CN102629330A - Rapid and high-precision matching method of depth image and color image - Google Patents
Rapid and high-precision matching method of depth image and color image Download PDFInfo
- Publication number
- CN102629330A CN102629330A CN2012100500338A CN201210050033A CN102629330A CN 102629330 A CN102629330 A CN 102629330A CN 2012100500338 A CN2012100500338 A CN 2012100500338A CN 201210050033 A CN201210050033 A CN 201210050033A CN 102629330 A CN102629330 A CN 102629330A
- Authority
- CN
- China
- Prior art keywords
- image
- depth image
- point
- coloured image
- depth
- 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
Images
Abstract
The invention discloses a rapid and high-precision matching method of a depth image and a color image. The method comprises the following steps: (1) Carrying out interframe smoothing and histogram equalization on the depth image shot by a depth camera; carrying out median filtering on the color image; (2) Carrying out multi-scale spatial Harris-Laplace corner detection on the depth image and the color image so as to obtain characteristic points; (3) Describing the detected characteristic points; (4) Using a KNN algorithm to match the characteristic points of the depth image and the color image so as to obtain candidate characteristic point matching pairs; (5) Using a RANSAC algorithm to eliminate the mismatched candidate characteristic point matching pairs so as to obtain the characteristic point matching pairs. The prior art is difficult to be used in the matching field of the depth image and an optical image. By using the method of the invention, the above problem can be solved. And robustness and accuracy can be increased.
Description
Technical field
The present invention relates to technical field of image processing, the quick high accuracy matching process of particularly a kind of depth image and coloured image.
Background technology
Along with continuing to bring out of novel sensor, the ability that people obtain image improves rapidly, and the image that sensor produced of different qualities also is on the increase.Because the view data obtained without sensor has limitation and certain defective, only through a sub-picture or one type of view data often can not be comprehensive and deep understanding, such data also are difficult to reach tests and demands of applications.For this reason, need the image synthesis that different sensors obtains be got up to use through images match and integration technology, even constitute three-dimensional model, understand and grasp thereby we are had more clearly target.
Image matching technology has been widely used in scientific research and people's production and life as one of the key of image processing techniques and basic technology.Accept widely and use and also more and more receive based on the image matching technology of unique point, wherein SIFT and Harris use representative the most widely in this class operator.Yet along with system to real-time, the requirement on robustness and the speed is more and more harsher.Especially to the object identification based on unique point, on the real time algorithm of location and tracking, traditional SIFT can not reach people's requirement on the unique point operator speed such as SURF and Harris.
And, make the depth transducer number of applications in scientific research along with the emergence of this degree of depth camera of the Kinect of Microsoft, and the degree of depth picture of taking out based on depth transducer is fully different in essence at picture with the optics picture before our.And traditional matching process basically all is the coupling that is applied to optical imagery, and the coupling of few certain applications between depth image arranged.And on the problem of depth image and two kinds of complete picture couplings of different nature of optical imagery, also seldom can find the good matching process of a kind of rate of exchange.
Summary of the invention
In order to overcome the above-mentioned deficiency of prior art, the object of the present invention is to provide the quick high accuracy matching process of a kind of depth image and coloured image.
The object of the invention is realized through following technical scheme:
The quick high accuracy matching process of a kind of depth image and coloured image may further comprise the steps:
(1) depth image of degree of depth camera being taken out carries out the level and smooth and histogram equalization of interframe; Coloured image is carried out medium filtering;
(2) respectively depth image and coloured image are carried out the Harris-Laplace Corner Detection of multiscale space, obtain unique point;
(3) detected unique point is described;
(4) with the KNN algorithm unique point of depth image and coloured image is mated, it is right to obtain candidate feature point coupling;
(5) the candidate feature point coupling with RANSAC algorithm elimination matching error is right, and it is right to obtain Feature Points Matching.
The said Harris-Laplace Corner Detection of respectively depth image and coloured image being carried out multiscale space of step (2) obtains unique point, is specially:
(2-1) definition space yardstick;
(2-2) with the gaussian kernel convolution of depth image and each yardstick, obtain the metric space of depth image; With the gaussian kernel convolution of coloured image and each yardstick, obtain the metric space of coloured image;
(2-3) on each yardstick of depth image metric space and coloured image metric space, calculate second-order matrix respectively;
(2-4) according to second moment matrix computations angle point degree value;
The local maximum point of (2-5) asking for the angle point degree value is as the candidate feature point;
(2-6) the normalization LOG value of the angle point degree value of calculated candidate unique point;
(2-7) ask for the local maximum point of normalization LOG value on dimension, i.e. the Harris angle point of metric space as unique point.
Step (3) is said to be described the unique point that detects, and is specially:
(3-1), calculate gaussian image according to the scale-value of each unique point of depth image and coloured image;
(3-2) adopt finite difference method, calculate the argument and the amplitude of the image gradient in unique point zone in the gaussian image;
(3-3) argument and the amplitude of the image gradient in use statistics with histogram unique point zone, histogrammic peak value is represented the principal direction of this unique point;
(3-4) be divided into a plurality of direction scopes, add up the image gradient strength information of each direction, obtain the SIFT eigenvector of multidimensional 0 °～360 °.
Step (4) is said matees the unique point of depth image and coloured image with the KNN algorithm, and it is right to obtain candidate feature point coupling, is specially:
At first calculate the Euclidean distance between the proper vector of two all unique points of width of cloth image, then the proper vector of proper vector and its arest neighbors is that the preliminary matching characteristic vector of confirming is right; Pick out the nearest neighbor distance and time nearest neighbor distance of proper vector; And threshold value is set; When the ratio of two distances during less than threshold value, the then preliminary matching characteristic vector of confirming is to be that candidate's matching characteristic vector is right, and characteristic of correspondence point is that candidate's matched feature points is right.
The said definition space yardstick of step (2-1) is specially: definition fixed size σ
1, σ
2... σ
n, σ wherein
n=k
nσ
0Defining integration yardstick σ
I, the variable of expression Harris angle point current scale; Definition differential yardstick σ
D, near the variable that differential value changes the expression Harris angle point; σ wherein
D=s σ
I, k, s confirm according to actual conditions.
Step (2-3) is said calculates second-order matrix on each yardstick of metric space, the formula of embodying is:
Wherein, g (σ
I) the expression yardstick is σ
IGaussian convolution nuclear, the position of ω presentation video; Image after L (ω) expression process Gauss is level and smooth, symbol
The expression convolution, L
x(ω, σ
D) and L
y(ω, σ
D) represent respectively image is used Gauss g (σ
D) function carry out level and smooth after, the result who gets its differential in x or y direction.
Step (2-4) is specially according to second moment matrix computations angle point degree value:
Angle point degree value HrF=det (μ (ω, σ
I, σ
D))-α trace
2(μ (ω, σ
I, σ
D)), α confirms that according to actual conditions det representes that determinant calculates, the straight mark of trace representing matrix, for example
Trace (X)=A+C;
The said local maximum point of normalization LOG value on dimension of asking for of step (2-7) is specially as unique point:
Make
then when normalization LOG value meets the following conditions, its corresponding candidate feature point is a unique point:
μ(ω，σ
n)＞μ(w，σ
n-1)
μ(ω，σ
n)＞μ(w，σ
n+1)
μ(ω，σ
n)＞threshold
L。
The said interframe of step (1) smoothly is that the interframe between continuous 11 frames is level and smooth.
The said medium filtering of step (1) is 7 * 7 medium filtering.
Compared with prior art, the present invention has the following advantages and technique effect:
1, through depth image and coloured image being carried out the Harris-Laplace Corner Detection of multiscale space; Obtain unique point; Accelerated image matching algorithm speed, under the situation of affined transformation and change of scale existence, having improved robustness and degree of accuracy greatly based on unique point.
2, in advance depth image is carried out the level and smooth and histogram equalization of interframe; Coloured image is carried out medium filtering; Solved the problem that prior art is difficult to be applied to depth image and optical imagery coupling field, in image co-registration, remote sensing image processing; Pattern-recognition, computer vision field has wide practical use.
Description of drawings
Fig. 1 is the process flow diagram of the quick high accuracy matching process of depth image of the present invention and coloured image.
Embodiment
Below in conjunction with embodiment and accompanying drawing, the present invention is done to specify further, but embodiment of the present invention is not limited thereto.
Embodiment
As shown in Figure 1, the quick high accuracy matching process of depth image of the present invention and coloured image may further comprise the steps:
(1) depth image of degree of depth camera being taken out carries out the level and smooth and histogram equalization of interframe between continuous 11 frames; Coloured image carried out 7 * 7 medium filtering;
(2) respectively depth image and coloured image are carried out the Harris-Laplace Corner Detection of multiscale space, obtain unique point:
(2-1) definition space yardstick: definition fixed size σ
1, σ
2... σ
n, σ wherein
n=k
nσ
0Defining integration yardstick σ
I, the variable of expression Harris angle point current scale; Definition differential yardstick σ
D, near the variable that differential value changes the expression Harris angle point; σ wherein
D=s σ
I, k=1.5; S=0.7;
(2-2) with the gaussian kernel convolution of depth image and each yardstick, obtain the metric space of depth image; With the gaussian kernel convolution of coloured image and each yardstick, obtain the metric space of coloured image;
(2-3) on each yardstick of metric space, calculate second-order matrix, the formula of embodying is:
Wherein, g (σ
I) the expression yardstick is σ
IGaussian convolution nuclear, the position of ω presentation video; Image after L (ω) expression process Gauss is level and smooth, symbol
The expression convolution, L
x(ω, σ
D) and L
y(ω, σ
D) represent respectively image is used Gauss g (σ
D) function carry out level and smooth after, the result who gets its differential in x or y direction;
(2-4), be specially according to second moment matrix computations angle point degree value:
Angle point degree value HrF=det (μ (ω, σ
I, σ
D))-α trace
2(μ (ω, σ
I, σ
D)), α=0.05, det representes that determinant calculates, the straight mark of trace representing matrix, for example
Trace (X)=A+C;
The local maximum point of (2-5) asking for the angle point degree value is as the candidate feature point;
(2-6) the normalization LOG value of the angle point degree value of calculated candidate unique point,
Make
then when normalization LOG value meets the following conditions, its corresponding candidate feature point is a unique point:
μ(ω，σ
n)＞μ(w，σ
n-1)
μ(ω，σ
n)＞μ(w，σ
n+1)
μ(ω，σ
n)＞threshold
L；
(2-7) ask for the local maximum point of normalization LOG value on dimension, i.e. the Harris angle point of metric space as unique point.
(3) detected unique point is described, is specially:
(3-1), calculate gaussian image according to the scale-value of each unique point of depth image and coloured image;
(3-2) adopt finite difference method, calculate that to calculate with the unique point in the gaussian image be the center, 3 * 1.5 σ be radius the zone image gradient argument m (x, y) with amplitude θ (x, y), concrete computing formula is following:
(3-3) argument and the amplitude of the image gradient in use statistics with histogram unique point zone, histogram will be divided into 36 posts, per 10 ° of posts; Histogrammic peak value is represented the principal direction of this unique point.
(3-4) be divided into 8 direction scopes with 0 °～360 °, each scope is 45 °, and the image gradient strength information of 8 directions of statistics owing to there are 4 * 4 sub regions, so have 4 * 4 * 8=128 data, finally forms the SIFT eigenvector of 128 dimensions.
(4) with the KNN algorithm unique point of depth image and coloured image is mated; It is right to obtain candidate feature point coupling: at first calculate the Euclidean distance between the proper vector of two all unique points of width of cloth image, then the proper vector of proper vector and its arest neighbors be that the matching characteristic vector of tentatively affirmation is right; Pick out the nearest neighbor distance and time nearest neighbor distance of proper vector; And threshold value is set; When the ratio of two distances during less than threshold value, the then preliminary matching characteristic vector of confirming is to be that candidate's matching characteristic vector is right, and characteristic of correspondence point is that candidate's matched feature points is right.
(5) the candidate feature point coupling with RANSAC algorithm elimination matching error is right; It is right to obtain Feature Points Matching: the matched feature points of choosing a part converses the transition matrix between two width of cloth images as observation data, and through translation; Rotation and amplification make two width of cloth images match; Iteration through limited number of times finally selects one group of match point that meets reality most again, thereby eliminates the mistake match point, and it is right to obtain Feature Points Matching.
The foregoing description is a preferred implementation of the present invention; But embodiment of the present invention is not limited by the examples; Other any do not deviate from change, the modification done under spirit of the present invention and the principle, substitutes, combination, simplify; All should be the substitute mode of equivalence, be included within protection scope of the present invention.
Claims (10)
1. the quick high accuracy matching process of depth image and coloured image is characterized in that, may further comprise the steps:
(1) depth image of degree of depth camera being taken out carries out the level and smooth and histogram equalization of interframe; Coloured image is carried out medium filtering;
(2) respectively depth image and coloured image are carried out the Harris-Laplace Corner Detection of multiscale space, obtain unique point;
(3) detected unique point is described;
(4) with the KNN algorithm unique point of depth image and coloured image is mated, it is right to obtain candidate feature point coupling;
(5) the candidate feature point coupling with RANSAC algorithm elimination matching error is right, and it is right to obtain Feature Points Matching.
2. the quick high accuracy matching process of depth image according to claim 1 and coloured image; It is characterized in that; The said Harris-Laplace Corner Detection of respectively depth image and coloured image being carried out multiscale space of step (2) obtains unique point, is specially:
(2-1) definition space yardstick;
(2-2) with the gaussian kernel convolution of depth image and each yardstick, obtain the metric space of depth image; With the gaussian kernel convolution of coloured image and each yardstick, obtain the metric space of coloured image;
(2-3) on each yardstick of depth image metric space and coloured image metric space, calculate second-order matrix respectively;
(2-4) according to second moment matrix computations angle point degree value;
The local maximum point of (2-5) asking for the angle point degree value is as the candidate feature point;
(2-6) the normalization LOG value of the angle point degree value of calculated candidate unique point;
(2-7) ask for the local maximum point of normalization LOG value on dimension, i.e. the Harris angle point of metric space as unique point.
3. the quick high accuracy matching process of depth image according to claim 2 and coloured image is characterized in that, step (3) is said to be described the unique point that detects, and is specially:
(3-1), calculate gaussian image according to the scale-value of each unique point of depth image and coloured image;
(3-2) adopt finite difference method, calculate the argument and the amplitude of the image gradient in unique point zone in the gaussian image;
(3-3) argument and the amplitude of the image gradient in use statistics with histogram unique point zone, histogrammic peak value is represented the principal direction of this unique point;
(3-4) be divided into a plurality of direction scopes, add up the image gradient strength information of each direction, obtain the SIFT eigenvector of multidimensional 0 °～360 °.
4. the quick high accuracy matching process of depth image according to claim 2 and coloured image is characterized in that, step (4) is said matees the unique point of depth image and coloured image with the KNN algorithm, and it is right to obtain candidate feature point coupling, is specially:
At first calculate the Euclidean distance between the proper vector of two all unique points of width of cloth image, then the proper vector of proper vector and its arest neighbors is that the preliminary matching characteristic vector of confirming is right; Pick out the nearest neighbor distance and time nearest neighbor distance of proper vector; And threshold value is set; When the ratio of two distances during less than threshold value, the then preliminary matching characteristic vector of confirming is to be that candidate's matching characteristic vector is right, and characteristic of correspondence point is that candidate's matched feature points is right.
5. the quick high accuracy matching process of depth image according to claim 1 and coloured image is characterized in that, the said definition space yardstick of step (2-1) is specially: definition fixed size σ
1, σ
2... σ
n, σ wherein
n=k
nσ
0Defining integration yardstick σ
I, the variable of expression Harris angle point current scale; Definition differential yardstick σ
D, near the variable that differential value changes the expression Harris angle point; σ wherein
D=s σ
I, k, s confirm according to actual conditions.
6. the quick high accuracy matching process of depth image according to claim 5 and coloured image is characterized in that, the said calculating second-order matrix of step (2-3), and the formula of embodying is:
Wherein, g (σ
I) the expression yardstick is σ
IGaussian convolution nuclear, the position of ω presentation video; Image after L (ω) expression process Gauss is level and smooth, symbol
The expression convolution, L
x(ω, σ
D) and L
y(ω, σ
D) represent respectively image is used Gauss g (σ
D) function carry out level and smooth after, the result who gets its differential in x or y direction.
7. the quick high accuracy matching process of depth image according to claim 6 and coloured image is characterized in that, step (2-4) is specially according to second moment matrix computations angle point degree value:
Angle point degree value HrF=det (μ (ω, σ
I, σ
D))-α trace
2(μ (ω, σ
I, σ
D)) α confirms that according to actual conditions det representes that determinant calculates the straight mark of trace representing matrix.
8. the quick high accuracy matching process of depth image according to claim 7 and coloured image is characterized in that, the said local maximum point of normalization LOG value on dimension of asking for of step (2-7) is specially as unique point:
Make
then when normalization LOG value meets the following conditions, its corresponding candidate feature point is a unique point:
μ(ω，σ
n)＞μ(w，σ
n-1)
μ(ω，σ
n)＞μ(w，σ
n+1)
μ(ω，σ
n)＞threshold
L。
9. the quick high accuracy matching process of depth image according to claim 1 and coloured image is characterized in that, the said interframe of step (1) smoothly is that the interframe between continuous 11 frames is level and smooth.
10. the quick high accuracy matching process of depth image according to claim 1 and coloured image is characterized in that, the said medium filtering of step (1) is 7 * 7 medium filtering.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012100500338A CN102629330A (en) | 2012-02-29 | 2012-02-29 | Rapid and high-precision matching method of depth image and color image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012100500338A CN102629330A (en) | 2012-02-29 | 2012-02-29 | Rapid and high-precision matching method of depth image and color image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102629330A true CN102629330A (en) | 2012-08-08 |
Family
ID=46587588
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012100500338A Pending CN102629330A (en) | 2012-02-29 | 2012-02-29 | Rapid and high-precision matching method of depth image and color image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102629330A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102915540A (en) * | 2012-10-10 | 2013-02-06 | 南京大学 | Image matching method based on improved Harris-Laplace and scale invariant feature transform (SIFT) descriptor |
| CN103164695A (en) * | 2013-02-26 | 2013-06-19 | 中国农业大学 | Fruit identification method based on multi-source image information fusion |
| CN103279315A (en) * | 2013-04-24 | 2013-09-04 | 电子科技大学 | Real-time desktop remote sharing method |
| CN103294813A (en) * | 2013-06-07 | 2013-09-11 | 北京捷成世纪科技股份有限公司 | Sensitive image search method and device |
| CN103985132A (en) * | 2014-05-30 | 2014-08-13 | 中国科学院遥感与数字地球研究所 | Mismatching point iterative detection method based on K neighbor graphs |
| CN105590108A (en) * | 2016-02-25 | 2016-05-18 | 云南大学 | Scene identification method used in noisy environment |
| CN105654423A (en) * | 2015-12-28 | 2016-06-08 | 西安电子科技大学 | Area-based remote sensing image registration method |
| CN104123532B (en) * | 2013-04-28 | 2017-05-10 | 浙江大华技术股份有限公司 | Target object detection and target object quantity confirming method and device |
| CN108335327A (en) * | 2017-01-19 | 2018-07-27 | 富士通株式会社 | Video camera Attitude estimation method and video camera attitude estimating device |
| CN108445443A (en) * | 2018-02-07 | 2018-08-24 | 武汉大学 | A kind of fingerprint point clustering method based on KNN |
| CN109712112A (en) * | 2018-11-22 | 2019-05-03 | 中北大学 | Taking photo by plane based on local feature is insulated the localization method of subgraph |
| CN110490923A (en) * | 2019-08-22 | 2019-11-22 | 江西渥泰环保科技有限公司 | The Pixel-level alignment methods of depth of field space-time 3-D image |
| CN111006676A (en) * | 2019-11-14 | 2020-04-14 | 广东博智林机器人有限公司 | Map construction method, device and system |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101777129A (en) * | 2009-11-25 | 2010-07-14 | 中国科学院自动化研究所 | Image matching method based on feature detection |
-
2012
- 2012-02-29 CN CN2012100500338A patent/CN102629330A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101777129A (en) * | 2009-11-25 | 2010-07-14 | 中国科学院自动化研究所 | Image matching method based on feature detection |
Non-Patent Citations (2)
| Title |
|---|
| 周志强等: "基于局部尺度不变特征的快速目标识别", 《光学技术》, vol. 34, no. 5, 15 September 2008 (2008-09-15), pages 742 - 745 * |
| 温文雅等: "一种基于特征点的图像匹配算法", 《2009系统仿真技术及其应用学术会议论文集》, 1 August 2009 (2009-08-01), pages 752 - 756 * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102915540A (en) * | 2012-10-10 | 2013-02-06 | 南京大学 | Image matching method based on improved Harris-Laplace and scale invariant feature transform (SIFT) descriptor |
| CN103164695A (en) * | 2013-02-26 | 2013-06-19 | 中国农业大学 | Fruit identification method based on multi-source image information fusion |
| CN103164695B (en) * | 2013-02-26 | 2016-02-17 | 中国农业大学 | A kind of fruit identification method based on multi-source image information fusion |
| CN103279315A (en) * | 2013-04-24 | 2013-09-04 | 电子科技大学 | Real-time desktop remote sharing method |
| CN104123532B (en) * | 2013-04-28 | 2017-05-10 | 浙江大华技术股份有限公司 | Target object detection and target object quantity confirming method and device |
| CN103294813A (en) * | 2013-06-07 | 2013-09-11 | 北京捷成世纪科技股份有限公司 | Sensitive image search method and device |
| CN103985132A (en) * | 2014-05-30 | 2014-08-13 | 中国科学院遥感与数字地球研究所 | Mismatching point iterative detection method based on K neighbor graphs |
| CN103985132B (en) * | 2014-05-30 | 2017-04-19 | 中国科学院遥感与数字地球研究所 | Mismatching point iterative detection method based on K neighbor graphs |
| CN105654423B (en) * | 2015-12-28 | 2019-03-26 | 西安电子科技大学 | Remote sensing image registration method based on region |
| CN105654423A (en) * | 2015-12-28 | 2016-06-08 | 西安电子科技大学 | Area-based remote sensing image registration method |
| CN105590108A (en) * | 2016-02-25 | 2016-05-18 | 云南大学 | Scene identification method used in noisy environment |
| CN105590108B (en) * | 2016-02-25 | 2018-08-17 | 云南大学 | A kind of scene recognition method under noisy environment |
| CN108335327A (en) * | 2017-01-19 | 2018-07-27 | 富士通株式会社 | Video camera Attitude estimation method and video camera attitude estimating device |
| CN108335327B (en) * | 2017-01-19 | 2021-09-24 | 富士通株式会社 | Camera attitude estimation method and camera attitude estimation device |
| CN108445443A (en) * | 2018-02-07 | 2018-08-24 | 武汉大学 | A kind of fingerprint point clustering method based on KNN |
| CN108445443B (en) * | 2018-02-07 | 2021-06-04 | 武汉大学 | KNN-based fingerprint point clustering method |
| CN109712112A (en) * | 2018-11-22 | 2019-05-03 | 中北大学 | Taking photo by plane based on local feature is insulated the localization method of subgraph |
| CN109712112B (en) * | 2018-11-22 | 2022-06-24 | 中北大学 | Aerial photography insulator image positioning method based on local features |
| CN110490923A (en) * | 2019-08-22 | 2019-11-22 | 江西渥泰环保科技有限公司 | The Pixel-level alignment methods of depth of field space-time 3-D image |
| CN111006676A (en) * | 2019-11-14 | 2020-04-14 | 广东博智林机器人有限公司 | Map construction method, device and system |
| CN111006676B (en) * | 2019-11-14 | 2021-08-31 | 广东博智林机器人有限公司 | Map construction method, device and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102629330A (en) | Rapid and high-precision matching method of depth image and color image | |
| CN103411553B (en) | The quick calibrating method of multi-linear structured light vision sensors | |
| Choi et al. | RGB-D edge detection and edge-based registration | |
| CN103020945B (en) | A kind of remote sensing image registration method of Multiple Source Sensor | |
| CN106204574B (en) | Camera pose self-calibrating method based on objective plane motion feature | |
| CN103700099B (en) | Rotation and dimension unchanged wide baseline stereo matching method | |
| CN101556692A (en) | Image mosaic method based on neighborhood Zernike pseudo-matrix of characteristic points | |
| CN102661708B (en) | High-density packaged element positioning method based on speeded up robust features (SURFs) | |
| CN101621711B (en) | Method for calibrating camera by adopting two same circles | |
| CN101383899A (en) | Video image stabilizing method for space based platform hovering | |
| CN102313536A (en) | Method for barrier perception based on airborne binocular vision | |
| CN104599258A (en) | Anisotropic characteristic descriptor based image stitching method | |
| CN105389774A (en) | Method and device for aligning images | |
| CN104121902A (en) | Implementation method of indoor robot visual odometer based on Xtion camera | |
| CN104036480A (en) | Surf algorithm based quick mismatching point eliminating method | |
| CN103679193A (en) | FREAK-based high-speed high-density packaging component rapid location method | |
| CN113160287B (en) | Complex component point cloud splicing method and system based on feature fusion | |
| CN101488224B (en) | Characteristic point matching method based on relativity measurement | |
| CN103632142A (en) | Local coordinate system feature description based image matching method | |
| CN112862768B (en) | Adaptive monocular VIO (visual image analysis) initialization method based on point-line characteristics | |
| CN109308715A (en) | A kind of optical imagery method for registering combined based on point feature and line feature | |
| CN108154066B (en) | Three-dimensional target identification method based on curvature characteristic recurrent neural network | |
| CN105513094A (en) | Stereo vision tracking method and stereo vision tracking system based on 3D Delaunay triangulation | |
| CN103700082B (en) | Image split-joint method based on dual quaterion relative orientation | |
| CN103914690A (en) | Shape matching method based on projective invariant |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120808 |