CN105866762B - Laser radar automatic calibrating method and device - Google Patents
Laser radar automatic calibrating method and device Download PDFInfo
- Publication number
- CN105866762B CN105866762B CN201610107997.XA CN201610107997A CN105866762B CN 105866762 B CN105866762 B CN 105866762B CN 201610107997 A CN201610107997 A CN 201610107997A CN 105866762 B CN105866762 B CN 105866762B
- Authority
- CN
- China
- Prior art keywords
- msub
- calibration
- parameter
- mrow
- radar
- 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.)
- Active
Links
Classifications
-
- 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
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
- G01S7/497—Means for monitoring or calibrating
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Optical Radar Systems And Details Thereof (AREA)
Abstract
A kind of laser radar calibration method and device, wherein method comprise the following steps：External parameter is calibrated；The external parameter calibration steps includes：Radar is moved with car, collects radar reflection data, and the data being collected into are substituted into dot matrix cloud energy equation, is solved and is caused energy equation value for minimum external calibration parameter；Prior art is different from, above-mentioned technical proposal can be analyzed the data that laser radar is passed back in real time, active correction laser radar inherent parameters, reached quick calibration of laser radar, improve scan data accuracy rate, improve the effect of scan image identification.
Description
Technical field
The present invention relates to unmanned vehicle navigation technical field, more particularly to a kind of mobile lidar automatic calibrating method and dress
Put.
Background technology
Laser radar sensor measurement performance due to its superior distance measurement function and accurately, has been widely used in work
Industry and smart field.Due to its detectivity to surrounding enviroment, laser radar has also been matched somebody with somebody as the standard of unmanned vehicle in recent years
One of put.Main application of the laser radar in unmanned vehicle field is to perceive car body surrounding enviroment, the other systems of it and vehicle
After (such as GPS, INS) is engaged, can be used for the drafting of map, the positioning of object (traffic lights and traffic marking detection) with
And condition of road surface detection.Laser radar traditionally only has a branch of rotation light, therefore its calibration method is also relatively simple.At present,
The laser radar of a new generation, which has been configured with, can realize the light beam rotated while different angle is strafed, so as to compared to tradition
Laser radar the data volume of one magnitude can be provided, realize that preferably drawing function, object detection and scene understand, can fit
Answer more advanced algorithm.
In order to fully adapt to and using this mass data, it is necessary to use the calibration method of higher level.First, compared to
For traditional laser radar, new calibration method will can support the distance of dozens or even hundreds of light beam and the reading of angle.
Secondly, function (such as drafting of Road Detection, map and fixed of laser radar intensity return value is used for unmanned vehicle others
Position), it is of great importance that intensity return value needs to be confirmed by different light beams.Parameter used in terms of above-mentioned two
Amount, all manually to train to seem and improper to calibrate laser radar.
The content of the invention
For this reason, it may be necessary to provide a kind of laser radar automatic calibrating method that can be adaptive, solves unmanned trailer-mounted radar certainly
The problem of master positioning.
To achieve the above object, a kind of laser radar calibration method is inventor provided, is comprised the following steps：External parameter
Calibration；
The external parameter calibration steps includes：Radar moves with car, collects radar reflection data, the data that will be collected into
Dot matrix cloud energy equation is substituted into, solves and causes energy equation value for minimum external calibration parameter；
Wherein dot matrix cloud energy equation is：
Wherein J is the energy of a cloud, B laser beams sum, and N is closes on light beam number, and k is the reflected beams number, pkIt is k-th point
Position under bodywork reference frame, mkIt is light beam biClosest to p in receiving the institute of reflection a littlekPoint or point set, ηkIt is mkPoint
Normal vector, wk| | pk-mk| | ＜ dmaxIt is 1 during establishment, is otherwise 0.
Further, in addition to step, inner parameter calibration, the inner parameter calibration steps include：Solution makes it possible to
Measure the minimum internal calibration parameter of equation value.
Specifically, the inner parameter is included per beam laser level and vertical angle and range correction value.
Specifically, the external calibration parameter include radar with respect to the plane coordinates parameter of car body, height, the angle of pitch, turn over
Roll angle or course angle.
Further, in addition to step, reception are calibrated, specifically, radar observation data are substituted into following formula：
Wherein j numbers for laser beam, and a is observed strength, and c is calibration result.
Further, in addition to step, Bayesian model is established to the noise properties and uncertainty of every beam laser beam, counted
Calculate in the case where given map grid is distributed, per the distribution of beam laser intensity return value.
A kind of laser radar calibrating installation, including collection module, external calibration module；
The collection module is used to collect radar reflection data when radar moves with car；
The external calibration module is used to the data being collected into substituting into dot matrix cloud energy equation, solves and causes energy equation
It is worth for the external calibration parameter of minimum；
Wherein dot matrix cloud energy equation is：
Wherein J is the energy of a cloud, B laser beams sum, and N is closes on light beam number, and k is the reflected beams number, pkIt is k-th point
Position under bodywork reference frame, mkIt is light beam biClosest to p in receiving the institute of reflection a littlekPoint or point set, ηkIt is mkPoint
Normal vector, wk| | pk-mk| | ＜ dmaxIt is 1 during establishment, is otherwise 0.
Further, in addition to internal calibration module, the internal calibration module are used for solution and cause energy equation value most
Small internal calibration parameter.
Specifically, the inner parameter is included per beam laser level and vertical angle and range correction value.
Specifically, the external calibration parameter include radar with respect to the plane coordinates parameter of car body, height, the angle of pitch, turn over
Roll angle or course angle.
Further, in addition to calibration module is received, the reception calibration module is used under radar observation data are substituted into
Formula：
Wherein j numbers for laser beam, and a is observed strength, and c is calibration result.
Further, in addition to bass leaf module, the bass leaf module be used for the noise properties of every beam laser beam and
Uncertainty establishes Bayesian model, calculates in the case where given map grid is distributed, point per beam laser intensity return value
Cloth.
Prior art is different from, above-mentioned technical proposal can be analyzed the data that laser radar is passed back in real time, actively
Calibration of laser radar inherent parameters, quick calibration of laser radar is reached, has improved scan data accuracy rate, improved scan image and distinguish
The effect of knowledge and magnanimity.
Brief description of the drawings
Fig. 1 is the laser radar automatic calibrating method flow chart described in the specific embodiment of the invention；
Fig. 2 is the adjacent two beams laser scanning figure described in the specific embodiment of the invention；
Fig. 3 is the normal vector schematic diagram described in the specific embodiment of the invention；
Fig. 4 is Bayes's calibrating patterns schematic diagram described in certain embodiment of the invention；
Fig. 5 is the Laser Radar Scanning effect diagram described in certain embodiment of the invention；
Fig. 6 is Laser Radar Scanning effect diagram after the calibration described in certain embodiment of the invention；
Fig. 7 is that the internal reference described in the specific embodiment of the invention calibrates effect diagram；
Fig. 8 is the parameter adjustment design sketch described in the specific embodiment of the invention；
Fig. 9 is the expection ambient intensity schematic diagram described in the specific embodiment of the invention；
Figure 10 is that strength map is just penetrated in the street described in the specific embodiment of the invention；
Figure 11 is the laser radar self-checking device module map described in the specific embodiment of the invention.
Description of reference numerals：
1100th, collection module；
1102nd, external calibration module；
1104th, internal calibration module；
1106th, calibration module is received；
1108th, Bayes's module.
Embodiment
To describe the technology contents of technical scheme, construction feature, the objects and the effects in detail, below in conjunction with specific reality
Apply example and coordinate accompanying drawing to be explained in detail.
First, general thought
Present invention proposition is a kind of brand-new automatically (to be swashed with Velodyne HD 64-E herein to multi-beam laser radar
Optical radar is example, but is used in the laser radar that any line is swept) inner parameter and external parameter carry out calibration method, and
During and need not use specific calibration target or human metering.After laser radar being loaded with a unmanned vehicle,
Along with the reading of Inertial Measurement Unit, our algorithm is not in the case where needing map auxiliary, only with regard in several seconds of machine
To the reading of any environmental scanning in periphery, with regard to individual laser radar sensor parameters up to a hundred can be calculated.
The main automatic calibration algorithm for being divided into three parts of the invention.First external calibration be primarily used to determine by
The laser radar caused by the difference of installation pose can be represented with six parameters compared to the relative position of car body, that is, led to
Cross translation matrix and spin matrix can realize conversion from laser radar coordinate to car body coordinate.The second part is related to laser
The inner parameter calibration of radar, is primarily used to estimate the optimal level and vertical angle of every beam light, and is read per beam light
Take the distance of scope to correct, i.e., the deviation of each generating laser and radar fix system is calibrated inside laser radar.
3rd part is to derive model with Bayes to enter reception value corresponding to different surfaces emissivity in environment every Shu Jiguang
Row modeling.So, the relative smooth on scanning distance of the intensity map after calibration correction, different objects or ground can be protruded
The correlated characteristic of shape.
2nd, external parameter is calibrated
For multi-line laser radar, what external calibration considered is phase of the whole laser radar for car body local Coordinate System
To installation site, and what internal calibration considered is the interior location of each single beam laser for radar in itself.In this section, we
Assume that in the case of known radar internal calibration, calculate external calibration.In a practical situation, if two kinds are calibrated not
Know, then two calibration procedure interactions can be carried out, until two models are all optimal parameter Estimation.
The calibration method of our two classes is all based under such hypotheses：Laser radar receives the intensity letter returned
When number being reflected in three-dimensional imaging, these observations can't show the pattern of random distribution in space.This premise is false
Being located in reality to reach.In fact, because the point of return represents the reflection of the physical surface of observed objects, one
The laser radar sensor moved along known trajectory after correct calibration is impossible to provide to present in three dimensions
The point cloud reading of random distribution.This relatively good understanding in reality, because laser can reflect shape when running into barrier
Into intensive point set, and spacious place is no reflection echo, is also blank on point cloud chart, such to every Shu Jiguang.Therefore, I
On the premise of the calibration method that proposes is the weak hypothesis that relative coherent condition is presented in space based on pip.
Referring to Fig. 1, be the flow chart of laser radar automatic calibrating method of the present invention, in certain embodiments, the present invention
Method starts from step S100 radars and moved with car, collects radar reflection data, the data substitute point that step S102 will be collected into
Battle array cloud energy equation, solve and cause energy equation value for minimum external calibration parameter.
Fig. 2 is illustrated after unmanned vehicle runs the several seconds along known trajectory, the whole that two adjacent beam laser were collected
Data, and be presented on 3D rendering.The data mark that wherein beam of laser is collected into red, mark that another beam of laser is collected into
White.Clearly because the full angle of laser is strafed, largely two beam laser have all got to identical surface, so into
As approximate.
Whole LIDAR can be represented relative to the installation site of bodywork reference frame with six parameters：X-axis (is used for weighing
Measure length), y-axis (is used for weighing width), and both can be as plane coordinates parameter of the laser radar with respect to car body, z
Axle (is used for weighing height), and roll angle, the angle of pitch and course angle.The origin (0,0,0) of coordinate system and the direction of origin
It is then to be determined by selected bodywork reference frame, such as can be using the three-dimensional coordinate of car body pose system and direction as just
Initial value.
Compared to existing calibration method, our method only needs to make simple hypothesis for vehicle-surroundings environment：
Surrounding enviroment geo-stationary, and contain some 3D features (not being a piece of spaciousness) in environment, without other conditions.To be terrible
To accurate calibration parameter, we allow unmanned vehicle to drive a segment distance along known trajectory and record laser radar data.Car
Running orbit can be arbitrary, but the change of course angle must be included, so can just measure laser radar and exist
Corrected value on x/y plane, therefore car body please not allowed only to keep straight and moved ahead.In addition, we are not intended to correct sensor
Height because our car body is to keep approximately parallel with ground, and by ranging and vehicle operating range, height
Calculate very simple.It is inessential that posture information of the car body under global coordinates system calibrates this step in laser radar, at me
Experiment in be not related to the use of existing map yet, so being only the need for the pose change information of car body in itself.
The posture information of car body itself can be obtained by a series of approach, for example be obtained from tire encoder and inertial guidance system
, or obtained from the GPS system of a real time correction.Reiterate, only vehicle travelled along track during phase
It is only the relevant information source of calibration algorithm to motion, the global posture information of car body is herein and need not.
Now, we define the energy equation of cloud, can be formed when reaching body surface by each Shu Jiguang intensive
Point group, our equation can punish those points thought of denoising (cluster) away from these point groups：
In above-mentioned formula, B is all laser beam sums, and N is near certain Shu Jiguang it is considered that belonging to neighbouring light
Beam sum, k is for light beam biReflect to be formed progressive alternate a little, pkIt is to be formed under current conversion formula
The k-th point of position under bodywork reference frame, mkIt is light beam biFrom p in sending formed institute a littlekNearest point or point
Collection, ηkIt is mkThe normal vector of point, wkIt is that 1 or 0 depend on | | pk-mk| | ＜ dmaxWhether set up.
In above-mentioned algorithm, we have used the ANN library of University of Maryland's announcement, and to calculate m_k, d_max exists
In follow-up experiment, the threshold values that we take is 20cm, and ours is that all radars return to the information next life received
Into every Shu Jiguang point cloud and normal vector, but in order to improve computational efficiency, just every 16 points we estimate once above-mentioned energy
Measure equation.The Velodyne equipment that we use can return to more than million points each second, and energy equation is carried out to each point
Calculating be unnecessary.If the initial setting up of laser radar is very inaccurate, if we with modern times desktop computer processing
Device, above-mentioned computing is carried out to the laser recording data of general 15 seconds length or so, laser can just be calculated by being respectively necessary for 1 hour
The internal calibration parameter and external calibration parameter of radar, the calibration to receiving intensity will be many soon, generally only need a few minutes.
It is proposed that energy equation and calculate point and have certain similar degree, but essential difference to the ICP error equations of plane
It is two places.First, ours is point of the surface for the point set formation that every beam of laser is generated with it adjacent to laser
Collection contrasts.The benefit done so is the point that wrong calibration between light beam will not significantly affect any single laser generation
The normal vector formed.Second and places different ICP be, what we were handled is not fixed electric cloud, Ren Heyi
The change of individual calibration parameter can all cause the point cloud that laser generates that complicated conversion occurs, because the point in each cloud is not
The same time, sensor different azimuth under observe come.
Normal vector will be individually calculated per Shu Jiguang, this is by making plane is chimeric to be accumulated from vehicle whole track
Generation a little in from each periphery nearest 20 points complete.Because the packing density of more line lasers is higher, often
The peripheral point of individual point is all very intensive, and neighborhood is small.We illustrate these normal vectors in figure 3, and which includes each normal direction
The x of amount, y, z parts.
The point of more line lasers generation is more intensive, it is such be advantageous in that almost any surface on a cloud resolution chart all
Substantially local upper plane.Therefore, the point that beam of laser generates is mapped to what is be made up of it adjacent to the point of laser generation
On surface, in the case of all calibrations are all accurate, only less error can be produced.
After having above-mentioned energy equation, remaining work is exactly that select can be so that the outside that total value is preferably minimized
Calibration parameter.Can although in theory, this be not necessarily a convex function, thus obtain optimal solution within the limited time
Do not ensure, but actually energy equation is still very smooth, the searching trial-and-error method of standard also has good result.
In our method, we alternately optimize translation matrix parameter and spin matrix parameter, until the two all
Reach convergence.For Optimization Solution each time, we search for a kind of grid type, that is, compare be present energy equation and
Value and the energy equation and value drawn after all solution parameters are carried out while adjusted on be possible to direction.Known to us
One coordinate declines time that iteration spent and linearly related with variable quantity, but grid type search for the spent time with
Variable quantity is that the relation of exponential increase is presented, and is possible to combine on direction because he considers.So run into part
During minimum, grid type search can't terminate, simultaneously because translation matrix and spin matrix are each no more than three
Variable, so this grid type way of search is feasible on a computer program.For example, the change of spin matrix is being considered
During change, each roll angle, the angle of pitch, course angle can be increases, constant or reduction, so, compared to current energy
Equation and value (i.e. three variables all remain unchanged), can produce 26 kinds of new energy equations and value.Separately there is the important point to be,
When considering every kind of possible calibration, all points will remap according to the pose of vehicle when each point is acquired
To 3d space.So, the change of calibration parameter will not just distort all points in an identical manner, because calibration parameter adjustment pair
The influence each put depends on the location of vehicle when returned data calculates.
Since larger interval, then iteration reduces to restraining and is spaced and repeats same steps for we, until we
Preferable minimum particle size is reached.And result of this last time optimization solution is our last calibration parameters.To sum up, it is outside
The problem of parametric calibration is solved in unmanned vehicle running, and laser radar is calibrated to own coordinate parameter automatically, reaches
The effect that positional information is quickly and accurately independently provided is arrived so that mobile lidar being capable of autonomous positioning.
3rd, inner parameter is calibrated
In fact, Velodyne laser radar product has the algorithm for providing internal reference calibration and Matlab modelings (to refer in itself
Product description), the method that he provides belongs to static debugging in a certain direction, and its result precision, which will be less than, to be presented herein
Algorithm.The method that Velodyne is provided mainly determines that plane sits calibration using two：One is according at 25.04 meters of radar
Plane, another is according to the plane at 2.4 meters of radar X-direction, 1.93 meters of Y direction, and the result calibrated out is entered into line
Property interpolation.
It can also be used in inner parameter calibration in the thinking for the external parameter calibration talked about in upper one, that is, an essence
True inner parameter calibration (the horizontal and vertical angle and range correction value that are used to calculate every Shu Jiguang) must be than not smart
True calibration draws lower energy value.Therefore in some embodiments to be introduced of this section, in addition to step S104, solve
So that the internal calibration parameter that energy equation value is minimum.
It is emphasized that this characteristic is the direct result that vehicle collects data band in traveling.It is quiet for one
Vehicle only, it is the calibration that impossible eliminate some ambiguities.In fact, many angle and distances are corrected in stationary vehicle all
It is seemingly correct, i.e., local solution is not unique.But when vehicle travels along known trajectory, accumulated with every Shu Jiguang return value
Gather and be mapped to 3d space up, inaccurate calibration can not re-form accurately puts cloud seemingly.
Although internal calibration is equally applicable to for the energy equation for calculating the calibration of laser radar external parameter, to 64
Shu Jiguang determines that the calculating of three parameters is excessively complicated using grid data service simultaneously.As replacement, we consider institute
Some level angles, all vertical angles, and all range corrections, until function convergence.In each step, join for solving
Number, we seek derivative of the energy equation at single parameter.Rethink energy equation：
When each iteration, for every beam of laser biAnd its neighbouring laser collection bj, we will be with bjRelevant accumulation
The point cloud and corresponding normal vector mapped out all remains unchanged, and then will solve parameter and increases or decreases certain intervals amount α, then weighs
It is new to calculate bjThe point mapped out.To then increasing or reducing both possibilities, the internal structure ∑ of energy equationkwk||
ηk·(pk-mk)||2It will recalculate, then we select the α in the direction of energy largest optimization solution, or if two sides
To selection result it is all even worse, remain unchanged.
With this method, our excessively all parameters of loop iteration and laser beam, all optimization aim equation in every step, directly
Reach to the iterations determined in advance or the change of whole equation result is very small.It is noted that although
This way of search is run fine in practice, unlike the grid type search used in external calibration, searching method here
The result of target equation can not be ensured can reduce in any iteration provided, because in every step iteration, this search
Method can be updated to some special parameter of all laser beams.However, it is contemplated that the space of search is extremely big, so
Approximation method be also reasonable, and finally, our method is run fine in practice.To sum up, by entering
Row inner parameter is calibrated, and improves the parameter degree of accuracy of every light beam, is preferably solved laser radar and is calibrated itself ginseng automatically
The problem of number.
4th, calibration is received
In addition to estimating the parameter of relative pose extremely laser beam of laser radar, we have also been devised a set of Bayes's life
Into model, every Shu Jiguang is calculated by maximum expected value method to there is the reception of the body surface of different reflectivity reaction.
In some embodiments shown in Fig. 1, in addition to step S106 receives calibration, calculates calibration result.
In the particular embodiment, with the continuous change of vehicle pose, T is made as observation { z1,…,znIntersection, wherein
ziFor four-tuple ＜ bi,ri,ai,ci＞, numbering, range measurement, ionization meter and the map net of the laser beam of observation have been corresponded to respectively
Lattice are numbered.Map can include 2D grids, that is, will project on floor a little；Or it can directly use 3D
Pattern, cover all information of data.The intensity observed is that the calibration result c of a laser beam j determination (j, a) may be used
To be calculated by following formula：
Intensity a map net is observed that is, the calibration result when laser beam j observes intensity a is laser beam j
The conditional expectation for the intensity that all laser beams of other in lattice observe.
On this basis further, we can amplify out a probability calibrating, including step S108, to every Shu Ji
The noise properties and uncertainty of light beam establish Bayesian model, calculate in the case where given map grid is distributed, per Shu Ji
The distribution of luminous intensity return value.Noise properties and uncertainties model with probability calibrating to every Shu Jiguang, this is in actual conditions
In be generally difficult to.It is considered that although the reflectivity of environment is successional, in order to calculate simplicity, we are by the value of emissivity
Scope is limited to the integer between 0 to 255, and this is also consistent with the range of readings of Velodyne laser radars.Thus for each
Grid map ci, we are distributed with P (m) to represent grid map ciIntensity be m probability, wherein m=0 to 255.
Now, for every beam of laser bi, we are intended to estimate distribution P (ai| m), this distribution is represented in grid
Map ciIntensity when be m, laser beam biIntensity a can be returnediProbability.The unified priori of each grid map
Intensity is as initial value, and for every Shu Jiguang, prior probability initial value is set to by we：
Wherein, η is normalisation coefft, and τ controls the kurtosis of distribution, and ε is the nonzero probability of a return random strength.With
This initial setting up, priori light beam are likely to return the numerical value at the bright place of map, as shown in Figure 4.
Since the above-mentioned initial setting up provided, we alternately to calculate each grid map P (m) (E-step) and
Every Shu Jiguang and map intensity P (ai|m)(M-step).It is understood that although each the intensity of map grid is howsoever
It is independent, but they are conditional sampling each other under given laser beam model, and this allows for us to be counted using EM
Calculate.Renewal equation is as follows.
E-step：
So as to which in E-step, we are under given present laser parameter, calculate intensity distribution.
M-step：
So as to which in M-step, we are the situations of the existing intensity distribution in known observation data and each grid map
Under, calculate most possible laser parameter.First, the intensity return value arrived in some known grid map per beam laser observations with
And in the case of the grid map intensity Distribution value, we can calculate the grid map to be possible to intensity with E-step
Probability.Then, with bayes rule, we can be calculated in the case of given map grid intensity distribution, per Shu Jiguang
Possible intensity return value distribution.
After EM convergences, we just have a complete raw forming model, can with estimation often Shu Jiguang to not
With the return value of the environmental surfaces of reflectivity.Preferably solve the problems, such as the calibration of laser radar inherent parameters, be advantageous to laser
Radar quickly and accurately perceives environment.
5th, result of the test
Reflect after the installation position difference of laser radar, also used originally in the embodiment described in Fig. 5 and Fig. 6
Quasi- setting is proofreaded outside, and radar accumulates the effect that the point cloud to be formed shows deformation, and body surface is also very fuzzy (Fig. 5).Shape therewith
That into sharp contrast is Fig. 6 after recalibration, and point cloud preferably flocks together.
In addition, our internal reference calibration is also very accurate.In the embodiment shown in fig. 7, illustrating our algorithm can incite somebody to action
Laser radar is progressively adjusted from maximum error to accurate model, has only used the data of 10 seconds.Wherein, usage vector mark color is put
.It is the point cloud under maximum error pattern to scheme a, and figure b is the point cloud after 40 iteration optimizations, and figure c is by 80 iteration
Optimization, figure d are by 300 iteration optimizations.
Moreover, our internal reference calibration effect will also be much better than the internal reference calibration algorithm result provided of producer.Fig. 8
It is shown that with producer calibration 64 beam laser level angle adjustment, since the same initial setting up, by 400 times I
Interative computation result.
In the embodiment shown in fig. 9, depict it is known per beam laser intensity return value in the case of, it is contemplated that environment
Intensity.All 64 beam laser all reflect on the diagram, it can be seen that have marked difference between them.
In embodiment shown in Figure 10, all calibration methods listed herein have been arrived in our integrated uses.In picture
Reflection is that strength map is just penetrated in street, the initial angle and range error of laser radar be all very big, is provided with us
After the calibration method of inside and outside, middle more clearly picture can be obtained, can be greatly improved after being eventually adding intensity calibration
Right figure afterwards.
As shown in figure 11, it is a kind of laser radar calibrating installation module map, wherein device includes collection module 1100, outside
Calibration module 1102；
The collection module 1100 is used to collect radar reflection data when radar moves with car；
The external calibration module 1102 is used to the data being collected into substituting into dot matrix cloud energy equation, solves and causes energy
Equation value is minimum external calibration parameter；
Wherein dot matrix cloud energy equation is：
Wherein J is the energy of a cloud, B laser beams sum, and N is closes on light beam number, and k is the reflected beams number, pkIt is k-th point
Position under bodywork reference frame, mkIt is light beam biClosest to p in receiving the institute of reflection a littlekPoint or point set, ηkIt is mkPoint
Normal vector, wk| | pk-mk| | ＜ dmaxIt is 1 during establishment, is otherwise 0.Designed by above-mentioned module, reached quick, accurate
Ground independently provides the effect of positional information so that mobile lidar being capable of autonomous positioning.
In some further embodiments, in addition to internal calibration module 1104, the internal calibration module are used to ask
Solution causes the minimum internal calibration parameter of energy equation value.Inner parameter calibration is carried out by designing internal calibration module 1104,
The parameter degree of accuracy of every light beam is improved, preferably solves the problems, such as that laser radar calibrates inherent parameters automatically.
Specifically, the inner parameter is included per beam laser level and vertical angle and range correction value.
Specifically, the external calibration parameter include radar with respect to the plane coordinates parameter of car body, height, the angle of pitch, turn over
Roll angle or course angle.
In some further embodiments, in addition to calibration module 1106 is received, the reception calibration module is used for will
Radar observation data substitute into following formula：
Wherein j numbers for laser beam, and a is observed strength, and c is calibration result.Above-mentioned module design solves radar observation
The problem of parametric calibration.
Further, in addition to bass leaf module 1108, the bass leaf module are used for special to the noise of every beam laser beam
Property and uncertainty establish Bayesian model, calculate in the case where given map grid is distributed, per beam laser intensity return value
Distribution.Preferably solve the problems, such as the calibration of laser radar inherent parameters, be advantageous to laser radar and quickly and accurately perceive ring
Border.
Prior art is different from, said apparatus module can be analyzed the data that laser radar is passed back in real time, actively
Calibration of laser radar inherent parameters, quick calibration of laser radar is reached, has improved scan data accuracy rate, improved scan image and distinguish
The effect of knowledge and magnanimity.
It should be noted that herein, such as first and second or the like relational terms are used merely to a reality
Body or operation make a distinction with another entity or operation, and not necessarily require or imply and deposited between these entities or operation
In any this actual relation or order.Moreover, term " comprising ", "comprising" or its any other variant are intended to
Nonexcludability includes, so that process, method, article or terminal device including a series of elements not only include those
Key element, but also the other element including being not expressly set out, or it is this process, method, article or end also to include
The intrinsic key element of end equipment.In the absence of more restrictions, limited by sentence " including ... " or " including ... "
Key element, it is not excluded that other key element in the process including the key element, method, article or terminal device also be present.This
Outside, herein, " being more than ", " being less than ", " exceeding " etc. are interpreted as not including this number；" more than ", " following ", " within " etc. understand
It is to include this number.
It should be understood by those skilled in the art that, the various embodiments described above can be provided as method, apparatus or computer program production
Product.These embodiments can use the embodiment in terms of complete hardware embodiment, complete software embodiment or combination software and hardware
Form.All or part of step in the method that the various embodiments described above are related to can by program come instruct the hardware of correlation come
Complete, described program can be stored in the storage medium that computer equipment can be read, for performing the various embodiments described above side
All or part of step described in method.The computer equipment, include but is not limited to：Personal computer, server, general-purpose computations
It is machine, special-purpose computer, the network equipment, embedded device, programmable device, intelligent mobile terminal, intelligent home device, wearable
Smart machine, vehicle intelligent equipment etc.；Described storage medium, include but is not limited to：RAM, ROM, magnetic disc, tape, CD, sudden strain of a muscle
Deposit, USB flash disk, mobile hard disk, storage card, memory stick, webserver storage, network cloud storage etc..
The various embodiments described above are with reference to method, equipment (system) and the computer program product according to embodiment
Flow chart and/or block diagram describe.It should be understood that can be by every in computer program instructions implementation process figure and/or block diagram
One flow and/or the flow in square frame and flow chart and/or block diagram and/or the combination of square frame.These computers can be provided
Programmed instruction is to the processor of computer equipment to produce a machine so that passes through the finger of the computing device of computer equipment
Order, which produces, to be used to realize what is specified in one flow of flow chart or multiple flows and/or one square frame of block diagram or multiple square frames
The device of function.
These computer program instructions may be alternatively stored in the computer that computer equipment can be guided to work in a specific way and set
In standby readable memory so that the instruction being stored in the computer equipment readable memory produces the manufacture for including command device
Product, the command device is realized to be referred in one flow of flow chart or multiple flows and/or one square frame of block diagram or multiple square frames
Fixed function.
These computer program instructions can be also loaded on computer equipment so that perform a system on a computing device
Row operating procedure is being flowed with producing computer implemented processing so as to which the instruction performed on a computing device provides for realizing
The step of function of being specified in one flow of journey figure or multiple flows and/or one square frame of block diagram or multiple square frames.
Although the various embodiments described above are described, those skilled in the art once know basic wound
The property made concept, then other change and modification can be made to these embodiments, so embodiments of the invention are the foregoing is only,
Not thereby the scope of patent protection of the present invention, every equivalent structure made using description of the invention and accompanying drawing content are limited
Or equivalent flow conversion, or other related technical areas are directly or indirectly used in, similarly it is included in the patent of the present invention
Within protection domain.
Claims (8)
1. a kind of laser radar calibration method, it is characterised in that comprise the following steps：External parameter is calibrated；
The external parameter calibration steps includes：Radar is moved with car, collects radar reflection data, and the data being collected into are substituted into
Dot matrix cloud energy equation, solve and cause energy equation value for minimum external calibration parameter；
Wherein dot matrix cloud energy equation is：
<mrow>
<mi>J</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<msub>
<mi>b</mi>
<mi>i</mi>
</msub>
<mi>B</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>b</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<msub>
<mi>b</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mi>N</mi>
</mrow>
<mrow>
<msub>
<mi>b</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mi>N</mi>
</mrow>
</munderover>
<munder>
<mo>&Sigma;</mo>
<mi>k</mi>
</munder>
<msub>
<mi>w</mi>
<mi>k</mi>
</msub>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>&eta;</mi>
<mi>k</mi>
</msub>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>p</mi>
<mi>k</mi>
</msub>
<mo>-</mo>
<msub>
<mi>m</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
</mrow>
Wherein J is the energy of a cloud, and B laser beams are total, biFor beam number, bjFor biNeighbouring laser collection, N is closes on light beam
Number, k is for light beam biReflect to be formed progressive alternate a little, pkIt is the k-th point of position under bodywork reference frame
Put, mkIt is light beam biClosest to p in receiving the institute of reflection a littlekPoint or point set, ηkIt is mkThe normal vector of point, WkIn ‖ pk-mk‖
＜ dmaxIt is 1 during establishment, is otherwise 0.
2. laser radar calibration method according to claim 1, it is characterised in that also calibrated including step, inner parameter,
The inner parameter calibration steps includes：Solve the internal calibration parameter for make it that energy equation value is minimum.
3. laser radar calibration method according to claim 2, it is characterised in that the inner parameter is included per Shu Jiguang
Horizontal and vertical angle and range correction value.
4. laser radar calibration method according to claim 1, it is characterised in that the external calibration parameter includes radar
With respect to plane coordinates parameter, height, the angle of pitch, roll angle or the course angle of car body.
5. a kind of laser radar calibrating installation, it is characterised in that including collection module, external calibration module；
The collection module is used to collect radar reflection data when radar moves with car；
The external calibration module is used to the data being collected into substituting into dot matrix cloud energy equation, solves and make it that energy equation value is
Minimum external calibration parameter；
Wherein dot matrix cloud energy equation is：
<mrow>
<mi>J</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<msub>
<mi>b</mi>
<mi>i</mi>
</msub>
<mi>B</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msub>
<mi>b</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<msub>
<mi>b</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mi>N</mi>
</mrow>
<mrow>
<msub>
<mi>b</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<mi>N</mi>
</mrow>
</munderover>
<munder>
<mo>&Sigma;</mo>
<mi>k</mi>
</munder>
<msub>
<mi>w</mi>
<mi>k</mi>
</msub>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>&eta;</mi>
<mi>k</mi>
</msub>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>p</mi>
<mi>k</mi>
</msub>
<mo>-</mo>
<msub>
<mi>m</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
</mrow>
Wherein J is the energy of a cloud, and B laser beams are total, biFor beam number, bjFor biNeighbouring laser collection, N is closes on light beam
Number, k is for light beam biReflect to be formed progressive alternate a little, pkIt is the k-th point of position under bodywork reference frame
Put, mkIt is light beam biClosest to p in receiving the institute of reflection a littlekPoint or point set, ηkIt is mkThe normal vector of point, wkIn ‖ pk-mk‖
＜ dmaxIt is 1 during establishment, is otherwise 0.
6. laser radar calibrating installation according to claim 5, it is characterised in that described also including internal calibration module
Internal calibration module is used to solve the internal calibration parameter for make it that energy equation value is minimum.
7. laser radar calibrating installation according to claim 6, it is characterised in that the inner parameter is included per Shu Jiguang
Horizontal and vertical angle and range correction value.
8. laser radar calibrating installation according to claim 5, it is characterised in that the external calibration parameter includes radar
With respect to plane coordinates parameter, height, the angle of pitch, roll angle or the course angle of car body.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610107997.XA CN105866762B (en) | 2016-02-26 | 2016-02-26 | Laser radar automatic calibrating method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610107997.XA CN105866762B (en) | 2016-02-26 | 2016-02-26 | Laser radar automatic calibrating method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105866762A CN105866762A (en) | 2016-08-17 |
| CN105866762B true CN105866762B (en) | 2018-02-23 |
Family
ID=56625488
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610107997.XA Active CN105866762B (en) | 2016-02-26 | 2016-02-26 | Laser radar automatic calibrating method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105866762B (en) |
Families Citing this family (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106405555B (en) * | 2016-09-23 | 2019-01-01 | 百度在线网络技术（北京）有限公司 | Obstacle detection method and device for Vehicular radar system |
| CN106524906B (en) * | 2016-10-13 | 2019-03-26 | 深圳市速腾聚创科技有限公司 | Laser scanner method for homing and laser scanner |
| US10591584B2 (en) * | 2016-10-25 | 2020-03-17 | GM Global Technology Operations LLC | Radar calibration with known global positioning of static objects |
| CN107153186A (en) * | 2017-01-06 | 2017-09-12 | 深圳市速腾聚创科技有限公司 | Laser radar scaling method and laser radar |
| US10509120B2 (en) * | 2017-02-16 | 2019-12-17 | GM Global Technology Operations LLC | Lidar-radar relative pose calibration |
| CN107167788B (en) * | 2017-03-21 | 2020-01-21 | 深圳市速腾聚创科技有限公司 | Method and system for obtaining laser radar calibration parameters and laser radar calibration |
| CN107388967B (en) * | 2017-08-14 | 2019-11-12 | 上海汽车集团股份有限公司 | A kind of outer parameter compensation method of vehicle-mounted three-dimensional laser sensor and device |
| CN108875114B (en) * | 2017-09-07 | 2021-06-04 | 湖南大学 | Method for identifying characteristic parameters of focused laser beam |
| CN109521403B (en) * | 2017-09-19 | 2020-11-20 | 百度在线网络技术（北京）有限公司 | Parameter calibration method, device and equipment of multi-line laser radar and readable medium |
| US10473772B2 (en) | 2017-10-12 | 2019-11-12 | Ford Global Technologies, Llc | Vehicle sensor operation |
| CN110308436B (en) * | 2018-03-27 | 2023-07-21 | 深圳砺剑天眼科技有限公司 | Laser optical axis calibration method and system for multi-line laser scanner |
| SG11201811410RA (en) * | 2018-06-25 | 2020-01-30 | Beijing Didi Infinity Technology & Development Co Ltd | Integrated sensor calibration in natural scenes |
| CN109084738B (en) * | 2018-07-06 | 2021-09-17 | 上海宾通智能科技有限公司 | Height-adjustable calibration system and calibration method |
| CN109375232B (en) * | 2018-10-17 | 2021-01-26 | 维沃移动通信（杭州）有限公司 | Distance measuring method and device |
| CN111114463B (en) * | 2018-10-30 | 2023-01-17 | 百度在线网络技术（北京）有限公司 | Method and device for acquiring blind area noise |
| CN110609268B (en) * | 2018-11-01 | 2022-04-29 | 驭势科技（北京）有限公司 | Laser radar calibration method, device and system and storage medium |
| CN111157971B (en) * | 2018-11-07 | 2024-01-19 | 湖南中车时代电动汽车股份有限公司 | Laser radar control method and device |
| JP2022510198A (en) * | 2018-11-29 | 2022-01-26 | エスゼット ディージェイアイ テクノロジー カンパニー リミテッド | A system for detecting the environment around a mobile platform and its method |
| CN109725303B (en) * | 2018-12-04 | 2021-07-02 | 北京万集科技股份有限公司 | Coordinate system correction method and device, and storage medium |
| CN111381224B (en) * | 2018-12-29 | 2022-02-22 | 深圳市优必选科技有限公司 | Laser data calibration method and device and mobile terminal |
| CN111458715A (en) * | 2019-01-22 | 2020-07-28 | 杭州智屹科技有限公司 | Distance measuring and calibrating method |
| CN109917355A (en) * | 2019-03-04 | 2019-06-21 | 合肥嘉东光学股份有限公司 | Laser radar range error compensation system |
| CN110058205B (en) * | 2019-05-24 | 2023-03-14 | 中国人民解放军陆军炮兵防空兵学院郑州校区 | Warning radar system error correction method based on iterative closest point algorithm |
| CN113156411B (en) * | 2021-05-03 | 2022-05-20 | 湖北汽车工业学院 | Vehicle-mounted laser radar calibration method |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001124854A (en) * | 1999-10-25 | 2001-05-11 | Hitachi Ltd | Laser range finder and its calibration mechanism |
| GB201116959D0 (en) * | 2011-09-30 | 2011-11-16 | Bae Systems Plc | Vehicle localisation with 2d laser scanner and 3d prior scans |
| CN103870845B (en) * | 2014-04-08 | 2017-02-15 | 重庆理工大学 | Novel K value optimization method in point cloud clustering denoising process |
| CN104408443B (en) * | 2014-12-15 | 2017-07-18 | 长春理工大学 | The method for recognizing road surface types and device based on laser radar of multisensor auxiliary |
| CN105301577B (en) * | 2015-10-10 | 2016-07-06 | 福州华鹰重工机械有限公司 | A kind of laser intensity calibration steps and device |
-
2016
- 2016-02-26 CN CN201610107997.XA patent/CN105866762B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN105866762A (en) | 2016-08-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105866762B (en) | Laser radar automatic calibrating method and device | |
| CN105160702B (en) | The stereopsis dense Stereo Matching method and system aided in based on LiDAR point cloud | |
| EP3631494B1 (en) | Integrated sensor calibration in natural scenes | |
| CN111046776B (en) | Method for detecting obstacle of path of mobile robot based on depth camera | |
| CN101509781B (en) | Walking robot positioning system based on monocular cam | |
| CN107632308B (en) | Method for detecting contour of obstacle in front of vehicle based on recursive superposition algorithm | |
| CN107850450A (en) | Method and system for generating and using locating reference datum | |
| CN106526593B (en) | Sub-pixel-level corner reflector automatic positioning method based on the tight imaging model of SAR | |
| US20190065824A1 (en) | Spatial data analysis | |
| CN112464812B (en) | Vehicle-based concave obstacle detection method | |
| Kellner et al. | Road curb detection based on different elevation mapping techniques | |
| CN114119920A (en) | Three-dimensional point cloud map construction method and system | |
| CN116109601A (en) | Real-time target detection method based on three-dimensional laser radar point cloud | |
| CN110989619B (en) | Method, apparatus, device and storage medium for locating objects | |
| CN114994635A (en) | Intelligent driving travelable area detection method and device | |
| CN114137562B (en) | Multi-target tracking method based on improved global nearest neighbor | |
| CN105469391A (en) | Cloud shadow detection method and cloud shadow detection system | |
| Kohara et al. | Obstacle detection based on occupancy grid maps using stereovision system | |
| Steinke et al. | Groundgrid: Lidar point cloud ground segmentation and terrain estimation | |
| CN110231035A (en) | Climb mobile robot path guide method | |
| Pfeiffer et al. | Ground truth evaluation of the Stixel representation using laser scanners | |
| Gong et al. | Active Suspension Control Based on Estimated Road Class for Off‐Road Vehicle | |
| Mandici et al. | Aggregate road surface based environment representation using digital elevation maps | |
| US20220404506A1 (en) | Online validation of lidar-to-lidar alignment and lidar-to-vehicle alignment | |
| CN112305558B (en) | Mobile robot track determination method and device using laser point cloud data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |