CN105654029A - Three-dimensional point cloud auricle identification method for increasing identification precision and efficiency - Google Patents
Three-dimensional point cloud auricle identification method for increasing identification precision and efficiency Download PDFInfo
- Publication number
- CN105654029A CN105654029A CN201510856611.0A CN201510856611A CN105654029A CN 105654029 A CN105654029 A CN 105654029A CN 201510856611 A CN201510856611 A CN 201510856611A CN 105654029 A CN105654029 A CN 105654029A
- Authority
- CN
- China
- Prior art keywords
- auricle
- model
- point cloud
- dimensional
- matrix
- 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
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- 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
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/64—Three-dimensional objects
- G06V20/653—Three-dimensional objects by matching three-dimensional models, e.g. conformal mapping of Riemann surfaces
Abstract
The invention discloses a three-dimensional point cloud auricle identification method for increasing identification precision and efficiency. The method comprises the following steps of based on PCA and SVD decomposition, carrying out normalization pretreatment on a position and an attitude of a three-dimensional auricle point cloud model; based on an Iannarelli classification system, extracting 4 local characteristic areas of the three-dimensional auricle point cloud model; using a Sparse ICP algorithm to match the local characteristic areas of the three-dimensional auricle point cloud model so as to acquire all the initial matching point pairs; based on the SVD, solving rotation transformation R and translation transformation T between matching point pairs and a minimum error function E till that a distance error between the models is a minimum value, and then acquiring optimal registration.
Description
Technical field
The present invention relates to a kind of three-dimensional point cloud Ear recognition method, especially a kind of three-dimensional point cloud Ear recognition method improving accuracy of identification and efficiency.
Background technology
Auricle is made up of the part such as helix, antihelix, fossa helicis, triangle nest, tragus, antitragus, ear-lobe, has the three-dimensional shape features of significant undulating profile and uniqueness, has more ubiquity, Du Texing, permanent and easy collection property. Ear recognition technology is biological identification technology emerging in recent years, and three-dimensional shape features identifies mainly to utilize the gully of people's ear profile tortuous etc., and human biological's features such as same palm, fingerprint, iris, DNA are the same, are permanent biological labels. Compared with two dimension Ear recognition technology, three-dimensional auricle attitude affects less by extraneous factors such as illumination, has significant stalwartness advantage. The three-dimensional shape features of auricle is not by the impact of the factors such as hair style, expression, beard, cosmetic, glasses, the colour of skin, illumination, and between 7 ~ 70 years old, obvious change can not occur for the structure of human pinnae and profile; Even if also there is measurable difference in the auricle of twins; Therefore, compared with other mankind's biological characteristics, the three-dimensional shape features of auricle has stability and the uniqueness of height.
Three-dimensional Ear recognition generally comprises three steps: auricle detection, feature are extracted and characteristic matching, the key problem that wherein feature is extracted, characteristic matching is three-dimensional Ear recognition. The full auricle of three-dimensional auricle model is mated by existing three-dimensional Ear recognition mostly based on iterative closest point (iterativeclosestpoint, ICP) algorithm. ICP algorithm is by the position of loop iteration, repeatedly intense adjustment point cloud model and attitude, the overall registration error of minimumization, thus realizes overall optimum matching. But, ICP algorithm computation complexity is higher, and efficiency is lower and accuracy of identification has much room for improvement.
The Iannarelli categorizing system in two dimensional image space is the anatomic characteristics based on auricle, calculate the similarity of auricle important component part, represent with the line segment on two dimensional image, as shown in Figure 1: line segment 1 ~ 4 is the width of outer helix, line segment 5 is the length of the triangle nest of part, line segment 6 ~ 8 for the distance between helix to outer helix, line segment 9 ~ 11 be the distance between helix pin to antihelix, line segment 12 is the length of ear-lobe.But the method not only on 2d each several part size all need hand dipping, cannot accurately locate, simultaneously can not directly apply to three-dimensional point cloud Ear recognition.
Summary of the invention
The present invention is to solve the above-mentioned technical problem existing for prior art, it is provided that a kind of three-dimensional point cloud Ear recognition method improving accuracy of identification and efficiency.
The technical solution of the present invention is: a kind of three-dimensional point cloud Ear recognition method improving accuracy of identification and efficiency, it is characterised in that carry out in accordance with the following steps:
A. Based PC A and SVD decomposes, and position and attitude to three-dimensional auricle point cloud model are normalized pre-treatment;
B. 4 local characteristic region of three-dimensional auricle point cloud model are extracted based on Iannarelli categorizing system;
C. utilize SparseICP algorithm to be mated by the local characteristic region of three-dimensional auricle point cloud model, obtain whole initial matching point pair; Solve based on SVD and between matching double points, rotate conversion R and translation conversion T, minimum error function E, until when model spacing error obtains minimum value, obtaining optimum registration.
Described a step is as follows:
If the vertex set of Arbitrary 3 D auricle point cloud model, then VRank geometric moment is:
Then three 1 rank geometric moments of V and six 2 rank geometric moments are respectively:
For any summit (x of vertex set Vi,yi,zi) do following conversion:
Thus by the barycenter of VMove to true origin;
As follows with the covariance matrix of six 2 rank geometric moment structure V:
If, namely covariance matrix M is carried out SVD decomposition, wherein U is the eigenvectors matrix of matrix M, and three main shafts of corresponding auricle model V, �� is the eigenvalue matrix of matrix M, to any summit (x of auricle model Vi,yi,zi) convert as follows:
Thus the first main shaft making auricle model aligns with y-axis, the 2nd main shaft aligns with x-axis, and the 3rd main shaft aligns with z-axis, it is achieved in database, all auricle point cloud models all normalize to basically identical attitude and position.
Described b step is as follows:
Using the first main shaft of Arbitrary 3 D auricle point cloud model and the 2nd main shaft as one pair of direction, simultaneously using the diagonal angle of these two main shafts as other one to direction, then do normal plane along these 4 directions and friendship asked by auricle model, finally extract 4 local characteristic region in auricle.
Described step c is as follows:
If X and Y is respectively three-dimensional auricle point cloud with reference to 4 local characteristic region of model and three-dimensional auricle point cloud target model, the alignment error vector between note X and Y is, then, the distance between note X and Y is, wherein, and, and the error function remembering X and Y between model is the lp norm of alignment error vector:
Wherein,;
Order is with reference to the upper some x of model XiThe point y nearest with this point of distance on target model YjComposition matching double points, iterative computation, obtains the whole initial matching points pair between X and Y;
Solve based on SVD and between matching double points, rotate conversion R and translation conversion T, minimum error function E, until when model spacing error obtains minimum value, obtaining optimum registration.
Described step c can also in accordance with the following steps:
If X and Y is respectively three-dimensional auricle point cloud with reference to 4 local characteristic region of model and three-dimensional auricle point cloud target model, the alignment error vector between note X and Y is, then, the distance between note X and Y is, wherein, and, and the error function remembering X and Y between model is the lp norm of alignment error vector, the lp norm of alignment error vector is defined as based on Lagrangian method:
Wherein,,For Lagrangian multiplier,For the punishment factor, utilize and exchange direction multiplier method, be broken down into 3 subproblems:
(1)
(2)
(3)
Wherein,,, pass through hiVector shrinks operator and is solved by above-mentioned subproblem, and algorithm flow is as follows:
Step 1. is with reference to 4 local characteristic region X, Y initial alignment of model and target model, and initialize Lagrange is taken advantage of, punishment factor mu and threshold tau;
Step 2. is for any point of X, and the Europe formula on Y of calculating, apart from the nearest point of this point, obtains whole initial matching point pair between X and Y;
Step 3. solves, pass through hiVector shrinks operator and solves, upgrade vector;
Step 4. upgrades, solved function, and calculate rotation matrix R and translation matrix t;
Step 5. solves, upgrade;
If step 6.It is greater than threshold tau, returns step 2, otherwise iteration terminates.
The present invention adopts principle component analysis (PrincipalComponentAnalysis, and SVD decomposes the position to all three-dimensional auricle point cloud models in auricle database, attitude is normalized PCA), then four local characteristic region in three-dimensional auricle model are extracted based on Iannarelli categorizing system, finally utilize SparseICP algorithm to be mated by the local characteristic region extracted, realize Ear recognition according to the divergence measurement between the Distance Judgment auricle between unique point. Experimental result shows, auricle local characteristic region is mated and has higher accuracy of identification and recognition efficiency by the present invention. Particularly introduce sparse degree to Model registration optimization, and utilizeNorm replaces Europe formula distance, and maximumization corresponding points spacing is the quantity of zero, avoids the problems such as local alignment and occurs. Simultaneously, algorithm adopts Lagrangian method to be redefined by error function, the problems such as the model surface nonconvex property that solution alignment thereof causes and non-smoothness, and utilize exchange direction multiplier method that the error function redefined is divided into three simple subproblems, and solve by shrinking operator, it is to increase the precision and stability of algorithm.
Accompanying drawing explanation
Fig. 1 is the Iannarelli categorizing system auricle segmentation schematic diagram in two dimensional image space.
Fig. 2 is contrast effect figure before and after the three-dimensional auricle point cloud model normalization method of the embodiment of the present invention.
Fig. 3 is that the embodiment of the present invention extracts auricle model 4 sub regions schematic diagram.
Fig. 4 is the embodiment of the present invention and prior art matching precision contrast schematic diagram.
Fig. 5 is the CMC curve comparison schematic diagram of the embodiment of the present invention and ICP algorithm.
Fig. 6 is the ROC curve contrast schematic diagram of the embodiment of the present invention and ICP algorithm.
Embodiment
The specific embodiment of the invention carries out on the three-dimensional auricle database of UND, and this database comprises the three-dimensional auricle of 1800 width from 415 people.
Specifically carry out in accordance with the following steps:
A. Based PC A and SVD decomposes, and position and attitude to three-dimensional auricle point cloud model are normalized pre-treatment:
Owing to three-dimensional auricle data set structure time span is bigger, the same auricle data obtained on different acquisition times by be subject between collected auricle and collection equipment distance, the factor such as angle impact, therefore first need the auricle model in three-dimensional auricle database is normalized pre-treatment.
Concrete steps are as follows:
If the vertex set of Arbitrary 3 D auricle point cloud model, then VRank geometric moment is:
Then three 1 rank geometric moments of V and six 2 rank geometric moments are respectively:
For any summit (x of vertex set Vi,yi,zi) do following conversion:
Thus by the barycenter of VMove to true origin;
As follows with the covariance matrix of six 2 rank geometric moment structure V:
If, namely covariance matrix M is carried out SVD decomposition, wherein U is the eigenvectors matrix of matrix M, and three main shafts of corresponding auricle model V, �� is the eigenvalue matrix of matrix M, to any summit (x of auricle model Vi,yi,zi) convert as follows:
Thus the first main shaft making auricle model aligns with y-axis, the 2nd main shaft aligns with x-axis, and the 3rd main shaft aligns with z-axis, it is achieved in database, all auricle point cloud models all normalize to basically identical attitude and position.
Before and after three-dimensional auricle point cloud model normalization method, contrast effect is as shown in Figure 2, and wherein a is the schematic diagram before three-dimensional auricle point cloud model normalization method; B is the design sketch before three-dimensional auricle point cloud model normalization method.
B. 4 local characteristic region of three-dimensional auricle point cloud model are extracted based on Iannarelli categorizing system:
Using the first main shaft of Arbitrary 3 D auricle point cloud model and the 2nd main shaft as one pair of direction, simultaneously using the diagonal angle of these two main shafts as other one to direction, then do normal plane along these 4 directions and friendship asked by auricle model, finally extract 4 local characteristic region in auricle.
As shown in Figure 3: these 4 local characteristic region contain 12 important geometric properties in Iannarelli categorizing system.
C. utilize SparseICP algorithm to be mated by the local characteristic region of three-dimensional auricle point cloud model, obtain whole initial matching point pair; Solve based on SVD and between matching double points, rotate conversion R and translation conversion T, minimum error function E, until when model spacing error obtains minimum value, obtaining optimum registration.
Specifically can as follows:
If X and Y is respectively three-dimensional auricle point cloud with reference to 4 local characteristic region of model and three-dimensional auricle point cloud target model, the alignment error vector between note X and Y is, then, the distance between note X and Y is, wherein, and, and the error function remembering X and Y between model is the lp norm of alignment error vector:
Wherein,;
Order is with reference to the upper some x of model XiThe point y nearest with this point of distance on target model YjComposition matching double points, iterative computation, obtains the whole initial matching points pair between X and Y;
Solve based on SVD and between matching double points, rotate conversion R and translation conversion T, minimum error function E, until when model spacing error obtains minimum value, obtaining optimum registration.
In order to optimize registration problems further, solving in minimum error function, based on the l of Lagrangian method by alignment error vectorpNorm is newly defined as:
Wherein,For Lagrangian multiplier,For the punishment factor.
Utilize and exchange direction multiplier method (alternatingdirectionmethodofmultipliers, ADMM), be broken down into 3 subproblems:
(1)
(2)
(3)
Wherein,,. Pass through hiVector shrinks operator (shrinkageoperator) and is solved by above-mentioned subproblem, and algorithm flow is as follows:
Step 1. is with reference to 4 local characteristic region X, Y initial alignment of model and target model, and initialize Lagrange is taken advantage of, punishment factor mu and threshold tau;
Step 2. is for any point of X, and the Europe formula on Y of calculating, apart from the nearest point of this point, obtains whole initial matching point pair between X and Y;
Step 3. solves, pass through hiVector shrinks operator and solves, upgrade vector;
Step 4. upgrades, solved function, and calculate rotation matrix R and translation matrix t;
Step 5. solves, upgrade;
If step 6.It is greater than threshold tau, returns step 2, otherwise iteration terminates.
Experimental result and analysis:
One. matching precision
Fig. 4 is the Europe formula square distance calculated after auricle model SparseICP algorithmic match in auricle model 05129d002ear to be measured and database and similarity distribution, wherein lines a represent the embodiment of the present invention use extract auricle region mate after result, lines b represent use complete auricle model mate after result. As can be seen from Figure 4, the embodiment of the present invention adopts the auricle model accuracy of regional area coupling higher.
Equally, taking 05129d002ear as auricle model to be measured, in database, other auricle models mate, mean distance between matching double points between computation model, that is:
Wherein, n represents the number of matching double points; MinDis represents the distance between matching double points; I represents the sequence number of matching double points; K represents the sequence number in region.Obviously, DisAvg is more little, and two distortions are more high; DisAvg is more big, and two distortions are more low. The mean distance in average each innings of territory, can obtain the similarity between auricle model. As shown in table 1. Obviously, the embodiment of the present invention is compared with ICP, ICNP algorithm, and precision is higher, and can effectively distinguish the model of the different data of identical auricle.
Between table 1 auricle, mean distance compares
Adopt cumulative matches characteristic (cumulativematchcharacteristics, CMC) curve carrys out the recognition efficiency of assessment algorithm, wherein the X-coordinate of CMC curve represents the model mating best front k in auricle Matching Experiment result, and ordinate zou represents the accuracy for three-dimensional Ear recognition. The CMC curve of the embodiment of the present invention and ICP algorithm is contrasted by Fig. 5. As can be seen from this figure, embodiment of the present invention recognition rate is higher than the recognition rate of ICP algorithm, and wherein rank-1 reaches 93.8%.
Adopt receiver operating characteristic (receiveroperatingcharacteristic, ROC) curve reflection correctly accepts rate (genuineacceptancerate, GAR) with false acceptance rate (falseacceptancerate, FAR) mutual relationship, the auricle that GAR represents correct is regarded as correct percentage, and FAR represents that the auricle of mistake is regarded as correct percentage. The ROC curve of the embodiment of the present invention and ICP algorithm is contrasted by Fig. 6. As can be seen from this figure, the recognition rate of the embodiment of the present invention is higher than the recognition rate of ICP algorithm.
Two. the coupling time
This experiment is based on Intel (R) Xeon (R) CPU of 2.40GHz, 16.0GBRAM, in the computing environment of 64 bit manipulation systems, 1800 three-dimensional auricle models in the three-dimensional auricle database of UND are extracted auricle character zone respectively, and compared for and utilize complete auricle model to carry out SparseICP, ICP, the match condition of ICNP algorithm, corresponding coupling used time contrast is as shown in table 2, can find out, the coupling used time based on auricle local characteristic region of the embodiment of the present invention is obviously less than the coupling used time based on complete auricle model, and as compared to ICP with ICNP algorithm, SparseICP algorithm time loss is less.
Table 2 mates the time and compares (s)
Conclusion: experiment proves, the embodiment of the present invention has very high accuracy of identification and efficiency compared with other algorithms.
Claims (5)
1. one kind can be improved the three-dimensional point cloud Ear recognition method of accuracy of identification and efficiency, it is characterised in that carry out in accordance with the following steps:
A. Based PC A and SVD decomposes, and position and attitude to three-dimensional auricle point cloud model are normalized pre-treatment;
B. 4 local characteristic region of three-dimensional auricle point cloud model are extracted based on Iannarelli categorizing system;
C. utilize SparseICP algorithm to be mated by the local characteristic region of three-dimensional auricle point cloud model, obtain whole initial matching point pair; Solve based on SVD and between matching double points, rotate conversion R and translation conversion T, minimum error function E, until when model spacing error obtains minimum value, obtaining optimum registration.
2. the three-dimensional point cloud Ear recognition method based on SparseICP according to claim 1, it is characterised in that described a step is as follows:
If the vertex set of Arbitrary 3 D auricle point cloud model, then VRank geometric moment is:
Then three 1 rank geometric moments of V and six 2 rank geometric moments are respectively:
For any summit (x of vertex set Vi,yi,zi) do following conversion:
Thus by the barycenter of VMove to true origin;
As follows with the covariance matrix of six 2 rank geometric moment structure V:
If, namely covariance matrix M is carried out SVD decomposition, wherein U is the eigenvectors matrix of matrix M, and three main shafts of corresponding auricle model V, �� is the eigenvalue matrix of matrix M, to any summit (x of auricle model Vi,yi,zi) convert as follows:
Thus the first main shaft making auricle model aligns with y-axis, the 2nd main shaft aligns with x-axis, and the 3rd main shaft aligns with z-axis, it is achieved in database, all auricle point cloud models all normalize to basically identical attitude and position.
3. the three-dimensional point cloud Ear recognition method based on SparseICP according to claim 1, it is characterised in that described b step is as follows:
Using the first main shaft of Arbitrary 3 D auricle point cloud model and the 2nd main shaft as one pair of direction, simultaneously using the diagonal angle of these two main shafts as other one to direction, then do normal plane along these 4 directions and friendship asked by auricle model, finally extract 4 local characteristic region in auricle.
4. the three-dimensional point cloud Ear recognition method based on SparseICP according to claim 1, it is characterised in that described step c is as follows:
If X and Y is respectively three-dimensional auricle point cloud with reference to 4 local characteristic region of model and three-dimensional auricle point cloud target model, the alignment error vector between note X and Y is, then, the distance between note X and Y is, wherein, and, and the error function remembering X and Y between model is the lp norm of alignment error vector:
Wherein,;
Order is with reference to the upper some x of model XiThe point y nearest with this point of distance on target model YjComposition matching double points, iterative computation, obtains the whole initial matching points pair between X and Y;
Solve based on SVD and between matching double points, rotate conversion R and translation conversion T, minimum error function E, until when model spacing error obtains minimum value, obtaining optimum registration.
5. the three-dimensional point cloud Ear recognition method based on SparseICP according to claim 1, it is characterised in that described step c is as follows:
If X and Y is respectively three-dimensional auricle point cloud with reference to 4 local characteristic region of model and three-dimensional auricle point cloud target model, the alignment error vector between note X and Y is, then, the distance between note X and Y is, wherein, and, and the error function remembering X and Y between model is the lp norm of alignment error vector, the lp norm of alignment error vector is defined as based on Lagrangian method:
Wherein,,For Lagrangian multiplier,For the punishment factor, utilize and exchange direction multiplier method, be broken down into 3 subproblems:
(1)
(2)
(3)
Wherein,,, pass through hiVector shrinks operator and is solved by above-mentioned subproblem, and algorithm flow is as follows:
Step 1. is with reference to 4 local characteristic region X, Y initial alignment of model and target model, and initialize Lagrange is taken advantage of, punishment factor mu and threshold tau;
Step 2. is for any point of X, and the Europe formula on Y of calculating, apart from the nearest point of this point, obtains whole initial matching point pair between X and Y;
Step 3. solves, pass through hiVector shrinks operator and solves, upgrade vector;
Step 4. upgrades, solved function, and calculate rotation matrix R and translation matrix t;
Step 5. solves, upgrade;
If step 6.It is greater than threshold tau, returns step 2, otherwise iteration terminates.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510856611.0A CN105654029B (en) | 2015-11-30 | 2015-11-30 | The three-dimensional point cloud Ear recognition method of accuracy of identification and efficiency can be improved |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510856611.0A CN105654029B (en) | 2015-11-30 | 2015-11-30 | The three-dimensional point cloud Ear recognition method of accuracy of identification and efficiency can be improved |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105654029A true CN105654029A (en) | 2016-06-08 |
CN105654029B CN105654029B (en) | 2019-04-05 |
Family
ID=56481905
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510856611.0A Expired - Fee Related CN105654029B (en) | 2015-11-30 | 2015-11-30 | The three-dimensional point cloud Ear recognition method of accuracy of identification and efficiency can be improved |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105654029B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106970956A (en) * | 2017-03-16 | 2017-07-21 | 天津大学 | A kind of method for searching three-dimension model based on tensor |
CN108871284A (en) * | 2018-05-08 | 2018-11-23 | 中国矿业大学 | Three-dimensional space similarity transformation model parameter based on line feature constraint without initial value method for solving |
CN109564693A (en) * | 2016-08-10 | 2019-04-02 | 株式会社高永科技 | Three-dimensional data integrating apparatus and method |
CN110222580A (en) * | 2019-05-09 | 2019-09-10 | 中国科学院软件研究所 | A kind of manpower 3 d pose estimation method and device based on three-dimensional point cloud |
CN111127505A (en) * | 2019-11-27 | 2020-05-08 | 天津津航技术物理研究所 | Online learning tracking and engineering realization method based on space planning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080143712A1 (en) * | 2006-12-19 | 2008-06-19 | Mcbagonluri Fred | Intelligent modeling method and system for earmold shell and hearing aid design |
CN101770566A (en) * | 2008-12-30 | 2010-07-07 | 复旦大学 | Quick three-dimensional human ear identification method |
CN103810751A (en) * | 2014-01-29 | 2014-05-21 | 辽宁师范大学 | Three-dimensional auricle point cloud shape feature matching method based on IsoRank algorithm |
CN105069403A (en) * | 2015-07-20 | 2015-11-18 | 同济大学 | Three-dimensional ear recognition based on block statistic features and dictionary learning sparse representation classification |
-
2015
- 2015-11-30 CN CN201510856611.0A patent/CN105654029B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080143712A1 (en) * | 2006-12-19 | 2008-06-19 | Mcbagonluri Fred | Intelligent modeling method and system for earmold shell and hearing aid design |
CN101770566A (en) * | 2008-12-30 | 2010-07-07 | 复旦大学 | Quick three-dimensional human ear identification method |
CN103810751A (en) * | 2014-01-29 | 2014-05-21 | 辽宁师范大学 | Three-dimensional auricle point cloud shape feature matching method based on IsoRank algorithm |
CN105069403A (en) * | 2015-07-20 | 2015-11-18 | 同济大学 | Three-dimensional ear recognition based on block statistic features and dictionary learning sparse representation classification |
Non-Patent Citations (1)
Title |
---|
王森 等: "基于Sparse ICP的三维点云耳廓识别", 《图学学报》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109564693A (en) * | 2016-08-10 | 2019-04-02 | 株式会社高永科技 | Three-dimensional data integrating apparatus and method |
US11481867B2 (en) | 2016-08-10 | 2022-10-25 | Koh Young Technology Inc. | Device and method for registering three-dimensional data |
CN109564693B (en) * | 2016-08-10 | 2024-04-09 | 株式会社高迎科技 | Three-dimensional data integration device and method |
CN106970956A (en) * | 2017-03-16 | 2017-07-21 | 天津大学 | A kind of method for searching three-dimension model based on tensor |
CN108871284A (en) * | 2018-05-08 | 2018-11-23 | 中国矿业大学 | Three-dimensional space similarity transformation model parameter based on line feature constraint without initial value method for solving |
CN110222580A (en) * | 2019-05-09 | 2019-09-10 | 中国科学院软件研究所 | A kind of manpower 3 d pose estimation method and device based on three-dimensional point cloud |
CN110222580B (en) * | 2019-05-09 | 2021-10-22 | 中国科学院软件研究所 | Human hand three-dimensional attitude estimation method and device based on three-dimensional point cloud |
CN111127505A (en) * | 2019-11-27 | 2020-05-08 | 天津津航技术物理研究所 | Online learning tracking and engineering realization method based on space planning |
CN111127505B (en) * | 2019-11-27 | 2024-03-26 | 天津津航技术物理研究所 | Online learning tracking and engineering realization method based on space planning |
Also Published As
Publication number | Publication date |
---|---|
CN105654029B (en) | 2019-04-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107194341B (en) | Face recognition method and system based on fusion of Maxout multi-convolution neural network | |
CN106682598B (en) | Multi-pose face feature point detection method based on cascade regression | |
CN105654029A (en) | Three-dimensional point cloud auricle identification method for increasing identification precision and efficiency | |
Gilani et al. | Deep, dense and accurate 3D face correspondence for generating population specific deformable models | |
CN103218609B (en) | A kind of Pose-varied face recognition method based on hidden least square regression and device thereof | |
CN102013011B (en) | Front-face-compensation-operator-based multi-pose human face recognition method | |
Ren et al. | Facial expression recognition based on AAM–SIFT and adaptive regional weighting | |
CN104598879A (en) | Three-dimensional face recognition method based on face contour lines of semi-rigid areas | |
Zhao et al. | Bisecting k-means clustering based face recognition using block-based bag of words model | |
CN103727930A (en) | Edge-matching-based relative pose calibration method of laser range finder and camera | |
CN104299003A (en) | Gait recognition method based on similar rule Gaussian kernel function classifier | |
CN109871818B (en) | Face recognition method based on normal vector distribution histogram and covariance descriptor | |
Ming | Rigid-area orthogonal spectral regression for efficient 3D face recognition | |
CN108734187A (en) | A kind of multiple view spectral clustering based on tensor singular value decomposition | |
CN101609507A (en) | Gait recognition method | |
CN104408405A (en) | Face representation and similarity calculation method | |
CN101770566A (en) | Quick three-dimensional human ear identification method | |
Fang et al. | Real-time hand posture recognition using hand geometric features and fisher vector | |
CN103984922A (en) | Face identification method based on sparse representation and shape restriction | |
Soltanpour et al. | Multimodal 2D–3D face recognition using local descriptors: pyramidal shape map and structural context | |
CN108182705A (en) | A kind of three-dimensional coordinate localization method based on machine vision | |
CN114648445B (en) | Multi-view high-resolution point cloud splicing method based on feature point extraction and fine registration optimization | |
Alsawwaf et al. | In your face: person identification through ratios and distances between facial features | |
CN105550642B (en) | Gender identification method and system based on multiple dimensioned linear Differential Characteristics low-rank representation | |
Hajati et al. | Surface geodesic pattern for 3D deformable texture matching |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
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: 20190405 Termination date: 20191130 |