CN107895167A - A kind of ir data classifying identification method based on rarefaction representation - Google Patents
A kind of ir data classifying identification method based on rarefaction representation Download PDFInfo
- Publication number
- CN107895167A CN107895167A CN201710948615.0A CN201710948615A CN107895167A CN 107895167 A CN107895167 A CN 107895167A CN 201710948615 A CN201710948615 A CN 201710948615A CN 107895167 A CN107895167 A CN 107895167A
- Authority
- CN
- China
- Prior art keywords
- sample
- forecast
- training
- matrix
- sparse
- 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.)
- Pending
Links
Classifications
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of ir data classifying identification method based on rarefaction representation, its technical scheme is to carry out Sparse to forecast sample with training sample using the algorithm of rarefaction representation to represent, realizes the Fast Classification of forecast sample.Concretely comprise the following steps：(1) training sample matrix is constructed；(2) training sample matrix is standardized；(3) sparse solution is solved；(4) forecast sample is estimated according to certain a kind of sparse solution and sample information.(5) classification is adjudicated.The current existing mode identification method to spectroscopic data generally requires smoothly, the pretreatment mode such as variables choice or needs space projection, Fourier's series scheduling algorithm to lift the validity of sorting algorithm.The invention provides a kind of do not need the spectroscopic data sorting technique that is pre-processed and changed accordingly, by sparsely linear expression with regard to can find with forecast sample similar in training sample realize classification, simple to operate, speed is fast.
Description
Technical field
The present invention relates to ir data classifying identification method, in particular to a kind of infrared light based on rarefaction representation
Modal data classifying identification method.
Background technology
Infrared spectrum detection has the characteristics of quick, safety, low cost, lossless, and material is carried out soon with infrared spectrum
Speed detection is a kind of effective method.But generally require to put down when spectroscopic data is carried out with algorithm quickly identification
Sliding, the pretreatment mode such as variables choice either needs space projection or Fourier's series scheduling algorithm to lift sorting algorithm
Validity, but these processes do not have actual meaning for classification.
Therefore, a kind of pre- place that can be reduced to ir data progress feature selecting or spatial alternation etc is needed badly
Reason process, ir data Fast Classification is identified so as to realize.
The content of the invention
Technical problem to be solved of the embodiment of the present invention is, there is provided a kind of infrared spectrum number based on rarefaction representation
According to classifying identification method, this method can be realized quickly carries out Classification and Identification to infrared data, while avoids infrared data need
Carry out the preprocessing process of feature selecting or spatial alternation etc.
To achieve the above object, the technical scheme is that comprising the following steps：
S01：Selection target sample y, some sample for needing to identify its ir data class label is specified, is done
Embark on journey be classified as dimension 1 matrix；
S02：Training sample matrix is constructed, the ir data for the training sample for having known class label is lined up
For dimension, the form of sample number is classified as, and similar sample is put together, the matrix of the A in construction linear system y=Ax, square
Matrix representation is as follows：A=[A1,A2,...Ak]=[v1,1,v1,2,...,vk,nk]；
S03：Training sample matrix is standardized, to the two of each training sample divided by corresponding training sample
Tie up norm；
S04：Sparse solution is solved, for predicting target sample y, linear system y=Ax solution is solved, for infrared spectrum
For data, its dimension is more than sample number, i.e. linear system y=Ax is overdetermination, by seeking the 1 rank norm for solving this most
It is small to approach the sparse effect of 0 rank Norm minimum, then obtain sparse solution x1；
S05：Forecast sample is estimated according to certain a kind of sparse solution and sample information, tried to achieve by previous step dilute
Training samples information before discongesting and being corresponding, tries to achieve the matrix product of training set and sparse solution corresponding to this kind of label, produces
The information of this raw a kind of training sample is indicated to forecast sample, obtains certain estimation of class training data to forecast sample,
It is exactly
S06：Classification is adjudicated, according to the difference of certain estimation of a kind of training set to forecast sample and actual prediction sample come
Make decisions, using the two-dimentional norm of certain a kind of sample estimates and forecast sample difference as standard, with actual prediction sample difference
Classification of the different minimum class label as forecast set, the two-dimentional norm are certain a kind of sample estimates and forecast sample difference
Euclidean distance；
S07：Export the class label of forecast sample.
Further setting is in described step S04, using single order Homotopy function bags, set iterations and
Fault-tolerant size, the solution of this single order Norm minimum is solved to approach the sparse solution of zeroth order Norm minimum, if the class of forecast sample
Wei the i-th class, then preferable sparse solution is x0=[0 ..., 0, αi,1,αi,2,…,αi,60,0,…,0]T。
Implement the embodiment of the present invention, have the advantages that：
Invention introduces the algorithm of rarefaction representation, is carried out according to unknown sample for the degree of approximation of known sample
Judgement, such process have actual meaning for classification.Step is to first pass through the sample known to by unknown sample to carry out
Rarefaction representation, corresponding sparse solution is obtained, unknown sample is then represented by the distribution of the sparse solution of known label information
Mode, the Fast Classification of spectroscopic data is realized with this, method of the invention to ir data without carrying out feature choosing
Select or the preprocessing process of spatial alternation etc, there is the effect quickly identified.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will to embodiment or
The required accompanying drawing used is briefly described in description of the prior art, it should be apparent that, drawings in the following description are only
Some embodiments of the present invention, for those of ordinary skill in the art, without having to pay creative labor,
Other accompanying drawings, which are obtained, according to these accompanying drawings still falls within scope of the invention.
The flow chart of the ir data classifying identification method of Fig. 1 present invention；
Fig. 2 embodiment of the present invention is implementing procedure figure.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, the present invention is made into one below in conjunction with accompanying drawing
It is described in detail on step ground.
The direction and position term that the present invention is previously mentioned, for example, " on ", " under ", "front", "rear", "left", "right", "
It is interior ", " outer ", " top ", " bottom ", " side " etc., be only refer to the attached drawing direction or position.Therefore, the direction and position used
It is to illustrate and understand the present invention to put term, rather than limiting the scope of the invention.
As shown in Figure 1 to Figure 2, it is in the embodiment of the present invention, there is provided a kind of ir data based on rarefaction representation point
Class recognition methods, comprises the following steps：
S01, selection target sample y.One shares five class mud blood clam samples, respectively mud blood clam a, mud blood clam b, mud blood clam c, mud blood clam d,
Mud blood clam e.The column matrix that the infrared data of each mud blood clam sample is 3000 × 1.A unknown mud blood clam sample is selected as target
Sample.
S02, construct training sample matrix.Every kind of mud blood clam sample number of known class label is 15, then by mud blood clam sample
Notebook data lines up behavior dimension, is classified as the form of sample number, and 1~15 is classified as first kind mud blood clam a, and 16~30 are classified as second
Class mud blood clam b, by that analogy, the matrix of the A in construction linear system y=Ax, A=[A1,A2,...A5]=[v1,1,v1,2,...,
v5,15*5]。
S03, training sample matrix is standardized.Column matrix point divided by this sample to each sample
Two-dimentional norm, is standardized, and turns into new training sample matrix, that is, vi,j=vi,j/||vi,j||2。
S04, solve sparse solution.Under conditions of known training sample matrix A and forecast sample y, linear system is solved
Y=Ax solution.Using single order Homotopy function bags, the solving condition parameter such as iterations, fault-tolerant size is set, is solved
The solution of this single order Norm minimum approaches the sparse solution of zeroth order Norm minimum, if the classification of forecast sample is the i-th class, that
Preferable sparse solution is x0=[0 ..., 0, αi,1,αi,2,…,αi,60,0,…,0]T。
S05, forecast sample is estimated according to certain a kind of sparse solution and corresponding sample information.Obtained by previous step
The sparse solution x arrived0, and sample information A before, the estimation to one forecast sample of every a kind of all generations,Because A before is the label information that exists sample, so the distribution of corresponding sparse solution is also to exist
Label information.Every a kind of predicted estimate is tried to achieve by every a kind of sparse solution, due to the presence of sparse solution, for be not with
The of a sort estimation of forecast set 0 matrix that very likely output is 3000 × 1.
S06, classification judgement.Sentenced according to the difference of the estimation of certain a kind of forecast sample and actual prediction sample
Certainly, otherness here is to be used as standard by the Euclidean distance between sample estimates and forecast sample, with that of error minimum
Classification of the class label as the forecast sample of reality.I.e.
S07, export the class label of forecast sample.Output
Can be with one of ordinary skill in the art will appreciate that realizing that all or part of step in above-described embodiment method is
The hardware of correlation is instructed to complete by program, described program can be stored in a computer read/write memory medium,
Described storage medium, such as ROM/RAM, disk, CD.
Above disclosure is only preferred embodiment of present invention, can not limit the right of the present invention with this certainly
Scope, therefore the equivalent variations made according to the claims in the present invention, still belong to the scope that the present invention is covered.
Claims (2)
1. a kind of ir data classifying identification method based on rarefaction representation, it is characterised in that comprise the following steps：
S01：Selection target sample y, some sample for needing to identify its ir data class label is specified, makes behavior
Dimension, the matrix for being classified as 1；
S02：Training sample matrix is constructed, the ir data for the training sample for having known class label is lined up into behavior dimension
Degree, the form of sample number is classified as, and similar sample is put together, the matrix of the A in construction linear system y=Ax, matrix table
Show as follows：A=[A1, A2... Ak]=[v1,1, v1,2..., vK, nk]；
S03：Training sample matrix is standardized, to each training sample divided by the two-dimentional model of corresponding training sample
Number；
S04：Sparse solution is solved, for predicting target sample y, linear system y=Ax solution is solved, for ir data
Speech, its dimension are more than sample number, i.e. linear system y=Ax is overdetermination, by asking the 1 rank Norm minimum for solving this to approach
The sparse effect of 0 rank Norm minimum, then obtains sparse solution x1；
S05：Forecast sample is estimated according to certain a kind of sparse solution and sample information, the sparse solution tried to achieve by previous step with
And the training samples information before corresponding to, the matrix product of training set and sparse solution corresponding to this kind of label is tried to achieve, produces this
The information of class training sample is indicated to forecast sample, obtains certain estimation of class training data to forecast sample, that is,
S06：Classification is adjudicated, and is sentenced according to the difference of certain estimation of a kind of training set to forecast sample and actual prediction sample
Certainly, standard is used as using the two-dimentional norm of certain a kind of sample estimates and forecast sample difference, with actual prediction differences between samples minimum
Classification of the class label as forecast set, the two-dimentional norm be certain a kind of sample estimates with forecast sample difference it is European away from
From；
S07：Export the class label of forecast sample.
2. according to the method for claim 1, it is characterised in that：In described step S04, single order Homotopy functions are utilized
Bag, sets iterations and fault-tolerant size, solves the solution of this single order Norm minimum to approach the sparse of zeroth order Norm minimum
Solution, if the classification of forecast sample is the i-th class, then preferable sparse solution is x0=[0 ..., 0, αI, 1, αI, 2..., αI, Ti,
0 ..., 0]T, wherein Ti is the number of certain class training sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710948615.0A CN107895167A (en) | 2017-10-12 | 2017-10-12 | A kind of ir data classifying identification method based on rarefaction representation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710948615.0A CN107895167A (en) | 2017-10-12 | 2017-10-12 | A kind of ir data classifying identification method based on rarefaction representation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107895167A true CN107895167A (en) | 2018-04-10 |
Family
ID=61803514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710948615.0A Pending CN107895167A (en) | 2017-10-12 | 2017-10-12 | A kind of ir data classifying identification method based on rarefaction representation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107895167A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109060771A (en) * | 2018-07-26 | 2018-12-21 | 温州大学 | A kind of common recognition model building method based on spectrum different characteristic collection |
| CN111930935A (en) * | 2020-06-19 | 2020-11-13 | 普联国际有限公司 | Image classification method, device, equipment and storage medium |
| CN111930935B (en) * | 2020-06-19 | 2024-06-07 | 普联国际有限公司 | Image classification method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550989A (en) * | 2015-12-09 | 2016-05-04 | 西安电子科技大学 | Image super-resolution method based on nonlocal Gaussian process regression |
| CN107067034A (en) * | 2017-04-15 | 2017-08-18 | 温州大学 | A kind of method and system of quick identification ir data classification |
-
2017
- 2017-10-12 CN CN201710948615.0A patent/CN107895167A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550989A (en) * | 2015-12-09 | 2016-05-04 | 西安电子科技大学 | Image super-resolution method based on nonlocal Gaussian process regression |
| CN107067034A (en) * | 2017-04-15 | 2017-08-18 | 温州大学 | A kind of method and system of quick identification ir data classification |
Non-Patent Citations (1)
| Title |
|---|
| 钱胡茜: "ℓ1范数极小化的稀疏表示模型与算法", 《宜宾学院学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109060771A (en) * | 2018-07-26 | 2018-12-21 | 温州大学 | A kind of common recognition model building method based on spectrum different characteristic collection |
| CN109060771B (en) * | 2018-07-26 | 2020-12-29 | 温州大学 | Consensus model construction method based on different characteristic sets of spectrum |
| CN111930935A (en) * | 2020-06-19 | 2020-11-13 | 普联国际有限公司 | Image classification method, device, equipment and storage medium |
| CN111930935B (en) * | 2020-06-19 | 2024-06-07 | 普联国际有限公司 | Image classification method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10956817B2 (en) | Unsupervised domain adaptation with similarity learning for images | |
| CN106462746B (en) | Analyzing digital holographic microscopy data for hematology applications | |
| CN102982349B (en) | A kind of image-recognizing method and device | |
| CN105224951B (en) | A kind of vehicle type classification method and sorter | |
| US20170032247A1 (en) | Media classification | |
| Sundara Sobitha Raj et al. | DDLA: dual deep learning architecture for classification of plant species | |
| Kaur et al. | A survey on machine learning algorithms | |
| CN1723468A (en) | Computer vision system and method employing illumination invariant neural networks | |
| Zhai et al. | A generative adversarial network based framework for unsupervised visual surface inspection | |
| KR20230107558A (en) | Model training, data augmentation methods, devices, electronic devices and storage media | |
| CN111178196B (en) | Cell classification method, device and equipment | |
| US20150242676A1 (en) | Method for the Supervised Classification of Cells Included in Microscopy Images | |
| Moate et al. | Vehicle detection in infrared imagery using neural networks with synthetic training data | |
| CA3002100A1 (en) | Unsupervised domain adaptation with similarity learning for images | |
| CN107895167A (en) | A kind of ir data classifying identification method based on rarefaction representation | |
| CN111783088A (en) | Malicious code family clustering method and device and computer equipment | |
| Nimitha et al. | An improved deep convolutional neural network architecture for chromosome abnormality detection using hybrid optimization model | |
| Wiliem et al. | A bag of cells approach for antinuclear antibodies HEp‐2 image classification | |
| CN116580272A (en) | Radar target classification method and system based on model fusion reasoning | |
| Pandey et al. | An image processing approach for analyzing assessment of pavement distress | |
| CN113468936A (en) | Food material identification method, device and equipment | |
| CN110879821A (en) | Method, device, equipment and storage medium for generating rating card model derivative label | |
| Khosla et al. | Integrating randomization and discrimination for classifying human-object interaction activities | |
| Ghazali et al. | Classification of images for automatic textual annotation: A review of techniques | |
| CN112766156B (en) | Riding attribute identification method and device and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180410 |