CN105354596B

CN105354596B - Coal-rock identification method based on structural sparse coding

Info

Publication number: CN105354596B
Application number: CN201510758330.1A
Authority: CN
Inventors: 伍云霞; 孙继平
Original assignee: China University of Mining and Technology Beijing CUMTB
Current assignee: China University of Mining and Technology Beijing CUMTB
Priority date: 2015-11-10
Filing date: 2015-11-10
Publication date: 2018-08-14
Anticipated expiration: 2035-11-10
Also published as: CN105354596A

Abstract

The invention discloses a kind of Coal-rock identification methods based on structural sparse coding, this method captures the airspace structure feature of coal petrography, the thus robustness with very strong distinguishing ability and to imaging circumstances variation, so that this method has very high identification stability and recognition correct rate, the production processes such as cash can be selected to provide reliable coal petrography identification information for automated mining, automatic coal discharge, automation.

Description

Coal-rock identification method based on structural sparse coding

Technical field

The present invention relates to a kind of Coal-rock identification methods based on structural sparse coding, belong to coal petrography identification field.

Background technology

It is coal or rock that coal petrography identification automatically identifies coal petrography object with a kind of method.In coal production process, coal Rock identification technology can be widely applied to roller coal mining, driving, top coal caving, raw coal select the production links such as spoil, for reducing Getting working face operating personnel realizes unmanned operation, mitigates labor intensity, improve operating environment, realize coal mine peace Overall height effect production is of great significance.

There are many methods to be applied to coal petrography identification, such as natural Gamma ray detection, radar detection, stress pick, infrared spy Survey, active power monitoring, shock detection, sound detection, dust detection, memory cut etc., but these methods have the following problems： 1. needing to install various kinds of sensors acquisition information additional on existing, cause apparatus structure complicated, it is of high cost.2. coalcutter rolls Stress is complicated in process of production for the equipment such as cylinder, development machine, it is big to vibrate violent, serious wear, dust, and sensor deployment is relatively more tired Difficulty is easy to cause mechanical component, sensor and electric wiring and is damaged, and device reliability is poor.3. for different type machinery Equipment, the selection of best type and the picking up signal point of sensor need to carry out personalized customization there are larger difference, system Bad adaptability.

To solve the above problems, image technique is also increasingly taken seriously and the coal petrography based on image technique that has developed some Recognition methods, however it is all the combination of the manually characteristics of image or characteristics of image of subjective design to carry out coal petrography to have method The feature of identification, engineer often can not accurately catch coal petrography image essential structure to cause to because image-forming condition variation causes Image data variation do not have tool robustness, thus identification stability and recognition correct rate on also have prodigious deficiency.

Need it is a kind of solving the problems, such as or at least improve one or more Coal-rock identification methods intrinsic in the prior art, with Improve coal petrography discrimination and identification stability.

Invention content

Therefore, the purpose of the present invention is to provide a kind of Coal-rock identification method based on structural sparse coding, this method The airspace structure feature of coal petrography, thus the robustness with very strong distinguishing ability and to imaging circumstances variation are captured, from And make this method that there is very high identification stability and recognition correct rate, can be automated mining, automatic coal discharge, automation The production processes such as cash are selected to provide reliable coal petrography identification information.

According to a kind of embodiment form, a kind of Coal-rock identification method encoded based on structural sparse is provided, including as follows Step：

A. the image of known coal (or rock) object of acquisition M+1；

B. N number of image block is extracted from every image respectively and amounts to (M+1) N number of image block [x₁, x₂... x_(M+1)N]=X ∈ R^p×(M+1)N, p is the dimension after image block vectorization；

C. the image block x of wherein M images is taken_iPrimitive matrix D=[d as coal (or rock) image₁, d₂... d_MN]∈ R^p×MN;

D. by solving optimization problemFind out remaining one expressed with D Open each image block x of figure_iPrimitive coefficient b_i∈R^MN×1,Figure As block x_iIt is characterized as and image block x_iThe mean value of the primitive coefficient of same position is：Wherein C is normaliztion constant, all image block x_iFeature constitutes matrix U, and image is special Collect the matrix diagonals element, i.e. F=diag (U)；

E. for coal petrography image to be identified, N number of image block y is extracted with method identical with step B_i, in step C D expresses image block y_i, images to be recognized feature V is found out with method identical with step D；

F. it usesThe similitude with known coal or rock object is measured, is when more than given threshold value Coal (or rock) is otherwise rock (or coal), and δ is given parameter.

Description of the drawings

By following explanation, attached drawing embodiment becomes aobvious and has seen, only preferred at least one being described in conjunction with the accompanying But the way of example of non-limiting embodiment provides.

Fig. 1 is the principle schematic of Coal-rock identification method of the present invention.

Specific embodiment

Fig. 1 is the principle schematic of Coal-rock identification method of the present invention, and main includes 3 layers：Image layer, coding layer and pond Change layer, image layer gives coding layer to provide input, and the present embodiment uses from gray level image abstract image block as the input of coding layer； Coding layer calculates expression coefficient when MK primitive expression of each image block, the present embodiment l₁- norm optimizes calculation expression Coefficient so that the nonzero element in expression coefficient is seldom, thus referred to as sparse coding；Pond layer calculates the system of all expression coefficients It counts characteristic and then obtains the feature representation of input picture, specific implementation steps are as follows：

A. it from the image of the scene of coal petrography identification mission such as M+1 coals (or rock) of coal working face acquisition, uniformly zooms to Suitable size such as 32*32 pixel sizes are as sample image；

B. N number of image block is respectively extracted from each sample image amounts to (M+1) N number of image block [x₁, x₂... x_(M+1)N]=X ∈R^p×(M+1)N, p be image block vectorization after dimension, if tile size takes 16*16 pixels, with step-length be 4 pixels in sample Sampled images block is slided in this image, and each image block vector is standardized：WithThe mean value for removing brightness of image is used with eliminating the influence of brightness changeImage block vector is normalized, wherein 1_pIndicate complete 1 vector of p dimensions, η is constant value；

D. by solving optimization problemFind out remaining one expressed with D Open each image block x of figure_iPrimitive coefficient b_i∈R^MN×1,Figure As block x_iIt is characterized as and image block x_iThe mean value of the primitive coefficient of same position is：Wherein C is normaliztion constant, all image block x_iFeature constitutes matrix U, and image is special Collect the matrix diagonals element, i.e. F=diag (U)；The optimization of wherein the problem of can use approximate gradient algorithm, to each image block x_iCorresponding coefficient u_iOptimized using the following steps：

1. assigning initial value to coefficient u, iterations are set；

2. in iteration each time：

u←u+ξD^T(x-Du), ξ is iteration step length；

3. repeating 2 until iteration terminates.

Claims

1. a kind of Coal-rock identification method based on structural sparse coding, it is characterised in that include the following steps：

A. the image of the known coal of acquisition M+1 or rock object；

B. N number of image block is extracted from every image respectively and amounts to (M+1) N number of image block [x₁, x₂... x_(M+1)N]=X ∈ R^p ^×(M+1)N, p is the dimension after image block vectorization；

C. take the image block of wherein M images as coal or primitive matrix D=[d of rock image₁, d₂... d_MN]∈R^p×MN；

D. by solving optimization problemλ is given parameter, is found out with D tables What is reached is left each image block x of a figure_iPrimitive coefficient b_i∈R^MN×1, Image block x_iIt is characterized as and image block x_iThe mean value of the primitive coefficient of same position is：I=1,2 ..., N, wherein C are normaliztion constant, all image block x_iFeature constitutes matrix U, image Feature takes the matrix diagonals element, i.e. F=diag (U)；

E. for coal petrography image to be identified, N number of image block y is extracted with method identical with step B_i, expressed with the D in step C Image block y_i, images to be recognized feature V is found out with method identical with step D；

F. it usesMeasurement and the similitude of known coal object are coal when more than given threshold value, otherwise for Rock；Alternatively, withThe similitude of measurement and known rock object is rock when more than given threshold value, no It is then coal, δ is given parameter.