CN105718932A - Colorful image classification method based on fruit fly optimization algorithm and smooth twinborn support vector machine and system thereof - Google Patents
Colorful image classification method based on fruit fly optimization algorithm and smooth twinborn support vector machine and system thereof Download PDFInfo
- Publication number
- CN105718932A CN105718932A CN201610038347.4A CN201610038347A CN105718932A CN 105718932 A CN105718932 A CN 105718932A CN 201610038347 A CN201610038347 A CN 201610038347A CN 105718932 A CN105718932 A CN 105718932A
- Authority
- CN
- China
- Prior art keywords
- vector machine
- support vector
- smooth
- image
- fruit bat
- 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
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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]
Abstract
The invention discloses a colorful image classification method based on a fruit fly optimization algorithm and a smooth twinborn support vector machine and a system thereof, and relates to the artificial intelligence and computer image processing technology field. The colorful image classification problem can be solved. The colorful image classification method is characterized in that weighted mixed characteristics can be formed by extracting color characteristics and texture characteristic of colorful images; the parameters of the smooth twinborn support vector machine can be selected by using the fruit fly optimization algorithm; the selected parameters and the weighted mixed characteristics of the colorful images can be used as the secondary planning problem, which can be input to construct the smooth twinborn support vector machine, and the secondary planning problem can be solved to acquire the smooth twinborn support vector machine classifier by using the smooth method; the weighted mixed characteristics of the new image of unknown category can be extracted, and can be input to the smooth twinborn support vector machine classifier to output the category corresponding to the new image.
Description
Technical field
The present invention relates to artificial intelligence and computer image processing technology field.Particularly to a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system.
Background technology
Along with the fast development of computer technology, the Internet and multimedia technology, image resource emerges in multitude.Particularly smart mobile phone is universal so that the collection of color digital image, stores and shares quantity and rapidly increase.Image is classified the most basic means being to obtain useful information from image, be also that people analyze, understand and the important channel of application image resource.Image classification is according to the different characteristic reflected in different classes of digital picture, the image processing method that different classes of target area is separated.Image Classification Studies is widely used, and such as the classification of image retrieval, mechanical workpieces, distinguishes scene from remote sensing images or satellite photo, assists diagnosis etc. according to CT image.
Conventional sorting technique includes Bayes classifier, artificial neural network and decision tree etc..But, Bayes classifier needs prior probability, and requires between attribute separate, and artificial neural network and decision tree exist over-fitting problem.Color image information often data volume is relatively larger, and image feature data structure is complicated, and above method tends not to process coloured image classification well.Therefore it provides a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system.
Summary of the invention
The present invention provides a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system, Swarm Intelligent Algorithm can be used to automatically determine the parameter of smooth twin support vector machine, and using smoothing method to solve the quadratic programming problem of smooth twin support vector machine, training obtains a smooth twin support vector machine Image Classifier.A kind of as follows based on the coloured image sorting technique of fruit bat optimized algorithm and smooth twin support vector machine and the main technical schemes of system:
A kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system mainly comprise the steps that
Step 1: color feature extracted.RGB color being converted to hsv color space, hsv color space uniform is quantified as 256 different colours, H is quantized into 16 grades, and S is quantized into 4 grades, and V is quantized into 4 grades.Adding up the different colours ratio shared in the picture color characteristic as this image, then the color characteristic of each image can with the vector representation of one 256 dimension.
Step 2: texture feature extraction.Each color component is converted into gray scale, obtains the gray matrix of coloured image, to gray-scale compression, it it is 16 grades by gray matrix uniform quantization, calculate gray level co-occurrence matrixes, according to gray level co-occurrence matrixes calculate energy, entropy, dependency, unfavourable balance from, adopt the average of these four indexs as textural characteristics.Then the textural characteristics of each image can with the vector representation of one 4 dimension.
Step 3: structure weighted blend feature.Color feature vector step 2 obtained with step 3 is normalized the composite character as coloured image after giving different weights with textural characteristics and combining.
Step 4: fruit bat optimized algorithm initializes.Set the maximum iteration time MAX_i of fruit bat optimized algorithm, population size, set fruit bat initial position (X_axis, Y_axis) and for initial point and set initial flight distance (X, Y) at random, setting initial best flavors concentration Best_Smell as 0, current iteration number of times i is 0.
Step 5: utilize fruit bat optimized algorithm to choose the parameter of smooth twin support vector machine.Run fruit bat optimized algorithm, calculate individual flavor concentration decision content vector S (i, 1) of current each fruit bat and S (i, 2), using vectorial for the concentration decision content parameter as smooth twin support vector machine.Randomly selecting 80% image as training set, residual image constitutes test set.Weighted blend characteristic that training set image is obtained in step 3 and by the given smooth twin support vector machine parameter of current fruit bat optimized algorithm as input, training obtains smooth twin supporting vector machine model, test set inputs smooth twin supporting vector machine model and calculates accuracy rate ACCi as flavor concentration Smell (i).If flavor concentration Smell (i) is more than the best flavors concentration Best_Smell in current record, then by value that best flavors concentration assignment is Smell (i) the fruit bat positional information and the taste decision content that record correspondence.Current iteration number of times i adds 1.If current iteration number of times is more than maximum iteration time, then perform step 6;The position otherwise updating all fruit bat positions is the fruit bat position that currently best accuracy rate is corresponding, if best flavors concentration does not update in this step, all fruit bat positions is added a Gauss disturbance, is otherwise not added with.Then random imparting one flying distance of each fruit bat, then re-executes step 5.
Step 6: the structure of smooth twin support vector machine classifier.Using the taste decision content that finally records of step 5 as the composite character data of parameter and all images as input, adopt the quadratic programming problem of the smooth twin support vector machine of smooth Algorithm for Solving, training obtains final smooth twin support vector machine classifier.
Step 7: categorised decision.When obtaining the new images of a unknown classification, quantify and perform step 2, step 3 and step 4 to obtain new images characteristic by image input, the smooth twin support vector machine classifier obtained by this feature input step 7, the classification that output new images is corresponding.
The invention have the advantages that and effect:
(1) this method uses fruit bat optimized algorithm to determine the parameter of grader, it is possible to obtain rational parameter value.
(2) this method uses smooth twin support vector machine as grader, has training speed faster.
Accompanying drawing explanation
Accompanying drawing 1 is the implementing procedure figure of a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine of the present invention and system.
Accompanying drawing 2 is a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine of the present invention and the fruit bat optimized algorithm flow chart in system.
Accompanying drawing 3 is that a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine of the present invention solves the flow chart of smooth twin supporting vector machine model with the Newton's algorithm that uses in system.
Accompanying drawing 4 is the module composition schematic diagram of a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine of the present invention and system.
Accompanying drawing 5 is the process schematic of a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system disclosed in the embodiment of the present invention.
Accompanying drawing 6 is the Performance comparision figure of a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine of the present invention and system.
Detailed description of the invention
The present invention is a kind of based on the coloured image sorting technique of fruit bat optimized algorithm and smooth twin support vector machine and the implementation process of system is:
The basic framework of the present invention is smooth twin support vector machine classifier, it it is a kind of supervised classification method, training smooth twin support vector machine process is sample characteristics two quadratic programming problems of structure utilizing input, two hyperplane are obtained by solving quadratic programming problem, the each corresponding classification of each hyperplane, classifying rules is new samples is just categorized into the category close to hyperplane corresponding to which class.Above-mentioned hyperplane and classifying rules just constitute smooth twin support vector machine classifier.Smooth twin support vector machine classifier needs to set two parameters, automatically determines parameter by fruit bat optimized algorithm.
The specific embodiment of the invention is described in detail below in conjunction with Fig. 1.
Step 1: S101 in Fig. 1, color of image feature extraction.The image of storage is the matrix using rgb space to represent, RGB color is converted to hsv color space, and for being expressed as the pixel in the coloured image of (R, G, B) in a RGB color, concrete conversion formula is:
V=max (R, G, B),
Order
Calculated (H, S, V) is pixel value in HSV space.Hsv color space uniform is quantified as 256 different colours, and H is quantized into 16 grades, and S is quantized into 4 grades, and V is quantized into 4 grades.Adding up the different colours ratio shared in the picture color characteristic as this image, then the color characteristic of each image can with the vector representation of one 256 dimension.
Step 2: image texture characteristic extracts, this step corresponds to S102.Colour picture being converted into grayscale mode, obtains the gray matrix of image, concrete is be the pixel of (R, G, B) for some rgb value in coloured image, and the gray scale Gray value of this point is
Gray=0.30R+0.59G+0.11B
The ash point calculating all pixels of width coloured image obtains the gray matrix of this image.To gray-scale compression, gray matrix uniform quantization is 16 grades and obtains gray scale f (x after compression of images, y), the element p (i of gray level co-occurrence matrixes P, gray level co-occurrence matrixes P is calculated, j, δ, θ) by add up f (x, in y) from gray scale be the pixel of i, distance is δ, and angle is θ place gray scale is that the number of times summation that obtains divided by all statistics again of number of times that the pixel of j occurs is tried to achieve.The co-occurrence matrix P so obtained has been normalized, can calculate energy, entropy according to gray level co-occurrence matrixes, dependency, unfavourable balance from.
Energy: ASM=ΣiΣjp2(i,j)
Entropy: ENT=-ΣiΣjp(i,j)lg(p(i,j))
Dependency:
Wherein
Unfavourable balance is from INV=ΣiΣjp(i,j)/(1+(i+j)2)
Taking distance for δ is 1 pixel, and angle is θ is 0 degree, 45 degree, 90 degree and 135 degree, obtain four gray level co-occurrence matrixes P1, P2, P3 and P4 by previously described method, according to gray level co-occurrence matrixes calculate energy, entropy, dependency, unfavourable balance from, then take the mean respectively as textural characteristics.Then the color characteristic of each image can with the vector representation of one 4 dimension.
Step 3: structure weighted blend feature, this step is corresponding to S103 in Fig. 1.Color feature vector step 2 obtained with step 3 needs to give different weights with textural characteristics according to actual motion and is then combined and is normalized obtaining composite character.According to practical situation, to the cromogram image set only comprising similar color within five kinds, the weights of color characteristic take 0.4, and the cromogram image set to various colors, the weights of color characteristic take 0.6.According to practical situation, image texture is unintelligible, the weights desirable 0.4 of textural characteristics, and image texture is clear, the weights desirable 0.6 of textural characteristics.
S104 correspondence step 4 and step 5 in Fig. 1.Fig. 2 is the flow chart of fruit bat optimized algorithm, and fruit bat optimized algorithm is that smooth twin support vector machine classifier finds suitable parameter through some step iteration after initializing.Concrete is as follows.
Step 4: initialize fruit bat optimized algorithm.Set the maximum iteration time MAX_i of fruit bat optimized algorithm as 500, population size is as 20 and initial fruit bat initial position (X_axis, Y_axis) for initial point random assignment initial flight distance (X, Y), setting initial best flavors concentration Best_Smell as 0, current iteration number of times i is 0.
Step 5: utilize fruit bat optimized algorithm to choose the parameter of smooth twin support vector machine.Run fruit bat optimized algorithm, utilize olfactory sensation random flight certain distance search food, namely calculate the flavor concentration decision content vector that current each fruit bat is individual,
Using vectorial for flavor concentration decision content two parameters as smooth twin support vector machine.Randomly selecting 80% image as training set, residual image constitutes test set.Composite character data C1 that training set image is obtained in step 4 and by the given smooth twin support vector machine parameter of current fruit bat optimized algorithm as input, constitute two quadratic programming problems:
Wherein A, B represent the image blend characteristic of+1 class and-1 class two class image pattern, and each provisional capital is the composite character of an image.K represents gaussian radial basis function kernel function.E1And e2Be element it is the vector of 1 entirely.C1=S (i, 1), c2=S (i, 2).Solving quadratic programming above and obtain smooth twin supporting vector machine model, the concrete mode of Quadratic Programming Solution is shown in step 6.Note solves the solution obtained and isWithThe vector composite character of image in test set formed in turn inputs and calculates
With
Relatively l1And l2Size, the former little then by this test spectral discrimination be similar with A image, i.e.+1 class, otherwise it is designated as another kind of, i.e.-1 class, whether the output then contrasting smooth twin support vector machine is consistent with the concrete class of correspondence image, and accuracy rate ACCi is as flavor concentration Smell (i) for statistics.If flavor concentration Smell (i) is more than the best flavors concentration Best_Smell in current record, then by value that Best_Smell assignment is Smell (i) the fruit bat positional information and the taste decision content that record correspondence.Current iteration number of times i adds 1, if current iteration number of times is more than maximum iteration time, then performs step 6, and the position otherwise updating all fruit bat positions is the fruit bat position that currently best accuracy rate is corresponding.If current best flavors concentration is constant in this step, fruit bat position adding a Gauss disturbance, namely Gauss disturbance meets the random number of NATURAL DISTRIBUTION;Otherwise it is not added with.Random imparting one flying distance of each fruit bat, then re-executes step 5.
Step 6: S105 in Fig. 1: the structure of smooth twin support vector machine classifier.Use taste decision content S (500,1) that finally records of step 5 and S (500,2) as the composite character data C of parameter and all images as input, be constructed as follows quadratic programming problem:
s.t-(K(B,CT)w1+e2b1)+ζ≥e2,ζ≥0
s.t(K(A,CT)w2+e1b2)+η≥e1,η≥0
A, B represent the coloured image composite character data of+1 class and-1 class two class image pattern in all images respectively.Adopt smoothing method to send out and solve quadratic programming problem.This quadratic programming is equivalent to following nothing constraint minima planning problem.
Plus () function is by element less than 0 in vector or matrix all with 0 replacement herein, is non-differentiable function.Integral function with unlimited continuously differentiable function Sigmoid function
Approximate replacement Plus () function, obtains two smooth nothing constraint minima planning problems:
By the smooth approximate solution obtaining quadratic programming problem without constraint minima planning problem of Newton method rapid solving the two.Concrete is, as it is shown on figure 3, seek first smooth nothing constraint minima planning problem, set w1, b1Initial value be 0, maximum iteration time is 500, and allowable error is 0.0001, calculate Φ1(w1,b1) derived functionIfEqual to 0, the w now obtained1, b1It is solutionOtherwise calculate Φ1(w1,b1) andThe mould of inverse product, and judge whether more than 0.0001.If it is not, the w now obtained1, b1It is solutionIf just willIt is assigned to (w1,b1), recalculate Φ1(w1,b1) with and subsequent flow process until try to achieve solution or iterative steps reach 500.Same, solved by Newton methodThen may determine that two hyperplane
With
Here it is the Optimal Separating Hyperplane of smooth twin support vector machine classifier.Decision function is
Namely the vector that new picture sample is converted to just is classified as the first kind close to the 1st hyperplane ,+1 class;Otherwise it is classified as-1 class.
Step 7: categorised decision, i.e. S106 in Fig. 1.When obtaining the new images of a unknown classification, quantify and perform step 2, step 3 and step 4 to obtain new images characteristic by image input, the smooth twin support vector machine classifier obtained by this feature input step 6, namely calculates decision function, is output as the classification that new images is corresponding.
A kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine of the present invention and system, in simulation process, use the independent simulation run respectively through 10 times, and compare with the method based on SVM in existing method and the method based on TWSVM, running the meansigma methods of the error rate obtained on 5 different cromogram image sets for 10 times as shown in Figure 6, this shows that colour picture can be classified by method provided by the invention exactly.
Fig. 4 reflects the present invention and is mainly made up of three modules, S410 coloured image characteristic extracting module is responsible for extracting weighted blend feature from coloured image, S402 is responsible for the parameter utilizing fruit bat optimized algorithm to choose smooth twin support vector machine based on the parameter optimization module of fruit bat algorithm, and S403 is responsible for training smooth twin support vector machine classifier.S401 and S402 is output as the input of S403, it is therefore necessary to first starts to perform S401 and S402 and just can perform S403.
Fig. 5 is the process schematic of a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine disclosed in the embodiment of the present invention and system.This process schematic reflects in the present invention between modules concrete cooperation relation in force and flow process.First by the image contract characteristics of image in the image set of the image set of first kind coloured image and Equations of The Second Kind coloured image, Matrix C AR and the MB of two phenogram pictures is obtained;Then this step is obtained Input matrix fruit bat optimized algorithm and obtains the parameter of smooth twin support vector machine;The parameter training finally utilizing Matrix C AR and MB that first two steps obtain and smooth twin support vector machine obtains smooth twin support vector machine classifier.
Claims (5)
1., based on the coloured image sorting technique of fruit bat optimized algorithm and smooth twin support vector machine and a system, specifically include that
Step 1: color feature extracted, is converted to hsv color space by RGB color, is quantified by hsv color space uniform, adds up the different colours ratio shared in the picture color characteristic as this image;
Step 2: texture feature extraction, is converted into gray scale by each color component, to gray-scale compression, calculates co-occurrence matrix, according to gray level co-occurrence matrixes calculate energy, entropy, dependency, unfavourable balance from, adopt the average of these four indexs as textural characteristics;
Step 3: structure weighted blend feature, color feature vector step 2 obtained with step 3 gives different weights with textural characteristics, then combines and is normalized the composite character as coloured image;
Step 4: fruit bat optimized algorithm initializes, sets the maximum iteration time of fruit bat optimized algorithm, population size and fruit bat initial position and sets initial flight distance at random, setting initial best flavors concentration as 0, and current iteration number of times is 0;
Step 5: utilize fruit bat optimized algorithm to choose the parameter of smooth twin support vector machine, calculate current each fruit bat individual flavor concentration decision content parameter as smooth twin support vector machine, randomly select 80% image as training set, residual image constitutes test set, the weighted blend characteristic obtained in step 3 by training set image is as input, training obtains smooth twin supporting vector machine model, test set inputs smooth twin supporting vector machine model and calculates accuracy rate ACCiAs flavor concentration Smell (i), if flavor concentration Smell (i) is more than best flavors concentration Best_Smell, by value that best flavors concentration assignment is Smell (i) the fruit bat positional information and the taste decision content that record correspondence, iterations adds 1, if current iteration number of times is more than maximum iteration time, then perform step 6, otherwise update the position of all fruit bats, if best flavors concentration does not update in this step, all fruit bat positions are added a Gauss disturbance, random imparting one flying distance of each fruit bat, then re-executes step 5;
Step 6: the structure of smooth twin support vector machine classifier, use the taste decision content that finally records of step 5 as the composite character data of parameter and all images as input, adopting the quadratic programming problem of the smooth twin support vector machine of smooth Algorithm for Solving, training obtains final smooth twin support vector machine classifier;
Step 7: when obtaining the new images of a unknown classification, quantify and perform step 2, step 3 and step 4 to obtain the characteristic of new images by image input, the smooth twin support vector machine classifier obtained by this feature input step 7, is output as the classification that new images is corresponding.
2. a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system according to claim 1, it is characterized in that: in step 3, adopt the weighted blend character representation image of color characteristic and textural characteristics, first coloured image is extracted color characteristic, image is converted into grayscale mode texture feature extraction again, color characteristic is multiplied by weights, to the cromogram image set only comprising similar color within five kinds, these weights take 0.4, cromogram image set to various colors, these weights take 0.6, textural characteristics is multiplied by weights, according to practical situation, image texture is unintelligible, these weights desirable 0.4, image texture is clear, these weights desirable 0.6, after weighting, color characteristic and textural characteristics collectively constitute the feature of coloured image.
3. a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system according to claim 1, it is characterized in that: described step 5, fruit bat optimized algorithm is used to choose the parameter of smooth twin support vector machine, for avoiding falling into local optimum, the fruit bat position updating process of fruit bat optimized algorithm adds Gauss disturbance.
4. a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system according to claim 1, it is characterized in that: step 6, adopting smooth twin support vector machine as grader framework, the mathematical model of smooth twin support vector machine is two quadratic programming problems:
A, B represent the composite character data of+1 class and-1 class two class image pattern respectively, and each provisional capital is the composite character of an image, and K represents gaussian radial basis function kernel function, and this quadratic programming problem is equivalent to following nothing constraint minima planning problem:
In Unconstrained optimization model, Plus () function is by element less than 0 in vector or matrix all with 0 replacement, is non-differentiable function, and use smoothing method solves quadratic programming problem.
5. a kind of coloured image sorting technique based on fruit bat optimized algorithm and smooth twin support vector machine and system according to claim 4, it is characterised in that: use smoothing method to solve the quadratic programming problem of smooth twin support vector machine: with the integral function of unlimited continuously differentiable function Sigmoid function
Non-differential term in the object function of the quadratic programming problem of approximate smooth twin support vector machine, obtains two smooth nothing constraint minima planning problems,
Solve the two unconstrained problem by Newton's algorithm, obtain the approximate solution of the quadratic programming of smooth twin support vector machine.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610038347.4A CN105718932A (en) | 2016-01-20 | 2016-01-20 | Colorful image classification method based on fruit fly optimization algorithm and smooth twinborn support vector machine and system thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610038347.4A CN105718932A (en) | 2016-01-20 | 2016-01-20 | Colorful image classification method based on fruit fly optimization algorithm and smooth twinborn support vector machine and system thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105718932A true CN105718932A (en) | 2016-06-29 |
Family
ID=56147318
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610038347.4A Pending CN105718932A (en) | 2016-01-20 | 2016-01-20 | Colorful image classification method based on fruit fly optimization algorithm and smooth twinborn support vector machine and system thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105718932A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106991434A (en) * | 2017-03-07 | 2017-07-28 | 中国矿业大学 | A kind of gray-scale map sorting technique and system based on the twin SVMs of small echo |
CN107678930A (en) * | 2017-09-11 | 2018-02-09 | 华东理工大学 | A kind of bank's automatic terminal abnormal alarm method based on Smooth Support Vector Machines |
CN108918587A (en) * | 2018-09-28 | 2018-11-30 | 山东建筑大学 | A kind of research method of heterogeneous material heat sustained release rule |
CN109993178A (en) * | 2017-12-29 | 2019-07-09 | 华为技术有限公司 | A kind of characteristic generates and feature matching method and device |
CN110033053A (en) * | 2019-04-22 | 2019-07-19 | 中国石油大学(华东) | Optimize the Favorable Reservoir prediction technique of twin support vector machines based on improvement drosophila |
CN110659482A (en) * | 2019-09-27 | 2020-01-07 | 吉林大学 | Industrial network intrusion detection method based on GAPSO-TWSVM |
CN111767803A (en) * | 2020-06-08 | 2020-10-13 | 北京理工大学 | Identification method for anti-target attitude sensitivity of synthetic extremely-narrow pulse radar |
CN112560672A (en) * | 2020-12-15 | 2021-03-26 | 安徽理工大学 | Fire image recognition method based on SVM parameter optimization |
CN113901965A (en) * | 2021-12-07 | 2022-01-07 | 广东省科学院智能制造研究所 | Liquid state identification method in liquid separation and liquid separation system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005122066A1 (en) * | 2004-06-04 | 2005-12-22 | Siemens Medical Solutions Usa, Inc. | Support vector classification with bounded uncertainties in input data |
CN104091216A (en) * | 2014-07-29 | 2014-10-08 | 吉林大学 | Traffic information predication method based on fruit fly optimization least-squares support vector machine |
CN104794368A (en) * | 2015-05-15 | 2015-07-22 | 哈尔滨理工大学 | Rolling bearing fault classifying method based on FOA-MKSVM (fruit fly optimization algorithm-multiple kernel support vector machine) |
-
2016
- 2016-01-20 CN CN201610038347.4A patent/CN105718932A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005122066A1 (en) * | 2004-06-04 | 2005-12-22 | Siemens Medical Solutions Usa, Inc. | Support vector classification with bounded uncertainties in input data |
CN104091216A (en) * | 2014-07-29 | 2014-10-08 | 吉林大学 | Traffic information predication method based on fruit fly optimization least-squares support vector machine |
CN104794368A (en) * | 2015-05-15 | 2015-07-22 | 哈尔滨理工大学 | Rolling bearing fault classifying method based on FOA-MKSVM (fruit fly optimization algorithm-multiple kernel support vector machine) |
Non-Patent Citations (3)
Title |
---|
于俊钊: "孪生支持向量机及其优化方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
朱志宾 等: "基于TWSVM的图像分类", 《南京师大学报(自然科学版)》 * |
黄华娟: "孪生支持向量机关键问题的研究", 《中国优秀博士学位论文全文数据库 信息科技辑》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106991434A (en) * | 2017-03-07 | 2017-07-28 | 中国矿业大学 | A kind of gray-scale map sorting technique and system based on the twin SVMs of small echo |
CN107678930A (en) * | 2017-09-11 | 2018-02-09 | 华东理工大学 | A kind of bank's automatic terminal abnormal alarm method based on Smooth Support Vector Machines |
CN109993178A (en) * | 2017-12-29 | 2019-07-09 | 华为技术有限公司 | A kind of characteristic generates and feature matching method and device |
CN109993178B (en) * | 2017-12-29 | 2024-02-02 | 华为技术有限公司 | Feature data generation and feature matching method and device |
CN108918587A (en) * | 2018-09-28 | 2018-11-30 | 山东建筑大学 | A kind of research method of heterogeneous material heat sustained release rule |
CN110033053A (en) * | 2019-04-22 | 2019-07-19 | 中国石油大学(华东) | Optimize the Favorable Reservoir prediction technique of twin support vector machines based on improvement drosophila |
CN110659482A (en) * | 2019-09-27 | 2020-01-07 | 吉林大学 | Industrial network intrusion detection method based on GAPSO-TWSVM |
CN111767803A (en) * | 2020-06-08 | 2020-10-13 | 北京理工大学 | Identification method for anti-target attitude sensitivity of synthetic extremely-narrow pulse radar |
CN112560672A (en) * | 2020-12-15 | 2021-03-26 | 安徽理工大学 | Fire image recognition method based on SVM parameter optimization |
CN113901965A (en) * | 2021-12-07 | 2022-01-07 | 广东省科学院智能制造研究所 | Liquid state identification method in liquid separation and liquid separation system |
CN113901965B (en) * | 2021-12-07 | 2022-05-24 | 广东省科学院智能制造研究所 | Liquid state identification method in liquid separation and liquid separation system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105718932A (en) | Colorful image classification method based on fruit fly optimization algorithm and smooth twinborn support vector machine and system thereof | |
CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
CN108132968B (en) | Weak supervision learning method for associated semantic elements in web texts and images | |
CN108647602B (en) | A kind of aerial remote sensing images scene classification method determined based on image complexity | |
CN112883839B (en) | Remote sensing image interpretation method based on adaptive sample set construction and deep learning | |
CN112347970B (en) | Remote sensing image ground object identification method based on graph convolution neural network | |
CN107680077A (en) | A kind of non-reference picture quality appraisement method based on multistage Gradient Features | |
CN107516103A (en) | A kind of image classification method and system | |
CN112232151B (en) | Iterative polymerization neural network high-resolution remote sensing scene classification method embedded with attention mechanism | |
CN111489370B (en) | Remote sensing image segmentation method based on deep learning | |
CN112115967B (en) | Image increment learning method based on data protection | |
CN111161244B (en) | Industrial product surface defect detection method based on FCN + FC-WXGboost | |
CN110807485B (en) | Method for fusing two-classification semantic segmentation maps into multi-classification semantic map based on high-resolution remote sensing image | |
CN105931241A (en) | Automatic marking method for natural scene image | |
CN106203448B (en) | A kind of scene classification method based on Nonlinear Scale Space Theory | |
CN111462162A (en) | Foreground segmentation algorithm for specific class of pictures | |
CN116912708A (en) | Remote sensing image building extraction method based on deep learning | |
CN114842264A (en) | Hyperspectral image classification method based on multi-scale spatial spectral feature joint learning | |
CN113435254A (en) | Sentinel second image-based farmland deep learning extraction method | |
CN111695436B (en) | High spatial resolution remote sensing image scene classification method based on target enhancement | |
Shukran et al. | Image classification technique using modified particle swarm optimization | |
CN111860601B (en) | Method and device for predicting type of large fungi | |
CN110807369B (en) | Short video content intelligent classification method based on deep learning and attention mechanism | |
CN111612090A (en) | Image emotion classification method based on content color cross correlation | |
CN115512207A (en) | Single-stage target detection method based on multipath feature fusion and high-order loss sensing sampling |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160629 |