CN102208033A - Data clustering-based robust scale invariant feature transform (SIFT) feature matching method - Google Patents
Data clustering-based robust scale invariant feature transform (SIFT) feature matching method Download PDFInfo
- Publication number
- CN102208033A CN102208033A CN2011101859288A CN201110185928A CN102208033A CN 102208033 A CN102208033 A CN 102208033A CN 2011101859288 A CN2011101859288 A CN 2011101859288A CN 201110185928 A CN201110185928 A CN 201110185928A CN 102208033 A CN102208033 A CN 102208033A
- Authority
- CN
- China
- Prior art keywords
- feature
- matching
- characteristic
- sift
- cluster
- 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
Images
Landscapes
- Image Analysis (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a data clustering-based robust scale invariant feature transform (SIFT) feature matching method, which comprises the following steps of: (1) acquiring reference image sequences, extracting SIFT feature sets, clustering all the SIFT feature sets by adopting a synthetic k-d data structure, and merging the repeated feature sets as clustering features, wherein the feature descriptors of the clustering features are expressed by adopting repeated feature average descriptors; (2) acquiring real-time image sequence feature sets by using an SIFT method, performing robust matching on the reference image sequences and the real-time image sequences, and selecting the corresponding reference image with maximum feature points as a key image to complete feature matching of a first stage; and (3) completing feature matching of a second stage by using the key image, removing exterior points by adopting the technologies of random sampling consistency (RANSAC), basic matrix and the like, and finally merging the feature matching results of the two stages. According to the matching method, the noise information interference can be reduced, and the robustness of feature matching is greatly improved.
Description
Technical field
The invention belongs to computing machine augmented reality technical field, specifically the SIFT feature set under the big data quantity condition is carried out the robust coupling, eliminate the noise effect of matching process, improve the robustness of characteristic matching.
Background technology
SIFT characteristic matching technology is the important research content in computer picture and the vision, is all obtaining application widely such as various fields such as image retrieval, Target Recognition, three-dimensional reconstruction and camera pose recoveries.
By SIFT feature description is regarded as higher-dimension Euclidean space Cartesian coordinates and adopted Euclidean distance as the similarity measurement standard, SIFT characteristic matching problem can be converted into the Euclidean space inquiry problem of contiguous geometric point.Define one of (branch and bound) method as most important branch, the k-d data structure is implementation space subdivision and be widely used in adjacent features inquiry work effectively.But along with the increase of characteristic dimension, owing to need each branch of traversal with accurate location matching characteristic, the matching efficiency of this method can descend rapidly.At this problem, people such as Beis have proposed a kind of approximate neighbor method of using Priority Queues, are called BBF (Best Bin First) method.This method can only lost the matching speed that greatly improves high dimensional feature under few precision prerequisite, is widely used in the coupling work of SIFT feature.But because SIFT feature description is very responsive to noise information, along with the increase of characteristic amount and the increase that is distributed in Euclidean space unique point density, the robustness of this method characteristic matching can descend rapidly.
Different with k-d data structure space partition patterns, the vocabulary data structure is cut apart by stratification k-means cluster mode implementation space, this mode makes vocabulary tree structure construction process in fact be equivalent to space nested voronoi cell generative process, thereby has further promoted the range of application of branch's confining method.People such as Nister are applied to the SIFT characteristic matching with the vocabulary data structure, provided a kind of large nuber of images querying method based on the vocabulary tree, but only there is statistical significance in this matching process, can't effectively realize the most contiguous inquiry work of SIFT feature.People such as Schindler further are applied to the vocabulary tree city orientation problem based on image, provided GNP (the Greedy N-Best Paths) method of a kind of similar BBF, in order to realize based on the adjacent features coupling of the SIFT feature of vocabulary tree.But along with the increase of data volume, this method also exists the problem that the coupling robustness reduces.Recently, people such as Dong have provided a kind of two stage matching process based on the vocabulary tree, this method is at first searched key images by the vocabulary tree, utilize key images to carry out characteristic matching between two images then, effectively avoided the problem that the characteristic matching robustness reduces under the big data quantity condition, but this method is the characteristic matching between two images in essence, still can't solve SIFT feature noise-sensitive problem.
Summary of the invention
The technical problem to be solved in the present invention: overcome the deficiencies in the prior art, a kind of SIFT feature matching method of eliminating noise and improving robustness that has is provided.
The technical solution used in the present invention: based on the robust SIFT feature matching method of data clusters, its characteristics are that step is as follows:
(1) extracts reference image sequence SIFT feature set, to whole SIFT feature sets, distinguish general feature and cluster feature, merging repetition general feature collection by cluster process is cluster feature, the cluster feature descriptor adopts and repeats the equal value representation of general feature descriptor, the average optimisation strategy can greatly be eliminated the influence of noise to the SIFT feature, improves the robustness of characteristic matching;
(2) based on the SIFT feature clustering process of synthetic k-d data structure,, at first choose reference picture 1 and comprise feature set FS[1 given n width of cloth reference picture] and give cluster feature collection FCS with it, generate synthetic k-d data structure.Travel through reference picture 2 then and comprise feature set FS[2] and carry out based on the coupling of synthesizing the k-d data structure with cluster feature collection FCS successively.If the match is successful, merge then that reference picture 2 comprises feature set and the cluster feature set pair should be characterized as new cluster feature.Otherwise this reference picture 2 is comprised feature add cluster feature collection FCS as general feature.Finish reference picture 2 and comprise feature set FS[2] with reference picture 1 comprise feature set FS[1] feature clustering after, comprise feature set FS[3 for reference picture 3～n]～FS[n] repeat this cluster process up to finishing all feature sets.This process only needs iterative step n-1 time, avoids the higher time complexity of images match process in twos;
(3) two stage SIFT feature matching methods, comprise respectively the phase one based on the cluster feature match selection key images process of synthetic k-d data structure and subordinate phase based on characteristic matching process between the image of key images.By the merging of two stage matching characteristic collection, further improved the robustness of characteristic matching.
The merging repetition general feature collection of described step (1) is that the method for cluster feature is: at first feature is divided into general feature and cluster feature.At a large amount of repeated characteristics that exist in the general feature, merging the repeated characteristic collection by the arithmetic mean process is cluster feature then, and the cluster feature descriptor adopts the arithmetic equal value of repeated characteristic descriptor to represent.
SIFT feature clustering method based on synthetic k-d data structure in the described step (2) is: to given n width of cloth image, at first choose reference picture 1 and comprise feature set FS[1] and give cluster feature collection FCS with it, generate synthetic k-d data structure.Travel through reference picture 2 then and comprise feature set FS[2] and carry out based on the coupling of synthesizing the k-d data structure with cluster feature collection FCS successively.If the match is successful, merge then that reference picture 2 comprises feature set and the cluster feature set pair should be characterized as new cluster feature.Otherwise this reference picture 2 is comprised feature add cluster feature collection FCS as general feature.Finish reference picture 2 and comprise feature set FS[2] with reference picture 1 comprise feature set FS[1] feature clustering after, comprise feature set FS[3 for reference picture 3～n]～FS[n] repeat this cluster process up to finishing all feature sets.
Two phase characteristic matching process are in the described step (3): the phase one is undertaken by the characteristic matching mode based on the k-d data structure.For the characteristics of image that the match is successful each time, at first judge its characteristic type, if general feature then only is image ballot under it; If cluster feature, then voting process need comprise whole general feature to it and carries out successively.After finishing whole matching characteristic ballots, select the maximum images of poll as key images.Subordinate phase characteristic matching process at first carries out then phase one matching characteristic collection being purified based on characteristic matching between the image of key images, and the matching characteristic collection that will belong to key images is directly passed to subordinate phase matching characteristic collection.Utilize technology such as RANSAC to do exterior point rejecting work at last, improve the robustness of characteristic matching.
The present invention's advantage compared with prior art is:
(1) allows to exist a large amount of repeated characteristic collection.Is cluster feature by distinguishing general feature point with cluster feature point and by cluster process merging repeated characteristic collection, effectively reaches and reduces the data volume purpose.Cluster feature descriptor strategy is found the solution in average optimization can eliminate the influence of noise to the SIFT feature, has greatly improved the robustness of characteristic matching;
(2) by expanding the k-d data structure so that it can hold cluster feature, than standard k-d data structure, synthetic k-d data structure has better isomeric data type adaptability, feature clustering process based on such data structure, not only can reach the purpose of effective cluster, and can improve cluster process efficient.
(3) not only key images can be correctly selected based on the key images system of selection of counting temporal voting strategy, and the work of SIFT characteristic matching can be effectively finished.By phase one matching characteristic collection is passed to subordinate phase matching characteristic collection, the cascade of two stage matching characteristic collection merges the robustness that can effectively improve characteristic matching under the big data quantity condition.
Description of drawings
Fig. 1 is the synthetic k-d data structure synoptic diagram of the present invention.
Fig. 2 compares synoptic diagram for characteristic matching robustness of the present invention.
Fig. 3 (a) and (b) are road administration coupling robustness of the present invention synoptic diagram directly perceived.
Embodiment
Table 1 has provided concrete steps of the present invention:
Table 1
1. the synthetic k-d data structure of off-line phase makes up and feature clustering
Fig. 1 has provided synthetic k-d data structure synoptic diagram.
Off-line phase is at first obtained reference picture SIFT feature set, and whole reference picture SIFT feature sets are carried out cluster.In the feature clustering method, general feature is directly obtained from image by the SIFT algorithm, can be expressed as F
s={ b
i, d
d, b wherein
iFor being subordinate to image, d
dIt is feature description.Cluster feature is the set of general feature, is expressed as F
c={ F
Ss, d
Md, F wherein
SsBe general feature set, d
MdBe the average descriptor of cluster, its value is:
Wherein k is that cluster feature comprises general feature number, (d
d)
iBe i general feature descriptor.This value is equivalent to the geometric center point that cluster feature comprises the general feature collection, therefore can express cluster feature preferably.As shown in Figure 1, each node can both hold cluster feature in the synthetic k-d data structure.
For n width of cloth image, the repeated characteristic collection can obtain by coupling calculating in twos, but this process need carries out
Inferior coupling.Along with the increase of picture number, this computation process is too complicated.Therefore this paper has provided a kind of feature clustering method based on synthetic k-d data structure.Shown in algorithm 1.
Clustering method adopts iterative manner to carry out, and at first chooses reference picture 1 and comprises SIFT feature set FS[1] and give cluster feature collection FCS with it, generate synthetic k-d data structure.Travel through reference picture 2 then and comprise SIFT feature set FS[2], carry out based on the coupling of synthesizing the k-d data structure with cluster feature collection FCS successively.If the match is successful, then merging the feature set that reference picture 2 comprises is new cluster feature with the concentrated character pair of cluster feature, and the cluster feature descriptor is the arithmetic mean of two feature description.Otherwise this feature is added cluster feature collection FCS as general feature.Finish reference picture 2 and comprise SIFT feature set FS[2] with reference picture 1 comprise SIFT feature set FS[1] feature clustering after, comprise SIFT feature set FS[3 for follow-up reference picture 3～n]～FS[n], repeat said process, carry out cluster process with cluster feature collection FCS successively.For n width of cloth image, the clustering method that increases progressively that this paper proposes only needs matching process n-1 time, has significantly saved computing cost with respect to mating in twos.
2. online two stages cascade coupling
The phase one characteristic matching is selected the key images process, shown in algorithm 2:
Algorithm is by carrying out based on the characteristic matching mode of synthetic k-d data structure.For the reference picture feature that the match is successful each time, algorithm is at first judged its characteristic type, if general feature then only is reference picture ballot under it; If cluster feature, then voting process need comprise whole general feature to it and carries out successively.After finishing whole matching characteristic ballots, select the maximum images of poll as key images.
Select key images can guarantee the accuracy that key images is selected preferably by above-mentioned probability optimization method, can further utilize this key images to carry out the subordinate phase coupling.Shown in algorithm 3.
The key images system of selection shows the concentrated general feature that belongs to key images in a large number that comprises of phase one coupling cluster feature in the phase one characteristic matching process.Algorithm 2 is when utilizing realtime graphic SIFT feature set and cluster feature collection match selection key images, in fact also partly finished the coupling work of the corresponding SIFT feature set of realtime graphic SIFT feature set with key images, therefore in the subordinate phase characteristic matching process of algorithm 3, for the phase one matching process feature set that the match is successful, it directly can be passed to subordinate phase matching characteristic collection, thus the time overhead of having avoided two stage repeated matching to cause.After finishing two phase characteristics couplings, can utilize technology such as RANSAC to do exterior point rejecting work, further obtain the robust features collection.
Fig. 2 is the robust SIFT characteristic matching effect synoptic diagram based on data clusters.The demonstration effect realizes that the hardware configuration environment is an Intel Core 2Duo 2.66GHz processor on the logical desktop PC device of a Daepori, software environment is a Windows XP platform, adopts the VC2005.NET platform development, has used the OpenCV storehouse in the performance history.Unified 640 * 480 pixel size pictures that adopt of image sequence.
For carrying out this demonstration, whole process has selected 60 width of cloth images as the reference image sequence, and 300 width of cloth images are as real-time matching image sequence, and every width of cloth image comprises about 1500～2000 unique points.There are a large amount of overlapping relations between image, guarantee to comprise in whole characteristic sets a large amount of repeated characteristic points.Be the robustness of verification method, we further therefrom picked at random two width of cloth matching effects synoptic diagram directly perceived, as shown in Figure 3.
In Fig. 3 (a) and (b), the matching characteristic collection that no arrow line segment representative image-image matching method generates, the new matching characteristic collection that the matching process that on behalf of this paper, the circular arrow line segment propose generates, image-image matching method matching characteristic number is 344 among Fig. 3 (a); The matching method matches characteristic number that this paper proposes among Fig. 3 (b) is 88.Above-mentioned data show that the inventive method can effectively reduce noise effect, greatly improve the robustness of SIFT characteristic matching.
Claims (4)
1. robust image SIFT feature matching method based on data clusters is characterized in that step is as follows:
(1) obtain reference image sequence, extract image SIFT feature set, merging the repeated characteristic that comprises in all images SIFT feature set is cluster feature, reduces noise, improves the characteristic matching robustness;
(2) expansion k-d data structure is synthetic k-d data structure, makes synthetic k-d data structure node can hold cluster feature, realizes the feature clustering process based on synthetic k-d data structure;
(3) based on the two stage SIFT characteristic matching processes of synthesizing the k-d data structure, phase one utilizes synthetic k-d data structure to carry out the cluster feature coupling, for the cluster feature that the match is successful, vote for the corresponding reference picture of its whole repeated characteristics that comprise based on the optimum probability temporal voting strategy, select at last to comprise the maximum reference pictures of unique point as key images, subordinate phase utilizes key images to finish characteristic matching; At the matching characteristic collection that generates in the two stage matching processs, merge all matching characteristic sequences that belong to identical key images and carry out exterior point rejecting work as final matching characteristic collection and to the matching error unique point.
2. according to the described robust image SIFT feature matching method based on data clusters of claim 1, it is characterized in that: the repeated characteristic that comprises in the merging all images SIFT feature set of described step (1) is that cluster feature is specially: at first feature is divided into general feature and cluster feature; At a large amount of repeated characteristics that exist in the general feature, merging the repeated characteristic collection by the arithmetic mean process is cluster feature then, and the cluster feature descriptor adopts the arithmetic equal value of repeated characteristic descriptor to represent.
3. according to the described robust image SIFT feature matching method of claim 1 based on data clusters, it is characterized in that: realize in the described step (2) being specially:, at first choose the feature set FS[1 that reference picture 1 comprises given n width of cloth reference picture based on the feature clustering process of synthetic k-d data structure] and give cluster feature collection FCS generation with it and synthesize the k-d data structure; Travel through the feature set FS[2 that reference picture 2 comprises then], and carry out based on the coupling of synthesizing the k-d data structure with cluster feature collection FCS successively; If the match is successful, then merge two and be characterized as new cluster feature, otherwise this feature is added cluster feature collection FCS as general feature; Finish the feature set FS[2 that reference picture 2 comprises] the feature set FS[1 that comprises with reference picture 1] feature clustering after, the feature set FS[3 that comprises for follow-up reference picture 3～n]～FS[n], repeat this cluster process up to finishing all feature sets.
4. according to the described robust image SIFT feature matching method based on data clusters of claim 1, it is characterized in that: two stage SIFT characteristic matching processes are in the described step (3): the phase one is specifically undertaken by the characteristic matching mode based on the k-d data structure; For the characteristics of image that the match is successful each time, at first judge its characteristic type, if general feature then only is image ballot under it; If cluster feature, then voting process need comprise whole general feature to it and carries out successively; After finishing whole matching characteristic ballots, select the maximum images of poll as key images; Subordinate phase characteristic matching process at first carries out then phase one matching characteristic collection being purified based on characteristic matching between the image of key images, and the matching characteristic collection that will belong to key images is directly passed to subordinate phase matching characteristic collection; Utilize the RANSAC technology to do exterior point rejecting work at last, improve the robustness of characteristic matching.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110185928 CN102208033B (en) | 2011-07-05 | 2011-07-05 | Data clustering-based robust scale invariant feature transform (SIFT) feature matching method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110185928 CN102208033B (en) | 2011-07-05 | 2011-07-05 | Data clustering-based robust scale invariant feature transform (SIFT) feature matching method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102208033A true CN102208033A (en) | 2011-10-05 |
| CN102208033B CN102208033B (en) | 2013-04-24 |
Family
ID=44696855
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110185928 Expired - Fee Related CN102208033B (en) | 2011-07-05 | 2011-07-05 | Data clustering-based robust scale invariant feature transform (SIFT) feature matching method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102208033B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102693311A (en) * | 2012-05-28 | 2012-09-26 | 中国人民解放军信息工程大学 | Target retrieval method based on group of randomized visual vocabularies and context semantic information |
| CN103034859A (en) * | 2012-12-13 | 2013-04-10 | 华为技术有限公司 | Method and device for obtaining posture model |
| CN103366387A (en) * | 2012-02-28 | 2013-10-23 | 索尼公司 | Selecting between clustering techniques for displaying images |
| CN103440652A (en) * | 2013-08-27 | 2013-12-11 | 电子科技大学 | Method for describing target detection area features based on merging between first order and second order |
| CN104216974A (en) * | 2014-08-28 | 2014-12-17 | 西北工业大学 | Unmanned aerial vehicle aerial image matching method based on vocabulary tree blocking and clustering |
| CN105809118A (en) * | 2016-03-03 | 2016-07-27 | 重庆中科云丛科技有限公司 | Three-dimensional object identifying method and apparatus |
| CN106327188A (en) * | 2016-08-15 | 2017-01-11 | 华为技术有限公司 | Binding method and device for bank cards in payment application |
| CN111968243A (en) * | 2020-06-28 | 2020-11-20 | 成都威爱新经济技术研究院有限公司 | AR image generation method, system, device and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5983224A (en) * | 1997-10-31 | 1999-11-09 | Hitachi America, Ltd. | Method and apparatus for reducing the computational requirements of K-means data clustering |
| US7475071B1 (en) * | 2005-11-12 | 2009-01-06 | Google Inc. | Performing a parallel nearest-neighbor matching operation using a parallel hybrid spill tree |
| CN101350101A (en) * | 2008-09-09 | 2009-01-21 | 北京航空航天大学 | Method for auto-registration of multi-amplitude deepness image |
| CN101388115A (en) * | 2008-10-24 | 2009-03-18 | 北京航空航天大学 | Depth image autoegistration method combined with texture information |
| CN101697232A (en) * | 2009-09-18 | 2010-04-21 | 浙江大学 | SIFT characteristic reducing method facing close repeated image matching |
| CN101944183A (en) * | 2010-09-02 | 2011-01-12 | 北京航空航天大学 | Method for identifying object by utilizing SIFT tree |
-
2011
- 2011-07-05 CN CN 201110185928 patent/CN102208033B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5983224A (en) * | 1997-10-31 | 1999-11-09 | Hitachi America, Ltd. | Method and apparatus for reducing the computational requirements of K-means data clustering |
| US7475071B1 (en) * | 2005-11-12 | 2009-01-06 | Google Inc. | Performing a parallel nearest-neighbor matching operation using a parallel hybrid spill tree |
| CN101350101A (en) * | 2008-09-09 | 2009-01-21 | 北京航空航天大学 | Method for auto-registration of multi-amplitude deepness image |
| CN101388115A (en) * | 2008-10-24 | 2009-03-18 | 北京航空航天大学 | Depth image autoegistration method combined with texture information |
| CN101697232A (en) * | 2009-09-18 | 2010-04-21 | 浙江大学 | SIFT characteristic reducing method facing close repeated image matching |
| CN101944183A (en) * | 2010-09-02 | 2011-01-12 | 北京航空航天大学 | Method for identifying object by utilizing SIFT tree |
Non-Patent Citations (1)
| Title |
|---|
| 宋金龙,等: "基于Feature Forest 的图像检索", 《计算机工程》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103366387A (en) * | 2012-02-28 | 2013-10-23 | 索尼公司 | Selecting between clustering techniques for displaying images |
| CN102693311A (en) * | 2012-05-28 | 2012-09-26 | 中国人民解放军信息工程大学 | Target retrieval method based on group of randomized visual vocabularies and context semantic information |
| CN102693311B (en) * | 2012-05-28 | 2014-07-23 | 中国人民解放军信息工程大学 | Target retrieval method based on group of randomized visual vocabularies and context semantic information |
| CN103034859A (en) * | 2012-12-13 | 2013-04-10 | 华为技术有限公司 | Method and device for obtaining posture model |
| CN103034859B (en) * | 2012-12-13 | 2016-03-30 | 华为技术有限公司 | A kind of method and device obtaining gesture model |
| CN103440652A (en) * | 2013-08-27 | 2013-12-11 | 电子科技大学 | Method for describing target detection area features based on merging between first order and second order |
| CN104216974A (en) * | 2014-08-28 | 2014-12-17 | 西北工业大学 | Unmanned aerial vehicle aerial image matching method based on vocabulary tree blocking and clustering |
| CN104216974B (en) * | 2014-08-28 | 2017-07-21 | 西北工业大学 | The method of unmanned plane images match based on words tree Block Cluster |
| CN105809118A (en) * | 2016-03-03 | 2016-07-27 | 重庆中科云丛科技有限公司 | Three-dimensional object identifying method and apparatus |
| CN106327188A (en) * | 2016-08-15 | 2017-01-11 | 华为技术有限公司 | Binding method and device for bank cards in payment application |
| US10937016B2 (en) | 2016-08-15 | 2021-03-02 | Huawei Technologies Co., Ltd. | Method and apparatus for binding bank card in payment application |
| CN111968243A (en) * | 2020-06-28 | 2020-11-20 | 成都威爱新经济技术研究院有限公司 | AR image generation method, system, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102208033B (en) | 2013-04-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102208033B (en) | Data clustering-based robust scale invariant feature transform (SIFT) feature matching method | |
| Shi et al. | From points to parts: 3d object detection from point cloud with part-aware and part-aggregation network | |
| CN101593205B (en) | Video-based method for searching three-dimension model | |
| CN102945289B (en) | Based on the image search method of CGCI-SIFT local feature | |
| CN101477529B (en) | Three-dimensional object retrieval method and apparatus | |
| CN105956560A (en) | Vehicle model identification method based on pooling multi-scale depth convolution characteristics | |
| CN102194133B (en) | Data-clustering-based adaptive image SIFT (Scale Invariant Feature Transform) feature matching method | |
| CN101859320B (en) | Massive image retrieval method based on multi-characteristic signature | |
| CN111027559A (en) | Point cloud semantic segmentation method based on expansion point convolution space pyramid pooling | |
| CN103336957A (en) | Network coderivative video detection method based on spatial-temporal characteristics | |
| CN109920050B (en) | Single-view three-dimensional flame reconstruction method based on deep learning and thin plate spline | |
| CN110097581B (en) | Method for constructing K-D tree based on point cloud registration ICP algorithm | |
| CN116128944A (en) | Three-dimensional point cloud registration method based on feature interaction and reliable corresponding relation estimation | |
| CN103839074A (en) | Image classification method based on matching of sketch line segment information and space pyramid | |
| CN111597367B (en) | Three-dimensional model retrieval method based on view and hash algorithm | |
| CN106951501B (en) | Three-dimensional model retrieval method based on multi-graph matching | |
| CN104699783A (en) | Social image searching method allowing adaptive adjustment and based on personalized vision dictionary | |
| CN108536769B (en) | Image analysis method, search method and device, computer device and storage medium | |
| CN116229406A (en) | Lane line detection method, system, electronic equipment and storage medium | |
| CN105740878A (en) | Symbol recognition method and device of geographic information system | |
| CN101894267A (en) | Three-dimensional object characteristic view selection method | |
| CN111382301B (en) | Three-dimensional model generation method and system based on generation countermeasure network | |
| CN108596034B (en) | Pedestrian re-identification method based on target center coding appearance model | |
| Lu et al. | Road vanishing-point detection: A multi-population genetic algorithm based approach | |
| Wu et al. | Ground-distance segmentation of 3D LiDAR point cloud toward autonomous driving |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130424 Termination date: 20170705 |