CN110097135B - Holographic diffraction label image recognition algorithm based on double tensors - Google Patents
Holographic diffraction label image recognition algorithm based on double tensors Download PDFInfo
- Publication number
- CN110097135B CN110097135B CN201910384032.9A CN201910384032A CN110097135B CN 110097135 B CN110097135 B CN 110097135B CN 201910384032 A CN201910384032 A CN 201910384032A CN 110097135 B CN110097135 B CN 110097135B
- Authority
- CN
- China
- Prior art keywords
- tensor
- similarity
- holographic diffraction
- original image
- image
- 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
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 39
- 238000012360 testing method Methods 0.000 claims abstract description 33
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 32
- 239000013598 vector Substances 0.000 claims abstract description 21
- 238000010219 correlation analysis Methods 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 25
- 230000009977 dual effect Effects 0.000 claims description 17
- 238000000513 principal component analysis Methods 0.000 claims description 17
- 230000004927 fusion Effects 0.000 claims description 14
- 238000000034 method Methods 0.000 claims description 12
- 238000012937 correction Methods 0.000 claims description 5
- 238000003708 edge detection Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 4
- 238000005266 casting Methods 0.000 claims 1
- 238000011524 similarity measure Methods 0.000 claims 1
- 238000000605 extraction Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 239000003086 colorant Substances 0.000 description 2
- 238000002591 computed tomography Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000002595 magnetic resonance imaging Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000007639 printing Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000002902 bimodal effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 238000003825 pressing Methods 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a holographic diffraction label image recognition algorithm based on double tensors, which comprises the following steps: s11, obtaining an original image of the holographic diffraction label, wherein the original image comprises an original image of a training sample and an original image of a test sample; s12, preprocessing the acquired original image of the test sample to generate an HSV tensor and extract an HOG tensor; preprocessing the acquired original image of the training sample to generate an HSV tensor and extract an HOG tensor; s13, forming a double tensor by the HSV tensor and the HOG tensor of the obtained test sample; forming the HSV tensor and the HOG tensor of the obtained training sample into a double tensor; measuring the similarity between different decomposition matrixes of the training sample and the test sample through canonical correlation analysis; s14, classifying the similarity between the different decomposition matrixes by using a nearest neighbor algorithm. The similarity vectors of different decomposition matrixes are projected to the PCA subspace to carry out KNN classification, and the identification performance among different samples is effectively improved.
Description
Technical Field
The invention relates to the technical field of computer image recognition, in particular to a holographic diffraction label image recognition algorithm based on double tensors.
Background
With the development of multimedia and printing technologies, digital images are widely used in the fields of science, industry, commerce, and the like. In the medical field, digital image storage technologies (e.g., CT (Computed Tomography) and MRI (Magnetic Resonance Imaging)) are used to image images on solid film, and automatic diagnosis is achieved by content-based image information retrieval [1]; the law enforcement department records facial features and fingerprint information of victims and suspects by using digital images, records crime scenes, and performs image similarity analysis in a database to quickly find criminals; a meteorological department acquires a remote sensing image of a satellite and analyzes the spectral characteristics of the image to monitor the environment and climate change [2]; the brand manufacturer uses the laser anti-counterfeiting technology to manufacture the commodity logo into a laser mould pressing holographic image anti-counterfeiting label, so that the anti-counterfeiting purpose is realized. The wide application of digital images makes image feature recognition more and more interesting. The laser holographic anti-counterfeiting label is a crystal of a high and new technology as a technical result of modern laser application, and has gained general attention at home and abroad in recent years. Compared with the common printing label, the laser holographic anti-counterfeiting label has the advantages of unique anti-counterfeiting function, large information amount, clear image, bright color, strong stereoscopic impression and one-time use.
The wide use of the holographic diffraction label brings great challenges to the traditional image recognition, and the traditional image recognition algorithm cannot obtain stable recognition effect due to the physical characteristics of the holographic diffraction label changing along with light.
For example, a patent with publication number CN1893694 discloses a camera phone with a barcode anti-counterfeiting query function, which includes a camera for acquiring images, a microprocessor, and a first wireless communication transceiver, where the microprocessor includes a main control module for implementing basic functions of the phone; an image acquisition module; the image processing module is used for identifying the acquired bar code image, analyzing the bar code image into a binary sequence and reversely obtaining data information corresponding to the bar code according to the bar code compiling method; the bar code unloading module is used for transferring bar code data information into the first wireless transmitting device and transmitting the bar code data information to a remote service center with anti-counterfeiting identification; and the anti-counterfeiting query result receiving module is used for receiving the query result of the remote service center through the first wireless receiving device. And the anti-counterfeiting inquiry system based on the camera mobile phone. The invention is based on a common camera mobile phone, has the effects of low cost, economy, practicability and convenient use, and can identify the authenticity of commodities or bills according to bar codes. Although the above patent can identify the authenticity of goods or bills based on bar codes, it employs a conventional image algorithm. Due to the physical property of the holographic diffraction label changing with light, the traditional image recognition algorithm cannot obtain a stable recognition effect.
Disclosure of Invention
The invention aims to provide a holographic diffraction label image identification algorithm based on double tensors aiming at the defects of the prior art. The automatic identification of the diffraction label can be more efficient and convenient, and the holographic diffraction label has better identification performance due to the complementation of the identification capabilities of different tensors.
In order to achieve the purpose, the invention adopts the following technical scheme:
a holographic diffraction label image recognition algorithm based on double tensors comprises the following steps:
s1, obtaining an original image of a holographic diffraction label, wherein the original image comprises an original image of a training sample and an original image of a test sample;
s2, preprocessing the acquired original image of the test sample to generate an HSV tensor and extract an HOG tensor;
preprocessing the acquired original image of the training sample to generate an HSV tensor and extract an HOG tensor;
s3, forming a double tensor by the HSV tensor and the HOG tensor of the obtained test sample; forming a double tensor by the HSV tensor and the HOG tensor of the obtained training sample; measuring the similarity between different decomposition matrixes of the training sample and the test sample through canonical correlation analysis;
and S4, classifying the similarity between the different decomposition matrixes by using a nearest neighbor algorithm.
Further, the step S2 specifically includes:
and respectively converting the original images of the test sample and the training sample into gray images.
Further, the step S2 further includes:
and converting the gray level image into a binary image by a maximum inter-class variance method, and performing Canny edge detection on the binary image.
Further, the step S2 further includes:
gradient information is calculated for all pixel positions of the gray scale image.
Further, the preprocessing of the original image in step S2 further includes rotation correction.
Further, the step S3 specifically includes:
a. generation of a decomposition matrix: using a high-order singular value decomposition algorithm to expand the high-order tensor to form a two-dimensional matrix;
b. typical correlation analysis similarity measurements: measuring the similarity between the dual tensor differential decomposition matrixes of the training sample and the test sample by using canonical correlation analysis;
c. similarity fusion: considering each decomposition matrix as an independent unit, similarity fusion is used to combine the similarity of the dual tensors of the training sample and the test sample.
Further, the typical correlation analysis similarity measurement specifically includes:
ρ＝maxcorr(u T x，v T y)；
where u and v represent typical variables, ρ represents typical correlation, and corr (x, y) represents correlation of x and y.
Further, the similarity fusion is specifically to find a projection subspace using principal component analysis.
Further, the searching the projection shadow space specifically includes:
will make Ψ 1 ，Ψ 2 ，……Ψ t Finding a subspace of principal component analysis as a similarity vector of training samples, the similarity vector Ψ 1 ，Ψ 2 ，……Ψ t Is defined as:
the scatter matrix is:
the scatter matrix is decomposed into:
S＝ΦΛΦ T ；
wherein,
the subspace of the principal component analysis is composed of eigenvectors corresponding to the eigenvalues of the scatter matrix S;
generating a test similarity vector:
Ψ T ＝{max(Ψ j，1 )，…，max(Ψ j ， n )|j＝1，2，…t}；
wherein, { max (Ψ) j，n ) I j =1,2, \8230; t } represents the maximum value of the training sample vector at the nth column;
all training sample vectors Ψ 1 ，Ψ 2 ，……Ψ t And Ψ T Projected into the subspace of the principal component analysis.
Further, the original image of the holographic diffraction label acquired in step S1 is acquired by a camera or a video camera.
Compared with the prior art, the invention designs a decomposition matrix similarity fusion strategy, projects the similarity vectors of different decomposition matrices to the principal component analysis subspace and classifies the similarity vectors through the nearest neighbor algorithm, thereby effectively improving the identification performance among different samples; the effectiveness of the fusion strategy is verified through experiments; the dual tensor identification algorithm provided by the invention has better identification performance for the holographic diffraction label because the identification capabilities of different tensors are complemented.
Drawings
Fig. 1 is a flowchart of an image identification algorithm based on a double-tensor holographic diffraction label according to an embodiment;
FIG. 2 is a flow chart of a dual tensor identification algorithm provided in accordance with an embodiment one;
FIG. 3 is a graph of the rotation correction results provided in the first embodiment;
FIG. 4 is a diagram of an example of HOG feature extraction provided in the first embodiment;
FIG. 5 is a matrix diagram developed by the third order tensor provided by the first embodiment;
FIG. 6 is a flow chart of a fusion algorithm provided in one embodiment;
FIG. 7 is a sample schematic view of different illumination conditions provided by the first embodiment;
FIG. 8 is a graph comparing the results of the experiments provided in the first example.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.
The invention aims to provide a holographic diffraction label image identification algorithm based on double tensors aiming at the defects of the prior art.
Example one
The embodiment provides a holographic diffraction label image recognition algorithm based on double tensors, as shown in fig. 1, including the steps of:
s11, obtaining an original image of the holographic diffraction label, wherein the original image comprises an original image of a training sample and an original image of a test sample;
s12, preprocessing the acquired original image of the test sample to generate an HSV tensor and extract an HOG tensor;
preprocessing the acquired original image of the training sample to generate an HSV tensor and extract an HOG tensor;
s13, forming a double tensor by the HSV tensor and the HOG tensor of the obtained test sample; forming the HSV tensor and the HOG tensor of the obtained training sample into a double tensor; measuring the similarity between different decomposition matrixes of the training sample and the test sample through canonical correlation analysis;
and S14, classifying the similarity between the different decomposition matrixes by using a nearest neighbor algorithm.
Because the tensor can keep the space-time characteristics of the original sample in the characteristic extraction process, the embodiment provides the holographic diffraction label identification algorithm fusing the dual-tensor characteristics. As shown in fig. 2, first, color information and gradient information of a sample are respectively expressed using HSV tensor and histogram of oriented gradient tensor of the sample; then, obtaining a decomposition matrix of the double tensors through tensor decomposition; finally, the similarity between different decomposition matrixes of the tensor is calculated by using typical correlation analysis.
In the embodiment, considering that different decomposition matrices have different recognition capabilities, a decomposition matrix similarity fusion strategy is designed, similarity vectors of different decomposition matrices are projected to a principal component analysis subspace to be classified through a nearest neighbor algorithm, and the recognition performance among different samples can be effectively improved; the effectiveness of the fusion strategy is verified through experiments; the dual tensor identification algorithm provided by the embodiment has better identification performance for the holographic diffraction label because the identification capabilities of different tensors are complemented.
In step S11, an original image of the holographic diffraction label is acquired, wherein the original image includes an original image of the training sample and an original image of the test sample.
Wherein the original image of the holographic diffraction label is acquired by a camera or a video camera.
In step S12, preprocessing the acquired original image of the test sample to generate an HSV tensor and extract an HOG tensor;
and preprocessing the acquired original image of the training sample to generate an HSV tensor and extract an HOG tensor.
Specifically, as shown in fig. 3, the preprocessing includes performing rotation correction on an original image, converting the image after the rotation correction into a gray image, converting the gray image into a binary image by a maximum inter-class variance method in order to remove the influence of an image background, and performing Canny edge detection on the binary image; gradient information is calculated for all pixel positions of the gray scale image.
And converting the preprocessed image into an HSV color space, thereby generating an HSV tensor.
And (3) carrying out normalization processing on the original data of the preprocessed image, and obtaining an HOG tensor through HOG feature extraction, as shown in figure 4.
In the traditional HOG feature extraction, an image is divided into small cell cells, the gradient direction of each cell is counted, a gradient direction histogram is constructed, and the direction histograms of several cells form a block histogram. The conventional HOG feature extraction is one-dimensional, and the present embodiment expresses the HOG feature extraction as three-dimensions to constitute the HOG tensor.
In step S13, the HSV tensor and the HOG tensor of the obtained test sample are combined into a double tensor; forming the HSV tensor and the HOG tensor of the obtained training sample into a double tensor; the similarity between different decomposition matrices of the training sample and the test sample is measured by a canonical correlation analysis, as shown in fig. 5 and 6.
Specifically, the obtained HOG tensor and HSV tensor are used as primary features of the image, and similarity between the test sample and the decomposition matrix of the training sample is measured through typical correlation analysis.
Wherein, a generation of decomposition matrix: and (3) unfolding the high-order tensor to form a two-dimensional matrix by using a high-order singular value decomposition algorithm.
b. Typical correlation analysis similarity measurements: measuring the similarity between the dual tensor differential decomposition matrixes of the training sample and the test sample by using canonical correlation analysis;
ρ＝maxcorr(u T x，v T y)；
where u and v represent typical variables, ρ represents typical correlation, and corr (x, y) represents correlation of x and y.
c. Similarity fusion: considering each decomposition matrix as an independent unit, similarity fusion is used to combine the similarity of the dual tensors of the training sample and the test sample.
Wherein the similarity fusion is specifically to find an optimal projection subspace using principal component analysis.
Will make Ψ 1 ，Ψ 2 ，……Ψ t Finding a subspace of principal component analysis as a similarity vector of training samples, the similarity vector Ψ 1 ，Ψ 2 ，……Ψ t Is defined as:
the scatter matrix is:
the scatter matrix is decomposed into:
S＝ΦΛΦ T ；
wherein,
the subspace of the principal component analysis is composed of eigenvectors corresponding to the eigenvalues of the scatter matrix S.
In this embodiment, the subspace of the principal component analysis is composed of eigenvectors corresponding to the first few larger eigenvalues of the dispersion matrix S.
And simultaneously generating a test similarity vector:
Ψ T ＝{max(Ψ j，1 )，…，max(Ψ j ， n )|j＝1，2，…t}；
wherein { max (Ψ) j，n ) L j =1,2, \8230t } represents the maximum value of the training sample vector for the nth column;
in the present embodiment, n =6, where { max (Ψ) j，1 ) L j =1,2, \8230lt } represents the maximum value of the training sample vector for the 1 st column.
Since the larger Ψ, the higher the similarity, the largest value per dimension is used as the test vector.
All training sample vectors Ψ 1 ，Ψ 2 ，……Ψ t And Ψ T Projection onto principal component analysisIn the space.
In step S14, the similarities between the different decomposition matrices are classified using a nearest neighbor algorithm.
FIG. 7 is a schematic diagram of samples under different lighting conditions.
Fig. 8 is a graph showing a comparison of experimental results.
Interpretation of professional terms:
histogram of Oriented Gradient (HOG) feature is a feature descriptor used in computer vision and image processing to perform object detection, and the HOG feature constitutes a feature by calculating and counting a Histogram of Oriented gradients of local regions of an image.
HSV (Value) is a color space created according to the intuitive characteristics of colors, also called a hexagonal pyramid Model (Hexcone Model); the parameters of the colors in this model are: hue (H), saturation (S), lightness (V).
Principal Component Analysis (PCA), a multivariate statistical Analysis method for selecting a small number of important variables from a plurality of variables by linear transformation; also known as principal component analysis.
A K-Nearest Neighbor (KNN) classification algorithm, wherein the method has the following steps: if a sample belongs to a certain class in the majority of the k most similar samples in feature space (i.e. the nearest neighbors in feature space), then the sample also belongs to this class. In the KNN algorithm, the selected neighbors are all objects that have been correctly classified. The method only determines the category of the sample to be classified according to the category of the nearest sample or a plurality of samples in the classification decision. The KNN method, although in principle also depends on the limit theorem, is only associated with a very small number of neighboring samples in the class decision. The KNN method mainly determines the class by the limited adjacent samples around, and not by the method of distinguishing the class domain.
The CCA (canonical correlation analysis) is an understanding of the cross-covariance matrix, and is a multivariate statistical analysis method that reflects the overall correlation between two sets of indexes by using the correlation between the synthetic variable pairs. The basic principle is as follows: in order to grasp the correlation between the two sets of indexes as a whole, two representative comprehensive variables U1 and V1 (each being a linear combination of each variable in the two variable sets) are extracted from the two sets of variables, respectively, and the overall correlation between the two sets of indexes is reflected by the correlation between the two comprehensive variables.
Canny aims to find an optimal edge detection algorithm, and the meaning of optimal edge detection is as follows:
(1) Optimal detection: the algorithm can identify actual edges in the image as much as possible, and the probability of missing detection of the actual edges and the probability of false detection of non-edges are both as small as possible;
(2) Optimal positioning criterion: the position of the detected edge point is closest to the position of the actual edge point, or the degree that the detected edge deviates from the real edge of the object due to the influence of noise is minimum;
(3) The detection points correspond to the edge points one by one: the edge points detected by the operator should have a one-to-one correspondence with the actual edge points.
Normalization is a simplified calculation mode, namely, a dimensional expression is transformed into a dimensionless expression to become a scalar.
The maximum between class variance (OTSU) is a method to automatically threshold from what is suitable for bimodal situations. It is to divide the image into background and object 2 parts according to the gray scale characteristics of the image. The larger the inter-class variance between the background and the object, the larger the difference of 2 parts constituting the image, and when part of the object is mistaken for the background or part of the background is mistaken for the object, the 2 parts of the difference becomes smaller. Thus, a segmentation that maximizes the inter-class variance means that the probability of false positives is minimized.
It is to be noted that the foregoing description is only exemplary of the invention and that the principles of the technology may be employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A holographic diffraction label image recognition algorithm based on double tensors is characterized by comprising the following steps:
s1, obtaining an original image of a holographic diffraction label, wherein the original image comprises an original image of a training sample and an original image of a test sample;
s2, preprocessing the acquired original image of the test sample to generate an HSV tensor and extract an HOG tensor;
preprocessing the acquired original image of the training sample to generate an HSV tensor and extract an HOG tensor;
s3, forming a double tensor by the HSV tensor and the HOG tensor of the obtained test sample; forming the HSV tensor and the HOG tensor of the obtained training sample into a double tensor; measuring the similarity between different decomposition matrixes of the training sample and the test sample through canonical correlation analysis;
and S4, classifying the similarity between the different decomposition matrixes by using a nearest neighbor algorithm.
2. The dual tensor-based holographic diffraction label image identification algorithm as claimed in claim 1, wherein the step S2 specifically comprises:
and respectively converting the original images of the test sample and the training sample into gray images.
3. The dual tensor-based holographic diffraction label image recognition algorithm as recited in claim 2, wherein the step S2 further comprises:
and converting the gray level image into a binary image by a maximum inter-class variance method, and performing Canny edge detection on the binary image.
4. The dual tensor-based holographic diffraction label image identification algorithm as recited in claim 2, wherein the step S2 further comprises:
gradient information is calculated for all pixel positions of the grayscale image.
5. The dual tensor-based holographic diffraction label image recognition algorithm as claimed in any one of claims 1 to 4, wherein the preprocessing of the original image in the step S2 further comprises rotation correction.
6. The dual tensor-based holographic diffraction label image identification algorithm as claimed in claim 5, wherein the step S3 specifically comprises:
a. and (3) generation of a decomposition matrix: unfolding the high-order tensor to form a two-dimensional matrix by using a high-order singular value decomposition algorithm;
b. typical correlation analysis similarity measurements: measuring similarity between the two tensor differential decomposition matrices of the training sample and the test sample by using typical correlation analysis;
c. similarity fusion: considering each decomposition matrix as an independent unit, similarity fusion is used to combine the similarity of the dual tensors of the training samples and the test samples.
7. The dual tensor-based holographic diffraction label image recognition algorithm as claimed in claim 6, wherein the canonical correlation analysis similarity measure is specifically:
ρ＝maxcorr(u T x，v T y)；
where u and v represent typical variables, ρ represents typical correlation, and corr (x, y) represents correlation of x and y.
8. The dual tensor-based holographic diffraction label image recognition algorithm as claimed in claim 6, wherein the similarity fusion is specifically to find a projection subspace using principal component analysis.
9. The dual tensor-based holographic diffraction label image recognition algorithm as claimed in claim 8, wherein the search shadow casting space is specifically:
will Ψ 1 ，Ψ 2 ，……Ψ t Finding a subspace of a principal component analysis as a similarity vector of training samples, the similarity vector Ψ 1 ，Ψ 2 ，……Ψ t Is defined as:
the scatter matrix is:
the scatter matrix is decomposed into:
S＝ΦΛΦ T ；
wherein,
the subspace of the principal component analysis is composed of eigenvectors corresponding to the eigenvalues of the scatter matrix S;
generating a test similarity vector:
Ψ T ＝{max(Ψ j，1 )，…，max(Ψ j，n )|j＝1，2，…t}；
wherein { max (Ψ) j，n ) I j =1,2, \8230; t } represents the maximum value of the training sample vector at the nth column;
all training sample vectors Ψ 1 ，Ψ 2 ，……Ψ t And Ψ T Projected into the subspace of the principal component analysis.
10. The dual tensor-based holographic diffraction label image recognition algorithm as claimed in claim 1, wherein the original image of the holographic diffraction label acquired in the step S1 is acquired by a camera or a video camera.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910384032.9A CN110097135B (en) | 2019-05-09 | 2019-05-09 | Holographic diffraction label image recognition algorithm based on double tensors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910384032.9A CN110097135B (en) | 2019-05-09 | 2019-05-09 | Holographic diffraction label image recognition algorithm based on double tensors |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110097135A CN110097135A (en) | 2019-08-06 |
| CN110097135B true CN110097135B (en) | 2022-12-09 |
Family
ID=67447398
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910384032.9A Active CN110097135B (en) | 2019-05-09 | 2019-05-09 | Holographic diffraction label image recognition algorithm based on double tensors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110097135B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016091017A1 (en) * | 2014-12-09 | 2016-06-16 | 山东大学 | Extraction method for spectral feature cross-correlation vector in hyperspectral image classification |
| CN106023065A (en) * | 2016-05-13 | 2016-10-12 | 中国矿业大学 | Tensor hyperspectral image spectrum-space dimensionality reduction method based on deep convolutional neural network |
| CN106778837A (en) * | 2016-11-29 | 2017-05-31 | 浙江工业大学 | SAR image target recognition method based on polyteny principal component analysis and tensor analysis |
-
2019
- 2019-05-09 CN CN201910384032.9A patent/CN110097135B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016091017A1 (en) * | 2014-12-09 | 2016-06-16 | 山东大学 | Extraction method for spectral feature cross-correlation vector in hyperspectral image classification |
| CN106023065A (en) * | 2016-05-13 | 2016-10-12 | 中国矿业大学 | Tensor hyperspectral image spectrum-space dimensionality reduction method based on deep convolutional neural network |
| CN106778837A (en) * | 2016-11-29 | 2017-05-31 | 浙江工业大学 | SAR image target recognition method based on polyteny principal component analysis and tensor analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110097135A (en) | 2019-08-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lu et al. | Joint dictionary learning for multispectral change detection | |
| Doretto et al. | Appearance-based person reidentification in camera networks: problem overview and current approaches | |
| US10074006B2 (en) | Methods and systems for disease classification | |
| Seo et al. | Training-free, generic object detection using locally adaptive regression kernels | |
| Elazary et al. | A Bayesian model for efficient visual search and recognition | |
| US8358855B2 (en) | Determining probabilities from compared covariance appearance models to detect objects of interest in images | |
| US9864923B2 (en) | Assisted surveillance of vehicles-of-interest | |
| US9700213B2 (en) | System and method for automatic polyp detection using global geometric constraints and local intensity variation patterns | |
| Yao et al. | Fast human detection from videos using covariance features | |
| US9165184B2 (en) | Identifying matching images | |
| Wang et al. | Multimodality semantic segmentation based on polarization and color images | |
| CN111259756A (en) | Pedestrian re-identification method based on local high-frequency features and mixed metric learning | |
| Abraham et al. | Automatically classifying crime scene images using machine learning methodologies | |
| Irhebhude et al. | A gender recognition system using facial images with high dimensional data | |
| Melek et al. | Datasets and methods of product recognition on grocery shelf images using computer vision and machine learning approaches: An exhaustive literature review | |
| CN110097135B (en) | Holographic diffraction label image recognition algorithm based on double tensors | |
| Akula et al. | Optimized bag of features framework for object recognition in thermal infrared images | |
| Al-Azzawy | Eigenface and SIFT for gender classification | |
| Fritz et al. | Rapid object recognition from discriminative regions of interest | |
| Cho et al. | Multi-classifier-based automatic polyp detection in endoscopic images | |
| Abdullah | Supervised learning algorithms for visual object categorization | |
| Nanni et al. | A comparison of multi-scale approaches for extracting image descriptors from the co-occurrence matrix | |
| Borovikov et al. | Face matching for post-disaster family reunification | |
| Dalka et al. | Selection of visual descriptors for the purpose of multi-camera object re-identification | |
| Soltani-Sarvestani et al. | Batch color classification using bag of colors and discriminative sparse coding |
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 |