CN105243401A - Coal rock recognition method based on coal structure element study - Google Patents

Abstract

The invention discloses a coal rock recognition method based on coal rock structure element study. The method is used for obtaining feature information of coal rock images through coal rock structure element study to recognize coal rocks. The method is composed of an image pre-processing module, a training process module and a recognition process module. The pre-processing module is used for simply pre-processing collected coal rock images to obtain a training sample set. The training module is used for training structure elements by use of the training sample set and extracting image feature information. The recognition module is used for extracting the feature information of an unknown coal rock object image according to the structure elements obtained by training and inputting the feature information in a classifier for classification and recognition. The images of coal rocks under different illuminance and different viewpoints are used as training samples in the method, and the method is higher in robustness on image noise, small in influence generated by illuminance and imaging viewpoint changes, high in recognition rate and good in stability.

Description

Based on the Coal-rock identification method of lithostructure meta learning

Technical field

The present invention relates to a kind of Coal-rock identification method based on lithostructure meta learning, belong to coal and rock identify technical field.

Background technology

Namely coal and rock identify automatically identifies coal or rock by a kind of method.In coal production process, coal and rock identify technology can be widely used in cylinder coal mining, driving, top coal caving, raw coal select the production links such as spoil, for minimizing getting working face operating personnel, alleviate labor strength, improve operating environment, to realize mine safety High-efficient Production significant.

Existing multiple Coal-rock identification method, as natural Gamma ray probe method, radar detection system, stress pick method, infrared detecting method, active power monitoring method, shock detection method, sound detection method, dust detection method, memory cut method etc., but there is following problem in these methods: 1. need to install various kinds of sensors obtaining information on existing additional, cause apparatus structure complicated, cost is high.2. the equipment such as coal mining machine roller, development machine in process of production stressed complexity, vibration is violent, serious wear, dust large, sensor deployment is more difficult, and easily cause mechanical component, sensor and electric wiring to be damaged, device reliability is poor.3. for dissimilar plant equipment, there is larger difference in the selection of the best type of sensor and picking up signal point, needs to carry out personalized customization, the bad adaptability of system.

For solving the problem, image technique more and more comes into one's own, and have developed some Coal-rock identification method based on image technique, but existing method needs the combination manually meticulously choosing characteristics of image or characteristics of image, this often needs very large effort and trial, but income approach does not always have robustness to changing the view data change caused because of image-forming condition, causes and is short of to some extent in identification stability and discrimination.

Need a kind of Coal-rock identification method solving or at least improve one or more problems intrinsic in prior art, to improve coal and rock identify rate and to identify stability.

Summary of the invention

Therefore, the object of the present invention is to provide a kind of Coal-rock identification method based on lithostructure meta learning, this recognition methods is without the need to manually meticulously choosing characteristics of image, and picture noise is had to the robustness of height, thus the method has higher identification stability and discrimination, can be automated mining, automatic coal discharge, robotization select the production runes such as cash to provide reliable coal and rock identify information.

According to a kind of embodiment form, a kind of Coal-rock identification method based on lithostructure meta learning is provided, comprises the steps:

S1. gather some width coals, rock image, composing training sample set Y, the category label collection of described sample set is set to L;

S2. utilize training sample set Y, structure coal petrography structural elements D and rarefaction representation matrix X, constructed fuction is as follows:

$\begin{array}{cc}\underset{D,X}{\min }\left\{\right||Y-DX|{|}_F^2\}& s.t.\∀i,\left|\right|x_i|{|}_0\≤T_0\end{array};$

S3. sorter training and test is carried out with X

Choose arbitrarily a part of column vector composing training collection training classifier in X, remaining part forms test set and is used for testing, and draws the value that each parameter of sorter is suitable, makes this sorter meet accuracy of identification;

S4. for unknown coal petrography object images y to be identified _i, with the structural elements D tried to achieve in step S2 as input, try to achieve y with greedy algorithm _irarefaction representation vector x _i, be the proper vector of this unknown images, the objective function of sign is as follows:

$\begin{array}{cc}\underset{x_i}{\min }\left\{\right||y_i-{\mathrm{Dx}}_i|{|}_2^2\}& s.t.\∀i,\left|\right|x_i|{|}_0\≤T_0\end{array};$

S5. by the proper vector x of unknown coal petrography object images _iin the sorter trained in (trying to achieve in step S4) input step S3, differentiate coal lithotypes.

Accompanying drawing explanation

By following explanation, accompanying drawing embodiment becomes aobvious and sees, its only with at least one described by reference to the accompanying drawings preferably but the way of example of non-limiting example provide.

Fig. 1 is the basic procedure of Coal-rock identification method of the present invention.

Specific implementation method

Fig. 1 is the basic procedure of the Coal-rock identification method based on lithostructure meta learning, is specifically described see Fig. 1.

S1. gather coal, rock object images, random selecting some width coal petrographys image carries out pre-service, composing training sample set Y, and marks classification, and this classification label set is set to L;

The different illumination come as coal-face collection at scene from coal and rock identify task, some coals of different points of view, rock sample image are (if coloured image, first be converted into gray level image), the center of image intercept pixel size as 64 × 64 subimage, and pull into row by often opening subimage, become the column vector that dimension is 4096, be normalized again, common composition training sample matrix Y.If gather coal, each 200 width of rock image altogether, then Y is the matrix of 4096 × 400, N=400.By numeral to each atom mark classification in Y, category label collection L to be then size be 1 × 400 matrix, deposit the category label that in Y, each image is corresponding.

S2. training sample set Y is utilized, structure coal petrography structural elements D and rarefaction representation matrix X;

This step is actual, and what will solve is following optimization problem:

$\begin{array}{cc}\underset{D,X}{\min }\left\{\right||Y-DX|{|}_F^2\}& s.t.\∀i,\left|\right|x_i|{|}_0\≤T_0\end{array}$

Wherein, x _ia certain atom in X, || || _ff norm, || || ₀l ₀norm, T ₀it is degree of rarefication.

For above-mentioned optimization problem, here, we adopt KSVD algorithm to solve, that part of training process in corresponding diagram 1, and comprising sparse coding and structural elements, to upgrade first latter two stage as follows:

In the sparse coding stage: first initialization is carried out to structural elements D, from Y, the individual atomic building D of s (as 100) can be chosen arbitrarily here; Use OMP Algorithm for Solving X again, namely solve following non-convex optimization problem:

$\begin{array}{ccc}i=1,2,...,N,& \underset{x_i}{\min }\left\{\right||y_i-{\mathrm{Dx}}_i|{|}_2^2\}& s.t.\left|\right|x_1|{|}_0\≤T_0.\end{array}$

Wherein, y _ia certain atom in Y, || || ₂l ₂norm.

Structural elements is the new stage more: structural elements D upgrades by column, supposes the kth row d that will upgrade structural elements D _k, make in sparse matrix X with d _kthe row k be multiplied is denoted as then objective function can be rewritten as:

$\begin{array}{c}\left|\right|Y-DX|{|}_F^2={||Y-\underset{j=1}{\overset{K}{\Σ}}{d}_j{x}_T^j||}_F^2\\ ={||(Y-\underset{j\≠k}{\Σ}{d}_j{x}_T^j)-d_k{x}_T^k||}_F^2\\ ={||E_k-d_k{x}_T^k||}_F^2\end{array}$

In above formula, E _krepresent and remove atom d _kthe error that causes in all samples of composition.

Openness for ensureing, can not directly to E _kcarry out SVD and decompose renewal structural elements, need E _kwith convert. represent and remove middle neutral element, only retains nonzero element, represent and only retain E _kmiddle corresponding d _kwith those of middle nonzero element product.Like this, former objective function can be converted into wherein F norm uses L ₂norm substitutes.Right carry out SVD decomposition, upgrade structural elements kth row d _kfor the first row element of U, upgrade for the first row element of V and the product of Δ (1,1).So just complete the once renewal of structural elements one row atom.

S3. training and the test of sorter is carried out with X;

L in step S1 and the X in step S2 is as input, and X is the eigenvectors matrix of training sample set Y.Set initial K value, KNN sorter is practiced in a part of column vector composing training training chosen arbitrarily in X, and remaining part forms test set and is used for testing, and draws discrimination.Adjustment K value, makes the accuracy of identification of this sorter reach more than 95%.If do not reach the precision of needs, from X, the training that some column vectors carry out sorter can be got by multiselect, also can strengthen the number of training sample.

S4. process unknown coal petrography object images to be identified, then characterize with structural elements D, obtain proper vector x _i;

The processing procedure of unknown coal petrography object images is identical with preprocessing process in step S1, and this unknown images is processed into normalized column vector y _i.Structural elements D in this step derives from the output of step S2, asks x _imethods and steps S2 in the sparse coding stage similar (can refer step S2), be also utilize OMP greedy algorithm to try to achieve.Right half part in this step corresponding diagram 1.

S5. identify unknown coal petrography object images, export its generic label.

By the proper vector x of unknown coal petrography object images _iin the KNN sorter trained in (trying to achieve in step S4) input step S3, according to majority vote principle, differentiate coal lithotypes, export the category label of this unknown coal petrography image.Identifying in this step corresponding diagram 1.

Claims (1)

1., based on the Coal-rock identification method of lithostructure meta learning, comprise the following steps:

S1. gather some width coals, rock image, composing training sample set Y, the category label collection of described sample set is set to L;

S2. utilize training sample set Y, structure coal petrography structural elements D and rarefaction representation matrix X, constructed fuction is as follows:

$\begin{array}{ccc}\underset{D,X}{min}\left\{\right||Y-DX|{|}_F^2\}& s.t.& \∀i,\left|\right|x_i|{|}_0\≤T_0\end{array};$

S3. sorter training and test is carried out with X

Choose arbitrarily a part of column vector composing training collection training classifier in X, remaining part forms test set and is used for testing, and draws the value that each parameter of sorter is suitable, makes this sorter meet accuracy of identification;

S4. for unknown coal petrography object images y to be identified _i, with the structural elements D tried to achieve in step S2 as input, try to achieve y with greedy algorithm _irarefaction representation vector x _i, be the proper vector of this unknown images, the objective function of sign is as follows:

$\begin{array}{ccc}\underset{x_i}{min}\left\{\right||y_i-{\mathrm{Dx}}_i|{|}_2^2\}& s.t.& \∀i,\left|\right|x_i|{|}_0\≤T_0\end{array};$

S5. by the proper vector x of unknown coal petrography object images _iin the sorter trained in (trying to achieve in step S4) input step S3, differentiate coal lithotypes.