value, judgement is currently at stationary state, to carry out drift correction, otherwise regards as motion state and keeps its former speed
Angle value is constant.
Using the result of above-mentioned drift correction as the result of inertial positioning.
Three, the fusion positioning based on PCA:
Using vision positioning result and inertial positioning result as two indices, their corresponding weights are determined using PCA,
The positioning result of vision and inertia is multiplied by obtaining final positioning result after corresponding Weight.
Embodiment 1
The technical solution of the embodiment of the present invention is further introduced below with reference to specific calculation formula, attached drawing,
It is described below:
Training sample set of the pretreatment creation comprising training input X and training output T is carried out to training vision data:
Firstly, pre-processing to visual information, SURF (accelerating robustness) feature is extracted to every frame training image, and will
The image I that number is iiThe image I for being i+1 with numberi+1It is matched, it is right using random sampling consistency (RANSAC algorithm)
Matching result is handled, and is removed Mismatching point, and calculate affine transformation matrix, is obtained the affine transformation square for being best suitable for match point
Battle array, formula are as follows:
In formula, r represents rotation angle, and A is scale vectors, Tx, TyRepresent translation vector.
Secondly, initialization ELM model, training sample is input in ELM neural network, using ELM regression algorithm to instruction
Practice sample set and establishes neural network model.
The step of constructing target output vector, by the relative displacement Δ T between adjacent two field picturesiAs output vector, Δ Ti
It can be acquired by following formula:
ΔTi=(Δ xi,Δyi)=(xi-xi-1,yi-yi-1) (2)
In formula: i ∈ (1 ..., N), xi,yiIt is the true x-axis and y-axis coordinate of every frame image, Δ x respectivelyiIt is adjacent in x-axis
Displacement between two field pictures, Δ yiFor the displacement between two field pictures adjacent in y-axis.
ELM model is initialized, training sample set is input in ELM neural network, using ELM regression algorithm to training sample
This collection establishes neural network model.
Wherein, the step of above-mentioned initialization is known to those skilled in the art, and the embodiment of the present invention does not repeat them here this.
Testing vision data are pre-processed, test sample collection is created, and test sample collection are input to trained
In ELM neural network model, obtain being displaced output accordingly.
The output position coordinate difference Δ T of each frame and next framek=(Δ xk,Δyk), k ∈ (1 ..., M) then passes through
Formula is to Δ TkIntegral calculation is carried out, each frame picture I is obtainedkOutput position.
In order to measure the fog-level of present image, this index of characteristic matching rate is introduced, calculation formula is as follows:
NS=Nc/Nq (4)
Wherein, NS represents characteristic matching rate, NcCharacteristic matching quantity between consecutive frame image, NqRepresent reference picture
Feature sum.
By setting a reasonable threshold value to NS, can be used to measure whether present image obscures.When picture blur,
That is when NS is less than threshold value, present frame is positioned using inertial data according to formula (5)-(6), it is upper using being obtained by ELM
The positioning result of one frame, in conjunction with acceleration, the Inertia informations such as gyroscope obtain the position of present frame.
xk=xk-1+Δsk-1·cosψk-1 (5)
yk=yk-1+Δsk-1·sinψk-1 (6)
Wherein, xkAnd ykIt is position coordinates of the k moment target in reference frame.Δsk-1It is target from k-1 moment to k
The distance run in this period at moment, ψk-1It is angle of the target at the k-1 moment.
Drift correction is carried out to original acceleration measuring signal using visual information, the acceleration after drift correction is carried out
Quadratic integral obtains the positioning result based on inertial data specifically:
Static exercise state inertial data is fed back to using visual information to be used to carry out drift correction.When in a period of time
When characteristic point pixel between successive frame hardly happens variation, just determine that object is in stationary motion state.
The embodiment of the present invention passes through the threshold value of translational movement between setting consecutive frame image, to distinguish motion state and static shape
State.When translation vector is less than threshold value, judgement is currently at stationary state, to carry out drift correction, otherwise regards as moving
State keeps its former acceleration value constant, carries out quadratic integral to the acceleration after drift correction and obtains final position.
Translational movement calculation formula is as follows:
Wherein, Tx, TyFor translation vector in affine transformation matrix.
Drift correction formula is as follows:
Wherein, akFor the acceleration information at k moment, FkTranslational movement between consecutive frame image, thFBetween consecutive frame image
The threshold value of translational movement.
Weight distribution finally is carried out to vision positioning result and inertial positioning result using PCA, to obtain final
Positioning result.
In conclusion the embodiment of the present invention merges inertia by PCA and visual information is more robust to provide, more accurately
Positioning result, the embodiment of the present invention introduce drift correction, and effective solution history inertial data accumulated error is melted by PCA
Inertia and vision positioning are closed as a result, fusion positioning result is more accurate.
Embodiment 2
The feasibility of scheme in embodiment 1 is verified below with reference to Fig. 2-Fig. 3, table 1 and specific example, is detailed in
It is described below:
To the effect of this method, using the algorithm steps in embodiment 1 as above to total duration 65 seconds, shift length 13m
Experiment carry out positioning analysis, the experiment include personnel arbitrarily pass in and out and scene be mutated etc. interference scene.Parameter setting is such as
Under: node in hidden layer 450, SURF characteristic NqIt is 30.
Qualitative angle, Fig. 2 are illustrated by the error accumulation distribution map comparison for the location algorithm that this method and document [5] propose
Figure;In order to reach the best locating effect of VL algorithm, therefore VL algorithm SURF number is set as 100 by this method in comparative experiments,
It, can it can be seen that this method has apparent advantage compared to VL algorithm in terms of reliability and stability from experimental result
By control errors within 2m, accurately positioning result can be still provided in the case where personnel's interference.
From quantitative angle, table 1 is that vision positioning, improved inertial positioning, PCA fusion positioning and the VL based on ELM are calculated
Each evaluation index result that method obtains.
Each evaluation index result of table 1
This method further improves the precision of inertial positioning system after drift correction, and with single vision positioning system
System is compared, and the RMSE of the method proposed is usually smaller.This means that emerging system is only more steady than single vision positioning system
It is fixed and steady.When being compared with VL, this algorithm is promoted in terms of positioning accuracy compared to VL, and the method proposed is in reality
Advantage in terms of when property is obvious.
In practical applications, the relevant parameter that this method is related to need to be configured.Fig. 3 is shown with hidden layer section
The increase position error of points is gradually reduced, when the increase of concealed nodes number to a certain extent when, position error no longer will obviously subtract
It is small.Locating effect when using sigmoid function as activation primitive when ratio sine function and radbas function is demonstrated simultaneously
More preferably.
The optimized parameter of this experiment is provided that node in hidden layer is 450, and activation primitive is sigmoid function, right
The quantity N of the SURF feature of each frame image zooming-outq30 are set as, image fuzzy Judgment threshold value is set as 0.8, while according to view
Feel that the threshold value of translational movement when information carries out static feedback is set as 0.06, using PCA to inertial positioning result and vision positioning knot
When fruit is weighted fusion, vision positioning result and the corresponding weight of inertial positioning result are 0.65 and 0.35.
The results show, under the parameter setting, this method real-time, stability and in terms of take
Obtain extraordinary effect.
Bibliography
[1] the WiFi indoor positioning technologies research Mianyang [D] of Chen Jiao based on RSSI ranging: Xinan Science and Technology Univ., 2015.
[2]Faulkner W T,Alwood R,Taylor D W A,et al.Altitude accuracy while
tracking pedestrians using a boot-mounted IMU[C].Position Location and
Navigation Symposium.IEEE,2010:90-96.
[3]Remazeilles A,Chaumette F.Image-based robot navigation from an
image memory[J].Robotics&Autonomous Systems,2007,55(4):345-356.
[4]Foxlin,Eric.Pedestrian Tracking with Shoe-Mounted Inertial Sensors
[J].IEEE Computer Graphics and Applications,2005,25(6):38-46.
[5]C.Piciarelli,‘Visual Indoor Localization in Known Environments’,
IEEE Signal Process.Lett.,2016,23,(10),pp.1330-1334.
[6]Fan Y G,Li P,Song Z H.KPCA based on feature samples for fault
detection[J].Kongzhi yu Juece/Control and Decision,2005,20(12):1415-1418+
1422.
The embodiment of the present invention to the model of each device in addition to doing specified otherwise, the model of other devices with no restrictions,
As long as the device of above-mentioned function can be completed.
It will be appreciated by those skilled in the art that attached drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Serial number is for illustration only, does not represent the advantages or disadvantages of the embodiments.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (4)
1. a kind of Weighted Fusion indoor orientation method based on PCA, which is characterized in that the described method comprises the following steps:
ELM model is initialized, training sample set is input in ELM neural network, using ELM regression algorithm to training sample set
Establish neural network model;
Test sample collection is input in trained neural network model, the relative displacement between consecutive frame image is obtained, to phase
Displacement result is integrated, the position of frame image is obtained;
Image fuzzy Judgment is introduced, calculates the position of current frame image as frame using the position of Inertia information and previous frame image
Framing result;
Drift correction is carried out to original acceleration measuring signal using visual information, the acceleration after drift correction is carried out secondary
Integral, obtains inertial positioning result;
Weight distribution is carried out to frame image, inertial positioning result using PCA, obtains final positioning result.
2. a kind of Weighted Fusion indoor orientation method based on PCA according to claim 1, which is characterized in that the instruction
Practice sample set specifically:
The SURF descriptor for extracting image, to the N number of SURF characteristic point of every frame image zooming-out, and carries out Feature Points Matching;Using
RANSAC algorithm handles matching result, removes Mismatching point, obtains affine transformation matrix;Each frame image is calculated under
Relative displacement between one frame image true coordinate;
It is inputted affine matrix as training, relative displacement establishes training sample set as training output.
3. a kind of Weighted Fusion indoor orientation method based on PCA according to claim 1, which is characterized in that described to draw
Enter image fuzzy Judgment, the position for calculating current frame image using the position of Inertia information and previous frame image is fixed as frame image
Position result specifically:
This index of introduced feature matching rate NS, by can be used for measuring whether current image obscures to NS given threshold;
When NS is less than threshold value, determine that the frame image is fuzzy, using the positioning result of the previous frame obtained by ELM, in conjunction with acceleration
Degree, the Inertia informations such as gyroscope obtain the position of present frame.
4. a kind of Weighted Fusion indoor orientation method based on PCA according to claim 1, which is characterized in that the benefit
Drift correction is carried out to original acceleration measuring signal with visual information, the acceleration after drift correction is subjected to quadratic integral tool
Body are as follows:
By the threshold value of translational movement between setting consecutive frame image, to distinguish motion state and stationary state;When translation vector is less than
When threshold value, judgement is currently at stationary state, to carry out drift correction, otherwise regards as motion state and keeps its former acceleration
Be worth it is constant, to after drift correction acceleration carry out quadratic integral obtain final position.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910636664.XA CN110458887B (en) | 2019-07-15 | 2019-07-15 | Weighted fusion indoor positioning method based on PCA |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910636664.XA CN110458887B (en) | 2019-07-15 | 2019-07-15 | Weighted fusion indoor positioning method based on PCA |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110458887A true CN110458887A (en) | 2019-11-15 |
CN110458887B CN110458887B (en) | 2022-12-06 |
Family
ID=68481248
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910636664.XA Active CN110458887B (en) | 2019-07-15 | 2019-07-15 | Weighted fusion indoor positioning method based on PCA |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110458887B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113111873A (en) * | 2021-03-25 | 2021-07-13 | 贵州电网有限责任公司 | PCA-based weighted fusion mobile robot positioning method |
Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104867138A (en) * | 2015-05-07 | 2015-08-26 | 天津大学 | Principal component analysis (PCA) and genetic algorithm (GA)-extreme learning machine (ELM)-based three-dimensional image quality objective evaluation method |
CN106447691A (en) * | 2016-07-19 | 2017-02-22 | 西安电子科技大学 | Weighted extreme learning machine video target tracking method based on weighted multi-example learning |
US20170103541A1 (en) * | 2015-10-12 | 2017-04-13 | Xsens Holding B.V. | Integration of Inertial Tracking and Position Aiding for Motion Capture |
CN106780570A (en) * | 2016-12-21 | 2017-05-31 | 中国航天科工集团第四研究院指挥自动化技术研发与应用中心 | A kind of real-time modeling method method based on on-line study |
CN107016319A (en) * | 2016-01-27 | 2017-08-04 | 北京三星通信技术研究有限公司 | A kind of key point localization method and device |
US20170345161A1 (en) * | 2016-05-31 | 2017-11-30 | Kabushiki Kaisha Toshiba | Information processing apparatus and method |
CN107615211A (en) * | 2015-05-23 | 2018-01-19 | 深圳市大疆创新科技有限公司 | Merged using the sensor of inertial sensor and imaging sensor |
CN108090921A (en) * | 2016-11-23 | 2018-05-29 | 中国科学院沈阳自动化研究所 | Monocular vision and the adaptive indoor orientation method of IMU fusions |
US10043076B1 (en) * | 2016-08-29 | 2018-08-07 | PerceptIn, Inc. | Visual-inertial positional awareness for autonomous and non-autonomous tracking |
US20180266828A1 (en) * | 2017-03-14 | 2018-09-20 | Trimble Inc. | Integrated vision-based and inertial sensor systems for use in vehicle navigation |
CN108716917A (en) * | 2018-04-16 | 2018-10-30 | 天津大学 | A kind of indoor orientation method merging inertia and visual information based on ELM |
CN108827341A (en) * | 2017-03-24 | 2018-11-16 | 快图有限公司 | The method of the deviation in Inertial Measurement Unit for determining image collecting device |
CN108921893A (en) * | 2018-04-24 | 2018-11-30 | 华南理工大学 | A kind of image cloud computing method and system based on online deep learning SLAM |
CN108981687A (en) * | 2018-05-07 | 2018-12-11 | 清华大学 | A kind of indoor orientation method that vision is merged with inertia |
CN109615056A (en) * | 2018-10-09 | 2019-04-12 | 天津大学 | A kind of visible light localization method based on particle group optimizing extreme learning machine |
CN109945858A (en) * | 2019-03-20 | 2019-06-28 | 浙江零跑科技有限公司 | It parks the multi-sensor fusion localization method of Driving Scene for low speed |
CN109993113A (en) * | 2019-03-29 | 2019-07-09 | 东北大学 | A kind of position and orientation estimation method based on the fusion of RGB-D and IMU information |
-
2019
- 2019-07-15 CN CN201910636664.XA patent/CN110458887B/en active Active
Patent Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104867138A (en) * | 2015-05-07 | 2015-08-26 | 天津大学 | Principal component analysis (PCA) and genetic algorithm (GA)-extreme learning machine (ELM)-based three-dimensional image quality objective evaluation method |
CN107615211A (en) * | 2015-05-23 | 2018-01-19 | 深圳市大疆创新科技有限公司 | Merged using the sensor of inertial sensor and imaging sensor |
US20170103541A1 (en) * | 2015-10-12 | 2017-04-13 | Xsens Holding B.V. | Integration of Inertial Tracking and Position Aiding for Motion Capture |
CN107016319A (en) * | 2016-01-27 | 2017-08-04 | 北京三星通信技术研究有限公司 | A kind of key point localization method and device |
US20170345161A1 (en) * | 2016-05-31 | 2017-11-30 | Kabushiki Kaisha Toshiba | Information processing apparatus and method |
CN106447691A (en) * | 2016-07-19 | 2017-02-22 | 西安电子科技大学 | Weighted extreme learning machine video target tracking method based on weighted multi-example learning |
US10043076B1 (en) * | 2016-08-29 | 2018-08-07 | PerceptIn, Inc. | Visual-inertial positional awareness for autonomous and non-autonomous tracking |
CN108090921A (en) * | 2016-11-23 | 2018-05-29 | 中国科学院沈阳自动化研究所 | Monocular vision and the adaptive indoor orientation method of IMU fusions |
CN106780570A (en) * | 2016-12-21 | 2017-05-31 | 中国航天科工集团第四研究院指挥自动化技术研发与应用中心 | A kind of real-time modeling method method based on on-line study |
US20180266828A1 (en) * | 2017-03-14 | 2018-09-20 | Trimble Inc. | Integrated vision-based and inertial sensor systems for use in vehicle navigation |
CN108827341A (en) * | 2017-03-24 | 2018-11-16 | 快图有限公司 | The method of the deviation in Inertial Measurement Unit for determining image collecting device |
CN108716917A (en) * | 2018-04-16 | 2018-10-30 | 天津大学 | A kind of indoor orientation method merging inertia and visual information based on ELM |
CN108921893A (en) * | 2018-04-24 | 2018-11-30 | 华南理工大学 | A kind of image cloud computing method and system based on online deep learning SLAM |
CN108981687A (en) * | 2018-05-07 | 2018-12-11 | 清华大学 | A kind of indoor orientation method that vision is merged with inertia |
CN109615056A (en) * | 2018-10-09 | 2019-04-12 | 天津大学 | A kind of visible light localization method based on particle group optimizing extreme learning machine |
CN109945858A (en) * | 2019-03-20 | 2019-06-28 | 浙江零跑科技有限公司 | It parks the multi-sensor fusion localization method of Driving Scene for low speed |
CN109993113A (en) * | 2019-03-29 | 2019-07-09 | 东北大学 | A kind of position and orientation estimation method based on the fusion of RGB-D and IMU information |
Non-Patent Citations (5)
Title |
---|
GYÖRGY DÁN; MUHAMMAD ALTAMASH KHAN; VIKTORIA FODOR: "《Characterization of SURF and BRISK Interest Point Distribution for Distributed Feature Extraction in Visual Sensor Networks》", 《IEEE TRANSACTIONS ON MULTIMEDIA》 * |
徐岩等: "一种融合加权ELM和AdaBoost的交通标志识别算法", 《小型微型计算机***》 * |
徐龙阳等: "基于神经网络的多传感器融合PDR定位方法", 《传感技术学报》 * |
汪剑鸣等: "室内惯性/视觉组合导航地面图像分割算法", 《中国惯性技术学报》 * |
王哲等: "蝗虫群优化和极限学习机相结合的RFID室内定位算法", 《计算机科学》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113111873A (en) * | 2021-03-25 | 2021-07-13 | 贵州电网有限责任公司 | PCA-based weighted fusion mobile robot positioning method |
Also Published As
Publication number | Publication date |
---|---|
CN110458887B (en) | 2022-12-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110125928A (en) | A kind of binocular inertial navigation SLAM system carrying out characteristic matching based on before and after frames | |
US20150235367A1 (en) | Method of determining a position and orientation of a device associated with a capturing device for capturing at least one image | |
CN112907678B (en) | Vehicle-mounted camera external parameter attitude dynamic estimation method and device and computer equipment | |
CN110553648A (en) | method and system for indoor navigation | |
CN111912416A (en) | Method, device and equipment for positioning equipment | |
CN110617821A (en) | Positioning method, positioning device and storage medium | |
Mseddi et al. | YOLOv5 based visual localization for autonomous vehicles | |
CN108303094A (en) | The Position Fixing Navigation System and its positioning navigation method of array are merged based on multiple vision sensor | |
CN109445599A (en) | Interaction pen detection method and 3D interactive system | |
Deigmoeller et al. | Stereo visual odometry without temporal filtering | |
CN112731503B (en) | Pose estimation method and system based on front end tight coupling | |
CN110458887A (en) | A kind of Weighted Fusion indoor orientation method based on PCA | |
Bonilla et al. | Pedestrian dead reckoning towards indoor location based applications | |
Irmisch et al. | Robust visual-inertial odometry in dynamic environments using semantic segmentation for feature selection | |
Yusefi et al. | A Generalizable D-VIO and Its Fusion with GNSS/IMU for Improved Autonomous Vehicle Localization | |
CN114608560A (en) | Passive combined indoor positioning system and method based on intelligent terminal sensor | |
Jonker et al. | Philosophies and technologies for ambient aware devices in wearable computing grids | |
Jung et al. | U-VIO: tightly coupled UWB visual inertial odometry for robust localization | |
Park et al. | A simulation based method for vehicle motion prediction | |
CN117593650B (en) | Moving point filtering vision SLAM method based on 4D millimeter wave radar and SAM image segmentation | |
Han | Optical flow/ins navigation system in four-rotor | |
Xia et al. | YOLO-Based Semantic Segmentation for Dynamic Removal in Visual-Inertial SLAM | |
Aufderheide et al. | A visual-inertial approach for camera egomotion estimation and simultaneous recovery of scene structure | |
Hashimoto et al. | Self-localization from a 360-Degree Camera Based on the Deep Neural Network | |
CN117576218B (en) | Self-adaptive visual inertial navigation odometer output 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 |