CN114358166A - Multi-target positioning method based on self-adaptive k-means clustering - Google Patents
Multi-target positioning method based on self-adaptive k-means clustering Download PDFInfo
- Publication number
- CN114358166A CN114358166A CN202111634014.5A CN202111634014A CN114358166A CN 114358166 A CN114358166 A CN 114358166A CN 202111634014 A CN202111634014 A CN 202111634014A CN 114358166 A CN114358166 A CN 114358166A
- Authority
- CN
- China
- Prior art keywords
- matching
- points
- clustering
- characteristic
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000003064 k means clustering Methods 0.000 title claims abstract description 17
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 35
- 230000009466 transformation Effects 0.000 claims abstract description 10
- 238000005286 illumination Methods 0.000 claims abstract description 7
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims 6
- 230000008569 process Effects 0.000 description 4
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 241000282320 Panthera leo Species 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
Images
Abstract
The invention provides a multi-target positioning method based on self-adaptive k-means clustering, which combines a density peak value clustering algorithm (DPC) and a k-means clustering algorithm to provide a self-adaptive k-means clustering algorithm, adaptively determines the number of targets to be positioned based on extracted feature points, and clusters feature point sets of different targets; coarse matching is carried out through a nearest neighbor ratio algorithm, and fine matching is carried out by constructing optimal geometric constraint through feature point voting, so that multi-target precise positioning is realized. The method can accurately position the targets to be positioned in different types and quantities under the complex environments of rotation, scale transformation, partial shielding, illumination transformation and the like, and has better robustness.
Description
Technical Field
The invention relates to the technical field of computer vision and image processing, in particular to a multi-target positioning method based on self-adaptive k-means clustering.
Background
Image matching is an important content of computer vision and pattern recognition, and is widely applied in the fields of image registration, image splicing, three-dimensional reconstruction and the like. Currently, image matching algorithms are mainly classified into two categories, namely, gray level matching and feature matching. The matching algorithm based on the gray level is used for matching through the region attributes in the image sampling window, the matching precision is high, the matching algorithm is easily influenced by the environment, and the matching algorithm is sensitive to the gray level change of the image; the method is based on a feature matching algorithm, the stable features in the image are detected, the neighborhood pixel information is used for feature description, and matching is completed according to the similarity of the calculated feature descriptors. Therefore, the scholars propose many excellent feature-based matching algorithms. Lowe proposes an SIFT algorithm, which can adapt to rotation and scale scaling transformation and is insensitive to illumination change, but because the image has a texture similar region, only neighborhood information is used as a descriptor to easily cause mismatching, and the application of the algorithm is limited to a certain extent. Kyork et al combine k-means with SIFT algorithm, and cluster the feature vector matrix by using the k-means algorithm, thereby improving the image matching rate; zhalihong et al first perform logarithmic transformation on the image, and then perform iterative matching on the image features to achieve multi-target matching of sonar images; in the matching process, the Dong brocade and the like only use the characteristic points of the single-layer Gaussian pyramid to carry out rough matching, and simultaneously combine the GMS algorithm and the RANSAC algorithm to calculate an affine transformation matrix, so that the real-time property of image matching is improved; wang stile et al propose that geometric constraint is combined with SIFT algorithm to match the image, and epipolar constraint is added in the matching stage, so that the affine invariance is good; liyunhong et al have advantages in matching time and precision by constructing a function model by iterative least square fitting to eliminate mismatching points.
Although the improved SIFT algorithm improves the matching efficiency, in the multi-target positioning process, the SIFT algorithm cannot accurately position all targets, the number of the targets to be positioned must be determined, and different target feature points are clustered.
Disclosure of Invention
Aiming at the problems, the invention provides a multi-target positioning algorithm based on self-adaptive k-means clustering, which comprises the following specific steps:
1. detecting a feature point set of an image to be matched and calculating a main direction of the feature point
(1) Convolving the image with Gaussian functions with different scale factors to construct a Gaussian scale space, and utilizing image subtraction in adjacent scale spaces to construct a Gaussian difference scale space;
(2) calculating local extreme points of a Gaussian difference scale space as characteristic points;
(3) calculating the gradient size and direction of pixels in a neighborhood with the feature point as the center, and counting the gradient size and direction of all the pixels to generate a gradient direction histogram, wherein the direction of the maximum value of the histogram is the main direction of the feature point;
(4) and dividing the sampling window into a plurality of sub-regions, counting gradient information in the sub-regions, constructing a characteristic vector, and normalizing the characteristic vector in order to further reduce the influence of illumination change.
2. The number of clusters and the initial cluster center are determined based on a density peak clustering algorithm (DPC).
(1) Calculating the truncation distance of the characteristic points and the local density and distance of each characteristic point;
(2) and calculating the gamma value of each characteristic point, performing descending arrangement on the gamma values, and selecting the first k data as an initial clustering center. The formula for γ is as follows:
γi＝ρi*δi
in the above formula rhoiIs the local density, deltaiIs a distance.
3. And performing iterative clustering on the clustering center by using a k-means clustering algorithm to obtain a final clustering result. Calculating the distances from all the feature points to the clustering centers, classifying the data to be classified according to the nearest principle, calculating the mean value of coordinates of all the feature points in each cluster after classifying all the feature points, taking the coordinate mean value as a new clustering center, continuously iterating and determining to finally obtain k clustering centers, and clustering the remaining feature points to the cluster where the feature points with the nearest distance and the local density higher than that of the remaining feature points are located.
4. Carrying out coarse matching on the feature points by using a nearest neighbor ratio algorithm, and carrying out coarse elimination on the feature points;
searching characteristic points which are closest to and next to characteristic points of the template image in a characteristic point set of the image to be matched, and calculating the closest Euclidean distance disntAnd a second nearest euclidean distance dissntIf the ratio of (A) to (B) satisfies:
the matching of the characteristic point is successful, otherwise, the characteristic point pair is removed, and T is a set threshold value.
5. And constructing optimal geometric constraint by using feature point voting to perform fine matching.
And (3) according to the descending order of the matching confidence degrees of the feature point pairs in the rough matching, randomly selecting 3 pairs of feature points from the previous n pairs of matching feature points through iteration, and selecting 3 pairs of feature point pairs with the best fitness to establish a local coordinate system. Based on the constructed local coordinate system, constructing a straight line element from the feature points and the coordinate origin, expressing the linear element by vector coordinates, calculating the coordinate similarity of the vector coordinates in the local coordinate system, comparing the similarity with a set threshold value, and eliminating mismatching points.
(1) The method for constructing the middle local coordinate system comprises the following steps:
selecting a first n pairs of feature point forming set R { (M) based on feature point matching confidence degree descending order in rough matchingi,Ni) 1, …, n, where M isiAnd NiSelecting 3 characteristic points from the set R to select 3 characteristic points from the set R to construct a local coordinate system, voting and scoring by utilizing the similarity of the rest characteristic points in the set in the local coordinate system, and finally selecting the 3 characteristic points with the highest score to construct the local coordinate system.
The feature point similarity calculation formula is as follows:
wherein { (L)ji,Lji') | i ═ 1,2,3} is the matching point pair MjAnd NjEuclidean distances to 3 matching feature points.
(2) The calculation method of the coordinate similarity of the characteristic point straight line elements is as follows:
the linear primitive vector is expressed by coordinates to form a coordinate set: q ═ pi(x,y),qi(x,y)|i＝1,…,N}
The linear primitive vector is represented by a calculation formula in coordinates:
wherein k is the slope of a straight line, (x)1,y1) Is the coordinates of the start point of the straight line element, (x)2,y2) Is the endpoint coordinate.
Secondly, a coordinate conversion formula for converting the vector coordinates into a local coordinate system is as follows:
where α, β are the coordinates of the vectorized straight-line primitive in the local coordinate system.
Computing coordinate similarity of straight line element
Wherein, PMAnd PNLocal coordinates of the straight line primitive.
6. And calculating a transformation model by using the correct matching points to realize the precise matching of multiple targets.
The invention has the following advantages: (1) the DPC algorithm is combined with the k-means, the clustering number and the clustering center are determined through the DPC algorithm, then the k-means algorithm is used for clustering the initial feature points, and the feature point set can be effectively separated. (2) The algorithm of the invention realizes the accurate and fast matching of multiple targets, can still realize the fast matching of the targets when the targets to be matched are subjected to rotation transformation and scale transformation and are influenced by illumination, shielding and other factors, has scale and rotation invariance, has a matching effect which is not influenced by shielding and illumination change, and has higher stability.
Drawings
FIG. 1 is a flow chart of the algorithm.
FIG. 2 is a schematic diagram of local coordinate calculation of a straight line element.
FIG. 3 is a drawing showing the matching result of the strip-shaped workpiece.
FIG. 4 is a diagram of the matching result of the L-shaped workpiece.
Detailed Description
1. Inputting an image to be matched, acquiring characteristic points of the image to be matched and calculating the main direction of the characteristic points
(1) Convolving an image to be matched with Gaussian functions with different scale factors to construct a Gaussian scale space, and in order to improve the stability and accuracy of extreme point detection, in the Gaussian scale space, utilizing image subtraction in adjacent scale spaces to construct a Gaussian difference scale space;
(2) and calculating local extreme points of the Gaussian difference scale space as characteristic points. Comparing the point to be detected with 26 pixels of 8 pixels of the neighborhood with the same scale and 9 multiplied by 2 pixels of the neighborhood at the corresponding position of the adjacent scale, if the gray value of the pixel to be detected is completely smaller or larger than the gray values of 26 pixel points of the neighborhood, the pixel is a local extreme point of a Gaussian scale space, and the local extreme point set is a feature point set
(3) In order to enable the descriptor to have rotation invariance, a reference direction needs to be allocated to the feature point, the gradient size and direction of pixels in a neighborhood taking the feature point as a center are calculated, the gradient size and direction of all the pixels are counted, a gradient direction histogram is generated, and the direction of the maximum value of the histogram is the main direction of the feature point;
(4) since the coordinate axis does not coincide with the principal direction, the sampling window needs to be rotated to the principal direction of the feature point. Dividing a sampling window into 4 multiplied by 4 sub-regions, counting gradient information of 8 directions in the sub-regions, constructing a 128-dimensional feature vector, and normalizing the feature vector in order to further reduce the influence of illumination change.
2. And determining the clustering number K and the initial clustering center based on a density peak value clustering algorithm (DPC).
In the multi-target matching process, the feature points of different targets in the feature point set must be clustered, so as to avoid the interference of other target feature points in the matching process. The DPC algorithm can determine the clustering centers and the clustering number by using the density peak value in the sample data, and solves the problems of selecting the initial clustering centers and the clustering number. The specific method comprises the following steps:
(1) calculating a feature point set of an image to be matched
And the local density p of each feature pointiAnd a distance deltai；
Any one feature point has two indexes: local density ρiAnd a distance deltaiIn order to determine the cluster center, two influencing factors of local density and distance are considered simultaneously.
(2) And calculating the gamma value of each characteristic point, wherein the calculation formula of gamma is as follows:
γi＝ρi*δi
the larger the gamma value is, the higher the probability that the characteristic point becomes a clustering center is, so that the characteristic points are sorted in a descending order, and the first k data are selected as initial clustering centers.
3. Iterative clustering is carried out on the clustering center by using a k-means clustering algorithm to obtain a final clustering result, and a characteristic point set is effectively separated;
based on the initial clustering center, calculating the distances from all the feature points to the clustering center, classifying the feature points to be classified according to the nearest principle, and classifying all the feature pointsThen, calculating the mean value of the coordinates of the feature points of each category, taking the mean value as a new clustering center, and continuously iterating and determining to finally obtain k clustering centers
Wherein
And clustering the residual feature points to the cluster which is closest to the residual feature points and has local density higher than that of the cluster in which the residual feature points are located.
4. Carrying out coarse matching on the feature points by using a nearest neighbor ratio algorithm, and carrying out coarse elimination on the feature points;
searching characteristic points which are closest to and next to characteristic points of the template image in a characteristic point set of the image to be matched, and calculating the closest Euclidean distance disntAnd a second nearest euclidean distance dissntIf the ratio of (A) to (B) satisfies:
the matching of the characteristic point is successful, otherwise, the characteristic point pair is removed, and T is a set threshold value.
5. And constructing optimal geometric constraint by using feature point voting to perform fine matching.
And (3) according to the descending order of the matching confidence degrees of the feature point pairs in the rough matching, randomly selecting 3 pairs of feature points from the previous n pairs of matching feature points through iteration, and selecting 3 pairs of feature point pairs with the best fitness to establish a local coordinate system. Based on the constructed local coordinate system, constructing a straight line element from the feature points and the coordinate origin, expressing the linear element by vector coordinates, calculating the coordinate similarity of the vector coordinates in the local coordinate system, comparing the similarity with a set threshold value, and eliminating mismatching points.
(1) Constructing a local coordinate system
Selecting a first n pairs of feature point forming set R { (M) based on feature point matching confidence degree descending order in rough matchingi,Ni) 1, …, n, where M isiAnd NiIs a pair of matching feature points, selected from the set RAnd constructing a local coordinate system by using the 3 characteristic points, performing voting scoring by using the similarity of the rest characteristic points in the set in the local coordinate system, and finally selecting the 3 characteristic points with the highest score to construct the local coordinate system.
The feature point similarity calculation formula is as follows:
wherein { (L)ji,Lji') | i ═ 1,2,3} is the matching point pair MjAnd NjEuclidean distances to 3 matching feature points.
(2) Calculating the vectorized coordinates of the straight line elements of the feature points
Suppose that the extracted 3 feature point pairs are { (M)j,Nj) 1,2,3, and M is given to each of the two images1And N1As origin of coordinates, connecting M1M2,M1M3And N1N2,N1N3Respectively establishing coordinate systems O-XY and O-X 'Y' by using linear elements
Vectorized coordinates are used as base information P of a coordinate system O-XYA(Pax,Pay) And PB(Pbx,Pby) The basis information of the coordinate system O-X 'Y' is calculated in the same way, as shown in FIG. 2.
The linear primitive vector is expressed in coordinates, forming a set of coordinates: q ═ pi(x,y),qi(x,y)|i＝1,…,N}
The linear primitive vector is represented by a calculation formula in coordinates:
wherein k is the slope of a straight line, (x)1,y1) Is the coordinates of the start point of the straight line element, (x)2,y2) Is the endpoint coordinate.
(3) Similarity calculation
To calculate the similarity of the straight line primitive, the vector coordinates are converted into coordinates in the local coordinate system, and the conversion formula is as follows:
where α, β are the coordinates of the vectorized straight-line primitive in the local coordinate system.
Calculating the coordinate similarity of the straight line elements, wherein the formula is as follows:
wherein, PMAnd PNLocal coordinates of the straight line primitive.
By comparing the similarity S with a set threshold tsComparing if S is less than tsThe matching of the characteristic point pairs is correct; otherwise, the point is a mismatching point; determined by multiple matching experiments, when t issWhen the value is 0.03, the matching effect is optimal.
6. And calculating a transformation model by using the correct matching points to realize the precise matching of multiple targets.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.
Furthermore, it should be understood that although the present description refers to embodiments, not every embodiment may contain only a single embodiment, and such description is for clarity only, and those skilled in the art should integrate the description, and the embodiments may be combined as appropriate to form other embodiments understood by those skilled in the art.
Claims (7)
1. A multi-target positioning method based on self-adaptive k-means clustering is characterized in that: the method comprises the following steps:
s1, detecting a feature point set of the image to be matched and calculating the main direction of the feature points;
s2, determining the clustering number and the initial clustering center based on the density peak clustering algorithm;
and S3, performing iterative clustering on the clustering centers by using a k-means clustering algorithm to obtain a final clustering result.
S4, carrying out coarse matching on the feature points by using a nearest neighbor ratio algorithm, and carrying out coarse elimination on the feature points;
s5, constructing optimal geometric constraint by using feature point voting to perform fine matching;
and S6, calculating a transformation model by using the correct matching points to realize the accurate matching of multiple targets.
2. The multi-target positioning method based on the adaptive k-means clustering as claimed in claim 1, characterized in that: s1 includes:
(1) convolving the image with Gaussian functions with different scale factors to construct a Gaussian scale space, and utilizing image subtraction in adjacent scale spaces to construct a Gaussian difference scale space;
(2) calculating local extreme points of a Gaussian difference scale space as characteristic points;
(3) calculating the gradient size and direction of pixels in a neighborhood with the feature point as the center, and counting the gradient size and direction of all the pixels to generate a gradient direction histogram, wherein the direction of the maximum value of the histogram is the main direction of the feature point;
(4) and dividing the sampling window into a plurality of sub-regions, counting gradient information in the sub-regions, constructing a characteristic vector, and normalizing the characteristic vector in order to further reduce the influence of illumination change.
3. The multi-target positioning method based on the adaptive k-means clustering as claimed in claim 1, characterized in that: s2 includes:
(1) calculating the truncation distance of the characteristic points and the local density and distance of each characteristic point;
(2) and calculating the gamma value of each characteristic point, performing descending arrangement on the gamma values, and selecting the first k data as an initial clustering center. The formula for γ is as follows:
γi＝ρi*δi
in the above formula rhoiIs the local density, deltaiIs a distance.
4. The multi-target positioning method based on the adaptive k-means clustering as claimed in claim 1, characterized in that: in the step S3, the first step,
calculating the distances from all the feature points to the clustering centers, classifying the data to be classified according to the nearest principle, calculating the mean value of coordinates of all the feature points in each cluster after classifying all the feature points, taking the coordinate mean value as a new clustering center, continuously iterating and determining to finally obtain k clustering centers, and clustering the remaining feature points to the cluster where the feature points with the nearest distance and the local density higher than that of the remaining feature points are located.
5. The multi-target positioning method based on the adaptive k-means clustering as claimed in claim 1, characterized in that: in S3, searching the characteristic points which are closest to and next to the Euclidean distance of the characteristic points of the template image in the characteristic point set of the image to be matched, and calculating the closest Euclidean distance disntAnd a second nearest euclidean distance dissntIf the ratio of (A) to (B) satisfies:
the matching of the characteristic point is successful, otherwise, the characteristic point pair is removed, and T is a set threshold value.
6. The multi-target positioning method based on the adaptive k-means clustering as claimed in claim 1, characterized in that: in S5, according to the matching confidence degree descending arrangement of the feature point pairs in the rough matching, randomly selecting 3 pairs of feature points from the previous n pairs of matching feature points through iteration to vote, and selecting 3 pairs of feature point pairs with the best fitness to establish a local coordinate system. Based on the constructed local coordinate system, constructing a straight line element from the feature points and the coordinate origin, expressing the linear element by vector coordinates, calculating the coordinate similarity of the vector coordinates in the local coordinate system, comparing the similarity with a set threshold value, and eliminating mismatching points.
7. The multi-objective positioning method based on adaptive k-means clustering as claimed in claim 6,
the method is characterized in that: the method for constructing the middle local coordinate system comprises the following steps:
selecting a first n pairs of feature point forming set R { (M) based on feature point matching confidence degree descending order in rough matchingi,Ni) 1, ·, n }, where MiAnd NiSelecting 3 characteristic points from the set R to select 3 characteristic points from the set R to construct a local coordinate system, voting and scoring by utilizing the similarity of the rest characteristic points in the set in the local coordinate system, and finally selecting the 3 characteristic points with the highest score to construct the local coordinate system.
The feature point similarity calculation formula is as follows:
wherein { (L)ji,Lji') | i ═ 1,2,3} is the matching point pair MjAnd NjEuclidean distances to 3 matching feature points;
the calculation method of the coordinate similarity of the characteristic point straight line elements is as follows:
the linear primitive vector is expressed by coordinates to form a coordinate set: q ═ pi(x,y),qiThe (x, y) | i ═ 1, …, N } straight line primitive vector is represented by a calculation formula in terms of coordinates:
wherein k is the slope of a straight line, (x)1,y1) Is the coordinates of the start point of the straight line element, (x)2,y2) Is a terminal coordinate;
secondly, a coordinate conversion formula for converting the vector coordinates into a local coordinate system is as follows:
wherein alpha and beta are coordinates of the vectorized straight line element in a local coordinate system;
computing coordinate similarity of straight line element
Wherein, PMAnd PNLocal coordinates of the straight line primitive.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111634014.5A CN114358166B (en) | 2021-12-29 | 2021-12-29 | Multi-target positioning method based on self-adaptive k-means clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111634014.5A CN114358166B (en) | 2021-12-29 | 2021-12-29 | Multi-target positioning method based on self-adaptive k-means clustering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114358166A true CN114358166A (en) | 2022-04-15 |
| CN114358166B CN114358166B (en) | 2023-11-07 |
Family
ID=81103034
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111634014.5A Active CN114358166B (en) | 2021-12-29 | 2021-12-29 | Multi-target positioning method based on self-adaptive k-means clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114358166B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116385480A (en) * | 2023-02-03 | 2023-07-04 | 腾晖科技建筑智能(深圳)有限公司 | Detection method and system for moving object below tower crane |
| CN116796214A (en) * | 2023-06-07 | 2023-09-22 | 南京北极光生物科技有限公司 | Data clustering method based on differential features |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103150737A (en) * | 2013-01-18 | 2013-06-12 | 西北工业大学 | Real-time space target feature point tracking method suitable for space tethered robot |
| WO2014067296A1 (en) * | 2012-11-05 | 2014-05-08 | 深圳市恩普电子技术有限公司 | Method for identifying, tracing and measuring internal and external membranes of vessel |
| CN104574347A (en) * | 2013-10-24 | 2015-04-29 | 南京理工大学 | On-orbit satellite image geometric positioning accuracy evaluation method on basis of multi-source remote sensing data |
| CN108154118A (en) * | 2017-12-25 | 2018-06-12 | 北京航空航天大学 | A kind of target detection system and method based on adaptive combined filter with multistage detection |
| WO2019245320A1 (en) * | 2018-06-22 | 2019-12-26 | 삼성전자주식회사 | Mobile robot device for correcting position by fusing image sensor and plurality of geomagnetic sensors, and control method |
| CN113177463A (en) * | 2021-04-27 | 2021-07-27 | 中国工商银行股份有限公司 | Target positioning method and device in mobile scene |
-
2021
- 2021-12-29 CN CN202111634014.5A patent/CN114358166B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014067296A1 (en) * | 2012-11-05 | 2014-05-08 | 深圳市恩普电子技术有限公司 | Method for identifying, tracing and measuring internal and external membranes of vessel |
| CN103150737A (en) * | 2013-01-18 | 2013-06-12 | 西北工业大学 | Real-time space target feature point tracking method suitable for space tethered robot |
| CN104574347A (en) * | 2013-10-24 | 2015-04-29 | 南京理工大学 | On-orbit satellite image geometric positioning accuracy evaluation method on basis of multi-source remote sensing data |
| CN108154118A (en) * | 2017-12-25 | 2018-06-12 | 北京航空航天大学 | A kind of target detection system and method based on adaptive combined filter with multistage detection |
| WO2019245320A1 (en) * | 2018-06-22 | 2019-12-26 | 삼성전자주식회사 | Mobile robot device for correcting position by fusing image sensor and plurality of geomagnetic sensors, and control method |
| CN113177463A (en) * | 2021-04-27 | 2021-07-27 | 中国工商银行股份有限公司 | Target positioning method and device in mobile scene |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116385480A (en) * | 2023-02-03 | 2023-07-04 | 腾晖科技建筑智能(深圳)有限公司 | Detection method and system for moving object below tower crane |
| CN116385480B (en) * | 2023-02-03 | 2023-10-20 | 腾晖科技建筑智能(深圳)有限公司 | Detection method and system for moving object below tower crane |
| CN116796214A (en) * | 2023-06-07 | 2023-09-22 | 南京北极光生物科技有限公司 | Data clustering method based on differential features |
| CN116796214B (en) * | 2023-06-07 | 2024-01-30 | 南京北极光生物科技有限公司 | Data clustering method based on differential features |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114358166B (en) | 2023-11-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110097093B (en) | Method for accurately matching heterogeneous images | |
| US8994723B2 (en) | Recognition and pose determination of 3D objects in multimodal scenes | |
| Su et al. | Fingerprint indexing with pose constraint | |
| US20110273442A1 (en) | Recognition and pose determination of 3d objects in 3d scenes | |
| CN110490913B (en) | Image matching method based on feature description operator of corner and single line segment grouping | |
| CN103136525B (en) | A kind of special-shaped Extended target high-precision locating method utilizing Generalized Hough Transform | |
| CN108509925B (en) | Pedestrian re-identification method based on visual bag-of-words model | |
| CN114358166A (en) | Multi-target positioning method based on self-adaptive k-means clustering | |
| CN107967477B (en) | Improved SIFT feature combined matching method | |
| CN110930456A (en) | Three-dimensional identification and positioning method of sheet metal part based on PCL point cloud library | |
| CN110443261B (en) | Multi-graph matching method based on low-rank tensor recovery | |
| CN111242221A (en) | Image matching method, system and storage medium based on image matching | |
| CN108182705A (en) | A kind of three-dimensional coordinate localization method based on machine vision | |
| CN111768447A (en) | Monocular camera object pose estimation method and system based on template matching | |
| CN110942473A (en) | Moving target tracking detection method based on characteristic point gridding matching | |
| CN111199558A (en) | Image matching method based on deep learning | |
| CN114332172A (en) | Improved laser point cloud registration method based on covariance matrix | |
| CN113902779A (en) | Point cloud registration method based on tensor voting method | |
| CN109829502B (en) | Image pair efficient dense matching method facing repeated textures and non-rigid deformation | |
| CN110705569A (en) | Image local feature descriptor extraction method based on texture features | |
| Han et al. | Accurate and robust vanishing point detection method in unstructured road scenes | |
| CN115690803A (en) | Digital image recognition method and device, electronic equipment and readable storage medium | |
| Ren et al. | SAR image matching method based on improved SIFT for navigation system | |
| CN111160433B (en) | High-speed matching method and system for high-resolution image feature points | |
| CN114494368A (en) | Low-overlapping-rate point cloud registration method combining dimensionality reduction projection and feature matching |
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 |