CN110261877A - A kind of air-ground coordination vision navigation method and device for scheming optimization SLAM based on improvement - Google Patents
A kind of air-ground coordination vision navigation method and device for scheming optimization SLAM based on improvement Download PDFInfo
- Publication number
- CN110261877A CN110261877A CN201910561547.1A CN201910561547A CN110261877A CN 110261877 A CN110261877 A CN 110261877A CN 201910561547 A CN201910561547 A CN 201910561547A CN 110261877 A CN110261877 A CN 110261877A
- Authority
- CN
- China
- Prior art keywords
- optimization
- intelligent body
- algorithm
- matrix
- information
- 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
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000005457 optimization Methods 0.000 title claims abstract description 46
- 230000006872 improvement Effects 0.000 title abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 53
- 239000011159 matrix material Substances 0.000 claims abstract description 50
- 230000000007 visual effect Effects 0.000 claims abstract description 10
- 230000005540 biological transmission Effects 0.000 claims description 25
- 238000004804 winding Methods 0.000 claims description 24
- 238000004891 communication Methods 0.000 claims description 20
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 11
- 230000003993 interaction Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 235000001674 Agaricus brunnescens Nutrition 0.000 claims description 6
- 230000000977 initiatory effect Effects 0.000 claims description 6
- 125000002015 acyclic group Chemical group 0.000 claims description 5
- 230000004927 fusion Effects 0.000 claims description 4
- 230000007704 transition Effects 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims 1
- 238000013461 design Methods 0.000 abstract description 4
- 238000005516 engineering process Methods 0.000 abstract description 4
- 239000003795 chemical substances by application Substances 0.000 description 12
- 238000011160 research Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
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
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S11/00—Systems for determining distance or velocity not using reflection or reradiation
- G01S11/12—Systems for determining distance or velocity not using reflection or reradiation using electromagnetic waves other than radio waves
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/42—Determining position
-
- 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
-
- 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
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/20112—Image segmentation details
- G06T2207/20164—Salient point detection; Corner detection
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Automation & Control Theory (AREA)
- Electromagnetism (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention discloses a kind of based on the air-ground coordination vision navigation method and device that improve figure optimization SLAM, belong to field of navigation technology, the present invention respectively constitutes aerial and ground intelligent body using sensor combination dynamic structure, in each intelligent body, it is equipped with the improvement figure optimization air-ground coordination vision navigation method that the present invention designs, this method includes four modules: signal acquisition module is used for the visual signal under acquisition position signal and location circumstances；Various signals are converted to matrix for handling collected information by front end processing block；Back end processing module, which is used, carries out seat estimation and state update by the matrix that front end provides；Improvement figure optimization algorithm module is for accelerating rear module calculating speed, reduce the calculating pressure of rear module, by using above method, the optimization that Position Fixing Navigation System is carried out to the multi-agent system being placed under circumstances not known improves the positioning accuracy of navigation system, accelerates locating speed, computation complexity is reduced simultaneously, the present invention solves the problems such as low in face of positioning accurate under circumstances not known in the prior art, positioning is slow, and error is big and calculated load is serious.
Description
Technical field
The invention belongs to field of navigation technology, and in particular to it is a kind of based on improvement figure optimization SLAM air-ground coordination vision lead
Boat method and device.
Background technique
Vision navigation system becomes research field more popular in navigation research in recent years.From 1986 immediately positioning with
Since map structuring (SLAM) proposes, swift and violent development has been obtained.And this technology is more applied to such as unmanned plane, unmanned vehicle etc.
In Intelligent unattended equipment.When mobile robot enters in a foreign environment, with needing the sensors towards ambient by itself
Figure is constructed, and is determined simultaneously from the position in map.Due to camera have it is small in size, it is light-weight, it is cheap etc.
Good characteristics, and video camera can obtain two-dimensional signal in scene, and obtain posture information and motion state by respective algorithms
Information, so that SLAM has significant progress.
Traditional monocular camera can not provide the information of enough dimensions for resolving, in precision side due to lacking depth
Face causes inefficient；Although and the problem of data dimension that binocular camera and depth camera solve, also increase hardware
Volume, so that can not play a role under the application scenes such as unmanned plane.Therefore, feature is improved on the basis of monocular camera
Simultaneously optimization algorithm can solve the problems, such as dimensional information to extracting method on the basis of not increasing equipment.Classical monocular vision
SLAM algorithm realizes positioning using the Kalman filtering (EKF) based on point feature and builds figure, and the main thought of this method is to make
With the three-dimensional coordinate of characteristic point in the posture information and map of state vector storage camera, observation is indicated with probability density function
Uncertainty, it is final to obtain the mean value and variance for updating state vector by the recursive calculation to observation model, but due to drawing
Into EKF, uncertain and linearization problem is brought for the Time & Space Complexity that SLAM algorithm calculates, meanwhile, make
The dimension of matrix is increased with point feature method, also increases the complexity of calculating.
It is influenced to make up the linearisation result bring of EKF, Unscented kalman filtering, particle filter etc. successively occurs
A variety of filtering modes.Although these methods solve the linearization problem of EKF, but still without significantly on computation complexity
It is promoted.Currently, SLAM technology is applied to mostly in single unmanned machine, it is multiple single under the scene of some more equipment collaborations
Unmanned machine can be repeated as many times in Same Scene and handle identical feature, this produces entire group in computing resource
Waste.For swarm intelligence equipment (multiple agent), research is more rested in the planning of pre-determined route at this stage, such as
Bee colony, the research such as ant colony has highly significant effect in path planning field, but under foreign environment, path planning system is difficult to
It exerts advantages of oneself, so that multiple agent still has ineffective systems when handling this kind of scene, system fluctuation of service,
The problems such as navigation accuracy is low.
Summary of the invention
The present invention provides a kind of based on the air-ground coordination vision navigation method and device for scheming optimization SLAM is improved, by drawing
The probability calculation for entering figure optimization is theoretical, and in the case where not abandoning the basic Computational frame of EKF, figure optimization can substantially reduce EKF prediction
The dimension of calculating matrix is needed in the process, so that matrix becomes sparse, is solved due to matrix dimensionality bring algorithm complexity
The problem of, and solving that be placed in navigation efficiency under the scene of position in face of multiple agent in the prior art low, system operation is unstable
The problems such as determining.
In order to achieve the above object, the invention adopts the following technical scheme:
It is a kind of based on improving the air-ground coordination vision guided navigation device for scheming optimization SLAM, comprising: signal acquisition module, at front end
Manage module, back end processing module, information communication module；The signal acquisition module includes monocular vision sensor；The front end
Processing module includes that signal processing system and number pass；The back end processing module includes data computing system；The information communication
Module includes digital transmission module and figure transmission module；The signal acquisition module is sent to after video signal collection at the front end
It manages module and carries out pre-processing, after obtaining key frame information and characteristic point information, the back end processing module is sent to, after described
It holds processing module to carry out pose resolving and state estimation to the characteristic point of corresponding key frame, and sends result to control system,
Contacting between modules and between modules and control system depends on the realization of information communication module.
In apparatus described above, the signal processing system in the front end processing block is the signal processing based on STM32
System, it is that 433Mhz number passes that the number, which passes,；Data computing system in the back end processing module is SCM Based data
Computing system；Digital transmission module in the information communication module is to be placed in the 433Mhz that data are used for transmission on each intelligent body
Digital transmission module, the figure transmission module be 5.8Ghz figure transmission module；The intelligent body includes: aerial intelligent body and ground intelligent
Body, the aerial intelligent body and ground intelligent body are passed by 433MHz number using Mavlink communications protocol and are carried out with mushroom antenna
Data interaction and transmission, the ground intelligent body are to carry GNSS receiver, inertial navigation sensors, monocular vision sensing
The unmanned intelligent carriage that device, 433MHz number pass and the processing system based on STM32 is constituted, the aerial intelligent body are to carry
GNSS receiver, acceleration transducer, gyroscope, monocular vision sensor, 433MHz number passes and the processor system based on STM32
The Intelligent unattended machine for composition of uniting；Each sensor by being connected to STM32, is transferred data in processor and is handled, needed
By being connected to, the number of STM32 communication port passes the information to be interacted and mushroom antenna carries out.
A kind of air-ground coordination vision navigation method for scheming optimization SLAM based on improvement, comprising the following steps:
(1) monocular vision sensor carries out the acquisition of video data to unknown scene, by video data according to it is certain when
Between interval carry out frame sampling, obtain each frame image handled；
(2) feature extraction of inflection point and line of demarcation etc. in the frame image is come out using algorithm, provides view for rear end
Characteristic in open country；
(3) obtained characteristic is transferred to the acquisition that figure optimization algorithm carries out posture information.
In step described above, algorithm described in step (2) is improved FAST Angular Point Extracting Method, described improved
FAST Angular Point Extracting Method is characterized increase principal direction information while extracting, and the main object for extracting characteristic point is image
The fast angle point of middle pixel transition or inflection point, i.e. foundation characteristic in image；And it is relatively clear to have a common boundary for planes some in space
Position, inflection point and angle point can fall into locally optimal solution during the extraction process, be described using Plucker linear feature, be extracted
Main object be apparent boundary in image, i.e., the advanced feature in image.The method combined by dotted line feature uses
STM32 is used as to handle carrier, the dotted line fusion feature of the available frame and the Initial state estimation of carrier, after being subsequent
End Processing Algorithm provides primary data and is initialized；The back-end algorithm is in monocular vision sensor and front-end algorithm pair
Position scene carries out the acquisition of video data and after processing obtains the image of frame unit, using FAST characteristic point Processing Algorithm,
Frame image features are mentioned and are described using matrix, and obtained characteristic is transferred to figure optimization algorithm and carries out posture information
Acquisition；
The specific steps of figure optimization algorithm in step (3) are as follows:
(a) using obtained Initial state estimation as the initiation parameter of algorithm, in conjunction with system model to initiation parameter
It is resolved to obtain spatiality parameter；
(b) according to System State Model, estimate the posture information of intelligent body subsequent time.Using feature Points And lines as in figure
Node, the relationship between each feature Points And lines is thought of as the Bayesian network of a directed acyclic, use is optimal as side
Path profile optimum theory reduces the dimension that matrix is used in calculating process, in this course, if there is new characteristic point and
Line, will the step for initialize；It will be removed in this step if there is characteristic point or line leave the visual field；
(c) it is updated by state estimation and covariance matrix of the computing system gain to system, completes this frame
Entire algorithm flow.
Above-described figure optimization algorithm uses Laplacian Matrix calculation method, is being estimated and updated operation
Before, the matrix after conversion is subjected to LS-SVM sparseness, so that the general features that the Hessian matrix calculated has coefficient matrix is participated in,
By the way that the difficulty of matrix calculating can be substantially reduced using this feature, accelerate the renewal speed of data.
Multi-agent Hierarchical SLAM method are as follows: under conditions of multiple agent is applied to same unknown scene, definition is connected
Coordinate system on intelligent body individual is local coordinate；Definition is world coordinate system with the coordinate system that the earth is connected；By each
The characteristic point information that single intelligent body is collected into is handled and is converted in local coordinate, due to system features there is still a need for
It is converted in world coordinate system, and it is complete mostly intelligent to be formed that the interaction for carrying out characteristic information is needed between multiple intelligent bodies
System system, therefore define a kind of new data merging scene: broad sense winding detection；The scene for designing the detection of broad sense winding includes two
Kind: one is the winding detection detected when identical single intelligent body returns to origin-location with winding, the other is working as a certain intelligence
When the position that energy body has already passed through to other intelligent bodies, the merging for the data that can also set out；System is just mentioned in local coordinate
The point got, the features such as line, is converted by coordinate, and data are transferred to world coordinate system by matrix operation etc., and with other intelligence
Body communication to reduce intelligent body winding detection number accelerates that convergence rate is accurately positioned.
The utility model has the advantages that the present invention provides a kind of based on the air-ground coordination vision navigation method and dress that improve figure optimization SLAM
It sets, has built the multiple agent vision guided navigation algorithm of view-based access control model navigation, by using the coordination between multiple vacant lot intelligent bodies,
Reduce single intelligent body characteristic point data to be treated, merges scene by setting data, simplify the meter for needing to carry out
Calculation and data interaction are solved the problems, such as channel occlusion caused by being transmitted due to mass data, accelerate whole system navigation data
Computing speed.
The present invention has merged dotted line feature, the feature extraction algorithm of dotted line Fusion Features, by the feature vector of whole system
Dimension has carried out preliminary reduction, when carrying out feature information processing, for indicate the matrix dimension of relationship between each point also with
Reduction, this can be substantially reduced required time when feature extraction, ensure that the instantaneity of feature extraction.
Present invention improves over figure optimization algorithms, optimize by using the figure for combining Laplacian Matrix calculation method and calculate
Method further decreases after being converted to relational graph on the basis of reducing feature vector dimension, for indicating the square of relational graph
The dimension of battle array.Due to being predicted and being updated in calculating process, need to carry out this matrix repeatedly complicated matrix operation.
The speed accordingly calculated is accelerated by reducing matrix dimension.Therefore improvement figure optimization algorithm also reduces on the basis of EKF
Algorithm complexity significantly mentions so that having in the speed predicted and updated using posture information of the characteristic to intelligent body
It is high.Navigation system based on multiple intelligent bodies of the invention can play a role in the multiple fields such as civilian, commercial, military.
Detailed description of the invention
Fig. 1 is the hardware system hardware structure diagram of the embodiment of the present invention；
Fig. 2 is the point of the embodiment of the present invention, line feature combination schematic diagram；
Fig. 3 is the system flow chart of the embodiment of the present invention；
Fig. 4 is the improvement figure optimization algorithm flow chart of the embodiment of the present invention；
Fig. 5 is the multi-agent Hierarchical system design drawing of the embodiment of the present invention.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments:
As shown in figure 3, a kind of based on the air-ground coordination vision guided navigation device for improving figure optimization SLAM, comprising: signal acquisition
Module, front end processing block, back end processing module, information communication module；The signal acquisition module includes monocular vision sensing
Device；The front end processing block includes that signal processing system and number pass；The back end processing module includes data computing system；Institute
Stating information communication module includes digital transmission module and figure transmission module；The signal acquisition module is sent to after video signal collection
The front end processing block carries out pre-processing, after obtaining key frame information and characteristic point information, sends the back-end processing to
Module, the back end processing module carries out pose resolving and state estimation to the characteristic point of corresponding key frame, and result is transmitted
To control system, it is real that contacting between modules and between modules and control system depends on information communication module
It is existing.The main function of the signal acquisition module is to detect scene and location information using sensor；Front end processing block will be believed
The information that number acquisition module obtains simply is pre-processed, and image information is converted to matrix information so as to operation；At rear end
The matrix information that reason module obtains front end processing block transmitting handles matrix；After improvement figure optimization algorithm module is subordinated to
Processing module is held, this module is used to accelerate the speed of information processing.
In apparatus described above, the signal processing system in the front end processing block is the signal processing based on STM32
System, it is that 433Mhz number passes that the number, which passes,；Data computing system in the back end processing module is SCM Based data
Computing system；Digital transmission module in the information communication module is to be placed in the 433Mhz that data are used for transmission on each intelligent body
Digital transmission module, the figure transmission module be 5.8Ghz figure transmission module；As shown in Figure 1, the intelligent body includes: aerial intelligent body
With ground intelligent body, the ground intelligent body be carry GNSS receiver, inertial navigation sensors, monocular vision sensor,
The unmanned intelligent carriage that 433MHz number passes and the processing system based on STM32 is constituted, the aerial intelligent body are to carry GNSS
Receiver, acceleration transducer, gyroscope, monocular vision sensor, 433MHz number passes and the processor system structure based on STM32
At Intelligent unattended machine；Each sensor by being connected to STM32, is transferred data in processor and is handled, need to hand over
By being connected to, the number of STM32 communication port passes mutual information and mushroom antenna carries out.The aerial intelligent body and ground intelligent
Body is passed by 433MHz number using Mavlink communications protocol and carries out data interaction and transmission with mushroom antenna, and devises data
Transmitting scene.When the triggering of data transmitting scene condition, it will data transmission event occurs；GNSS receiver, inertial navigation pass
Sensor is used to provide the location information under unknown scene and determines that position provides data with when detecting winding to be rough；Visual sensor
For acquiring the image under unknown scene, by carrying out preliminary treatment to image, the point feature in image, line feature are extracted.It is logical
The process that algorithm completes feature extraction is crossed, the characteristic in the visual field is provided for rear end；STM32 is as algorithm carrier and processing
Device, by writing program for algorithm programming into flash memory, after the characteristic that visual sensor provides in the visual field, according to improvement
Figure optimization algorithm is completed pose and is estimated, coordinates number biography and updates the functions such as local and global data；Multiple intelligent bodies pass through layering
SLAM collaboration carries out the interaction of environmental characteristic related data and information.
As shown in Fig. 2, a kind of based on the air-ground coordination vision navigation method for improving figure optimization SLAM, comprising the following steps:
(1) monocular vision sensor carries out the acquisition of video data to unknown scene, by video data according to it is certain when
Between interval carry out frame sampling, obtain each frame image handled；
(2) feature extraction of inflection point and line of demarcation etc. in the frame image is come out using algorithm, provides view for rear end
Characteristic in open country；
(3) obtained characteristic is transferred to the acquisition that figure optimization algorithm carries out posture information.
In step described above, algorithm described in step (2) is improved FAST Angular Point Extracting Method, described improved
FAST Angular Point Extracting Method is characterized increase principal direction information while extracting, and the main object for extracting characteristic point is image
The fast angle point of middle pixel transition or inflection point, i.e. foundation characteristic in image；And it is relatively clear to have a common boundary for planes some in space
Position, inflection point and angle point can fall into locally optimal solution during the extraction process, be described using Plucker linear feature, be extracted
Main object be apparent boundary in image, i.e., the advanced feature in image.The method combined by dotted line feature uses
STM32 is used as to handle carrier, the dotted line fusion feature of the available frame and the Initial state estimation of carrier, after being subsequent
End Processing Algorithm provides primary data and is initialized；The back-end algorithm is in monocular vision sensor and front-end algorithm pair
Position scene carries out the acquisition of video data and after processing obtains the image of frame unit, using FAST characteristic point Processing Algorithm,
Frame image features are mentioned and are described using matrix, and obtained characteristic is transferred to figure optimization algorithm and carries out posture information
Acquisition；
The specific steps of figure optimization algorithm in step (3) are as follows:
(a) using obtained Initial state estimation as the initiation parameter of algorithm, in conjunction with system model to initiation parameter
It is resolved to obtain spatiality parameter；
(b) according to System State Model, estimate the posture information of intelligent body subsequent time.Using feature Points And lines as in figure
Node, the relationship between each feature Points And lines is thought of as the Bayesian network of a directed acyclic, use is optimal as side
Path profile optimum theory reduces the dimension that matrix is used in calculating process, in this course, if there is new characteristic point and
Line, will the step for initialize；It will be removed in this step if there is characteristic point or line leave the visual field；
(c) it is updated by state estimation and covariance matrix of the computing system gain to system, completes this frame
Entire algorithm flow.
Above-described figure optimization algorithm uses Laplacian Matrix calculation method, is being estimated and updated operation
Before, the matrix after conversion is subjected to LS-SVM sparseness, so that the general features that the Hessian matrix calculated has coefficient matrix is participated in,
By the way that the difficulty of matrix calculating can be substantially reduced using this feature, accelerate the renewal speed of data.
In the improvement figure optimization algorithm, the characteristics of due to circumstances not known, the characteristic point and Eigenvector that can extract
Quantity cannot predict, if this quantity is very huge, the quantity on its node and side is very considerable after being converted into figure.And
In calculating below, the calculating carried out to these matrixes can increase with the increase of matrix dimension, so that calculating speed becomes
Must be very slow, therefore present invention employs Laplacian Matrix calculation methods, before being estimated and updating operation, will turn
Matrix after changing carries out LS-SVM sparseness, so that participating in the general features that the Hessian matrix calculated has sparse matrix, passes through benefit
It can be substantially reduced the difficulty of matrix calculating with this feature, accelerate the renewal speed of data.
As shown in figure 4, the process of improvement figure optimization algorithm is, after characteristic is converted matrix by front end, figure optimization is calculated
Connection between each feature is abstracted as the side of figure by method by all the points, the node that line feature abstraction is figure.It will be between reference point
It is connect with side chain, so that it may obtain the Bayesian network of a directed acyclic.The Bayesian network of this directed acyclic is taken out
Relational matrix sets a threshold value, it is 1 that correlation, which is greater than its side right value of this threshold value, otherwise is 0.Then the optimal calculation in path is used
Method, carries out optimizing to Bayesian network, and the node relationships matrix after being optimized is carried out using Laplacian Matrix calculation method
Calculation processing finally obtains the vector that can be used for more new system covariance and state.
The multi-agent Hierarchical SLAM method is designed as, and is applied to the condition of same unknown scene in multiple agent
Under, the coordinate system that definition is connected on intelligent body individual is local coordinate；The coordinate system that definition is connected with the earth is world's seat
Mark system.Local map information passes through transition matrix for this mapIt is transformed into global figureFor any one intelligence
For body, pose vector is defined as in its local coordinate
Wherein, RiIndicate i-th of intelligent body posture information,Indicate the road sign point letter in map
Breath.The posture of robot is consecutive variations, posture by robot in the local map at current time posture with before when
Quarter, the Relative Transformation of state obtained, can be expressed as
Posture information in global level map can be defined as
Same characteristic point provides information by L intelligent body in global figure, thus using least square method to information at
Reason
It is handled and is converted in local coordinate by the characteristic point information that each single intelligent body is collected into.Due to being
System feature in world coordinate system there is still a need for being converted, and the interaction for needing to carry out characteristic information between multiple intelligent bodies is with shape
At complete multi-agent system, therefore define a kind of new data merging scene: broad sense winding detection.Design the inspection of broad sense winding
It is the winding detection detected when identical single intelligent body returns to origin-location with winding that the scene of survey, which includes two kinds: one, separately
One is the merging for the data that can also set out when the position that a certain intelligent body has already passed through to other intelligent bodies.System has just been existed
Posture information is transformed into world coordinate system by the above operation by the point extracted in local coordinate, the features such as line, and and its
Its agent communication shares the characteristic point information of the position.To reduce intelligent body winding detection number, accelerate that convergence is accurately positioned
Speed.
As shown in figure 5, multi-agent Hierarchical SLAM design cycle is to be adopted first using sensor to circumstances not known
Then collection carries out information resolving in front end, obtains position and visual signature information vector, at this time by detection local coordinate
Location information illustrates that the intelligent physical examination measures winding, i.e., local winding is completed if there is the location information；If do not examined
It measures, the information for whether having this position under world coordinate system is detected, if so, illustrating that other intelligent bodies once reached this position, i.e.,
Global winding is completed.No matter local winding or global winding can be unified for broad sense winding, and when winding completion can be used
This information and raw information carry out unknown prediction and update by least square method, and if winding inspection can not be completed
It surveys, then carries out unknown prediction and update using this metrical information.
The foregoing is merely the preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto, ability
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by field technique personnel, should all cover of the invention
Within protection scope.
Claims (10)
1. a kind of based on the air-ground coordination vision guided navigation device for improving figure optimization SLAM characterized by comprising signal acquisition mould
Block, front end processing block, back end processing module, information communication module；The signal acquisition module includes monocular vision sensor；
The front end processing block includes that signal processing system and number pass；The back end processing module includes data computing system；It is described
Information communication module includes digital transmission module and figure transmission module；The signal acquisition module sends institute to for after video signal collection
It states front end processing block and carries out pre-processing, after obtaining key frame information and characteristic point information, send the back-end processing mould to
Block, the back end processing module carries out pose resolving and state estimation to the characteristic point of corresponding key frame, and result is sent to
Control system, contacting between modules and between modules and control system depend on the realization of information communication module.
2. according to claim 1 based on the air-ground coordination vision guided navigation device for improving figure optimization SLAM, which is characterized in that
Signal processing system in the front end processing block is the signal processing system based on STM32, and it is 433Mhz number that the number, which passes,
It passes；Data computing system in the back end processing module is SCM Based data computing system；The information communication mould
Digital transmission module in block is the digital transmission module for being placed in the 433Mhz that data are used for transmission on each intelligent body, the figure transmission module
For the figure transmission module of 5.8Ghz.
3. according to claim 2 based on the air-ground coordination vision guided navigation device for improving figure optimization SLAM, which is characterized in that
The intelligent body includes: aerial intelligent body and ground intelligent body, and the aerial intelligent body and ground intelligent body are logical using Mavlink
Agreement is interrogated to pass and mushroom antenna progress data interaction and transmission by 433MHz number.
4. according to claim 3 based on the air-ground coordination vision guided navigation device for improving figure optimization SLAM, which is characterized in that
The ground intelligent body is to carry GNSS receiver, inertial navigation sensors, monocular vision sensor, 433MHz number biography and base
In the unmanned intelligent carriage that the processing system of STM32 is constituted, the aerial intelligent body is to carry GNSS receiver, acceleration biography
The Intelligent unattended machine that sensor, gyroscope, monocular vision sensor, 433MHz number pass and the processor system based on STM32 is constituted；
Each sensor by being connected to STM32, transfers data in processor and is handled, interactive information is needed to pass through connection
It is passed in the number of STM32 communication port and mushroom antenna carries out.
5. a kind of based on the air-ground coordination vision navigation method for improving figure optimization SLAM, which comprises the following steps:
(1) monocular vision sensor carries out the acquisition of video data to unknown scene, by video data according between the regular hour
Every carrying out frame sampling, each frame image handled is obtained；
(2) feature extraction of inflection point and line of demarcation etc. in the frame image is come out using algorithm, is provided in the visual field for rear end
Characteristic；
(3) obtained characteristic is transferred to the acquisition that figure optimization algorithm carries out posture information.
6. according to claim 5 based on the air-ground coordination vision navigation method for improving figure optimization SLAM, which is characterized in that
Algorithm described in step (2) is improved FAST Angular Point Extracting Method, and the improved FAST Angular Point Extracting Method is in extraction
It is characterized simultaneously and increases principal direction information, the main object for extracting characteristic point the fast angle point of pixel transition or is turned in image
Point, i.e. foundation characteristic in image；And have a common boundary relatively clear position for planes some in space, inflection point and angle point are mentioning
Locally optimal solution can be fallen into during taking, is described using Plucker linear feature, and the main object of extraction is apparent in image
Boundary, i.e., the advanced feature in image, the method combined by dotted line feature use STM32 as to handle carrier, can obtain
To the dotted line fusion feature of the frame and the Initial state estimation of carrier, primary data is provided for subsequent back-end processing algorithm and is carried out
Initialization.
7. according to claim 6 based on the air-ground coordination vision navigation method for improving figure optimization SLAM, which is characterized in that
The back-end algorithm is that the acquisition and processing of video data are carried out to position scene in monocular vision sensor and front-end algorithm
After obtaining the image of frame unit, using FAST characteristic point Processing Algorithm, frame image features is mentioned and are described using matrix, and
Obtained characteristic is transferred to the acquisition that figure optimization algorithm carries out posture information.
8. according to claim 5 based on the air-ground coordination vision navigation method for improving figure optimization SLAM, which is characterized in that
The specific steps of figure optimization algorithm in step (3) are as follows:
(a) using obtained Initial state estimation as the initiation parameter of algorithm, initiation parameter is carried out in conjunction with system model
Resolving obtains spatiality parameter；
(b) according to System State Model, the posture information of intelligent body subsequent time is estimated, using feature Points And lines as the section in figure
Point, the relationship between each feature Points And lines are thought of as the Bayesian network of a directed acyclic, use optimal path as side
Figure optimum theory reduces the dimension that matrix is used in calculating process, in this course, will if there is new feature Points And lines
The step for initialized；It will be removed in this step if there is characteristic point or line leave the visual field；
(c) it is updated by state estimation and covariance matrix of the computing system gain to system, completes the entire of this frame
Algorithm flow.
9. feature exists based on the air-ground coordination vision navigation method for scheming optimization SLAM is improved according to claim 5 or 8
In the figure optimization algorithm uses Laplacian Matrix calculation method and will convert before being estimated and updating operation
Matrix afterwards carries out LS-SVM sparseness, so that participating in the general features that the Hessian matrix calculated has sparse matrix.
10. according to claim 5 based on the air-ground coordination vision navigation method for scheming optimization SLAM is improved, feature exists
In multi-agent Hierarchical SLAM method are as follows: under conditions of multiple agent is applied to same unknown scene, definition is connected in intelligence
Coordinate system on body individual is local coordinate；Definition is world coordinate system with the coordinate system that the earth is connected；By each single intelligence
The characteristic point information that energy body is collected into is handled and is converted in local coordinate, since there is still a need for sit in the world for system features
It is converted in mark system, and needs the interaction for carrying out characteristic information between multiple intelligent bodies to form complete multiple agent system
System defines a kind of new data merging scene: broad sense winding detection；The scene for designing the detection of broad sense winding includes two kinds: one
It is the winding detection detected with winding when identical single intelligent body returns to origin-location, the other is when a certain intelligent body arrives it
When the position that his intelligent body has already passed through, the merging for the data that can also set out；The point that system has just been extracted in local coordinate,
The features such as line, are converted by coordinate, and data are transferred to world coordinate system by matrix operation etc., and with other agent communications, with
It reduces intelligent body winding and detects number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910561547.1A CN110261877B (en) | 2019-06-26 | 2019-06-26 | Ground-air collaborative visual navigation method and device based on improved graph optimization SLAM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910561547.1A CN110261877B (en) | 2019-06-26 | 2019-06-26 | Ground-air collaborative visual navigation method and device based on improved graph optimization SLAM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110261877A true CN110261877A (en) | 2019-09-20 |
| CN110261877B CN110261877B (en) | 2024-06-11 |
Family
ID=67921876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910561547.1A Active CN110261877B (en) | 2019-06-26 | 2019-06-26 | Ground-air collaborative visual navigation method and device based on improved graph optimization SLAM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110261877B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112714165A (en) * | 2020-12-22 | 2021-04-27 | 声耕智能科技(西安)研究院有限公司 | Distributed network cooperation strategy optimization method and device based on combination mechanism |
| CN112948411A (en) * | 2021-04-15 | 2021-06-11 | 深圳市慧鲤科技有限公司 | Pose data processing method, interface, device, system, equipment and medium |
| CN112987032A (en) * | 2019-12-17 | 2021-06-18 | 无锡市电子仪表工业有限公司 | Internet of things multidata collaborative protocol based on Beidou positioning |
| CN113470089A (en) * | 2021-07-21 | 2021-10-01 | 中国人民解放军国防科技大学 | Cross-domain cooperative positioning and mapping method and system based on three-dimensional point cloud |
| CN114942029A (en) * | 2022-05-31 | 2022-08-26 | 哈尔滨工业大学 | Multi-robot mutual positioning method and system based on anonymous relative angle measurement |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101625573A (en) * | 2008-07-09 | 2010-01-13 | 中国科学院自动化研究所 | Digital signal processor based inspection robot monocular vision navigation system |
| CN101839721A (en) * | 2010-03-12 | 2010-09-22 | 西安电子科技大学 | Visual navigation method in autonomous rendezvous and docking |
| US20120121161A1 (en) * | 2010-09-24 | 2012-05-17 | Evolution Robotics, Inc. | Systems and methods for vslam optimization |
| CN104764440A (en) * | 2015-03-12 | 2015-07-08 | 大连理工大学 | Rolling object monocular pose measurement method based on color image |
| CN105388908A (en) * | 2015-12-11 | 2016-03-09 | 国网四川省电力公司电力应急中心 | Machine vision-based unmanned aerial vehicle positioned landing method and system |
| CN106447579A (en) * | 2016-10-10 | 2017-02-22 | 成都理工大学 | Space-ground-air integrated coordinated search and rescue system suitable for complex mountain scenic areas |
| CN107144281A (en) * | 2017-06-30 | 2017-09-08 | 飞智控（天津）科技有限公司 | Unmanned plane indoor locating system and localization method based on cooperative target and monocular vision |
| CN107687850A (en) * | 2017-07-26 | 2018-02-13 | 哈尔滨工业大学深圳研究生院 | A kind of unmanned vehicle position and orientation estimation method of view-based access control model and Inertial Measurement Unit |
| CN109307508A (en) * | 2018-08-29 | 2019-02-05 | 中国科学院合肥物质科学研究院 | A kind of panorama inertial navigation SLAM method based on more key frames |
| CN211878189U (en) * | 2019-06-26 | 2020-11-06 | 南京航空航天大学 | Ground-air cooperative vision navigation device based on improved map optimization SLAM |
-
2019
- 2019-06-26 CN CN201910561547.1A patent/CN110261877B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101625573A (en) * | 2008-07-09 | 2010-01-13 | 中国科学院自动化研究所 | Digital signal processor based inspection robot monocular vision navigation system |
| CN101839721A (en) * | 2010-03-12 | 2010-09-22 | 西安电子科技大学 | Visual navigation method in autonomous rendezvous and docking |
| US20120121161A1 (en) * | 2010-09-24 | 2012-05-17 | Evolution Robotics, Inc. | Systems and methods for vslam optimization |
| CN104764440A (en) * | 2015-03-12 | 2015-07-08 | 大连理工大学 | Rolling object monocular pose measurement method based on color image |
| CN105388908A (en) * | 2015-12-11 | 2016-03-09 | 国网四川省电力公司电力应急中心 | Machine vision-based unmanned aerial vehicle positioned landing method and system |
| CN106447579A (en) * | 2016-10-10 | 2017-02-22 | 成都理工大学 | Space-ground-air integrated coordinated search and rescue system suitable for complex mountain scenic areas |
| CN107144281A (en) * | 2017-06-30 | 2017-09-08 | 飞智控（天津）科技有限公司 | Unmanned plane indoor locating system and localization method based on cooperative target and monocular vision |
| CN107687850A (en) * | 2017-07-26 | 2018-02-13 | 哈尔滨工业大学深圳研究生院 | A kind of unmanned vehicle position and orientation estimation method of view-based access control model and Inertial Measurement Unit |
| CN109307508A (en) * | 2018-08-29 | 2019-02-05 | 中国科学院合肥物质科学研究院 | A kind of panorama inertial navigation SLAM method based on more key frames |
| CN211878189U (en) * | 2019-06-26 | 2020-11-06 | 南京航空航天大学 | Ground-air cooperative vision navigation device based on improved map optimization SLAM |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112987032A (en) * | 2019-12-17 | 2021-06-18 | 无锡市电子仪表工业有限公司 | Internet of things multidata collaborative protocol based on Beidou positioning |
| CN112714165A (en) * | 2020-12-22 | 2021-04-27 | 声耕智能科技(西安)研究院有限公司 | Distributed network cooperation strategy optimization method and device based on combination mechanism |
| CN112948411A (en) * | 2021-04-15 | 2021-06-11 | 深圳市慧鲤科技有限公司 | Pose data processing method, interface, device, system, equipment and medium |
| CN112948411B (en) * | 2021-04-15 | 2022-10-18 | 深圳市慧鲤科技有限公司 | Pose data processing method, interface, device, system, equipment and medium |
| CN113470089A (en) * | 2021-07-21 | 2021-10-01 | 中国人民解放军国防科技大学 | Cross-domain cooperative positioning and mapping method and system based on three-dimensional point cloud |
| CN113470089B (en) * | 2021-07-21 | 2022-05-03 | 中国人民解放军国防科技大学 | Cross-domain cooperative positioning and mapping method and system based on three-dimensional point cloud |
| CN114942029A (en) * | 2022-05-31 | 2022-08-26 | 哈尔滨工业大学 | Multi-robot mutual positioning method and system based on anonymous relative angle measurement |
| CN114942029B (en) * | 2022-05-31 | 2024-09-10 | 哈尔滨工业大学 | Multi-robot mutual positioning method and system based on anonymous relative angle measurement |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110261877B (en) | 2024-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110261877A (en) | A kind of air-ground coordination vision navigation method and device for scheming optimization SLAM based on improvement | |
| CN104766320B (en) | Many Bernoulli Jacob under thresholding is measured filter Faint target detection and tracking | |
| CN113223045B (en) | Vision and IMU sensor fusion positioning system based on dynamic object semantic segmentation | |
| CN110223348A (en) | Robot scene adaptive bit orientation estimation method based on RGB-D camera | |
| CN111880207B (en) | Visual inertial satellite tight coupling positioning method based on wavelet neural network | |
| Jia et al. | A Survey of simultaneous localization and mapping for robot | |
| CN110097584A (en) | The method for registering images of combining target detection and semantic segmentation | |
| CN102155955A (en) | Stereoscopic vision mile meter and measuring method | |
| CN115240047A (en) | Laser SLAM method and system fusing visual loopback detection | |
| CN110555412A (en) | End-to-end human body posture identification method based on combination of RGB and point cloud | |
| CN111476089B (en) | Pedestrian detection method, system and terminal for multi-mode information fusion in image | |
| CN113284144B (en) | Tunnel detection method and device based on unmanned aerial vehicle | |
| CN109978919A (en) | A kind of vehicle positioning method and system based on monocular camera | |
| CN111914615A (en) | Fire-fighting area passability analysis system based on stereoscopic vision | |
| CN116608873A (en) | Multi-sensor fusion positioning mapping method for automatic driving vehicle | |
| CN113050083A (en) | Ultra-wideband radar human body posture reconstruction method based on point cloud | |
| CN108519083A (en) | A kind of capture of space non-cooperative multiple target and track algorithm | |
| CN113516232B (en) | Self-attention mechanism-based wall-penetrating radar human body posture reconstruction method | |
| CN102510512B (en) | Binocular-vision-node based self-positioning method for dynamic three-dimensional multimedia sensor network | |
| CN118010024A (en) | Point-line-surface decoupling-based 6-DoF pose estimation optimization method and system | |
| CN211878189U (en) | Ground-air cooperative vision navigation device based on improved map optimization SLAM | |
| CN109961476A (en) | The localization method of the underground parking of view-based access control model | |
| Yin | Research on natural disaster target change detection method based on deep learning | |
| Song et al. | Improved FCM algorithm for fisheye image cluster analysis for tree height calculation | |
| CN117289298B (en) | Multi-machine collaborative online mapping method, system and terminal equipment based on laser radar |
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 |