CN110298891A - The method and device that Camera extrinsic precision is assessed automatically - Google Patents
The method and device that Camera extrinsic precision is assessed automatically Download PDFInfo
- Publication number
- CN110298891A CN110298891A CN201910557081.8A CN201910557081A CN110298891A CN 110298891 A CN110298891 A CN 110298891A CN 201910557081 A CN201910557081 A CN 201910557081A CN 110298891 A CN110298891 A CN 110298891A
- Authority
- CN
- China
- Prior art keywords
- aperture
- verifying plate
- coordinate
- plate
- point cloud
- 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.)
- Pending
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention provides a kind of methods that Camera extrinsic precision is assessed automatically, comprising: obtains laser point cloud;The point cloud in the first ROI region is traversed, according to the depth difference and reflectivity difference of the first and second verifying plates, obtains the marginal point cloud of each aperture;The marginal point cloud of multiple apertures is fitted;Obtain the first nodal point coordinate of each aperture;Obtain image information;It is obtained under image coordinate system according to the gradient of pixel in the second ROI region using Hough transformation, the second center point coordinate of each aperture;Second center point coordinate of each aperture and third center point coordinate are compared, when the offset of the second center point coordinate of each aperture and third center point coordinate is not more than preset offset threshold value, determine that Camera extrinsic is available.Thereby, it is possible to the stated accuracies to Camera extrinsic accurately to be assessed and be determined, have good consistency and robustness, be able to satisfy the flow of task of volume production process, simple operation and high efficiency requirement.
Description
Technical field
The present invention relates to technical field of data processing more particularly to a kind of method that Camera extrinsic precision is assessed automatically and dresses
It sets.
Background technique
In the aware scheme of autonomous driving vehicle, laser radar and camera are two kinds of most important detecting sensors.Laser
Radar accuracy is high and performance is stablized, adapt to complex environment such as night etc., but ambience color information can not be provided etc..Camera can
The information such as color and texture abundant in environment are provided, can be used to carry out target detection and classification, but can not provide accurately
The information such as obstacle distance.So the aware scheme of most automatic Pilots is to use laser radar and camera fusion.
The data fusion of two sensors needs to carry out sensor point-device outer ginseng calibration, to guarantee laser radar and camera
Perception data accurately merged in the same coordinate reference system.
Currently, external calibration parameter (outer ginseng) calibration result appraisal procedure between laser radar and camera depends primarily on
The principle of multi-camera calibration algorithm itself, including the outer ginseng result verification under special scenes and nonspecific scene.
When assessing best calibration result under special scenes, generally pass through error projection point cloud and total projection point cloud ratio system
For number to verify, the ratio is smaller, illustrates that result is more accurate.Outcome evaluation, but the assessment can also be carried out by Feature Points Matching
The premise of method needs to guarantee that characteristic point does not have error hiding, and otherwise this method can fail.
Under nonspecific scene, when assessing best calibration result, mostly take between laser point cloud and image pixel
Relevance function is calculated, and the optimal solution of gradient decline is found after successive ignition, by the optimal solution after multiple calculate into
Row output.The method for wherein calculating point cloud and image pixel correlation mainly has the side such as most relevance information approach and edge detection
Method, but since cloud and pixel interdependence function are all that height is non-convex, so being readily available locally optimal solution, that is, make
It is not globally optimal solution after the verification of algorithm at the outer ginseng result finally obtained.
For being equipped with low line beam laser radar volume production automatic driving vehicle, in the outer ginseng mark for completing laser radar and camera
After fixed, it is insecure for only assessing by the calibrating parameters of calibration algorithm itself, is limited to calibration algorithm itself, not can guarantee output
The result is that globally optimal solution, or since the influences such as environmental factor lead to output error result.
The difficult point of the combined calibrating of laser radar and camera is that result consistency and algorithm robustness are lower, so certainly
In the dynamic application driven, for the perception safety of support vehicles, for calibrating parameters result accuracy assessment generally all by people
Eye judges, cloud is projected on image by the outer ginseng of calibration and camera internal reference, then human eye come judge cloud in image
The registration of real-world object.This assessment mode is influenced and inefficiency by subjective consciousness, wasting manpower and material resources.
After laser radar and camera combined calibrating, program can export calibration result, but whether result can be used, it is necessary to pass through
Cross assessment verification.Since assessment of the calibration algorithm to output result itself is unreliable, so in general, verifying by artificial
It completes, i.e., judges whether calibration result meets expection, but waste of manpower and inefficiency in this way by experienced engineer.
Summary of the invention
The purpose of the embodiment of the present invention is that a kind of method and device that Camera extrinsic precision is assessed automatically is provided, it is existing to solve
There is the problem of appraisal procedure wasting manpower and material resources in technology, inefficiency.
To solve the above problems, in a first aspect, the present invention provides a kind of method that Camera extrinsic precision is assessed automatically, institute
The method of stating includes:
Obtain laser point cloud;The laser point cloud is that the laser radar on vehicle is incident upon the first verifying plate and passes through
After multiple apertures on first verifying plate, the point cloud of the second verifying plate is projected;First verifying plate and described second
Verifying plate laid out in parallel, first verifying plate are set between vehicle and the second verifying plate;
According to the position of first verifying plate and size, described cloud is extracted in the first ROI of first verifying plate
Region;
Point every harness of cloud in first ROI region is orderly, according to consecutive points cloud in same harness described first
The depth difference and reflectivity difference of verifying plate and second verifying plate, to first ROI region in a manner of dynamic sliding window
Interior point cloud is traversed, and the marginal point cloud of each aperture is obtained;
The marginal point cloud of multiple apertures is fitted;
It when being fitted successfully, obtains under laser coordinate system, the first nodal point coordinate of each aperture;
Obtain image information;Described image information includes the first verifying plate of the camera shooting on vehicle and passes through institute
The image of the second verifying plate after stating multiple apertures on the first verifying plate;
Described image information is handled, the second ROI region of the first verifying plate in described image information is extracted;
The picture of aperture under image coordinate system is obtained using Hough transformation according to the gradient of pixel in second ROI region
Plain shape, to acquire the second center point coordinate of each aperture;
According to the internal reference parameter of the first nodal point coordinate, second center point coordinate and camera, by described first
Center point coordinate projects to image coordinate system, obtains under image coordinate system, the third center point coordinate of each aperture;
Second center point coordinate of each aperture and the third center point coordinate are compared, when each aperture
Second center point coordinate and the offset of third center point coordinate when being all not more than preset offset threshold value, really
The fixed Camera extrinsic is available.
In one possible implementation, the center of the coordinate origin of the laser radar and first verifying plate
In the same horizontal line.
In one possible implementation, the shape of the aperture is circle, and the quantity of the aperture is 4,4 institutes
The circle for stating same size is arranged on first verifying plate at equal intervals.
In one possible implementation, when the aperture is round, the shape of the marginal point cloud composition of each aperture
Shape is annulus, and the marginal point cloud to multiple apertures is fitted, and is specifically included:
According to stochastical sampling consistency algorithm, Spatial Sphere fitting is carried out to the marginal point cloud of four annulus.
In one possible implementation, the stochastical sampling consistency algorithm, to the marginal point of four annulus
Cloud carries out after Spatial Sphere fitting conjunction, further includes:
Judge ball quantity and relative position, if meet theory relation threshold value;
When being unsatisfactory for, next frame point cloud is traversed, the marginal point cloud of each aperture is obtained.
In one possible implementation, the gradient according to pixel in second ROI region, is become using Hough
It changes, obtains under image coordinate system, after the second center point coordinate of each aperture, further includes:
Judge whether round quantity and relative position meet theory relation threshold value;
When being unsatisfactory for, next frame image is extracted, obtains the second ROI region of next frame image.
Second aspect, the present invention provides a kind of device that Camera extrinsic precision is assessed automatically, described device includes:
Acquiring unit, the acquiring unit is for obtaining laser point cloud;The laser point cloud is the laser radar on vehicle
It is incident upon the first verifying plate and after multiple apertures on first verifying plate, projects the point cloud of the second verifying plate;
First verifying plate and the second verifying plate laid out in parallel, first verifying plate be set to vehicle and the second verifying plate it
Between;
Extraction unit, the extraction unit are used for position and size according to first verifying plate, extract described cloud
In the first ROI region of first verifying plate;
Computing unit, the point every harness of cloud of the computing unit in first ROI region is orderly, according to same
In harness consecutive points cloud first verifying plate and second verifying plate depth difference and reflectivity difference, dynamically to slide
Window mode traverses the point cloud in first ROI region, obtains the marginal point cloud of each aperture;
Fitting unit, the fitting unit is for being fitted the marginal point cloud of multiple apertures;
The computing unit is also used to, and when being fitted successfully, is obtained under laser coordinate system, the first nodal point of each aperture
Coordinate;
The acquiring unit is also used to obtain image information;Described image information includes first of the camera shooting on vehicle
The image of verifying plate and the second verifying plate after multiple apertures on first verifying plate;
The extraction unit is also used to, and is handled described image information, and the first verification in described image information is extracted
Second ROI region of plate;
The computing unit is also used to, and is obtained according to the gradient of pixel in second ROI region using Hough transformation
The primitive shape of aperture under image coordinate system, to acquire the second center point coordinate of each aperture;
The computing unit is also used to, according to the first nodal point coordinate, second center point coordinate and camera
Internal reference parameter obtains under image coordinate system by the first nodal point coordinate projection to image coordinate system, the third of each aperture
Center point coordinate;
Comparing unit, the comparing unit are used for second center point coordinate of each aperture and the third center
Point coordinate be compared, when each aperture second center point coordinate and the third center point coordinate offset not
When greater than preset offset threshold value, determine that the Camera extrinsic is available.
The third aspect, the present invention provides a kind of equipment, including memory and processor, the memory is for storing journey
Sequence, the processor are used to execute any method of first aspect.
Fourth aspect, the present invention provides a kind of computer program products comprising instruction, when the computer program produces
When product are run on computers, so that the computer executes the method as described in first aspect is any.
5th aspect, the present invention provides a kind of computer readable storage medium, on the computer readable storage medium
It is stored with computer program, the method as described in first aspect is any is realized when the computer program is executed by processor.
It, can be to camera by the method and device assessed automatically using Camera extrinsic precision provided in an embodiment of the present invention
The stated accuracy of outer ginseng is accurately assessed and is determined, has good consistency and robustness, has ensured ginseng outside camera
The safety of several precision and vehicle sensory perceptual system.In towards automatic driving vehicle volume production, which automates assessment side
Case replaces manual verification, is able to satisfy the flow of task of volume production process, the requirement such as simple operation and high efficiency.
Detailed description of the invention
Fig. 1 is the method flow schematic diagram that the head Camera extrinsic precision that the embodiment of the present invention one provides is assessed automatically;
Fig. 2 is the external condition schematic diagram that the embodiment of the present invention one provides;
Fig. 3 is the first verifying plate schematic diagram that the embodiment of the present invention one provides;
Fig. 4 is the schematic diagram that the laser radar that the embodiment of the present invention one provides projects the first verifying plate partial region;
Fig. 5 is the annular edge point cloud schematic diagram that the embodiment of the present invention one provides;
Fig. 6 is schematic diagram of the center of circle that provides of the embodiment of the present invention one under same referential;
Fig. 7 is the apparatus structure schematic diagram that Camera extrinsic precision provided by Embodiment 2 of the present invention is assessed automatically.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that for just
Part relevant to related invention is illustrated only in description, attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
In automatic driving vehicle, there are multiple sensors, such as laser radar, camera, ultrasonic radar etc., in volume production
Before automatic driving vehicle comes into operation, need to demarcate sensor, to guarantee the accuracy of sensor, so guarantee certainly
The dynamic safety for driving vehicle.But it is whether accurate for calibration result, it is also necessary to further verification, for example, in camera mark
Determine link, it is whether accurate in order to sentence Camera extrinsic calibration result, joint laser radar is needed, Camera extrinsic calibration result is carried out
Verification.
Wherein, Camera extrinsic includes: spin matrix and translation matrix.Wherein, spin matrix describes world coordinate system
Direction of the reference axis relative to camera coordinates axis, translation matrix describe under camera coordinates system, the position of space origins.Rotation
Matrix and translation matrix describe how that a point is transformed into camera coordinates system from world coordinate system jointly
Fig. 1 is the method flow schematic diagram that the head Camera extrinsic precision that the embodiment of the present invention one provides is assessed automatically.The party
The executing subject of method can be vehicle automatic driving vehicle control unit (Automated Vehicle Control Unit,
AVCU), AVCU is the processor of automatic driving vehicle, is equivalent to " brain " of vehicle.As shown in Figure 1, this method includes following
Step:
Step 101, laser point cloud is obtained;Laser point cloud be vehicle on laser radar be incident upon the first verifying plate and
After multiple apertures on the first verifying plate, the point cloud of the second verifying plate is projected;First verifying plate and the second verifying plate are simultaneously
Column arrangement, the first verifying plate are set between vehicle and the second verifying plate.
In one example, example and it is non-limiting, the shape of aperture is circle, and the quantity of aperture is 4,4 identical big
Small circle is arranged on the first verifying plate at equal intervals.The center of the coordinate origin of laser radar and the first verifying plate is same
On one horizontal line.The position of the dimensional parameters of the first verifying plate, aperture can according to actual needs, be arranged in those skilled in the art
And quantity, the application do not limit this.
Before executing this step, need to set external condition.Fig. 2 is the Camera extrinsic that the embodiment of the present invention one provides
Calibration result verifies schematic diagram.The external condition such as 2 of this method is executed, automatic driving vehicle to be verified is in validation region,
There are locating chucking tools in rear vehicle end and two sides to guarantee consistency that vehicle is put, immediately ahead of laser radar and camera sensor
There is the first verifying plate in fixed range, for the first verifying plate perpendicular to ground, the second verifying plate can be metope, first verifying plate
It is the important calibration tool of this programme.First verifying plate schematic diagram is as shown in Figure 3.First verifying plate is that there are four hollow out circles for band
Square plate, flatness and certain roughness with degree of precision, color are dark but non-black.
Laser point cloud projects on the first verifying plate, and since there are four hollow out circles on the first verifying plate, and first verifies
There is also a face walls in certain distance behind plate, therefore laser can project below on the position of four hollow outs circle across annulus
Wall on.
In the following, being that circle is illustrated the application for the quantity of aperture is 4 with the shape of aperture.
Step 102, according to the position of the first verifying plate and size, point cloud is extracted in the first region of interest of the first verifying plate
The region domain (region of interest, ROI).
Specifically, extracting a cloud in the first verification as shown in figure 4, Fig. 4 is position according to the first verifying plate and size
First ROI region of plate.Black dot is the point cloud on laser projection to the first verifying plate in figure, and grey dot passes through for laser
After annulus on first verifying plate, the point cloud on the second verifying plate is projected.
Step 103, point every harness of cloud in the first ROI region is orderly, according to consecutive points cloud in same harness described
The depth difference and reflectivity difference of first verifying plate and second verifying plate, to the first ROI region in a manner of dynamic sliding window
Interior point cloud is traversed, and the marginal point cloud of each aperture is obtained.
Specifically, with continued reference to Fig. 4, since the material of the first verifying plate and the second verifying plate is different, reflectivity
Difference, and the depth value for putting cloud is also different, according to this otherness, the order of the cloud that puts a spot itself can be to the first area ROI
A cloud harness is successively traversed each of in domain, compares the depth difference and reflectivity difference of two neighboring point, if the difference
Greater than some threshold value, then it is reasonable that, the two consecutive points are located at the edge of hollow out circle, have thus extracted four circles
The marginal point cloud at ring edge.The marginal point cloud of four annulus is as shown in black color dots in Fig. 5.
Step 104, the marginal point cloud of multiple apertures is fitted.
Step 105, it when being fitted successfully, obtains under laser coordinate system, the first nodal point coordinate of each aperture.
Specifically, the shape that the marginal point cloud of each aperture is constituted is annulus, according to stochastical sampling when aperture is round
Consistency algorithm carries out Spatial Sphere fitting to four annulus respectively, and circle of dotted line indicates four successful spheres of fitting in Fig. 5.Such as
Fruit is fitted successfully, then exports under laser coordinate system, the sphere centre coordinate of each centre of sphere, if it fails, then continuing to next frame point cloud
Continue same work, until the quantity of ball and relative position, meet preset theory relation threshold value.
Wherein, the sphere centre coordinate of each centre of sphere, i.e., the first nodal point coordinate of each aperture.Theory relation threshold value, can be with
Including the distance between round quantity and each diameter of a circle, any two circle, the side length and height of the first verifying plate.Referring to
Fig. 3, the height of the first verifying plate are h+w, and the side length of the first verifying plate is w.
Step 106, image information is obtained;Image information includes the first verifying plate of the camera shooting on vehicle and wears
The image of the second verifying plate after crossing multiple apertures on the first verifying plate.
Specifically, the first verifying plate color is dark color, subsequent second verifying plate, i.e. wall color are white, this makes
The round color with the first verifying plate itself of hollow out has high contrast on the image.
Step 107, image information is handled, extracts the second ROI region of the first verifying plate in image information.
Specifically, extracting the second ROI region locating for the first verifying plate on image according to the high contrast of color.
Step 108, according to the gradient of pixel in the second ROI region, using Hough transformation, aperture under image coordinate system is obtained
Primitive shape, to acquire the second center point coordinate of each aperture.
Specifically, using Hough transformation principle, extracting Hough circle respectively according to the pixel gradient in the 2nd ROI, being opened
The primitive shape in hole.
Then, the quantity of output Hough circle and relative position, according to the relative positional relationship in four centers of circle itself, to detection
As a result judged, it is when the quantity of Hough circle is 4, and the relative position of 4 Houghs circle meets theory relation threshold value, then defeated
Coordinate value of four round centers of circle under image coordinate system out, if it fails, then continuing to continue same work to next frame image
Make, until the satisfactory sufficient theory relation threshold value of the Hough that extracts.
Wherein, step 101- step 105 and step 106- step 108 execution sequence are unrelated with number order, can be simultaneously
What column executed.
Step 109, according to the internal reference parameter of first nodal point coordinate, the second center point coordinate and camera, by the first center
Point coordinate projection obtains under image coordinate system, the third center point coordinate of each aperture to image coordinate system.
Specifically, having respectively obtained the three-dimensional of four centers of circle under laser coordinate system by above step 101-108 and having sat
The two-dimensional coordinate value of mark and four centers of circle under image coordinate system.At this moment according to the outer ginseng of laser radar and camera combined calibrating
As a result with the internal reference parameter of camera itself, the center of circle three-dimensional system of coordinate under laser coordinate system is projected under image coordinate system, is thrown
The schematic diagram of movie queen is as shown in Figure 6.Two square vertex respectively indicate the coordinate position relationship in two groups of centers of circle in Fig. 6, reason
By upper, the coordinate in the center of circle of laser point cloud after projecting to image should on the image with four coordinate positions in the image center of circle point
It is not overlapped, but due to the error or error value of outer ginseng calibration result, leads to the positional relationship presence of two groups of central coordinate of circle after projection
Deviation, i.e., as shown in the offset in two groups of centers of circle in figure.The deviant in the corresponding center of circle is respectively such as the m1, m2, m3 in figure, m4 institute
Show, the position deviation where the center of circle at straight-line intersection is as shown in m0.
Wherein, camera internal reference includes the parameters such as focal length related with camera, distortion parameter.
Step 110, the second center point coordinate of each aperture and third center point coordinate are compared, when each aperture
The second center point coordinate and the offset of third center point coordinate when being all not more than preset offset threshold value, determine outside camera
Ginseng is available.
In one example, this method, can be respectively to Fig. 6 in order to verify the accuracy and availability of Camera extrinsic result
In five offsets calculated and judged, when offset exceeds respective threshold value respectively, it is believed that Camera extrinsic is unavailable;When
The value of offset respectively in respective threshold value when, it is believed that in error range, Camera extrinsic meets to be made the Camera extrinsic result
With requiring.
In another example, the position deviation at straight-line intersection where the center of circle can not also be judged, only to upper
Four offsets are stated to be judged.
Specifically, continuing to connect example, work as m0, m1, m2, m3, m4 are respectively no more than respective preset offset threshold value Δ
When m0, Δ m1, Δ m2, Δ m3, Δ m4, it can determine that Camera extrinsic is available.When any one offset is corresponding greater than its partially
When shifting amount threshold value, that is, it can determine that Camera extrinsic is unavailable.
It is understood that the method that the Camera extrinsic precision of the application is assessed automatically, can also apply in other scenes
In, for example, being equipped in other terminals of laser radar and camera, the application does not limit this.
The method assessed automatically by the Camera extrinsic precision that the application embodiment of the present invention one provides, can be to Camera extrinsic
Stated accuracy accurately assessed and determined, have good consistency and robustness.This method has ensured outside camera
The precision of parameter and the safety of vehicle sensory perceptual system.In towards automatic driving vehicle volume production, calibration result automation assessment
Scheme replaces manual verification, is able to satisfy the flow of task of volume production process, the requirement such as simple operation and high efficiency.
Fig. 7 is the apparatus structure schematic diagram that Camera extrinsic precision provided by Embodiment 2 of the present invention is assessed automatically, the device
In the method assessed automatically using Camera extrinsic precision in example 1.As shown in fig. 7, the Camera extrinsic precision is commented automatically
The device 700 estimated includes: acquiring unit 701, extraction unit 702, computing unit 703, fitting unit 704 and comparing unit 705.
Acquiring unit 701 is for obtaining laser point cloud;Laser point cloud is that the laser radar on vehicle is incident upon the first verification
Plate and after multiple apertures on the first verifying plate, projects the point cloud of the second verifying plate;First verifying plate and the second school
Plate laid out in parallel is tested, the first verifying plate is set between vehicle and the second verifying plate;
Extraction unit 702 is used for position and size according to the first verifying plate, extracts point cloud the first of the first verifying plate
ROI region;
Computing unit, the point every harness of cloud of computing unit in first ROI region is orderly, according to same harness
Interior consecutive points cloud first verifying plate and second verifying plate depth difference and reflectivity difference, with dynamic sliding window side
Formula traverses the point cloud in first ROI region, obtains the marginal point cloud of each aperture;
Fitting unit 704 is for being fitted the marginal point cloud of multiple apertures;
Computing unit 703 is also used to, and when being fitted successfully, is obtained under laser coordinate system, the first nodal point of each aperture
Coordinate;
Acquiring unit 701 is also used to obtain image information;Image information includes the first verification of the camera shooting on vehicle
The image of plate and the second verifying plate after multiple apertures on the first verifying plate;
Extraction unit 702 is also used to, and is handled image information, and second of the first verifying plate in image information is extracted
ROI region;
Computing unit 703 is also used to, and obtains image seat using Hough transformation according to the gradient of pixel in the second ROI region
Mark is the primitive shape of lower aperture, to acquire the second center point coordinate of each aperture;
Computing unit 703 is also used to, according to the internal reference parameter of first nodal point coordinate, the second center point coordinate and camera,
First nodal point coordinate projection to image coordinate system is obtained under image coordinate system, the third center point coordinate of each aperture;
Comparing unit 705 is used to for the second center point coordinate of each aperture and third center point coordinate being compared, when
When second center point coordinate of each aperture and the offset of third center point coordinate are not more than preset offset threshold value, really
It is available to determine Camera extrinsic.
The concrete function of each unit is similar to described in embodiment one, and details are not described herein again.
It, can be to Camera extrinsic by the device assessed automatically using Camera extrinsic precision provided by Embodiment 2 of the present invention
Stated accuracy accurately assessed and determined, have good consistency and robustness.The device has ensured outside camera
The precision of parameter and the safety of vehicle sensory perceptual system.In towards automatic driving vehicle volume production, calibration result automation assessment
Scheme replaces manual verification, is able to satisfy the flow of task of volume production process, the requirement such as simple operation and high efficiency.
The embodiment of the present invention three provides a kind of equipment, including memory and processor, and memory is deposited for storing program
Reservoir can be connect by bus with processor.Memory can be nonvolatile storage, such as hard disk drive and flash memory, storage
Software program and device driver are stored in device.Software program is able to carry out the above method provided in an embodiment of the present invention
Various functions;Device driver can be network and interface drive program.Processor is for executing software program, the software journey
Sequence is performed, the method that can be realized the offer of the embodiment of the present invention one.
The embodiment of the present invention four provides a kind of computer program product comprising instruction, when computer program product is being counted
When being run on calculation machine, so that computer executes the method that the embodiment of the present invention one provides.
The embodiment of the present invention five provides a kind of computer readable storage medium, is stored on computer readable storage medium
Computer program realizes the method that the embodiment of the present invention one provides when computer program is executed by processor.
Professional should further appreciate that, described in conjunction with the examples disclosed in the embodiments of the present disclosure
Unit and algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, hard in order to clearly demonstrate
The interchangeability of part and software generally describes each exemplary composition and step according to function in the above description.
These functions are implemented in hardware or software actually, the specific application and design constraint depending on technical solution.
Professional technician can use different methods to achieve the described function each specific application, but this realization
It should not be considered as beyond the scope of the present invention.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can be executed with hardware, processor
The combination of software module or the two is implemented.Software module can be placed in random access memory (RAM), memory, read-only memory
(ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technical field
In any other form of storage medium well known to interior.
Above specific embodiment has carried out further in detail the purpose of the present invention, technical scheme and beneficial effects
Illustrate, it should be understood that the above is only a specific embodiment of the invention, the protection model that is not intended to limit the present invention
It encloses, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should be included in the present invention
Protection scope within.
Claims (10)
1. a kind of method that Camera extrinsic precision is assessed automatically, which is characterized in that the described method includes:
Obtain laser point cloud;The laser point cloud is that the laser radar on vehicle is incident upon the first verifying plate and passes through described
After multiple apertures on first verifying plate, the point cloud of the second verifying plate is projected;First verifying plate and second verification
Plate laid out in parallel, first verifying plate are set between vehicle and the second verifying plate;
According to the position of first verifying plate and size, described cloud is extracted in the first ROI region of first verifying plate;
Point every harness of cloud in first ROI region is orderly, according to consecutive points cloud in same harness in first verification
The depth difference and reflectivity difference of plate and second verifying plate, in first ROI region in a manner of dynamic sliding window
Point cloud is traversed, and the marginal point cloud of each aperture is obtained;
The marginal point cloud of multiple apertures is fitted;
It when being fitted successfully, obtains under laser coordinate system, the first nodal point coordinate of each aperture;
Obtain image information;Described image information includes the first verifying plate of the camera shooting on vehicle and passes through described the
The image of the second verifying plate after multiple apertures on one verifying plate;
Described image information is handled, the second ROI region of the first verifying plate in described image information is extracted;
According to the gradient of pixel in second ROI region, using Hough transformation, the pixel shape of aperture under image coordinate system is obtained
Shape, to acquire the second center point coordinate of each aperture;
According to the internal reference parameter of the first nodal point coordinate, second center point coordinate and camera, by first center
Point coordinate projection obtains under image coordinate system, the third center point coordinate of each aperture to image coordinate system;
Second center point coordinate of each aperture and the third center point coordinate are compared, when the institute of each aperture
When stating the offset of the second center point coordinate and the third center point coordinate and being all not more than preset offset threshold value, institute is determined
It is available to state Camera extrinsic.
2. the method according to claim 1, wherein the coordinate origin of the laser radar and first school
Test the center of plate in the same horizontal line.
3. the method according to claim 1, wherein the shape of the aperture is circle, the quantity of the aperture
It is 4, the circle of 4 same sizes is arranged on first verifying plate at equal intervals.
4. according to the method described in claim 3, it is characterized in that, when the aperture is round, the marginal point of each aperture
The shape that cloud is constituted is annulus, and the marginal point cloud to multiple apertures is fitted, specifically includes:
According to stochastical sampling consistency algorithm, Spatial Sphere fitting is carried out to the marginal point cloud of four annulus.
5. according to the method described in claim 3, it is characterized in that, the stochastical sampling consistency algorithm, to four circles
The marginal point cloud of ring carries out after Spatial Sphere fitting conjunction, further includes:
Judge ball quantity and relative position, if meet theory relation threshold value;
When being unsatisfactory for, next frame point cloud is traversed, the marginal point cloud of each aperture is obtained.
6. the method according to claim 1, wherein the gradient according to pixel in second ROI region,
It using Hough transformation, obtains under image coordinate system, after the second center point coordinate of each aperture, further includes:
Judge whether round quantity and relative position meet theory relation threshold value;
When being unsatisfactory for, next frame image is extracted, obtains the second ROI region of next frame image.
7. a kind of device that Camera extrinsic precision is assessed automatically, which is characterized in that described device includes:
Acquiring unit, the acquiring unit is for obtaining laser point cloud;The laser point cloud is the laser radar projection on vehicle
In the first verifying plate and after multiple apertures on first verifying plate, the point cloud of the second verifying plate is projected;It is described
First verifying plate and the second verifying plate laid out in parallel, first verifying plate are set between vehicle and the second verifying plate;
Extraction unit, the extraction unit are used for position and size according to first verifying plate, extract the point Yun Suo
State the first ROI region of the first verifying plate;
Computing unit, the point every harness of cloud of the computing unit in first ROI region is orderly, according to same harness
Interior consecutive points cloud first verifying plate and second verifying plate depth difference and reflectivity difference, with dynamic sliding window side
Formula traverses the point cloud in first ROI region, obtains the marginal point cloud of each aperture;
Fitting unit, the fitting unit is for being fitted the marginal point cloud of multiple apertures;
The computing unit is also used to, and when being fitted successfully, is obtained under laser coordinate system, and the first nodal point of each aperture is sat
Mark;
The acquiring unit is also used to obtain image information;Described image information includes the first verification of the camera shooting on vehicle
The image of plate and the second verifying plate after multiple apertures on first verifying plate;
The extraction unit is also used to, and is handled described image information, and the first verifying plate in described image information is extracted
Second ROI region;
The computing unit is also used to, and obtains image using Hough transformation according to the gradient of pixel in second ROI region
Under coordinate system, the second center point coordinate of each aperture;
The computing unit is also used to, according to the first nodal point coordinate, the internal reference of second center point coordinate and camera
Parameter obtains under image coordinate system by the first nodal point coordinate projection to image coordinate system, the third center of each aperture
Point coordinate;
Comparing unit, the comparing unit are used to sit second center point coordinate of each aperture and the third central point
Mark is compared, when second center point coordinate of each aperture and the offset of the third center point coordinate are not more than
When preset offset threshold value, determine that the Camera extrinsic is available.
8. a kind of equipment, which is characterized in that the equipment includes memory and processor, and the memory is used to store program,
The processor requires any method of 1-6 for perform claim.
9. a kind of computer program product comprising instruction, which is characterized in that when the computer program product on computers
When operation, so that the computer executes the method as described in claim 1-6 is any.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program realizes the method as described in claim 1-6 is any when the computer program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910557081.8A CN110298891A (en) | 2019-06-25 | 2019-06-25 | The method and device that Camera extrinsic precision is assessed automatically |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910557081.8A CN110298891A (en) | 2019-06-25 | 2019-06-25 | The method and device that Camera extrinsic precision is assessed automatically |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110298891A true CN110298891A (en) | 2019-10-01 |
Family
ID=68028861
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910557081.8A Pending CN110298891A (en) | 2019-06-25 | 2019-06-25 | The method and device that Camera extrinsic precision is assessed automatically |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110298891A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111784783A (en) * | 2020-08-14 | 2020-10-16 | 支付宝(杭州)信息技术有限公司 | System and method for calibrating external parameters of camera |
CN111830667A (en) * | 2020-07-30 | 2020-10-27 | 广东博智林机器人有限公司 | Lens focusing device |
CN112612016A (en) * | 2020-12-07 | 2021-04-06 | 深兰人工智能(深圳)有限公司 | Sensor combined calibration method and device, electronic equipment and storage medium |
CN112991454A (en) * | 2019-12-18 | 2021-06-18 | 动态Ad有限责任公司 | Calibration and verification of camera to LiDAR |
CN113739712A (en) * | 2020-05-29 | 2021-12-03 | 同方威视技术股份有限公司 | Vehicle wheel base measuring method and device and vehicle wheel base measuring instrument |
WO2022193828A1 (en) * | 2021-03-17 | 2022-09-22 | 苏州一径科技有限公司 | Method, apparatus, and device for verifying precision of calibration angle, and storage medium |
CN115267746A (en) * | 2022-06-13 | 2022-11-01 | 广州文远知行科技有限公司 | Positioning method for laser radar point cloud projection error and related equipment |
CN115267746B (en) * | 2022-06-13 | 2024-06-28 | 广州文远知行科技有限公司 | Positioning method for laser radar point cloud projection errors and related equipment |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140063189A1 (en) * | 2012-08-28 | 2014-03-06 | Digital Signal Corporation | System and Method for Refining Coordinate-Based Three-Dimensional Images Obtained from a Three-Dimensional Measurement System |
CN103745452A (en) * | 2013-11-26 | 2014-04-23 | 理光软件研究所(北京)有限公司 | Camera external parameter assessment method and device, and camera external parameter calibration method and device |
CN109300162A (en) * | 2018-08-17 | 2019-02-01 | 浙江工业大学 | A kind of multi-line laser radar and camera combined calibrating method based on fining radar scanning marginal point |
CN109636837A (en) * | 2018-12-21 | 2019-04-16 | 浙江大学 | A kind of evaluation method of monocular camera and ginseng calibration accuracy outside millimetre-wave radar |
CN109919893A (en) * | 2019-03-20 | 2019-06-21 | 湖北亿咖通科技有限公司 | Point cloud modification method, device and readable storage medium storing program for executing |
-
2019
- 2019-06-25 CN CN201910557081.8A patent/CN110298891A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140063189A1 (en) * | 2012-08-28 | 2014-03-06 | Digital Signal Corporation | System and Method for Refining Coordinate-Based Three-Dimensional Images Obtained from a Three-Dimensional Measurement System |
CN103745452A (en) * | 2013-11-26 | 2014-04-23 | 理光软件研究所(北京)有限公司 | Camera external parameter assessment method and device, and camera external parameter calibration method and device |
CN109300162A (en) * | 2018-08-17 | 2019-02-01 | 浙江工业大学 | A kind of multi-line laser radar and camera combined calibrating method based on fining radar scanning marginal point |
CN109636837A (en) * | 2018-12-21 | 2019-04-16 | 浙江大学 | A kind of evaluation method of monocular camera and ginseng calibration accuracy outside millimetre-wave radar |
CN109919893A (en) * | 2019-03-20 | 2019-06-21 | 湖北亿咖通科技有限公司 | Point cloud modification method, device and readable storage medium storing program for executing |
Non-Patent Citations (1)
Title |
---|
韩正勇等: "一种针孔相机与三维激光雷达外参标定方法", 《传感器与微系统》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112991454A (en) * | 2019-12-18 | 2021-06-18 | 动态Ad有限责任公司 | Calibration and verification of camera to LiDAR |
US11940539B2 (en) | 2019-12-18 | 2024-03-26 | Motional Ad Llc | Camera-to-LiDAR calibration and validation |
CN113739712A (en) * | 2020-05-29 | 2021-12-03 | 同方威视技术股份有限公司 | Vehicle wheel base measuring method and device and vehicle wheel base measuring instrument |
CN113739712B (en) * | 2020-05-29 | 2022-06-07 | 同方威视技术股份有限公司 | Vehicle wheel base measuring method and device and vehicle wheel base measuring instrument |
CN111830667A (en) * | 2020-07-30 | 2020-10-27 | 广东博智林机器人有限公司 | Lens focusing device |
CN111784783A (en) * | 2020-08-14 | 2020-10-16 | 支付宝(杭州)信息技术有限公司 | System and method for calibrating external parameters of camera |
CN112612016A (en) * | 2020-12-07 | 2021-04-06 | 深兰人工智能(深圳)有限公司 | Sensor combined calibration method and device, electronic equipment and storage medium |
WO2022193828A1 (en) * | 2021-03-17 | 2022-09-22 | 苏州一径科技有限公司 | Method, apparatus, and device for verifying precision of calibration angle, and storage medium |
CN115267746A (en) * | 2022-06-13 | 2022-11-01 | 广州文远知行科技有限公司 | Positioning method for laser radar point cloud projection error and related equipment |
CN115267746B (en) * | 2022-06-13 | 2024-06-28 | 广州文远知行科技有限公司 | Positioning method for laser radar point cloud projection errors and related equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110298891A (en) | The method and device that Camera extrinsic precision is assessed automatically | |
CN102369413B (en) | A method for determining the relative position of a first and a second imaging device and devices therefore | |
JP4245963B2 (en) | Method and system for calibrating multiple cameras using a calibration object | |
JP2919284B2 (en) | Object recognition method | |
CN106340044B (en) | Join automatic calibration method and caliberating device outside video camera | |
CN109816704A (en) | The 3 D information obtaining method and device of object | |
AU2011362799B2 (en) | 3D streets | |
CN108475433A (en) | Method and system for determining RGBD camera postures on a large scale | |
CN103562934B (en) | Face location detection | |
CN106560835B (en) | A kind of guideboard recognition methods and device | |
CN113146073B (en) | Vision-based laser cutting method and device, electronic equipment and storage medium | |
CN107025663A (en) | It is used for clutter points-scoring system and method that 3D point cloud is matched in vision system | |
CN111080662A (en) | Lane line extraction method and device and computer equipment | |
JP2022528301A (en) | Calibration method, positioning method, equipment, electronic devices and storage media | |
CN111383285A (en) | Millimeter wave radar and camera sensor fusion calibration method and system | |
US10186051B2 (en) | Method and system for calibrating a velocimetry system | |
CN105894511B (en) | Demarcate target setting method, device and parking assistance system | |
Yan et al. | Joint camera intrinsic and lidar-camera extrinsic calibration | |
CN107680125A (en) | The system and method that three-dimensional alignment algorithm is automatically selected in vision system | |
CN107590444A (en) | Detection method, device and the storage medium of static-obstacle thing | |
CN113034605B (en) | Target object position determining method and device, electronic equipment and storage medium | |
CN107025647A (en) | Distorted image evidence collecting method and device | |
CN106067160B (en) | Large screen merges projecting method | |
WO2019012004A1 (en) | Method for determining a spatial uncertainty in images of an environmental area of a motor vehicle, driver assistance system as well as motor vehicle | |
GB2551970A (en) | Determining the position of an object in a scene |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191001 |
|
RJ01 | Rejection of invention patent application after publication |