CN108568868A - A kind of the screen of trees cleaning air-robot and barrier-avoiding method of automatic obstacle-avoiding - Google Patents
A kind of the screen of trees cleaning air-robot and barrier-avoiding method of automatic obstacle-avoiding Download PDFInfo
- Publication number
- CN108568868A CN108568868A CN201810192078.6A CN201810192078A CN108568868A CN 108568868 A CN108568868 A CN 108568868A CN 201810192078 A CN201810192078 A CN 201810192078A CN 108568868 A CN108568868 A CN 108568868A
- Authority
- CN
- China
- Prior art keywords
- robot
- air
- barrier
- trees
- avoiding
- 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 43
- 238000004140 cleaning Methods 0.000 title claims abstract description 20
- 230000004888 barrier function Effects 0.000 claims abstract description 69
- 210000000245 forearm Anatomy 0.000 claims abstract description 22
- 238000012544 monitoring process Methods 0.000 claims abstract description 8
- 238000001514 detection method Methods 0.000 claims description 28
- 230000000694 effects Effects 0.000 claims description 22
- 230000005540 biological transmission Effects 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 16
- 238000004891 communication Methods 0.000 claims description 11
- 230000007613 environmental effect Effects 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 7
- 239000012212 insulator Substances 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 6
- 230000000007 visual effect Effects 0.000 claims description 6
- 238000009434 installation Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 230000009471 action Effects 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000009432 framing Methods 0.000 claims description 3
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 claims description 3
- 239000010931 gold Substances 0.000 claims description 3
- 229910052737 gold Inorganic materials 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 3
- 238000004064 recycling Methods 0.000 claims description 3
- 230000001373 regressive effect Effects 0.000 claims description 3
- 239000007787 solid Substances 0.000 claims description 3
- 230000002146 bilateral effect Effects 0.000 claims description 2
- 239000002023 wood Substances 0.000 claims description 2
- 238000002156 mixing Methods 0.000 claims 1
- 230000008447 perception Effects 0.000 description 8
- 238000006073 displacement reaction Methods 0.000 description 7
- 238000006243 chemical reaction Methods 0.000 description 6
- 239000006096 absorbing agent Substances 0.000 description 5
- 230000000712 assembly Effects 0.000 description 4
- 238000000429 assembly Methods 0.000 description 4
- 230000005611 electricity Effects 0.000 description 4
- 230000033001 locomotion Effects 0.000 description 4
- 230000006641 stabilisation Effects 0.000 description 4
- 238000011105 stabilization Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000005484 gravity Effects 0.000 description 3
- 238000013017 mechanical damping Methods 0.000 description 3
- 238000005096 rolling process Methods 0.000 description 3
- RZVHIXYEVGDQDX-UHFFFAOYSA-N 9,10-anthraquinone Chemical compound C1=CC=C2C(=O)C3=CC=CC=C3C(=O)C2=C1 RZVHIXYEVGDQDX-UHFFFAOYSA-N 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000005452 bending Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 230000003139 buffering effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 206010020718 hyperplasia Diseases 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B27—WORKING OR PRESERVING WOOD OR SIMILAR MATERIAL; NAILING OR STAPLING MACHINES IN GENERAL
- B27C—PLANING, DRILLING, MILLING, TURNING OR UNIVERSAL MACHINES FOR WOOD OR SIMILAR MATERIAL
- B27C5/00—Machines designed for producing special profiles or shaped work, e.g. by rotary cutters; Equipment therefor
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01G—HORTICULTURE; CULTIVATION OF VEGETABLES, FLOWERS, RICE, FRUIT, VINES, HOPS OR SEAWEED; FORESTRY; WATERING
- A01G3/00—Cutting implements specially adapted for horticultural purposes; Delimbing standing trees
- A01G3/08—Other tools for pruning, branching or delimbing standing trees
-
- 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
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
- G01S17/933—Lidar systems specially adapted for specific applications for anti-collision purposes of aircraft or spacecraft
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02G—INSTALLATION OF ELECTRIC CABLES OR LINES, OR OF COMBINED OPTICAL AND ELECTRIC CABLES OR LINES
- H02G1/00—Methods or apparatus specially adapted for installing, maintaining, repairing or dismantling electric cables or lines
- H02G1/02—Methods or apparatus specially adapted for installing, maintaining, repairing or dismantling electric cables or lines for overhead lines or cables
Landscapes
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Forests & Forestry (AREA)
- Physics & Mathematics (AREA)
- Aviation & Aerospace Engineering (AREA)
- Biodiversity & Conservation Biology (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Wood Science & Technology (AREA)
- Mechanical Engineering (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Manipulator (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention discloses a kind of screen of trees of automatic obstacle-avoiding cleaning air-robot and barrier-avoiding methods, include more rotor platforms that flying power is provided for air-robot, the fuselage being fixedly connected on immediately below more rotor platforms, it is arranged in two longitudinal thrusters for being located at fuselage both sides below more rotor platforms, with fuselage longitudinal axes parallel and it is fixed on the forearm of front fuselage, with fuselage longitudinal axes parallel and it is fixed on the postbrachium of back body, it is fixed on the toolbox of forearm front end, it is fixed on the battery pack of postbrachium rear end, it is fixed on the laser radar of more rotor platform uppers, it is arranged in the height finding radar of fuselage bottom, on front side of more rotor platforms, it is arranged with camera on rear side of both sides and battery pack.The present invention realizes the real-time monitoring of barrier, is conducive to implement the high-precision avoidance obstacle of air-robot, is suitable for entering screen of trees and the intensive intersecting area of conducting wire implements operation of removing obstacles.
Description
Technical field
The present invention relates to a kind of screen of trees of automatic obstacle-avoiding cleaning air-robot and barrier-avoiding methods, belong to transmission line of electricity screen of trees
Cleaning plant technical field.
Background technology
Screen of trees is a kind of security risk existing for electric transmission line channel, shows as the gradual prestige of continuous hyperplasia of trees in channel
Coerce the operational safety of transmission line of electricity.For this purpose, power departments at different levels will put into a large amount of human and material resources with financial resources to linchpin every year
Channel screen of trees in area carries out cleaning regulation.Current screen of trees cleaning depends on manual cleaning, and there is inefficient, peaces
Full blast nearly big deficiency, therefore there is an urgent need for a kind of power circuit channel screens of trees to clear up air-robot automatically.
Air-robot when carrying out prune job to screen of trees, often with the environment such as transmission line of electricity, building, other trees
Object distance is closer, easily causes air-robot to damage and environmental objects especially power circuits once colliding
Damage, leads to serious safety accident.Therefore, screen of trees cleaning air-robot is made to have environment sensing and the ability of automatic obstacle avoiding
It is extremely important.
Invention content
Present invention solves the technical problem that being:Screen of trees cleaning air-robot and the avoidance side of a kind of automatic obstacle-avoiding are provided
Method can realize line channel Obstacle Position perception nearby and automatic obstacle avoiding, remove obstacles efficient that security risk is low, to solve
Problems of the prior art.
The technical solution that the present invention takes is:A kind of screen of trees cleaning air-robot of automatic obstacle-avoiding, including be aerial machine
Device people provides more rotor platforms of flying power, the fuselage being fixedly connected on immediately below more rotor platforms, is installed on fuselage
Longitudinal thruster, with fuselage longitudinal axes parallel and be fixed on front fuselage forearm, with fuselage longitudinal axes parallel and be fixed on machine
The postbrachium held behind, the battery pack for being fixed on postbrachium rear end, is installed on more rotor platforms at the toolbox for being fixed on forearm front end
The laser radar of central upper portion, the height finding radar for being arranged in fuselage bottom, more rotor platforms front sides, both sides and battery pack rear side are
It is disposed with camera.
Preferably, above-mentioned air-robot further includes signal processor, flight controller and the communication module being connected with each other;
Laser radar, camera and height finding radar are connected to the signal input part of signal processor, and flight controller is connected to more rotors
The power plant module of platform, the cutter controller of the driving motor of longitudinal thruster and toolbox and Inertial Measurement Unit,
Barometertic altimeter, satellite navigation receiver, communication module is for ground station transmission flying quality and onboard image.
Preferably, above-mentioned fuselage is front and back narrow structure, and air-robot is symmetrical.
Preferably, above-mentioned battery pack include to toolbox, to more rotor platforms, to longitudinal thruster, to signal processing
Device, the battery powered to flight controller and to airborne sensor.
Preferably, above-mentioned forearm is two-stage structure, and is connected as one by Saving cortilage, and Saving cortilage includes fixing
Fork, cross axle, activity fork), cylindrical sleeve, spring) and screw, fixation fork, activity fork, cylindrical sleeve are hollow cylindrical, cross
Axis is connected by the rear portion of the front of bearing and fixation fork, activity fork respectively, thus constitutes universal joint, the rear portion of fixation fork is with before
The back segment of arm is connected, and the front of cylindrical sleeve and the leading portion of forearm are connected, the rear portion of cylindrical sleeve can slide axially, can relative rotation
Form of sleeve with activity fork front connect, spring be cylinder, be installed in the form of package fixation fork, activity pitch and cylinder
The both ends of the outside of set, spring are connected with fixation fork and cylindrical sleeve respectively by two screws.
A kind of barrier-avoiding method of the screen of trees cleaning air-robot of automatic obstacle-avoiding, steps are as follows for this method:
1) installation laser radar makes its laser scanning line be slightly above more rotor wing rotation planes, by laser radar to ring around
Border carry out 360 degree of horizontal sweeps, detect and perceive air-robot surrounding object to laser radar distance;
2) camera obtains air-robot environmental images all around and is sent into signal processor, then uses machine
Vision algorithm is to the conducting wire inside and outside electric transmission line channel, shaft tower, trees and other objects are identified and framing, wherein
Shaft tower includes insulator and gold utensil, and other objects include building, bridge and chain of mountains landform;
3) the two-dimensional distance information that step 1) is perceived is merged with the barrier that step 2) is identified into row information, is obtained
The range information of barrier carries out air-robot automatic obstacle avoidance:
If --- the barrier perceived is conducting wire, shaft tower, channel outer trees and other objects, and these barriers and sky
The distance of middle robot≤preset safety distance threshold, even air-robot implements automatic obstacle avoidance control, while earthward
Monitoring station sends security alarm information;
If --- the barrier perceived is the trees in channel, and air-robot implementation is removed obstacles, while earthward being supervised
Control station sends action message of removing obstacles;
--- other than above-mentioned two situations, it is not necessarily to avoidance obstacle;
4) height of the air-robot with respect to ground is detected by the height finding radar installed downwards, this relative altitude is used for
The height of air-robot is kept and the foundation of air-robot height avoidance.
Preferably, step 2) is identified environmental objects using machine vision algorithm as follows with the method for positioning:
A) for conducting wire barrier, it is detected using straight line parameter method:First, maximum between-cluster variance is utilized
All barriers are partitioned into from background and carry out binaryzation by method, are not 0 picture to all pixels value in the plane of delineation then
Vegetarian refreshments carries out straight line modeling, it is assumed that it is (x, y) that one, which is not 0 pixel point coordinates, then it necessarily belongs to straight line, the straight line
Equation be represented by:
ρ=xcos θ+ysin θ
Wherein ρ indicates that distance of the straight line relative to origin belonging to the point, θ indicate normal of the straight line relative to origin
With the angle of positive direction of the x-axis, by the spaces ρ-θ maximizing, converting all the points in image to linear equation;Most
Afterwards, it is power transmission line according to the Straight Line Identification that angle and length are met to certain threshold condition and feeds back to its two-dimentional geological information
Information fusion algorithm;
B) for shaft tower and trees, object detection is carried out using depth convolutional neural networks method:
First, shaft tower and the Tree image library of magnanimity are established, model training when for target detection and detection.It is establishing
During image library, need to acquire the object images under various illumination, angle, focal length, weather environment, and mark;
Secondly, detection network model is established.Detection model is made of 5 convolutional layers, 3 pond layers and 2 full articulamentums;
Detection model is classified into three candidate regions selection, feature extraction, the classification of object and the recurrence of detection block part compositions;Candidate regions
Selection is using SelectiveSearch methods to 2000 RoIs of every image zooming-out;Then these RoIs are subjected to convolution sum
Pondization operates, to extract the feature of RoIs;The classification and detection of object will there are two outputs:One output is used for describing each
Probability value p=(the p of the affiliated types of RoI0,p1,...,pu), Probability p is calculated by the full articulamentum combination softmax of last layer
It obtains;Another output layer is that boundingbox returns layer, this layer exports the co-ordinate position information t of target objectu=(tx u,
ty u,tw u,th u), wherein u is that the number of barrier indexes, and (x, y) indicates that the top left co-ordinate of target object, (w, h) indicate mesh
The width and height of object are marked, RoI is region of interest;
It is all that output one is corresponding when making classification training and position recurrence for the RoIs of each tape label
The boundingbox regressive objects v of ground-truth type labels u and ground-truth.Then multitask loss is calculated
Function L:
L(p,u,tu, v) and=Lcls(p,u)+λ[u≥1]Lloc(tu,v)
Wherein, Lcls(p, u)=- logpuIt is lost for the log of true classification u;Section 2 returns loss function LlocIt is true
Object location information v=(vx,vy,vw,vh) and prediction object location informationBetween error loss;Parameter
λ can take λ=1 for balancing two loss functions;
For LlocUsing the smooth loss functions of L1:
Wherein
In above formula, d=tu-v;
Finally, trained network is transplanted on the embedded image processing computer accelerated with GPU, in real time to figure
Insulator as in is labeled and its two-dimentional geological information is fed back to information fusion algorithm;
C) trees outside line channel and other barriers, are the target of air-robot automatic obstacle avoidance, therefore are not required to know
Not specific object, therefore detected by laser radar for this class object, using distance be less than the object of threshold value as
Barrier;
Preferably, the obstacle distance information acquisition method of step 3) information fusion is as follows:
First, by coordinate transform by the collected range information of laser radar and four-way camera the image collected into
Row alignment, obtains each physical coordinates of the pixel relative to laser radar in image;
Secondly, the two-dimensional geometry information of the conducting wire found out in step 2), shaft tower, trees and other barriers is passed through into coordinate
It is transformed to three-dimensional information, transform method is:
Wherein (x, y) is the two-dimensional coordinate of a pixel in camera, (u0,v0) indicate camera photocentre coordinate, fxAnd fyPoint
Not Biao Shi the focal length of camera in the x and y direction, the distance at the practical relative image center of barrier component in the x and y direction point
Not Wei X and Y, Z be the affiliated straight line (Z of conducting wirel) or insulator (Zd) belonging to rectangular area average distance;
Again, the maximum distance region that laser radar detects is found out, the average distance in the region is denoted as Zr, compare Zr、
ZlAnd ZdSize, minimum distance is denoted as Zp, then ZpThe range information of barrier as in step 3);
Finally, Z is obtainedpThe size of opposite image rectangle or linear region, i.e. (X, Y, Zp), and as avoidance
Foundation.
Preferably, the barrier-avoiding method of the screen of trees cleaning air-robot of a kind of automatic obstacle-avoiding, step 3) or step 4) are hollow
The control method of middle robot obstacle-avoiding is:
If 1) only preceding find barrier, and obstacle distance≤safety distance threshold to camera, then carry out immediately tight
Anxious braking, and ground station sends a warning message, then by operating personnel by earth station's monitor video or site observation, to sky
Middle robot frontal scene carries out implementing further avoidance operation after again judging:1. for conducting wire or shaft tower class barrier, according to
It slowly climbs or is displaced sideways upwards according to former course until barrier and to be removed from camera view or Autonomous landing waits for work
Personnel recycle or manual control air-robot is leapt or cut-through object;2. for the trees to be cleaned in channel,
Then continue to implement operation of removing obstacles forward;
If 2) only left or right finds barrier to camera, and obstacle distance≤safety distance threshold, then immediately into
Then row emergency braking is slowly moved to barrier opposite direction, until barrier is removed from camera view;
If 3) preceding, left and right has been found that barrier, and obstacle distance≤safety distance threshold to camera, then immediately into
Row emergency braking, and ground station sends a warning message, then by operating personnel by earth station's monitor video or site observation,
To before air-robot, left and right carries out to scene implementing the operation of further avoidance after again judging:Slowly vertically climb until
Barrier is removed from camera view or Autonomous landing waits for staff's recycling, or by manually taking over manual implementation comprehensively
It removes obstacles operation;
If 4) height finding radar detection air-robot with respect to ground height≤safe altitude threshold value, even aerial machine
People rises to implement active height avoidance obstacle, and earthward monitoring station sends security alarm information by communication module.
Beneficial effects of the present invention:Compared with prior art, effect of the invention is as follows:
1) the present invention is based on more rotor platforms, using the environment sensing side of laser radar+multi-cam+height finding radar combination
Formula, detection ambient enviroment object realize the real-time detection of barrier, are conducive to implement aerial machine at a distance from air-robot
The high-precision automatic obstacle avoiding control of people;
2) present invention can realize transmission line of electricity tree by more rotor platforms, longitudinal thruster and operation toolbox
The aerial of barrier quickly cleans, and avoids high voltage transmission line of the operating personnel at screen of trees, and operation is safer, can effectively be promoted clear
It manages the efficiency of operation and reduces operational risk, it is inefficient big with security risk to solve manual cleaning existing in the prior art
The problem of;
3) Saving cortilage of forearm of the invention setting has the mechanical damping degree of freedom of four direction, can effectively weaken tree
Hinder influence of the vibration of reaction force or torque and operation cutter to air-robot flight attitude;
4) battery pack is placed in the rear portion of air-robot, plays good counterweight effect, while forearm machine opposite with postbrachium
Body can be elastic and be locked, and the center of gravity for quickly adjusting air-robot is conducive to.
Description of the drawings
Fig. 1 looks up dimensional structure diagram for the present invention's;
Fig. 2 is the top perspective structural schematic diagram of the present invention;
Fig. 3 is Saving cortilage structural schematic diagram;
Fig. 4 is Saving cortilage overall schematic.
In figure, more than 1-rotor platform, 2-fuselages, 3-longitudinal thrusters, 4-forearms, 5-postbrachiums, 6-toolboxs,
7-battery packs, 8-laser radars, 9-cameras, 10-height finding radars, 12-Saving cortilages;
1201-fixation forks, 1202-cross axles, 1203-activity forks, 1204-cylindrical sleeves, 1205-springs, 1206-
Screw.
Specific implementation mode
In the following, the present invention is described further in conjunction with attached drawing and specific embodiment.
Embodiment 1:As Figure 1-Figure 4, a kind of screen of trees of automatic obstacle-avoiding clears up air-robot, including is aerial machine
People provides more rotor platforms 1 of flying power, the fuselage 2 being fixedly connected on immediately below more rotor platforms 1, is installed on fuselage 2
Longitudinal thruster 3, with 2 longitudinal axes parallel of fuselage and be fixed on 2 front end of fuselage forearm 4, with 2 longitudinal axes parallel of fuselage and solid
It is connected in the postbrachium 5 of 2 rear end of fuselage, the toolbox 6 for being fixed on 4 front end of forearm, the battery pack 7 for being fixed on 5 rear end of postbrachium, installation
In more 1 central upper portions of rotor platform laser radar 8, be arranged in the height finding radar 10 of 2 bottom of fuselage, 1 front side of more rotor platforms,
Both sides and 7 rear side of battery pack are arranged with camera 9.More 1 shapes of rotor platform are symmetrical, and height finding radar 10 is radio thunder
It reaches or ultrasonic radar, toolbox 6 is connected to 4 front end of forearm by dismountable connector, ring flange connection can be used in connector
Or nut-screw rod quickly connects, corresponding connecting portion is equipped with electric connector head, thus with mechanically and electrically dual connection work(
Can, it is compact-sized, it can be achieved that toolbox 6 quick-replaceable.
Preferably, above-mentioned air-robot further includes signal processor, flight controller and the communication module being connected with each other;
Laser radar 8, camera 9 and height finding radar 10 are connected to the signal input part of signal processor, the physical interface packet of input terminal
Include the video interface (such as HDMI, DP, CVBS) accessed for camera and the digital bus interface that accesses for radar (such as
USB, RS-232/485/422, CAN etc.), high-performance neural network chip and image processor built in signal processor;Flight control
Device processed is connected to the cutter controller of the power plant module of more rotor platforms 1, the driving motor of longitudinal thruster and toolbox 6
And Inertial Measurement Unit, barometertic altimeter, satellite navigation receiver, communication module for ground station transmission flying quality with
Onboard image, flight controller are built in more rotor platforms 1 or fuselage 2, hard with existing multi-rotor unmanned aerial vehicle flight controller
Part is similar, including flight-control computer and with Inertial Measurement Unit, barometertic altimeter, satellite navigation receiver, Tool Control
Device is connected with cutter motor, cutter speed probe and tool temperature sensor.
Preferably, above-mentioned fuselage 2 is front and back narrow structure, and air-robot is bilateral symmetry.
Preferably, above-mentioned battery pack 7 include to toolbox 6, to more rotor platforms 1, to longitudinal thruster 3, to signal
Processor, the battery powered to flight controller and to airborne sensor.
Preferably, the barrier-avoiding method of the screen of trees cleaning air-robot of a kind of automatic obstacle-avoiding, steps are as follows for this method:
1) installation laser radar 8 makes its laser scanning line be slightly above more rotor wing rotation planes, by laser radar 8 to surrounding
Environment carry out 360 degree of horizontal sweeps, detect and perceive air-robot surrounding object to laser radar 8 distance;
2) camera 9 obtains air-robot environmental images all around and is sent into signal processor, then uses machine
Device vision algorithm to the conducting wire inside and outside electric transmission line channel, shaft tower, trees and other objects are identified and framing,
In, shaft tower includes insulator and gold utensil, and other objects include building, bridge and chain of mountains landform;
3) the two-dimensional distance information that step 1) is perceived is merged with the obstacle information that step 2) is identified, is obtained
The range information of barrier carries out air-robot automatic obstacle avoidance:
If --- the barrier perceived is conducting wire, shaft tower, channel outer trees and other objects, and these barriers and sky
The distance of middle robot≤preset safety distance threshold, even air-robot implements automatic obstacle avoidance control, while earthward
Monitoring station sends security alarm information;
If --- the barrier perceived is the trees in channel, and air-robot implementation is removed obstacles, while earthward being supervised
Control station sends action message of removing obstacles;
--- other than above-mentioned two situations, it is not necessarily to avoidance obstacle;
4) height of the air-robot with respect to ground is detected by the height finding radar 10 installed downwards, this relative altitude is used
It is kept in the height of air-robot, and carries out air-robot height avoidance.
Preferably, above-mentioned steps 2) using machine vision algorithm to environmental objects be identified with position method it is as follows:
A) for conducting wire barrier, it is detected using straight line parameter method:First, maximum between-cluster variance is utilized
All barriers are partitioned into from background and carry out binaryzation by method;Then, to all pixels value in the plane of delineation be 0 picture
Vegetarian refreshments carries out straight line modeling, it is assumed that it is (x, y) that one, which is not 0 pixel point coordinates, then it necessarily belongs to straight line, the straight line
Equation be represented by:
ρ=xcos θ+ysin θ
Wherein ρ indicates that distance of the straight line relative to origin belonging to the point, θ indicate normal of the straight line relative to origin
With the angle of positive direction of the x-axis, by the spaces ρ-θ maximizing, converting all the points in image to linear equation;Most
Afterwards, it is power transmission line according to the Straight Line Identification that angle and length are met to certain threshold condition and feeds back to its two-dimentional geological information
Information fusion algorithm;
B) for shaft tower and trees, object detection is carried out using depth convolutional neural networks method:
First, shaft tower and the Tree image library of magnanimity are established, model training when for target detection and detection.It is establishing
During image library, need to acquire the object images under various illumination, angle, focal length, weather environment, and mark;
Secondly, detection network model is established.Detection model is made of 5 convolutional layers, 3 pond layers and 2 full articulamentums;
Detection model is classified into three candidate regions selection, feature extraction, the classification of object and the recurrence of detection block part compositions;Candidate regions
Selection is using SelectiveSearch methods to 2000 RoIs of every image zooming-out;Then these RoIs are subjected to convolution sum
Pondization operates, to extract the feature of RoIs;The classification and detection of object will there are two outputs:One output is used for describing each
Probability value p=(the p of the affiliated types of RoI0,p1,...,pu), Probability p is calculated by the full articulamentum combination softmax of last layer
It obtains;Another output layer is that boundingbox returns layer, this layer exports the co-ordinate position information t of target objectu=(tx u,
ty u,tw u,th u), wherein u is that the number of barrier indexes, and (x, y) indicates that the top left co-ordinate of target object, (w, h) indicate mesh
The width and height of object are marked, RoI is region of interest;
It is all that output one is corresponding when making classification training and position recurrence for the RoIs of each tape label
The boundingbox regressive objects v of ground-truth type labels u and ground-truth.Then multitask loss is calculated
Function L:
L(p,u,tu, v) and=Lcls(p,u)+λ[u≥1]Lloc(tu,v)
Wherein, Lcls(p, u)=- logpuIt is lost for the log of true classification u;Section 2 returns loss function LlocIt is true
Object location information v=(vx,vy,vw,vh) and prediction object location informationBetween error loss;Parameter
λ can take λ=1 for balancing two loss functions;
For LlocUsing the smooth loss functions of L1:
Wherein
In above formula, d=tu-v;
Finally, trained network is transplanted on the embedded image processing computer accelerated with GPU, in real time to figure
Insulator as in is labeled and its two-dimentional geological information is fed back to information fusion algorithm;
C) trees outside line channel and other barriers, are the target of air-robot automatic obstacle avoidance, therefore are not required to know
Not specific object, therefore detected by laser radar 8 for this class object, using distance be less than the object of threshold value as
Barrier.
Preferably, the obstacle distance information acquisition method of step 3) information fusion is as follows:
First, by coordinate transform by 8 collected range information of laser radar and four-way camera the image collected
It is aligned, obtains each physical coordinates of the pixel relative to laser radar 8 in image;
Secondly, the two-dimensional geometry information of the conducting wire found out in step 2), shaft tower, trees and other barriers is passed through into coordinate
It is transformed to three-dimensional information, transform method is:
Wherein (x, y) is the two-dimensional coordinate of a pixel in camera, (u0,v0) indicate camera photocentre coordinate, fxAnd fyPoint
Not Biao Shi the focal length of camera in the x and y direction, the distance at the practical relative image center of barrier component in the x and y direction point
Not Wei X and Y, Z be the affiliated straight line (Z of conducting wirel) or shaft tower barrier (Zd) belonging to rectangular area average distance;
Again, the maximum distance region that laser radar 8 detects is found out, the average distance in the region is denoted as Zr, compare
Zr、ZlAnd ZdSize, minimum distance is denoted as Zp, then ZpThe range information of barrier as in step 3);
Finally, Z is obtainedpThe size of opposite image rectangle or linear region, i.e. (X, Y, Zp), and as avoidance
Foundation.
Preferably, the barrier-avoiding method of the screen of trees cleaning air-robot of a kind of automatic obstacle-avoiding, the control of air-robot avoidance
Method processed is:
Need implement automatic obstacle avoidance barrier include:Power line, shaft tower in electric power corridor, the trees outside electric power corridor
With other barriers;
If 1) only preceding find barrier, and obstacle distance≤safety distance threshold to camera 9, then carry out immediately tight
Anxious braking, and ground station sends a warning message, then by operating personnel by earth station's monitor video or site observation, to sky
Middle robot frontal scene carries out implementing further avoidance operation after again judging:1. for conducting wire or shaft tower class barrier, according to
It slowly climbs or is displaced sideways upwards according to former course until barrier is removed from 9 visual field of camera or Autonomous landing waits for work
It is recycled as personnel or manual control air-robot is leapt or cut-through object;2. for the tree to be cleaned in channel
Wood then continues to implement operation of removing obstacles forward;
If 2) only left or right finds barrier to camera 9, and obstacle distance≤safety distance threshold, then immediately into
Then row emergency braking is slowly moved to barrier opposite direction, until barrier is removed from 9 visual field of camera;
3) if preceding, left and right has been found that barrier, and obstacle distance≤safety distance threshold to camera 9, then immediately
Emergency braking is carried out, and ground station sends a warning message, and is then seen by earth station's monitor video or on the spot by operating personnel
Examine, to before air-robot, left and right to scene carries out implementing further avoidance after again judging operating:It slowly vertically climbs straight
It is removed from 9 visual field of camera to barrier or Autonomous landing waits for staff's recycling, or by manually taking over manual reality comprehensively
Apply operation of removing obstacles;
If 4) height finding radar 10 detect air-robot with respect to ground height≤safe altitude threshold value, even aerial machine
Device people rises to implement active height avoidance obstacle, and earthward monitoring station sends security alarm information by communication module.
Preferably, above-mentioned forearm 4 is two-stage structure, and is connected as one by Saving cortilage 12;As shown in Figure 3, Figure 4,
Saving cortilage 12 has stress buffering and operation dynamics perceptional function, including fixation fork 1201, cross axle 1202, activity fork
1203, cylindrical sleeve 1204, spring 1205 and screw 1206, fixation fork 1201, activity fork 1203, cylindrical sleeve 1204 are hollow circle
Cylindricality, cross axle 1202 are connect by bearing with the rear portion of the front of fixation fork 1201, activity fork 1203 respectively, thus constitute tool
There are the universal joint for being rotated up and down (pitching) and rotating left and right (course) degree of freedom, the back segment at the rear portion and forearm 4 of fixation fork 1201
Be connected, the front of cylindrical sleeve 1204 and the leading portion of forearm 4 are connected, the rear portion of cylindrical sleeve 1204 can slide axially, can opposite rotation
The form of sleeve for turning (rolling) is connect with the front of activity fork 1203, and spring 1205 is cylinder, is installed in the form of package solid
The outside of fixed fork 1201, activity fork 1203 and cylindrical sleeve 1204, the both ends of spring 1205 by two screws 1206 respectively with fixation
Fork 1201 and cylindrical sleeve 1204 are connected.Saving cortilage 12 possesses the mechanical damping degree of freedom of four direction, can effectively weaken screen of trees
Influence of the vibration of reaction force or torque and cutter to air-robot flight attitude.
The opposite course angle for rotating left and right (course) amplitude of both perception is equipped between cross axle 1202 and fixation fork 1201
Sensor is spent, the opposite pitch angle for being rotated up and down (pitching) amplitude of both perception is equipped between cross axle 1202 and activity fork 1203
Sensor is spent, the axial position of both perception axial relative movement (axial direction) amplitude is equipped between cylindrical sleeve 1204 and activity fork 1203
The roll angle sensor of both displacement sensor, perception relative rotary motion (rolling) amplitude, therefore Saving cortilage 12 can perceive
The screen of trees reaction force or torque of four direction suffered by toolbox 6, and as tool feeding or exit, air-robot posture
Or the control input of high fine-tuning, making to remove obstacles, it is more accurate to control.Wherein, photoelectric coding can be used in above-mentioned each angular transducer
Slide rheostat or grating scale etc., the calculating of active force or torque can be used in device or potentiometer etc., upper displacement sensors:Pass through
Tensible rigidity, bending stiffness and the torsion stiffness of displacement and spring that each displacement sensor and angular transducer measure carry out
It calculates and obtains each active force (stretch or compress) or torque (pitching moment, course torque, torsional moment).
Opposite stress-the displacement in both ends (cylindrical sleeve 1204 and fixation fork 1201) of Saving cortilage 12 is demarcated using scaling method
Or the curve by torque-angle, the axial rigidity curve, pitch stiffness curve, course rigidity to obtain Saving cortilage 12 are bent
Line and torsion stiffness curve pass through each stiffness curve and corresponding displacement or angle, you can acquire 12 both ends of Saving cortilage by
Power or torque.
The flight controller be directed to above-mentioned angular transducer and displacement sensor, be equipped with corresponding analog quantity (voltage or
Electric current) or the style interfaces such as digital quantity (including bus), pulsed quantity, frequency quantity, PWM is set also directed to rotor assemblies or bus connects
Mouthful, for signal processor, communication module and cutter controller, bus interface is set.The bus includes CAN, RS-485/
422/232, the types such as Ethernet or airborne-bus.
Preferably, the air-robot control method of Saving cortilage 12 is as follows:
1) information such as attitude angle, angular speed, acceleration, satellite positioning, height and speed of air-robot are acquired in real time,
Combined ground telecommand (passes through the wireless connection of flight controller and ground remote control device), calculates the rotary speed instruction of all rotors
And it exports to above-mentioned rotor assemblies, to realize stabilization and the control of air-robot posture and position;
2) the output control instruction that longitudinal thruster 3 is generated according to ground telecommand realizes aerial machine when screen of trees cleaning
Feeding forward under device people's horizontal attitude exits back control;
3) the screen of trees reaction force (axial direction) or torque (pitching, boat suffered by toolbox 6 are perceived by Saving cortilage 12
To, rolling), once meeting or exceeding scheduled protection thresholding, can determine that cutter is in overload, even cutter controller and
Flight controller automatic synchronization enters protected mode:Control toolbox 6 first brakes and inverts afterwards, at the same control air-robot to
Operation is exited in movement afterwards;
If 4) above-mentioned reaction force or torque are less than scheduled protection thresholding, by the control as air-robot movement fine tuning
System input, specific control method are as follows:
A the axial force of the perception of Saving cortilage 12 when removing obstacles) is set as X, backward for just, corresponding operation thresholding is λX, it is ineffective
Quick area is δX, wherein λX> 0,0≤δX< λX, have:
If --- X<0, judgement air-robot by screen of trees forward direction pulling force, flight controller can take following measure it
One:The fine tuning 1. control air-robot travels forward, continues operation of currently removing obstacles if X forward direction increases, if X is constant or negative sense increases
Add, goes to 2.;2. control air-robot enters floating state, while earthward personnel send safety report by communication module
Alert information, in the hope of manual intervention;
If --- X < λX-δX, flight controller control air-robot travels forward fine tuning, so that axial force is increased, realization
Axially self operation is fed;
If --- | X- λX|≤δX, flight controller control air-robot holding hovering, axial feeding zero;
If --- X > λX+δX, flight controller control air-robot move fine tuning backward, so that axial force is reduced, realization
Axially self protection retracts.
B the course torque of the perception of Saving cortilage 12 when removing obstacles) is set as N, and it is just that corresponding operation thresholding is to the right to overlook
λN, dead band δN, wherein λN> 0,0≤δN< λN, have:
If --- | N | < λN-δN, flight controller controls air-robot to making | N | increased direction move fine tuning and navigates
To the horizontal lateral automatic job feeding of realization;
If --- | | N |-λN|≤δN, the current course of flight controller control air-robot holding, the horizontal lateral amount of feeding
It is zero;
If --- | N | > λN+δN, flight controller controls air-robot to making | N | the direction of reduction move fine tuning and navigates
To realizing level, laterally automatic protection retracts.
C the pitching moment of the perception of Saving cortilage 12 when removing obstacles) is set as M, upwards for just, corresponding dead band is δM,
In, δM>=0, have:
If --- | M | > δM, flight controller controls air-robot to making | M | the direction of reduction moves height-fine adjustment;
If --- | M |≤δM, flight controller control air-robot holding present level.
Preferably, more rotor platforms 1 include platform support, the multiple rotor assemblies being fixedly connected on platform support (number
The even number that amount is >=4), rotor assemblies include rotor and rotor motor, and rotor is fixedly connected on the output shaft of rotor motor, rotation
Wing motor is fixedly connected on platform support.
Preferably, toolbox 6 includes tool bar, the operation cutter being connected on tool bar, insertion or is installed on cutter
The cutter controller of bar, operation cutter are connected with driving motor, and driving motor is fixedly connected on the front end of tool bar.
Preferably, the thrust axis of above-mentioned longitudinal thruster 3 is overlapped with the Plane of rotation of operation cutter so that is longitudinally propelling
The feeding thrust forward direction of device 3 is applied on operation cutter, and the pitch attitude to air-robot is avoided to generate additional moment, to
Be conducive to stabilization and the control of air-robot.
Preferably, the lower end of above-mentioned fuselage 2 is provided with vibration absorber, and above-mentioned forearm 4 and postbrachium 5 are distinguished by vibration absorber
It is connected to the lower end of fuselage 2, vibration absorber is spring-dampers structure.
Preferably, above-mentioned forearm 4 and postbrachium 5 actively pass through vibration absorber, and are locked stretch out by clamping screw respectively
Thus length can be adjusted the whole center of gravity of air-robot.
Preferably, undercarriage is equipped with below air-robot.
Advantages of the present invention is as follows:
1) laser radar, multi-cam and height finding radar are carried on more rotor platforms, are constructed three and are combined environment sense
Know pattern, detection ambient enviroment object is conducive to implement the high-precision avoidance control of air-robot at a distance from air-robot
System, the framework and pilotless automobile technology are similar;
2) the tandem compound framework for using signal processor+flight controller, by the soft or hard of environment detection and avoidance obstacle
Part function is rationally divided, and system function and the modularization of realization are realized, and the independent of modules is facilitated to realize, upgrade and tie up
Shield;
3) air-robot provides robot lift by more rotor platforms and implements attitude stabilization and controlled with position, by longitudinal direction
Propeller implements the operation tool feeding removed obstacles and exits control, and the control mode decoupling of this combination driving is advantageous to
Project Realization;
4) Saving cortilage set by has the mechanical damping degree of freedom of four direction, can effectively weaken screen of trees reaction force
Or influence of the vibration of torque and operation cutter to air-robot flight attitude;
5) thrust axis of longitudinal thruster is arranged and is overlapped with the Plane of rotation of operation cutter so that longitudinal thruster
Feeding thrust forward direction is applied on operation cutter, avoids the pitch attitude to air-robot from generating additional moment, to advantageous
Stabilization in air-robot and control;
6) influence of the vibration to body of operation cutter can be effectively filtered out by vibration absorber;
7) battery pack is placed in the rear portion of air-robot, plays good counterweight effect, while forearm machine opposite with postbrachium
Body can be elastic and be locked, and the center of gravity for quickly adjusting air-robot is conducive to;
8) air-robot with automatic obstacle avoidance functions will be with broader practice foreground;
9) effect that can be cleared up the formalness of screen of trees, screen of trees by forward sight camera carries out close-ups, and real
When monitor operation cutter working condition.
The above description is merely a specific embodiment example, scope of protection of the present invention is not limited thereto.It is ripe
It knows the those skilled in the art in the technical scope disclosed by the present invention, change or replacement mode can be easily found, these
It should be covered by the protection scope of the present invention.For this purpose, protection scope of the present invention should be with the protection model of the claim
Subject to enclosing.
Claims (9)
1. a kind of screen of trees of automatic obstacle-avoiding clears up air-robot, it is characterised in that:It is dynamic including providing flight for air-robot
More rotor platforms (1) of power, are installed on indulging on fuselage (2) at the fuselage (2) being fixedly connected on immediately below more rotor platforms (1)
To propeller (3) and fuselage (2) longitudinal axes parallel and it is fixed on the forearm (4) of fuselage (2) front end, is put down with fuselage (2) longitudinal axis
Row and after being fixed on the postbrachium (5) of fuselage (2) rear end, the toolbox (6) for being fixed on forearm (4) front end, being fixed on postbrachium (5)
The battery pack (7) at end, the laser radar (8) for being installed on more rotor platform (1) central upper portions, the survey for being arranged in fuselage (2) bottom
High radar (10), more rotor platforms (1) front sides, both sides and battery pack (7) rear side are arranged with camera (9).
2. a kind of screen of trees of automatic obstacle-avoiding according to claim 1 clears up air-robot, it is characterised in that:It further include phase
Signal processor, flight controller and the communication module to connect;Laser radar (8), camera (9) and height finding radar (10) are even
It is connected to the signal input part of signal processor, flight controller is connected to power plant module, the longitudinal thruster of more rotor platforms (1)
(3) cutter controller and Inertial Measurement Unit of driving motor, toolbox (6), barometertic altimeter, satellite navigation receive
Machine, communication module is for ground station transmission flying quality and onboard image.
3. a kind of screen of trees of automatic obstacle-avoiding according to claim 1 clears up air-robot, it is characterised in that:Fuselage (2)
For front and back narrow structure, air-robot is bilateral symmetry.
4. a kind of screen of trees of automatic obstacle-avoiding according to claim 1 clears up air-robot, it is characterised in that:Battery pack
(7) include being controlled to toolbox (6), to more rotor platforms (1), to longitudinal thruster (3), to signal processor, to flight
Device and the battery powered to airborne sensor.
5. a kind of screen of trees of automatic obstacle-avoiding according to claim 1 clears up air-robot, it is characterised in that:Forearm (4)
It for two-stage structure, and is connected as one by Saving cortilage (12), Saving cortilage (12) includes fixation fork (1201), cross axle
(1202), activity fork (1203), cylindrical sleeve (1204), spring (1205) and screw (1206), fixation fork (1201), activity fork
(1203), cylindrical sleeve (1204) is hollow cylindrical, cross axle (1202) respectively by bearing and fixation fork (1201) before
The rear portion connection in portion, activity fork (1203), thus constitutes universal joint, and the rear portion of fixation fork (1201) is solid with the back segment of forearm (4)
Even, the front of cylindrical sleeve (1204) and the leading portion of forearm (4) are connected, the rear portion of cylindrical sleeve (1204) can slide axially, can phase
The form of sleeve of rotation is connect with the front of activity fork (1203), spring (1205) is cylinder, is installed in the form of package
Fixation fork (1201), activity pitch the outside of (1203) and cylindrical sleeve (1204), and the both ends of spring (1205) pass through two screws
(1206) it is connected respectively with fixation fork (1201) and cylindrical sleeve (1204).
6. the barrier-avoiding method of air-robot is cleared up according to a kind of screen of trees of any automatic obstacle-avoidings of claim 1-5,
It is characterized in that:Steps are as follows for this method:
1) installation laser radar (8) makes its laser scanning line be higher than more rotor wing rotation planes, by laser radar (8) to ring around
Border carry out 360 degree of horizontal sweeps, detect and perceive air-robot surrounding object to laser radar (8) distance;
2) camera (9) obtains air-robot environmental images all around and is sent into signal processor, then uses machine
Vision algorithm is to the conducting wire inside and outside electric transmission line channel, shaft tower, trees and other objects are identified and framing, wherein
Shaft tower includes insulator and gold utensil, and other objects include building, bridge and chain of mountains landform;
3) the two-dimensional distance information that step 1) is perceived is merged with the obstacle information that step 2) is identified, acquired disturbance
The range information of object carries out air-robot automatic obstacle avoidance:
If --- the barrier perceived is conducting wire, shaft tower, channel outer trees and other objects, and these barriers and aerial machine
The distance of device people≤preset safety distance threshold even air-robot implements automatic obstacle avoidance control, while earthward monitoring
It stands and sends security alarm information;
If --- the barrier perceived is the trees in channel, and air-robot implementation is removed obstacles, while earthward monitoring station
Transmission is removed obstacles action message;
--- other than above-mentioned two situations, it is not necessarily to avoidance obstacle;
4) height of the air-robot with respect to ground is detected by the height finding radar (10) installed downwards, this relative altitude is used for
The height of air-robot is kept, and carries out air-robot height avoidance.
7. a kind of barrier-avoiding method of the screen of trees cleaning air-robot of automatic obstacle-avoiding according to claim 6, feature exist
In:Step 2) is identified environmental objects using machine vision algorithm as follows with the method for positioning:
A) for conducting wire barrier, it is detected using straight line parameter method:It first, will using maximum variance between clusters
All barriers are partitioned into from background and carry out binaryzation;Then, to all pixels value in the plane of delineation be 0 pixel
Carry out straight line modeling, it is assumed that it is (x, y) that one, which is not 0 pixel point coordinates, then it necessarily belongs to straight line, the side of the straight line
Journey is represented by:
ρ=x cos θ+y sin θs
Wherein ρ indicates that distance of the straight line relative to origin belonging to the point, θ indicate normal and x of the straight line relative to origin
The angle of axis positive direction, by the spaces ρ-θ maximizing, converting all the points in image to linear equation;Finally, root
It is power transmission line according to the Straight Line Identification that angle and length are met to given threshold condition and its two-dimentional geological information is fed back into information
Blending algorithm;
B) for shaft tower and trees, object detection is carried out using depth convolutional neural networks method:
First, shaft tower and Tree image library are established, image library process is being established in model training when for target detection and detection
In, it needs to acquire the object images under various illumination, angle, focal length, weather environment, and mark;
Secondly, detection network model is established, detection model is made of 5 convolutional layers, 3 pond layers and 2 full articulamentums, is detected
Model is divided into three candidate regions selection, feature extraction, the classification of object and the recurrence of detection block part compositions, and candidate regions selection is adopted
With SelectiveSearch methods to 2000 RoIs of every image zooming-out;Then these RoIs are subjected to convolution sum pondization behaviour
Make, to extract the feature of RoIs;There are two outputs for the classification and detection of object:One output is used for describing the affiliated classes of each RoI
Probability value p=(the p of type0,p1,...,pu), Probability p is calculated by the full articulamentum combination softmax of last layer;It is another
A output layer is that boundingbox returns layer, this layer exports the co-ordinate position information t of target objectu=(tx u,ty u,tw u,th u),
Wherein, u is that the number of barrier indexes, and (x, y) indicates that the top left co-ordinate of target object, (w, h) indicate the width of target object
Degree and height, RoI is region of interest;
When making classification training and position recurrence for the RoIs of each tape label, a corresponding ground- will be all exported
The boundingbox regressive objects v of truth type labels u and ground-truth.Then multitask loss function L is calculated:
L(p,u,tu, v) and=Lcls(p,u)+λ[u≥1]Lloc(tu,v)
Wherein, Lcls(p, u)=- log puIt is lost for the log of true classification u;Section 2 returns loss function LlocIt is real object
Location information v=(vx,vy,vw,vh) and prediction object location informationBetween error loss;Parameter lambda is used
In balancing two loss functions, λ=1 is taken;
For LlocUsing the smooth loss functions of L1:
Wherein
In above formula, d=tu-v;
Finally, trained network is transplanted on the embedded image processing computer accelerated with GPU, in real time in image
Insulator be labeled and its two-dimentional geological information fed back into information fusion algorithm;
C) trees outside line channel and other barriers, are the target of air-robot automatic obstacle avoidance, pass through laser radar
It is detected, distance is less than the object of threshold value as barrier.
8. a kind of barrier-avoiding method of the screen of trees cleaning air-robot of automatic obstacle-avoiding according to claim 6, feature exist
In:The obstacle distance information acquisition method that information is merged in step 3) is as follows:
First, the collected range information of laser radar and four-way camera the image collected are carried out pair by coordinate transform
Standard obtains each physical coordinates of the pixel relative to laser radar in image;
Secondly, the two-dimensional geometry information of the conducting wire found out in step 2), shaft tower, trees and other barriers is passed through into coordinate transform
For three-dimensional information, transform method is:
Wherein (x, y) is the two-dimensional coordinate of a pixel in camera, (u0,v0) indicate camera photocentre coordinate, fxAnd fyTable respectively
Show the focal length of camera in the x and y direction, the component of the distance at the practical relative image center of barrier in the x and y direction is respectively
X and Y, Z are the affiliated straight line (Z of conducting wirel) or shaft tower barrier (Zd) belonging to rectangular area average distance;
Again, the maximum distance region that laser radar detects is found out, the average distance in the region is denoted as Zr, compare Zr、ZlWith
ZdSize, minimum distance is denoted as Zp, then ZpThe range information of barrier as in step 3);
Finally, Z is obtainedpThe size of opposite image rectangle or linear region, i.e. (X, Y, Zp), and as the foundation of avoidance.
9. a kind of barrier-avoiding method of the screen of trees cleaning air-robot of automatic obstacle-avoiding according to claim 6, feature exist
In:Step 3) and the control method of the hollow middle robot obstacle-avoiding of step 4) are:
Need implement automatic obstacle avoidance barrier include:Power line, shaft tower in electric power corridor, trees outside electric power corridor and its
Its barrier;
If 1) only preceding find barrier, and obstacle distance≤safety distance threshold to camera (9), then carry out immediately urgent
Braking, and ground station sends a warning message, then by operating personnel by earth station's monitor video or site observation, to aerial
Robot frontal scene carries out implementing further avoidance operation after again judging:1. for conducting wire or shaft tower class barrier, foundation
It slowly climbs or is displaced sideways upwards until barrier is removed from camera (9) visual field or Autonomous landing waits for work in former course
It is recycled as personnel or manual control air-robot is leapt or cut-through object;2. for the tree to be cleaned in channel
Wood then continues to implement operation of removing obstacles forward;
2) if only left or right finds barrier, and obstacle distance≤safety distance threshold to camera (9), then carry out immediately
Then emergency braking is slowly moved to barrier opposite direction, until barrier is removed from camera (9) visual field;
If 3) preceding, left and right has been found that barrier, and obstacle distance≤safety distance threshold to camera (9), then immediately into
Row emergency braking, and ground station sends a warning message, then by operating personnel by earth station's monitor video or site observation,
To before air-robot, left and right carries out to scene implementing the operation of further avoidance after again judging:Slowly vertically climb until
Barrier is removed from camera (9) visual field or Autonomous landing waits for staff's recycling, or by manually taking over manual reality comprehensively
Apply operation of removing obstacles;
If 4) height finding radar (10) detection air-robot with respect to ground height≤safe altitude threshold value, even aerial machine
People rises to implement active height avoidance obstacle, and earthward monitoring station sends security alarm information by communication module.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810192078.6A CN108568868B (en) | 2018-03-08 | 2018-03-08 | Automatic obstacle avoidance tree obstacle cleaning aerial robot and obstacle avoidance method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810192078.6A CN108568868B (en) | 2018-03-08 | 2018-03-08 | Automatic obstacle avoidance tree obstacle cleaning aerial robot and obstacle avoidance method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108568868A true CN108568868A (en) | 2018-09-25 |
CN108568868B CN108568868B (en) | 2024-03-19 |
Family
ID=63576101
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810192078.6A Active CN108568868B (en) | 2018-03-08 | 2018-03-08 | Automatic obstacle avoidance tree obstacle cleaning aerial robot and obstacle avoidance method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108568868B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109960880A (en) * | 2019-03-26 | 2019-07-02 | 上海交通大学 | A kind of industrial robot obstacle-avoiding route planning method based on machine learning |
CN110265915A (en) * | 2019-06-19 | 2019-09-20 | 广东电网有限责任公司 | One kind tying up tree device |
CN110752540A (en) * | 2019-10-21 | 2020-02-04 | 黄碧茶 | Cable maintenance loading device capable of crossing barrier |
WO2020082947A1 (en) * | 2018-10-22 | 2020-04-30 | 科沃斯机器人股份有限公司 | Travel control method, device, and storage medium |
CN113312992A (en) * | 2021-05-18 | 2021-08-27 | 中山方显科技有限公司 | Dynamic object sensing and predicting method based on multi-source sensor information fusion |
WO2021226776A1 (en) * | 2020-05-11 | 2021-11-18 | 华为技术有限公司 | Vehicle drivable area detection method, system, and automatic driving vehicle using system |
CN114097450A (en) * | 2021-11-23 | 2022-03-01 | 南方电网电力科技股份有限公司 | Robot for cleaning tree obstacles on power line |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102412530A (en) * | 2011-12-23 | 2012-04-11 | 北京国网富达科技发展有限责任公司 | Line navigation amphibious power circuit comprehensive maintenance robot and circuit maintenance method thereof |
CN104682267A (en) * | 2015-03-26 | 2015-06-03 | 国家电网公司 | Power transmission line fault removal device based on multi-axle air vehicle |
CN104724288A (en) * | 2015-03-26 | 2015-06-24 | 国网辽宁省电力有限公司抚顺供电公司 | Overhead conductor foreign matter processing rotorcraft |
US20160114887A1 (en) * | 2002-10-01 | 2016-04-28 | Dylan T X Zhou | Amphibious vertical takeoff and landing unmanned system and flying car with multiple aerial and aquatic flight modes for capturing panoramic virtual reality views, interactive video and transportation with mobile and wearable application |
CN106125092A (en) * | 2016-08-08 | 2016-11-16 | 成都希德电子信息技术有限公司 | A kind of unmanned plane automatic obstacle-avoiding system and method based on two-dimensional laser radar |
CN106143890A (en) * | 2016-08-05 | 2016-11-23 | 微普安全科技(徐州)股份有限公司 | A kind of fire-fighting intelligent search robot |
CN106444819A (en) * | 2016-12-02 | 2017-02-22 | 国网四川省电力公司电力应急中心 | UWB-array-based automatic obstacle avoidance system and method for drone |
CN106774405A (en) * | 2016-12-30 | 2017-05-31 | 华南农业大学 | Orchard plant protection unmanned plane obstacle avoidance apparatus and method based on three-level avoidance mechanism |
CN107703519A (en) * | 2017-10-17 | 2018-02-16 | 杭州瓦屋科技有限公司 | Plant protection unmanned plane obstacle detection method based on laser displacement sensor |
CN107706816A (en) * | 2017-10-27 | 2018-02-16 | 广东电网有限责任公司揭阳供电局 | A kind of electric line foreign matter remove device |
-
2018
- 2018-03-08 CN CN201810192078.6A patent/CN108568868B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160114887A1 (en) * | 2002-10-01 | 2016-04-28 | Dylan T X Zhou | Amphibious vertical takeoff and landing unmanned system and flying car with multiple aerial and aquatic flight modes for capturing panoramic virtual reality views, interactive video and transportation with mobile and wearable application |
CN102412530A (en) * | 2011-12-23 | 2012-04-11 | 北京国网富达科技发展有限责任公司 | Line navigation amphibious power circuit comprehensive maintenance robot and circuit maintenance method thereof |
CN104682267A (en) * | 2015-03-26 | 2015-06-03 | 国家电网公司 | Power transmission line fault removal device based on multi-axle air vehicle |
CN104724288A (en) * | 2015-03-26 | 2015-06-24 | 国网辽宁省电力有限公司抚顺供电公司 | Overhead conductor foreign matter processing rotorcraft |
CN106143890A (en) * | 2016-08-05 | 2016-11-23 | 微普安全科技(徐州)股份有限公司 | A kind of fire-fighting intelligent search robot |
CN106125092A (en) * | 2016-08-08 | 2016-11-16 | 成都希德电子信息技术有限公司 | A kind of unmanned plane automatic obstacle-avoiding system and method based on two-dimensional laser radar |
CN106444819A (en) * | 2016-12-02 | 2017-02-22 | 国网四川省电力公司电力应急中心 | UWB-array-based automatic obstacle avoidance system and method for drone |
CN106774405A (en) * | 2016-12-30 | 2017-05-31 | 华南农业大学 | Orchard plant protection unmanned plane obstacle avoidance apparatus and method based on three-level avoidance mechanism |
CN107703519A (en) * | 2017-10-17 | 2018-02-16 | 杭州瓦屋科技有限公司 | Plant protection unmanned plane obstacle detection method based on laser displacement sensor |
CN107706816A (en) * | 2017-10-27 | 2018-02-16 | 广东电网有限责任公司揭阳供电局 | A kind of electric line foreign matter remove device |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020082947A1 (en) * | 2018-10-22 | 2020-04-30 | 科沃斯机器人股份有限公司 | Travel control method, device, and storage medium |
US11487297B2 (en) | 2018-10-22 | 2022-11-01 | Ecovacs Robotics Co., Ltd. | Method of travel control, device and storage medium |
CN109960880A (en) * | 2019-03-26 | 2019-07-02 | 上海交通大学 | A kind of industrial robot obstacle-avoiding route planning method based on machine learning |
CN109960880B (en) * | 2019-03-26 | 2023-01-03 | 上海交通大学 | Industrial robot obstacle avoidance path planning method based on machine learning |
CN110265915A (en) * | 2019-06-19 | 2019-09-20 | 广东电网有限责任公司 | One kind tying up tree device |
CN110752540A (en) * | 2019-10-21 | 2020-02-04 | 黄碧茶 | Cable maintenance loading device capable of crossing barrier |
WO2021226776A1 (en) * | 2020-05-11 | 2021-11-18 | 华为技术有限公司 | Vehicle drivable area detection method, system, and automatic driving vehicle using system |
CN113312992A (en) * | 2021-05-18 | 2021-08-27 | 中山方显科技有限公司 | Dynamic object sensing and predicting method based on multi-source sensor information fusion |
CN114097450A (en) * | 2021-11-23 | 2022-03-01 | 南方电网电力科技股份有限公司 | Robot for cleaning tree obstacles on power line |
Also Published As
Publication number | Publication date |
---|---|
CN108568868B (en) | 2024-03-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108321722A (en) | Automatic obstacle-avoiding can buckling screen of trees cleaning air-robot and barrier-avoiding method | |
CN108568868A (en) | A kind of the screen of trees cleaning air-robot and barrier-avoiding method of automatic obstacle-avoiding | |
CN108292140B (en) | System and method for automatic return voyage | |
US10475209B2 (en) | Camera calibration | |
McGee et al. | Obstacle detection for small autonomous aircraft using sky segmentation | |
CN109388150B (en) | Multi-sensor environment mapping | |
JP6312178B2 (en) | Navigation system, method for evaluating state information of transport aircraft, and apparatus for evaluating state information of transport aircraft | |
CN206057974U (en) | A kind of obstacle avoidance system applied on rotor wing unmanned aerial vehicle | |
CN112799426B (en) | Unmanned aerial vehicle navigation control system and method based on big data analysis | |
CN110525650B (en) | Unmanned aerial vehicle and control method thereof | |
WO2017160192A1 (en) | Method for precision landing an unmanned aerial vehicle | |
CN107783547A (en) | Post disaster relief rotor wing unmanned aerial vehicle obstacle avoidance system and method | |
CN113240881B (en) | Fire identification system based on multi-feature fusion | |
Meivel et al. | Remote sensing for UREA Spraying Agricultural (UAV) system | |
Flores et al. | Aerial photography for 3D reconstruction in the Peruvian Highlands through a fixed-wing UAV system | |
CN116661497A (en) | Intelligent aerocar | |
CN208675850U (en) | A kind of automatic obstacle-avoiding can buckling screen of trees clear up air-robot | |
CN108154715B (en) | Lateral collision monitoring method | |
Lacroix et al. | The autonomous blimp project of LAAS/CNRS-Current status and research challenges | |
Cho et al. | Stabilized UAV flight system design for structure safety inspection | |
Priambodo et al. | A Vision and GPS Based System for Autonomous Precision Vertical Landing of UAV Quadcopter | |
KR102585428B1 (en) | An automatic landing system to guide the drone to land precisely at the landing site | |
CN210113177U (en) | Electric power patrols line unmanned aerial vehicle based on VR and multisensor | |
CN113778134A (en) | Ground station for coal mine environment | |
JPH0757045B2 (en) | Power line monitoring system |
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 | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20211213 Address after: No.17 Binhe Road, Nanming District, Guiyang City, Guizhou Province Applicant after: Guizhou Power Grid Co.,Ltd. Address before: No.17 Binhe Road, Nanming District, Guiyang City, Guizhou Province Applicant before: Guizhou Power Grid Co.,Ltd. Applicant before: NANJING TAISIDE INTELLIGENT TECHNOLOGY Co.,Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |