CN110148086A - The depth polishing method, apparatus and three-dimensional rebuilding method of sparse depth figure, device - Google Patents
The depth polishing method, apparatus and three-dimensional rebuilding method of sparse depth figure, device Download PDFInfo
- Publication number
- CN110148086A CN110148086A CN201910353712.4A CN201910353712A CN110148086A CN 110148086 A CN110148086 A CN 110148086A CN 201910353712 A CN201910353712 A CN 201910353712A CN 110148086 A CN110148086 A CN 110148086A
- Authority
- CN
- China
- Prior art keywords
- depth
- sparse
- characteristic point
- image
- dense
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 107
- 238000005498 polishing Methods 0.000 title claims abstract description 65
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 32
- 230000000717 retained effect Effects 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims description 33
- 230000015654 memory Effects 0.000 claims description 23
- 230000010354 integration Effects 0.000 claims description 14
- 238000001914 filtration Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 6
- 230000001052 transient effect Effects 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000011084 recovery Methods 0.000 description 3
- 238000005267 amalgamation Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002146 bilateral effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000012466 permeate Substances 0.000 description 1
- 238000004080 punching Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of depth polishing method, apparatus of sparse depth figure and three-dimensional rebuilding method, device, the depth polishing method of the sparse depth figure includes: that at least two interpolation algorithms are respectively adopted to handle sparse depth figure to obtain corresponding processed sparse depth figure；Depth value of each characteristic point in each processed sparse depth figure in acquisition sparse depth figure respectively；The difference for retaining depth value is less than the characteristic point of predetermined threshold；Dense depth map is obtained according to the characteristic point retained.The straightforward procedure of dense depth map is obtained by applying the present invention, realizing a kind of pair of sparse depth figure and carrying out depth polishing, provides accurate data basis to carry out three-dimensional reconstruction using dense depth map.
Description
Technical field
The present invention relates to technical field of computer vision, and in particular to the depth polishing method, apparatus of sparse depth figure and
Three-dimensional rebuilding method, device.
Background technique
Three-dimensional reconstruction is that a kind of pair of three-dimension object establishes the technology for being suitable for the mathematical model of computer disposal, and is counting
The basis for being handled three-dimension object and being analyzed under calculation machine environment, is even more established in a computer for expressing objective world
The key technology of virtual reality.Three-dimensional reconstruction refers to that the three-dimensional information (shape etc.) of the object restores in the two-dimensional projection using object
Mathematical procedure and computer technology, including data acquisition, pretreatment, point cloud (fusion) and signature analysis.
Currently, acquiring the image of target scene in different location by image capture device, the sparse of target scene is obtained
Depth image, then by carrying out depth polishing to sparse depth image, dense depth image is obtained, and then according to dense depth map
As the threedimensional model for constructing target scene is a kind of one of conventional three-dimensional rebuilding method.Wherein, to sparse depth image into
The method of row depth polishing directly determines the matching degree to dense depth image and real scene image, and then direct shadow
Ring the reconstruction effect of final threedimensional model.
Summary of the invention
In view of this, the embodiment of the invention provides a kind of depth polishing method, apparatus of sparse depth figure and Three-dimensional Gravities
Construction method, device obtain dense depth image to realize to the depth polishing of sparse depth figure.
According in a first aspect, the embodiment of the invention provides a kind of depth polishing methods of sparse depth figure, comprising: respectively
At least two interpolation algorithms are used to handle the sparse depth figure to obtain corresponding processed sparse depth figure；Point
Each characteristic point is not obtained in the sparse depth figure in the depth value of each processed sparse depth figure；Retain the depth
The difference of angle value is less than the characteristic point of predetermined threshold；Dense depth map is obtained according to the characteristic point retained.
Optionally, the difference of the depth value includes: the difference of depth value and/or the difference of two squares of depth value.
Optionally, it states and the characteristic point composition that dense depth map includes: each reservation is obtained according to the characteristic point retained
Complete depth map；The corresponding original image of the sparse depth figure is obtained, and according to the original image to the complete depth
Figure carries out depth polishing, obtains dense depth map.
Optionally, described to obtain the corresponding original image of the sparse depth figure, and according to the original image to described
Complete depth map carries out depth polishing, obtains dense depth map, comprising: according to the resolution ratio of the sparse depth figure to the original
Beginning image is handled；According to treated, original image is filtered the complete depth image；After the processing
Original image and filtered complete depth map changed using default filtering algorithm to the filtered complete depth map
Generation filtering, obtains the dense depth map.
Optionally, the depth polishing method of the sparse depth figure further include: judge the resolution ratio of the dense depth map
Whether the resolution ratio of the original image is lower than；When the resolution ratio of the dense depth map is lower than the original image, by institute
It states sparse depth figure and is updated to the dense depth map, and return at least two default interpolation algorithms of the use respectively to described
The step of sparse depth figure is up-sampled, and each preset difference value algorithm corresponding intermediate sparse depth figure is obtained, until institute
The resolution ratio for stating dense depth map is identical as the resolution ratio of the original image.
According to second aspect, at least one set of equivalent pair comprising scene objects is obtained the embodiment of the invention also provides a kind of
Mesh image information；Corresponding sparse depth figure is established according to the equivalent binocular image information of each group；Using above method embodiment institute
The depth polishing method for the sparse depth figure stated carries out depth polishing to each sparse depth figure, obtains each dense depth map；
Each dense depth map is subjected to depth integration according to the equivalent binocular image information, obtains the three-dimensional of the scene objects
Model.
Optionally, equivalent binocular image information described in every group includes the first image and correspondence in the acquisition of the first predeterminated position
The first pose and the second predeterminated position obtain the second image and corresponding second pose；Equivalent binocular figure described in every group
As the acquisition methods of information include: first predeterminated position and second predeterminated position according to the scene objects, obtain
To the equivalent baseline currently organized；First pose and the second of the scene objects are determined according to the equivalent baseline
Appearance；The first image is obtained in first predeterminated position according to the first Pose Control image acquisition equipment；According to institute
It states the imaged acquisition equipment of second attitude control and obtains second image, the first image and institute in second predeterminated position
State one group of equivalent binocular image of the second image construction.
It is optionally, described that establish corresponding sparse depth figure according to the equivalent binocular image information of each group include: to each group
The first image and second image in the equivalent binocular image information carry out characteristic matching, obtain multiple groups characteristic point
Information；The parallax information of characteristic point is calculated separately according to the characteristic point information；According to the parallax information and the equivalent base
Line obtains the depth information of each characteristic point；The corresponding sparse depth figure is established according to the depth information of each characteristic point.
Optionally, it is described according to the parallax information and the equivalent baseline obtain each characteristic point depth information it
Afterwards, before the depth information according to each characteristic point establishes the corresponding sparse depth figure, the three-dimensional rebuilding method
Further include: the depth information for according to preset coordinate range and currently organizing the characteristic point obtains multiple characteristic point regions；It counts respectively
Calculate the mean depth information for each characteristic point that the characteristic point region is included；The corresponding characteristic point in each characteristic point region,
The mean depth information is determined as to the depth information of the corresponding characteristic point in the characteristic point region.
According to the third aspect, the embodiment of the invention also provides a kind of depth polishing devices of sparse depth figure, comprising: the
One processing module is handled to obtain corresponding warp the sparse depth figure at least two interpolation algorithms to be respectively adopted
The sparse depth figure of processing；Second processing module, for obtaining in the sparse depth figure each characteristic point respectively each described
The depth value of processed sparse depth figure；Third processing module, the difference for retaining the depth value are less than predetermined threshold
Characteristic point；Fourth processing module, for obtaining dense depth map according to the characteristic point retained.
According to fourth aspect, the embodiment of the invention also provides a kind of three-dimensional reconstruction apparatus, comprising: the 5th processing module,
For obtaining at least one set of equivalent binocular image information comprising scene objects；6th processing module, for equivalent according to each group
Binocular image information establishes corresponding sparse depth figure；7th processing module, for using as claimed in claim 10 sparse
The depth polishing device of depth map carries out depth polishing to each sparse depth figure, obtains each dense depth map；8th processing
Module obtains the scene for each dense depth map to be carried out depth integration according to the equivalent binocular image information
The threedimensional model of target.
According to the 5th aspect, the embodiment of the invention provides a kind of electronic equipment, comprising: memory and processor, it is described
Connection is communicated with each other between memory and the processor, computer instruction is stored in the memory, and the processor is logical
Cross and execute the computer instruction, thereby executing in a first aspect, described in its any one optional embodiment it is sparse
The depth polishing method of depth map perhaps executes three-dimensional described in second aspect or its any one optional embodiment
Method for reconstructing.
It is described computer-readable the embodiment of the invention provides a kind of computer readable storage medium according to the 6th aspect
Storage medium stores computer instruction, and the computer instruction is for executing the computer in a first aspect, it is any
A kind of depth polishing method of sparse depth figure described in optional embodiment, perhaps executes second aspect or it is any
Three-dimensional rebuilding method described in a kind of optional embodiment.
Technical solution of the present invention has the following beneficial effects:
The depth polishing method of sparse depth figure provided in an embodiment of the present invention, by using different interpolation algorithms to dilute
Thin depth map is handled, by obtaining in sparse depth figure each characteristic point in the depth of each treated sparse depth figure
Value retains the characteristic point that depth value difference is less than preset threshold, obtains dense depth map according to the characteristic point of reservation.To provide
A kind of pair of sparse depth figure carries out depth polishing and obtains the straightforward procedure of dense depth map, to utilize dense depth map progress three
Dimension rebuilds and provides accurate data basis.
Three-dimensional rebuilding method provided in an embodiment of the present invention, by obtaining at least one set of equivalent binocular comprising scene objects
Image information, and corresponding sparse depth figure is established according to the equivalent binocular image information of each group, then inserted by using different
Value-based algorithm handles sparse depth figure, by obtaining in sparse depth figure each characteristic point in each treated sparse depth
The depth value of figure is spent, retains the characteristic point that depth value difference is less than preset threshold, dense depth is obtained according to the characteristic point of reservation
Figure finally carries out depth integration to dense depth map and obtains the threedimensional model of scene objects.To provide a kind of simple field
The three-dimensional rebuilding method of scape target, this method can be realized without relying on complicated image capture device and computer equipment, should
Three-dimensional rebuilding method has the characteristics that be simple and efficient.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of flow chart of the depth polishing method of sparse depth figure according to an embodiment of the present invention；
Fig. 2 is a kind of flow chart of three-dimensional rebuilding method according to an embodiment of the present invention；
Fig. 3 is the schematic diagram of equivalent baseline and shooting direction according to an embodiment of the present invention；
Fig. 4 is another schematic diagram of equivalent baseline and shooting direction according to an embodiment of the present invention；
Fig. 5 is another schematic diagram of equivalent baseline and shooting direction according to an embodiment of the present invention；
Fig. 6 is a kind of structural schematic diagram of the depth polishing device of sparse depth figure according to an embodiment of the present invention；
Fig. 7 is a kind of structural schematic diagram of three-dimensional reconstruction apparatus according to an embodiment of the present invention；
Fig. 8 is the structural schematic diagram of a kind of electronic equipment of the embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those skilled in the art are not having
Every other embodiment obtained under the premise of creative work is made, shall fall within the protection scope of the present invention.
As long as technical characteristic involved in invention described below different embodiments does not constitute punching each other
It is prominent to be combined with each other.
The embodiment of the invention provides a kind of depth polishing methods of sparse depth figure, as shown in Figure 1, the sparse depth figure
Depth polishing method include:
Step S1: be respectively adopted at least two interpolation algorithms sparse depth figure is handled it is corresponding through handling to obtain
Sparse depth figure.Specifically, which is sparse depth figure corresponding to scene objects image, used difference
Algorithm includes: nearest neighbour's interpolation algorithm, bilinear interpolation algorithm, cube sum algorithm etc..
Step S2: depth value of each characteristic point in each processed sparse depth figure in acquisition sparse depth figure respectively.
Specifically, after sparse depth figure is up-sampled by different interpolation algorithms, the difference of different characteristic point after treatment is sparse
Depth value in depth map may change, if the depth value in different sparse depth figures differs greatly, illustrate this
The depth value of characteristic point differs greatly with the depth value really put represented by this feature point, and the depth value of this feature point is inaccurate
Really, the influence in order to avoid the characteristic point depth value of inaccuracy to final result, by this feature point deletion.
Step S3: the difference for retaining depth value is less than the characteristic point of predetermined threshold.Specifically, it is small only to retain depth value difference
In the characteristic point of predetermined threshold, the depth information of the dense depth map constituted with these characteristic points and original image keeps height
Consistency.
Step S4: dense depth map is obtained according to the characteristic point retained.In practical applications, using original image to institute
The depth map that the characteristic point of reservation is constituted carries out depth polishing, to obtain dense depth map.
S1 to step S4 through the above steps, the depth polishing method of sparse depth figure provided in an embodiment of the present invention are led to
It crosses and sparse depth figure is handled using different interpolation algorithms, by each characteristic point in acquisition sparse depth figure each
The depth value of treated sparse depth figure retains the characteristic point that depth value difference is less than preset threshold, according to the feature of reservation
Point obtains dense depth map.The simple side of dense depth map is obtained to provide a kind of pair of sparse depth figure progress depth polishing
Method, this method have simple and quick characteristic, can obtain the dense depth map for accurately reflecting the pixel depth of original image, be
Three-dimensional reconstruction, which is carried out, using dense depth map provides accurate data basis.
Specifically, in one embodiment, at least two interpolation algorithms are respectively adopted to sparse depth figure in above-mentioned step S1
It is handled to obtain corresponding processed sparse depth figure.In practical applications, such as a given resolution ratio is H × W
And the sparse depth figure D of only N number of depth value can be respectively adopted arest neighbors interpolation, bilinear interpolation and cube convolution and insert
The mode of value up-samples sparse depth figure D, respectively obtains sparse depth figure D1, D2 and D3.
Specifically, in one embodiment, above-mentioned step S2, respectively obtain sparse depth figure in each characteristic point in each warp
The depth value of the sparse depth figure of processing.In practical applications, by obtaining above-mentioned N number of characteristic point respectively in D1, D2 and D3
Depth value.
Specifically, in one embodiment, above-mentioned step S3, the difference for retaining depth value are less than the feature of predetermined threshold
Point.A in practical applications, the depth value by same characteristic point in above-mentioned D1, D2 and D3 in different images carries out numerical value respectively
It is indicated by way of comparison, such as difference and/or squared differences calculating the resulting depth value of any two kinds of interpolation algorithms
The difference of each characteristic point depth value, if the difference and/or squared differences are less than predetermined threshold, then it is assumed that this feature point is to stablize
Characteristic point, it is consistent with the depth information of original image, retained.
In an optional embodiment, above-mentioned step S4 obtains dense depth map according to the characteristic point retained, specifically
Include the following steps:
Step S41: the characteristic point respectively retained constitutes complete depth map.All characteristic point roots that will retain in above-mentioned steps S3
According to the depth information of each characteristic point, complete depth map D* is constituted.
Step S42: the corresponding original image of sparse depth figure is obtained, and complete depth map is carried out deeply according to original image
Polishing is spent, dense depth map is obtained.Specifically, step S42 includes the following steps:
Step S421: original image is handled according to the resolution ratio of sparse depth figure.In practical applications, above-mentioned dilute
The resolution ratio for dredging depth map is probably well below the resolution ratio of original image, and therefore, it is necessary to first carry out down-sampling behaviour to original image
Make, obtains the original image with sparse image equal resolution.
Step S422: according to treated, original image is filtered complete depth image.In practical applications, by phase
Original image with rate respectively carries out guidance filtering to above-mentioned complete depth map D* as guide image, such as: it can use
The polishing that depth map is completed in guidance filtering is carried out by joint bilateral filtering or Steerable filter, obtains image D '.
Step S423: according to treated original image and filtered complete depth map using default filtering algorithm to filter
Complete depth map after wave is iterated filtering, obtains dense depth map.In practical applications, with above-mentioned identical rate respectively
Original image and D ' are used as guide image, are iterated filtering to D ' using the filtering of three sides of joint, obtain more smooth dense
Depth map.
Step S424: judge whether the resolution ratio of dense depth map is lower than the resolution ratio of original image.In practical applications,
The dense depth map for restoring original image as far as possible in order to obtain, needs to obtain the dense depth with original image equal resolution
Figure, it is therefore desirable to judge whether currently available dense depth map meets the resolution ratio of original image, be executed if being unsatisfactory for
Step S425.
Step S425: when the resolution ratio of dense depth map is lower than original image, sparse depth figure is updated to dense depth
Degree figure, and return and sparse depth figure is up-sampled respectively using at least two default interpolation algorithms, obtain each preset difference value
The step of algorithm corresponding intermediate sparse depth figure, until the resolution ratio of dense depth map is identical as the resolution ratio of original image.
It in practical applications, will be current if the resolution ratio of obtained current dense depth map is lower than the resolution ratio of original image
For dense depth map as new sparse depth figure, return re-executes above-mentioned step S1, until obtained dense depth map
Resolution ratio it is identical as the resolution ratio of original image.To ensure the accuracy of dense depth map, improve subsequent thick according to this
Close depth map carries out the accuracy of three-dimensional reconstruction.
S1 to step S4 through the above steps, the depth polishing method of sparse depth figure provided in an embodiment of the present invention are led to
It crosses and sparse depth figure is handled using different interpolation algorithms, by each characteristic point in acquisition sparse depth figure each
The depth value of treated sparse depth figure retains the characteristic point that depth value difference is less than preset threshold, according to the feature of reservation
Point obtains dense depth map.The simple side of dense depth map is obtained to provide a kind of pair of sparse depth figure progress depth polishing
Method, this method have simple and quick characteristic, can obtain the dense depth map for accurately reflecting the pixel depth of original image, be
Three-dimensional reconstruction, which is carried out, using dense depth map provides accurate data basis.
The embodiment of the invention also provides a kind of three-dimensional rebuilding methods, as shown in Fig. 2, the three-dimensional rebuilding method includes:
Step S101: at least one set of equivalent binocular image information comprising scene objects is obtained.In practical applications, each
The image that equivalent binocular image information of the group comprising scene objects can be obtained captured by two known locations by camera.
Step S102: corresponding sparse depth figure is established according to the equivalent binocular image information of each group.Specifically, the sparse depth
Degree figure is the characteristic point including depth information obtained by passing through binocular ranging in equivalent binocular image information, projects to each face
It is obtained under close shooting visual angle.
Step S103: using the depth polishing method of the sparse depth figure in above-described embodiment, to each sparse depth figure into
Row depth polishing, obtains each dense depth map.Specifically, the process of depth polishing is carried out referring to above-mentioned sparse to sparse depth figure
The associated description of the depth polishing embodiment of the method for depth map, is no longer repeated herein.
Step S104: each dense depth map is carried out by depth integration according to equivalent binocular image information, obtains scene objects
Threedimensional model.In practical applications, then can will by the way that each dense depth map is respectively converted into a threedimensional model
Each threedimensional model is put under the same coordinate system and is merged, and the complete three-dimensional scenic about scene objects can be obtained.
S101 to step S104 through the above steps, three-dimensional rebuilding method provided in an embodiment of the present invention, by obtaining extremely
Few one group includes the equivalent binocular image information of scene objects, and is established according to the equivalent binocular image information of each group corresponding sparse
Then depth map is handled sparse depth figure by using different interpolation algorithms, each in sparse depth figure by obtaining
A characteristic point retains the characteristic point that depth value difference is less than preset threshold in the depth value of each treated sparse depth figure,
Dense depth map is obtained according to the characteristic point of reservation, depth integration finally is carried out to dense depth map and obtains the three-dimensional of scene objects
Model.To provide a kind of three-dimensional rebuilding method of simple scene objects, this method is without relying on complicated Image Acquisition
Equipment and computer equipment can be realized, which has the characteristics that be simple and efficient.
Specifically, in one embodiment, above-mentioned step S101 obtains at least one set of equivalent binocular comprising scene objects
Image information.In practical applications, every group of equivalent binocular image information include the first predeterminated position obtain the first image and
Corresponding first pose and the second image and corresponding second pose obtained in the second predeterminated position.Every group of equivalent binocular figure
As the acquisition methods of information specifically comprise the following steps:
Step S201: according to the first predeterminated position and the second predeterminated position of scene objects, the equivalent base currently organized
Line.In practical applications, first predeterminated position and the second predeterminated position are indicated by A and B respectively, then the equivalent baseline such as Fig. 3
It is shown, detailed process are as follows: two given positions are respectively A, B two o'clock, and specifically, the location information of A and B two o'clock can basis
What camera fuselage size and steering engine control command were calculated.Camera can be mobile from A point with arbitrary motion track
To B point (or being moved to A point from B point), equivalent baseline is the line of A and B two o'clock, and length can be calculated by the position of A and B
Out.Provide camera A and B two o'clock shooting direction perpendicular to equivalent baseline, and the shooting direction at A, B two is mutually flat
Row, thus constructs an equivalent biocular systems, and the picture that camera obtains at A, B two constitutes one group of equivalent binocular figure
Picture.
Step S202: the first pose and the second pose of scene objects are determined according to equivalent baseline.In practical applications, it takes the photograph
As the pose that head is shot in two positions above-mentioned A and B, can be obtained by the fuselage size and steering engine control command of video camera
It arrives, still, due in shooting process it cannot be guaranteed that the absolute rest of camera, it is therefore desirable to be believed by characteristic point obtained
Breath carries out feature point extraction and feature to estimate the pose of camera, to image obtained in above-mentioned steps S201
Match, the parallax and depth value at characteristic point can be calculated.Then information is carried out with fuselage size and steering engine control command to merge,
Obtain accurate camera pose when shooting equivalent binocular image to get to A point the first pose and B point the
Two poses.
Step S203: the first image is obtained in the first predeterminated position according to the first Pose Control image acquisition equipment.
Step S204: according to the imaged acquisition equipment of second attitude control the second predeterminated position obtain the second image, first
One group of equivalent binocular image of image and the second image construction.In practical applications, above-mentioned camera be one can be with shooting figure
The camera of picture, lens focus, photosensitive unit, camera lens type are without limitation, such as can be the camera of standard, can also be with
It is fish-eye camera or other camera lenses.By taking fish-eye camera as an example, it can be demarcated with camera calibration tool, then
Acquisition image is used further to after being corrected to the image that it takes；An equivalent flake biocular systems can also be constructed,
Extracting and matching feature points are carried out by flake binocular ranging algorithm, calculate parallax and depth value.
In practical applications, as shown in figure 4, by taking three positions as an example, the acquisition of above-mentioned every group of equivalent binocular image information
Method detailed process are as follows: given tri- positions A, B, C, and a shooting direction is specified for two positions A, C, for B
Specified 2 shooting direction are set, step A is respectively adopted at A, B and B, C and constructs equivalent biocular systems to obtain the three of characteristic point
Tie up the pose of coordinate and camera.It can get the pose of two groups of three-dimensional feature points and camera at A, B, C as a result,.It needs
It should be noted that the practical motion track of camera needs not be straight line, the arbitrary trajectory as shown in filled arrows can be.
As a kind of alternative embodiments, as shown in figure 5, to give multiple positions, and given multiple positions can be with
For the position for constituting annular locus, the acquisition methods detailed process of the equivalent binocular image information of above-mentioned every group are as follows: camera
In moving process successively by position A, B, C ..., H, be then return to position A, on each position towards perpendicular to etc.
The direction for imitating baseline shoots equivalent binocular image, thus to obtain the pose of one group of camera, and by double under each pose
Some characteristic points for having real depth information that mesh matching obtains.It should be noted that the practical motion track of camera is not required to
If straight line, the arbitrary trajectory as shown in filled arrows can be.It, can since these positions constitute an annular locus
To establish pose figure, then these poses are optimized using the methods of figure optimization, reduce accumulated error, it is more accurate to obtain
One group of pose.
Specifically, in one embodiment, above-mentioned step S102 is established corresponding according to the equivalent binocular image information of each group
Sparse depth figure, specifically comprises the following steps:
Step S301: in each group of equivalent binocular image information the first image and the second image carry out characteristic matching,
Obtain multiple groups characteristic point information.In practical applications, above-mentioned two images shot in A point and the point position B are subjected to binocular spy
Sign matching, obtains the information of the characteristic point of each group of equivalent binocular image.
Step S302: the parallax information of characteristic point is calculated separately according to characteristic point information.Specifically, the step can adopt
It is calculated with the method for calculating parallax in the prior art, is no longer repeated herein.
Step S303: the depth information of each characteristic point is obtained according to parallax information and equivalent baseline.Specifically, according to above-mentioned
The equivalent baseline of the parallax information of each characteristic point and above-mentioned A and B two o'clock construction, can be according to characteristic point depth in the prior art
The calculation method of value calculates the depth information of this feature point, and the calculating process of specific depth value refers to the calculating of the prior art
Journey is no longer repeated herein.
Step S304: according to the depth information of preset coordinate range and current group characteristic point, multiple characteristic point regions are obtained.
Step S305: the mean depth information for each characteristic point that characteristic point region is included is calculated separately.
Step S306: mean depth information is determined as characteristic point region by the corresponding characteristic point in each characteristic point region
The depth information of corresponding characteristic point.
Step S307: corresponding sparse depth figure is established according to the depth information of each characteristic point.Each group characteristic point is carried out
The projecting characteristic points for closing on visual angle obtain the sparse depth figure at multiple and different visual angles.
In practical applications, it in order to improve the accuracy of sparse depth figure, needs to click through multiple groups feature obtained above
The processing of row time-domain stability, to obtain most characterizing the characteristic point of original image depth information, detailed process is, by each group of image
Projecting characteristic points to adjacent shooting pose (visual angle) under, two groups of characteristic points under adjacent view often will appear certain friendship
Collection.Since there are errors for depth measurement, the characteristic point in these intersections, even if the actually same point, can also go out
Now different positions.In this regard, it is proposed that carrying out time-domain stability processing, specific step to the Neighbor Points occurred in three dimensions
Suddenly it can be and delimit a preset zonule, the characteristic point region which is constituted is actually adjacent sparse depth figure view
The same point that angle is seen, this point position under different sparse depth angle of field may have deviation, and characteristic point region is exactly these
The view field of point (the only one point actually in physical space) spatially.Then in characteristic point region
Point is handled, and looks for most representational point as representing, we will recognize the multiple characteristic points being located in this zonule
To be the same characteristic point, an estimated value of the average value of these characteristic point three-dimensional coordinates as this feature point is taken；Such as: ginseng
It examines in Embedding Temporally Consistent Depth Recovery for Real-time (IROS2018)
Characteristic point antihunt means delimit a zonule for Neighbor Points, will be located at multiple characteristic point coordinate arrangements in this zonule
It is each to arrange the coordinate for representing a point for a matrix, value (such as the of every a line different characteristic point in the same reference axis
A line is x coordinate value, and the second row is y-coordinate value, and the third line is z coordinate value).Low-rank sparse decomposition method is used to the matrix
(LRSD) the biggish characteristic point of error is filtered out, is averaged to the three-dimensional coordinate of the lesser characteristic point of error, obtains this feature
One estimated value of point.By carrying out above-mentioned processing to characteristic points all in intersection, the stable spy of one group of three-dimensional coordinate can get
Sign point.
Specifically, in one embodiment, above-mentioned step S103, using the depth of the sparse depth figure in above-described embodiment
Polishing method carries out depth polishing to each sparse depth figure, obtains each dense depth map.Specifically, sparse depth figure is carried out
The process of depth polishing is no longer gone to live in the household of one's in-laws on getting married herein referring to the associated description of the depth polishing embodiment of the method for above-mentioned sparse depth figure
It states.
Specifically, in one embodiment, above-mentioned step S104, according to equivalent binocular image information by each dense depth map
Depth integration is carried out, the threedimensional model of scene objects is obtained.It in practical applications, can be according to all equivalent binocular image information
Dense depth map under multidigit appearance is put under global coordinate system and merges, obtains one by the pose at the multiple visual angles for being included
A complete three-dimensional scenic.The mode of depth integration is determined by the representation of three-dimensional scenic, can be a cloud, gridding mould
Type, voxel or volume indicate, respectively correspond different amalgamation modes, the present invention is not limited this time.It is listed below two kinds of generations
The amalgamation mode of table:
Fusion based on cloud: three-dimensional point cloud is converted according to camera internal reference by dense depth map first, then basis
Pose is placed on point cloud under the same coordinate system, for the standardized a small range of each three-dimensional point for being overlapped partial dot cloud, is located at
Several three-dimensional points within the scope of this permeate point, and fused coordinate is being averaged for the point coordinate within the scope of this
Value；After having merged all the points cloud, a threedimensional model with point cloud representation can be obtained.
The fusion indicated based on volume: a volumetric region is defined for entire scene objects first, volumetric region is divided
For several voxels, Stage Symbol function (Truncated Signed Distance is defined in whole volume region
Function, TSDF).Then with the method for KinectFusion according to the depth value of each depth map to corresponding voxel value into
Row updates, and can be obtained the threedimensional model that a complete volume indicates after updating according to all depth maps.
Below in conjunction with concrete application example, three-dimensional rebuilding method provided in an embodiment of the present invention is illustrated.
Desktop human-computer interaction that the present invention is directed to a desk lamp shape, that camera can be moved according to control signal fills
Set, the default camera site of camera uses station acquisition image information as shown in Figure 5, then by computer processor into
Row three-dimensional reconstruction, the specific implementation process is as follows:
Step 1) solves the control of each mechanical part of camera according to given camera position and shooting towards sequence
Parameter processed；
Step 2), according to control parameter control mechanical part movement, and calculate camera posture and equivalent baseline it is long
Degree；
Step 3), camera shoots image according to specified direction in given position, equivalent etc. to being made up of equivalent baseline
Two images of effect binocular image carry out the binocular ranging based on ORB feature, obtain sparse depth map；
Step 4) calculates the camera in the shooting of each sub-picture according to match point, fuselage size and control parameter
Pose；
Step 5), repeat step above-mentioned steps 2) -4), until all positions on image have been used for binocular ranging with
Depth recovery；
Step 6) establishes pose figure according to the winding information in position sequence, and solves packet to estimated using G2O
The pose of camera come optimizes；
Three-dimensional feature point is placed under global coordinate system by step 7) according to the pose after optimization, and it is steady to carry out time domain
It is fixed；
Projecting characteristic points to the image coordinate system for the shooting angle closed on are generated the sparse depth of multi-angle of view by step 8)
Figure；
Step 9) is that guidance carries out multiple dimensioned recovery to depth map with image, obtains the dense depth map of multi-angle of view；
Dense depth map is carried out the depth integration based on TSDF by step 10), obtains dense scene threedimensional model.
S101 to step S104 through the above steps, three-dimensional rebuilding method provided in an embodiment of the present invention, by obtaining extremely
Few one group includes the equivalent binocular image information of scene objects, and is established according to the equivalent binocular image information of each group corresponding sparse
Then depth map is handled sparse depth figure by using different interpolation algorithms, each in sparse depth figure by obtaining
A characteristic point retains the characteristic point that depth value difference is less than preset threshold in the depth value of each treated sparse depth figure,
Dense depth map is obtained according to the characteristic point of reservation, depth integration finally is carried out to dense depth map and obtains the three-dimensional of scene objects
Model.To provide a kind of three-dimensional rebuilding method of simple scene objects, this method is without relying on complicated Image Acquisition
Equipment and computer equipment can be realized, which has the characteristics that be simple and efficient.It is provided in an embodiment of the present invention
Single camera is used only as sensor in three-dimensional rebuilding method, low to equipment requirement and manufacturing cost, convenient for miniaturization, according to
Steering engine angle and fuselage size determine the scale of three-dimensional reconstruction, and the multi-angle of view of scene is obtained by a series of equivalent binocular measurement
Under sparse depth figure obtain the dense threedimensional model of captured scene after carrying out depth polishing and depth integration.This three
Tie up method for reconstructing can on computer CPU single thread real time execution, calculate cost is relatively low, from measurement accuracy, equipment cost, meter
It calculates resource and angularly all has good effect.
The embodiment of the invention also provides a kind of depth polishing devices of sparse depth figure, as shown in fig. 6, the sparse depth
The depth polishing device of figure includes:
First processing module 1 is handled to obtain sparse depth figure at least two interpolation algorithms to be respectively adopted
Corresponding processed sparse depth figure.Detailed content referring to step S1 in above-described embodiment associated description.
Second processing module 2, for each characteristic point in acquisition sparse depth figure respectively in each processed sparse depth
The depth value of figure.Detailed content referring to step S2 in above-described embodiment associated description.
Third processing module 3, the difference for retaining depth value are less than the characteristic point of predetermined threshold.Detailed content is referring to upper
State the associated description of step S3 in embodiment.
Fourth processing module 4, for obtaining dense depth map according to the characteristic point retained.Detailed content is referring to above-mentioned reality
Apply the associated description of step S4 in example.
By the cooperative cooperating of above-mentioned each component part, the depth polishing of sparse depth figure provided in an embodiment of the present invention
Device is handled sparse depth figure by using different interpolation algorithms, by obtaining each feature in sparse depth figure
Point retains the characteristic point that depth value difference is less than preset threshold, according to guarantor in the depth value of each treated sparse depth figure
The characteristic point stayed obtains dense depth map.Dense depth map is obtained to provide a kind of pair of sparse depth figure progress depth polishing
Straightforward procedure, this method have simple and quick characteristic, the dense of the pixel depth for accurately reflecting original image can be obtained
Depth map provides accurate data basis to carry out three-dimensional reconstruction using dense depth map.
The embodiment of the invention also provides a kind of three-dimensional reconstruction apparatus, as shown in fig. 7, the three-dimensional reconstruction apparatus includes:
5th processing module 101, for obtaining at least one set of equivalent binocular image information comprising scene objects.In in detail
Hold the associated description referring to step S101 in above-described embodiment.
6th processing module 102, for establishing corresponding sparse depth figure according to the equivalent binocular image information of each group.In detail
Content referring to step S102 in above-described embodiment associated description.
7th processing module 103, for the depth polishing device using the sparse depth figure such as claim 10, to each dilute
It dredges depth map and carries out depth polishing, obtain each dense depth map.Detailed content referring to step S103 in above-described embodiment correlation
Description.
8th processing module 104 is obtained for each dense depth map to be carried out depth integration according to equivalent binocular image information
To the threedimensional model of scene objects.Detailed content referring to step S104 in above-described embodiment associated description.
Pass through the cooperative cooperating of above-mentioned each component part, three-dimensional reconstruction apparatus provided in an embodiment of the present invention, by obtaining
Take it is at least one set of include the equivalent binocular image information of scene objects, and established according to the equivalent binocular image information of each group corresponding
Then sparse depth figure is handled sparse depth figure by using different interpolation algorithms, by obtaining sparse depth figure
In each characteristic point in the depth value of each treated sparse depth figure, retain the feature that depth value difference is less than preset threshold
Point obtains dense depth map according to the characteristic point of reservation, finally carries out depth integration to dense depth map and obtains scene objects
Threedimensional model.To provide a kind of three-dimensional rebuilding method of simple scene objects, this method is without relying on complicated image
Acquisition equipment and computer equipment can be realized, which has the characteristics that be simple and efficient.
The embodiment of the invention also provides a kind of electronic equipment, as shown in figure 8, the electronic equipment may include processor
901 and memory 902, wherein processor 901 can be connected with memory 902 by bus or other modes, with logical in Fig. 8
It crosses for bus connection.
Processor 901 can be central processing unit (Central Processing Unit, CPU).Processor 901 may be used also
Think other general processors, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), field programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
The combination of the chips such as discrete hardware components or above-mentioned all kinds of chips.
Memory 902 is used as a kind of non-transient computer readable storage medium, can be used for storing non-transient software program, non-
Transient computer executable program and module, the journey as corresponding to the access device fault handling method in the embodiment of the present invention
Sequence instruction/module.Non-transient software program, instruction and the module that processor 901 is stored in memory 902 by operation,
Thereby executing the various function application and data processing of processor, the i.e. depth of thin depth map in realization above method embodiment
Polishing method is spent, alternatively, realizing the three-dimensional rebuilding method in above method embodiment.
Memory 902 may include storing program area and storage data area, wherein storing program area can store operation system
Application program required for system, at least one function；It storage data area can the data etc. that are created of storage processor 901.In addition,
Memory 902 may include high-speed random access memory, can also include non-transient memory, and a for example, at least disk is deposited
Memory device, flush memory device or other non-transient solid-state memories.In some embodiments, it includes opposite that memory 902 is optional
In the remotely located memory of processor 901, these remote memories can pass through network connection to processor 901.Above-mentioned net
The example of network includes but is not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.
One or more module is stored in memory 902, and when being executed by processor 901, it is real to execute the above method
The depth polishing method of the thin depth map in example is applied, alternatively, executing the three-dimensional rebuilding method in above method embodiment.
Above-mentioned electronic equipment detail can be corresponded to refering to associated description corresponding in above method embodiment and effect
Understood, details are not described herein again.
It is that can lead to it will be understood by those skilled in the art that realizing all or part of the process in above-described embodiment method
Computer program is crossed to instruct relevant hardware and complete, program can be stored in a computer-readable storage medium, should
Program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, storage medium can be magnetic disk, CD, read-only
Storage memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM),
Flash memory (Flash Memory), hard disk (Hard Disk Drive, abbreviation: HDD) or solid state hard disk (Solid-
State Drive, SSD) etc.；Storage medium can also include the combination of the memory of mentioned kind.
Although being described in conjunction with the accompanying the embodiment of the present invention, those skilled in the art can not depart from the present invention
Spirit and scope in the case where various modifications and variations can be made, such modifications and variations are each fallen within by appended claims institute
Within the scope of restriction.
Claims (13)
1. a kind of depth polishing method of sparse depth figure characterized by comprising
Be respectively adopted at least two interpolation algorithms the sparse depth figure is handled it is corresponding processed sparse to obtain
Depth map；
Each characteristic point is obtained in the sparse depth figure respectively in the depth value of each processed sparse depth figure；
The difference for retaining the depth value is less than the characteristic point of predetermined threshold；
Dense depth map is obtained according to the characteristic point retained.
2. the depth polishing method of sparse depth figure according to claim 1, which is characterized in that the difference of the depth value
It include: the difference of depth value and/or the difference of two squares of depth value.
3. the depth polishing method of sparse depth figure according to claim 1, which is characterized in that described according to being retained
Characteristic point obtains dense depth map
The characteristic point of each reservation constitutes complete depth map；
The corresponding original image of the sparse depth figure is obtained, and the complete depth map is carried out deeply according to the original image
Polishing is spent, dense depth map is obtained.
4. the depth polishing method of sparse depth figure according to claim 3, which is characterized in that the acquisition is described sparse
The corresponding original image of depth map, and depth polishing is carried out to the complete depth map according to the original image, it obtains dense
Depth map, comprising:
The original image is handled according to the resolution ratio of the sparse depth figure；
According to treated, original image is filtered the complete depth image；
According to treated the original image and filtered complete depth map using default filtering algorithm to the filtering after
Complete depth map be iterated filtering, obtain the dense depth map.
5. the depth polishing method of sparse depth figure according to claim 1, which is characterized in that further include:
Judge whether the resolution ratio of the dense depth map is lower than the resolution ratio of the original image；
When the resolution ratio of the dense depth map is lower than the original image, the sparse depth figure is updated to described dense
Depth map, and return at least two default interpolation algorithms of the use and the sparse depth figure is up-sampled respectively, it obtains
The step of each preset difference value algorithm corresponding intermediate sparse depth figure, until the resolution ratio of the dense depth map with it is described
The resolution ratio of original image is identical.
6. a kind of three-dimensional rebuilding method characterized by comprising
Obtain at least one set of equivalent binocular image information comprising scene objects；
Corresponding sparse depth figure is established according to the equivalent binocular image information of each group；
Using the depth polishing method of sparse depth figure as described in any one of claims 1-3, to each sparse depth figure
Depth polishing is carried out, each dense depth map is obtained；
Each dense depth map is subjected to depth integration according to the equivalent binocular image information, obtains the scene objects
Threedimensional model.
7. three-dimensional rebuilding method according to claim 6, which is characterized in that equivalent binocular image information described in every group includes
In the first image of the first predeterminated position acquisition and corresponding first pose and the second image obtained in the second predeterminated position
With corresponding second pose；
The acquisition methods of equivalent binocular image information described in every group include:
According to first predeterminated position and second predeterminated position of the scene objects, the equivalent base currently organized
Line；
First pose and second pose of the scene objects are determined according to the equivalent baseline；
The first image is obtained in first predeterminated position according to the first Pose Control image acquisition equipment；
Second image is obtained in second predeterminated position according to the imaged acquisition equipment of the second attitude control, described the
One image and second image construction, one group of equivalent binocular image.
8. three-dimensional rebuilding method according to claim 7, which is characterized in that described according to the equivalent binocular image information of each group
Establishing corresponding sparse depth figure includes:
The first image and second image in the equivalent binocular image information described in each group carry out characteristic matching, obtain
To multiple groups characteristic point information；
The parallax information of characteristic point is calculated separately according to the characteristic point information；
The depth information of each characteristic point is obtained according to the parallax information and the equivalent baseline；
The corresponding sparse depth figure is established according to the depth information of each characteristic point.
9. three-dimensional rebuilding method according to claim 8, which is characterized in that described according to the parallax information and described
After equivalent baseline obtains the depth information of each characteristic point, established in the depth information according to each characteristic point corresponding described
Before sparse depth figure, the three-dimensional rebuilding method further include:
The depth information for according to preset coordinate range and currently organizing the characteristic point, obtains multiple characteristic point regions；
Calculate separately the mean depth information for each characteristic point that the characteristic point region is included；
The corresponding characteristic point in each characteristic point region, it is corresponding to be determined as the characteristic point region for the mean depth information
The depth information of characteristic point.
10. a kind of depth polishing device of sparse depth figure characterized by comprising
First processing module is handled the sparse depth figure at least two interpolation algorithms to be respectively adopted to obtain pair
The processed sparse depth figure answered；
Second processing module, for obtaining in the sparse depth figure each characteristic point respectively in each processed sparse depth
Spend the depth value of figure；
Third processing module, the difference for retaining the depth value are less than the characteristic point of predetermined threshold；
Fourth processing module, for obtaining dense depth map according to the characteristic point retained.
11. a kind of three-dimensional reconstruction apparatus characterized by comprising
5th processing module, for obtaining at least one set of equivalent binocular image information comprising scene objects；
6th processing module, for establishing corresponding sparse depth figure according to the equivalent binocular image information of each group；
7th processing module, for the depth polishing device using sparse depth figure as claimed in claim 10, to each described
Sparse depth figure carries out depth polishing, obtains each dense depth map；
8th processing module, for each dense depth map to be carried out depth integration according to the equivalent binocular image information,
Obtain the threedimensional model of the scene objects.
12. a kind of electronic equipment characterized by comprising
Memory and processor communicate with each other connection, are stored in the memory between the memory and the processor
Computer instruction, the processor are described in any item thereby executing claim 1-5 by executing the computer instruction
The depth polishing method of sparse depth figure, alternatively, perform claim requires the described in any item three-dimensional rebuilding methods of 6-9.
13. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer to refer to
It enables, the computer instruction is for making the computer thereby executing the described in any item sparse depth figures of claim 1-5
Depth polishing method, alternatively, perform claim requires the described in any item three-dimensional rebuilding methods of 6-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910353712.4A CN110148086B (en) | 2019-04-28 | 2019-04-28 | Depth filling method and device for sparse depth map and three-dimensional reconstruction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910353712.4A CN110148086B (en) | 2019-04-28 | 2019-04-28 | Depth filling method and device for sparse depth map and three-dimensional reconstruction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110148086A true CN110148086A (en) | 2019-08-20 |
| CN110148086B CN110148086B (en) | 2023-02-17 |
Family
ID=67594762
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910353712.4A Active CN110148086B (en) | 2019-04-28 | 2019-04-28 | Depth filling method and device for sparse depth map and three-dimensional reconstruction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110148086B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113780349A (en) * | 2021-08-09 | 2021-12-10 | 深圳奥锐达科技有限公司 | Method for acquiring training sample set, model training method and related device |
| CN118212383A (en) * | 2024-05-21 | 2024-06-18 | 成都唐米科技有限公司 | Depth sparse map-based backhaul charging method and system for small unmanned equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160012633A1 (en) * | 2014-07-09 | 2016-01-14 | Google Inc. | High-Quality Stereo Reconstruction Featuring Depth Map Alignment and Outlier Identification |
| CN106651938A (en) * | 2017-01-17 | 2017-05-10 | 湖南优象科技有限公司 | Depth map enhancement method blending high-resolution color image |
| CN108062769A (en) * | 2017-12-22 | 2018-05-22 | 中山大学 | A kind of fast deep restoration methods for three-dimensional reconstruction |
| US20190007671A1 (en) * | 2017-06-28 | 2019-01-03 | Siemens Healthcare Gmbh | System For Dense Registration Of Two-Dimensional Depth Images |
-
2019
- 2019-04-28 CN CN201910353712.4A patent/CN110148086B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160012633A1 (en) * | 2014-07-09 | 2016-01-14 | Google Inc. | High-Quality Stereo Reconstruction Featuring Depth Map Alignment and Outlier Identification |
| CN106651938A (en) * | 2017-01-17 | 2017-05-10 | 湖南优象科技有限公司 | Depth map enhancement method blending high-resolution color image |
| US20190007671A1 (en) * | 2017-06-28 | 2019-01-03 | Siemens Healthcare Gmbh | System For Dense Registration Of Two-Dimensional Depth Images |
| CN108062769A (en) * | 2017-12-22 | 2018-05-22 | 中山大学 | A kind of fast deep restoration methods for three-dimensional reconstruction |
Non-Patent Citations (1)
| Title |
|---|
| 贺秉安等: "基于稀疏激光点云数据和单帧图像融合的三维重构算法", 《计测技术》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113780349A (en) * | 2021-08-09 | 2021-12-10 | 深圳奥锐达科技有限公司 | Method for acquiring training sample set, model training method and related device |
| WO2023015880A1 (en) * | 2021-08-09 | 2023-02-16 | 深圳奥锐达科技有限公司 | Acquisition method for training sample set, model training method and related apparatus |
| CN113780349B (en) * | 2021-08-09 | 2023-07-11 | 深圳奥锐达科技有限公司 | Training sample set acquisition method, model training method and related device |
| CN118212383A (en) * | 2024-05-21 | 2024-06-18 | 成都唐米科技有限公司 | Depth sparse map-based backhaul charging method and system for small unmanned equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110148086B (en) | 2023-02-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109003325B (en) | Three-dimensional reconstruction method, medium, device and computing equipment | |
| CN115100339B (en) | Image generation method, device, electronic equipment and storage medium | |
| CN115082639B (en) | Image generation method, device, electronic equipment and storage medium | |
| CN108537876A (en) | Three-dimensional rebuilding method, device, equipment based on depth camera and storage medium | |
| WO2020206903A1 (en) | Image matching method and device, and computer readable storage medium | |
| CN105378796B (en) | Scalable volume 3D reconstruct | |
| CN110570522B (en) | Multi-view three-dimensional reconstruction method | |
| CN108898630A (en) | A kind of three-dimensional rebuilding method, device, equipment and storage medium | |
| CN108898676B (en) | Method and system for detecting collision and shielding between virtual and real objects | |
| CN106023147B (en) | The method and device of DSM in a kind of rapidly extracting linear array remote sensing image based on GPU | |
| US20110128286A1 (en) | Image restoration apparatus and method thereof | |
| CN107707899B (en) | Multi-view image processing method, device and electronic equipment comprising moving target | |
| WO2023015409A1 (en) | Object pose detection method and apparatus, computer device, and storage medium | |
| CN116977522A (en) | Rendering method and device of three-dimensional model, computer equipment and storage medium | |
| CN116051747A (en) | House three-dimensional model reconstruction method, device and medium based on missing point cloud data | |
| CN115272608A (en) | Human hand reconstruction method and equipment | |
| JPH11504452A (en) | Apparatus and method for reproducing and handling a three-dimensional object based on a two-dimensional projection | |
| CN110415322A (en) | The generation method and device of the action command of virtual objects model | |
| CN110148086A (en) | The depth polishing method, apparatus and three-dimensional rebuilding method of sparse depth figure, device | |
| CN107845126A (en) | A kind of three-dimensional animation manufacturing method and device | |
| CN110378948B (en) | 3D model reconstruction method and device and electronic equipment | |
| CN115619986B (en) | Scene roaming method, device, equipment and medium | |
| CN111105489A (en) | Data synthesis method and apparatus, storage medium, and electronic apparatus | |
| CN115082636B (en) | Single image three-dimensional reconstruction method and device based on mixed Gaussian network | |
| CN116843867A (en) | Augmented reality virtual-real fusion method, electronic device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |