CN107589093A - A kind of ature of coal on-line checking analysis method based on regression analysis - Google Patents
A kind of ature of coal on-line checking analysis method based on regression analysis Download PDFInfo
- Publication number
- CN107589093A CN107589093A CN201710807824.3A CN201710807824A CN107589093A CN 107589093 A CN107589093 A CN 107589093A CN 201710807824 A CN201710807824 A CN 201710807824A CN 107589093 A CN107589093 A CN 107589093A
- Authority
- CN
- China
- Prior art keywords
- coal
- module
- ature
- image
- data
- 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
Abstract
The invention belongs to coal property test field, discloses a kind of ature of coal on-line checking analysis method based on regression analysis, and the graphic feature that each region inside coal is automatically extracted using the algorithm with regress analysis method of machine learning carries out image characteristics extraction;Using machine learning using the characteristics of image extracted as according to ature of coal on-line checking result is provided, testing result is shown by display module;Specifically include:Coal scan image is obtained by the data acquisition equipment for being built-in near infrared from detecting module and laser acquisition module;Coal scan image is obtained using pretreatment module to be pre-processed, and coal scan image is divided into multiple blocks.The present invention carries out coal data information detection by near infrared from detecting module and laser acquisition module various ways, and detection data are accurate;The advantages that present invention is using into subnetwork nonlinear prediction, strong adaptability, zmodem, can both realize the full elemental analysis of ature of coal, can improve measurement accuracy again.
Description
Technical field
The invention belongs to coal property test field, more particularly to a kind of ature of coal on-line checking analysis side based on regression analysis
Method.
Background technology
The chemical examination carried out to understand the quality and combustion characteristics of coal with the method for physics and chemistry to coal sample and test work
Make.Coal analysis is carried out by national technical standard or special test technology, it be for the design about equipment and technical process and
Operation provides the basic work of foundation;, must be in time in order to carry out real-time monitoring to production process in the coal unit such as power plant
The specific composition of coal on belt conveyor is understood, so as to adjust the manufacturing parameter of correlation according to the change of coal constituent.Than
Such as in coal-burning power plant, coal cost account for the 80% of totle drilling cost, so understanding and studying shadow of the factors such as ature of coal to production efficiency
Sound is vital.This requires that constituent analysis can be carried out to the coal on conveyer belt in real time online.
Method based on data mining and machine learning has the potentiality for solving this problem.Have a variety of be based at present
The brain image analysis method of computerized algorithm.Conventional method has the VBM methods using pixel as fundamental analysis unit, with big
Each displacement structure of brain is the TBM methods of fundamental analysis unit, using the image high-level characteristic of manual extraction as fundamental analysis unit
FBM methods etc..Various statistical models or machine learning model can be used fundamental analysis unit, such as Bayesian analysis, recently
Adjacent algorithm, neutral net, deep learning etc.
But existing algorithm still can not effectively handle problems faced is needed in practical application at present.
VBM methods are based on image conversion and are analyzed, and this method is one of current most widely used method.But
It is that this method is very sensitive to parameter setting, therefore the data for being unfavorable for integrating a variety of different coals carry out model training with surveying
Examination.In practical application, alignment algorithm usually can not effectively distinguish the image difference caused by quality changes.Force to every
Individual pictorial element, which carries out alignment, can cause the loss of comparison information.
TBM methods base has the defects of similar with VBM.
In existing patent, using a kind of special image characteristic extracting method, characteristics of image is converted into one kind and is more conducive to
The data of Vector Machine learning method processing.The method of this patent by suitable for including dispersion tensor image it is various not
Same MRI.But image processing effect is not excellent.Analyzed for ature of coal on-line checking and combine excellent image procossing
Have much room for improvement.
In summary, the problem of prior art is present be:Existing detection and analysis mode is single, and prohibited data detection is true.
The content of the invention
The problem of existing for prior art, the invention provides a kind of ature of coal on-line checking point based on regression analysis
Analysis method.
The present invention is achieved in that a kind of ature of coal on-line checking analysis method based on regression analysis, described to be based on back
The ature of coal on-line checking analysis method of analysis is returned to automatically extract each region inside coal using the algorithm with regress analysis method of machine learning
Graphic feature carry out image characteristics extraction;Ature of coal is provided using machine learning as foundation using the characteristics of image extracted to examine online
Result is surveyed, testing result is shown by display module.
Further, the ature of coal on-line checking analysis method based on regression analysis includes:
Coal scanning figure is obtained by the data acquisition equipment for being built-in near infrared from detecting module and laser acquisition module
Picture;
Coal scan image is obtained using pretreatment module to be pre-processed, and coal scan image is divided into multiple blocks;
Using ature of coal image characteristics extraction module an image feature vector is extracted from each block;
It is by the high dimensional feature vector compression of each block with the unsupervised machine learning module built in data acquisition equipment
One dimensional numerical;The one dimensional numerical of each block is represented to be connected as a vector, the vector as whole coal scan image is retouched
State;
Supervised machine learning is used using the detection module built in data acquisition equipment, using coal scan vector to be defeated
Enter, carry out image prediction.
Further, described the step of extracting an image feature vector from each block, includes:
Different coal regions are tentatively snapped to standard coal region masterplate using affine transformation;
It is that can cover the multiple three-dimensional cuboids or spheroid of complete image by the picture breakdown for coal sector scanning of having alignd
Block, partly overlap between block;
The pixel intensity of each block is converted into characteristics of image description, each feature using image characteristics extraction algorithm
Description is expressed as a high dimension vector.
Further, as it is described with unsupervised machine learning be a dimension by the high dimensional feature vector compression of each block
The step of value, includes:
The characteristics of image clustering algorithm of each image block to be gathered for two classifications, a classification is similar with poor ature of coal,
Another classification is similar with good ature of coal;
To each image block, the image feature vector of the block is divided into described two classifications by one grader of training
A classification in cluster result;
Classification results are converted into a real number, the tile images characteristic vector is represented and is divided into the related spy of poor ature of coal
The probability of sign.
Further, the ature of coal on-line checking analysis method based on regression analysis also includes:
Step 1, the detection module carry out the data after image prediction and are sent to data acquisition module, data acquisition module
The analog electric signal that block obtains is converted to digital quantity signal, and is sent to data processing module;
The laser acquisition module is gone out and template to a certain frame video image Q collected using surf Algorithm for Solving
All characteristic point P={ p matched somebody with somebody1,p2,...,pn, wherein, piFor the characteristic point in image Q;
From whole matching characteristic point P={ p1,p2,...,pn4 most accurate matching characteristic point P of middle selection0={ pj1,
pj2,pj3,pj4},jk∈ { 1,2 ..., n }, k=1,2,3,4, record the image coordinate value (u of these characteristic pointsi,vi), i=
j1,j2,j3,j4, and wherein for world coordinates origin, to record the world coordinates of other characteristic points
Utilize the coordinate of world coordinate systemWith the pixel coordinate (u of its subpointi,vi) between relation
Formula calculates the outer parameter matrix H of camera, wherein, i=j1,j2,j3,j4;
Using the outer parameter matrix H calculated, using current video two field picture as in d engine three in graphics engine
The background of scene is tieed up, required position renders threedimensional model in the scene, realizes that real-time three-dimensional is superimposed;
Using the interactive function in graphics engine, the interaction between the dummy object of three-dimensional overlay and real world object is realized;
Step 2, data processing module obtain near infrared spectrum data and laser spectrum spectral line from data acquisition module, and
Row is input to operational module, connects correcting module by criterion module, correcting module is connected to be formed with operational module and is recycled back to again
The Nonlinear Mapping and laser spectrum the intensity of spectral line matrix to coal near infrared spectrum are completed in road jointly, and it is revised to obtain ature of coal
Parameter;
The data processing module is by reception signal data R (x), according to formula r (x)=sign (Re (R (x)))+j*sign
(Im (R (x))) obtains the result r (x) mapped out to reception signal reality imaginary part by sign bit, then by local training sequence data C (k),
Obtain mapping out training sequence data reality imaginary part by sign bit using formula c (k)=sign (Re (C (k)))+j*sign (Im (C (k)))
Result c (k), formula is utilized according to obtained r (x) and c (k)
Timing slip estimation function is generated, N=2* (NFFT+CP) represents the length of associated window and local sequence in formula, and x, which is represented, to be slided
The original position of associated window;
Step 3, wireless communication module is wirelessly connected, realize remote online connection and from data processing module
Obtain ature of coal parameter;
The timing slip estimation function F (x) of the wireless communication module, according to formula
Dynamic threshold is obtained, wherein G (m) represents the value of m moment dynamic thresholds,Represent the M started counting up from the m moment
The average value of individual timing slip estimation function value, mul represent a constant;Acquisition methods specifically include:
Training sequence is mapped by sign bit, and using result as local sequence, the data received flow into slip successively
In window, the data in sliding window are subjected to conjugation related operation by sign bit and local sequence, obtained on sliding window start bit
The timing slip estimation function value put;
The procedural representation of computing is:First according to the sign bit information of reception signal reality imaginary data, formula r (x) is utilized
=sign (Re (R (x)))+j*sign (Im (R (x))), map receiving data, wherein R (x) represents reception signal, Re
() represents to take the value of real part of complex data, and Im () represents to take the imaginary values of complex data, and sign () represents to take a data
Sign bit, if data be more than 0 output result be 1, be that -1, r (x) is that reception signal reality imaginary part is taken less than 0 output result
The result mapped out after symbol, there are four kinds of numerical value ± 1 ± j, then utilize formula c (k)=sign (Re (C (k)))+j*sign
(Im (C (k))) is mapped training sequence, and wherein C (k) represents local training sequence, and c (k) is to local sequence reality imaginary part
The complex result that data are gone out by sign bit information MAP, there are four kinds of numerical value ± 1 ± j;Finally according to formulaTiming slip estimation function is asked for, wherein, F (x)
Timing slip estimation function value is represented, N=2* (NFFT+CP) represents the length of associated window and local sequence;
The link stability and energy hybrid model of the method for repairing route of the wireless communication module:
Internet of Things topological structure regards the network model G=(V, E) of a non-directed graph as, and wherein V represents a group node, E tables
Show the side collection of one group of connecting node, P (u, v)={ P0,P1,P2,L,PnIt is all possible paths between node u and node v
Set, PiIt is node u and v possible path, selects egress u to node v optimal path,
The formula of link stability and residue energy of node is as follows:
Wherein, EisAnd Ei0For the dump energy and gross energy of node i, EthFor the energy threshold of node;
Link stability formula and residue energy of node formula change into the optimization formula of a totality, and the formula provides
Two important parameter (w1And w2), shown in its expression formula such as formula (4):
Wherein w1And w2The coefficient of setting between node energy and link stationary value, w1+w2=1;
The maximum of the target summation is taken, is represented with formula below (5):
MRFact(Pi)=max { RFact (P1),RFact(P2),L RFact(Pn)} (5)
Node calculates the stationary value of outgoing link according to formula (1) and formula (2) respectively when receiving data packet information
With the dump energy of node, optimal path then is chosen using formula (5), to complete the selected of route.
Further, the near-infrared generation module is made up of near-infrared light source, optical splitter, and generation wavelength is 1000-
2500nm near infrared light.
Further, the laser acquisition module detection method:
First, one group of coal sample is as calibration sample known to selection each element mass concentration, using installed in defeated coal
Laser induced plasma spectroscopic system above belt detects to calibration sample, obtains the spectrum spectrum of this group of calibration sample
Line, laser induced plasma characteristic spectrum the intensity of spectral line of various elements in every kind of calibration sample is obtained, form the intensity of spectral line square
Battle array E0, the structure of E0 matrixes are as follows:
Wherein, Ii λjRepresent i-th kind of sample the intensity of spectral line, i=1,2, L, n corresponding at the wavelength X j;J=1,2, L, m;
Then, from matrix E0Middle extraction principal component, obtains matrix E0Covariance matrix, ask for the feature of covariance matrix
Value, characteristic value are followed successively by A from big to small1, A2, L, Ah;Eigenvalue of maximum λ1Corresponding characteristic vector is the first main shaft a1, second
Eigenvalue λ2Corresponding characteristic vector is the second main shaft a2, thus try to achieve first, second principal component t1、t2;
t1=E0a1;t2=E0a2;The like can be in the hope of h-th of principal component th;
Finally, the number of principal components of acquisition is handled according to logical data processing module and all kinds of calibration samples is established respectively
Composition network model, the input layer into subnetwork are characterized the intensity of spectral line, and input layer number is characterized the number of spectral line;
Output layer is each element concentration, and output layer interstitial content is the element species number for needing to determine.
Another object of the present invention is to provide a kind of ature of coal on-line checking analysis system based on regression analysis.
Advantages of the present invention and good effect are:Entered by near infrared from detecting module and laser acquisition module various ways
Row coal data information detection, detection data are accurate;Bulk information contained in spectrum is taken full advantage of simultaneously to all kinds of coals point
Subnetwork model is not created as, can eliminate error caused by the variety classes of coal, using into subnetwork nonlinear prediction, suitable
The advantages that Ying Xingqiang, zmodem, the full elemental analysis of ature of coal can be both realized, measurement accuracy can be improved again.Laser acquisition
Module realizes the target object real-time tracking of no specific markers, real-time three-dimensional superposition to the data detected, to every frame video
Image calculates the three-dimensional coordinate information of target in real time, improves the accuracy of detection data;Data processing module is inclined into timing
Estimation function is moved, the acquisition of the timing slip estimation function of wireless communication module improves the stability of signal transmission, helps to carry
The accuracy rate of high on-line analysis.
In the imaging detection method of the present invention, characteristics of image automatic Extraction Algorithm is automatically extracted using machine learning algorithm
The graphic feature in each region of coal;Using machine learning prediction result is provided by foundation of characteristics of image.The view data of acquisition
Accurately, the processing for postorder provides guarantee.
Brief description of the drawings
Fig. 1 is the ature of coal on-line checking analysis method flow chart based on regression analysis that the present invention implements to provide;
Fig. 2 is the ature of coal on-line checking analysis system structural representation based on regression analysis that the present invention implements to provide;
In figure:1st, near infrared from detecting module;2nd, laser acquisition module;3rd, data acquisition module;4th, data processing module;
5th, wireless communication module;6th, pretreatment module;7th, ature of coal image characteristics extraction module;8th, unsupervised machine learning module;9th, examine
Survey module.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to embodiments, to this hair
It is bright to be further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, not
For limiting the present invention.
The application principle of the present invention is explained in detail below in conjunction with the accompanying drawings.
Ature of coal on-line checking analysis method provided in an embodiment of the present invention based on regression analysis, it is described based on recurrence point
The ature of coal on-line checking analysis method of analysis automatically extracts the figure in each region inside coal using the algorithm with regress analysis method of machine learning
Shape feature carries out image characteristics extraction;Using machine learning ature of coal on-line checking knot is provided using the characteristics of image extracted as foundation
Fruit, testing result is shown by display module.
As shown in figure 1, the ature of coal on-line checking analysis method provided in an embodiment of the present invention based on regression analysis includes:
S101:Data acquisition equipment by being built-in near infrared from detecting module and laser acquisition module obtains coal and swept
Trace designs picture.
S102:Coal scan image is obtained using pretreatment module to be pre-processed, and coal scan image is divided into multiple
Block.
S103:Using ature of coal image characteristics extraction module an image feature vector is extracted from each block.
S104:It is with the unsupervised machine learning module built in data acquisition equipment that the high dimensional feature of each block is vectorial
Boil down to one dimensional numerical;The one dimensional numerical of each block is represented to be connected as a vector, as whole coal scan image to
Amount description.
S105:Supervised machine learning is used using the detection module built in data acquisition equipment, with coal scan vector
For input, image prediction is carried out.
S106:Data after the detection module progress image prediction are sent to data acquisition module, data acquisition module
The analog electric signal of acquisition is converted to digital quantity signal, and is sent to data processing module.
S107:Data processing module obtains near infrared spectrum data and laser spectrum spectral line from data acquisition module, parallel
Operational module is input to, correcting module is connected by criterion module, correcting module is connected to form circulation loop with operational module again
The common Nonlinear Mapping and laser spectrum the intensity of spectral line matrix completed to coal near infrared spectrum, after obtaining ature of coal amendment
Parameter.
S108:Wireless communication module is wirelessly connected, realizes that remote online is connected and obtained from data processing module
Take ature of coal parameter.
Described the step of extracting an image feature vector from each block, includes:
Different coal regions are tentatively snapped to standard coal region masterplate using affine transformation;
It is that can cover the multiple three-dimensional cuboids or spheroid of complete image by the picture breakdown for coal sector scanning of having alignd
Block, partly overlap between block;
The pixel intensity of each block is converted into characteristics of image description, each feature using image characteristics extraction algorithm
Description is expressed as a high dimension vector.
As it is described with unsupervised machine learning be one dimensional numerical by the high dimensional feature vector compression of each block the step of
Including:
The characteristics of image clustering algorithm of each image block to be gathered for two classifications, a classification is similar with poor ature of coal,
Another classification is similar with good ature of coal;
To each image block, the image feature vector of the block is divided into described two classifications by one grader of training
A classification in cluster result;
Classification results are converted into a real number, the tile images characteristic vector is represented and is divided into the related spy of poor ature of coal
The probability of sign.
The laser acquisition module is gone out and template to a certain frame video image Q collected using surf Algorithm for Solving
All characteristic point P={ p matched somebody with somebody1,p2,...,pn, wherein, piFor the characteristic point in image Q;
From whole matching characteristic point P={ p1,p2,...,pn4 most accurate matching characteristic point P of middle selection0={ pj1,pj2,pj3,
pj4},jk∈ { 1,2 ..., n }, k=1,2,3,4, record the image coordinate value (u of these characteristic pointsi,vi), i=j1,j2,j3,j4, and with it
In a little be world coordinates origin, record the world coordinates of other characteristic points
Utilize the coordinate of world coordinate systemWith the pixel coordinate (u of its subpointi,vi) between relation
Formula calculates the outer parameter matrix H of camera, wherein, i=j1,j2,j3,j4;
Using the outer parameter matrix H calculated, using current video two field picture as in d engine three in graphics engine
The background of scene is tieed up, required position renders threedimensional model in the scene, realizes that real-time three-dimensional is superimposed;
Using the interactive function in graphics engine, the interaction between the dummy object of three-dimensional overlay and real world object is realized;
The data processing module is by reception signal data R (x), according to formula r (x)=sign (Re (R (x)))+j*sign (Im
(R (x))) the result r (x) mapped out to reception signal reality imaginary part by sign bit is obtained, then by local training sequence data C (k), profit
Obtain mapping out training sequence data reality imaginary part by sign bit with formula c (k)=sign (Re (C (k)))+j*sign (Im (C (k)))
Result c (k), formula is utilized according to obtained r (x) and c (k)
Timing slip estimation function is generated, N=2* (NFFT+CP) represents the length of associated window and local sequence in formula, and x, which is represented, to be slided
The original position of associated window;
The timing slip estimation function F (x) of the wireless communication module, according to formula
Dynamic threshold is obtained, wherein G (m) represents the value of m moment dynamic thresholds,Represent the M started counting up from the m moment
The average value of individual timing slip estimation function value, mul represent a constant;Acquisition methods specifically include:
Training sequence is mapped by sign bit, and using result as local sequence, the data received flow into slip successively
In window, the data in sliding window are subjected to conjugation related operation by sign bit and local sequence, obtained on sliding window start bit
The timing slip estimation function value put;
The procedural representation of computing is:First according to the sign bit information of reception signal reality imaginary data, formula r (x) is utilized
=sign (Re (R (x)))+j*sign (Im (R (x))), map receiving data, wherein R (x) represents reception signal, Re
() represents to take the value of real part of complex data, and Im () represents to take the imaginary values of complex data, and sign () represents to take a data
Sign bit, if data be more than 0 output result be 1, be that -1, r (x) is that reception signal reality imaginary part is taken less than 0 output result
The result mapped out after symbol, there are four kinds of numerical value ± 1 ± j, then utilize formula c (k)=sign (Re (C (k)))+j*sign
(Im (C (k))) is mapped training sequence, and wherein C (k) represents local training sequence, and c (k) is to local sequence reality imaginary part
The complex result that data are gone out by sign bit information MAP, there are four kinds of numerical value ± 1 ± j;Finally according to formulaTiming slip estimation function is asked for, wherein, F (x)
Timing slip estimation function value is represented, N=2* (NFFT+CP) represents the length of associated window and local sequence;
The link stability and energy hybrid model of the method for repairing route of the wireless communication module:
Internet of Things topological structure regards the network model G=(V, E) of a non-directed graph as, and wherein V represents a group node, E tables
Show the side collection of one group of connecting node, P (u, v)={ P0,P1,P2,L,PnIt is all possible paths between node u and node v
Set, PiIt is node u and v possible path, selects egress u to node v optimal path,
The formula of link stability and residue energy of node is as follows:
Wherein, EisAnd Ei0For the dump energy and gross energy of node i, EthFor the energy threshold of node;
Link stability formula and residue energy of node formula change into the optimization formula of a totality, and the formula provides
Two important parameter (w1And w2), shown in its expression formula such as formula (4):
Wherein w1And w2The coefficient of setting between node energy and link stationary value, w1+w2=1;
The maximum of the target summation is taken, is represented with formula below (5):
MRFact(Pi)=max { RFact (P1),RFact(P2),L RFact(Pn)} (5)
Node calculates the stationary value of outgoing link according to formula (1) and formula (2) respectively when receiving data packet information
With the dump energy of node, optimal path then is chosen using formula (5), to complete the selected of route.
Laser acquisition module detection method:
First, one group of coal sample is as calibration sample known to selection each element mass concentration, using installed in defeated coal
Laser induced plasma spectroscopic system above belt detects to calibration sample, obtains the spectrum spectrum of this group of calibration sample
Line, that is, laser induced plasma characteristic spectrum the intensity of spectral line of various elements in every kind of calibration sample is obtained, it is strong to form spectral line
The structure for spending matrix E0, E0 matrix is as follows:
Wherein, Ii λjRepresent i-th kind of sample the intensity of spectral line, i=1,2, L, n corresponding at the wavelength X j;J=1,2, L, m.
Then, from matrix E0Middle extraction principal component, obtains matrix E0Covariance matrix, ask for the feature of covariance matrix
Value, characteristic value are followed successively by A from big to small1, A2, L, Ah;Eigenvalue of maximum λ1Corresponding characteristic vector is the first main shaft a1, second
Eigenvalue λ2Corresponding characteristic vector is the second main shaft a2, thus try to achieve first, second principal component t1、t2,
t1=E0a1;t2=E0a2;The like can be in the hope of h-th of principal component th。
Finally, the number of principal components of acquisition is handled according to logical data processing module and all kinds of calibration samples is established respectively
Composition network model, the input layer into subnetwork are characterized the intensity of spectral line, and input layer number is characterized the number of spectral line;
Output layer is each element concentration, and output layer interstitial content is the element species number for needing to determine.
As shown in Fig. 2 the ature of coal on-line checking analysis system provided in an embodiment of the present invention based on regression analysis includes:
Near infrared from detecting module 1, laser acquisition module 2, data acquisition module 3, data processing module 4, wireless communication module 5, pre- place
Manage module 6, ature of coal image characteristics extraction module 7, unsupervised machine learning module 8, detection module 9.
Near infrared from detecting module 1 connects data acquisition module 3 by circuit line respectively with laser acquisition module 2;At data
Managing module 4 includes operational module, criterion module, correcting module, and is interconnected by circuit line;Data acquisition module 3 passes through
Operational module in circuit line connection data processing module 4;Criterion module in data processing module 4 is connected by circuit line
Wireless communication module 5.
Coal scanning figure is obtained by the data acquisition equipment for being built-in near infrared from detecting module and laser acquisition module
Picture.
Coal scan image is obtained using pretreatment module 6 to be pre-processed, and coal scan image is divided into multiple areas
Block.
Using ature of coal image characteristics extraction module 7 image feature vector is extracted from each block.
With the unsupervised machine learning module 8 built in data acquisition equipment by the high dimensional feature vector compression of each block
For one dimensional numerical;The one dimensional numerical of each block is represented to be connected as a vector, the vector as whole coal scan image is retouched
State.
Supervised machine learning is used using the detection module 9 built in data acquisition equipment, using coal scan vector to be defeated
Enter, carry out image prediction.
Near infrared from detecting module 1, it is connected with data acquisition module 3, for passing through the near-infrared photosensitive tube on coal
Obtain detection data.
Laser acquisition module 2, it is connected with data acquisition module 3, for being lured by the laser above coal conveyer belt
Lead plasma light spectra system to detect calibration sample, obtain the laser spectrum spectral line of this group of calibration sample.
Data acquisition module 3, it is connected, is used for near infrared from detecting module 1, laser acquisition module 2, data processing module 4
The analog electric signal that nearly infrared detection module 1 and laser acquisition module 2 obtain is converted to digital quantity signal, and is sent to
Data processing module 4.
Data processing module 4, it is connected with data acquisition module 3, wireless communication module 5, for from data acquisition module 3
Near infrared spectrum data and laser spectrum spectral line are obtained, is input to operational module parallel, amendment mould is connected by criterion module
Block, correcting module be connected again with operational module to be formed circulation loop complete jointly to the Nonlinear Mapping of coal near infrared spectrum and
Laser spectrum the intensity of spectral line matrix, so as to obtain the revised parameter of ature of coal, and production parameter is passed through into the nothing of wireless communication module 5
Line mode transfers out.
Wireless communication module 5, it is connected, is connected for remote online and from data processing module 4 with data processing module 4
Obtain ature of coal parameter.
Near-infrared occurs module 1 and is made up of near-infrared light source, optical splitter, and it is the near of 1000-2500nm that can produce wavelength
Infrared light.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement made within refreshing and principle etc., should be included in the scope of the protection.
Claims (8)
- A kind of 1. ature of coal on-line checking analysis method based on regression analysis, it is characterised in that the coal based on regression analysis Matter on-line checking analysis method automatically extracts the graphic feature in each region inside coal using the algorithm with regress analysis method of machine learning Carry out image characteristics extraction;Using machine learning using the characteristics of image extracted as according to ature of coal on-line checking result is provided, will examine Result is surveyed to be shown by display module.
- 2. the ature of coal on-line checking analysis method based on regression analysis as claimed in claim 1, it is characterised in that described to be based on The ature of coal on-line checking analysis method of regression analysis includes:Coal scan image is obtained by the data acquisition equipment for being built-in near infrared from detecting module and laser acquisition module;Coal scan image is obtained using pretreatment module to be pre-processed, and coal scan image is divided into multiple blocks;Using ature of coal image characteristics extraction module an image feature vector is extracted from each block;With the unsupervised machine learning module built in data acquisition equipment by the high dimensional feature vector compression of each block to be one-dimensional Numerical value;The one dimensional numerical of each block is represented to be connected as a vector, the vector description as whole coal scan image;Supervised machine learning is used using the detection module built in data acquisition equipment, using coal scan vector as input, is entered Row image prediction.
- 3. the ature of coal on-line checking analysis method based on regression analysis as claimed in claim 2, it is characterised in that described from every The step of individual one image feature vector of block extraction, includes:Different coal regions are tentatively snapped to standard coal region masterplate using affine transformation;It is that can cover the multiple three-dimensional cuboids or spheroid block of complete image by the picture breakdown for coal sector scanning of having alignd, Partly overlapped between block;The pixel intensity of each block is converted into characteristics of image description, each feature description using image characteristics extraction algorithm It is expressed as a high dimension vector.
- 4. the ature of coal on-line checking analysis method based on regression analysis as claimed in claim 2, it is characterised in that as described The step of with unsupervised machine learning by the high dimensional feature vector compression of each block being one dimensional numerical, includes:The characteristics of image clustering algorithm of each image block is gathered for two classifications, a classification is similar with poor ature of coal, another Individual classification is similar with good ature of coal;To each image block, the image feature vector of the block is divided into described two categorical clusters knots by one grader of training A classification in fruit;Classification results are converted into a real number, the tile images characteristic vector is represented and is divided into the general of poor ature of coal correlated characteristic Rate.
- 5. the ature of coal on-line checking analysis method based on regression analysis as claimed in claim 2, it is characterised in that described to be based on The ature of coal on-line checking analysis method of regression analysis also includes:Step 1, the detection module carry out the data after image prediction and are sent to data acquisition module, and data acquisition module obtains The analog electric signal taken is converted to digital quantity signal, and is sent to data processing module;The laser acquisition module is gone out and template matches to a certain frame video image Q collected using surf Algorithm for Solving All characteristic point P={ p1,p2,...,pn, wherein, piFor the characteristic point in image Q;From whole matching characteristic point P={ p1,p2,...,pn4 most accurate matching characteristic point P of middle selection0={ pj1,pj2,pj3,pj4}, jk∈ { 1,2 ..., n }, k=1,2,3,4, record the image coordinate value (u of these characteristic pointsi,vi), i=j1,j2,j3,j4, and with wherein It is some world coordinates origin, records the world coordinates of other characteristic pointsUtilize the coordinate of world coordinate systemWith the pixel coordinate (u of its subpointi,vi) between relational expression calculate The outer parameter matrix H of camera, wherein, i=j1,j2,j3,j4;Using the outer parameter matrix H calculated, using current video two field picture as three dimensional field in d engine in graphics engine The background of scape, in the scene required position render threedimensional model, realize real-time three-dimensional be superimposed;Using the interactive function in graphics engine, the interaction between the dummy object of three-dimensional overlay and real world object is realized;Step 2, data processing module obtains near infrared spectrum data and laser spectrum spectral line from data acquisition module, parallel defeated Enter to operational module, correcting module is connected by criterion module, correcting module is connected with operational module again to be formed circulation loop and be total to With the Nonlinear Mapping and laser spectrum the intensity of spectral line matrix completed to coal near infrared spectrum, the revised parameter of ature of coal is obtained;The data processing module is by reception signal data R (x), according to formula r (x)=sign (Re (R (x)))+j*sign (Im (R (x) the result r (x) mapped out to reception signal reality imaginary part by sign bit)) is obtained, then by local training sequence data C (k), utilizes public affairs Formula c (k)=sign (Re (C (k)))+j*sign (Im (C (k))) obtains the knot mapped out to training sequence data reality imaginary part by sign bit Fruit c (k), formula is utilized according to obtained r (x) and c (k) Timing slip estimation function is generated, N=2* (NFFT+CP) represents the length of associated window and local sequence in formula, and x, which is represented, slides phase Close the original position of window;Step 3, wireless communication module is wirelessly connected, realize remote online connection and obtained from data processing module Ature of coal parameter;The timing slip estimation function F (x) of the wireless communication module, according to formulaObtain Dynamic threshold, wherein G (m) represent the value of m moment dynamic thresholds,Represent to start counting up from the m moment M determine The average value of hour offset estimation function value, mul represent a constant;Acquisition methods specifically include:Training sequence is mapped by sign bit, and using result as local sequence, the data received are flowed into sliding window successively, Data in sliding window are subjected to conjugation related operation by sign bit and local sequence, obtain determining on sliding window original position Hour offset estimation function value;The procedural representation of computing is:First according to the sign bit information of reception signal reality imaginary data, formula r (x)=sign (Re are utilized (R (x)))+j*sign (Im (R (x))), map receiving data, wherein R (x) represents reception signal, and Re () represents to take plural number According to value of real part, Im () represents to take the imaginary values of complex data, and sign () represents to take the sign bit of a data, if data are more than 0 output result is 1, is that -1, r (x) is to take the result mapped out after symbol to reception signal reality imaginary part less than 0 output result, there is four kinds Numerical value ± 1 ± j, then training sequence is mapped using formula c (k)=sign (Re (C (k)))+j*sign (Im (C (k))), its Middle C (k) represents local training sequence, and c (k) is the complex result gone out to local sequence reality imaginary data by sign bit information MAP, is had Four kinds of numerical value ± 1 ± j;Finally according to formulaAsk for Timing slip estimation function, wherein, F (x) represents timing slip estimation function value, and N=2* (NFFT+CP) represents associated window and this The length of ground sequence;The link stability and energy hybrid model of the method for repairing route of the wireless communication module:Internet of Things topological structure regards the network model G=(V, E) of a non-directed graph as, and wherein V represents a group node, and E represents one The side collection of group connecting node, P (u, v)={ P0,P1,P2,L,PnBe all possible paths between node u and node v set, Pi It is node u and v possible path, selects egress u to node v optimal path,The formula of link stability and residue energy of node is as follows:Wherein, EisAnd Ei0For the dump energy and gross energy of node i, EthFor the energy threshold of node;Link stability formula and residue energy of node formula change into the optimization formula of a totality, and the formula provides two weights Want parameter (w1And w2), shown in its expression formula such as formula (4):Wherein w1And w2The coefficient of setting between node energy and link stationary value, w1+w2=1;The maximum of the target summation is taken, is represented with formula below (5):MRFact(Pi)=max { RFact (P1),RFact(P2),L RFact(Pn)} (5)Node calculates the stationary value and node of outgoing link according to formula (1) and formula (2) respectively when receiving data packet information Dump energy, then optimal path is chosen using formula (5), to complete the selected of route.
- 6. the ature of coal on-line checking analysis method based on regression analysis as claimed in claim 5, it is characterised in that described near red Outer generation module is made up of near-infrared light source, optical splitter, produces the near infrared light that wavelength is 1000-2500nm.
- 7. the ature of coal on-line checking analysis method based on regression analysis as claimed in claim 5, it is characterised in that the laser Detecting module detection method:First, one group of coal sample is as calibration sample known to selection each element mass concentration, using installed in coal conveyer belt The laser induced plasma spectroscopic system of top detects to calibration sample, obtains the optic spectrum line of this group of calibration sample, obtains Laser induced plasma characteristic spectrum the intensity of spectral line of various elements into every kind of calibration sample, the intensity of spectral line matrix E0 is formed, The structure of E0 matrixes is as follows:Wherein, Ii λjRepresent i-th kind of sample the intensity of spectral line, i=1,2, L, n corresponding at the wavelength X j;J=1,2, L, m;Then, from matrix E0Middle extraction principal component, obtains matrix E0Covariance matrix, ask for the characteristic value of covariance matrix, it is special Value indicative is followed successively by A from big to small1, A2, L, Ah;Eigenvalue of maximum λ1Corresponding characteristic vector is the first main shaft a1, Second Eigenvalue λ2Corresponding characteristic vector is the second main shaft a2, thus try to achieve first, second principal component t1、t2;t1=E0a1;t2=E0a2;The like can be in the hope of h-th of principal component th;Finally, the number of principal components of acquisition is handled according to logical data processing module and subnetting is created as respectively to all kinds of calibration samples Network model, the input layer into subnetwork are characterized the intensity of spectral line, and input layer number is characterized the number of spectral line;Output layer is Each element concentration, the element species number that output layer interstitial content determines for needs.
- A kind of 8. coal based on regression analysis of the ature of coal on-line checking analysis method based on regression analysis as claimed in claim 1 Matter on-line checking analysis system.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710807824.3A CN107589093A (en) | 2017-09-08 | 2017-09-08 | A kind of ature of coal on-line checking analysis method based on regression analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710807824.3A CN107589093A (en) | 2017-09-08 | 2017-09-08 | A kind of ature of coal on-line checking analysis method based on regression analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107589093A true CN107589093A (en) | 2018-01-16 |
Family
ID=61051095
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710807824.3A Pending CN107589093A (en) | 2017-09-08 | 2017-09-08 | A kind of ature of coal on-line checking analysis method based on regression analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107589093A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108298139A (en) * | 2018-01-29 | 2018-07-20 | 曹凯 | A kind of carbon-based bio-bacterial manure packing processing unit |
CN108519330A (en) * | 2018-04-03 | 2018-09-11 | 北京英思创科技有限公司 | Coal elemental composition real-time and precise detecting system and method |
CN108764600A (en) * | 2018-04-03 | 2018-11-06 | 北京英思创科技有限公司 | A kind of method of artificial intelligence deep learning |
CN110927193A (en) * | 2019-10-22 | 2020-03-27 | 北京浩然科诺科技有限公司 | Coal quality online detection and analysis system and method based on deep learning |
CN115508500A (en) * | 2022-10-27 | 2022-12-23 | 江苏未来智慧信息科技有限公司 | Visual layered detection method and system for carbon emission measurement |
CN116297272A (en) * | 2023-05-22 | 2023-06-23 | 北京易兴元石化科技有限公司 | On-line coal quality analysis system and method |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1912588A (en) * | 2006-08-23 | 2007-02-14 | 太原市海通自动化技术有限公司 | Coal on-line analyse equipment based on laser induced spectral and nerve network technology |
CN102004088A (en) * | 2010-11-09 | 2011-04-06 | 清华大学 | Method for measuring coal property on line based on neural network |
CN102364497A (en) * | 2011-05-06 | 2012-02-29 | 北京师范大学 | Image semantic extraction method applied in electronic guidance system |
CN102831401A (en) * | 2012-08-03 | 2012-12-19 | 樊晓东 | Method and system for tracking, three-dimensionally superposing and interacting target object without special mark |
CN103728268A (en) * | 2013-12-17 | 2014-04-16 | 沈阳工程学院 | Near-infrared on-line coal quality measurement device |
CN103986648A (en) * | 2014-05-06 | 2014-08-13 | 安徽理工大学 | Internet-of-Things route repairing method based on link stability and energy sensing |
CN104125190A (en) * | 2014-08-18 | 2014-10-29 | 西安电子科技大学 | OFDM (orthogonal frequency division multiplexing) system symbol timing synchronization realizing method suitable for low-signal-to-noise-ratio channel environments |
US20170256057A1 (en) * | 2016-03-04 | 2017-09-07 | Disney Enterprises, Inc. | Systems and Methods for Re-Identifying Objects in Images |
-
2017
- 2017-09-08 CN CN201710807824.3A patent/CN107589093A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1912588A (en) * | 2006-08-23 | 2007-02-14 | 太原市海通自动化技术有限公司 | Coal on-line analyse equipment based on laser induced spectral and nerve network technology |
CN102004088A (en) * | 2010-11-09 | 2011-04-06 | 清华大学 | Method for measuring coal property on line based on neural network |
CN102364497A (en) * | 2011-05-06 | 2012-02-29 | 北京师范大学 | Image semantic extraction method applied in electronic guidance system |
CN102831401A (en) * | 2012-08-03 | 2012-12-19 | 樊晓东 | Method and system for tracking, three-dimensionally superposing and interacting target object without special mark |
CN103728268A (en) * | 2013-12-17 | 2014-04-16 | 沈阳工程学院 | Near-infrared on-line coal quality measurement device |
CN103986648A (en) * | 2014-05-06 | 2014-08-13 | 安徽理工大学 | Internet-of-Things route repairing method based on link stability and energy sensing |
CN104125190A (en) * | 2014-08-18 | 2014-10-29 | 西安电子科技大学 | OFDM (orthogonal frequency division multiplexing) system symbol timing synchronization realizing method suitable for low-signal-to-noise-ratio channel environments |
US20170256057A1 (en) * | 2016-03-04 | 2017-09-07 | Disney Enterprises, Inc. | Systems and Methods for Re-Identifying Objects in Images |
Non-Patent Citations (1)
Title |
---|
LIRONG TAN ET AL.: "A semi-supervised Support Vector Machine model for predicting the language outcomes following cochlear implantation based on pre-implant brain fMRI imaging", 《BRAIN AND BEHAVIOR》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108298139A (en) * | 2018-01-29 | 2018-07-20 | 曹凯 | A kind of carbon-based bio-bacterial manure packing processing unit |
CN108519330A (en) * | 2018-04-03 | 2018-09-11 | 北京英思创科技有限公司 | Coal elemental composition real-time and precise detecting system and method |
CN108764600A (en) * | 2018-04-03 | 2018-11-06 | 北京英思创科技有限公司 | A kind of method of artificial intelligence deep learning |
CN108764600B (en) * | 2018-04-03 | 2022-05-10 | 北京英思创科技有限公司 | Artificial intelligence deep learning method |
CN110927193A (en) * | 2019-10-22 | 2020-03-27 | 北京浩然科诺科技有限公司 | Coal quality online detection and analysis system and method based on deep learning |
CN110927193B (en) * | 2019-10-22 | 2022-03-25 | 北京浩然科诺科技有限公司 | Coal quality online detection and analysis system and method based on deep learning |
CN115508500A (en) * | 2022-10-27 | 2022-12-23 | 江苏未来智慧信息科技有限公司 | Visual layered detection method and system for carbon emission measurement |
CN115508500B (en) * | 2022-10-27 | 2024-01-05 | 江苏未来智慧信息科技有限公司 | Visual layering detection method and system for carbon emission metering |
CN116297272A (en) * | 2023-05-22 | 2023-06-23 | 北京易兴元石化科技有限公司 | On-line coal quality analysis system and method |
CN116297272B (en) * | 2023-05-22 | 2023-08-18 | 北京易兴元石化科技有限公司 | On-line coal quality analysis system and method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107589093A (en) | A kind of ature of coal on-line checking analysis method based on regression analysis | |
Koukoulas et al. | Mapping individual tree location, height and species in broadleaved deciduous forest using airborne LIDAR and multi‐spectral remotely sensed data | |
CN104700404B (en) | A kind of fruit positioning identifying method | |
CN106356757A (en) | Method for inspecting electric power lines by aid of unmanned aerial vehicle on basis of human vision characteristics | |
CN109544501A (en) | A kind of transmission facility defect inspection method based on unmanned plane multi-source image characteristic matching | |
CN102308306A (en) | A constraint generator for use in image segregation | |
CN111639587B (en) | Hyperspectral image classification method based on multi-scale spectrum space convolution neural network | |
CN105469098A (en) | Precise LINDAR data ground object classification method based on adaptive characteristic weight synthesis | |
CN108256588A (en) | A kind of several picture identification feature extracting method and system | |
Xie et al. | Trainable spectral difference learning with spatial starting for hyperspectral image denoising | |
Song et al. | Deep domain adaptation based multi-spectral salient object detection | |
Yuan et al. | Multi-attention DeepEMD for few-shot learning in remote sensing | |
Chu et al. | MFGAN: Multi feature guided aggregation network for remote sensing image | |
Wang et al. | Fusion of hyperspectral and lidar data based on dual-branch convolutional neural network | |
Acito et al. | CWV-Net: A deep neural network for atmospheric column water vapor retrieval from hyperspectral VNIR data | |
Shang et al. | Spatiotemporal reflectance fusion using a generative adversarial network | |
Song et al. | Remote sensing image spatiotemporal fusion via a generative adversarial network with one prior image pair | |
Kowaleczko et al. | A real-world benchmark for sentinel-2 multi-image super-resolution | |
CN117011759A (en) | Method and system for analyzing multi-element geological information of surrounding rock of tunnel face by drilling and blasting method | |
CN116824488A (en) | Target detection method based on transfer learning | |
CN109993071A (en) | The method and system of discoloration forest automatic identification and investigation based on remote sensing image | |
Costa et al. | Segmentation of optical remote sensing images for detecting homogeneous regions in space and time. | |
Xue et al. | Self-supervised feature learning and few-shot land cover classification for cross-modal remote sensing images | |
Nyberg | Exploring the Capabilities of Generative Adversarial Networks in Remote Sensing Applications | |
Zhu et al. | Method for generating SPOT natural-colour composite images based on spectrum machine learning |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180116 |
|
RJ01 | Rejection of invention patent application after publication |