CN108931621A - A kind of zinc ore grade flexible measurement method of Kernel-based methods textural characteristics - Google Patents
A kind of zinc ore grade flexible measurement method of Kernel-based methods textural characteristics Download PDFInfo
- Publication number
- CN108931621A CN108931621A CN201810446656.4A CN201810446656A CN108931621A CN 108931621 A CN108931621 A CN 108931621A CN 201810446656 A CN201810446656 A CN 201810446656A CN 108931621 A CN108931621 A CN 108931621A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- value
- textural characteristics
- zinc
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/20—Metals
-
- 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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medicinal Chemistry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Food Science & Technology (AREA)
- Software Systems (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
A kind of method of the hard measurement of the zinc Floatation Concentrate Grade of Kernel-based methods feature, the present invention combines expertise and Data Modeling Method, see that observation when bubble brings forward the single frames textural characteristics based on image statistics feature and characterizes froth images according to field worker first, according to field worker need to observe in a period of time foam state come proposition the characteristics of judging current production status with texture sequence come the current production status of mathematicization, and propose the modeling method of a kind of pair of texture sequence, reduce the dimension of feature vector.It is effective to inhibit to improve generalization ability due to learning too fast caused overfitting problem using improved promotion decision Tree algorithms in prediction algorithm.Experiments have shown that the method for the present invention calculates simply, fast speed is executed, prediction accuracy is higher, is convenient for practical operation at the scene, can instruct execute-in-place immediately, optimizes production process, solves the problems, such as that existing zinc ore grade on-line checking is difficult.
Description
Technical field
The invention belongs to froth flotation technical fields, and in particular to a kind of prediction technique of zinc Floatation Concentrate Grade.
Background technique
Froth flotation is one of most important beneficiation method in zinc abstraction now, and flotation is according to mineral particle surface physics
The difference of chemical property, by the method that the difference of mineral floatability is sorted, froth flotation be one will be smashed useful
The process that mineral are separated with the gangue of its symbiosis is formed big by constantly stirring and the air blast in floatation process in ore pulp
Measurer have different sizes, form, Texture eigenvalue bubble, bubble carry mineral grain rise to tank surface formed foam
Layer, to realize the separation of mineral and gangue.The industrial process of complexity such a for froth flotation, because process flow is long,
Sub- process severe conjunction coupling, Partial key parameter are difficult to the reasons such as on-line checking, and floatation process work condition state lacks effective
Synthesis perception means, depend critically upon artificial inspection back and forth, by virtue of experience substantially judge whether current production is in normal shape
State, further to implement corresponding operation strategy.This single rough, the perception of heavy dependence artificial experience method, often produces
Raw and inappropriate production operation, not can guarantee the stable optimal operation of production.Although selecting factory can be by offline assay
Accurate concentrate grade is obtained to determine the production status of floatation process, however this generally requires several hours, checkout procedure is multiple
It is miscellaneous and at high cost seriously lag behind production process.Since flotation process is long, influence factor is more, concentrate grade cannot achieve
On-line checking affects the instant adjustment to dosage and other parameters, finally affects the rate of recovery of mineral.Therefore, it studies
The real-time on-line detecting method of floatation process production target has important meaning to Instructing manufacture operation and the optimization of process operation
Justice.
With the fast development of computer technology, digital image processing techniques, by the soft-measuring technique based on machine vision
Real-time monitoring applied to floatation process to floatation indicators brings new breakthrough.Machine vision is a kind of imitation mankind itself view
Sensing capability is felt to realize the important means of industrial process automation measurement and control, because it has high-precision, modularization, intelligence
A variety of advantages such as change, lossless perception, may be implemented the on-line checking of floatation process concentrate grade.It can be with by image capture device
The foam video under a large amount of different grades is obtained, these videos are combined with collected corresponding creation data, shape
At a raw data set, the data model of froth images and concentrate grade is established using the method that data-driven models, and is realized
The on-line checking of concentrate grade.Existing mine concentration grade forecast method mainly use B-spline Partial Least Squares Regression, support to
The methods of amount machine, neural network, the defect of these methods all various degrees, they are difficult to handle the data of large sample,
And it is more sensitive to the data with noise, some problems are being still had using upper.
Summary of the invention
Difficult for concentrate grade on-line checking in zinc floatation process, at high cost, delay is big and in zinc flotation concentrate product
Deficiency on position prediction, the present invention propose a kind of lead zinc flotation using the Heuristics of field worker and the creation data of accumulation
The building method of froth images process feature, while a kind of prediction technique of concentrate grade is constructed, this method has good pre-
Precision is surveyed, anti-interference ability and has the faster speed of service.
Steps are as follows for the technical solution adopted by the present invention：
S1：The foam video and creation data for collecting the zinc flotation under different grades, to collected zinc flotation data with
And creation data carries out data prediction, process is as follows：
1) wrong data that the data value measured exceeds variation range is rejected；
2) unmatched data are rejected and there are the data of vacancy value；
S2：RGB froth images are read using flotation site image capturing system foam video obtained, by froth images
HSI color space is transformed by RGB color, and extract light intensity level obtains an image sequence I=as source images
[I1,I2,...,Iq], q is the frame number of video；
S3：Extract the i-th frame froth images I in image sequence IiTextural characteristics, be denoted as Ti, to each in image sequence I
Frame image all does same treatment, obtains texture sequence T=[T1,T2,...,Tq], wherein Ti=[βix,μix,γix,βiy,μiy,
γiy]；
S4：Gauss Markov autoregressive moving-average model, letter are established to size distribution series T obtained in S3
Number expression formula is as follows：
Wherein：X (k), x (k+1) are the state vectors of n dimension；
Y (k) is the output vector of m dimension, y (k)=TkRepresent textural characteristics；
V (k) is the stochastic variable of a Gaussian distributed, covariance matrix V；
W (k) is the stochastic variable of a Gaussian distributed, covariance matrix W；
K=1,2,3 ..., q；
Parameter A is estimated, the value of C, V are arranged into the mistake that column vector a F, F are referred to as this period of time floatation process
Journey textural characteristics；
S5：F obtained in S4 and corresponding concentrate grade G are combined, as a sample point Di={ Fi,
Gi)}；All videos of collection are found out into process grain feature, are combined with concentrate grade, all sample point set D are found out
={ (F1,G1),(F2,G2),...,(FN,GN), use Fi (j)Indicate FiJ-th of component；
S6：The sample set obtained in S5 is denoted as f using CART algorithm training decision-tree model0；
S7：Prediction model is established, a loss function Lf is designed and comes model output value and reality that quantitative calculating generates
The deviation of measured value generates one gradually and incrementally according to loss function Lf with f0Based on promotion tree-model, step is such as
Under：
1) by fiIt is updated to fi'=fi+φi+1(F), wherein：I=0,1,2 ..., φi+1For the weak learner being newly added；
2) allowable loss function Lf is enabledWherein,
η is a constant less than 1；
3) it solvesIt obtains so that losing letter
The φ that number Lf is minimizedi+1Numerical solution；
4) with φi+1For target value, decision-tree model is established using CART algorithm and obtains fi+1；
5) decision tree is repeatedly generated, symbiosis is set at L, and get a promotion decision tree：
fboost(F)=f0(F)+ηf1(F)+...+ηfL(F)；The value of η is determined by verifying collection；
S8：Zinc flotation froth sample data input industrial computer to be detected is obtained, computer is calculated according to step S3, S4
Obtained process grain feature is inputted model obtained in S7, the zinc concentrate grade of prediction can be obtained by process grain feature；
Weibull distribution function described in S3 isC is normalized constant, β, μ, γ
For parameter；
The invention proposes a kind of methods of the hard measurement of the zinc Floatation Concentrate Grade of Kernel-based methods textural characteristics, solve
The problem of live antimony ore grade on-line checking hardly possible；The characteristics of image of picture extraction is relied solely on for conventional method to characterize
The problem of current foam state, the experience of present invention combination field worker propose a kind of process size characteristic related to time.
See that observation when bubble brings forward the single frames textural characteristics based on image statistics feature and characterizes foam characteristics according to field worker,
Expertise has been merged, the state of current foam can be more accurately represented；It is needed to observe a period of time according to field worker
Interior foam state proposes the modeling methods of a kind of pair of size distribution series the characteristics of judging current production status, reduce feature
The dimension of vector, while also can reflect the rule of size distribution series dynamic change；It is difficult to for traditional grade prediction technique
The data of large sample are handled, and sensitive issue is compared to the data with noise, product are established in proposition in the way of data-driven
The promotion decision-tree model of position prediction；Due to tree-model be a nonlinear model therefore can be very good fitting grade predict this
One intensive nonlinear functions of sample, and tree-model is branched structure, arithmetic speed quickly, can rapid solving go out current foam-like
The corresponding concentrate grade of state, is easy to implement on-line checking；Hold for promoting decision-tree model and being applied in the prediction of zinc concentrate grade
Over-fitting easily occur leads to the problem of the model generalization ability difference trained, introduces learning rate η, in the process that each forward direction is incremented by
It is middle that a constant less than 1 is added, it can effectively inhibit to improve extensive energy due to learning too fast caused overfitting problem
Power；This method calculates simply, executes fast speed, and prediction accuracy is higher, is convenient for practical operation at the scene, can instruct immediately existing
Field operation, optimizes production process.
Detailed description of the invention
Fig. 1 is the flow chart that the present invention implements zinc Floatation Concentrate Grade prediction technique.
Specific embodiment
Here is to combine attached drawing of the present invention, in further detail, is clearly made that technical solution employed in the present invention
It describes and explains.The present invention relies solely on single frames picture for conventional method and is difficult to accurately reflect the limitation proposition of foam state
A kind of process size characteristic extracting method related to time, and zinc concentrate product are realized using improved promotion decision-tree model
The on-line checking of position.Obviously, described embodiment is only a part in the embodiment of the present invention, is not the complete of embodiment
Portion.Based on the embodiments of the present invention, those skilled in the relevant art are in the obtained institute of premise for not making creative work
There are other embodiments all to should be protection scope of the present invention.
As shown in Figure 1, for the soft survey of Pb-Zn deposits grade of one of embodiment of the present invention Kernel-based methods size distribution characteristics
The flow chart of amount method, this method comprise the following specific steps that：
S1：The foam video and creation data for collecting the zinc flotation under different grades, to collected zinc flotation data with
And production
Data carry out data prediction, and process is as follows：
1) wrong data that the data value measured exceeds variation range is rejected；
2) unmatched data are rejected and there are the data of vacancy value；
S2：RGB froth images are read using flotation site image capturing system foam video obtained, by froth images
HSI color space is transformed by RGB color, and extract light intensity level obtains an image sequence I=as source images
[I1,I2,...,Iq], q is the frame number of video；
S3：To the froth images I in image sequenceiTexture feature extraction is carried out, the process of texture feature extraction is as follows：
1) with median filtering filter to IiMedian filtering is done, noise is removed, obtains Ii'；3*3 mould is used in the method
The median filter of plate,
2) using the Gaussian derivative filter in the direction x to Ii' filtering, obtain the image B comprising image detail edgeix；
3) B is countedixThe probability histogram H of gray scaleix, traverse the pixel that all pixel statistics adhere to different brightness separately
Number, divided by image resolution ratio, Hix={ (- 255, H-255),(-254,H-254),…,(255,H255)}；
4) by HixGrey linear transformation to section [- 1 ,+1], obtain standardized probability histogram Hix, using weber
Distribution function is fitted Hix, solveObtain the parameter beta of Weibull distribution functionix,μix,γix；
5) using the Gaussian derivative filter in the direction y to Ii' filtering, obtain the image B comprising image detail edgeiy；
6) B is countediyThe probability histogram H of gray scaleiy, traverse the pixel that all pixel statistics adhere to different brightness separately
Number, obtain H divided by image resolution ratioiy；
7) by HiyProbability histogram H of the grey linear transformation to [- 1 ,+1], after being standardizedix, with weber point
Cloth Function Fitting Hiy, solveObtain the parameter beta of Weibull distribution functioniy,μiy,γiy；
By Ti=[βix,μix,γix,βiy,μiy,γiy] it is used as image IiTextural characteristics, to each frame in image sequence I
Image all does same treatment, obtains texture sequence T=[T1,T2,...,Tq]；
S4：Gauss Markov autoregressive moving-average model, function table are established to texture sequence T obtained in S3
It is as follows up to formula：
Wherein：X (k), x (k+1) are the state vectors of n dimension；
Y (k) is the output vector of m dimension, y (k)=TkRepresent textural characteristics；
V (k) is the stochastic variable of a Gaussian distributed, is desired for 0, covariance matrix V；
W (k) is the stochastic variable of a Gaussian distributed, is desired for 0, covariance matrix W；
K=1,2,3 ..., q；
Parameter A is estimated, the value of C, V are arranged into the mistake that column vector a F, F are referred to as this period of time floatation process
Journey textural characteristics, steps are as follows：
1) by the y (k) in the Gauss Markov sliding autoregression model in S4, following form of doing, Y=[y (1), y are write
(2),y(3),...,y(q)]；
2) all output y (k) of model are found out, k=1,2 ..., the average value y of qm,Enable Yq=Y-
ym；
3) using a mapping function from m*q dimension space to real number field as probability density function, to simulate y (k),
Mathematic(al) representation is as follows：
Wherein：μ is one and YqThe identical matrix of shape, ∑ is a symmetrical matrix and element is positive number, and z is one and returns
One changes constant；
4) it solves so that p (Yq) it is maximum when parameter value, i.e.,Specific steps
It is as follows：
1. former formula is written as follow form：Y1 τ=[y (1), y (2) ..., y (k)],W1 τ
=[w (1), w (2) ..., W (k)], then
2. to Y1 τY can be obtained by doing singular value decomposition1 τ=U ∑ VT, then
3. solvingObtain the estimated value of A；
④Wherein
5) A, C, V are arranged in a column vector FiIf its dimension is n；
S5：F obtained in S4 and corresponding concentrate grade G are combined, as a sample point Di={ Fi,
Gi)}；All videos of collection are found out into process grain feature, are combined with concentrate grade, all sample point set D are found out
={ (F1,G1),(F2,G2),...,(FN,GN)}；
S6：F is denoted as using CART algorithm training decision-tree model with the sample set obtained in S50；
S7：Prediction model is established, a loss function Lf is designed and comes model output value and reality that quantitative calculating generates
The deviation of measured value generates one gradually and incrementally according to loss function Lf with f0Based on promotion tree-model, step is such as
Under：
1) by fiIt is updated to fi'=fi+φi+1(F), wherein：I=0,1,2 ..., φi+1For the weak learner being newly added；
2) allowable loss function Lf, taking loss function in the method is L2 loss function, is enabled
Wherein, η is a constant less than 1；
3) it solvesIt obtains so that losing letter
The φ that number Lf is minimizedi+1Numerical solution, enableFind out φi+1；
4) with φi+1For target value, decision-tree model is established using CART algorithm and obtains fi+1；
5) decision tree is repeatedly generated, is continuously added to newly set the feelings that will not result in prediction error and reduce if there is 10 times
Condition, it is believed that have reached stop condition, symbiosis is set at L, and get a promotion decision tree：
fboost(F)=f0(F)+ηf1(F)+...+ηfL(F)；
S8：Zinc flotation froth sample data input computer to be detected is obtained, computer is according to step S3, S4 calculating process
Obtained process grain feature is inputted model obtained in S7, the zinc concentrate grade of prediction can be obtained by textural characteristics.
Claims (2)
1. a kind of Pb-Zn deposits grade flexible measurement method of Kernel-based methods textural characteristics, it is characterised in that include the following steps：
S1：The foam video and creation data for collecting the zinc flotation under different grades, to collected zinc flotation data and life
It produces data and carries out data prediction, process is as follows：
1) wrong data that the data value measured exceeds variation range is rejected；
2) unmatched data are rejected and there are the data of vacancy value；
S2：Read RGB froth images using flotation site image capturing system foam video obtained, by froth images by
RGB color is transformed into HSI color space, and extract light intensity level obtains an image sequence I=[I as source images1,
I2,...,Iq], q is the frame number of video；
S3：Extract the i-th frame froth images I in image sequence IiTextural characteristics, be denoted as Ti, to frame figure each in image sequence I
As all doing above-mentioned processing, texture sequence T=[T is obtained1,T2,...,Tq], wherein Ti=[βix,μix,γix,βiy,μiy,γiy]；
S4：Gauss Markov autoregressive moving-average model, function table are established to size distribution series T obtained in S3
It is as follows up to formula：
Wherein：X (k), x (k+1) are the state vectors of n dimension；
Y (k) is the output vector of m dimension, y (k)=TkRepresent textural characteristics；
V (k) is the stochastic variable of a Gaussian distributed, its covariance matrix is V；
W (k) is the stochastic variable of a Gaussian distributed, covariance matrix W；
K=1,2,3 ..., q；
Parameter A is estimated, A, C, V are arranged in the mistake that column vector a F, F are referred to as this period of time floatation process by the value of C, V
Journey textural characteristics；
S5：F obtained in S4 and corresponding concentrate grade G are combined, as a sample point Di={ Fi,
Gi)}；All videos of collection are found out into process grain feature, are combined with concentrate grade, all sample point set D are found out
={ (F1,G1),...,(Fi,Gi),...,(FN,GN), use Fi (j)Indicate FiJ-th of component；
S6：F is denoted as using CART algorithm training decision-tree model with the sample set obtained in S50；
S7：Prediction model is established, a loss function Lf is designed and comes model output value and actual measurement that quantitative calculating generates
The deviation of value generates one gradually and incrementally according to loss function Lf with f0Based on promotion tree-model, its step are as follows：
1) by fiIt is updated to fi'=fi+φi+1(F), wherein：I=0,1,2 ..., φi+1For the weak learner being newly added；
2) allowable loss function Lf is enabledWherein, η is
One constant less than 1；
3) it solvesIt obtains so that loss function Lf takes
The φ of minimum valuei+1Numerical solution；
4) with φi+1For target value, decision-tree model is established according to step S6 and obtains fi+1；
5) it repeats the above steps, symbiosis is set at L, and get a promotion decision tree：
fboost(F)=f0(F)+ηf1(F)+...+ηfL(F)；
S8：Zinc flotation froth sample data input computer to be detected is obtained, computer is according to step S3, S4 calculating process texture
Obtained process grain feature is inputted model obtained in S7, the zinc concentrate grade of prediction can be obtained by feature.
2. the Pb-Zn deposits grade flexible measurement method of Kernel-based methods textural characteristics according to claim 1, it is characterised in that：
Estimate parameter A, the value of C, V are arranged into a column vector F, and steps are as follows：
1) by the y (k) in the Gauss Markov sliding autoregression model in S4, following form of doing, Y=[y (1), y (2), y are write
(3),...,y(q)]；
2) all output y (k) of model are found out, k=1,2 ..., the average value y of qm,Enable Yq=Y-ym；
3) using a mapping function from m*q dimension space to real number field as probability density function, to simulate y (k), mathematics
Expression formula is as follows：
Wherein：μ is one and YqThe identical matrix of shape, ∑ is a symmetrical matrix and element is positive number, and z is a normalization
Constant；
4) it solves so that p (Yq) it is maximum when parameter value, i.e.,
A, C, V are arranged in a column vector FiIf its dimension is n, Fi (j)Indicate FiJ-th of component.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810446656.4A CN108931621B (en) | 2018-05-11 | 2018-05-11 | Zinc ore grade soft measurement method based on process texture characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810446656.4A CN108931621B (en) | 2018-05-11 | 2018-05-11 | Zinc ore grade soft measurement method based on process texture characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108931621A true CN108931621A (en) | 2018-12-04 |
| CN108931621B CN108931621B (en) | 2020-10-02 |
Family
ID=64448193
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810446656.4A Active CN108931621B (en) | 2018-05-11 | 2018-05-11 | Zinc ore grade soft measurement method based on process texture characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108931621B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110109446A (en) * | 2019-05-28 | 2019-08-09 | 中南大学 | A kind of zinc floatation process Fuzzy Fault Diagnosis based on time series feature |
| CN110728677A (en) * | 2019-07-22 | 2020-01-24 | 中南大学 | Texture roughness defining method based on sliding window algorithm |
| CN110766673A (en) * | 2019-07-22 | 2020-02-07 | 中南大学 | Texture roughness defining method based on Euclidean distance judgment |
| CN115049165A (en) * | 2022-08-15 | 2022-09-13 | 北矿机电科技有限责任公司 | Flotation concentrate grade prediction method, device and equipment based on deep learning |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0644278B2 (en) * | 1984-03-27 | 1994-06-08 | 株式会社ニレコ | Method and apparatus for automatic quantitative measurement of tissue by image analysis |
| CN101036904A (en) * | 2007-04-30 | 2007-09-19 | 中南大学 | Flotation froth image recognition device based on machine vision and the mine concentration grade forecast method |
| CN101334366A (en) * | 2008-07-18 | 2008-12-31 | 中南大学 | Flotation recovery rate prediction method based on image characteristic analysis |
| CN101334844A (en) * | 2008-07-18 | 2008-12-31 | 中南大学 | Critical characteristic extraction method for flotation foam image analysis |
| CN101404722A (en) * | 2008-11-13 | 2009-04-08 | 中南大学 | Floatation foam image vision monitoring apparatus |
| US20090252413A1 (en) * | 2008-04-04 | 2009-10-08 | Microsoft Corporation | Image classification |
| CN103530621A (en) * | 2013-11-04 | 2014-01-22 | 中国矿业大学(北京) | Coal and rock image identification method based on back propagation (BP) neural network |
| CN103839057A (en) * | 2014-03-28 | 2014-06-04 | 中南大学 | Antimony floatation working condition recognition method and system |
| CN104331714A (en) * | 2014-11-28 | 2015-02-04 | 福州大学 | Image data extraction and neural network modeling-based platinum flotation grade estimation method |
| CN105260805A (en) * | 2015-11-16 | 2016-01-20 | 中南大学 | Antimony ore grade soft-measurement method based on selective fusion of heterogeneous classifier |
| CN106257498A (en) * | 2016-07-27 | 2016-12-28 | 中南大学 | Zinc flotation work condition state division methods based on isomery textural characteristics |
-
2018
- 2018-05-11 CN CN201810446656.4A patent/CN108931621B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0644278B2 (en) * | 1984-03-27 | 1994-06-08 | 株式会社ニレコ | Method and apparatus for automatic quantitative measurement of tissue by image analysis |
| CN101036904A (en) * | 2007-04-30 | 2007-09-19 | 中南大学 | Flotation froth image recognition device based on machine vision and the mine concentration grade forecast method |
| US20090252413A1 (en) * | 2008-04-04 | 2009-10-08 | Microsoft Corporation | Image classification |
| CN101334366A (en) * | 2008-07-18 | 2008-12-31 | 中南大学 | Flotation recovery rate prediction method based on image characteristic analysis |
| CN101334844A (en) * | 2008-07-18 | 2008-12-31 | 中南大学 | Critical characteristic extraction method for flotation foam image analysis |
| CN101404722A (en) * | 2008-11-13 | 2009-04-08 | 中南大学 | Floatation foam image vision monitoring apparatus |
| CN103530621A (en) * | 2013-11-04 | 2014-01-22 | 中国矿业大学(北京) | Coal and rock image identification method based on back propagation (BP) neural network |
| CN103839057A (en) * | 2014-03-28 | 2014-06-04 | 中南大学 | Antimony floatation working condition recognition method and system |
| CN104331714A (en) * | 2014-11-28 | 2015-02-04 | 福州大学 | Image data extraction and neural network modeling-based platinum flotation grade estimation method |
| CN105260805A (en) * | 2015-11-16 | 2016-01-20 | 中南大学 | Antimony ore grade soft-measurement method based on selective fusion of heterogeneous classifier |
| CN106257498A (en) * | 2016-07-27 | 2016-12-28 | 中南大学 | Zinc flotation work condition state division methods based on isomery textural characteristics |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110109446A (en) * | 2019-05-28 | 2019-08-09 | 中南大学 | A kind of zinc floatation process Fuzzy Fault Diagnosis based on time series feature |
| CN110109446B (en) * | 2019-05-28 | 2020-08-25 | 中南大学 | Zinc flotation process fuzzy fault diagnosis method based on time series characteristics |
| CN110728677A (en) * | 2019-07-22 | 2020-01-24 | 中南大学 | Texture roughness defining method based on sliding window algorithm |
| CN110766673A (en) * | 2019-07-22 | 2020-02-07 | 中南大学 | Texture roughness defining method based on Euclidean distance judgment |
| CN110728677B (en) * | 2019-07-22 | 2021-04-02 | 中南大学 | Texture roughness defining method based on sliding window algorithm |
| CN110766673B (en) * | 2019-07-22 | 2021-04-30 | 中南大学 | Texture roughness defining method based on Euclidean distance judgment |
| CN115049165A (en) * | 2022-08-15 | 2022-09-13 | 北矿机电科技有限责任公司 | Flotation concentrate grade prediction method, device and equipment based on deep learning |
| CN115049165B (en) * | 2022-08-15 | 2022-11-22 | 北矿机电科技有限责任公司 | Flotation concentrate grade prediction method, device and equipment based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108931621B (en) | 2020-10-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108647722A (en) | A kind of zinc ore grade flexible measurement method of Kernel-based methods size characteristic | |
| CN110210463B (en) | Precise ROI-fast R-CNN-based radar target image detection method | |
| CN108931621A (en) | A kind of zinc ore grade flexible measurement method of Kernel-based methods textural characteristics | |
| CN108830188B (en) | Vehicle detection method based on deep learning | |
| CN105741267B (en) | The multi-source image change detecting method of cluster guidance deep neural network classification | |
| CN109977808A (en) | A kind of wafer surface defects mode detection and analysis method | |
| CN110033002A (en) | Detection method of license plate based on multitask concatenated convolutional neural network | |
| CN109919934A (en) | A kind of liquid crystal display panel defect inspection method based on the study of multi-source domain depth migration | |
| CN109461141A (en) | A kind of workpiece starved detection method | |
| Liu et al. | Recognition of the operational statuses of reagent addition using dynamic bubble size distribution in copper flotation process | |
| CN108537102A (en) | High Resolution SAR image classification method based on sparse features and condition random field | |
| CN109242829A (en) | Liquid crystal display defect inspection method, system and device based on small sample deep learning | |
| CN103324937A (en) | Method and device for labeling targets | |
| CN111127499A (en) | Security inspection image cutter detection segmentation method based on semantic contour information | |
| Wang et al. | Multiscale feature fusion and semi-supervised temporal-spatial learning for performance monitoring in the flotation industrial process | |
| CN103745238B (en) | Pantograph identification method based on AdaBoost and active shape model | |
| CN111860459A (en) | Gramineous plant leaf stomata index measuring method based on microscopic image | |
| CN109508753A (en) | A kind of on-line prediction method of Mineral Floating Process index | |
| CN105976397A (en) | Target tracking method based on half nonnegative optimization integration learning | |
| Cao et al. | A new froth image classification method based on the MRMR-SSGMM hybrid model for recognition of reagent dosage condition in the coal flotation process | |
| CN106340007A (en) | Image processing-based automobile body paint film defect detection and identification method | |
| CN107886539A (en) | High class gear visible detection method under a kind of industrial scene | |
| Nakhaei et al. | Column flotation performance prediction: PCA, ANN and image analysis-based approaches | |
| Xu et al. | Multi-model soft measurement method of the froth layer thickness based on visual features | |
| CN104634265A (en) | Soft measurement method for thickness of mineral floating foam layer based on multivariate image feature fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |