CN109285117A - A kind of more maps splicing blending algorithm based on map feature - Google Patents
A kind of more maps splicing blending algorithm based on map feature Download PDFInfo
- Publication number
- CN109285117A CN109285117A CN201811029781.1A CN201811029781A CN109285117A CN 109285117 A CN109285117 A CN 109285117A CN 201811029781 A CN201811029781 A CN 201811029781A CN 109285117 A CN109285117 A CN 109285117A
- Authority
- CN
- China
- Prior art keywords
- map
- point
- feature
- algorithm
- rigid body
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000002156 mixing Methods 0.000 title claims abstract description 13
- 238000013178 mathematical model Methods 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 16
- 230000004044 response Effects 0.000 claims description 9
- 238000000034 method Methods 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000006467 substitution reaction Methods 0.000 claims description 3
- 230000001174 ascending effect Effects 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 238000010845 search algorithm Methods 0.000 claims description 2
- 239000000284 extract Substances 0.000 claims 2
- 238000002844 melting Methods 0.000 claims 1
- 230000008018 melting Effects 0.000 claims 1
- 230000008901 benefit Effects 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- 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
- G06T3/4038—Image mosaicing, e.g. composing plane images from plane sub-images
-
- 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
- G06T3/60—Rotation of whole images or parts thereof
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of, and more maps based on map feature splice blending algorithm, include the following steps: the edge pixel point for extracting grating map to be spliced, establish rigid body translation mathematical model, convert image registration minimization problem for map Bonding Problem；Based on improved SURF algorithm, feature is extracted from map to be spliced, determines optimal initial splicing parameter；Parameter will initially be spliced and substitute into mathematical model, improved ICP algorithm is based on, solves accurate rigid body translation result；It rotated, translated and is merged accordingly as a result, treating stitching image according to accurate rigid body translation, obtain the accurate splicing result of grating map.The present invention has initial splicing speed faster, is conducive to fairly large map splicing；There is stronger robustness and higher precision to biggish map environment.
Description
Technical field
The invention belongs to robot control fields, and in particular to a kind of more maps splicing fusion calculation based on map feature
Method.
Background technique
Mobile robot (Robot) is the automatic installations for executing work, it can not only receive mankind commander, but also can be with
The program of preparatory layout is run, can also can assist in or replace according to principle program action formulated with artificial intelligence technology
The mankind complete dangerous work, such as can put into mobile robot in this extreme environment of mountain area substation.Therefore draw
Domestic and foreign scholars have been played more and more to pay attention to.Robot, which wants to realize, to be explored and navigates in circumstances not known, needs first to complete map
Creation.The prior art is mostly to create smaller or medium scale environmental map, however, with the continuous expansion of environment scale,
As still carried out map building using the prior art, the map of creation precision, robustness and in terms of will all deposit
In problem, the demand of extensive environment can not be adapted to.
Summary of the invention
The purpose of the present invention is to provide a kind of, and more maps based on map feature splice blending algorithm.
The technical solution for realizing the aim of the invention is as follows: a kind of more maps splicing blending algorithm based on map feature,
Include the following steps:
Step 1, the edge pixel point for extracting grating map to be spliced, establish rigid body translation mathematical model, map are spliced
Problem is converted into image registration minimization problem；
Step 2 is based on improved SURF (Speeded Up Robust Features) algorithm, mentions from map to be spliced
Feature is taken, determines optimal initial splicing parameter；
Step 3 will initially splice parametric results substitution mathematical model, be based on improved ICP (iterative closest
Point) algorithm solves accurate rigid body translation result；
Step 4 is rotated accordingly, translated and is merged as a result, treating stitching image according to accurate rigid body translation, is obtained
Obtain the accurate splicing result of grating map.
Compared with prior art, the present invention its remarkable advantage are as follows: 1) present invention has initial splicing speed faster, favorably
Splice in fairly large map；2) present invention has stronger robustness and higher essence for biggish map environment
Degree；3) present invention utilize SURF algorithm, available more effective feature to optimal initial splicing parameter.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow charts of more maps of map feature splicing blending algorithm.
Fig. 2 is the flow chart of the improved SURF algorithm of the present invention.
Fig. 3 is the flow chart that the present invention obtains optimal solution.
Fig. 4 is the scale space figure that the present invention constructs.
Specific embodiment
The principle of the present invention and scheme are further illustrated with reference to the accompanying drawings and detailed description.
As shown in Figure 1, the process of more maps splicing blending algorithm based on map feature are as follows:
Step 1: extracting the edge pixel point of grating map to be spliced, establish rigid body translation mathematical model, map is spliced
Problem is converted into image registration minimization problem.
For there are overlapping region two-dimensional grid map G and W, Boundary extracting algorithm is utilized from grating map to be spliced
Edge pixel point is extracted respectivelyWithNgAnd NwRespectively indicate the element of edge pixel point set G and W
Number.It will be in map G and the edge pixel point set G of the overlapping region map WεIt indicates, it is the subset of G, overlapping percentages ε
To indicate.
Map Bonding Problem can be regarded as image registration problem, and calculating acquires rigid body translation parameter T={ R, t }, and R is one
A Two Dimensional Rotating transformation matrix, t ∈ R is two-dimension translational transformation vector, so that transformed treated point set T (G) can be well
It is matched with point set W, and then grating map Bonding Problem is expressed as minimization problem, such as following formula:
s.t.RTR=I2×2, Δ (R)=1
In formula, λ is control parameter, | | represent element number in set, minimum overlay percentage εminIt indicates.
Step 2: using the vector for improving SURF algorithm and extracting the SURF feature of grating map to be spliced, determine it is optimal just
Begin splicing parameter；
In order to solve the rigid body translation in above formula, preferable feature extraction is needed initially to splice parameter (R0,t0), therefore adopt
Initially splice parameter based on to improvement SURF algorithm with a kind of to obtain.SURF algorithm is improved to filter using integral image, box
With Haar small echo accelerating algorithm, stable margin is all had to the complicated variation such as scale, noise, illumination and visual angle difference, is had
Body realizes that steps are as follows:
Step 2.1, building scale space
Since box filtering and integral image is utilized in SURF algorithm, it is unrelated with picture size to calculate time-consuming, therefore in spy
SURF algorithm ratio SIFT efficiency of algorithm is higher when sign detection, and SURF algorithm is exactly that original image and box filter are accelerated using the property
The convolution algorithm of wave.The scale space pyramid bottom of SURF algorithm is filtered using the box that a size is 9x9, box filter
Minimum 6 pixel step lengths of wave size conversion, the 2nd layer of minimum 15x 15.The box of current size filters corresponding Gauss scale-value
Calculation formula it is as follows:Current box filtering size is indicated with N.As shown in figure 4, as a kind of specific reality
Mode is applied, constructs 4 degree 4 layers of scale space, the size variation of cassette filter is respectively 6,12,24,48 in each layer.
Step 2.2, detection characteristic point
For the scale space of building, candidate extreme point is detected by Hessian matrix, finds candidate extreme point postscript
Its positions and dimensions is recorded, image space and scale space are then found to extreme point neighborhood territory pixel interpolation using Harriss algorithm
In sub-pixel precision characteristic point, eliminate the extreme point of wherein low contrast, remaining is the position of characteristic point；
Step 2.3 determines characteristic point direction
After the position for determining characteristic point, to make characteristic point that there are the characteristics such as invariable rotary, a determination one need to be characterized
Principal direction, while robustness is improved using Haar small echo.First centered on characteristic point, by the 6 δ (scale where characterized by δ
Value) it is that radius work is justified, the Haar small echo response that the point in border circular areas is 4 δ in the x and y direction is calculated, and with Gaussian template pair
Obtained response results do the cumulative principal direction to get this feature point, more meet reality.As a kind of specific embodiment, with
Centered on characteristic point, the fan-shaped region that central angle is 60 ° be unit, traverse entire response image with 5 ° for step-length, 6 fans can be obtained
Shape region, the haar small echo response in each fan-shaped region on the direction accumulation calculating x and y generate a new vector, will be new
Principal direction of the direction of modulus value maximum vector as this feature point in vector.
Step 2.4 generates feature point description symbol
To ensure that rotational invariance rotates reference axis to SURF characteristic point principal direction, along principal direction with characteristic point it is
The heart constructs the square window that 20 δ are side length, which is uniformly divided into 16 sub-blocks, with 2 δ in each subregion
Haar small echo template with 5x5 size for 1 fritter interval sampling, calculate in each fritter test point relative to principal direction x and y
The response dx and d of axis directionyThe response and its absolute value of 25 sampled points in each fritter are added up, just formed
14 dimensional vector of the subregion describe sub- ν=(∑ dx,∑dy,∑|dx|,∑|dy|).By the processing of the above process, often
A characteristic point just obtains 1 64 dimension description vectors, can remove the influence of brightness change after feature vector is normalized, finally obtains
The vector descriptor of SURF feature.
After obtaining the vector descriptor of feature, validity feature pair is obtained using RANSAC algorithm, determines initial spell
Connect parameter.Grating map G and W are extracted and are established N group SURF feature pairWherein Pi,GAnd Pi,WIt respectively indicates
I-th group of feature of map G and W, they use P in the position of grating map respectivelyi,GAnd Pi,WIt indicates.If T={ R, t } is splicing
The approximate solution of parameter, and { Pi,G,Pi,W}I=1For one group of effective feature pair, that is, they are for the same position in two-dimensional coordinate
It sets, then | | Rpi,G+t-pi,W||2≈0.Two groups of features such as { P are randomly selected from SURF feature pairi,G,Pi,W}I=n, mIt calculates newest
Parameter T to be spliceds={ Rs,ts, wherein variable s is the number of iterations, initial value 0；Calculate di=| | Rpi,G+t-pi,W||2, root
According to the threshold value d of settingth, count di≤dthSURF feature pair group number；It repeats above step and reaches set maximum value to s, benefit
Optimal initial splicing parameter T is obtained to calculating with the N number of effective SURF feature foundbest。
Step 3 will initially splice parameter substitution mathematical model, be based on improved ICP algorithm, solve accurate rigid body translation
As a result；
Step 3.1 establishes point to relationship
(R is changed by the rigid body that preceding an iteration obtainsk-1,tk-1), establish the corresponding relationship put between point set:
In order to accelerate ICP algorithm speed, using the nearest neighbor search algorithm based on k-d tree, by giWith corresponding points
Between Euclidean distanceRetain, establishes point to relationship
Step 3.2 calculates εkAnd it updates
According to the point of foundation to relationship, overlapping percentages ε is calculatedk, and update corresponding subset
According to the Euclidean distance retainedBy ascending order mode to all-pairIt is ranked up, cumulative time
Point after going through sequence determines the corresponding point of minimum target functional value to position in the queue the target function value in simultaneously calculating formula
Set label Ck, for calculating the current optimum superposing percentage ε of grating map to be splicedk=(Ck/Ng), and use preceding CkA point is to more
New subset
Step 3.3 calculates rigid body translation result
To updated subsetNewest rigid body translation result (R is calculated using least square methodk,tk):
IfUsing above formula respectively to parameter (θ, tx,ty) seek local derviation
Number, then can get three equations, and three equations of simultaneous can be obtained current newest rigid body translation result.
Step 3.4 repeats above step, until | ξk-ξk-1| < ξ or the number of iterations k reaches the maximum times K of formulation
Then stop iteration, whereinAccurate rigid body translation knot can be obtained
Fruit (εk,Rk,tk)。
Step 4 is rotated accordingly, translated and is merged as a result, treating stitching image according to accurate rigid body translation, is obtained
Obtain the accurate splicing result of grating map；
Splicing parameter is obtained by previous step algorithm and solves rigid body translation as a result, map G can thus be carried out
Corresponding rotation and translation obtains transformed map T (G).If directly transformed map T (G) is added to map W
On, it is possible that then splicing result is inevitably wrong the case where partial occlusion.Therefore, it in order to splice the integrality of map, needs
Transformed map T (G) and W are merged.Before map fusion, a width is defined first for storing map fusion knot to be spliced
The blank map L of fruit is determined the size of map L by map T (G) and the direct stack result of W, then selected reference point, by map W
Coordinate origin align with this point, one by one traverse and determine map L in each pixel pixel value, finally can get grating map
Accurate splicing result, complete the fusion of map.
Claims (6)
1. more maps based on map feature splice blending algorithm, which comprises the steps of:
Step 1, the edge pixel point for extracting grating map to be spliced, establish rigid body translation mathematical model, by map Bonding Problem
It is converted into image registration minimization problem；
Step 2 is based on improved SURF algorithm, extracts feature from map to be spliced, determines optimal initial splicing parameter；
Step 3 will initially splice parameter substitution mathematical model, be based on improved ICP algorithm, solve accurate rigid body translation knot
Fruit；
Step 4 is rotated accordingly, translated and is merged as a result, treating stitching image according to accurate rigid body translation, and grid are obtained
The accurate splicing result of lattice map.
2. more maps according to claim 1 based on map feature splice blending algorithm, which is characterized in that in step 1,
For there are overlapping region two-dimensional grid map G and W, if being with the edge pixel point set of map W in map GWithNgAnd NwRespectively indicate the element number of edge pixel point set G and W, in the map G and overlapping region map W
Edge pixel point set is Gε, it is the subset of G, ε is overlapping percentages, and by calculating rigid body translation parameter T={ R, t }, R is one
Two Dimensional Rotating transformation matrix, t ∈ R are two-dimension translational transformation vectors, and grating map Bonding Problem is converted into image registration problem,
So that transformed, treated that point set T (G) can be matched with point set W well, such as following formula:
s.t.RTR=I2×2, Δ (R)=1
In formula, λ is control parameter, | | represent element number in set, minimum overlay percentage εminIt indicates.
3. more maps according to claim 2 based on map feature splice blending algorithm, which is characterized in that in step 2,
Improved SURF algorithm pass through respectively building scale space, detect characteristic point, determine characteristic point direction, extract SURF feature to
Amount, specifically:
Step 2.1, building scale space: the scale space pyramid bottom is filtered using the box that a size is 9x9, box
Minimum 6 pixel step lengths of size conversion, the 2nd layer of minimum 15x15 are filtered, box filters the calculation formula of corresponding Gauss scale-value
It is as follows:Wherein current box filtering size is indicated with N；
Step 2.2, detection characteristic point: for the scale space of building, candidate extreme point is detected by Hessian matrix, is found
Its positions and dimensions is recorded after candidate extreme point, using Harriss algorithm to extreme point neighborhood territory pixel interpolation, finds image space
With the sub-pixel precision characteristic point in scale space, the extreme point of wherein low contrast is eliminated, remaining is the position of characteristic point
It sets；
Step 2.3 determines characteristic point direction: centered on characteristic point, justifying by radius work of 6 δ, the point calculated in border circular areas exists
It is responded on the direction x and y for the Haar small echo of 4 δ, and obtained response results is made of Gaussian template and are added up to get this feature point
Principal direction, the scale-value where wherein δ is characterized；
Step 2.4, the vector descriptor for forming SURF feature: constructing 20 δ centered on characteristic point along principal direction is side
The region is uniformly divided into 16 sub-blocks by long square window, in each subregion with the Haar small echo template of 2 δ with
5x5 size is 1 fritter interval sampling, calculates response d of the test point relative to principal direction x and y-axis direction in each fritterx
And dy, the response and its absolute value of 25 sampled points in each fritter are added up, 14 dimension of the subregion is formed
The sub- ν of vector description=(∑ dx,∑dy,∑|dx|,∑|dy|), by the processing of the above process, each characteristic point just obtains 1
64 dimension description vectors, after feature vector is normalized, finally obtain the vector descriptor of SURF feature.
4. more maps according to claim 3 based on map feature splice blending algorithm, which is characterized in that in step 2,
Obtain initial splicing parameter method particularly includes:
Firstly, randomly selecting two groups of feature such as P from SURF feature pairi,G,Pi,W}I=n, m, calculate newest parameter T to be spliceds=
{Rs,ts, wherein variable s is the number of iterations, initial value 0；
Then, it calculatesAccording to the threshold value d of settingth, count di≤dthSURF feature pair group
Number；
Finally, repeating to utilize the N number of effective SURF feature pair found up to reaching maximum number of iterations with above-mentioned steps, count
It calculates and obtains optimal initial splicing parameter Tbest。
5. more maps according to claim 1 based on map feature splice blending algorithm, which is characterized in that in step 4,
Based on improved ICP algorithm, accurate rigid body translation is solved as a result, being specifically:
Step 3.1 changes (R by the rigid body that preceding an iteration obtainsk-1,tk-1), establish the corresponding relationship put between point set:
In order to accelerate ICP algorithm speed, using the nearest neighbor search algorithm based on k-dtree, by giWith corresponding pointsBetween
Euclidean distanceRetain, establishes point to relationship
Step 3.2, according to the point of foundation to relationship, calculate overlapping percentages εk, and update corresponding subset
According to the Euclidean distance retainedBy ascending order mode to all-pairIt is ranked up, the traversal that adds up row
Point after sequence determines that the corresponding point of minimum target functional value marks position in the queue to the target function value in simultaneously calculating formula
Number Ck, for calculating the current optimum superposing percentage ε of grating map to be splicedk=(Ck/Ng), and use preceding CkA point is sub to updating
Collection；
Step 3.3, to updated subsetFormula is calculated using least square method and calculates newest rigid body translation result (Rk,
tk):
IfUsing above formula respectively to parameter (θ, tx,ty) partial derivative is sought,
Three equations are obtained, three equations of simultaneous are up to current newest rigid body translation result；
Step 3.4 repeats step 3.1 to 3.3, until ξk-ξk-1| < ξ or the number of iterations k reach maximum times K, obtain accurate
Rigid body translation result (εk,Rk,tk), wherein
6. more maps according to claim 1 based on map feature splice blending algorithm, which is characterized in that in step 4,
Before map fusion, the blank map L that a width is used to store map fusion results to be spliced is defined first, it is straight by map T (G) and W
The size that stack result determines map L is connect, then selected reference point, the coordinate origin of map W is aligned with this point, is traversed one by one
And determine the pixel value of each pixel in map L, it finally can get the accurate splicing result of grating map, complete melting for map
It closes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811029781.1A CN109285117A (en) | 2018-09-05 | 2018-09-05 | A kind of more maps splicing blending algorithm based on map feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811029781.1A CN109285117A (en) | 2018-09-05 | 2018-09-05 | A kind of more maps splicing blending algorithm based on map feature |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109285117A true CN109285117A (en) | 2019-01-29 |
Family
ID=65184444
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811029781.1A Pending CN109285117A (en) | 2018-09-05 | 2018-09-05 | A kind of more maps splicing blending algorithm based on map feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109285117A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109839153A (en) * | 2019-01-30 | 2019-06-04 | 江苏理工学院 | A kind of computer system applied to city stratification environmental monitoring and simulation |
| CN109978767A (en) * | 2019-03-27 | 2019-07-05 | 集美大学 | The ground laser SLAM drawing method based on multirobot collaboration |
| CN110321398A (en) * | 2019-06-26 | 2019-10-11 | 浙江吉利控股集团有限公司 | A kind of dead space is parked method for building up, device and the terminal of map |
| CN110415174A (en) * | 2019-07-31 | 2019-11-05 | 达闼科技（北京）有限公司 | Map amalgamation method, electronic equipment and storage medium |
| WO2020259361A1 (en) * | 2019-06-28 | 2020-12-30 | Oppo广东移动通信有限公司 | Map update method and apparatus, and terminal and storage medium |
| CN112991187A (en) * | 2021-04-28 | 2021-06-18 | 四川大学 | Convolution twin-point network blade profile splicing system based on multiple spatial similarities |
| CN114581498A (en) * | 2022-05-05 | 2022-06-03 | 环球数科集团有限公司 | Iterative model registration system combining vector data and raster image |
| CN114699013A (en) * | 2022-03-29 | 2022-07-05 | 深圳优地科技有限公司 | Fixed-point cleaning method and device, electronic equipment and storage medium |
-
2018
- 2018-09-05 CN CN201811029781.1A patent/CN109285117A/en active Pending
Non-Patent Citations (1)
| Title |
|---|
| 唐宏伟: "基于SURF特征的多机器人栅格地图拼接方法", 《电子测量与仪器学报》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109839153A (en) * | 2019-01-30 | 2019-06-04 | 江苏理工学院 | A kind of computer system applied to city stratification environmental monitoring and simulation |
| CN109978767A (en) * | 2019-03-27 | 2019-07-05 | 集美大学 | The ground laser SLAM drawing method based on multirobot collaboration |
| CN109978767B (en) * | 2019-03-27 | 2023-09-15 | 集美大学 | Laser SLAM map method based on multi-robot cooperation |
| CN110321398A (en) * | 2019-06-26 | 2019-10-11 | 浙江吉利控股集团有限公司 | A kind of dead space is parked method for building up, device and the terminal of map |
| WO2020259361A1 (en) * | 2019-06-28 | 2020-12-30 | Oppo广东移动通信有限公司 | Map update method and apparatus, and terminal and storage medium |
| CN110415174A (en) * | 2019-07-31 | 2019-11-05 | 达闼科技（北京）有限公司 | Map amalgamation method, electronic equipment and storage medium |
| CN112991187A (en) * | 2021-04-28 | 2021-06-18 | 四川大学 | Convolution twin-point network blade profile splicing system based on multiple spatial similarities |
| CN114699013A (en) * | 2022-03-29 | 2022-07-05 | 深圳优地科技有限公司 | Fixed-point cleaning method and device, electronic equipment and storage medium |
| CN114581498A (en) * | 2022-05-05 | 2022-06-03 | 环球数科集团有限公司 | Iterative model registration system combining vector data and raster image |
| CN114581498B (en) * | 2022-05-05 | 2022-07-29 | 环球数科集团有限公司 | Iterative model registration system combining vector data and raster image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109285117A (en) | A kind of more maps splicing blending algorithm based on map feature | |
| CN108564616B (en) | Fast robust RGB-D indoor three-dimensional scene reconstruction method | |
| Ma et al. | A review of 3D reconstruction techniques in civil engineering and their applications | |
| Zheng et al. | Non-local scan consolidation for 3D urban scenes | |
| WO2021119024A1 (en) | Interior photographic documentation of architectural and industrial environments using 360 panoramic videos | |
| Dal Poz et al. | Automated extraction of road network from medium-and high-resolution images | |
| Zhao et al. | Deep lucas-kanade homography for multimodal image alignment | |
| CN104036523A (en) | Improved mean shift target tracking method based on surf features | |
| CN104050675B (en) | Feature point matching method based on triangle description | |
| CN105389774A (en) | Method and device for aligning images | |
| CN103700101A (en) | Non-rigid brain image registration method | |
| WO2022179002A1 (en) | Image matching method and apparatus, electronic device, and storage medium | |
| CN104881029A (en) | Mobile robot navigation method based on one point RANSAC and FAST algorithm | |
| CN104754182A (en) | Stationary phase method for aerial video with high resolution based on self-adaption motion filtering | |
| CN116452644A (en) | Three-dimensional point cloud registration method and device based on feature descriptors and storage medium | |
| CN112053441A (en) | Full-automatic layout recovery method for indoor fisheye image | |
| CN113192200A (en) | Method for constructing urban real scene three-dimensional model based on space-three parallel computing algorithm | |
| Seo et al. | A direct method for robust model-based 3D object tracking from a monocular RGB image | |
| Wang et al. | Terrainfusion: Real-time digital surface model reconstruction based on monocular slam | |
| Bülow et al. | Fast and robust photomapping with an unmanned aerial vehicle (uav) | |
| CN112221143A (en) | Method, device and storage medium for controlling movement of virtual object | |
| Dong et al. | Learning regional purity for instance segmentation on 3d point clouds | |
| CN110838130B (en) | Fast fuzzy distance transformation method | |
| Cárdenas-Donoso et al. | Modeling of the 3D tree skeleton using real-world data: A survey | |
| CN106709942A (en) | Panoramic image mistaken matching elimination method based on characteristic azimuth |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190129 |