CN108537234A - The demarcation method of rule background dictionary-based learning and rectangular target image - Google Patents
The demarcation method of rule background dictionary-based learning and rectangular target image Download PDFInfo
- Publication number
- CN108537234A CN108537234A CN201810238135.XA CN201810238135A CN108537234A CN 108537234 A CN108537234 A CN 108537234A CN 201810238135 A CN201810238135 A CN 201810238135A CN 108537234 A CN108537234 A CN 108537234A
- Authority
- CN
- China
- Prior art keywords
- matrix
- image
- column
- class
- target 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 239000011159 matrix material Substances 0.000 claims abstract description 87
- 239000013598 vector Substances 0.000 claims description 47
- 238000005070 sampling Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 13
- ZAMOUSCENKQFHK-UHFFFAOYSA-N Chlorine atom Chemical compound [Cl] ZAMOUSCENKQFHK-UHFFFAOYSA-N 0.000 claims description 3
- 239000000460 chlorine Substances 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 2
- 238000000926 separation method Methods 0.000 abstract description 4
- 238000001514 detection method Methods 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
- G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of demarcation method of rule background dictionary-based learning and rectangular target image：(1) RGB pictures are converted to gray scale picture to store；(2) it is sampled respectively using the priori fringe region horizontal and vertical to picture；(3) lateral, longitudinal sample is trained respectively using dictionary learning algorithm；(4) geometric position of horizontal boundary, longitudinal boundary is obtained using dictionary learning grader；(5) demarcation of rule background and rectangular target image is completed by horizontal, longitudinal boundary geometric position.The classification and Detection of the present invention combination dictionary learning model and coding specification device, classified to dictionary matrix and sparse matrix using least squares estimate, and delimited for rule background in general picture and the separation of rectangular target image, the intelligence of picture is delimited and image rule background separation has great significance and influences.
Description
Technical Field
The invention relates to a novel delimiting method based on dictionary learning and a code classifier, in particular to a delimiting method based on regular background and rectangular target images of dictionary learning.
Background
With the increasing popularity and wide use of mobile terminals, the photographing function of the mobile terminal is also spreading. With the integration of mobile terminal devices and image sensors, this has made the world's information exchange increasingly convenient and efficient. While the sensor technology of the mobile terminal is continuously developed and improved, the real-time acquisition and transmission of pictures in the information exchange of the mobile terminal are more and more approved and widely put into practical use.
At present, in picture information exchange, a plurality of situations exist in which regular stripes are used as backgrounds, the acquisition of target information needs to be delimited manually due to the existence of the backgrounds, and the regular backgrounds and target images are separated usually by using an SIFT algorithm, i.e. a derivative algorithm thereof, or some machine learning algorithms specific to specific signals.
And taking a picture of the mobile terminal equipment, and taking a rectangular target image with regular stripes as a background, wherein the rotation angle of the rectangular target image is within 10 degrees. For the acquisition of target image information with regular stripes as backgrounds, if a manual delimiting mode is used, the investment of labor cost needs to be greatly increased; the SIFT and its derivative algorithms or some other machine learning algorithms require too long training time or cannot achieve the accuracy required by the application.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a regular background and rectangular target image delimitation method based on dictionary learning, which combines a dictionary learning model and classification detection of a code classifier, utilizes a least square estimation method to classify a dictionary matrix and a sparse matrix, is used for separating and delimitating the regular background and the target image in a common picture, and has important significance and influence on intelligent delimitation of the picture and separation of the picture background.
The invention adopts the following technical scheme for solving the technical problems:
the invention provides a delimiting method of a regular background and a rectangular target image based on dictionary learning, which specifically comprises the following steps:
step 1, converting an RGB image into a gray image and storing the gray image;
step 2, respectively carrying out horizontal and longitudinal sampling on the edge of the gray level image by using a priori to serve as a horizontal sample and a longitudinal sample of a regular background; sampling the center of the gray image transversely and longitudinally to obtain a transverse sample and a longitudinal sample of the target image;
step 3, training a regular background and a target image horizontal sample, and training a regular background and a target image vertical sample respectively by using a dictionary learning algorithm;
step 4, obtaining a regular background and the transverse and longitudinal boundaries of the target image by using a dictionary learning classifier;
and 5, delimitation of the regular background and the target image is completed through the geometric positions of the regular background and the transverse and longitudinal boundaries of the target image.
As a further technical scheme of the invention, the storage form of the gray scale image in the step 1 is a matrix with element ranges of 0-255.
As a further technical solution of the present invention, in step 2, the whole row and whole column sampling is performed on the gray image matrix in a vector form: taking a horizontal sample of a regular background as a starting N line and a tail N line of a gray image matrix, and taking a longitudinal sample as a starting N column and a tail N column of the gray image matrix; taking 2N rows at the center of a gray level image matrix as a horizontal sample of the target image, and taking 2N columns at the center of the gray level image matrix as a vertical sample; wherein N is a set value.
As a further technical scheme of the invention, the objective function of the dictionary learning model in the step 3 is as follows:
where, C ═ 1,2, C ═ 1 indicates that the image belongs to a regular background class, C ═ 2 indicates that the image belongs to a rectangular target image class, k ═ 1,2,a transpose matrix representing a matrix of vertical samples belonging to a regular background class,a transpose matrix representing a matrix of transverse samples belonging to a regular background class,a transpose matrix representing a matrix of vertical samples belonging to a rectangular object image class,a transpose matrix representing a matrix of transverse samples belonging to a rectangular object image class,to representCorresponding sparse representation matrix, D1Dictionary matrices representing correspondences of longitudinal samples, D2and representing dictionary matrixes corresponding to the transverse samples, and beta represents a regular term coefficient.
As a further technical solution of the present invention,2K lines in total, wherein the first K lines correspond to the 1 st type, and the K +1 th to 2K lines correspond to the 2 nd type; dkThe total number of columns is 2K, the first K column corresponds to the 1 st class, and the K +1 st to 2K columns correspond to the 2 nd class.
As a further technical scheme of the invention, a dictionary matrix DkThe training process of (2) is as follows:
first, a matrix D is initialized using a random number methodkAnd
secondly, by the least square method, the matrix D is obtainedkAndthe iterative formula of (a) is:
wherein,represents XkI denotes an identity transformation matrix, diRepresents DkI th column of (d)iLeast squares estimation of To representThe number of the ith row of (a), LCrepresents DkAll columns or X in (1) belonging to class CkA set of rows in class C;denotes all D not belonging to class CkEach column of (1) and XkThe sum of the products of the corresponding rows in (b),means all belonging to class C, except DkColumn i of DkEach column of (1) and XkThe sum of products of corresponding rows in;
finally, iteration is carried out according to the set iteration times by utilizing the iteration formula to obtain the trained dictionary matrix Dk。
As a further technical solution of the present invention, step 4 specifically is: and traversing the row vectors and the column vectors of the gray image matrix by using a dictionary learning classifier to obtain a regular background and the transverse and longitudinal boundaries of the target image.
As a further technical solution of the present invention, a column vector of a gray-scale image matrix is traversed, and it is determined that the column belongs to a regular background class or a rectangular target image class by the following formula:
traversing the transpose of the row vector of the gray level image matrix, and judging whether the row belongs to a regular background class or a rectangular target image class according to the following formula:
wherein |. non chlorinejRepresenting the absolute value of the element of row j of the column direction, BiRepresents the ith column of the grayscale image matrix B, (B)T)iAn ith column representing a transpose of the grayscale image matrix B;
and according to the traversal result, obtaining row generic vectors and column generic vectors of the gray image matrix B according to the geometric sequence, wherein the positions of elements of the row generic vectors and the column generic vectors, which are changed from 1 to 2 and are changed from 2 to 1, in the corresponding vectors are the coordinate positions of the line boundary and the column boundary of the target image.
Compared with the prior art, the invention adopting the technical scheme has the following technical effects: the regular background and rectangular target image delimitation method based on dictionary learning can effectively delimitate the regular background and the rectangular target image in the picture shot by the mobile terminal, greatly improves delimitation efficiency and accuracy, replaces manual interactive supervision with prior supervision, and avoids labor cost investment.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
fig. 2 is a photograph taken by the mobile terminal under No. 1 regular background;
fig. 3 is a photograph taken by the mobile terminal under No. 2 regular background;
FIG. 4 is a drawing of the delimiting effect of the regular background and the rectangular target image based on dictionary learning under No. 1 regular background;
fig. 5 is a delimiting effect diagram of a regular background and a rectangular target image based on dictionary learning under the No. 2 regular background.
Detailed Description
The technical scheme of the invention is further described in detail by combining the drawings and the specific embodiments:
the invention provides a method for priori sampling and dictionary learning aiming at the problem of distinguishing regular backgrounds and rectangular target images in pictures, which comprises the following steps:
(1) the edge area of the default picture is a regular background, the central area of the default picture is a rectangular target image, and the edge area and the central area of the default picture are divided into a horizontal area and a vertical area to be sampled respectively;
(2) respectively carrying out row and column sample training by utilizing a dictionary learning model based on least square optimization to obtain a corresponding dictionary matrix column dictionary D1Line dictionary D2。
According to a trained dictionary D1、D2And respectively traversing rows and columns of the storage matrix of the computer corresponding to the gray-scale image, and determining the class attribution of each row and the class attribution of each column through a dictionary learning coding classifier so as to determine that each row and each column belong to a regular background class or a rectangular target image class.
According to the traversal result, obtaining row generic vectors and column generic vectors of the gray level image matrix according to the geometric sequence, wherein in the returned two vectors, the element can be 1 or 2, wherein 1 represents that the row or column corresponding to the picture belongs to the regular background class, and 2 represents that the row or column corresponding to the picture belongs to the rectangular target image class. The positions of the elements of the two returned vectors, namely the coordinate positions of the line and column boundaries of the target image, which are changed from 1 to 2 and from 2 to 1, in the vectors are the coordinate positions of the line and column boundaries of the target image.
Extraction of dictionary learning samples
For an initial picture obtained by a mobile device, a priori is used for supervision, the edge of the picture belongs to a regular background class and the central area of the picture belongs to a rectangular target image class in the prior supervision. After grayscaling the image, a matrix in the form of a range of elements 0-255 is stored. And respectively carrying out whole-row sampling and whole-column sampling on the regular background class and the rectangular target image class in a vector form. Taking line sampling as an example, a regular background sample takes an initial N-row vector, a final N-row vector, a rectangular target image sample takes a central 2N-row vector, similarly, a column vector regular background sample takes an initial N-column vector, a final N-column vector, and a rectangular target image sample takes a central 2N-column vector, according to the following target formula:
wherein, C ═ 1,2, C ═ 1 indicates that the image belongs to the regular background class, C ═ 2 indicates that the image belongs to the rectangular target image class, k ═ 1,2,a transpose matrix representing a matrix of vertical samples belonging to a regular background class,a transpose matrix representing a matrix of transverse samples belonging to a regular background class,a transpose matrix representing a matrix of vertical samples belonging to a rectangular object image class,a transpose matrix representing a matrix of transverse samples belonging to a rectangular object image class,to representCorresponding sparse representation matrix, D1Dictionary matrices representing correspondences of longitudinal samples, D2and representing dictionary matrixes corresponding to the transverse samples, and beta represents a regular term coefficient.2K lines in total, wherein the first K lines correspond to the 1 st type, and the K +1 th to 2K lines correspond to the 2 nd type; dkThe total number of columns is 2K, the first K column corresponds to the 1 st class, and the K +1 st to 2K columns correspond to the 2 nd class.
Training of dictionaries
Initialization of D Using random number methodkMatrix sumMatrix, obtained by least square methodMatrix, DkThe iterative formula of the matrix is as follows, where DkThe matrix is updated in sequence by columns:
wherein,represents XkIs determined by the least-squares estimation of (c),representation matrix DkI denotes an identity transformation matrix, diRepresents DkThe (c) th column of (a),is diIs determined by the least-squares estimation of (c),to representTranspose of the ith row of (1);represents a correspondence diThe residual matrix of (a) is determined,LCrepresents DkAll columns or X in (1) belonging to class CkA set of rows in class C;denotes all D not belonging to class CkEach column of (1) and XkThe sum of the products of the corresponding rows in (b),means all belonging to class C, except DkColumn i of DkEach column of (1) and XkThe sum of the products of the corresponding rows in (a).
Performing 15-20 iterations by using the formula to obtain the final training completed DkAnd (4) matrix.
Classifier delimitation
According to the row and column samples of the gray level image matrix B, the row samples are transposed and applied to a learning model, and two corresponding dictionaries can be obtained through dictionary training. After obtaining the corresponding dictionary, traversing the transpose of the row vector and the column vector of the matrix B, taking the column vector as an example, judging whether the column of the matrix B belongs to a regular background class or a rectangular target image class by the following formula:
wherein |. non chlorinejRepresenting the absolute value of the element of the jth column of the column vector. B isiRepresents the ith column of the B matrix,is a column vector, and the number of rows and XkAre the same number of columns, so that the analogy of XkThe first K rows belong to the first class, the K +1 to 2K rows belong to the second class,meaning that the absolute values of the elements of all rows of the same class are summed separately.
Similarly, for a row vector, the corresponding ith row of B is transposed, using the following classification formula:
and respectively returning each row and each column of B to belong to a regular background or a rectangular target image by using the traversal result. According to the traversal result, obtaining row generic vectors and column generic vectors of the gray level image matrix according to the geometric sequence, wherein in the returned two vectors, the element can be 1 or 2, wherein 1 represents that the row or column corresponding to the picture belongs to the regular background class, and 2 represents that the row or column corresponding to the picture belongs to the rectangular target image class. The positions of the elements of the two returned vectors, namely the coordinate positions of the line and column boundaries of the target image, which are changed from 1 to 2 and from 2 to 1, in the vectors are the coordinate positions of the line and column boundaries of the target image.
Examples
Two stripe backgrounds are selected, one is a regular background parallel to a target image, the other is a diagonal stripe regular background, meanwhile, the method is suitable for a rectangular target image, and an invoice is selected as the target image. The experimental method includes the steps that rectangular invoices with different lengths are placed on different regular backgrounds, the mobile phone terminal is used for shooting the invoices, the invoices are uploaded to a computer to serve as illustration pictures (such as pictures shot by the mobile terminal under the No. 1 regular background and the No. 2 regular background shown in the figures 2 and 3), the invention method is used for delimiting the pictures, and the delimiting results are shown in the figures 4 and 5. From the delimited display results, it is demonstrated that our inventive method can fully achieve very high accuracy and practical standards.
The invention is realized on an MATLAB experimental platform, as shown in figure 1, and mainly comprises the following steps:
step 1: and shooting a certain rectangular target image on a regular background by utilizing the mobile terminal equipment. Executing the step 2;
step 2: and importing the pictures into MATLAB and processing the pictures into a double data type matrix. Executing the step 3;
and 3, step 3: sampling by using prior supervision, wherein the edge is a regular background sample, the center is a rectangular target image sample, and the step 4 is executed;
and 4, step 4: training the sample according to the formulas (2), (3) and (4), iterating for about 15-20 times, and executing the step 5;
and 5, step 5: and (5) traversing the picture matrix in longitudinal and transverse dimensions by using a formula (5), determining the class to which the vector belongs, returning the vectors of the classes to which the rows and the columns belong, and executing the step 6.
And 6, step 6: and eliminating error data of the returned vector by using difference to obtain the geometric position information of the row and column boundaries of the regular background and the rectangular target image.
The invention discloses a regular background and rectangular target image delimitation method based on dictionary learning. The method comprises the following steps: (1) converting the RGB picture into a gray picture for storage; (2) respectively sampling the horizontal and longitudinal edge regions of the picture by using prior as regular background samples and respectively sampling the horizontal and longitudinal central regions of the picture as rectangular target image samples; (3) training a regular background sample and a horizontal sample and a longitudinal sample of a rectangular target image respectively by using a dictionary learning algorithm; (4) using a dictionary learning classifier to obtain the geometric positions of the transverse boundary and the longitudinal boundary of the regular background and the rectangular target image; (5) and delimitation of the regular background and the rectangular target image is completed through the geometric positions of the transverse boundary and the longitudinal boundary. The invention combines the dictionary learning model and the classification detection of the encoding classifier, utilizes the least square estimation method to classify the dictionary matrix and the sparse matrix, is used for separating and delimiting regular background and rectangular target image in general picture, and has important significance and influence on the intelligent delimitation of picture and the regular background separation of picture.
The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can understand that the modifications or substitutions within the technical scope of the present invention are included in the scope of the present invention, and therefore, the scope of the present invention should be subject to the protection scope of the claims.
Claims (8)
1. The regular background and rectangular target image delimiting method based on dictionary learning is characterized by specifically comprising the following steps of:
step 1, converting an RGB image into a gray image and storing the gray image;
step 2, respectively carrying out horizontal and longitudinal sampling on the edge of the gray level image by using a priori to serve as a horizontal sample and a longitudinal sample of a regular background; sampling the center of the gray image transversely and longitudinally to obtain a transverse sample and a longitudinal sample of the target image;
step 3, training a regular background and a target image horizontal sample, and training a regular background and a target image vertical sample respectively by using a dictionary learning algorithm;
step 4, obtaining a regular background and the transverse and longitudinal boundaries of the target image by using a dictionary learning classifier;
and 5, delimitation of the regular background and the target image is completed through the geometric positions of the regular background and the transverse and longitudinal boundaries of the target image.
2. The method for delimiting regular background and rectangular target images based on dictionary learning according to claim 1, wherein the storage form of the gray level image in step 1 is a matrix with element range 0-255.
3. The method for delimiting regular background and rectangular target image based on dictionary learning as claimed in claim 2, wherein in step 2, the whole row and column sampling is performed on the gray image matrix in vector form respectively: taking a horizontal sample of a regular background as a starting N line and a tail N line of a gray image matrix, and taking a longitudinal sample as a starting N column and a tail N column of the gray image matrix; taking 2N rows at the center of a gray level image matrix as a horizontal sample of the target image, and taking 2N columns at the center of the gray level image matrix as a vertical sample; wherein N is a set value.
4. The method for delimiting regular backgrounds and rectangular target images based on dictionary learning according to claim 3, wherein the target function of the dictionary learning model in step 3 is as follows:
where, C ═ 1,2, C ═ 1 indicates that the image belongs to a regular background class, C ═ 2 indicates that the image belongs to a rectangular target image class, k ═ 1,2,representing longitudinal sample moments belonging to regular background classesThe transpose of the matrix is the matrix,a transpose matrix representing a matrix of transverse samples belonging to a regular background class,a transpose matrix representing a matrix of vertical samples belonging to a rectangular object image class,a transpose matrix representing a matrix of transverse samples belonging to a rectangular object image class,to representCorresponding sparse representation matrix, D1Dictionary matrices representing correspondences of longitudinal samples, D2and representing dictionary matrixes corresponding to the transverse samples, and beta represents a regular term coefficient.
5. The method for delimiting regular background and rectangular target images based on dictionary learning according to claim 4,2K lines in total, wherein the first K lines correspond to the 1 st type, and the K +1 th to 2K lines correspond to the 2 nd type; dkThe total number of columns is 2K, the first K column corresponds to the 1 st class, and the K +1 st to 2K columns correspond to the 2 nd class.
6. The method for delimiting regular background and rectangular target image based on dictionary learning as claimed in claim 4, wherein the dictionary matrix DkThe training process of (2) is as follows:
first, a matrix D is initialized using a random number methodkAnd
secondly, by the least square method, the matrix D is obtainedkAndthe iterative formula of (a) is:
wherein,represents XkI denotes an identity transformation matrix, diRepresents DkI th column of (d)iLeast squares estimation of To representThe number of the ith row of (a), LCrepresents DkAll columns or X in (1) belonging to class CkA set of rows in class C;denotes all D not belonging to class CkEach column of (1) and XkThe sum of the products of the corresponding rows in (b),means all belonging to class C, except DkColumn i of DkEach column of (1) and XkThe sum of products of corresponding rows in;
finally, iteration is carried out according to the set iteration times by utilizing the iteration formula to obtain the trained dictionary matrix Dk。
7. The method for delimiting regular backgrounds and rectangular target images based on dictionary learning according to claim 6, wherein step 4 specifically comprises: and respectively traversing the transpose of the row vector and the column vector of the gray image matrix by using a dictionary learning classifier to obtain a regular background and the transverse and longitudinal boundaries of the target image.
8. The method for delimiting regular backgrounds and rectangular target images based on dictionary learning according to claim 7, characterized in that, traversing the column vector of the gray image matrix, and determining that the column belongs to the regular background class or the rectangular target image class according to the following formula:
traversing the transpose of the row vector of the gray level image matrix, and judging whether the row belongs to a regular background class or a rectangular target image class according to the following formula:
wherein |. non chlorinejRepresenting the absolute value of the element of row j of the column direction, BiRepresents the ith column of the grayscale image matrix B, (B)T)iAn ith column representing a transpose of the grayscale image matrix B;
and according to the traversal result, obtaining row generic vectors and column generic vectors of the gray image matrix B according to the geometric sequence, wherein the positions of elements of the row generic vectors and the column generic vectors, which are changed from 1 to 2 and are changed from 2 to 1, in the corresponding vectors are the coordinate positions of the line boundary and the column boundary of the target image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810238135.XA CN108537234A (en) | 2018-03-22 | 2018-03-22 | The demarcation method of rule background dictionary-based learning and rectangular target image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810238135.XA CN108537234A (en) | 2018-03-22 | 2018-03-22 | The demarcation method of rule background dictionary-based learning and rectangular target image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108537234A true CN108537234A (en) | 2018-09-14 |
Family
ID=63484096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810238135.XA Pending CN108537234A (en) | 2018-03-22 | 2018-03-22 | The demarcation method of rule background dictionary-based learning and rectangular target image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108537234A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101283364A (en) * | 2005-09-16 | 2008-10-08 | 索尼电子有限公司 | Extracting a moving object boundary |
| CN101582159A (en) * | 2009-06-26 | 2009-11-18 | 哈尔滨工业大学 | Infrared image background suppression method based on unsupervised kernel regression analysis |
| CN107832786A (en) * | 2017-10-31 | 2018-03-23 | 济南大学 | A kind of recognition of face sorting technique based on dictionary learning |
-
2018
- 2018-03-22 CN CN201810238135.XA patent/CN108537234A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101283364A (en) * | 2005-09-16 | 2008-10-08 | 索尼电子有限公司 | Extracting a moving object boundary |
| CN101582159A (en) * | 2009-06-26 | 2009-11-18 | 哈尔滨工业大学 | Infrared image background suppression method based on unsupervised kernel regression analysis |
| CN107832786A (en) * | 2017-10-31 | 2018-03-23 | 济南大学 | A kind of recognition of face sorting technique based on dictionary learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104008538B (en) | Based on single image super-resolution method | |
| CN108108764B (en) | Visual SLAM loop detection method based on random forest | |
| CN110503680B (en) | Unsupervised convolutional neural network-based monocular scene depth estimation method | |
| CN108564025A (en) | A kind of infrared image object identification method based on deformable convolutional neural networks | |
| CN108510504B (en) | Image segmentation method and device | |
| CN111160229B (en) | SSD network-based video target detection method and device | |
| CN112967341B (en) | Indoor visual positioning method, system, equipment and storage medium based on live-action image | |
| CN111369487A (en) | Hyperspectral and multispectral image fusion method, system and medium | |
| CN109376641B (en) | Moving vehicle detection method based on unmanned aerial vehicle aerial video | |
| CN111709980A (en) | Multi-scale image registration method and device based on deep learning | |
| CN107680043B (en) | Single image super-resolution output method based on graph model | |
| CN111831844A (en) | Image retrieval method, image retrieval device, image retrieval apparatus, and medium | |
| CN105335952B (en) | Matching power flow computational methods and device and parallax value calculating method and equipment | |
| CN107944459A (en) | A kind of RGB D object identification methods | |
| CN107862680B (en) | Target tracking optimization method based on correlation filter | |
| CN113011253B (en) | Facial expression recognition method, device, equipment and storage medium based on ResNeXt network | |
| CN111444923A (en) | Image semantic segmentation method and device under natural scene | |
| CN108182694A (en) | A kind of estimation based on interpolation and adaptive video method for reconstructing | |
| CN107705295B (en) | Image difference detection method based on robust principal component analysis method | |
| CN113436220B (en) | Image background estimation method based on depth map segmentation | |
| KR101864000B1 (en) | Multi-purpose image processing core | |
| CN114663880A (en) | Three-dimensional target detection method based on multi-level cross-modal self-attention mechanism | |
| CN110992441A (en) | Writing track processing method and device | |
| CN108764233B (en) | Scene character recognition method based on continuous convolution activation | |
| CN112396167A (en) | Loop detection method for fusing appearance similarity and spatial position information |
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 | ||
| AD01 | Patent right deemed abandoned | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20221206 |