CN107909073A - Multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning - Google Patents
Multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning Download PDFInfo
- Publication number
- CN107909073A CN107909073A CN201710971988.XA CN201710971988A CN107909073A CN 107909073 A CN107909073 A CN 107909073A CN 201710971988 A CN201710971988 A CN 201710971988A CN 107909073 A CN107909073 A CN 107909073A
- Authority
- CN
- China
- Prior art keywords
- lbp
- spectral line
- pixel
- mrow
- music score
- 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
- 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
- G06V10/273—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 removing elements interfering with the pattern to be recognised
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
-
- 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
- G06V10/467—Encoded features or binary features, e.g. local binary patterns [LBP]
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)
- Auxiliary Devices For Music (AREA)
Abstract
The present invention relates to hand-written music score to identify field, spectral line this very noisy is removed to be embodied as note identification module, so as to improve the accuracy rate of note identification.For this reason, the technical solution adopted by the present invention is, multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning, step are as follows:1) feature extraction:Spectral line is deleted and regards a kind of classification task as, i.e., foreground pixel is subjected to two classification:Spectral line and note, a pixel, which belongs to spectral line or note, to be judged by the pixel of its neighborhood, and feature extraction is carried out using local binary patterns LBP operators;2) xgboost models are built;3) tuning is tested:Using the model preserved, tested on test set, analyzed by Comparative result, xgboost parameters are adjusted, the model before optimizing.Present invention is mainly applied to manufacture and design occasion.
Description
Technical field
The present invention relates to hand-written music score to identify field, and the spectral line of music score is completed by technologies such as image procossing and machine learning
Deletion task.Concretely relate to multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning.
Background technology
Music score is by the sound property of music, as pitch, interval, beat etc. show by using visual mark
Music recording means.The presence of music score allows music worldwide to carry out propagation exchange, while is also music interest
" textbook " of person's study.Before printing music score is widely used, substantial amounts of musical works is all in the form of hand-written music score
Carry out in store.However, hand-written music score is highly susceptible to damage, and there is the risk lost.With the popularization of computer, letter
The exchange speed of breath has the lifting of leap.At this moment, the propagation of hand-written music score just becomes very slow.Therefore, it is necessary to will be hand-written
Music score is converted to digital information, and storage is in the database.Due to there is the hand-written music score of magnanimity, being accomplished manually this work
Become abnormal difficult.So need a kind of automatic music score identifying system --- hand-written music score is converted to what computer " can be understood "
Digital information, therefore generate Optical Music Recognition (Optical Music Recognition, OMR) system.
It is nearly all to grind since the presence of spectral line causes very big obstacle to the note identification module behind OMR systems
The personnel for studying carefully OMR systems are proposed a kind of spectral line deletion algorithm.The music score spectral line of canonical form shows as one group (generally five
Bar) horizontal filament, but for hand-written music score, spectral line becomes complicated and changeable, for example, flexural deformation, line width be not fixed and
Interruption missing etc..The task that so spectral line is deleted becomes to be not easy to.Therefore, it is badly in need of a kind of very strong spectral line of robustness and deletes calculation
Method, can be suitable for polytype hand-written music score.
Local binary patterns (Local Binary Pattern, LBP) are a kind of operators for describing Local textural feature.It is logical
The pixel value size of comparison object pixel and its surrounding pixel point is crossed, comparative result is converted to binary system is indicated, and makees
For the texture eigenvalue of target pixel points.On the basis of original LBP, it can be improved.The size of LBP windows can not
Together, and the shape of window can also have a variety of changes, such as circular, square, oval and straight line etc..It can use different
LBP windows extract the LBP values of target point, these LBP values are formed to the feature vector of a multidimensional, for representing target point
Feature.
Machine learning is a branch of artificial intelligence field.Target is to find a kind of model of suitable data, passes through instruction
Practice collection and carry out training pattern parameter, model parameter is preserved, is predicted in unknown data.Main task has classification
And recurrence.The algorithm of classification contains many kinds again.Xgboost is a kind of 2014 sorting algorithms occurred, and in 2016
Formally delivered by tianqiChen et al..Xgboost algorithms are one kind of boosting algorithms, in data contest in recent years
Yield unusually brilliant results.By the way that parameter in xgboost algorithms is adjusted, can be very good to adapt to the data of oneself, adjustable parameter bag
Include loss function, learning rate, punishment term coefficient etc..
The content of the invention
For overcome the deficiencies in the prior art, it is contemplated that being embodied as note identification module removes spectral line this very noisy,
So as to improve the accuracy rate of note identification.For this reason, the technical solution adopted by the present invention is, multidimensional local binary patterns and engineering
Hand-written music score spectral line delet method is practised, step is as follows:
1) feature extraction:Spectral line is deleted and regards a kind of classification task as, i.e., foreground pixel is subjected to two classification:Spectrum
Line and note, a pixel, which belongs to spectral line or note, to be judged by the pixel of its neighborhood, using local binary mould
Formula LBP operators carry out feature extraction, and the sliding window size of LBP operators uses different sizes, and the sliding window of each size includes
These LBP values are formed a multidimensional characteristic vectors as input by the LBP operators of various shapes;
2) xgboost models are built:Using the multidimensional LBP feature vectors of extraction as input, deleted using known in data set
Except rear image is as true value label, xgboost models are trained simultaneously to be preserved;
3) tuning is tested:Using the model preserved, tested on test set, analyzed by Comparative result, it is right
Xgboost parameters are adjusted, the model before optimizing.
Using 3*3 sliding window to central pixel point carry out feature extraction, central pixel point with surrounding pixel successively into
The size of row pixel value compares, if central point pixel value is less than the point compared with it, binary coding takes 1, otherwise takes 0.
Then this Binary Conversion is just obtained to the LBP values of central point, calculation formula for the decimal system:
Wherein (xc,yc) represent central point, gcRepresent the pixel value of central point, giRepresent the pixel value of 8 points of surrounding.
Using improved LBP operators:The size of window uses 7*7,9*9,11*11 and 13*13 totally 4 kinds of size windows, and
And contain the LBP sliding windows of 16 kinds of shapes again for each window, and design LBP sliding windows are linear pattern, since straight up,
Often rotating 22.5 degree takes straight line so to rotate a circle and the LBP sliding windows in 16 directions are obtained as LBP sliding windows, and final one
4*16 LBP value has been obtained, these LBP values are combined into the feature vector of 1 higher-dimension, the input as grader.
7*7 first LBP feature of window, i.e., LBP features straight up are calculated first, it is assumed that central point position (x, y), that
Pixel straight up is followed successively by (x-1, y), (x-2, y), (x-3, y), the pixel value of central point successively with above 3
Pixel point value is compared, and then obtains the binary number of one 3 according to standard LBP operators recited above, then basis
LBP is worth calculation formula, this binary number is converted to decimal number, has obtained the LBP features of central point straight up;Connect
Get off to continue the LBP features for calculating other directions below, that is, at interval of 22.5 degree of calculating, one LBP feature, so rotate
It will obtain within one week the LBP characteristic values in 16 directions;The computational methods 7*7 of 9*9,11*11,13*13 window is the same, so each
Window can all obtain 16 LBP features;And then by the LBP combinations of features of 4 windows together, finally obtain the features of 64 dimensions to
Amount.
Build xgboost models:The target selection two of model is classified, the L2 regular terms parameters of Controlling model complexity
Lambda=4, controls whether the parameter gamma=0.1 of rear beta pruning, builds the depth parameter max depth=6 of decision tree, control
The percentage subsample=0.7 of stochastical sampling training sample processed, generates row sampling parameter colsample during decision tree
Bytree=0.7, while xgboost supports multithreading operation, and maximum thread may be selected according to allocation of computer.
The features of the present invention and beneficial effect are:
The spectral line that the present invention carries out hand-written music score by improved LBP operators and xgboost algorithms is deleted, classical with some
Method compare, advantage is mainly reflected in:
Novelty:Multiple dimensioned multidimensional LBP features and xgboost algorithms are combined first and carry out hand-written music score spectral line deletion work
Make, the window selection of LBP is carried out according to the characteristics of hand-written music score.
Robustness:Various hand-written music score complicated and changeable can be applicable in using the algorithm of the present invention, some biographies can be overcome
Algorithm of uniting shows the shortcomings that very poor to certain non-ideal music score, has very strong stability.
Brief description of the drawings:
Mono- hand-written musical score image of Fig. 1.
Fig. 2 removes the true value image of spectral line.
The original LBP operators figures of Fig. 3.
The multiple dimensioned LBP operators figure of the improved multidimensional of Fig. 4.
Flow charts of the Fig. 5 based on improved LBP and xgboost algorithms.
Fig. 6 carries out the result figure of spectral line deletion using the present invention.
Embodiment
The main object of the present invention is to apply the spectral line in hand-written music score to delete image processing techniques and machine learning techniques
Except in task, as the preposition processing module of OMR systems, spectral line this very noisy is eliminated for note identification module below,
So as to improve the accuracy rate of note identification.
The method that the present invention uses supervised learning, trains disaggregated model by training set first, then carries out model
Preserve, apply on unknown music score.Grader uses xgboost algorithms, and feature is using multiple dimensioned multidimensional LBP as target picture
The feature vector of vegetarian refreshments.In the test of hand-written musical score image, it can be very good to complete spectral line deletion work.
To achieve these goals, the present invention adopts the following technical scheme that:
1) feature extraction.Spectral line is deleted and regards a kind of classification task as by the present invention, i.e., foreground pixel is carried out two
Classification:Spectral line and note.One pixel belongs to spectral line or note can be judged that LBP is exactly by the pixel of its neighborhood
The operator of Local textural feature is described.Therefore, feature extraction is carried out using LBP operators, the sliding window size of LBP operators is using different
Size, and the sliding window of each size includes the LBP operators of various shapes, by these LBP values form a multidimensional characteristic to
Amount is as input.
2) xgboost models are built.Using the multidimensional LBP feature vectors of extraction as input, deleted using known in data set
Except rear image is as true value label, xgboost models are trained simultaneously to be preserved.
3) tuning is tested.Using the model preserved, tested on test set.Analyzed by Comparative result, it is right
Xgboost parameters are adjusted, the model before optimizing.
The invention will be further described with example below in conjunction with the accompanying drawings.
1) feature extraction.
Original LBP operators are as shown in Figure 3.Feature extraction, center pixel are carried out to central pixel point using the sliding window of 3*3
The size that point carries out pixel value with the pixel of surrounding successively compares, if central point pixel value is less than the point compared with it,
Binary coding takes 1, otherwise takes 0.Then this Binary Conversion is just obtained to the LBP values of central point for the decimal system.Calculate public
Formula:
Wherein (xc,yc) represent central point, gcRepresent the pixel value of central point, giRepresent the pixel value of 8 points of surrounding.
The present invention uses improved LBP operators, and the size of window employs 7*7,9*9,11*11 and 13*13 totally 4 kinds of sizes
Window, and contain the LBP sliding windows of 16 kinds of shapes again for each window.For music score, note and spectral line have
Very strong directionality, can utilize this priori, design LBP sliding windows are linear pattern, since straight up, are often rotated
22.5 degree take straight line so to rotate a circle and the LBP sliding windows in 16 directions are obtained as LBP sliding windows.Final one is obtained
These LBP values, are combined into the feature vector of 1 higher-dimension, the input as grader by 4*16 LBP value.
It will be made below illustrating:
7*7 first LBP feature of window, i.e., LBP features straight up are calculated first, it is assumed that central point position (x, y), that
Pixel straight up is followed successively by (x-1, y), (x-2, y), (x-3, y), the pixel value of central point successively with above 3
Pixel point value is compared, and then obtains the binary number of one 3 according to standard LBP operators recited above, then basis
LBP is worth calculation formula, this binary number is converted to decimal number, has obtained the LBP features of central point straight up.Connect
Get off to continue the LBP features for calculating other directions below, that is, at interval of 22.5 degree of calculating, one LBP feature, so rotate
It will obtain within one week the LBP characteristic values in 16 directions.The computational methods 7*7 of 9*9,11*11,13*13 window is the same, so each
Window can all obtain 16 LBP features.And then by the LBP combinations of features of 4 windows together, finally obtain the features of 64 dimensions to
Amount.
2) xgboost models are built.
The task of the present invention is the spectral line deletion of hand-written music score, the prospect color dot of music score should be divided into two class of spectral line and note,
So the target selection two of model is classified.The L2 regular terms parameter lambda=4 of Controlling model complexity, control whether rear beta pruning
Parameter gamma=0.1, build the depth parameter max depth=6 of decision tree, control the percentage of stochastical sampling training sample
Than subsample=0.7, row sampling parameter colsample bytree=0.7 during decision tree, while xgboost branch are generated
Multithreading operation is held, maximum thread may be selected according to allocation of computer.
3) tuning is tested.
Fig. 6 is the result figure that spectral line deletion is carried out using inventive algorithm, it is seen that the validity of algorithm.And in tree evidence greatly
It is tested on collection, includes polytype hand-written music score, F-measure is evaluated more than 96%, illustrates the present invention
Algorithm has very strong robustness.Meanwhile by test error, the model parameter of xgboost can be further adjusted, carries out model
Optimization.
Claims (5)
1. a kind of multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning, it is characterized in that, step is as follows:
1) feature extraction:Spectral line is deleted and regards a kind of classification task as, i.e., foreground pixel is subjected to two classification:Spectral line and
Note, a pixel, which belongs to spectral line or note, to be judged by the pixel of its neighborhood, using local binary patterns LBP
Operator carries out feature extraction, and the sliding window size of LBP operators uses different sizes, and the sliding window of each size includes a variety of shapes
These LBP values are formed a multidimensional characteristic vectors as input by the LBP operators of shape;
2) xgboost models are built:Using the multidimensional LBP feature vectors of extraction as input, after known deletion in data set
Image is simultaneously preserved as true value label, training xgboost models;
3) tuning is tested:Using the model preserved, tested on test set, analyzed by Comparative result, to xgboost
Parameter is adjusted, the model before optimizing.
2. multidimensional local binary patterns as claimed in claim 1 and the hand-written music score spectral line delet method of machine learning, its feature
It is feature extraction to be carried out to central pixel point using the sliding window of 3*3, central pixel point carries out pixel successively with the pixel of surrounding
The size of value compares, if central point pixel value is less than the point compared with it, binary coding takes 1, otherwise takes 0.Then will
This Binary Conversion just obtains the LBP values of central point, calculation formula for the decimal system:
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>c</mi>
</msub>
<mo>,</mo>
<msub>
<mi>y</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mn>7</mn>
</munderover>
<mi>s</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>g</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>g</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<msup>
<mn>2</mn>
<mi>i</mi>
</msup>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>s</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
<mo>,</mo>
<mi>x</mi>
<mo>></mo>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
<mo>,</mo>
<mi>x</mi>
<mo>&le;</mo>
<mn>0</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein (xc,yc) represent central point, gcRepresent the pixel value of central point, giRepresent the pixel value of 8 points of surrounding.
3. multidimensional local binary patterns as claimed in claim 1 and the hand-written music score spectral line delet method of machine learning, its feature
It is, using improved LBP operators:The size of window using 7*7,9*9,11*11 and 13*13 totally 4 kinds of size windows, and for
Each window contains the LBP sliding windows of 16 kinds of shapes again, and design LBP sliding windows are linear pattern, since straight up, are often rotated
22.5 degree take straight line so to rotate a circle and the LBP sliding windows in 16 directions are obtained, final one is obtained as LBP sliding windows
These LBP values, are combined into the feature vector of 1 higher-dimension, the input as grader by 4*16 LBP value.
4. multidimensional local binary patterns as claimed in claim 3 and the hand-written music score spectral line delet method of machine learning, its feature
It is to calculate 7*7 first LBP feature of window, i.e., LBP features straight up, it is assumed that central point position (x, y) first, then perpendicular
Straight upward pixel is followed successively by (x-1, y), (x-2, y), (x-3, y), the pixel value of central point successively with 3 pixels above
Point value is compared, and then the binary number of one 3 is obtained according to standard LBP operators recited above, then according to LBP values
Calculation formula is obtained, this binary number is converted into decimal number, has obtained the LBP features of central point straight up;Next
Continue to calculate the LBP features in other directions below, that is, a LBP feature is calculated at interval of 22.5 degree, so rotate a circle
It will obtain the LBP characteristic values in 16 directions;The computational methods 7*7 of 9*9,11*11,13*13 window is the same, so each window
It will obtain 16 LBP features;And then by the LBP combinations of features of 4 windows together, finally obtain the feature vector of 64 dimensions.
5. multidimensional local binary patterns as claimed in claim 3 and the hand-written music score spectral line delet method of machine learning, its feature
It is to build xgboost models:The target selection two of model is classified, the L2 regular terms parameters lambda=of Controlling model complexity
4, control whether the parameter gamma=0.1 of rear beta pruning, build the depth parameter max depth=6 of decision tree, control is adopted at random
The percentage subsample=0.7 of sample training sample, generates row sampling parameter colsample bytree=during decision tree
0.7, while xgboost supports multithreading operation, and maximum thread may be selected according to allocation of computer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710971988.XA CN107909073A (en) | 2017-10-18 | 2017-10-18 | Multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710971988.XA CN107909073A (en) | 2017-10-18 | 2017-10-18 | Multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107909073A true CN107909073A (en) | 2018-04-13 |
Family
ID=61840638
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710971988.XA Pending CN107909073A (en) | 2017-10-18 | 2017-10-18 | Multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107909073A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109711409A (en) * | 2018-11-15 | 2019-05-03 | 天津大学 | A kind of hand-written music score spectral line delet method of combination U-net and ResNet |
CN110569716A (en) * | 2019-07-26 | 2019-12-13 | 浙江工业大学 | Goods shelf image copying detection method |
CN110705608A (en) * | 2019-09-12 | 2020-01-17 | 杭州惠合信息科技有限公司 | Retail terminal display shelf reproduction identification method and device |
CN110852178A (en) * | 2019-10-17 | 2020-02-28 | 天津大学 | Piano music score difficulty identification method based on decision tree lifting |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101964049A (en) * | 2010-09-07 | 2011-02-02 | 东南大学 | Spectral line detection and deletion method based on subsection projection and music symbol structure |
CN102663423A (en) * | 2012-03-28 | 2012-09-12 | 北京航空航天大学 | Method for automatic recognition and playing of numbered musical notation image |
US20120250941A1 (en) * | 2011-03-31 | 2012-10-04 | Masanori Katsuta | Sound reproduction program and sound reproduction device |
CN106250936A (en) * | 2016-08-16 | 2016-12-21 | 广州麦仑信息科技有限公司 | Multiple features multithreading safety check contraband automatic identifying method based on machine learning |
CN106525027A (en) * | 2016-11-02 | 2017-03-22 | 上海航天控制技术研究所 | Star sensor star point extracting method based on local binary pattern |
CN106548168A (en) * | 2016-10-25 | 2017-03-29 | 天津大学 | A kind of detection of staff spectral line and delet method based on low-rank structure |
CN106570508A (en) * | 2016-11-05 | 2017-04-19 | 天津大学 | Music score line detecting and deleting method based on local binary pattern |
CN106874912A (en) * | 2016-12-20 | 2017-06-20 | 银江股份有限公司 | A kind of image object detection method based on improvement LBP operators |
-
2017
- 2017-10-18 CN CN201710971988.XA patent/CN107909073A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101964049A (en) * | 2010-09-07 | 2011-02-02 | 东南大学 | Spectral line detection and deletion method based on subsection projection and music symbol structure |
US20120250941A1 (en) * | 2011-03-31 | 2012-10-04 | Masanori Katsuta | Sound reproduction program and sound reproduction device |
CN102663423A (en) * | 2012-03-28 | 2012-09-12 | 北京航空航天大学 | Method for automatic recognition and playing of numbered musical notation image |
CN106250936A (en) * | 2016-08-16 | 2016-12-21 | 广州麦仑信息科技有限公司 | Multiple features multithreading safety check contraband automatic identifying method based on machine learning |
CN106548168A (en) * | 2016-10-25 | 2017-03-29 | 天津大学 | A kind of detection of staff spectral line and delet method based on low-rank structure |
CN106525027A (en) * | 2016-11-02 | 2017-03-22 | 上海航天控制技术研究所 | Star sensor star point extracting method based on local binary pattern |
CN106570508A (en) * | 2016-11-05 | 2017-04-19 | 天津大学 | Music score line detecting and deleting method based on local binary pattern |
CN106874912A (en) * | 2016-12-20 | 2017-06-20 | 银江股份有限公司 | A kind of image object detection method based on improvement LBP operators |
Non-Patent Citations (3)
Title |
---|
CHEN GENFANG 等: "Detecting the Staff-lines of Musical Score with Hough Transform and Mathematical Morphology", 《INTERNATIONAL CONFERENCE ON MULTIMEDIA TECHNOLOGY》 * |
刘晓翔 等: "一种快速稳健的乐谱图像谱线检测与删除方法", 《计算机工程与应用》 * |
孟凡奥 等: "基于局部二进制模式的乐谱谱线检测与删除", 《计算机科学与探索》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109711409A (en) * | 2018-11-15 | 2019-05-03 | 天津大学 | A kind of hand-written music score spectral line delet method of combination U-net and ResNet |
CN110569716A (en) * | 2019-07-26 | 2019-12-13 | 浙江工业大学 | Goods shelf image copying detection method |
CN110705608A (en) * | 2019-09-12 | 2020-01-17 | 杭州惠合信息科技有限公司 | Retail terminal display shelf reproduction identification method and device |
CN110852178A (en) * | 2019-10-17 | 2020-02-28 | 天津大学 | Piano music score difficulty identification method based on decision tree lifting |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111126386B (en) | Sequence domain adaptation method based on countermeasure learning in scene text recognition | |
CN108229550B (en) | Cloud picture classification method based on multi-granularity cascade forest network | |
CN107194418B (en) | Rice aphid detection method based on antagonistic characteristic learning | |
CN107909073A (en) | Multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning | |
CN107103317A (en) | Fuzzy license plate image recognition algorithm based on image co-registration and blind deconvolution | |
CN103093240A (en) | Calligraphy character identifying method | |
CN107341523A (en) | Express delivery list information identifying method and system based on deep learning | |
CN110827260B (en) | Cloth defect classification method based on LBP characteristics and convolutional neural network | |
CN109657612B (en) | Quality sorting system based on facial image features and application method thereof | |
CN108154158B (en) | Building image segmentation method for augmented reality application | |
CN106340016A (en) | DNA quantitative analysis method based on cell microscope image | |
CN107945200A (en) | Image binaryzation dividing method | |
CN104598885A (en) | Method for detecting and locating text sign in street view image | |
CN111178177A (en) | Cucumber disease identification method based on convolutional neural network | |
CN107784321A (en) | Numeral paints this method for quickly identifying, system and computer-readable recording medium | |
CN105930792A (en) | Human action classification method based on video local feature dictionary | |
CN105117707A (en) | Regional image-based facial expression recognition method | |
CN110599463A (en) | Tongue image detection and positioning algorithm based on lightweight cascade neural network | |
CN106228136A (en) | Panorama streetscape method for secret protection based on converging channels feature | |
CN111539330A (en) | Transformer substation digital display instrument identification method based on double-SVM multi-classifier | |
CN114140665A (en) | Dense small target detection method based on improved YOLOv5 | |
CN112488128A (en) | Bezier curve-based detection method for any distorted image line segment | |
CN110598581B (en) | Optical music score recognition method based on convolutional neural network | |
CN113657414B (en) | Object identification method | |
CN109325487B (en) | Full-category license plate recognition method based on target detection |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180413 |
|
WD01 | Invention patent application deemed withdrawn after publication |