CN110223252A - A kind of depth image reparation algorithm based on compound adaptive region growth criterion - Google Patents
A kind of depth image reparation algorithm based on compound adaptive region growth criterion Download PDFInfo
- Publication number
- CN110223252A CN110223252A CN201910495178.0A CN201910495178A CN110223252A CN 110223252 A CN110223252 A CN 110223252A CN 201910495178 A CN201910495178 A CN 201910495178A CN 110223252 A CN110223252 A CN 110223252A
- Authority
- CN
- China
- Prior art keywords
- point
- pixel
- repaired
- depth
- image
- 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
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 24
- 150000001875 compounds Chemical class 0.000 title claims abstract description 24
- 238000004364 calculation method Methods 0.000 claims abstract description 13
- 230000008439 repair process Effects 0.000 claims abstract description 12
- 230000002146 bilateral effect Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 5
- 238000003384 imaging method Methods 0.000 claims description 4
- 241000208340 Araliaceae Species 0.000 claims description 3
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims description 3
- 235000003140 Panax quinquefolius Nutrition 0.000 claims description 3
- 235000008434 ginseng Nutrition 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000034 method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000007850 degeneration Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000000376 reactant Substances 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
- G06T7/85—Stereo camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20172—Image enhancement details
- G06T2207/20192—Edge enhancement; Edge preservation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention proposes a kind of depth images based on compound adaptive region growth criterion to repair algorithm, which is proposed the calculation method of confidence item and Weighting type priority, the reparation order of depth image is determined with Weighting type priority;Under the guidance of Weighting type priority, criterion is grown by compound adaptive region, the pixel more like with point feature to be repaired is found and constructs vertex neighborhood to be repaired;Under the guidance of RGB image, the estimation of Depth of complex point to be repaired is determined with the linear weighted combination of pixel in vertex neighborhood to be repaired.The algorithm can obtain higher reparation precision in the case where keeping the complete situation of boundary information.
Description
Technical field
The present invention relates to depth image reparations, using the reparation order of Weighting type priority guidance depth image, and utilize
RGB image information finds neighborhood of a point to be repaired, and the bilateral weighted by repairing pixel in vertex neighborhood obtains complex point to be repaired
Estimation of Depth repairs the pixel missing of depth image with this.
Background technique
Being constantly progressive and develop with the epoch, Digital image technology is all widely used in various fields.Its
As the important component in computer graphics and machine vision, accurate image information often determines final acquired
As a result accuracy, and usually there is the problems such as fuzzy or even damaged in most of image, image restoration technology becomes therewith to be worked as
Preceding research hotspot.The appearance of depth information further enriches the neighborhood that machine vision is related to, and depth image refers to quilt
Surveying object indicates image generated to the distance between depth camera with pixel value, which describes object in three-dimensional space
Between in position and characteristic information.Depth information is in 3 D scene rebuilding, human-computer interaction, target following, free viewpoint video wash with watercolours
The various aspects such as dye, augmented reality have vital effect.
The progress and development of optical technology make it is more convenient it is more universal using depth sensing equipment sampling depth information at
It is possible.Up to the present, depth sensing equipment substantially can be divided into passive depth according to the acquisition principle of depth information to pass
Sense and two class of active depth sensing.Active depth sensing mainly obtains required information using the transmitting-receiving of detectable signal,
Laser ranging is based on the flight time, based on technologies such as structure lights.Based on structure light and based on the sensing technology of flight time
Then the new lover in development as depth sensing technology all has portability although the measuring principle of the two is less identical
Various advantages such as height, robustness are high, image taking speed is fast, and the dramatic of scene can be obtained in real time, it is adapted to
It is such in more applications, but because of various influences such as hardware condition restriction, environment influence and measuring principle restrictions
Depth camera depth image obtained has a certain defect, includes apparent noise more, cannot complete reactant feature or
React wrong.Therefore, it probes into the degeneration of depth image and more reasonable reparation depth image becomes inevitable.
In terms of depth image repairing research, although domestic and foreign scholars propose numerous correcting strategies, such as more images
Fusion, machine learning, depth image and combination of intensity image etc..But these methods or unsuitable dynamic scene or preceding
Set that work is sufficiently complex, or because excessively high time complexity is dfficult to apply to actual scene.
Summary of the invention
Weighting type priority is introduced into neighborhood and repaired in algorithm by the present invention, real as the criterion for determining image repair order
The correct extension of image texture is showed.In the case where correctly repairing order, propose that comprehensively consider pixel similarity similar with texture
Property compound adaptive region grow criterion, with obtain with the higher vertex neighborhood to be repaired of points correspondence degree to be repaired.Pass through
The bilateral weighted of pixel obtains the depth value estimated value of complex point to be repaired in vertex neighborhood to be repaired.
Innovative point of the invention first is that the proposition of Weighting type priority calculation method.Consider Criminisi priority
Deficiency, to improve the robustness that priority calculates, new confidence item and priority calculation method are as follows:
(1) the confidence item of p point indicates that calculation method is as follows with C (p):
In formula, ψpFor p neighborhood of a point, q is neighborhood ψpIn non-empty pixel, Φ is non-empty pixel in depth image
The set of point, (xp,yp), (xq,yq) it is respectively p, the pixel coordinate of q two o'clock.
(2) mentioned confidence item in (1) is utilized, the calculation expression of priority P (p) is as follows:
P (p)=β C (p)+γ D (p)
In formula, D (p) is the data item of p point, and β and γ are the coefficient of balance of confidence item and data item.
Another innovation of the invention is a kind of building of new vertex neighborhood to be repaired.Utilize pixel value tag and office
The constraint of portion's textural characteristics, compound adaptive region growth criterion are as follows:
In formula, m is corresponding pixel after p point is mapped in RGB image, and as initial seed point, Ω is the neighbour of m point
The set put in domain, τ are a threshold value, dist, Х(m,n)It is shown below:
In formula, R, G, B are the triple channel of RGB image,WithRespectively represent the pixel value of m point and the channel n point c, (xn,
yn) be n point pixel coordinate, (xm,ym) be m point pixel coordinate, κ and λ are weight.
In formula, ULBPmAnd ULBPnFor the uniform LBP code of m, n two o'clock.
The utility model has the advantages that priority calculation method proposed by the present invention, improves the shortcoming of Criminisi priority,
So that the reparation order of image is more reasonable, the correct propagation of image texture characteristic in repair process ensure that.And the present invention mentions
The construction method of the adaptive neighborhood based on region growing out, can be improved the reparation precision of image, and inhibit object edge
The blurring for repairing result, guarantees the accuracy of image feature information.
Detailed description of the invention
Fig. 1: the depth image based on compound adaptive region growth criterion repairs algorithm structure block diagram
Fig. 2: the depth image based on compound adaptive region growth criterion repairs algorithm flow chart
Specific embodiment
1 to attached drawing 2 with reference to the accompanying drawing, and the invention will be further described.
With reference to attached drawing 1, the present invention is firstly the need of obtaining depth image and corresponding the RGB image after being registrated, to depth
All empty pixels, which do priority and calculate to obtain, in degree image has top-priority pixel as complex point to be repaired, passes through
Compound adaptive region growth criterion obtains neighborhood of a point to be repaired, is estimated with repairing the bilateral weighted of pixel in vertex neighborhood
Complex point depth value to be repaired, final updating hole region repeat step to repair complete.
The algorithm includes the following steps:
S1: calibration RGB-D camera obtains internal reference and the outer ginseng of RGB camera and depth camera, complete using pinhole imaging system principle
At the registration of depth image and RGB image;
S2: traversal depth image finds all empty pixels, obtains priority by Weighting type priority calculation method
Highest cavity pixel obtains initial seed point in RGB image after complex point to be repaired to be mapped to registration as complex point to be repaired;
S3: criterion is grown using compound adaptive region, indexes the pixel in RGB image in initial seed neighborhood of a point
The pixel for meeting compound adaptive region growth criterion is included into a set by point, by the compound mapping into depth image,
Obtain neighborhood of a point to be repaired.
S4: complex point to be repaired in depth image and neighborhood of a point to be repaired are mapped in RGB image together, obtain corresponding picture
The codomain weight of vegetarian refreshments obtains the estimating depth value of complex point to be repaired with the bilateral weighted of codomain and spatial domain;
S5: repeating step S2-S4, until all empty pixels be repaired it is complete.
With reference to attached drawing 2, the depth image based on compound adaptive region growth criterion repairs algorithm, and specific step is as follows:
Step S1 specifically comprises the following steps:
S11: it demarcates to obtain the internal reference K of depth camera and RGB camera by Zhang Shi1、K2With outer ginseng R1、R2(camera coordinates arrive
The spin matrix of world coordinates experience), T1、T2(translation vector that camera coordinates to world coordinates are undergone).
S12: the mapping relations between depth camera coordinate are obtained according to pinhole imaging system principle:
In formula, P1、P2For same point under world coordinates in two camera coordinates corresponding point.
S13: according to camera imaging model, pass through the mapping between the available pixel coordinate of camera internal reference and camera coordinates
Relationship:
In formula, p1、p2For this o'clock in two pixel coordinates corresponding point.
S14: the mapping relations between depth image and RGB image can be obtained by formula (1) (2):
Step S2 specifically comprises the following steps:
S21: traversal depth image finds all pixels point that pixel value is 0.
S22: by the RGB image gray processing after registration and seeking gradient information, calculates empty pixel using gradient information
Data item calculates the priority of empty pixel using designed confidence item and Weighting type priority calculation method.It finds excellent
Maximum cavity pixel p is first weighed as complex point to be repaired.Designed confidence item C (p) and Weighting type priority P (p) calculation method
It is as follows:
In formula, ψpFor p neighborhood of a point, q is neighborhood ψpIn non-empty pixel, Φ is non-empty pixel in depth image
The set of point, (xp,yp), (xq,yq) it is respectively p, the pixel coordinate of q two o'clock.
P (p)=β C (p)+γ D (p) (5)
In formula, D (p) is the data item of p point, and β and γ are the coefficient of balance of confidence item and data item.
In formula,Indicate p point isophote direction, npIndicate p point normal vector, α is normalization factor, for 8-bit amount
α=255 for the digital picture of change.
S23: in the RGB image after p point to be mapped to registration, using corresponding pixel as initial seed point.
Step S3 specifically comprises the following steps:
S31: doing region growing to initial seed point, in the case where compound adaptive region grows criterion to initial seed vertex neighborhood
Pixel makes a decision, and the pixel for meeting criterion incorporates into a set.Compound adaptive region growth criterion is as follows:
In formula, m is corresponding pixel after p point is mapped in RGB image, and as initial seed point, Ω is the neighbour of m point
The set put in domain, τ are a threshold value, dist, Х(m,n)It is shown below:
In formula, R, G, B are the triple channel of RGB image,WithRespectively represent the pixel value of m point and the channel n point c, (xn,
yn) be n point pixel coordinate, (xm,ym) be m point pixel coordinate, κ and λ are weight.
In formula, ULBPmAnd ULBPnFor the uniform LBP code of m, n two o'clock.
S32: using the pixel for meeting criterion in step S31 as new seed point, continue in new seed neighborhood of a point
It finds the pixel for meeting compound adaptive region growth criterion and is included into identity set, until meeting termination condition.It terminates
Condition is described as follows:
1) when being grown again using the pixel obtained for meeting growth criterion as seed point, in seed vertex neighborhood
It inside can not find the pixel that can satisfy criterion.
2) the pixel number for meeting growth criterion has reached a upper limit ρ.
S33: the region growing pixel collection obtained for meeting adaptive region growth criterion is mapped to depth map
As in, neighborhood of a point to be repaired is obtained.
Step S4 specifically comprises the following steps:
S41: by RGB image information, the codomain weight W of pixel in vertex neighborhood to be repaired is obtainedrAnd spatial domain weight
Wd, as shown in formula 10,11.
In formula, I is the pixel value of corresponding pixel points in RGB image, and (x, y) is the pixel coordinate of p point, (i, j) is that p point is adjacent
The pixel coordinate of pixel, σ in domainrSize control pixel because of the size with p point value differences cause codomain weight
Change number.
In formula, σdSize control be used to estimate the size of neighborhood of a point to be repaired.
S42: according to the weight of pixel in the available vertex neighborhood to be repaired of formula 10,11:
W=Wr·Wd (12)
Linear weighted function is done to pixel in vertex neighborhood to be repaired and obtains the estimation of Depth value D of complex point to be repairedp':
In formula, Dp、DqFor the depth pixel value of pixel in complex point p to be repaired and p vertex neighborhood.
Experimental result and analysis:
To verify the performance that the depth image of the present invention for growing criterion based on compound adaptive region repairs algorithm, adopt
Use the depth image data collection in Middlebury as identifying object, with root-mean-square error (RMSE) and Y-PSNR
(PSRN) it is bilateral to compare reparation algorithm based on Navier_Stokes equation, joint for the standard that result superiority and inferiority is repaired as evaluation
Filtering algorithm and algorithm of the present invention, Art etc. are the different scenes image in data set, and the results are shown in Table 1 for reparation.
For multiple depth scenes, the present invention all has preferably RMSE and PSRN, and algorithm has robustness.
The different algorithm reparation results of repairing of table 1 compare
It is verified through emulation experiment, the present invention achieves ideal repairing effect in depth image reparation.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
Specific implementation of the invention is only limited to these instructions.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, several simple deduction or replace can also be made, all shall be regarded as belonging to of the invention
Protection scope.
Claims (3)
1. a kind of depth image based on compound adaptive region growth criterion repairs algorithm, it is characterised in that: the algorithm includes
Following steps:
S1: calibration RGB-D camera obtains internal reference and the outer ginseng of RGB camera and depth camera, is completed using pinhole imaging system principle deep
Spend the registration of image and RGB image;
S2: traversal depth image finds all empty pixels, obtains priority highest by Weighting type priority calculation method
Empty pixel as complex point to be repaired, obtain initial seed point in the RGB image after complex point to be repaired to be mapped to registration;
S3: growing criterion using compound adaptive region, index the pixel in RGB image in initial seed neighborhood of a point, will
The pixel for meeting compound adaptive region growth criterion is included into a set, by the compound mapping into depth image, is obtained
Neighborhood of a point to be repaired.
S4: complex point to be repaired in depth image and neighborhood of a point to be repaired are mapped in RGB image together, obtain corresponding pixel points
Codomain weight, the estimating depth value of complex point to be repaired is obtained with the bilateral weighted of codomain and spatial domain;
S5: repeating step S2-S4, until all empty pixels be repaired it is complete.
2. a kind of depth image based on compound adaptive region growth criterion according to claim 1 repairs algorithm,
It is characterized in that: Weighting type priority calculation method in the step S2:
(1) complex point to be repaired in depth image, is indicated with p, and the confidence item of p point indicates that calculation method is as follows with C (p):
In formula, ψpFor p neighborhood of a point, q is neighborhood ψpIn non-empty pixel, Φ is non-empty pixel in depth image
Set, (xp,yp), (xq,yq) it is respectively p, the pixel coordinate of q two o'clock.
(2) mentioned confidence item C (p) in (1) is utilized, the calculation expression of priority P (p) is as follows:
P (p)=β C (p)+γ D (p)
In formula, D (p) is the data item of p point, and β and γ are the coefficient of balance of confidence item and data item.
3. a kind of depth image based on compound adaptive region growth criterion according to claim 1 repairs algorithm,
It is characterized in that: compound adaptive region growth criterion, expression formula in the step S3 are as follows:
In formula, m is corresponding pixel after p point is mapped in RGB image, and as initial seed point, Ω is in m neighborhood of a point
The set of point, τ are a threshold value, dist, Х(m,n)It is shown below:
In formula, R, G, B are the triple channel of RGB image,WithRespectively represent the pixel value of m point and the channel n point c, (xn,yn) it is n
The pixel coordinate of point, (xm,ym) be m point pixel coordinate, κ and λ are weight.
In formula, ULBPmAnd ULBPnFor the uniform LBP code of m, n two o'clock.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910495178.0A CN110223252B (en) | 2019-06-10 | 2019-06-10 | Depth image restoration algorithm based on composite self-adaptive region growing criterion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910495178.0A CN110223252B (en) | 2019-06-10 | 2019-06-10 | Depth image restoration algorithm based on composite self-adaptive region growing criterion |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110223252A true CN110223252A (en) | 2019-09-10 |
CN110223252B CN110223252B (en) | 2022-12-06 |
Family
ID=67815923
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910495178.0A Active CN110223252B (en) | 2019-06-10 | 2019-06-10 | Depth image restoration algorithm based on composite self-adaptive region growing criterion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110223252B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114066779A (en) * | 2022-01-13 | 2022-02-18 | 杭州蓝芯科技有限公司 | Depth map filtering method and device, electronic equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100278426A1 (en) * | 2007-12-14 | 2010-11-04 | Robinson Piramuthu | Systems and methods for rule-based segmentation for objects with full or partial frontal view in color images |
US20120313937A1 (en) * | 2010-01-18 | 2012-12-13 | Disney Enterprises, Inc. | Coupled reconstruction of hair and skin |
CN103996174A (en) * | 2014-05-12 | 2014-08-20 | 上海大学 | Method for performing hole repair on Kinect depth images |
CN107316283A (en) * | 2017-07-14 | 2017-11-03 | 天津大学 | Algorithm is repaired based on the digital picture that multiwindow is merged |
-
2019
- 2019-06-10 CN CN201910495178.0A patent/CN110223252B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100278426A1 (en) * | 2007-12-14 | 2010-11-04 | Robinson Piramuthu | Systems and methods for rule-based segmentation for objects with full or partial frontal view in color images |
US20120313937A1 (en) * | 2010-01-18 | 2012-12-13 | Disney Enterprises, Inc. | Coupled reconstruction of hair and skin |
CN103996174A (en) * | 2014-05-12 | 2014-08-20 | 上海大学 | Method for performing hole repair on Kinect depth images |
CN107316283A (en) * | 2017-07-14 | 2017-11-03 | 天津大学 | Algorithm is repaired based on the digital picture that multiwindow is merged |
Non-Patent Citations (3)
Title |
---|
YING LIU等: "A Novel Exemplar-based Image Inpainting Algorithm", 《2015 INTERNATIONAL CONFERENCE ON INTELLIGENT NETWORKING AND COLLABORATIVE SYSTEMS》 * |
兰小丽等: "基于纹理块与梯度特征的图像修复改进算法", 《计算机工程与应用》 * |
吴亚娟等: "图像修复算法中匹配点搜索和修复顺序研究", 《西华师范大学学报(自然科学版)》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114066779A (en) * | 2022-01-13 | 2022-02-18 | 杭州蓝芯科技有限公司 | Depth map filtering method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110223252B (en) | 2022-12-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN1871622B (en) | Image collation system and image collation method | |
CN114782691B (en) | Robot target identification and motion detection method based on deep learning, storage medium and equipment | |
US11830216B2 (en) | Information processing apparatus, information processing method, and storage medium | |
KR100831187B1 (en) | Object posture estimation/correlation system using weight information | |
CN109974743B (en) | Visual odometer based on GMS feature matching and sliding window pose graph optimization | |
CN108010081B (en) | RGB-D visual odometer method based on Census transformation and local graph optimization | |
CN111723691B (en) | Three-dimensional face recognition method and device, electronic equipment and storage medium | |
CN111462207A (en) | RGB-D simultaneous positioning and map creation method integrating direct method and feature method | |
CN110766024B (en) | Deep learning-based visual odometer feature point extraction method and visual odometer | |
CN107563323A (en) | A kind of video human face characteristic point positioning method | |
JP7479324B2 (en) | Information processing device, information processing method, and program | |
CN114677323A (en) | Semantic vision SLAM positioning method based on target detection in indoor dynamic scene | |
CN111998862A (en) | Dense binocular SLAM method based on BNN | |
CN110428461B (en) | Monocular SLAM method and device combined with deep learning | |
CN114612393A (en) | Monocular vision-based reflective part pose estimation method | |
CN110223252A (en) | A kind of depth image reparation algorithm based on compound adaptive region growth criterion | |
CN105339981B (en) | Method for using one group of primitive registration data | |
KR101766823B1 (en) | Robust visual odometry system and method to irregular illumination changes | |
CN113642397A (en) | Object length measuring method based on mobile phone video | |
CN113720323B (en) | Monocular vision inertial navigation SLAM method and device based on point-line feature fusion | |
CN109377524A (en) | A kind of single image depth recovery method and system | |
CN114972937A (en) | Feature point detection and descriptor generation method based on deep learning | |
CN110211239B (en) | Augmented reality method, apparatus, device and medium based on label-free recognition | |
Liu et al. | Visual odometry algorithm based on deep learning | |
CN113256711A (en) | Pose estimation method and system of monocular camera |
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 |