CN108830281A - A kind of multiimage matching process based on localized variation detection and spatial weighting - Google Patents
A kind of multiimage matching process based on localized variation detection and spatial weighting Download PDFInfo
- Publication number
- CN108830281A CN108830281A CN201810527166.7A CN201810527166A CN108830281A CN 108830281 A CN108830281 A CN 108830281A CN 201810527166 A CN201810527166 A CN 201810527166A CN 108830281 A CN108830281 A CN 108830281A
- Authority
- CN
- China
- Prior art keywords
- projective transformation
- point
- matching
- image
- match point
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Abstract
The invention discloses a kind of multiimage matching process based on localized variation detection and spatial weighting, specific steps include：Image to be matched information is extracted by SIFT algorithm；Carry out projective transformation；Three indexs of detection image：Distance, abnormal point, projective transformation amplitude between match point；Objective function is solved with least square method：It matches dot position information similarity maximum and projective transformation amplitude is minimum, obtain projective transformation parameter；The matching dot position information similarity under this projective transformation parameter is calculated, the multilevel iudge images match situation with threshold value is passed through.Influence of the tiny transformation of shooting angle to result is simulated by introducing projective transformation, it is contemplated that influence of the spatial distribution of pixel to result, while influence of the abnormal point to matching result is added, have great importance to the multiimage detection of monitor supervision platform.
Description
Technical field
The invention belongs to technical field of image detection, and in particular to a kind of weight based on localized variation detection and spatial weighting
Complex pattern matching process.
Background technique
Continuous improvement with the development and people of scientific and technological information technology to demand for security, video monitoring system are much being led
Domain is all widely used.As the important component of security product, video monitoring product accounts for its ratio on 50% left side
The right side, during 2014-2016, ratio of the video monitoring in security product application is respectively 47.06%, 48.33% and
50.63%.During 2010-2017, China's video monitoring market scale rises to 112,400,000,000 yuan from 24,200,000,000 yuan, average annual compound
Growth rate is up to 24.53%.
For the multiimage detection in monitor video, domestic and foreign scholars have done a large amount of research work.Main image
Matching algorithm can be classified as four classes：Matching algorithm, feature-based matching algorithm, the matching algorithm based on model based on region
With the matching algorithm based on transform domain.The characteristic point in image to be matched, such as brightness, edge, angle point, profile information are extracted,
It carries out matching to be at present most generally being also most effective multiimage detection method in conjunction with measuring similarity criterion.
SIFT (Scale invariant features transform) is a kind of local feature description's, and SIFT feature matching algorithm can handle two
The matching problem in the case of translation, rotation, affine transformation occurs between width image, there is very strong matching capacity；SIFT algorithm
The matching characteristic point extracted lacks location information, does not consider the influence of spatial distribution and abnormal point, is directly used in multiimage
Matching lack certain reasonability.
Clustering is also known as cluster analysis, is a kind of multi-variate statistical analysis that quantitative classification is carried out to multiple samples (or index)
Method.Q type clustering is also known as into classification to sample.
The present invention introduces two-dimensional Gaussian function and Q type clustering, proposes one on the basis of SIFT algorithm as a result,
Multiimage matching process of the kind based on localized variation detection and spatial weighting.
Summary of the invention
To solve deficiency in the prior art, the present invention provides a kind of repetition based on localized variation detection and spatial weighting
Image matching method, the spatial distribution of pixel and abnormal point are lacked when solving to measure image repeated matching degree leads to matching not
Accurate problem.
In order to achieve the above objectives, the present invention adopts the following technical scheme that：One kind is added based on localized variation detection and space
The multiimage matching process of power, it is characterised in that：Including step：
Step 1：Image to be matched is inputted, by the match point in SIFT algorithm extraction image to be matched and its accordingly
Location information；
Step 2：Piece image in selected image to be matched carries out projective transformation；
Step 3：After projective transformation, using between match point distance, abnormal point, projective transformation amplitude as measure repeat scheme
Three indexs of the matching degree of picture；
Step 4：In conjunction with distance, abnormal point, the projective transformation amplitude between match point in step 3, establish to match point
Confidence ceases similarity maximum and the smallest objective function of projective transformation amplitude；
Step 5：Projective transformation parameter is solved by least square method to objective function, substituting into objective function will be final
Obtained target function value obtains matching result compared with threshold value.
A kind of multiimage matching process based on localized variation detection and spatial weighting above-mentioned, it is characterised in that：Institute
It states in step 1, the match point and its corresponding location information in image to be matched is extracted by SIFT algorithm, obtain formula：
Wherein, n is the logarithm of total matching characteristic point, (xi, yi) withIndicate match point respectively in two images
In position, i=1~n；P withRespectively indicate all matching dot position informations of image to be matched, P withBetween for one by one
Corresponding relationship.
A kind of multiimage matching process based on localized variation detection and spatial weighting above-mentioned, it is characterised in that：Institute
It states the piece image that step 2 is selected in image to be matched and carries out projective transformation, specially：
Wherein, a, b, c, d are projective transformation parameter,Indicate all matching dot position informations of the P after projective transformation, P
Indicate all matching dot position informations in the piece image of image to be matched.
A kind of multiimage matching process based on localized variation detection and spatial weighting above-mentioned, it is characterised in that：Step
Rapid three：After projective transformation, using between match point distance, abnormal point, projective transformation amplitude as measure multiimage matching degree
Three indexs, specially：
Step 3.1：The distance between corresponding match point of image to be matched after calculating projective transformation：
Wherein,Images match point after expression projective transformationWith match point before projective transformationIt is European between correspondence
Distance；
After being weighted in such a way that Euclidean distance is combined with two-dimensional Gaussian functionWithBetween Corresponding matching point
Euclidean distance ei, specially：
Wherein, ωiIndicate the weight that obtains by two-dimensional Gaussian function, m and l respectively represent the row pixel number of image with
Column pixel number；
Step 3.2：Detect the abnormal point between match point；
It introduces Q type clustering and looks for potential abnormal point, using the Euclidean distance of match point as sorting criterion, by match point
It is divided into and fluctuates larger and fluctuate smaller two class, wherein fluctuating biggish match point is considered as potential abnormal point, especially by 0-1 variable
fiIt indicates, such as formula (6)：
Wherein, Z indicates to fluctuate the distance set of smaller match point,Indicate the distance set of the larger match point of fluctuation, i.e.,
The distance set of potential abnormal point；
Step 3.3：Projective transformation amplitude is detected, the amplitude g of projective transformation is indicated by formula (7)：
G=a2+b2+c2+d2 (7)
Wherein, a, b, c, d are projective transformation parameter, and amplitude g is bigger, and the degree of projective transformation is bigger, and images match degree is got over
It is low.
A kind of multiimage matching process based on localized variation detection and spatial weighting above-mentioned, it is characterised in that：Institute
State step 4：In conjunction with the distance, abnormal point, projective transformation amplitude of match point in step 3, establish to match dot position information phase
It is specially like degree maximum and the smallest objective function of projective transformation amplitude, objective function J：
Wherein, n is the logarithm of total matching characteristic point, and λ is correction factor.
A kind of multiimage matching process based on localized variation detection and spatial weighting above-mentioned, it is characterised in that：Institute
Belong to step 5：Projective transformation parameter is solved by least square method to objective function, substituting into objective function will finally obtain
Target function value matching result is obtained compared with threshold value, specially：
It is that objective function is solved with least square method with formula (8)：
Partial derivative is asked to parameter a, b, c, d respectively, is indicated as shown in equation group (9)：
Solving equations (9), can find out parameter a, b, c, d, as shown in formula (10)：
Wherein,
The expression formula that the projective transformation parameter formula (10) solved is substituted into J in formula (8), will be by finally obtained target letter
Numerical value is compared with threshold value and then judges the matching degree of image.
The device have the advantages that：Repetition figure based on localized variation detection and spatial weighting proposed by the invention
As matching process, the redundant image of monitor supervision platform can be effectively detected, a large amount of memory space is saved, become by introducing projection
Change the influence for simulating the tiny transformation of shooting angle to result, it is contemplated that pixel space is distributed the influence to result, is added simultaneously
Influence of the abnormal point to matching result improves the reasonability and reliability of multiimage detection.
Detailed description of the invention
Fig. 1 is the flow diagram of the method for the present invention；
Fig. 2 is the implementation process schematic diagram of SIFT algorithm.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.Following embodiment is only used for clearly illustrating the present invention
Technical solution, and not intended to limit the protection scope of the present invention.
As shown in Figure 1, a kind of multiimage matching process based on localized variation detection and spatial weighting, including following step
Suddenly：
Step 1：Image to be matched is inputted, as shown in Fig. 2, extracting the match point in image to be matched by SIFT algorithm
And its corresponding location information, as shown in formula (1),
Wherein, n is the logarithm of total matching characteristic point, (xi, yi) withIndicate match point respectively in two images
In position, i=1~n；P withRespectively indicate all matching dot position informations of image to be matched, P withBetween for one by one
Corresponding relationship；
Step 2：Piece image in selected image to be matched carries out projective transformation, simulates the slight change of shooting angle
Influence to matching result；
Most of multiimages all inevitably will receive the influence of shooting angle, thus introduces projective transformation and goes to simulate
Angular transformation influences matching result bring.Assuming that the projective transformation as shown in formula (2) exists, then it is to be matched after projective transformation
The match point distribution of image is roughly the same, and the piece image specifically selected in image to be matched carries out following projective transformation：
Wherein, a, b, c, d are projective transformation parameter,Indicate all matching dot position informations of the P after projective transformation；
Step 3：After projective transformation, using between match point distance, abnormal point, projective transformation amplitude as measure repeat scheme
Three indexs of the matching degree of picture, detailed process are：
Step 3.1：The distance between corresponding match point of image to be matched after calculating projective transformation；
After projective transformation, detecting projective transformation by the distance between match point whether there is, if illustrating apart from excessive
Projective transformation appropriate is not present between image to be matched, i.e. matching degree is lower, characterizes in particular by Euclidean distance：
Wherein,Images match point after expression projective transformationWith match point before projective transformationBetween correspondence it is European away from
From；
Most of image main information concentrates on intermediate region, and what most of surrounding pixel point embodied is background information,
Influence to matching result is far away from intermediate region.Two-dimensional Gaussian function is introduced as a result, describes the spatial distribution of pixel to knot
The center of two-dimensional Gaussian function, is corresponded to central pixel point (0.5m, 0.5l) (its of image by the important implementations that fruit influences
In, m and l respectively represent the row pixel number and column pixel number of image) at, two-dimensional Gaussian function value is in spoke by center around
It penetrates shape to successively decrease, final match point distance e is obtained in such a way that Euclidean distance is combined with two-dimensional Gaussian functioni, specially：
Wherein, eiAfter indicating weightingWithEuclidean distance between Corresponding matching point, ωiExpression passes through dimensional Gaussian
The weight that function obtains, m and l respectively represent the row pixel number and column pixel number of image.
Step 3.2：Detect the abnormal point between match point；
Consider as a whole all match points apart from when often ignore the presence of abnormal point, therefore, introduce Q type cluster point
Analysis look for potential abnormal point, using the Euclidean distance of match point as sorting criterion, by match point be divided into fluctuation it is larger and fluctuate compared with
Small two class, wherein fluctuating biggish match point is considered as potential abnormal point, their distance value is most important to matching result, specifically
Pass through 0-1 variable fiIt indicates, such as formula (6)：
Wherein, Z indicates to fluctuate the distance set of smaller match point,Indicate the distance set of the biggish match point of fluctuation,
The distance set of i.e. potential abnormal point；
Step 3.3：Projective transformation amplitude is detected, distance, the pixel when measuring image to be matched, in addition to considering match point
Spatial distribution and abnormal point, it is also necessary to guarantee projective transformation amplitude in a certain range, lose it if projective transformation is excessive
Existing meaning not can guarantee the similitude of image, and the amplitude g of projective transformation is indicated in particular by formula (7)：
A=a2+b2+c2+d2 (7)
Wherein, a, b, c, d are projective transformation parameter, and amplitude g is bigger, and the degree of projective transformation is bigger, and images match degree is got over
It is low；
Step 4：In conjunction with three indexs (distance, abnormal point, the projective transformation amplitude of match point) in step 3, establish
To match dot position information similarity maximum and the smallest objective function of projective transformation amplitude, specially：
There are two the main sources for influencing matching result：The location information of match point and projective transformation, wherein match point
Location information is measured by Euclidean distance, and the influence of the spatial distribution and abnormal point of pixel is added.Finally, with potential exception
Point Gauss Weighted distance and come characterize match point location information influence.For image higher for matching degree, distance
Small as far as possible it should just can guarantee that projective transformation appropriate exists, while projective transformation also should ensure that in the range of tolerable,
Otherwise projective transformation will lose meaning, thus obtain objective function J：
Wherein, n is the logarithm of total matching characteristic point, and λ is correction factor, according to experimental results, selectes the value of λ
Range is：0.1~10；
Step 5：Projective transformation parameter is solved by least square method to objective function, substitutes into objective function, it will be final
Target function value obtains matching result compared with threshold value；
Estimation for projective transformation parameter is that objective function is solved with least square method with formula (8)：
Partial derivative is asked to parameter a, b, c, d respectively, is indicated as shown in equation group (9)：
Solving equations (9), can find out parameter a, b, c, d, as shown in formula (10)：
Wherein,
The expression formula that the projective transformation parameter formula (10) solved is substituted into J in formula (8), by finally obtained objective function
Value is compared with threshold value and then judges the matching degree of image.
To sum up, the present invention can simulate influence of the subtle shooting angle variation to result, it is contemplated that the space of pixel point
Influence of the cloth to result, while influence of the abnormal point to matching result is added, there is higher robustness.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, several improvement and deformations can also be made, these improvement and deformations
Also it should be regarded as protection scope of the present invention.
Claims (6)
1. a kind of multiimage matching process based on localized variation detection and spatial weighting, it is characterised in that：Including step：
Step 1：Image to be matched is inputted, match point and its corresponding position in image to be matched are extracted by SIFT algorithm
Information；
Step 2：Piece image in selected image to be matched carries out projective transformation；
Step 3：After projective transformation, using between match point distance, abnormal point, projective transformation amplitude is as measuring multiimage
Three indexs of matching degree；
Step 4：In conjunction with distance, abnormal point, the projective transformation amplitude between match point in step 3, establish with match point position letter
Cease similarity maximum and the smallest objective function of projective transformation amplitude；
Step 5：Projective transformation parameter is solved by least square method to objective function, substituting into objective function will finally obtain
Target function value matching result is obtained compared with threshold value.
2. a kind of multiimage matching process based on localized variation detection and spatial weighting according to claim 1,
It is characterized in that：In the step 1, the match point and its corresponding location information in image to be matched are extracted by SIFT algorithm,
Obtain formula：
Wherein, n is the logarithm of total matching characteristic point, (xi, yi) withIndicate the position of match point respectively in two images
It sets, i=1~n；P withRespectively indicate all matching dot position informations of image to be matched, P withBetween for correspond close
System.
3. a kind of multiimage matching process based on localized variation detection and spatial weighting according to claim 1,
It is characterized in that：The step 2 selectes the piece image in image to be matched and carries out projective transformation, specially：
Wherein, a, b, c, d are projective transformation parameter,Indicate that all matching dot position informations of the P after projective transformation, P indicate
All matching dot position informations in the piece image of image to be matched.
4. a kind of multiimage matching process based on localized variation detection and spatial weighting according to claim 3,
It is characterized in that：Step 3：After projective transformation, using between match point distance, abnormal point, projective transformation amplitude as measure repeat scheme
Three indexs of the matching degree of picture, specially：
Step 3.1：The distance between corresponding match point of image to be matched after calculating projective transformation：
Wherein,Images match point after expression projective transformationWith match point before projective transformationEuclidean distance between correspondence；
After being weighted in such a way that Euclidean distance is combined with two-dimensional Gaussian functionWithEurope between Corresponding matching point
Formula distance ei, specially：
Wherein, ωiIndicate the weight obtained by two-dimensional Gaussian function, m and l respectively represent the row pixel number and column picture of image
Vegetarian refreshments number；
Step 3.2：Detect the abnormal point between match point；
It introduces Q type clustering and looks for potential abnormal point, using the Euclidean distance of match point as sorting criterion, match point is divided into
It fluctuates larger and fluctuates smaller two class, wherein fluctuating biggish match point is considered as potential abnormal point, especially by 0-1 variable fiTable
Show, such as formula (6)：
Wherein, Z indicates to fluctuate the distance set of smaller match point,Indicate the distance set of the larger match point of fluctuation, i.e., it is potential
The distance set of abnormal point；
Step 3.3：Projective transformation amplitude is detected, the amplitude g of projective transformation is indicated by formula (7)：
G=a2+b2+c2+d2 (7)
Wherein, a, b, c, d are projective transformation parameter, and amplitude g is bigger, and the degree of projective transformation is bigger, and images match degree is lower.
5. a kind of multiimage matching process based on localized variation detection and spatial weighting according to claim 4,
It is characterized in that：The step 4：In conjunction with the distance, abnormal point, projective transformation amplitude of match point in step 3, establish with match point
Location information similarity maximum and the smallest objective function of projective transformation amplitude, objective function J are specially：
Wherein, n is the logarithm of total matching characteristic point, and λ is correction factor.
6. a kind of multiimage matching process based on localized variation detection and spatial weighting according to claim 5,
It is characterized in that：Belonging step 5：Projective transformation parameter is solved by least square method to objective function, substitutes into objective function
Finally obtained target function value is compared with threshold value and obtains matching result, specially：
It is that objective function is solved with least square method with formula (8)：
Partial derivative is asked to parameter a, b, c, d respectively, is indicated as shown in equation group (9)：
Solving equations (9), can find out parameter a, b, c, d, as shown in formula (10)：
Wherein,
By the projective transformation parameter formula (10) solved substitute into formula (8) in J expression formula, by finally obtained target function value with
Threshold value is compared and then judges the matching degree of image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810527166.7A CN108830281B (en) | 2018-05-29 | 2018-05-29 | Repeated image matching method based on local change detection and spatial weighting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810527166.7A CN108830281B (en) | 2018-05-29 | 2018-05-29 | Repeated image matching method based on local change detection and spatial weighting |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108830281A true CN108830281A (en) | 2018-11-16 |
| CN108830281B CN108830281B (en) | 2021-09-28 |
Family
ID=64145888
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810527166.7A Active CN108830281B (en) | 2018-05-29 | 2018-05-29 | Repeated image matching method based on local change detection and spatial weighting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830281B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111832910A (en) * | 2020-06-24 | 2020-10-27 | 陕西法士特齿轮有限责任公司 | Method and system for determining multi-index abnormal sound judgment threshold value and computer equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101354796A (en) * | 2008-09-05 | 2009-01-28 | 浙江大学 | Omnidirectional stereo vision three-dimensional rebuilding method based on Taylor series model |
| CN102968777A (en) * | 2012-11-20 | 2013-03-13 | 河海大学 | Image stitching method based on overlapping region scale-invariant feather transform (SIFT) feature points |
| US20140064554A1 (en) * | 2011-11-14 | 2014-03-06 | San Diego State University Research Foundation | Image station matching, preprocessing, spatial registration and change detection with multi-temporal remotely-sensed imagery |
| CN105654421A (en) * | 2015-12-21 | 2016-06-08 | 西安电子科技大学 | Projection transform image matching method based on transform invariant low-rank texture |
-
2018
- 2018-05-29 CN CN201810527166.7A patent/CN108830281B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101354796A (en) * | 2008-09-05 | 2009-01-28 | 浙江大学 | Omnidirectional stereo vision three-dimensional rebuilding method based on Taylor series model |
| US20140064554A1 (en) * | 2011-11-14 | 2014-03-06 | San Diego State University Research Foundation | Image station matching, preprocessing, spatial registration and change detection with multi-temporal remotely-sensed imagery |
| CN102968777A (en) * | 2012-11-20 | 2013-03-13 | 河海大学 | Image stitching method based on overlapping region scale-invariant feather transform (SIFT) feature points |
| CN105654421A (en) * | 2015-12-21 | 2016-06-08 | 西安电子科技大学 | Projection transform image matching method based on transform invariant low-rank texture |
Non-Patent Citations (1)
| Title |
|---|
| 姚国标等: "融合互补仿射不变特征的倾斜立体影像高精度自动配准方法", 《测绘学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111832910A (en) * | 2020-06-24 | 2020-10-27 | 陕西法士特齿轮有限责任公司 | Method and system for determining multi-index abnormal sound judgment threshold value and computer equipment |
| CN111832910B (en) * | 2020-06-24 | 2024-03-12 | 陕西法士特齿轮有限责任公司 | Multi-index abnormal sound judgment threshold value determining method, system and computer equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108830281B (en) | 2021-09-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103035013B (en) | A kind of precise motion shadow detection method based on multi-feature fusion | |
| CN103632158B (en) | Forest fire prevention monitor method and forest fire prevention monitor system | |
| CN104809463B (en) | A kind of high-precision fire disaster flame detection method for converting dictionary learning based on intensive scale invariant feature | |
| CN110148162A (en) | A kind of heterologous image matching method based on composition operators | |
| CN108765470A (en) | One kind being directed to the improved KCF track algorithms of target occlusion | |
| CN103544499B (en) | The textural characteristics dimension reduction method that a kind of surface blemish based on machine vision is detected | |
| CN105224921A (en) | A kind of facial image preferentially system and disposal route | |
| CN107579846B (en) | Cloud computing fault data detection method and system | |
| CN110580510B (en) | Clustering result evaluation method and system | |
| CN115272652A (en) | Dense object image detection method based on multiple regression and adaptive focus loss | |
| CN112488211A (en) | Fabric image flaw classification method | |
| CN109711267A (en) | A kind of pedestrian identifies again, pedestrian movement's orbit generation method and device | |
| CN112288758B (en) | Infrared and visible light image registration method for power equipment | |
| CN109509188A (en) | A kind of transmission line of electricity typical defect recognition methods based on HOG feature | |
| CN109472770A (en) | A kind of image characteristic point Fast Match Algorithm in printed circuit board (PCB) detecting | |
| CN103366373B (en) | Multi-time-phase remote-sensing image change detection method based on fuzzy compatible chart | |
| CN108830281A (en) | A kind of multiimage matching process based on localized variation detection and spatial weighting | |
| CN113191359A (en) | Small sample target detection method and system based on support and query samples | |
| CN109583307A (en) | A kind of Cashmere and Woolens fiber recognition method based on local feature Yu word packet model | |
| CN108986083A (en) | SAR image change detection based on threshold optimization | |
| CN106326927B (en) | A kind of shoes print new category detection method | |
| CN108257148A (en) | The target of special object suggests window generation method and its application in target following | |
| CN107886060A (en) | Pedestrian's automatic detection and tracking based on video | |
| CN108710886B (en) | Repeated image matching method based on SIFT algorithm | |
| KR101133225B1 (en) | System and method for recognizing face using pose estimation |
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 |