CN108981693A - VIO fast joint initial method based on monocular camera - Google Patents
VIO fast joint initial method based on monocular camera Download PDFInfo
- Publication number
- CN108981693A CN108981693A CN201810685385.8A CN201810685385A CN108981693A CN 108981693 A CN108981693 A CN 108981693A CN 201810685385 A CN201810685385 A CN 201810685385A CN 108981693 A CN108981693 A CN 108981693A
- Authority
- CN
- China
- Prior art keywords
- acceleration
- gravity
- imu
- speed
- rotation
- 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
-
- 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
-
- 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
- G06T7/75—Determining position or orientation of objects or cameras using feature-based methods involving models
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Navigation (AREA)
Abstract
The invention discloses a kind of VIO fast joint initial method based on monocular camera, includes the following steps: that (1) handles several seconds videos with ORB-SLAM to obtain initial pose and several key frames；(2) pass through the key frame and IMU pre-integration building cost function calculation gyroscope deviation in step (1)；(3) gravitational accelerometer biasing, acceleration of gravity, acceleration of gravity calibration and gravitational accelerometer biasing are sought by resolving the linear model of continuous interframe；(4) scale and velocity information are found out with parameters separated by the acceleration of gravity calibration in step (3).Initial method of the invention has initialization speed faster, and to location precision very little, method proposed by the present invention can will be controlled within 10 seconds initialization time；The speed for improving joint initialization is of great significance for improving real-time.
Description
Technical field
The present invention relates to Robot visual location field of navigation technology, especially a kind of VIO based on monocular camera is quick
Joint initial method.
Background technique
The state estimation of intelligent mobile terminal has become computer vision and the hot topic of robot field, because it
It can be applied to the emerging technologies such as minute vehicle, autonomous driving vehicle, virtual reality and augmented reality.Traditional intelligent mobile is whole
End positioning is divided into outdoor positioning and indoor positioning.Outdoor positioning includes satellite positioning and base station location, and indoor positioning relies primarily on
Multi-sensor Fusion.In different sensing systems, vision inertia system due to its is small in size, at low cost, potentiality are big and by
Extensive concern is arrived.There are two types of the main methods of estimation pose for vision inertia system: nonlinear optimization method and recursive filtering
Method.Single camera vision system can provide enough environmental informations to complete 3D and build figure and positioning, but cannot restore dimensional information,
Therefore its application in real world in terms of robot is limited.Since IMU sensor provides displacement information, and
It can estimate gravity direction and Observable absolute pitch angle, this allows monocular vision inertia system to restore dimensional information.
Initial state estimation is an important step of vision inertia odometer, and most of existing monocular VINS methods exist
Do not have to be that can not carry out pinpoint in the case where good Initial state estimation, and if estimation is initial due to lacking
Condition and failure, they can not restore movement.The initial method that existing VIO is provided is camera posture, scale, speed
Degree, gravity, gyroscope and accelerometer deviation provide best solution, but initialize that convergence needs 10 seconds or more when
Between.Another initial method that the VINS that port University of Science and Technology proposes is introduced provides the initialization procedure of a robust, can be never
The original state guidance system known, but a large amount of computing resources are needed to calculate gravity direction.
Summary of the invention
Technical problem to be solved by the present invention lies in provide a kind of VIO fast joint initialization based on monocular camera
There is method faster initialization speed can will control initialization time within 10 seconds location precision very little.
In order to solve the above technical problems, the present invention provides a kind of VIO fast joint initial method based on monocular camera,
Include the following steps:
(1) several seconds videos are handled with ORB-SLAM to obtain initial pose and several key frames；
(2) pass through the key frame and IMU pre-integration building cost function calculation gyroscope deviation in step (1)；
(3) gravitational accelerometer biasing, acceleration of gravity, gravity acceleration are sought by resolving the linear model of continuous interframe
Degree calibration is biased with gravitational accelerometer；
(4) scale and velocity information are found out with parameters separated by the acceleration of gravity calibration in step (3).
Preferably, in step (3), gravitational accelerometer biasing estimation specifically: the movement between two continuous key frames
The pre-integration Δ R measured by two interframe, Δ v, Δ p are described, by pre-integration equation:
I and i+1 is frame number respectively,WithIt is rotation of the i+1 frame of IMU under world coordinate system respectively
Turn, speed and translation, Δ Ri,i+1、Δvi,i+1With Δ pi,i+1Be respectively the i-th and i+1 interframe of IMU world coordinate system under rotation
Turn, speed and translation increment, Δ ti,i+1It is the time difference,WithIt is to translate relative to the refined each of angular speed and acceleration respectively
Than matrix,WithIt is Jacobin matrix of the speed relative to angular speed and acceleration respectively,It is rotation relative to angle speed
The Jacobin matrix of degree,WithIndicate the accelerometer and offset of gyroscope of the i-th frame, gwIt is that gravity in world coordinate system adds
Velocity vector, w are table systems of the world, and b is IMU coordinate system, biAnd bi+1It is the IMU coordinate system of i-th and i+1 frame respectively；Join in the world
It examines and defines the rotation that IMU measurement obtains in coordinate systemTranslationAnd speed
Having measured rotation and gyroscope deviation can have been estimated according to two continuous key frames, for all continuous passes
Key frame pair is optimized by minimizing the difference between the relative rotation that gyroscope calculates IMU pre-integration and ORB-SLAM is calculated
Offset of gyroscope:
N is the quantity of key frame,It is the spin matrix from IMU frame to world's frame, is gone to from camera coordinates
The spin matrix of world coordinate systemIt is calculated by ORB-SLAM, Δ Ri,i+1It is calculated by pre-integration two continuous
Spin matrix between key frame, the present invention resolve formula (2) with gauss-newton method.
Preferably, in step (3), acceleration of gravity estimation specifically: compensation can be passed through later by obtaining gyroscope deviation
Offset of gyroscope accurately calculates position, speed and the rotation of camera with pre-integration again, the video camera rail calculated by ORB-SLAM
The scale-value of mark is a relative value, so needing to introduce a scale when interframe of the actual estimated from IMU to camera is displaced
Value s,WithIt is the translation and rotation under camera to world coordinate system respectively,It is the translation parameters between IMU and camera；
In conjunction with formula (3) and formula (1), it is available to ignore acceleration biasing:
Target is to estimate gravity and unknown scale with system of linear equations, in order to reduce the complexity of operation, in calculating process
The length velocity relation in two relationships (4) and (3) between three continuous key frames of consideration is with release rate, and variables separation is such as
Under:
All relationships between three continuous key frames (5) are added to measurement model A3(N-2)×4x4×1=B3(N-2)×1In,
Can by singular value decomposition SVD come solving model to obtain scale factor s*And gravitational vectorsBecause there is 3 (N-2) a sides
Journey and 4 unknown numbers, at least need 4 key frames.
Preferably, in step (3), acceleration of gravity calibration is biased with gravitational accelerometer specifically: considerable in order to increase
The property surveyed increases accelerometer deviation and gravity constant G, introduces reference to gravitational acceleration directionIt has been computed
Gravity directionSpin matrix can be calculated as follows:
Acceleration of gravity calibration are as follows:
Since the rotation in the direction z does not interfere with gw, RWIIt can be parameterized by the rotation of two axis of x and y, pass through introducing
One small disturbance δ θ is to gwIt optimizes
By first order Taylor approximation:
Convolution (9) and (4) simultaneously consider that accelerometer biases:
Joint type (5) after three key frames is observed, available:
All relational expressions (11) between three continuous key frames are added to by the wherein same formula of the expression way of λ (i) (5)
Measurement model A3(N-2)×6x6×1=B3(N-1)×1In, it again may be by SVD and decompose to obtain dimensional information S*, acceleration of gravity correction
InformationAnd acceleration biasingDue to having used a equation of 3 (N-2) and 6 unknown quantitys, clearing equation at least needs
4 key frames.
Preferably, in step (4), scale and velocity information are estimated specifically: by speed, acceleration of gravity and scale connection
It stands as state vector,Etc. be IMU at the i moment in biSpeed under coordinate system:
WhereinFor the speed under i-th of IMU frame, consider that two continuous key frame formulas (4) can indicate (12) are as follows:
By linear measurement model available after parameters separated:
By calculating least square model:
The velocity information of scale and each frame can be obtained.
The invention has the benefit that initial method of the invention has initialization speed faster, to positioning accuracy
Very little is influenced, method proposed by the present invention can will control within 10 seconds initialization time；Improve the speed of joint initialization
It is of great significance for improving real-time.
Detailed description of the invention
Fig. 1 is method flow schematic diagram of the invention.
Fig. 2 is that scale, gravity and the IMU that the present invention estimates bias schematic diagram.
Fig. 3 is VIO track and real trace of the present invention using initial method and conventional initialization method of the invention
Between root-mean-square error (RMSE) contrast schematic diagram.
Specific embodiment
As shown in Figure 1, the VIO fast joint initial method based on monocular camera, includes the following steps:
(1) several seconds videos are handled with ORB-SLAM to obtain initial pose and several key frames；
(2) pass through the key frame and IMU pre-integration building cost function calculation gyroscope deviation in step (1)；
(3) gravitational accelerometer biasing, acceleration of gravity, gravity acceleration are sought by resolving the linear model of continuous interframe
Degree calibration is biased with gravitational accelerometer；
(4) scale and velocity information are found out with parameters separated by the acceleration of gravity calibration in step (3).
Step 1: gravitational accelerometer biasing estimation
Movement the pre-integration Δ R measured by two interframe, Δ v, Δ p between two continuous key frames are described.
By pre-integration equation:
I and i+1 is frame number respectively,WithIt is rotation of the i+1 frame of IMU under world coordinate system respectively
Turn, speed and translation, Δ Ri,i+1、Δvi,i+1With Δ pi,i+1Be respectively the i-th and i+1 interframe of IMU world coordinate system under rotation
Turn, speed and translation increment, Δ ti,i+1It is the time difference,WithIt is to translate relative to the refined each of angular speed and acceleration respectively
Than matrix,WithIt is Jacobin matrix of the speed relative to angular speed and acceleration respectively,It is rotation relative to angle speed
The Jacobin matrix of degree,WithIndicate the accelerometer and offset of gyroscope of the i-th frame, gwIt is that gravity in world coordinate system adds
Velocity vector, w are table systems of the world, and b is IMU coordinate system, biAnd bi+1It is the IMU coordinate system of i-th and i+1 frame respectively；Join in the world
It examines and defines the rotation that IMU measurement obtains in coordinate systemTranslationAnd speed
Having measured rotation and gyroscope deviation can have been estimated according to two continuous key frames.For all continuous passes
Key frame pair, the present invention is by minimizing the difference between the relative rotation that gyroscope calculates IMU pre-integration and ORB-SLAM is calculated
To optimize offset of gyroscope:
N is the quantity of key frame.It is the spin matrix from IMU frame to world's frame.It is gone to from camera coordinates
The spin matrix of world coordinate systemIt is calculated by ORB-SLAM.ΔRi,i+1It is calculated by pre-integration two continuous
Spin matrix between key frame.The present invention resolves formula (2) with gauss-newton method.
Step 2: acceleration of gravity is estimated
The position of camera can accurately be calculated with pre-integration again by compensating offset of gyroscope by obtaining gyroscope deviation later
It sets, speed and rotation.A relative value by the scale-value of video camera track that ORB-SLAM is calculated, thus when actual estimated from
The interframe of IMU to camera needs to introduce a scale-value s when being displaced,WithIt is flat under camera to world coordinate system respectively
It moves and rotates,It is the translation parameters between IMU and camera.
In conjunction with formula (3) and formula (1), it is available to ignore acceleration biasing:
This target calculated is to estimate gravity and unknown scale with system of linear equations.In order to reduce the complexity of operation,
In calculating process consider three continuous key frames between two relationships (4) and (3) in length velocity relation with release rate, and
Variables separation is as follows:
All relationships between three continuous key frames (5) are added to measurement model A3(N-2)×4x4×1=B3(N-2)×1In.
We can by singular value decomposition (SVD) come solving model to obtain scale factor s*And gravitational vectorsBecause we have 3
(N-2) a equation and 4 unknown numbers, we at least need 4 key frames.
Step 3: acceleration of gravity calibration and acceleration biasing estimation
In order to increase observability, consider to increase accelerometer deviation and gravity constant G.Introduce reference to gravitational acceleration side
ToThe gravity direction being computedWe can calculate as follows spin matrix:
Acceleration of gravity calibration are as follows:
Since the rotation in the direction z does not interfere with gw, RWIIt can be parameterized by the rotation of two axis of x and y.The present invention is logical
It crosses and introduces a small disturbance δ θ to gwIt optimizes
By first order Taylor approximation:
Convolution (9) and (4) simultaneously consider that accelerometer biases:
Joint type (5) after three key frames is observed, available:
The wherein same formula of the expression way of λ (i) (5).All relational expressions (11) between three continuous key frames are added to
Measurement model A3(N-2)×6x6×1=B3(N-1)×1In.It again may be by SVD to decompose to obtain dimensional information s*, acceleration of gravity correction
InformationAnd acceleration biasingDue to having used a equation of 3 (N-2) and 6 unknown quantitys, clearing equation at least needs
4 key frames
Step 4: scale and velocity estimation
By speed, acceleration of gravity and scale simultaneous are state vector,Etc. be IMU at the i moment in biUnder coordinate system
Speed:
WhereinFor the speed under i-th of IMU frame, consider that two continuous key frame formulas (4) can indicate (12) are as follows:
By linear measurement model available after parameters separated:
By calculating least square model:
The velocity information of scale and each frame can be obtained.
The present invention is using the data in EuRoC data set to proposition based on the fast of Visual-Inertial ORB-SLAM
Fast initial method is tested.The data set is collected in the micro air in two rooms and in industrial environment that flies
Device (MAV) institute's acquired image information and IMU information.Data set includes image, synchronous IMU measurement (ADIS16448,
200Hz) and accurately movement and structural floor are live (VICON and Leica MS50).It includes 11 sequences, according to illumination
Situation, texture and movement velocity are classified as simply, medium and difficult.
Present invention employs sequence V1_02_medium to run fast initializing method.Fig. 2 be estimation scale, gravity and
IMU biasing.As can be seen that all variables converge to stabilization in 10 seconds, and VIO needs 15 seconds.For EuRoC data set, I
Observe that MAV can realize accurate initialization after the 10s that flown.
Fig. 3 compared the track VIO using initial method of the invention and conventional initialization method and real trace it
Between root-mean-square error (RMSE).
Table 1 counted the track VIO using initial method of the invention and conventional initialization method and real trace it
Between root-mean-square error (RMSE).It can be seen that the method applied in the present invention precision differed with tradition VIO precision 2 centimetres with
It is interior, and the present invention tests and carries out the data group V1_03_diffcult that initialization will fail using traditional approach, finally may be used
To see that the initialization mode that uses of the present invention successfully initializes data sequence, thus demonstrate of the invention initial
Change mode has better robustness.
Root-mean-square error (RMSE) statistics between table 1 and real trace
Claims (5)
1. the VIO fast joint initial method based on monocular camera, which comprises the steps of:
(1) several seconds videos are handled with ORB-SLAM to obtain initial pose and several key frames；
(2) pass through the key frame and IMU pre-integration building cost function calculation gyroscope deviation in step (1)；
(3) gravitational accelerometer biasing, acceleration of gravity, acceleration of gravity school are sought by resolving the linear model of continuous interframe
It is quasi- to be biased with gravitational accelerometer；
(4) scale and velocity information are found out with parameters separated by the acceleration of gravity calibration in step (3).
2. the VIO fast joint initial method based on monocular camera as described in claim 1, which is characterized in that step (3)
In, gravitational accelerometer biasing estimation specifically: the movement between two continuous key frames can be pre- measured by two interframe
Δ R is integrated, Δ v, Δ p are described, by pre-integration equation:
I and i+1 is the moment respectively,WithBe respectively rotation of the i+1 moment of IMU under world coordinate system,
Speed and translation, Δ Ri,i+1、Δvi,i+1With Δ pi,i+1The rotation under world coordinate system between i the and i+1 moment for being respectively IMU,
Speed and translation increment, Δ ti,i+1It is the time difference,WithIt is the Jacobi translated relative to angular speed and acceleration respectively
Matrix,WithIt is Jacobin matrix of the speed relative to angular speed and acceleration respectively,It is rotation relative to angular speed
Jacobin matrix, baAnd bgIndicate accelerometer and offset of gyroscope, gwIt is the acceleration of gravity vector in world coordinate system, w
It is table system of the world, b is IMU coordinate system, biAnd bi+1It is the IMU coordinate system at i-th and i+1 moment respectively；In world's reference frame
The middle rotation for defining IMU measurement and obtainingTranslationAnd speed
Having measured rotation and gyroscope deviation can have been estimated according to two continuous key frames, for all continuous key frames
It is right, optimize gyro by minimizing the difference between the relative rotation that gyroscope calculates IMU pre-integration and ORB-SLAM is calculated
Instrument biasing:
N is the quantity of key frame,It is the spin matrix from IMU frame to world's frame, goes to the world from camera coordinates
The spin matrix of coordinate systemIt is calculated by ORB-SLAM, Δ Ri,i+1It is the two continuous keys calculated by pre-integration
Spin matrix between frame, the present invention resolve formula (2) with gauss-newton method.
3. the VIO fast joint initial method based on monocular camera as described in claim 1, which is characterized in that step (3)
In, acceleration of gravity estimation specifically: obtaining gyroscope deviation can use pre-integration quasi- again by compensating offset of gyroscope later
Position, speed and the rotation for really calculating camera, the scale-value by the video camera track calculated ORB-SLAM is a relative value,
So need to introduce a scale-value s when interframe of the actual estimated from IMU to camera is displaced,WithIt is that camera arrives respectively
Translation and rotation under world coordinate system,It is the translation parameters between IMU and camera；
In conjunction with formula (3) and formula (1), it is available to ignore acceleration biasing:
Target is to estimate that gravity and unknown scale consider in calculating process to reduce the complexity of operation with system of linear equations
The length velocity relation in two relationships (4) and (3) between three continuous key frames is with release rate, and variables separation is as follows:
All relationships between three continuous key frames (5) are added to measurement model A3(N-2)×4x4×1=B3(N-2)×1In, it can be with
By singular value decomposition SVD come solving model to obtain scale factor s*And gravitational vectorsBecause having a equation of 3 (N-2) and 4
A unknown number at least needs 4 key frames.
4. the VIO fast joint initial method based on monocular camera as described in claim 1, which is characterized in that step (3)
In, acceleration of gravity calibration is biased with gravitational accelerometer specifically: in order to increase observability, increase accelerometer deviation and
Gravity constant G introduces reference to gravitational acceleration directionThe gravity direction being computedSpin matrix can be calculated as follows:
Acceleration of gravity calibration are as follows:
Since the rotation in the direction z does not interfere with gw, RWIIt can be parameterized by the rotation of two axis of x and y, by introducing one
Small disturbance δ θ is to gwIt optimizes
By first order Taylor approximation:
Convolution (9) and (4) simultaneously consider that accelerometer biases:
Joint type (5) after three key frames is observed, available:
All relational expressions (11) between three continuous key frames are added to measurement by the wherein same formula of the expression way of λ (i) (5)
Model A3(N-2)×6x6×1=B3(N-1)×1In, it again may be by SVD and decompose to obtain dimensional information s*, acceleration of gravity correction informationAnd acceleration biasingDue to having used a equation of 3 (N-2) and 6 unknown quantitys, clearing equation at least needs 4
Key frame.
5. the VIO fast joint initial method based on monocular camera as described in claim 1, which is characterized in that step (4)
In, scale and velocity information are estimated specifically: by speed, acceleration of gravity and scale simultaneous are state vector,Etc. being
IMU is at the i moment in biSpeed under coordinate system:
WhereinFor the speed under i-th of IMU frame, consider that two continuous key frame formulas (4) can indicate (12) are as follows:
By linear measurement model available after parameters separated:
By calculating least square model:
The velocity information of scale and each frame can be obtained.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810241538 | 2018-03-22 | ||
| CN201810241538X | 2018-03-22 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108981693A true CN108981693A (en) | 2018-12-11 |
| CN108981693B CN108981693B (en) | 2021-10-29 |
Family
ID=64539248
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810685385.8A Active CN108981693B (en) | 2018-03-22 | 2018-06-28 | VIO rapid joint initialization method based on monocular camera |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108981693B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109767470A (en) * | 2019-01-07 | 2019-05-17 | 浙江商汤科技开发有限公司 | A kind of tracking system initial method and terminal device |
| CN109887003A (en) * | 2019-01-23 | 2019-06-14 | 亮风台（上海）信息科技有限公司 | A kind of method and apparatus initialized for carrying out three-dimensional tracking |
| CN110030994A (en) * | 2019-03-21 | 2019-07-19 | 东南大学 | A kind of robustness vision inertia close coupling localization method based on monocular |
| CN110068875A (en) * | 2019-04-30 | 2019-07-30 | 深圳市万普拉斯科技有限公司 | Geomagnetic sensor calibration method, mobile terminal and computer readable storage medium |
| CN110084832A (en) * | 2019-04-25 | 2019-08-02 | 亮风台（上海）信息科技有限公司 | Correcting method, device, system, equipment and the storage medium of camera pose |
| CN110702107A (en) * | 2019-10-22 | 2020-01-17 | 北京维盛泰科科技有限公司 | Monocular vision inertial combination positioning navigation method |
| CN111578937A (en) * | 2020-05-29 | 2020-08-25 | 天津工业大学 | Visual inertial odometer system capable of optimizing external parameters simultaneously |
| CN111665826A (en) * | 2019-03-06 | 2020-09-15 | 北京奇虎科技有限公司 | Depth map acquisition method based on laser radar and monocular camera and sweeping robot |
| CN112862768A (en) * | 2021-01-28 | 2021-05-28 | 重庆邮电大学 | Adaptive monocular VIO (visual image analysis) initialization method based on point-line characteristics |
| WO2022061799A1 (en) * | 2020-09-27 | 2022-03-31 | 深圳市大疆创新科技有限公司 | Pose estimation method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103033189A (en) * | 2012-12-26 | 2013-04-10 | 北京航空航天大学 | Inertia/vision integrated navigation method for deep-space detection patrolling device |
| CN105931275A (en) * | 2016-05-23 | 2016-09-07 | 北京暴风魔镜科技有限公司 | Monocular and IMU fused stable motion tracking method and device based on mobile terminal |
| CN107025668A (en) * | 2017-03-30 | 2017-08-08 | 华南理工大学 | A kind of design method of the visual odometry based on depth camera |
| CN107193279A (en) * | 2017-05-09 | 2017-09-22 | 复旦大学 | Robot localization and map structuring system based on monocular vision and IMU information |
| CN107607111A (en) * | 2017-09-07 | 2018-01-19 | 驭势科技（北京）有限公司 | Acceleration biases method of estimation and device, vision inertia odometer and its application |
| CN107767425A (en) * | 2017-10-31 | 2018-03-06 | 南京维睛视空信息科技有限公司 | A kind of mobile terminal AR methods based on monocular vio |
-
2018
- 2018-06-28 CN CN201810685385.8A patent/CN108981693B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103033189A (en) * | 2012-12-26 | 2013-04-10 | 北京航空航天大学 | Inertia/vision integrated navigation method for deep-space detection patrolling device |
| CN105931275A (en) * | 2016-05-23 | 2016-09-07 | 北京暴风魔镜科技有限公司 | Monocular and IMU fused stable motion tracking method and device based on mobile terminal |
| CN107025668A (en) * | 2017-03-30 | 2017-08-08 | 华南理工大学 | A kind of design method of the visual odometry based on depth camera |
| CN107193279A (en) * | 2017-05-09 | 2017-09-22 | 复旦大学 | Robot localization and map structuring system based on monocular vision and IMU information |
| CN107607111A (en) * | 2017-09-07 | 2018-01-19 | 驭势科技（北京）有限公司 | Acceleration biases method of estimation and device, vision inertia odometer and its application |
| CN107767425A (en) * | 2017-10-31 | 2018-03-06 | 南京维睛视空信息科技有限公司 | A kind of mobile terminal AR methods based on monocular vio |
Non-Patent Citations (3)
| Title |
|---|
| ANASTASIOS I.等: "A Multi-State Constraint Kalman Filter for Vision-aided Inertial Navigation", 《2007 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION》 * |
| LAURENT KNEIP等: "Deterministic Initialization of Metric State Estimation Filters for Loosely-Coupled Monocular Vision-Inertial Systems", 《2011 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS》 * |
| 逯建军等: "惯性/双目视觉里程计深组合导航方法", 《导航定位与授时》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109767470B (en) * | 2019-01-07 | 2021-03-02 | 浙江商汤科技开发有限公司 | Tracking system initialization method and terminal equipment |
| CN109767470A (en) * | 2019-01-07 | 2019-05-17 | 浙江商汤科技开发有限公司 | A kind of tracking system initial method and terminal device |
| CN109887003A (en) * | 2019-01-23 | 2019-06-14 | 亮风台（上海）信息科技有限公司 | A kind of method and apparatus initialized for carrying out three-dimensional tracking |
| CN109887003B (en) * | 2019-01-23 | 2021-11-19 | 亮风台（上海）信息科技有限公司 | Method and equipment for carrying out three-dimensional tracking initialization |
| CN111665826A (en) * | 2019-03-06 | 2020-09-15 | 北京奇虎科技有限公司 | Depth map acquisition method based on laser radar and monocular camera and sweeping robot |
| CN110030994A (en) * | 2019-03-21 | 2019-07-19 | 东南大学 | A kind of robustness vision inertia close coupling localization method based on monocular |
| CN110084832A (en) * | 2019-04-25 | 2019-08-02 | 亮风台（上海）信息科技有限公司 | Correcting method, device, system, equipment and the storage medium of camera pose |
| CN110084832B (en) * | 2019-04-25 | 2021-03-23 | 亮风台（上海）信息科技有限公司 | Method, device, system, equipment and storage medium for correcting camera pose |
| CN110068875A (en) * | 2019-04-30 | 2019-07-30 | 深圳市万普拉斯科技有限公司 | Geomagnetic sensor calibration method, mobile terminal and computer readable storage medium |
| CN110068875B (en) * | 2019-04-30 | 2021-06-08 | 深圳市万普拉斯科技有限公司 | Geomagnetic sensor calibration method, mobile terminal and computer-readable storage medium |
| CN110702107A (en) * | 2019-10-22 | 2020-01-17 | 北京维盛泰科科技有限公司 | Monocular vision inertial combination positioning navigation method |
| CN111578937A (en) * | 2020-05-29 | 2020-08-25 | 天津工业大学 | Visual inertial odometer system capable of optimizing external parameters simultaneously |
| CN111578937B (en) * | 2020-05-29 | 2024-01-09 | 上海新天策数字科技有限公司 | Visual inertial odometer system capable of simultaneously optimizing external parameters |
| WO2022061799A1 (en) * | 2020-09-27 | 2022-03-31 | 深圳市大疆创新科技有限公司 | Pose estimation method and device |
| CN112862768A (en) * | 2021-01-28 | 2021-05-28 | 重庆邮电大学 | Adaptive monocular VIO (visual image analysis) initialization method based on point-line characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108981693B (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108981693A (en) | VIO fast joint initial method based on monocular camera | |
| CN110030994B (en) | Monocular-based robust visual inertia tight coupling positioning method | |
| Weiss et al. | Real-time onboard visual-inertial state estimation and self-calibration of MAVs in unknown environments | |
| Cao et al. | Accurate position tracking with a single UWB anchor | |
| WO2021180128A1 (en) | Rgbd sensor and imu sensor-based positioning method | |
| CN110095116A (en) | A kind of localization method of vision positioning and inertial navigation combination based on LIFT | |
| CN108036785A (en) | A kind of aircraft position and orientation estimation method based on direct method and inertial navigation fusion | |
| Xiong et al. | G-VIDO: A vehicle dynamics and intermittent GNSS-aided visual-inertial state estimator for autonomous driving | |
| CN105953796A (en) | Stable motion tracking method and stable motion tracking device based on integration of simple camera and IMU (inertial measurement unit) of smart cellphone | |
| CN108846857A (en) | The measurement method and visual odometry of visual odometry | |
| Omari et al. | Metric visual-inertial navigation system using single optical flow feature | |
| Kang et al. | Vins-vehicle: A tightly-coupled vehicle dynamics extension to visual-inertial state estimator | |
| CN109798891A (en) | Inertial Measurement Unit calibration system based on high-precision motion capture system | |
| CN109238277B (en) | Positioning method and device for visual inertial data depth fusion | |
| CN110533719A (en) | Augmented reality localization method and device based on environmental visual Feature point recognition technology | |
| CN114690229A (en) | GPS-fused mobile robot visual inertial navigation method | |
| CN112284379B (en) | Inertial pre-integration method of combined motion measurement system based on nonlinear integral compensation | |
| CN112945233A (en) | Global drift-free autonomous robot simultaneous positioning and map building method | |
| Hu et al. | Tightly coupled visual-inertial-UWB indoor localization system with multiple position-unknown anchors | |
| Irmisch et al. | Simulation framework for a visual-inertial navigation system | |
| Ling et al. | RGB-D inertial odometry for indoor robot via keyframe-based nonlinear optimization | |
| Kehoe et al. | Partial aircraft state estimation from optical flow using non-model-based optimization | |
| Meier et al. | Detection and characterization of moving objects with aerial vehicles using inertial-optical flow | |
| CN111539982B (en) | Visual inertial navigation initialization method based on nonlinear optimization in mobile platform | |
| Ready et al. | Inertially aided visual odometry for miniature air vehicles in gps-denied environments |
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 |