CN106023303B - A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud - Google Patents
A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud Download PDFInfo
- Publication number
- CN106023303B CN106023303B CN201610298507.9A CN201610298507A CN106023303B CN 106023303 B CN106023303 B CN 106023303B CN 201610298507 A CN201610298507 A CN 201610298507A CN 106023303 B CN106023303 B CN 106023303B
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- cloud
- derivation
- frame
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Processing Or Creating Images (AREA)
Abstract
It Three-dimensional Gravity is improved based on profile validity lays foundations the method for the dense degree of cloud the invention discloses a kind of comprising have：1, its contour of object is extracted, corresponding effective coverage graphic sequence is generated；2, point cloud is calculated in x, y, the extension scale in z-axis；Each of 3 pairs of initial point clouds point is extended, and is obtained one and is derived from point cloud；4, it will derive under point Cloud transform to camera coordinates system, and in back projection to effective coverage figure, retain the point in effective coverage；5, the initial point of the derivation point of calculation processing is to the dot product of the vector and the normal vector of the point of the point, point of the retention point product value more than zero；6, it checks whether the dense degree for deriving from point cloud reaches demand, the derivation point cloud is regard as initial point cloud when being unsatisfactory for demand, again the operation after progress step 2 to meet demand.The present invention is not limited to specifically not be too dependent on the adjustment of parameter around image sequence is clapped, and can improve the dense degree of available point cloud in a relatively short period of time in the case of relatively low calculation amount.
Description
Technical field
The present invention relates to computer vision fields, and in particular to a kind of that a cloud can be effectively improved within the time faster
The method of the dense degree of cloud of being laid foundations based on profile validity raising Three-dimensional Gravity of dense degree.
Background technology
Computer vision is related to multiple subjects, is the inverse process of video camera imaging process, and research range is quite extensive, main
Including：Object detection and recognition, edge extracting, feature extraction and three-dimensional reconstruction etc..Three-dimensional reconstruction is also based on
The modeling technique of image, is just concerned at the beginning of birth, and this method only needs two frame adjacent images more accurate extensive
It appears again the three-dimensional relationship of matching characteristic point and camera in image.In this process, the quantity of matching characteristic point is directly determined
The quality for having determined the point cloud of three-dimensional reconstruction acquisition, so that it is determined that the quality of reconstruction model.
Common three-dimensional rebuilding method has three classes：(1) Stereo Vision.This method simulates human visual system to objective
The perceptive mode of three-dimension object is imaged the same scenery in different location using more than two cameras, further according to two frames
Disparity map between image, is converted to depth map, obtains the depth information of object.The geometrical model file that the method generates is logical
It is often smaller, it is easy to be used in virtual reality.But this method needs to overcome the problems, such as that object features are sparse, when texture is flat
When, there is large stretch of white space in the disparity map being calculated, the dense degree for putting cloud is very low.(2) motion structure method.To object
It is identical that body, which carries out the movement that the point around any position in bat, rigid objects occurs between two field pictures, by two frames
Extracted between image it is several to characteristic point and match, the transformation matrix that object moves can be calculated, according to change
It changes matrix and can determine that the position relationship between two cameras can restore characteristic point in world coordinate system by pinhole imaging system principle
In coordinate.The method develops comparative maturity, the movement of camera can be calculated in the case where camera internal reference is demarcated, to sparse point
Cloud, which carries out processing, to be obtained compared with dense point cloud, and recovers more accurately threedimensional model.But it requires adjacent two interframe
Matching characteristic point quantity is more, therefore in the negligible amounts of the flat region available point of feature.(3) side based on depth image
Method.The point cloud of the object under Current camera coordinate system, adjacent two frame can be generated with depth map by the RGB figures of every frame image
Two groups of point clouds that RGBD figures generate are matched, and calculate the transformation matrix of two frame cameras, so that it may be fused to generation with two groups of point clouds
Boundary's coordinate system.The point cloud that the method is calculated is more accurate, and the dense degree for putting cloud is higher.But it needs depth camera
Assistance, and it is very sensitive to the precision of depth map, rebuild in scene on a large scale, the precision of depth camera is always limited, and
The precision of depth camera will be directly linked the precision of reconstruction point cloud.
In above method, the calculation amount that stereo vision method needs is smaller, but obtained in image texture flat site
There are white spaces for disparity map, therefore the dense degree of point cloud being calculated is very low；Motion structure method has higher universality,
Wherein include derivation history of the sparse cloud to dense point cloud, but the dense degree of the dense point cloud obtained still depends on figure
The texture complexity degree of picture, for the image that texture is flat, the dense degree of point cloud of acquisition is also relatively low；Based on depth image
Method for reconstructing precision is higher, and does not require the texture complexity degree of image, but this method is to the quick of depth camera precision
Sense degree is higher, not applicable at present to be rebuild with a wide range of object dimensional.
In view of the above problems, needing a kind of new method at present, make it the dense degree for improving available point cloud, and make
The influence that the point cloud of acquisition overcomes texture flat to a certain extent.
Invention content
The present invention in view of the above shortcomings of the prior art, provides and a kind of improving Three-dimensional Gravity based on profile validity and lay foundations cloud
The method of dense degree is not limited to that specifically around image sequence is clapped, is not too dependent on the adjustment of parameter, can be relatively low
In the case of calculation amount, the dense degree of available point cloud is improved in a relatively short period of time, while can delete the mistake in origin cloud
Overdue cloud so that the influence that the point cloud of acquisition overcomes texture flat to a certain extent.
The method of the present invention is pre-processed to rebuilding image sequence, obtains one group of corresponding effective coverage figure；Then will
Initial point cloud is extended the cloud that is expanded；According to the transformation matrix obtained in the calculating process of initial point cloud, will extend a little
In cloud back projection to each frame effective coverage figure；Back projection position is fallen into the point deletion in the figure of effective coverage outside effective coverage,
Aforesaid operations are repeated to one group of image around bat, the point cloud after being expanded once also needs to the point cloud inside deleting, only at this time
Reservation external point cloud, the point being located at below body surface is deleted according to the normal vector of initial point cloud, that is, completes point in deleting
Work.Point cloud of the extension after primary is extended again as initial point cloud, and interior point is filtered out after back projection's deletion, is obtained more
The point cloud for thickening the soup close；Above-mentioned extended operation is repeated, when the quantity of existing cloud reaches dense enough, the extension of end point cloud,
The point cloud obtained at this time has very high dense degree, and do not depend on the adjustment with parameter, and the customer service flat shadow of texture
It rings.
To solve problems of the prior art, the specific technical solution that the present invention uses is：
A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud comprising following steps：
S1, one group is obtained by photographic equipment around image sequence is clapped, to every frame around bat image zooming-out contour of object, and will take turns
Pixel value in wide region is set as 255, and the pixel value outside profile is set as 0, obtains a frame bianry image, referred to as effective coverage
Figure；
S2, opposing connection clap image sequence and carry out three-dimensional reconstruction step, obtain the very low point cloud of a consistency, referred to as initial point
Cloud, while spin matrix R and translation vector t of each frame camera relative to world coordinate system are also obtained, spin matrix and translation
Vector combines to form transformation matrix M；
Each point in S3, traversal initial point cloud, obtaining all the points in initial point cloud, value is most on three axis of x, y, z
Big value and minimum value, and the distance between maxima and minima difference on each axis is calculated, it is denoted as x_dis, y_dis, z_ respectively
Dis, respectively by this three range difference divided by 100, three obtained amount, referred to as the derivation scale of initial point cloud are denoted as x_
scalar,y_scalar,z_scalar；
S4, using a point in initial point cloud as source point, respectively along x, the positive negative direction respectively extension pair in tri- directions y, z
The derivation scale size calculated in S3 is answered, a cuboid centered on source point is obtained, the length, width and height of the cuboid are respectively
2*x_scalar, 2*y_scalar, 2*z_scalar, the source point center toward extending 26 directions altogether around cuboid,
A new point is derived in each direction, takes the normal vector of the new point identical as the normal vector of source point, and each point that derives from is remembered
Record its source point；
S5, the derivation operation described in a step S4 is carried out to each point in initial point cloud, a group will be obtained
Raw point cloud, the quantity at this cloud midpoint are 26 times of initial point cloud quantity；
S6, opposing connection clap the i-th frame image in image sequence, take out its transformation matrix M being calculated in step s 2i,
By the derivation point cloud obtained in step S5 according to transformation matrix MiIt transforms under corresponding camera coordinates system, and according to projection theory
It will be on the effective coverage figure that the i-th frame obtained in Dian Yun back projections to step S1 be derived from；
S7, according to the step in S6, to projecting to the point in the inactive area in the i-th frame effective coverage figure, by it from group
It is deleted in raw point cloud, the point in the effective coverage projected in the i-th frame effective coverage figure then retains；
S8, opposing connection clap the operation that each frame in image sequence is performed both by above-mentioned steps S6 and S7, by deriving from point cloud
Around projecting and deleting, three-dimensional reconstruction obtains the derivation point cloud containing interior point；
Each of derivation point cloud that S9, traversal step S8 are obtained derives from point, judge it is each derive from point be interior point or
Exterior point, delete be interior point derivation point cloud, reservation is the derivation point cloud of exterior point；The point cloud finally retained is to derive from once
Available point cloud；
The quantity of available point cloud obtained by S10, statistic procedure S9, if dense degree reaches demand, this available point cloud is most
Terminal cloud；If dense degree is not up to demand, using the available point cloud as initial point cloud, above-mentioned S3 is repeated to current procedures,
Until the available point cloud of acquisition meets consistency requirement.
Preferred scheme, the derivation scale described in step S3 be in initial point cloud all point coordinates respectively in x, y, z three
On axis between maxima and minima range difference 1 percent, which is the actual parameter after repeatedly testing, and is not required to
It to be adjusted according to concrete application.
Further preferred scheme, in step S4 one of using in initial point cloud point as source point to cuboid 26
A direction, which is derived from, obtains new point, wherein the calculation formula newly put is：
Wherein, x_org, y_org, z_org are respectively that some in initial point cloud puts the coordinate on x, y, z axis, x_
Scalar, y_scalar, z_scalar are respectively the derivation scale in three directions of x, y, z being calculated,
The case where 3*3*3 that above formula is calculated new point coordinates in addition to source point increment of coordinate are (0,0,0), it will group
Bear 26 new point clouds described in step S4.
Scheme still more preferably, in step S6, will derive from point Cloud transform to the camera coordinates system of the i-th frame image meter
Calculate formula：
(x_cami,y_cam,z_cami)=(x_world, y_world, z_world) * Ri+ti
Wherein, (x_world, y_world, z_world) is the coordinate for deriving from point cloud in world coordinate system, Ri, tiRespectively
For the spin matrix and translation vector of the i-th frame camera, by RiWith tiTransformation, the point cloud in world coordinate system has been gone to i-th
Under frame camera coordinates system, i.e. the coordinate of point cloud after being converted in the i-th frame camera coordinates system is (x_cami,y_cami,z_cami)；
Point cloud in camera coordinates system is subjected to back projection, by each spot projection to the i-th frame effective coverage figure, projects position
The calculation formula set：
Wherein, f is camera focus, Cx、CyRespectively 0.5 times of image resolution ratio, u, the v being calculated are the spot projection
Position on to image, i.e. u rows on image, v arrange corresponding location of pixels.
Scheme still further preferably, in step S8, by after clapping image and filtering out to deriving from a point cloud, derivation point at this time
Cloud has not had external miscellaneous point, while filtering out the point partly repeated in initial point cloud yet, but is 26 due to deriving from direction
A direction, therefore there is also the points positioned at interior surface, i.e., interior point.
A kind of method of dense degree of cloud of being laid foundations based on profile validity raising Three-dimensional Gravity according to claim 1,
It is characterized in that, each derivation point of judgement described in step S9 is a method for interior point or exterior point, it is the side according to dot product
Method determines：Since the outside of the surface tangential plane of the normal vector directed towards object of source point calculates the derivation to each derivation point
Point derive from source point to the derivation point vector, if the dot product result of the vector and the normal vector of the derivation source point of the derivation point is
Just, illustrate that the derivation point is exterior point, retain the point；If the dot product result of the vector and the normal vector of the derivation source point of the derivation point
It is negative, illustrates that the source point that derives from of the derivation point is more than 90 degree to the angle of the vector and normal vector that derive from point, illustrate that the point is interior
Point deletes the point.
It should be noted that now widely used sift etc. can be used in the method that contour of object extracts in the step S1
Algorithm, finally obtained effective coverage figure is bianry image.
It should be noted that the three-dimensional of current comparative maturity can be used in the three-dimensional reconstruction step referred in the step S2
Algorithm for reconstructing, transformation matrix M are the integration M=[ of spin matrix R and translation vector t;R|t].
It should be further noted that in step S7, verify the point projected position whether effective coverage method, first
The coordinate being located on image that the point is found by above-mentioned back projection detects the pixel position in the figure of corresponding i-th frame effective coverage
Whether the pixel value set is 255.Retain the point if pixel value is 255, otherwise deletes the point.
It should be noted that in step s 8, after being filtered out to a derivation point cloud around bat image, derivation point cloud at this time is
Through the miscellaneous point in no outside, while also the point of partial error in initial point cloud being filtered out, but is 26 sides due to deriving from direction
To, therefore there is also the points positioned at interior surface.
It should be noted that in step slo, putting the consistency of cloud can voluntarily select, and million ranks be reached as high as, when thick
Density can stop iteration after reaching demand.
By using above technical scheme, the present invention is a kind of to be improved Three-dimensional Gravity based on profile validity and lays foundations the dense degree of cloud
Method be compared with the prior art, have technical effect that：
1, compared to the method for obtaining point cloud based on stereoscopic vision：The method for obtaining point cloud based on stereoscopic vision needs to provide line
Complicated image sequence is managed, does not have the region of disparity map not have a little in reconstruction process, reconstruction error is solved error by disparity map
It influences.And the requirement that texture of the present invention to image be not excessive, only initial point cloud to be offered can be closer to real-world object
Shape, just can restore to a certain extent initial point cloud loss most information.
2, compared to the method for obtaining point cloud based on motion structure：The point cloud obtained based on the method that motion structure obtains point cloud
Quantity depend on the quantity that characteristic point pair is effectively matched between adjacent two frame, the derivation of the sparse cloud taken to dense point cloud
The calculating of mode is complicated.And the quantity of derivation point cloud and the texture of image generated in the present invention does not contact directly, to initial
Point cloud does not require excessively, as long as being closer to real-world object, can make that putting cloud originally is distributed sparse ground by derivation mode
The point cloud quantity of side increases, and increases the quantity of available point cloud.
3, compared to the method based on depth image：Method based on depth image needs to provide the depth map per frame image,
Algorithm is higher to the susceptibility of the accuracy of depth map, and the matching between two clouds uses iterative algorithm so that required meter
Calculation amount is very big, and there are many matrix operation, needs to calculate on GPU.And method proposed by the present invention does not require depth map, with
Filtering out for point cloud is derived from, the quantity of calculative point is also being reduced, and calculating speed is speeded, and need not be calculated on GPU and also can
There is faster speed.
Description of the drawings
Fig. 1 is one group of image around shooting；
Fig. 2 is corresponding effective term area figure that a frame is calculated around bat image and the present invention；
Fig. 3 is the flow chart that the present invention improves point cloud quality.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, in conjunction with the accompanying drawings and embodiments to this hair
It is bright to be described in further detail, it should be understood that described herein specific examples are only used to explain the present invention, and does not have to
It is of the invention in limiting.
As shown in figure 3, first choosing one group around image sequence is clapped, as shown in Figure 1, and selecting a basic three-dimensional reconstruction side
Method claps image by opposing connection and carries out rebuilding acquisition initial point cloud, by initial point cloud, around image sequence is clapped, per the transformation of frame image
Matrix handles it using the method for the present invention, is as follows as input：
Step 1：Pixel value 255 is inserted to the profile of every frame image zooming-out wherein object, and in object region, in nothing
Object area inserts pixel value 0, obtains one group of effective image corresponding with around image is clapped, as shown in Figure 2.
Step 2：Initial point cloud is traversed, the point coordinates in record point cloud is in x respectively, y, the interval in z-axis, each axis
The centesimal size of upper interval is as the extension scale on each axis.
Step 3：According to extension scale, it regard each of initial point cloud point as source point, increases and reduce along extension scale
Directional Extension, 26 dimensions altogether, expansion obtains 26 new points, and the normal vector newly put is identical as the normal vector of source point.
The calculation formula of inflexion point：
Wherein, x_org, y_org, z_org are respectively that some in initial point cloud puts the coordinate on x, y, z axis, x_
Scalar, y_scalar, z_scalar are respectively the derivation scale in three directions of x, y, z being calculated.
Step 4：According to the transformation matrix of the first frame image of input, point Cloud transform will be derived to corresponding camera coordinates system
Under, and in back projection to first frame effective coverage figure, spot projection falls the place that pixel value is 255 in the figure of effective coverage, then protects
Stay the point；Spot projection falls the place that pixel value is 0 in the figure of effective coverage, then deletes the point.The derivation deleted point cloud is made
For new derivation point cloud, aforesaid operations are carried out to the transformation matrix of the second frame image；It operates successively, until having traversed all figures
Picture obtains final derivation point cloud at this time.Wherein, calculation formula of the Dian Yun back projections to image coordinate：
In above-mentioned formula, f is camera focus, Cx、CyRespectively 0.5 times of image resolution ratio, u, the v being calculated are should
Position on spot projection to image, i.e. u rows on image, v arrange corresponding location of pixels.
Step 5：Up to the present, the derivation point cloud of acquisition is that the point cloud of not external error passes through each derivation point
Its source point is calculated to the vector of the derivation point and the point product value of source point, which is more than zero, illustrates that the derivation point is located at the outer of source point
Side retains the point, which is less than zero, illustrates that deriving from point is located on the inside of source point, then is interior point, deletes the point.Pass through this operation energy
Delete the interior point generated in derivation history.
Step 6：Check whether the dense degree of the derivation point cloud in deleting after point reaches demand, to dense requirement is not achieved
Derivation point cloud, using the derivation point cloud as initial point cloud, the transformation matrix sequence and image sequence with input are together as defeated
Enter, the operation to repeat the above steps again after 2, until dense degree reaches demand.
The point cloud that consistency meets demand is finally obtained according to above step.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement etc., should all be included in the protection scope of the present invention made by within refreshing and principle.
Claims (4)
1. a kind of method for the dense degree of cloud of being laid foundations based on profile validity raising Three-dimensional Gravity, which is characterized in that it includes following
Step：
S1, one group is obtained by photographic equipment around clapping image sequence, to every frame around clapping image zooming-out contour of object, and by profile region
Pixel value in domain is set as 255, and the pixel value outside profile is set as 0, obtains a frame bianry image, referred to as effective coverage figure；
S2, opposing connection clap image sequence and carry out three-dimensional reconstruction step, obtain the very low point cloud of a consistency, referred to as initial point cloud,
Spin matrix R and translation vector t, spin matrix and translation vector of each frame camera relative to world coordinate system are also obtained simultaneously
Combine to form transformation matrix M；
Each point in S3, traversal initial point cloud, obtains the maximum value of all the points value on three axis of x, y, z in initial point cloud
With minimum value, and the distance between maxima and minima difference on each axis is calculated, is denoted as x_dis, y_dis, z_dis respectively,
Respectively by this three range difference divided by 100, three obtained amount, referred to as the derivation scale of initial point cloud are denoted as x_scalar, y_
scalar,z_scalar；
S4, using a point in initial point cloud as source point, corresponding step is respectively extended along the positive negative direction in three directions of x, y, z respectively
The derivation scale size calculated in rapid S3, obtains a cuboid centered on source point, the length, width and height of the cuboid are respectively
2*x_scalar, 2*y_scalar, 2*z_scalar, the source point center toward extending 26 directions altogether around cuboid,
A new point is derived in each direction, takes the normal vector of the new point identical as the normal vector of source point, and each point that derives from is remembered
Record its source point；
S5, in initial point cloud each point carry out a step S4 described in derivation operate, by obtain one derivation
Point cloud, the quantity at this cloud midpoint is 26 times of initial point cloud quantity；
S6, opposing connection clap the i-th frame image in image sequence, take out its transformation matrix M being calculated in step s 2i, by step
The derivation point cloud obtained in S5 is according to transformation matrix MiIt transforms under corresponding camera coordinates system, and will be derived from according to projection theory
On the effective coverage figure of the i-th frame obtained in each of point cloud point back projection to step S1；
S7, according to the step in S6, to projecting to the point in the inactive area in the i-th frame effective coverage figure, by its from derive from point
It is deleted in cloud, the point in the effective coverage projected in the i-th frame effective coverage figure then retains；
S8, opposing connection clap the operation that each frame in image sequence is performed both by above-mentioned steps S6 and S7, by being surround to deriving from point cloud
It projects and deletes, three-dimensional reconstruction obtains the derivation point cloud containing interior point；
Each of derivation point cloud that S9, traversal step S8 are obtained derives from point, judges that each point that derives from is interior point or exterior point,
Delete be interior point derivation point cloud, reservation is the derivation point cloud of exterior point；The point cloud finally retained is to derive from once effective
Point cloud；
The quantity of available point cloud obtained by S10, statistic procedure S9, if dense degree reaches demand, this available point cloud is maximal end point
Cloud；If dense degree is not up to demand, using the available point cloud as initial point cloud, above-mentioned S3 is repeated to current procedures, until
The available point cloud of acquisition meets consistency requirement；
Derivation scale described in step S3 be in initial point cloud all point coordinates respectively on three axis of x, y, z maximum value with it is minimum
1 the percent of range difference between value, the derivation scale are the actual parameters after repeatedly testing, need not be according to specifically answering
With adjustment；
The new point of acquisition is derived from 26 directions of the point as source point to cuboid one of using in initial point cloud in step S4,
In the calculation formula newly put be：
Wherein, x_org, y_org, z_org are respectively that some in initial point cloud puts coordinate on x, y, z axis, x_scalar,
Y_scalar, z_scalar are respectively the derivation scale in three directions of x, y, z being calculated,
The case where 3*3*3 that above formula is calculated new point coordinates in addition to source point increment of coordinate are (0,0,0), it will derive
26 new point clouds described in step S4；
In step S6, calculation formula of the point Cloud transform to the camera coordinates system of the i-th frame image will be derived from：
(x_cami,y_cami,z_cami)=(x_world, y_world, z_world) * Ri+ti
Wherein, (x_world, y_world, z_world) is the coordinate for deriving from point cloud in world coordinate system, Ri, tiRespectively
The spin matrix and translation vector of i frame cameras, by RiWith tiTransformation, the point cloud in world coordinate system has been gone into the i-th frame phase
Under machine coordinate system, i.e. the coordinate of point cloud after being converted in the i-th frame camera coordinates system is (x_cami,y_cami,z_cami)；
Point cloud in camera coordinates system is subjected to back projection, by each spot projection to the i-th frame effective coverage figure, projected position
Calculation formula：
Wherein, f is camera focus, Cx、CyRespectively 0.5 times of image resolution ratio, u, the v being calculated are the spot projection to figure
As upper position, i.e., u rows, v on image arrange corresponding location of pixels；
In step S8, after being filtered out to a derivation point cloud around bat image, derivation point cloud at this time has not had external miscellaneous point, simultaneously
Also the point of partial error in initial point cloud is filtered out, but is 26 directions due to deriving from direction, there is also positioned at table
Point inside face, i.e., interior point；
The each derivation point of judgement described in step S9 is a method for interior point or exterior point, is determined according to the method for dot product
's：Since the outside of the surface tangential plane of the normal vector directed towards object of source point calculates the derivation of the derivation point to each derivation point
Source point to the derivation point vector, if the dot product result of the normal vector of the derivation source point of the vector and the derivation point is just explanation
The derivation point is exterior point, retains the point；If the dot product result of the vector and the normal vector of the derivation source point of the derivation point is negative, say
The angle for deriving from source point to the vector and normal vector that derive from point of the bright derivation point is more than 90 degree, illustrates that the point is interior point, deletes
This point；
The method of extraction contour of object described in the step S1 is using sift algorithms, and finally obtained effective coverage figure
It is bianry image.
2. a kind of method of dense degree of cloud of being laid foundations based on profile validity raising Three-dimensional Gravity according to claim 1,
It is characterized in that, the three-dimensional reconstruction step described in the step S2 is to use three-dimensional reconstruction algorithm, and transformation matrix M is spin matrix
The integration M=[ of R and translation vector t;R|t].
3. a kind of method of dense degree of cloud of being laid foundations based on profile validity raising Three-dimensional Gravity according to claim 1,
It is characterized in that, in step S7, the projected position of point is derived from verification, and whether the method in the effective coverage in the figure of effective coverage is：
The coordinate being located on image that back projection finds the derivation point is first passed through, the pixel in the figure of corresponding i-th frame effective coverage is detected
Whether the pixel value of position is 255, if pixel value is 255, retains the point, otherwise deletes the point.
4. a kind of method of dense degree of cloud of being laid foundations based on profile validity raising Three-dimensional Gravity according to claim 1,
It is characterized in that, in step slo, putting the consistency of cloud can voluntarily select, and can stop iteration after consistency reaches demand.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610298507.9A CN106023303B (en) | 2016-05-06 | 2016-05-06 | A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610298507.9A CN106023303B (en) | 2016-05-06 | 2016-05-06 | A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106023303A CN106023303A (en) | 2016-10-12 |
| CN106023303B true CN106023303B (en) | 2018-10-26 |
Family
ID=57081878
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610298507.9A Active CN106023303B (en) | 2016-05-06 | 2016-05-06 | A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106023303B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570507B (en) * | 2016-10-26 | 2019-12-27 | 北京航空航天大学 | Multi-view-angle consistent plane detection and analysis method for monocular video scene three-dimensional structure |
| CN106887043A (en) * | 2017-03-08 | 2017-06-23 | 景致三维（江苏）股份有限公司 | The method of the method, device and three-dimensional modeling of the removal of three-dimensional modeling exterior point |
| EP3467789A1 (en) * | 2017-10-06 | 2019-04-10 | Thomson Licensing | A method and apparatus for reconstructing a point cloud representing a 3d object |
| GB2569656B (en) * | 2017-12-22 | 2020-07-22 | Zivid Labs As | Method and system for generating a three-dimensional image of an object |
| CN108655571A (en) * | 2018-05-21 | 2018-10-16 | 广东水利电力职业技术学院（广东省水利电力技工学校） | A kind of digital-control laser engraving machine, control system and control method, computer |
| CN108735279B (en) * | 2018-06-21 | 2022-04-12 | 广西虚拟现实科技有限公司 | Virtual reality upper limb rehabilitation training system for stroke in brain and control method |
| CN108846896A (en) * | 2018-06-29 | 2018-11-20 | 南华大学 | A kind of automatic molecule protein molecule body diagnostic system |
| CN109081524A (en) * | 2018-09-25 | 2018-12-25 | 江西理工大学 | A kind of intelligence mineral processing waste water reuse change system, detection method |
| CN109293434A (en) * | 2018-11-15 | 2019-02-01 | 关静 | A kind of agricultural environmental protection base manure and preparation method |
| CN111819602A (en) * | 2019-02-02 | 2020-10-23 | 深圳市大疆创新科技有限公司 | Method for increasing point cloud sampling density, point cloud scanning system and readable storage medium |
| CN110070608B (en) * | 2019-04-11 | 2023-03-31 | 浙江工业大学 | Method for automatically deleting three-dimensional reconstruction redundant points based on images |
| CN110888143B (en) * | 2019-10-30 | 2022-09-13 | 中铁四局集团第五工程有限公司 | Bridge through measurement method based on unmanned aerial vehicle airborne laser radar |
| CN111553985B (en) * | 2020-04-30 | 2023-06-13 | 四川大学 | O-graph pairing European three-dimensional reconstruction method and device |
| CN113436242B (en) * | 2021-07-22 | 2024-03-29 | 西安电子科技大学 | Method for obtaining high-precision depth value of static object based on mobile depth camera |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101271591A (en) * | 2008-04-28 | 2008-09-24 | 清华大学 | Interactive multi-vision point three-dimensional model reconstruction method |
| CN101533529A (en) * | 2009-01-23 | 2009-09-16 | 北京建筑工程学院 | Range image-based 3D spatial data processing method and device |
| US7843448B2 (en) * | 2005-11-21 | 2010-11-30 | Leica Geosystems Ag | Identification of occluded edge regions from 3D point data |
| CN102592284A (en) * | 2012-02-27 | 2012-07-18 | 上海交通大学 | Method for transforming part surface appearance three-dimensional high-density point cloud data into grayscale image |
-
2016
- 2016-05-06 CN CN201610298507.9A patent/CN106023303B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7843448B2 (en) * | 2005-11-21 | 2010-11-30 | Leica Geosystems Ag | Identification of occluded edge regions from 3D point data |
| CN101271591A (en) * | 2008-04-28 | 2008-09-24 | 清华大学 | Interactive multi-vision point three-dimensional model reconstruction method |
| CN101533529A (en) * | 2009-01-23 | 2009-09-16 | 北京建筑工程学院 | Range image-based 3D spatial data processing method and device |
| CN102592284A (en) * | 2012-02-27 | 2012-07-18 | 上海交通大学 | Method for transforming part surface appearance three-dimensional high-density point cloud data into grayscale image |
Non-Patent Citations (2)
| Title |
|---|
| Fitting range data to primitives for rapid local 3D modeling using sparse range point clouds;Soon-Wook Kwon等;《Automation in Construction》;20041231;第67-81页 * |
| 实验三 插值;yxyxyxyx41;《百度文库 https://wenku.baidu.com/view/1fb468e8f8c75fbfc77db2ab.html》;20140429;第2页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106023303A (en) | 2016-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106023303B (en) | A method of Three-dimensional Gravity is improved based on profile validity and is laid foundations the dense degree of cloud | |
| CN103345736B (en) | A kind of virtual viewpoint rendering method | |
| CN110288642B (en) | Three-dimensional object rapid reconstruction method based on camera array | |
| CN106683173B (en) | A method of Three-dimensional Gravity is improved based on neighborhood Block- matching and is laid foundations the dense degree of cloud | |
| CN103810685B (en) | A kind of super-resolution processing method of depth map | |
| CN107833270A (en) | Real-time object dimensional method for reconstructing based on depth camera | |
| CN105279789B (en) | A kind of three-dimensional rebuilding method based on image sequence | |
| CN109754459B (en) | Method and system for constructing human body three-dimensional model | |
| KR20180054487A (en) | Method and device for processing dvs events | |
| CN107025660B (en) | Method and device for determining image parallax of binocular dynamic vision sensor | |
| CN106023230B (en) | A kind of dense matching method of suitable deformation pattern | |
| CN103456043B (en) | A kind of viewpoint internetwork roaming method and apparatus based on panorama sketch | |
| CN110567441B (en) | Particle filter-based positioning method, positioning device, mapping and positioning method | |
| CN106778660B (en) | A kind of human face posture bearing calibration and device | |
| CN110310331A (en) | A kind of position and orientation estimation method based on linear feature in conjunction with point cloud feature | |
| CN111563952A (en) | Method and system for realizing stereo matching based on phase information and spatial texture characteristics | |
| CN111325828B (en) | Three-dimensional face acquisition method and device based on three-dimensional camera | |
| JP6285686B2 (en) | Parallax image generation device | |
| CN110516639B (en) | Real-time figure three-dimensional position calculation method based on video stream natural scene | |
| CN116310131A (en) | Three-dimensional reconstruction method considering multi-view fusion strategy | |
| Hung et al. | Multipass hierarchical stereo matching for generation of digital terrain models from aerial images | |
| CN112102504A (en) | Three-dimensional scene and two-dimensional image mixing method based on mixed reality | |
| CN110148168B (en) | Three-eye camera depth image processing method based on size double baselines | |
| CN108090930A (en) | Barrier vision detection system and method based on binocular solid camera | |
| CN113808185B (en) | Image depth recovery method, electronic device and storage medium |
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 |