CN102880877A - Target identification method based on contour features - Google Patents

Target identification method based on contour features Download PDF

Info

Publication number
CN102880877A
CN102880877A CN2012103724595A CN201210372459A CN102880877A CN 102880877 A CN102880877 A CN 102880877A CN 2012103724595 A CN2012103724595 A CN 2012103724595A CN 201210372459 A CN201210372459 A CN 201210372459A CN 102880877 A CN102880877 A CN 102880877A
Authority
CN
China
Prior art keywords
detected
image
point
edge
profile
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
Application number
CN2012103724595A
Other languages
Chinese (zh)
Other versions
CN102880877B (en
Inventor
赵雪专
陈斌
张绍兵
裴利沈
廖世鹏
成苗
何莲
张元�
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chengdu Information Technology Co Ltd of CAS
Original Assignee
Chengdu Information Technology Co Ltd of CAS
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chengdu Information Technology Co Ltd of CAS filed Critical Chengdu Information Technology Co Ltd of CAS
Priority to CN201210372459.5A priority Critical patent/CN102880877B/en
Publication of CN102880877A publication Critical patent/CN102880877A/en
Application granted granted Critical
Publication of CN102880877B publication Critical patent/CN102880877B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Image Analysis (AREA)

Abstract

The invention discloses a target identification method based on contour features. The method comprises steps of establishing a feature bank of object template contours; extracting a complete contour of an object template; extracting feature points and central points from the contour; establishing a distance matrix to describe the contour by using feature points and central points; conducting calculation of the distance matrix by aiming to all pixels on the contour; conducting target identification for an image to be detected; extracting an edge of the image to be detected; extracting feature points on the edge; calculating feature description composed of feature points; matching the feature description of the image to be detected with features in the feature bank of object template contours; estimating central points of the contour of the image to be detected; and estimating the contour of the image to be detected. Compared with the prior art, the method has the advantages that the size problem in the contour matching is solved, and the size in the contour matching is not changed. The method can be applied to target identification of images effectively.

Description

A kind of target identification method based on contour feature
Technical field
The invention belongs to machine vision and mode identification technology, particularly relate to a kind of target identification method based on the contour feature in the picture shape feature.
Background technology
The method of current goal identification can be divided three classes: based on the method for external performance, the method that combines based on the method for shape facility, based on presentation and shape facility.Target identification method based on shape facility is a focus of current research, has obtained in the last few years significant progress.Based on Fourier transform or the method such as bending moment not, develop into the current shape matching method that mostly belongs to based on profile and two kinds of shape descriptors of skeleton from traditional.With respect to point set, profile has more abundant information, and profile is not easy to be subject to the impact of illumination, object color and texture variations, the most important thing is effectively to describe the object structures of large scale scope.
A lot of Target detection and identification methods based on profile were suggested in the last few years, wherein mainly comprised two large steps: descriptor and the similarity between the descriptor of extracting profile are calculated.Based on this 2 point, scholar both domestic and external conducts in-depth research this.The people such as Bai Xiang describe the topological structure of object what international computer vision conference (IEEE International Conference on Computer Vision 2009) proposed with skeleton, unite tree the profile of various piece is effectively organized by setting up skeleton, and then with the shape similarity calculating of carrying out profile with the Chamfer of direction coupling.The people such as MA propose a kind of new shape descriptor based on the Shape context geometry in international computer vision and pattern-recognition meeting (IEEE Conference on Computer Vision and Pattern Recognition 2011), the coupling of profile are converted into the similarity computational problem of matrix.The people such as Wang are proposing fan-shaped model in international computer vision and pattern-recognition meeting (IEEE Conference on Computer Vision and Pattern Recognition 2012), unique point on the profile and the line of central point are equivalent to the lath of fan, and the deformation of satisfying object has certain robustness.Also propose simultaneously a kind of method of determining fast target scale, and then put and put the coupling between the feature.
Yet said method all fails to find a kind of to the simple and effective description of profile, to solve the scale problem in the outline process.
Summary of the invention
For the problem that prior art exists, it is a kind of at the constant target identification method based on contour feature of matching process mesoscale that fundamental purpose of the present invention is to provide.
For achieving the above object, the invention provides a kind of embodiment of the target identification method based on contour feature, the method comprises the steps: step 1, sets up the feature database of object template contours; Step 2, treat detected image and carry out target identification.
Wherein the step 1 feature database of setting up the object template contours comprises following steps (1.1)～step (1.4):
(1.1) integrity profile of extract phantom plate;
(1.2) central point of extract minutiae and profile on the profile of object template;
(1.3) utilizing unique point and central point to set up distance matrix is described profile;
(1.4) carry out the calculating of distance matrix for all pixels on the profile, the set of result of calculation is the feature database of object template contours.
Step 2 is treated detected image and is carried out target identification and comprise following steps (2.1)～step (2.6):
(2.1) edge of extraction image to be detected;
(2.2) extract minutiae on the edge of image to be detected;
(2.3) calculate the feature description that image characteristic point to be detected forms;
(2.4) Characteristic of Image description to be detected is mated with the feature in the object template contours feature database;
(2.5) central point of estimation image outline to be detected;
(2.6) distance relation of central point and profile in the central point of estimating according to image outline to be detected and the distance relation at correspondence image edge and the object template contours estimates the profile of image to be detected.
Utilize the integrity profile of background subtraction method extract phantom plate in the step (1.1), the result of profile is expressed as S={P iI=1...N}, wherein S represents the profile of object template, P iPixel on the expression profile, N represents the number of pixel in the profile.
In the step (1.2) on the profile of object template the central point of extract minutiae and profile comprise the steps: that (1) is to the some P on the profile i(x i, y i) calculate the characteristic of correspondence point that it satisfies certain condition, specifically comprise following steps: establish the arbitrfary point P on the object template contours j(x j, y j), get a P k(x k, y k), wherein k=2*j-i obtains D represents a P jTo a P iWith a P kOrganize straight distance, Ax+By+C=0 represents a P iWith a P kThe straight-line equation at place, wherein A=y k-y j, B=x i-x k, C=x ky j-x iy k, some P iTo a P kDistance can be expressed as:
Figure BDA00002210895600032
Figure BDA00002210895600033
T is d and d IkThe ratio of these two distances when t 〉=T, is then got the most close P on the order iThe P of point jPoint is P iCharacteristic of correspondence point, T are the threshold parameter of setting according to different objects; (2) on the profile have a few and calculate successively the characteristic of correspondence point, obtain corresponding relation C table: C={P i, P mI=1...N, m ∈ 1 ..., N}}, wherein P mBe P iThe characteristic of correspondence point, N represents the number of pixel in the profile; (3) central point of calculating profile is: P center ( x , y ) = 1 N Σ i = 1 N P i ( x i , y i ) .
Step (1.3) is utilized unique point and central point to set up distance matrix and profile is described is specifically comprised following steps: calculate with the some P on the profile (1) iBe the Description Matrix D in the object template contours feature database of starting point i, specifically comprise following steps: utilize corresponding relation C table to find P iCharacteristic of correspondence point P jUtilize corresponding relation C table to find P jCharacteristic of correspondence point P kCalculate D i = d i , i d i , j d i , k d j , i d j , j d j , k d k , i d k , j d k , k , D iFor with P iBe starting point, P kBe the profile Description Matrix of terminal point, wherein
Figure BDA00002210895600042
d I, jP iAnd P jEuclidean distance between the point is in the distance expression of log space.(2) calculate each unique point P on the profile iAnd center point P CenterRelation, with vector representation be
Figure BDA00002210895600043
Wherein N represents the number of pixel in the profile.
The edge that step (2.1) is extracted image to be detected specifically comprises following steps: the image transitions to be detected that (1) will collect is gray-scale map; (2) treat detected image with the canny operator and carry out edge extracting; (3) with edge linking algorithm the edge of close together is coupled together, obtain the edge aggregation E={e (i) of image to be detected; I=1...M}, wherein M represents the number at edge.
Step (2.2) extract minutiae on the edge of image to be detected specifically comprises following steps: (1) is for any edge e (i) of image to be detected, on the edge have a few and calculate successively the characteristic of correspondence point, obtain the corresponding relation C table of edge e (i); (2) treat detected image edge aggregation E={e (i); All edges among the i=1...M} carry out respectively the calculating of unique point, and all edges are obtained corresponding corresponding relation C table, E c={ e c(i); I=1...M}, wherein M represents the number at edge, and the corresponding relation C of edge i is expressed as e c(i)={ P j, P mJ=1...N i, m ∈ 1 ..., N i, N wherein iRepresent the number of the upper pixel of edge i, P mBe P jThe characteristic of correspondence point.
The feature description that step (2.3) is calculated image characteristic point composition to be detected specifically comprises following steps: calculate with the some P on the edge i (1) m(i) be the feature description of starting point, specifically comprise following steps: utilize corresponding relation C table set E cIn e c(i) find P m(i) characteristic of correspondence point P j(i); Utilize corresponding relation C table e c(i) find P j(i) characteristic of correspondence point P k(i); Calculate D e ( i ) m = d m , m d m , j d m , k d j , m d j , j d j , k d k , m d k , j d k , k ,
Figure BDA00002210895600052
For edge i goes up with P m(i) be starting point, P k(i) be the profile description of terminal point, wherein d I, j=log (1+||P i-P j|| 2), d I, jP iAnd P jEuclidean distance between the point is in the distance expression of log space; (2) edge set E={e (i); Each edge among the i=1...M} respectively calculated characteristics is described,
Figure BDA00002210895600053
Wherein Be illustrated on the edge i feature Description Matrix of some j place's beginning, N iRepresent the number of the upper pixel of edge i, M represents the number at edge.
Step (2.4) with Characteristic of Image to be detected describe with object template contours feature database in feature mate mainly and to carry out respectively similarity by the Characteristic of Image Description Matrix to be detected that will extract and the Description Matrix in the object template contours feature database and calculate to realize that key step is as follows: (1) Characteristic of Image Description Matrix to be detected is D e, the Description Matrix in the object template contours feature database is D m, to two matrix D e, D mThe diagonal line both sides carry out respectively normalized, to D eObtain after the processing Wherein
Figure BDA00002210895600056
D eSome i on (i, j) expression image outline e to be detected is to the Euclidean distance of some j, to D mObtain after the processing D m ′ = 1 s m D m , Wherein s m = 1 2 Σ i = 1 3 Σ j = 1 3 D m ( i , j ) , D mSome i on (i, j) expression object template contours m is to the Euclidean distance of some j; (2) to matrix D ' eAnd matrix D ' mCarry out similarity and calculate, the value of similarity is A ( i , j ) = max ( exp ( - ( D e ′ ( i , j ) - D m ′ ( i , j ) ) 2 ( D e ′ ( i , j ) σ ) 2 ) , exp ( - ( D e ′ ( i , j ) - D m ′ ( 3 - i , 3 - j ) ) 2 ( D e ′ ( i , j ) σ ) 2 ) , I wherein, j=1,2,3, i ≠ j; σ is the punishment parameter of error of adjusting the distance, when i=j, and A (i, j)=1, then the value of similarity is: ψ ( D e ′ , D m ′ ) = 1 9 Σ i = 1 3 Σ j = 1 3 A ( i , j ) .
Step (2.5) estimates that the central point of image outline to be detected specifically comprises following steps: (1) according to similarity value ψ (D ' e, D ' m) judgement 1-ψ (D ' e, D ' mWhether)＞ψ τ sets up, wherein ψ τThreshold parameter for the restriction similarity; (2) if 1-ψ (D ' e, D ' m)＞ψ τ, i.e. Characteristic of Image Description Matrix D to be detected eWith the Description Matrix D in the object template contours feature database mMate unsuccessfully, then give up D e(3) if 1-ψ (D ' e, D ' m)≤ψ τ, i.e. Characteristic of Image Description Matrix D to be detected eWith the Description Matrix D in the object template contours feature database mThe match is successful, then according to S e, S mAnd vector
Figure BDA00002210895600062
Estimate and form D eThree some P m(i), P j(i), P k(i) central point corresponding to difference
Figure BDA00002210895600063
P wherein iBe the arbitrfary point on the image border to be detected, P ' iFor on the image border to be detected with P iCorresponding point, P Center' be the central point of estimating; (4) calculate the marginate center point P of image to be detected Center'; (4) by the marginate center point P of image to be detected Center' be formed in the image to be detected ballot figure to central point; (5) in the statistics image to be detected to the voting results of central point, with mean shift algorithm voting results are carried out cluster, the cluster centre that obtains is the central point of image outline to be detected.
The distance relation at central point and edge in the central point that step (2.6) is estimated according to image outline to be detected and the distance relation at correspondence image edge and the object template contours, the profile that estimates image to be detected specifically comprises following steps: (1) gets the center point P of image outline to be detected Center(i), wherein i is the some targets in the image to be detected, finds with P Center(i) centered by, radius is the series of features Description Matrix D of ballot in the σ zone e, wherein σ refers to the radius value according to image size to be detected and scene setting; (2) statistics characteristics of image Description Matrix D to be detected e(i) with object template contours feature database in corresponding Description Matrix D m(j) the scaling relation V between (i) is namely according to S m(j) and S e(i) value obtains ratio Wherein e represents a certain edge of image to be detected, and i represents the starting point of characteristics of image Description Matrix to be detected on edge e, and m represents the jobbie template contours, the starting point of Description Matrix on profile m of correspondence in the j representative body template contours feature database; (3) statistics is with P Center(i) centered by, radius is all characteristics of image Description Matrix D to be detected of ballot in the σ zone eWith the Description Matrix D in the object template contours feature database mBetween scaling relation, obtain in characteristics of image Description Matrix to be detected and the object template contours feature database Description Matrix than value set V={V (i); I=1...m}, wherein the m representative is with P Center(i) centered by, radius is the quantity of the characteristics of image Description Matrix to be detected of ballot in the σ zone; (4) carry out cluster with the ratio among the mean shift algorithm pair set V, determine the ratio V (i) of the Description Matrix in characteristics of image Description Matrix to be detected and the object template contours feature database according to cluster result, determine to satisfy the characteristics of image Description Matrix D to be detected of described ratio according to this ratio V (i) e(i), according to characteristics of image Description Matrix D to be detected e(i) determine point on the edge of composition characteristic Description Matrix, these points are with P Center(i) centered by, with the point in the object template contours image outline to be detected that the match is successful; (5) according to ratio V, with P Center(i) centered by, the point in the image outline to be detected that does not match is filled the profile of the complete image to be detected that obtains estimating.
The present invention has solved the scale problem in the outline process with respect to prior art, makes the yardstick in the outline process constant, and effectively is applied to the target identification in the image.
Description of drawings
Fig. 1 is the target identification method process flow diagram based on contour feature of the present invention.
Embodiment
Below in conjunction with accompanying drawing, describe the specific embodiment of the present invention in detail.
As shown in Figure 1, the target identification method based on contour feature of the present invention is used for identifying image target to be detected, if a plurality of targets are arranged in the image to be detected, then identifies one by one the target in the image to be detected.Comprise following steps S1～S2:S1, set up the feature database of object template contours; S2, treat detected image and carry out target identification.
Wherein step S1 comprises following steps S11～S14:
The integrity profile of S11, extract phantom plate mainly under static background, utilizes the integrity profile of background subtraction method extract phantom plate, and the result of profile is expressed as S={P iI=1...N}, wherein S represents the profile of object template, P iPixel on the expression profile, N represents the number of pixel in the profile.
S12, on the profile of object template the central point of extract minutiae and profile, comprise the steps: that (1) is to the some P on the profile i(x i, y i) calculate the characteristic of correspondence point that it satisfies certain condition, specifically comprise following steps: establish the arbitrfary point P on the object template contours j(x j, y j), get a P k(x k, y k), wherein k=2*j-i obtains
Figure BDA00002210895600081
D represents a P jTo a P iWith a P kOrganize straight distance, Ax+By+C=0 represents a P iWith a P kThe straight-line equation at place, wherein A=y k-y j, B=x i-x k, C=x ky j-x iy k, some P iTo a P kDistance can be expressed as: d ik = ( x i - x k ) 2 + ( y i - y k ) 2 , t = d d ik , T is d and d IkThe ratio of these two distances when t 〉=T, is then got the most close P on the order iThe P of point jPoint is P iCharacteristic of correspondence point, T are the threshold parameter of setting according to different objects, and the setting of T is relevant with the crooked situation of object objective contour, and the value of T can be 0.12; (2) all pixels on the profile are calculated the characteristic of correspondence point successively, obtain corresponding relation C table: C={P i, P mI=1...N, m ∈ 1 ..., N}}, wherein P mBe P iThe characteristic of correspondence point, N represents the number of pixel in the profile; (3) central point of calculating profile is:
Figure BDA00002210895600084
S13, utilize unique point and central point to set up distance matrix profile to be described, to comprise following steps:
(1) calculates with the some P on the profile iBe the Description Matrix D in the object template contours feature database of starting point i, specifically comprise following steps: utilize corresponding relation C table to find P iCharacteristic of correspondence point P jUtilize corresponding relation C table to find P jCharacteristic of correspondence point P kCalculate D i = d i , i d i , j d i , k d j , i d j , j d j , k d k , i d k , j d k , k , D iFor with P iBe starting point, P kBe the profile Description Matrix of terminal point, wherein d I, j=log (1+||P i-P j|| 2), d I, j, be P iAnd P jEuclidean distance between the point is at the distance expression of log space, d I, k=log (1+||P i-P k|| 2), d I, kP iAnd P kEuclidean distance between the point is at the distance expression of log space, d J, k=log (1+||P j-P k|| 2), d J, kP jAnd P kEuclidean distance between the point is at the distance expression of log space, D iOther value in the matrix is such as d I, i, D J, i, d K, iImplication by that analogy; (2) calculate each unique point P on the profile iAnd center point P CenterRelation, namely calculate pixel P all on the profile iWith center point P CenterVector relations, with vector representation be
Figure BDA00002210895600092
Wherein N represents the number of pixel in the profile.
S14, carry out the calculating of distance matrix for all pixels on the profile, the set of result of calculation is the feature database of object template contours, and the Description Matrix in the object template contours feature database is D m, D mThe D that is calculated by all pixels on the profile iConsist of.For same class clarification of objective storehouse, a lot of training samples can be arranged, for the inhomogeneity things, difference composition characteristic storehouse, D={D ObjObj=people, horse, car ....
Step S2 comprises following steps S21～S26:
The edge of S21, extraction image to be detected, specifically comprise following steps: the image transitions to be detected that (1) will collect is gray-scale map; (2) treat detected image with the canny operator and carry out edge extracting; (3) with edge linking algorithm the edge of close together is coupled together, obtain the edge aggregation E={e (i) of image to be detected; I=1...M}, wherein M represents the number at edge.
S22, on the edge of image to be detected extract minutiae, with among the step S12 on the profile of object template the method for extract minutiae identical, specifically comprise following steps: (1) is for any edge e (i) of image to be detected, on the edge have a few and calculate successively the characteristic of correspondence point, obtain the corresponding relation C table of edge e (i); (2) treat detected image edge aggregation E={e (i); All edges of i=1...M} carry out respectively the calculating of unique point, and all edges are obtained corresponding corresponding relation C table, E c={ e c(i); I=1...M}, wherein M represents the number at edge, and the corresponding relation C of edge i is expressed as e c(i)={ P j, P mJ=1...N i, m ∈ 1 ..., N i, N wherein iRepresent the number of the upper pixel of edge i, P mBe P jThe characteristic of correspondence point.
S23, calculate the feature that image characteristic point to be detected forms and describe, with to utilize unique point and central point to set up distance matrix among the step S13 identical to the method that profile is described, specifically comprise following steps: calculate with the some P on the edge i (1) m(i) be the feature description of starting point, specifically comprise following steps: utilize corresponding relation C table set E cIn e c(i) find P m(i) characteristic of correspondence point P j(i); Utilize corresponding relation C table e c(i) find P j(i) characteristic of correspondence point P k(i); Calculate D e ( i ) m = d m , m d m , j d m , k d j , m d j , j d j , k d k , m d k , j d k , k ,
Figure BDA00002210895600102
For edge i goes up with P m(i) be starting point, P k(i) be the profile description of terminal point, wherein d I, j=log (1+||P i-P j|| 2), d I, jP iAnd P jEuclidean distance between the point is in the distance expression of log space; (2) edge set E={e (i); Each edge among the i=1...M} respectively calculated characteristics is described,
Figure BDA00002210895600103
Wherein
Figure BDA00002210895600104
Be illustrated on the edge i feature Description Matrix of some j place's beginning, N iRepresent the number of the upper pixel of edge i, M represents the number at edge.
S24, with Characteristic of Image to be detected describe with object template contours feature database in feature mate, mainly carry out respectively similarity by the Characteristic of Image Description Matrix to be detected that will extract and the Description Matrix in the object template contours feature database and calculate to realize that key step is as follows: (1) Characteristic of Image Description Matrix to be detected is D e, the Description Matrix in the object template contours feature database is D m, to two matrix D e, D mThe diagonal line both sides carry out respectively normalized, to D eObtain after the processing
Figure BDA00002210895600111
Wherein D eSome i on (i, j) expression image outline e to be detected is to the Euclidean distance of some j, to D mObtain after the processing D m ′ = 1 s m D m , Wherein s m = 1 2 Σ i = 1 3 Σ j = 1 3 D m ( i , j ) , D mSome i on (i, j) expression object template contours m is to the Euclidean distance of some j; (2) to matrix D ' eAnd matrix D ' mCarry out similarity and calculate, the value of similarity is A ( i , j ) = max ( exp ( - ( D e ′ ( i , j ) - D m ′ ( i , j ) ) 2 ( D e ′ ( i , j ) σ ) 2 ) , exp ( - ( D e ′ ( i , j ) - D m ′ ( 3 - i , 3 - j ) ) 2 ( D e ′ ( i , j ) σ ) 2 ) , I wherein, j=1,2,3, i ≠ j; σ is the punishment parameter of error of adjusting the distance, and industry is general unifiedly gets 0.2, when i=j, and A (i, j)=1, then the value of similarity is: ψ ( D e ′ , D m ′ ) = 1 9 Σ i = 1 3 Σ j = 1 3 A ( i , j ) .
S25, estimate the central point of image outline to be detected specifically to comprise following steps: (1) according to similarity value ψ (D ' e, D ' m) judgement 1-ψ (D ' e, D ' m)＞ψ τWhether set up, wherein ψ τBe the threshold parameter of restriction similarity, this threshold parameter can be 0.1; (2) if 1-ψ (D ' e, D ' M)＞ψ τ, i.e. Characteristic of Image Description Matrix D to be detected eWith the Description Matrix D in the object template contours feature database mMate unsuccessfully, then give up D e(3) if 1-ψ (D ' e, D ' m)≤ψ τ, i.e. Characteristic of Image Description Matrix D to be detected eWith the Description Matrix D in the object template contours feature database mThe match is successful, then according to S e, S mAnd vector
Figure BDA00002210895600117
Estimate and form D eThree some P m(i), P j(i), P k(i) central point corresponding to difference
Figure BDA00002210895600118
P wherein iBe the arbitrfary point on the image border to be detected, P ' iFor on the image border to be detected with P iCorresponding point, P Center' be the central point of estimating; (4) calculate the marginate center point P of image to be detected Center'; (4) by the marginate center point P of image to be detected Center' be formed in the image to be detected ballot figure to central point; (5) in the statistics image to be detected to the voting results of central point, with mean shift algorithm voting results are carried out cluster, the cluster centre that obtains is the central point of image outline to be detected.
The distance relation of central point and profile in the distance relation at S26, the central point of estimating according to image outline to be detected and correspondence image edge and the object template contours, estimate the profile of image to be detected, specifically comprise following steps: (1) gets the center point P of image outline to be detected Center(i), wherein i is the some targets (because if a plurality of targets are arranged in the image to be detected, then may there be a plurality of central points in the corresponding central point of each target) in the image to be detected, finds with P Center(i) centered by, radius is the series of features Description Matrix D of ballot in the σ zone e, wherein σ refers to can get 5*5 according to the radius value of image size to be detected and scene setting; (2) statistics characteristics of image Description Matrix D to be detected e(i) with object template contours feature database in corresponding Description Matrix D m(j) the scaling relation V between (i) is namely according to S m(j) and S e(i) value obtains ratio
Figure BDA00002210895600121
Wherein e represents a certain edge of image to be detected, and i represents the starting point of characteristics of image Description Matrix to be detected on edge e, and m represents the jobbie template contours, the starting point of Description Matrix on profile m of correspondence in the j representative body template contours feature database; (3) statistics is with P Center(i) centered by, radius is all characteristics of image Description Matrix D to be detected of ballot in the σ zone eWith the Description Matrix D in the object template contours feature database mBetween scaling relation, obtain in characteristics of image Description Matrix to be detected and the object template contours feature database Description Matrix than value set V={V (i); I=1...m}, wherein the m representative is with P Center(i) centered by, radius is the quantity of the characteristics of image Description Matrix to be detected of ballot in the σ zone; (4) carry out cluster with the ratio among the mean shift algorithm pair set V, determine the ratio V (i) of the Description Matrix in characteristics of image Description Matrix to be detected and the object template contours feature database according to cluster result, determine to satisfy the characteristics of image Description Matrix D to be detected of described ratio according to this ratio V (i) e(i), according to characteristics of image Description Matrix D to be detected e(i) determine point on the edge of composition characteristic Description Matrix, these points are with P Center(i) centered by, with the point in the object template contours image outline to be detected that the match is successful; (5) according to ratio V, with P Center(i) centered by, the point in the image outline to be detected that does not match is filled the profile of the complete image to be detected that obtains estimating.
More than introduced a kind of target identification method based on contour feature, the method has estimated the definite position of target when finishing target identification.The present invention is not limited to above embodiment, and any technical solution of the present invention that do not break away from is namely only carried out improvement or the change that those of ordinary skills know to it, all belongs within protection scope of the present invention.

Claims (10)

1. target identification method based on contour feature, described method comprises the steps: step 1, sets up the feature database of object template contours; Step 2, treat detected image and carry out target identification, it is characterized in that described step 1 comprises following steps:
(1.1) integrity profile of extract phantom plate;
(1.2) central point of extract minutiae and profile on the profile of object template;
(1.3) utilizing unique point and central point to set up distance matrix is described profile;
(1.4) carry out the calculating of distance matrix for all pixels on the profile, the set of result of calculation is the feature database of object template contours,
Described step 2 comprises following steps:
(2.1) edge of extraction image to be detected;
(2.2) extract minutiae on the edge of image to be detected;
(2.3) calculate the feature description that image characteristic point to be detected forms;
(2.4) Characteristic of Image description to be detected is mated with the feature in the object template contours feature database;
(2.5) central point of estimation image outline to be detected;
(2.6) according to the central point of image outline estimation to be detected and distance relation and the thing at correspondence image edge
The distance relation of central point and profile in the body template contours estimates the profile of image to be detected.
2. the target identification method based on contour feature according to claim 1 is characterized in that, utilizes the integrity profile of background subtraction method extract phantom plate in the described step (1.1), and the result of profile is expressed as S={P iI=1...N}, wherein S represents the profile of object template, P iPixel on the expression profile, N represents the number of pixel in the profile.
3. the target identification method based on contour feature according to claim 2 is characterized in that, in the described step (1.2) on the profile of object template the central point of extract minutiae and profile comprise the steps: the some P on the profile i(x i, y i) calculate the characteristic of correspondence point that it satisfies certain condition, specifically comprise following steps: establish the arbitrfary point P on the object template contours J (x j, y j), get a P k(x k, y k), wherein k=2*j-i obtains
Figure FDA00002210895500021
D represents a P jTo a P iWith a P kOrganize straight distance, Ax+By+C=0 represents a P iWith a P kThe straight-line equation at place, wherein A=y k-y j, B=x i-x k, C=x ky j-x iy k, some P iTo a P kDistance can be expressed as:
Figure FDA00002210895500022
Figure FDA00002210895500023
T is d and d IkThe ratio of these two distances when t 〉=T, is then got the most close P on the order iThe P of point jPoint is P iCharacteristic of correspondence point, T are the threshold parameter of setting according to different objects; To on the profile have a few and calculate successively the characteristic of correspondence point, obtain corresponding relation C table: C={P i, P mI=1...N, m ∈ 1 ..., N}}, wherein P mBe P iThe characteristic of correspondence point, N represents the number of pixel in the profile;
The central point that calculates profile is: P center ( x , y ) = 1 N Σ i = 1 N P i ( x i , y i ) .
4. the target identification method based on contour feature according to claim 3 is characterized in that, described step (1.3) is utilized unique point and central point to set up distance matrix and profile is described is specifically comprised following steps:
Calculating is with the some P on the profile iBe the Description Matrix D in the object template contours feature database of starting point i, specifically comprise following steps: utilize corresponding relation C table to find P iCharacteristic of correspondence point P jUtilize corresponding relation C table to find P jCharacteristic of correspondence point P kCalculate D i = d i , i d i , j d i , k d j , i d j , j d j , k d k , i d k , j d k , k , D iFor with P iBe starting point, P kBe the profile Description Matrix of terminal point, wherein d I, j=log (1+||P i-P j|| 2), d I, jP iAnd P jEuclidean distance between the point is in the distance expression of log space;
Calculate each unique point P on the profile iAnd center point P CenterRelation, with vector representation be
Figure FDA00002210895500031
Wherein N represents the number of pixel in the profile.
5. the target identification method based on contour feature according to claim 4 is characterized in that, the edge that described step (2.1) is extracted image to be detected specifically comprises following steps:
Be gray-scale map with the image transitions to be detected that collects;
Treat detected image with the canny operator and carry out edge extracting;
With edge linking algorithm the edge of close together is coupled together, obtain the edge aggregation E={e (i) of image to be detected; I=1...M}, wherein M represents the number at edge.
6. the target identification method based on contour feature according to claim 5 is characterized in that, described step (2.2) extract minutiae on the edge of image to be detected specifically comprises following steps:
For any edge e (i) of image to be detected, on the edge have a few and calculate successively the characteristic of correspondence point, obtain the corresponding relation C table of edge e (i);
Treat detected image edge aggregation E={e (i); All edges of i=1...M} carry out respectively the calculating of unique point, and all edges are obtained corresponding corresponding relation C table, E c={ e c(i); I=1...M}, wherein M represents the number at edge, and the corresponding relation C of edge i is expressed as e c(i)={ P j, P mJ=1...N i, m ∈ 1 ... N i, N wherein iRepresent the number of the upper pixel of edge i, P mBe P jThe characteristic of correspondence point.
7. the target identification method based on contour feature according to claim 6 is characterized in that, the feature description that described step (2.3) is calculated image characteristic point composition to be detected specifically comprises following steps: calculate with the some P on the edge i m(i) be the feature description of starting point, specifically comprise following steps: utilize corresponding relation C table set E cIn e c(i) find P m(i) characteristic of correspondence point P j(i); Utilize corresponding relation C table e c(i) find P j(i) characteristic of correspondence point P k(i); Calculate D e ( i ) m = d m , m d m , j d m , k d j , m d j , j d j , k d k , m d k , j d k , k ,
Figure FDA00002210895500033
For edge i goes up with P m(i) be starting point, P k(i) be the profile description of terminal point, wherein d I, j=log (1+||P i-P j|| 2), d I, jP iAnd P jEuclidean distance between the point is in the distance expression of log space; Edge set E={e (i); Each edge among the i=1...M} respectively calculated characteristics is described,
Figure FDA00002210895500041
Wherein
Figure FDA00002210895500042
Be illustrated on the edge i feature Description Matrix of some j place's beginning, N iRepresent the number of the upper pixel of edge i, M represents the number at edge.
8. the target identification method based on contour feature according to claim 7, it is characterized in that, described step (2.4) with Characteristic of Image to be detected describe with object template contours feature database in feature mate mainly and to carry out respectively similarity by the Characteristic of Image Description Matrix to be detected that will extract and the Description Matrix in the object template contours feature database and calculate to realize that key step is as follows: Characteristic of Image Description Matrix to be detected is D e, the Description Matrix in the object template contours feature database is D m, to two matrix D e, D mThe diagonal line both sides carry out respectively normalized, to D eObtain after the processing D e ′ = 1 s e D e , Wherein s e = 1 2 Σ i = 1 3 Σ j = 1 3 D e ( i , j ) , D eSome i on (i, j) expression image outline e to be detected is to the Euclidean distance of some j, to D mObtain after the processing
Figure FDA00002210895500045
Wherein
Figure FDA00002210895500046
D mSome i on (i, j) expression object template contours m is to the Euclidean distance of some j;
To matrix D ' eAnd matrix D ' mCarry out similarity and calculate, the value of similarity is A ( i , j ) = max ( exp ( - ( D e ′ ( i , j ) - D m ′ ( i , j ) ) 2 ( D e ′ ( i , j ) σ ) 2 ) , exp ( - ( D e ′ ( i , j ) - D m ′ ( 3 - i , 3 - j ) ) 2 ( D e ′ ( i , j ) σ ) 2 ) , I wherein, j=1,2,3, i ≠ j; σ is the punishment parameter of error of adjusting the distance, when i=j, and A (i, j)=1, then the value of similarity is: ψ ( D e ′ , D m ′ ) = 1 9 Σ i = 1 3 Σ j = 1 3 A ( i , j ) .
9. the target based on contour feature according to claim 8 is identified ten thousand methods, it is characterized in that, described step (2.5) estimates that the central point of image outline to be detected specifically comprises following steps:
According to similarity value ψ (D ' e, D ' m) judgement 1-ψ (D ' e, D ' m)＞ψ τWhether set up, wherein ψ τThreshold parameter for the restriction similarity;
If 1-ψ (D ' e, D ' m)＞ψ τ, i.e. Characteristic of Image Description Matrix D to be detected eWith the Description Matrix D in the object template contours feature database mMate unsuccessfully, then give up D e
If 1-ψ (D ' e, D ' m)≤ψ τ, i.e. Characteristic of Image Description Matrix D to be detected eWith the Description Matrix D in the object template contours feature database mThe match is successful, then according to s e, s mAnd vector
Figure FDA00002210895500051
Estimate and form D eThree some P m(i), P j(i), P k(i) central point corresponding to difference
Figure FDA00002210895500052
P wherein iBe the arbitrfary point on the image border to be detected, P i' be on the image border to be detected with P iCorresponding point, P Center' be the central point of estimating;
Calculate the marginate center point P of image to be detected Center';
By the marginate center point P of image to be detected Center' be formed in the image to be detected ballot figure to central point;
Add up in the image to be detected the voting results of central point, with mean shift algorithm voting results are carried out cluster, the cluster centre that obtains is the central point of image outline to be detected.
10. the target identification method based on contour feature according to claim 9, it is characterized in that, the distance relation at central point and edge in the central point that described step (2.6) is estimated according to image outline to be detected and the distance relation at correspondence image edge and the object template contours, the profile that estimates image to be detected specifically comprises following steps:
Get the center point P of image outline to be detected Center(i), wherein i is the some targets in the image to be detected, finds with P CenteCentered by the r (i), radius is the series of features Description Matrix D of ballot in the σ zone e, wherein σ refers to the radius value according to image size to be detected and scene setting;
Add up characteristics of image Description Matrix D to be detected e(i) with object template contours feature database in corresponding Description Matrix D m(j) the scaling relation V between (i) is namely according to s m(j) and s e(i) value obtains ratio
Figure FDA00002210895500061
Wherein e represents a certain edge of image to be detected, and i represents the starting point of characteristics of image Description Matrix to be detected on edge e, and m represents the jobbie template contours, the starting point of Description Matrix on profile m of correspondence in the j representative body template contours feature database;
Statistics is with P Center(i) centered by, radius is all characteristics of image Description Matrix D to be detected of ballot in the σ zone eWith the Description Matrix D in the object template contours feature database mBetween scaling relation, obtain in characteristics of image Description Matrix to be detected and the object template contours feature database Description Matrix than value set V={V (i); I=1...m}, wherein the m representative is with P Center(i) centered by, radius is the quantity of the characteristics of image Description Matrix to be detected of ballot in the σ zone;
Carry out cluster with the ratio among the mean shift algorithm pair set V, determine the ratio V (i) of the Description Matrix in characteristics of image Description Matrix to be detected and the object template contours feature database according to cluster result, determine to satisfy the characteristics of image Description Matrix D to be detected of described ratio according to this ratio V (i) e(i), according to characteristics of image Description Matrix D to be detected e(i) determine point on the edge of composition characteristic Description Matrix, these points are with P Center(i) centered by, with the point in the object template contours image outline to be detected that the match is successful; According to ratio V, with P Center(i) centered by, the point in the image outline to be detected that does not match is filled the profile of the complete image to be detected that obtains estimating.
CN201210372459.5A 2012-09-28 2012-09-28 Target identification method based on contour features Active CN102880877B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201210372459.5A CN102880877B (en) 2012-09-28 2012-09-28 Target identification method based on contour features

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201210372459.5A CN102880877B (en) 2012-09-28 2012-09-28 Target identification method based on contour features

Publications (2)

Publication Number Publication Date
CN102880877A true CN102880877A (en) 2013-01-16
CN102880877B CN102880877B (en) 2015-07-15

Family

ID=47482194

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201210372459.5A Active CN102880877B (en) 2012-09-28 2012-09-28 Target identification method based on contour features

Country Status (1)

Country Link
CN (1) CN102880877B (en)

Cited By (35)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103093245A (en) * 2013-01-21 2013-05-08 信帧电子技术（北京）有限公司 Signal lamp identifying method in video image
CN103984942A (en) * 2014-05-28 2014-08-13 深圳市中兴移动通信有限公司 Object recognition method and mobile terminal
CN104090581A (en) * 2014-04-16 2014-10-08 上海热像机电科技有限公司 Intelligent realization of scene matching of thermal infrared imager
CN104978582A (en) * 2015-05-15 2015-10-14 苏州大学 Contour chord angle feature based identification method for blocked target
CN106028160A (en) * 2016-06-03 2016-10-12 腾讯科技（深圳）有限公司 Image data processing method and device
CN106297292A (en) * 2016-08-29 2017-01-04 苏州金螳螂怡和科技有限公司 Based on highway bayonet socket and the Trajectory System of comprehensively monitoring
CN106295649A (en) * 2015-05-15 2017-01-04 中云智慧（北京）科技有限公司 Target identification method based on contour features
CN106447695A (en) * 2016-09-23 2017-02-22 广州视源电子科技股份有限公司 Method and device for judging same object in multi-object tracking
CN103793528B (en) * 2014-02-25 2017-05-03 张文虎 Method and device for obtaining network addresses
CN106651878A (en) * 2016-12-21 2017-05-10 福建师范大学 Method for extracting straight line from local invariant feature points
CN104123712B (en) * 2013-04-26 2017-07-28 富士通株式会社 Similarity Measure apparatus and method and object identification device and method
CN107103323A (en) * 2017-03-09 2017-08-29 广东顺德中山大学卡内基梅隆大学国际联合研究院 A kind of target identification method based on image outline feature
CN107784309A (en) * 2017-11-01 2018-03-09 深圳汇生通科技股份有限公司 A kind of realization method and system to vehicle cab recognition
CN107909032A (en) * 2017-11-15 2018-04-13 重庆邮电大学 A kind of behavioral value and recognition methods based on single sample
CN108549874A (en) * 2018-04-19 2018-09-18 广州广电运通金融电子股份有限公司 A kind of object detection method, equipment and computer readable storage medium
CN109344900A (en) * 2018-09-30 2019-02-15 中民筑友科技投资有限公司 A kind of component identification method, apparatus, equipment and computer readable storage medium
WO2019033509A1 (en) * 2017-08-16 2019-02-21 歌尔科技有限公司 Interior and exterior identification method for image contour and device
CN109363770A (en) * 2018-12-06 2019-02-22 安徽埃克索医疗机器人有限公司 A kind of surgical navigational robot index point automatic identification localization method
CN109544573A (en) * 2017-09-21 2019-03-29 卡西欧计算机株式会社 Contour detecting device, printing equipment, profile testing method and recording medium
CN109623223A (en) * 2019-02-02 2019-04-16 河北宏龙环保科技有限公司 Pipeline groove detection method
CN109664321A (en) * 2018-12-27 2019-04-23 四川文理学院 Mechanical arm, explosive trolley and method for searching
CN109697441A (en) * 2017-10-23 2019-04-30 杭州海康威视数字技术股份有限公司 A kind of object detection method, device and computer equipment
CN109886213A (en) * 2019-02-25 2019-06-14 湖北亿咖通科技有限公司 Fatigue state judgment method, electronic equipment and computer readable storage medium
CN110443826A (en) * 2019-07-10 2019-11-12 佛山科学技术学院 A kind of virtual reality fusion emulation experiment mistake householder method and system
WO2020034672A1 (en) * 2018-08-16 2020-02-20 北京七鑫易维信息技术有限公司 Method and apparatus for determining region of interest of user in image
CN111079575A (en) * 2019-11-29 2020-04-28 拉货宝网络科技有限责任公司 Material identification method and system based on packaging image characteristics
CN111539259A (en) * 2020-03-31 2020-08-14 广州富港万嘉智能科技有限公司 Target object recognition method, artificial neural network training method, computer-readable storage medium, and manipulator
CN111612014A (en) * 2020-04-26 2020-09-01 东风汽车集团有限公司 Windshield glass identification error-proofing method in gluing process
CN111738166A (en) * 2020-06-24 2020-10-02 平安科技（深圳）有限公司 Target contour delineation method, device, computer system and readable storage medium
CN112288820A (en) * 2020-12-29 2021-01-29 深圳市六合智能感知系统科技有限公司 Contour calculation method, device and equipment for regular object and storage medium
CN112949625A (en) * 2021-01-29 2021-06-11 西安电子科技大学 Target identification method and system based on centroid contour distance
CN112966138A (en) * 2021-02-22 2021-06-15 济南大学 Two-dimensional shape retrieval method and system based on contour feature point matching
CN113033592A (en) * 2019-12-25 2021-06-25 山东理工大学 Shape matching and object identification method based on slope difference distribution
CN114693862A (en) * 2020-12-29 2022-07-01 北京万集科技股份有限公司 Three-dimensional point cloud data model reconstruction method, target re-identification method and device
CN117484497A (en) * 2023-11-03 2024-02-02 广州尚纳智能科技有限公司 Intelligent auxiliary three-dimensional teaching method based on shape contour matching

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040141003A1 (en) * 2003-01-21 2004-07-22 Dell Products, L.P. Maintaining a user interest profile reflecting changing interests of a customer
CN1996346A (en) * 2005-12-29 2007-07-11 电子科技大学 Vehicle recognition algorithm based on contour
CN101140623A (en) * 2007-09-29 2008-03-12 华为技术有限公司 Video frequency objects recognition method and system based on supporting vectors machine
CN101635031A (en) * 2009-08-27 2010-01-27 上海交通大学 Method for extracting and identifying small sample character contour feature
CN102509085A (en) * 2011-11-19 2012-06-20 江苏大学 Pig walking posture identification system and method based on outline invariant moment features

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040141003A1 (en) * 2003-01-21 2004-07-22 Dell Products, L.P. Maintaining a user interest profile reflecting changing interests of a customer
CN1996346A (en) * 2005-12-29 2007-07-11 电子科技大学 Vehicle recognition algorithm based on contour
CN101140623A (en) * 2007-09-29 2008-03-12 华为技术有限公司 Video frequency objects recognition method and system based on supporting vectors machine
CN101635031A (en) * 2009-08-27 2010-01-27 上海交通大学 Method for extracting and identifying small sample character contour feature
CN102509085A (en) * 2011-11-19 2012-06-20 江苏大学 Pig walking posture identification system and method based on outline invariant moment features

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
周正杰 等: "基于轮廓的形状特征提取与识别方法", 《计算机工程与应用》 *

Cited By (45)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103093245B (en) * 2013-01-21 2016-01-20 信帧电子技术（北京）有限公司 The method of marker lamp in video image
CN103093245A (en) * 2013-01-21 2013-05-08 信帧电子技术（北京）有限公司 Signal lamp identifying method in video image
CN104123712B (en) * 2013-04-26 2017-07-28 富士通株式会社 Similarity Measure apparatus and method and object identification device and method
CN103793528B (en) * 2014-02-25 2017-05-03 张文虎 Method and device for obtaining network addresses
CN104090581A (en) * 2014-04-16 2014-10-08 上海热像机电科技有限公司 Intelligent realization of scene matching of thermal infrared imager
CN103984942A (en) * 2014-05-28 2014-08-13 深圳市中兴移动通信有限公司 Object recognition method and mobile terminal
CN104978582A (en) * 2015-05-15 2015-10-14 苏州大学 Contour chord angle feature based identification method for blocked target
CN106295649A (en) * 2015-05-15 2017-01-04 中云智慧（北京）科技有限公司 Target identification method based on contour features
CN104978582B (en) * 2015-05-15 2018-01-30 苏州大学 Shelter target recognition methods based on profile angle of chord feature
CN106028160A (en) * 2016-06-03 2016-10-12 腾讯科技（深圳）有限公司 Image data processing method and device
CN106297292A (en) * 2016-08-29 2017-01-04 苏州金螳螂怡和科技有限公司 Based on highway bayonet socket and the Trajectory System of comprehensively monitoring
CN106447695A (en) * 2016-09-23 2017-02-22 广州视源电子科技股份有限公司 Method and device for judging same object in multi-object tracking
CN106651878A (en) * 2016-12-21 2017-05-10 福建师范大学 Method for extracting straight line from local invariant feature points
CN106651878B (en) * 2016-12-21 2019-06-11 福建师范大学 A method of for extracting straight line from local invariant feature point
CN107103323A (en) * 2017-03-09 2017-08-29 广东顺德中山大学卡内基梅隆大学国际联合研究院 A kind of target identification method based on image outline feature
CN107103323B (en) * 2017-03-09 2020-06-16 广东顺德中山大学卡内基梅隆大学国际联合研究院 Target identification method based on image contour features
WO2019033509A1 (en) * 2017-08-16 2019-02-21 歌尔科技有限公司 Interior and exterior identification method for image contour and device
CN109544573B (en) * 2017-09-21 2022-11-25 卡西欧计算机株式会社 Contour detection device, printing device, contour detection method, and recording medium
CN109544573A (en) * 2017-09-21 2019-03-29 卡西欧计算机株式会社 Contour detecting device, printing equipment, profile testing method and recording medium
CN109697441A (en) * 2017-10-23 2019-04-30 杭州海康威视数字技术股份有限公司 A kind of object detection method, device and computer equipment
WO2019080743A1 (en) * 2017-10-23 2019-05-02 杭州海康威视数字技术股份有限公司 Target detection method and apparatus, and computer device
US11288548B2 (en) 2017-10-23 2022-03-29 Hangzhou Hikvision Digital Technology Co., Ltd. Target detection method and apparatus, and computer device
CN107784309A (en) * 2017-11-01 2018-03-09 深圳汇生通科技股份有限公司 A kind of realization method and system to vehicle cab recognition
CN107909032A (en) * 2017-11-15 2018-04-13 重庆邮电大学 A kind of behavioral value and recognition methods based on single sample
CN108549874A (en) * 2018-04-19 2018-09-18 广州广电运通金融电子股份有限公司 A kind of object detection method, equipment and computer readable storage medium
CN108549874B (en) * 2018-04-19 2021-11-23 广州广电运通金融电子股份有限公司 Target detection method, target detection equipment and computer-readable storage medium
WO2020034672A1 (en) * 2018-08-16 2020-02-20 北京七鑫易维信息技术有限公司 Method and apparatus for determining region of interest of user in image
CN109344900A (en) * 2018-09-30 2019-02-15 中民筑友科技投资有限公司 A kind of component identification method, apparatus, equipment and computer readable storage medium
CN109363770A (en) * 2018-12-06 2019-02-22 安徽埃克索医疗机器人有限公司 A kind of surgical navigational robot index point automatic identification localization method
CN109664321A (en) * 2018-12-27 2019-04-23 四川文理学院 Mechanical arm, explosive trolley and method for searching
CN109623223A (en) * 2019-02-02 2019-04-16 河北宏龙环保科技有限公司 Pipeline groove detection method
CN109886213A (en) * 2019-02-25 2019-06-14 湖北亿咖通科技有限公司 Fatigue state judgment method, electronic equipment and computer readable storage medium
CN110443826A (en) * 2019-07-10 2019-11-12 佛山科学技术学院 A kind of virtual reality fusion emulation experiment mistake householder method and system
CN111079575A (en) * 2019-11-29 2020-04-28 拉货宝网络科技有限责任公司 Material identification method and system based on packaging image characteristics
CN113033592A (en) * 2019-12-25 2021-06-25 山东理工大学 Shape matching and object identification method based on slope difference distribution
CN111539259A (en) * 2020-03-31 2020-08-14 广州富港万嘉智能科技有限公司 Target object recognition method, artificial neural network training method, computer-readable storage medium, and manipulator
CN111612014A (en) * 2020-04-26 2020-09-01 东风汽车集团有限公司 Windshield glass identification error-proofing method in gluing process
WO2021258991A1 (en) * 2020-06-24 2021-12-30 平安科技（深圳）有限公司 Object contour delineation method and apparatus, and computer system and readable storage medium
CN111738166A (en) * 2020-06-24 2020-10-02 平安科技（深圳）有限公司 Target contour delineation method, device, computer system and readable storage medium
CN111738166B (en) * 2020-06-24 2024-03-01 平安科技（深圳）有限公司 Target contour defining method, device, computer system and readable storage medium
CN112288820A (en) * 2020-12-29 2021-01-29 深圳市六合智能感知系统科技有限公司 Contour calculation method, device and equipment for regular object and storage medium
CN114693862A (en) * 2020-12-29 2022-07-01 北京万集科技股份有限公司 Three-dimensional point cloud data model reconstruction method, target re-identification method and device
CN112949625A (en) * 2021-01-29 2021-06-11 西安电子科技大学 Target identification method and system based on centroid contour distance
CN112966138A (en) * 2021-02-22 2021-06-15 济南大学 Two-dimensional shape retrieval method and system based on contour feature point matching
CN117484497A (en) * 2023-11-03 2024-02-02 广州尚纳智能科技有限公司 Intelligent auxiliary three-dimensional teaching method based on shape contour matching

Also Published As

Publication number Publication date
CN102880877B (en) 2015-07-15

Similar Documents

Publication Publication Date Title
CN102880877B (en) Target identification method based on contour features
CN106529468B (en) A kind of finger vein identification method and system based on convolutional neural networks
CN103268497B (en) A kind of human face posture detection method and the application in recognition of face
US9489561B2 (en) Method and system for estimating fingerprint pose
CN105224937B (en) Fine granularity semanteme color pedestrian recognition methods again based on human part position constraint
CN102156884B (en) Straight segment detecting and extracting method
CN105956560A (en) Vehicle model identification method based on pooling multi-scale depth convolution characteristics
CN102270308B (en) Facial feature location method based on five sense organs related AAM (Active Appearance Model)
CN104102904B (en) A kind of static gesture identification method
CN105160317A (en) Pedestrian gender identification method based on regional blocks
CN105320950A (en) A video human face living body detection method
Cao et al. Fingerprint matching by incorporating minutiae discriminability
US8953852B2 (en) Method for face recognition
CN104636721A (en) Palm print identification method based on contour and edge texture feature fusion
CN103020614B (en) Based on the human motion identification method that space-time interest points detects
CN103903238A (en) Method for fusing significant structure and relevant structure of characteristics of image
CN104778472B (en) Human face expression feature extracting method
CN104036229A (en) Regression-based active appearance model initialization method
CN111488798B (en) Fingerprint identification method, fingerprint identification device, electronic equipment and storage medium
CN102831445B (en) Target detection method based on semantic Hough transformation and partial least squares
CN107644203A (en) A kind of feature point detecting method of form adaptive classification
CN114581990A (en) Intelligent running test method and device
CN102930242A (en) Bus type identifying method
CN107909049A (en) Pedestrian's recognition methods again based on least square discriminant analysis distance study
CN101894154B (en) Method for extracting key pattern from image file

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C53 Correction of patent of invention or patent application
CB02 Change of applicant information

Address after: 610041 Chengdu South Road, Sichuan, No. four, No. 9

Applicant after: Chengdu Information Technology of Chinese Academy of Sciences Co., Ltd.

Address before: 610041 Chengdu South Road, Sichuan, No. four, No. 9

Applicant before: Chengdu Information Technology Co., Ltd., CAS

COR Change of bibliographic data

Free format text: CORRECT: APPLICANT; FROM: CHENGDU INFORMATION TECHNOLOGY CO., LTD., CAS TO: CHENGDU INFORMATION TECHNOLOGY OF CHINESE ACADEMY OF SCIENCES CO., LTD.

C53 Correction of patent of invention or patent application
CB02 Change of applicant information

Address after: Room 18, building 1, building No. 360, No. 610041, Fai Road, high tech Zone, Sichuan, Chengdu, China

Applicant after: Chengdu Information Technology of Chinese Academy of Sciences Co., Ltd.

Address before: 610041 Chengdu South Road, Sichuan, No. four, No. 9

Applicant before: Chengdu Information Technology of Chinese Academy of Sciences Co., Ltd.

C14 Grant of patent or utility model
GR01 Patent grant