CN106886754B - Object identification method and system under a kind of three-dimensional scenic based on tri patch - Google Patents
Object identification method and system under a kind of three-dimensional scenic based on tri patch Download PDFInfo
- Publication number
- CN106886754B CN106886754B CN201710034186.6A CN201710034186A CN106886754B CN 106886754 B CN106886754 B CN 106886754B CN 201710034186 A CN201710034186 A CN 201710034186A CN 106886754 B CN106886754 B CN 106886754B
- Authority
- CN
- China
- Prior art keywords
- feature vector
- tri patch
- characteristic value
- dimensional
- histogram
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Geometry (AREA)
- Computer Graphics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the object identification methods under a kind of three-dimensional scenic based on tri patch.This method is on the basis of traditional two dimensional image recognition methods, in view of the triangle gridding storage mode of 3 D scene rebuilding information, 3-D image is subjected to tri patch division, it is counted by color neighborhood histogram feature, spatial pyramid multi-scale information and neighborhood space information are merged, it is trained with SVM, obtains disaggregated model and the meadow dough sheet in three-dimensional scenic is identified.The present invention also provides the object identification systems under a kind of three-dimensional scenic based on tri patch, technical solution of the present invention solves the problems, such as in three-dimensional reconstruction scene the problem of local message cannot express characteristics of image completely in the doubt problem of identification region and image in identification, significant for resurfacing and scene understanding etc..
Description
Technical field
The invention belongs to technical field of computer vision, more particularly, to a kind of three-dimensional scenic based on tri patch
Under object identification method and system.
Background technique
With the development of human society, image recognition has rapidly developed the technological means particularly important as one,
Goal in research is to confer to the visual capacity of the computer mankind, makes it through image cognition surrounding enviroment information, including environment-identification
Geometry, position and posture of middle objective etc..Image recognition needs integrated use computer science, pattern-recognition, machine
The subjects such as device vision and image understanding, and advance with the development of these subjects.
In general, image identification system is made of following components.First part is the acquisition of image information, to figure
As being exactly that picture, egative film and character and graphic etc. are converted to electric signal in case subsequent processing with photoelectric scanning device for identification.
Second part is the pretreatment of image, and main purpose is to eliminate extraneous features and reinforce feature of interest.Third step is that feature mentions
It takes, common feature includes color, texture, Scale invariant features transform (Scale-invariant feature
Transform, SIFT), histogram of gradients feature (Histogram of Oriented Gradient, HOG) etc..Part IV
It is judgement or classification.Wherein training method can using support vector machine method (Support Vector Machine, SVM),
Improved repetitive exercise method (Adaptive Boostint, AdaBoost) etc..
Currently, three kinds of Three-dimension object recognition Research Thinkings of most mainstream are: 1) method based on model or geometry;2) it is based on
The method of appearance and view;3) it is based on the matched method of local feature.
The principle of method based on model or geometry is exactly the three-dimensional information of real goal to be obtained using sensor and to letter
Breath is analyzed and processed, and obtains the description of a kind of surface, boundary and connection relationship.This method advantage is comparison intuitively and is easy to
Understand, but the operand of the algorithm generally used is all larger, and needs manually by CAD (Computer
Aided Design, CAD) etc. softwares generate model.
Object is identified by visual similarity based on the method for view, identifying system design is relatively easy, without explicit
Ground calculates object dimensional model.This method is generally divided into two steps: firstly, by different illumination conditions and three-dimension object two
The different postures showed in dimension image, automatically to learn expression or the training system of object;Then, at one unknown two
Object is judged whether there is in dimension image.The restrictive condition of this method is that interested object must can be preferable with background
It distinguishes, therefore more sensitive between the overlapping object, and needs preferable image segmentation.But work as the Geometric Modeling of object very
It is difficult or when not being likely to be obtained, then the method based on view can be utilized to identify three-dimension object.
It mainly include that point, line and the local feature in face, its advantages are based on the matched method of local feature, because of visual angle
The deformation of object appearance caused by changing can be by simply becoming on the scale of part although global apparently extremely complex
Change to estimate;Simultaneously as being matched without all local features, this method has overlapping and complex background feelings in object
There is preferable robustness under condition.Because establishing interregional similitude, localization has also been accomplished in object identification;Two dimensional image
In identification in the pretreatment stage of image often through super-pixel segmentation at different regions, then to region carry out feature extraction
And classify, and three-dimensional scenic cannot be identified due to its structure is complicated property by dividing super-pixel region.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides a kind of three dimensional fields based on tri patch
Object identification method and system under scape, its object is to which three-dimensional scenic to be identified is divided into different size of tri patch,
It is identified using the clarification of objective in tri patch, is thus extended two dimensional image recognition methods, propose a kind of base
In the meadow recognition methods of tri patch.
To achieve the above object, according to one aspect of the present invention, a kind of three-dimensional scenic based on tri patch is provided
Under object identification method, method includes the following steps:
(1) three-dimensional scenic to be identified is subjected to three-dimensional reconstruction, by the triangulation tri patch on three-dimensional scenic surface;
(2) the HSI feature of all tri patch units is extracted;
(3) histogram is carried out according to the H characteristic value for the adjacent triangular faces for selecting tri patch and its shared side and S characteristic value
Figure statistics, statistical data is put into feature vector, by the feature of the feature vector of tri patch and its adjacent triangular faces to
Amount is overlapped and (is described in detail in specification), feature vector of the stack result as selected tri patch;
(4) feature vector of selected tri patch is input in SVM identification model, obtains recognition result.
Further, histogram is carried out in the step (3) to be specifically configured to:
The group number that H feature value histogram and S feature value histogram is arranged is 8, and group is away from respectively H characteristic value value range
The 1/8 of 1/8 and S characteristic value value range.
Further, statistical data is put into feature vector in the step (3) specifically:
The H characteristic value of tri patch is subjected to statistics with histogram and obtains one group of 8 dimensional vector, by the S characteristic value of tri patch
It carries out statistics with histogram and also obtains one group of 8 dimensional vector, 8 dimensional vectors of 8 dimensional vectors of H characteristic value and S characteristic value are joined end to end
The feature vector of 16 dimension of composition.
Further, feature vector is overlapped in the step (3) specifically:
Wherein, { H11,H12,H13,...,H18,S11,S12,...,S18And { H21,H22,H23,...,H28,S21,S22,...,
S28Indicate 16 feature vectors tieed up.
Further, in the step (4) SVM identification model training process specifically:
(41) setting positive sample and setting negative sample are collected as sample set, positive sample selection is all object to be identified
Two-dimension picture, negative sample choose be free of object to be identified two-dimension picture;
(42) be by the size adjusting of all samples it is consistent, select multiple preset scales to carry out tri patch to sample
It divides, extracts the HSI feature of all tri patch;
(43) it is carried out according to the H characteristic value for the adjacent triangular faces for selecting tri patch and its shared side and S characteristic value straight
Side's figure statistics, statistical data is put into feature vector, by the feature of the feature vector of tri patch and its adjacent triangular faces
Vector is overlapped, feature vector of the stack result as selected tri patch;
(44) feature vector of all samples SVM is input to be trained;The decision function of SVM are as follows:
Wherein, L presentation class sum, value 2;ciRepresentative function coefficient, ci∈R;γ indicates the parameter of cross check, γ
∈R+;Expression is chosen for trained support vector;h0Indicate the non-supported vector of training;Using the chi-Square measure of histogramAs kernel function, is defined as:
Wherein, N indicates feature vector total quantity in sample set;Indicate k-th of feature vector it is corresponding support to
Amount;h0(k) k-th of feature vector is indicated;
(45) SVM identification model parameter is obtained after training, constructs SVM identification model.
It is another aspect of this invention to provide that the object identification system under providing a kind of three-dimensional scenic based on tri patch
System, changes system and comprises the following modules:
Tri patch division module, for three-dimensional scenic to be identified to be carried out three-dimensional reconstruction, by the three of three-dimensional scenic surface
Angle grid dividing tri patch;
HSI characteristic extracting module, for extracting the HSI feature of all tri patch units;
Feature vector computing module, for the H feature according to the adjacent triangular faces for selecting tri patch and its shared side
Value and S characteristic value carry out statistics with histogram, statistical data are put into feature vector, by the feature vector of tri patch and its phase
The feature vector of adjacent tri patch is overlapped, feature vector of the stack result as selected tri patch;
Identification module obtains identification knot for the feature vector of selected tri patch to be input in SVM identification model
Fruit.
Further, histogram is carried out in described eigenvector computing module to be specifically configured to:
The group number that H feature value histogram and S feature value histogram is arranged is 8, and group is away from respectively H characteristic value value range
The 1/8 of 1/8 and S characteristic value value range.
Further, statistical data is put into feature vector in described eigenvector computing module specifically:
The H characteristic value of tri patch is subjected to statistics with histogram and obtains one group of 8 dimensional vector, by the S characteristic value of tri patch
It carries out statistics with histogram and also obtains one group of 8 dimensional vector, 8 dimensional vectors of 8 dimensional vectors of H characteristic value and S characteristic value are joined end to end
The feature vector of 16 dimension of composition.
Further, feature vector is overlapped in the step (3) specifically:
Wherein, { H11,H12,H13,...,H18,S11,S12,...,S18And { H21,H22,H23,...,H28,S21,S22,...,
S28Indicate 16 feature vectors tieed up.
Further, the training system of SVM identification model includes with lower unit in the identification module:
Sample collection unit, for collecting multiple positive samples and multiple negative samples as sample set, positive sample selection is all
The two-dimension picture of object to be identified, negative sample choose the two-dimension picture for being free of object to be identified;
Tri patch division unit, for being consistent, the multiple preset scales pair of selection by the size adjusting of all samples
Sample carries out the division of tri patch, extracts the HSI feature of all tri patch;
Feature vector computing unit, for the H feature according to the adjacent triangular faces for selecting tri patch and its shared side
Value and S characteristic value carry out statistics with histogram, statistical data are put into feature vector, by the feature vector of tri patch and its phase
The feature vector of adjacent tri patch is overlapped, feature vector of the stack result as selected tri patch;
Training unit is trained for the feature vector of all samples to be input to SVM;The decision function of SVM are as follows:
Wherein, L presentation class sum, value 2;ciRepresentative function coefficient, ci∈R;γ indicates the parameter of cross check, γ
∈R+;Expression is chosen for trained support vector;h0Indicate the non-supported vector of training;Using the chi-Square measure of histogramAs kernel function, is defined as:
Wherein, N indicates feature vector total quantity in sample set;Indicate k-th of feature vector it is corresponding support to
Amount;h0(k) k-th of feature vector is indicated;
Model construction unit constructs SVM identification model for obtaining SVM identification model parameter after training.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, have following technology special
Sign and the utility model has the advantages that
(1) triangle gridding in the three-dimensional scenic reconstructed is utilized in technical solution of the present invention, and three-dimensional scenic is divided into
Multiple two dimension tri patch avoid carrying out the problem that super-pixel extracts excessively complexity in three-dimensional scenic;
(2) technical solution of the present invention uses Analysis On Multi-scale Features information and neighboring information fusion, simple and effective, and discrimination is high,
The problems such as can also being further used for three-dimensional scenic resurfacing and scene understanding analysis.
Detailed description of the invention
Fig. 1 is meadow of embodiment of the present invention recognition methods process;
Fig. 2 is the flow chart of the SVM training of meadow of embodiment of the present invention identification;
Fig. 3 a is meadow of embodiment of the present invention sample H data statistic histogram;
Fig. 3 b is sky sample H data statistic histogram of the embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
As shown in Figure 1 for using the meadow recognition methods process of the three-dimensional scenic of technical solution of the present invention, including following step
It is rapid:
(1) three-dimensional scenic to be identified is subjected to three-dimensional reconstruction, by the triangulation tri patch on three-dimensional scenic surface.
(2) the HSI feature of all tri patch units is extracted.
(3) histogram is carried out according to the H characteristic value for the adjacent triangular faces for selecting tri patch and its shared side and S characteristic value
Figure statistics, statistical data is put into feature vector, by the feature of the feature vector of tri patch and its adjacent triangular faces to
Amount is overlapped, feature vector of the stack result as selected tri patch.
(4) feature vector of selected tri patch is input in SVM identification model, obtains recognition result.
It is illustrated in figure 2 the process of the meadow SVM identification model of the building three-dimensional scenic using technical solution of the present invention:
(1) 50 positive samples and 200 negative samples are collected, select the image for being all meadow as positive sample in order to carrying out
Training set sampling and feature extraction select Pascal's data set (PASCAL since the negative sample type to be covered is complicated here
VOC all kinds of non-meadow targets in) are as negative sample.
(2) it is 162*162 by the size adjusting of positive negative training sample, training sample is made into regular right angle trigonometry dough sheet and is divided
The size of processing, tri patch is 3 kinds, is 3 pixels of right angle side length, 6 pixels and 9 pixels respectively;Due to sample
This image is RGB mode, but at different conditions due to RGB mode, the color Distribution value dispersion of RGB is measured, in rgb space two
The non-linear relationship of Euclidean distance and color distance of point is unfavorable for carrying out the classification of color characteristic.It, will to improve recognition effect
Color indicates to be converted into the HSI color space closer to eye-observation mode from RGB color, to improve under different illumination
The robustness of identifying system under environment.In HSI color space: H represents coloration;S represents saturation degree;I represents brightness.3 points
In amount, being illuminated by the light influences maximum to be I component, and H and S is usually the light absorption and reflectance signature for constituting the raw material of object
It determines.In order to reduce influence of the illumination to accuracy of identification, when identifying target, using H and S as the feature extracted.From RGB to
The conversion formula of HSI is as follows:
I=(R+G+B)/3
S=1-3min (R, G, B)/I
If B > G, H=360-H
(3) histogram is carried out according to the H characteristic value for the adjacent triangular faces for selecting tri patch and its shared side and S characteristic value
Figure statistics, statistical data is put into feature vector, by the feature of the feature vector of tri patch and its adjacent triangular faces to
Amount is overlapped, feature vector of the stack result as selected tri patch.H feature value histogram and S characteristic value histogram are set
The group number of figure is 8, the 1/8 of 1/8 and S characteristic value value range of the group away from respectively H characteristic value value range.By tri patch
H characteristic value carries out statistics with histogram and obtains one group of 8 dimensional vector, and the S characteristic value of tri patch is carried out statistics with histogram and is also obtained
8 dimensional vectors of 8 dimensional vectors of H characteristic value and S characteristic value are joined end to end and form the feature vector of 16 dimensions by one group of 8 dimensional vector.It will
The feature vector of tri patch, which is saved, to be used at mat file for SVM.It is as shown in Figure 3a some tri patch in sky sample
H statistics with histogram is as a result, be as shown in Figure 3b the H statistics with histogram result of some tri patch in the sample of meadow.
(4) feature vector of all samples SVM is input to be trained;The decision function of SVM are as follows:
Wherein, L presentation class sum, value 2;ciRepresentative function coefficient, ci∈R;γ indicates the parameter of cross check, γ
∈R+;Expression is chosen for trained support vector;h0Indicate the non-supported vector of training;Using the chi-Square measure of histogramAs kernel function, is defined as:
Wherein, N indicates feature vector total quantity in sample set;Indicate k-th of feature vector it is corresponding support to
Amount;h0(k) k-th of feature vector is indicated.
According to the chi-square distance calculation covariance matrix space of histogram.Mirror image change is carried out to this covariance matrix space
Get corresponding hyperplane matrix in return, that is to say, that by a vector transformation be the mirror image reflected by a hyperplane, be a kind of
Linear transformation, SVM are exactly that optimal hyperlane is constructed in feature space.Obtain covariance matrix and its corresponding hyperplane square
Battle array, the characteristic coefficient of each feature is calculated separately according to the two matrixes, and characteristic coefficient zooms in and out covariance matrix.
After inverting to the covariance matrix after scaling, computation model parameter.It will be in the parameter deposit matrix file of trained SVM model
For the identifying call of subsequent three-dimensional scenic meadow.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (6)
1. a kind of meadow recognition methods of the three-dimensional scenic based on tri patch, which is characterized in that method includes the following steps:
(1) three-dimensional scenic to be identified is subjected to three-dimensional reconstruction, by the triangulation tri patch on three-dimensional scenic surface;
(2) the HSI feature of all tri patch units is extracted;
(3) histogram system is carried out according to the H characteristic value for the adjacent triangular faces for selecting tri patch and its shared side and S characteristic value
Meter, statistical data is put into feature vector, by the feature vector of the feature vector of tri patch and its adjacent triangular faces into
Row superposition, feature vector of the stack result as selected tri patch;
The statistical data is put into feature vector specifically:
The H characteristic value of tri patch is subjected to statistics with histogram and obtains one group of 8 dimensional vector, the S characteristic value of tri patch is carried out
Statistics with histogram also obtains one group of 8 dimensional vector, and 8 dimensional vectors of 8 dimensional vectors of H characteristic value and S characteristic value are joined end to end composition
The feature vector of 16 dimensions;
The feature vector of the feature vector by tri patch and its adjacent triangular faces is overlapped specifically:
Wherein, { H11,H12,H13,...,H18,S11,S12,...,S18And { H21,H22,H23,...,H28,S21,S22,...,S28Table
Show the feature vector of 16 dimensions;
(4) feature vector of selected tri patch is input in SVM identification model, obtains recognition result.
2. a kind of meadow recognition methods of three-dimensional scenic based on tri patch according to claim 1, which is characterized in that
Histogram is carried out in the step (3) to be specifically configured to:
The group number that H feature value histogram and S feature value histogram is arranged is 8, group away from respectively H characteristic value value range 1/8
With the 1/8 of S characteristic value value range.
3. a kind of meadow recognition methods of three-dimensional scenic based on tri patch according to claim 1, which is characterized in that
The training process of SVM identification model in the step (4) specifically:
(41) multiple positive samples and multiple negative samples are collected as sample set, positive sample chooses the X-Y scheme for being all object to be identified
Piece, negative sample choose the two-dimension picture for being free of object to be identified;
(42) be by the size adjusting of all samples it is consistent, select multiple preset scales to carry out tri patch to sample and draw
Point, extract the HSI feature of all tri patch;
(43) histogram is carried out according to the H characteristic value for the adjacent triangular faces for selecting tri patch and its shared side and S characteristic value
Statistics, statistical data is put into feature vector, by the feature vector of the feature vector of tri patch and its adjacent triangular faces
It is overlapped, feature vector of the stack result as selected tri patch;
The statistical data is put into feature vector specifically:
The H characteristic value of tri patch is subjected to statistics with histogram and obtains one group of 8 dimensional vector, the S characteristic value of tri patch is carried out
Statistics with histogram also obtains one group of 8 dimensional vector, and 8 dimensional vectors of 8 dimensional vectors of H characteristic value and S characteristic value are joined end to end composition
The feature vector of 16 dimensions;
The feature vector of the feature vector by tri patch and its adjacent triangular faces is overlapped specifically:
Wherein, { H11,H12,H13,...,H18,S11,S12,...,S18And { H21,H22,H23,...,H28,S21,S22,...,S28Table
Show the feature vector of 16 dimensions;
(44) feature vector of all samples SVM is input to be trained;The decision function of SVM are as follows:
Wherein, L presentation class sum, value 2;ciRepresentative function coefficient, ci∈R;γ indicates the parameter of cross check, γ ∈ R+;Expression is chosen for trained support vector;h0Indicate the non-supported vector of training;Using the chi-Square measure of histogramAs kernel function, is defined as:
Wherein, N indicates feature vector total quantity in sample set;Indicate the corresponding support vector of k-th of feature vector;h0
(k) k-th of feature vector is indicated;
(45) SVM identification model parameter is obtained after training, constructs SVM identification model.
4. a kind of meadow identifying system of the three-dimensional scenic based on tri patch, which is characterized in that the system comprises the following modules:
Tri patch division module, for three-dimensional scenic to be identified to be carried out three-dimensional reconstruction, by the triangulation network on three-dimensional scenic surface
Lattice divide tri patch;
HSI characteristic extracting module, for extracting the HSI feature of all tri patch units;
Feature vector computing module, for the H characteristic value and S according to the adjacent triangular faces for selecting tri patch and its shared side
Characteristic value carries out statistics with histogram, statistical data is put into feature vector, by the feature vector of tri patch adjacent with its three
The feature vector of edged surface piece is overlapped, feature vector of the stack result as selected tri patch;
The statistical data is put into feature vector specifically:
The H characteristic value of tri patch is subjected to statistics with histogram and obtains one group of 8 dimensional vector, the S characteristic value of tri patch is carried out
Statistics with histogram also obtains one group of 8 dimensional vector, and 8 dimensional vectors of 8 dimensional vectors of H characteristic value and S characteristic value are joined end to end composition
The feature vector of 16 dimensions;
The feature vector of the feature vector by tri patch and its adjacent triangular faces is overlapped specifically:
Wherein, { H11,H12,H13,...,H18,S11,S12,...,S18And { H21,H22,H23,...,H28,S21,S22,...,S28Table
Show the feature vector of 16 dimensions;
Identification module obtains recognition result for the feature vector of selected tri patch to be input in SVM identification model.
5. a kind of meadow identifying system of three-dimensional scenic based on tri patch according to claim 4, which is characterized in that
Histogram is carried out in described eigenvector computing module to be specifically configured to:
The group number that H feature value histogram and S feature value histogram is arranged is 8, group away from respectively H characteristic value value range 1/8
With the 1/8 of S characteristic value value range.
6. a kind of meadow identifying system of three-dimensional scenic based on tri patch according to claim 4, which is characterized in that
The training system of SVM identification model includes with lower unit in the identification module:
Sample collection unit, for collecting multiple positive samples and multiple negative samples as sample set, positive sample selection is all wait know
The two-dimension picture of other object, negative sample choose the two-dimension picture for being free of object to be identified;
Tri patch division unit, for be by the size adjusting of all samples it is consistent, select multiple preset scales to sample
The division for carrying out tri patch, extracts the HSI feature of all tri patch;
Feature vector computing unit, for the H characteristic value and S according to the adjacent triangular faces for selecting tri patch and its shared side
Characteristic value carries out statistics with histogram, statistical data is put into feature vector, by the feature vector of tri patch adjacent with its three
The feature vector of edged surface piece is overlapped, feature vector of the stack result as selected tri patch;
The statistical data is put into feature vector specifically:
The H characteristic value of tri patch is subjected to statistics with histogram and obtains one group of 8 dimensional vector, the S characteristic value of tri patch is carried out
Statistics with histogram also obtains one group of 8 dimensional vector, and 8 dimensional vectors of 8 dimensional vectors of H characteristic value and S characteristic value are joined end to end composition
The feature vector of 16 dimensions;
The feature vector of the feature vector by tri patch and its adjacent triangular faces is overlapped specifically:
Wherein, { H11,H12,H13,...,H18,S11,S12,...,S18And { H21,H22,H23,...,H28,S21,S22,...,S28Table
Show the feature vector of 16 dimensions;
Training unit is trained for the feature vector of all samples to be input to SVM;The decision function of SVM are as follows:
Wherein, L presentation class sum, value 2;ciRepresentative function coefficient, ci∈R;γ indicates the parameter of cross check, γ ∈ R+;Expression is chosen for trained support vector;h0Indicate the non-supported vector of training;Using the chi-Square measure of histogramAs kernel function, is defined as:
Wherein, N indicates feature vector total quantity in sample set;Indicate the corresponding support vector of k-th of feature vector;h0
(k) k-th of feature vector is indicated;
Model construction unit constructs SVM identification model for obtaining SVM identification model parameter after training.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710034186.6A CN106886754B (en) | 2017-01-17 | 2017-01-17 | Object identification method and system under a kind of three-dimensional scenic based on tri patch |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710034186.6A CN106886754B (en) | 2017-01-17 | 2017-01-17 | Object identification method and system under a kind of three-dimensional scenic based on tri patch |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106886754A CN106886754A (en) | 2017-06-23 |
CN106886754B true CN106886754B (en) | 2019-07-09 |
Family
ID=59176518
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710034186.6A Active CN106886754B (en) | 2017-01-17 | 2017-01-17 | Object identification method and system under a kind of three-dimensional scenic based on tri patch |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106886754B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108921952B (en) * | 2018-06-15 | 2022-09-06 | 深圳大学 | Object functionality prediction method, device, computer equipment and storage medium |
CN109344798B (en) * | 2018-10-24 | 2022-08-12 | 艾凯克斯(嘉兴)信息科技有限公司 | Three-dimensional object identification and detection method based on three-dimensional grid and neural network |
CN110059213A (en) * | 2019-03-20 | 2019-07-26 | 杭州电子科技大学 | A kind of threedimensional model classification retrieving method based on Density Estimator |
CN114882272A (en) * | 2022-04-22 | 2022-08-09 | 成都飞机工业(集团)有限责任公司 | Fusion analysis method for aerial manufacturing full-angle projection image surface patch attributes |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104598889A (en) * | 2015-01-30 | 2015-05-06 | 北京信息科技大学 | Human action recognition method and device |
CN105006021A (en) * | 2015-06-30 | 2015-10-28 | 南京大学 | Color mapping method and device suitable for rapid point cloud three-dimensional reconstruction |
CN105608421A (en) * | 2015-12-18 | 2016-05-25 | 中国科学院深圳先进技术研究院 | Human movement recognition method and device |
CN105678278A (en) * | 2016-02-01 | 2016-06-15 | 国家电网公司 | Scene recognition method based on single-hidden-layer neural network |
CN105912999A (en) * | 2016-04-05 | 2016-08-31 | 中国民航大学 | Human behavior identification method based on depth information |
CN106022228A (en) * | 2016-05-11 | 2016-10-12 | 东南大学 | Three-dimensional face recognition method based on vertical and horizontal local binary pattern on the mesh |
CN106204749A (en) * | 2016-07-07 | 2016-12-07 | 北京航空航天大学 | The threedimensional model of sparse low-rank feature representation is divided into segmentation method |
-
2017
- 2017-01-17 CN CN201710034186.6A patent/CN106886754B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104598889A (en) * | 2015-01-30 | 2015-05-06 | 北京信息科技大学 | Human action recognition method and device |
CN105006021A (en) * | 2015-06-30 | 2015-10-28 | 南京大学 | Color mapping method and device suitable for rapid point cloud three-dimensional reconstruction |
CN105608421A (en) * | 2015-12-18 | 2016-05-25 | 中国科学院深圳先进技术研究院 | Human movement recognition method and device |
CN105678278A (en) * | 2016-02-01 | 2016-06-15 | 国家电网公司 | Scene recognition method based on single-hidden-layer neural network |
CN105912999A (en) * | 2016-04-05 | 2016-08-31 | 中国民航大学 | Human behavior identification method based on depth information |
CN106022228A (en) * | 2016-05-11 | 2016-10-12 | 东南大学 | Three-dimensional face recognition method based on vertical and horizontal local binary pattern on the mesh |
CN106204749A (en) * | 2016-07-07 | 2016-12-07 | 北京航空航天大学 | The threedimensional model of sparse low-rank feature representation is divided into segmentation method |
Non-Patent Citations (1)
Title |
---|
多特征和SVM相融合的三维物体识别方法;湛宁;《计算机仿真》;20130331;第30卷(第3期);第380-383、407页 |
Also Published As
Publication number | Publication date |
---|---|
CN106886754A (en) | 2017-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9639748B2 (en) | Method for detecting persons using 1D depths and 2D texture | |
CN104182765B (en) | Internet image driven automatic selection method of optimal view of three-dimensional model | |
CN106384126B (en) | Clothes fashion recognition methods based on contour curvature characteristic point and support vector machines | |
CN108549873A (en) | Three-dimensional face identification method and three-dimensional face recognition system | |
CN113177977B (en) | Non-contact three-dimensional human body size measuring method | |
CN111160269A (en) | Face key point detection method and device | |
CN106886754B (en) | Object identification method and system under a kind of three-dimensional scenic based on tri patch | |
CN104835175B (en) | Object detection method in a kind of nuclear environment of view-based access control model attention mechanism | |
CN108319957A (en) | A kind of large-scale point cloud semantic segmentation method based on overtrick figure | |
CN110569782A (en) | Target detection method based on deep learning | |
CN104240264A (en) | Height detection method and device for moving object | |
CN106709452B (en) | Instrument position detection method based on intelligent inspection robot | |
CN103295025A (en) | Automatic selecting method of three-dimensional model optimal view | |
CN105488541A (en) | Natural feature point identification method based on machine learning in augmented reality system | |
CN108182705A (en) | A kind of three-dimensional coordinate localization method based on machine vision | |
Li et al. | High-throughput phenotyping analysis of maize at the seedling stage using end-to-end segmentation network | |
CN110533648A (en) | A kind of blackhead identifying processing method and system | |
CN113012129A (en) | System and device for counting area positioning and marked nerve cells of brain slice image | |
CN110910497B (en) | Method and system for realizing augmented reality map | |
CN102708589B (en) | Three-dimensional target multi-viewpoint view modeling method on basis of feature clustering | |
Zhang et al. | Fully automatic system for fish biomass estimation based on deep neural network | |
Gao et al. | Improved binocular localization of kiwifruit in orchard based on fruit and calyx detection using YOLOv5x for robotic picking | |
Zhang et al. | Pedestrian detection with EDGE features of color image and HOG on depth images | |
CN117079125A (en) | Kiwi fruit pollination flower identification method based on improved YOLOv5 | |
Yang et al. | Intelligent collection of rice disease images based on convolutional neural network and feature matching |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |