CN102680050A - Sulfur flotation liquid level measuring method based on foam image characteristic and air volume - Google Patents

Sulfur flotation liquid level measuring method based on foam image characteristic and air volume Download PDF

Info

Publication number
CN102680050A
CN102680050A CN201210120613XA CN201210120613A CN102680050A CN 102680050 A CN102680050 A CN 102680050A CN 201210120613X A CN201210120613X A CN 201210120613XA CN 201210120613 A CN201210120613 A CN 201210120613A CN 102680050 A CN102680050 A CN 102680050A
Authority
CN
China
Prior art keywords
liquid level
characteristic
value
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.)
Granted
Application number
CN201210120613XA
Other languages
Chinese (zh)
Other versions
CN102680050B (en
Inventor
阳春华
朱红求
何明芳
桂卫华
陈志鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Priority to CN201210120613.XA priority Critical patent/CN102680050B/en
Publication of CN102680050A publication Critical patent/CN102680050A/en
Application granted granted Critical
Publication of CN102680050B publication Critical patent/CN102680050B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B03—SEPARATION OF SOLID MATERIALS USING LIQUIDS OR USING PNEUMATIC TABLES OR JIGS; MAGNETIC OR ELECTROSTATIC SEPARATION OF SOLID MATERIALS FROM SOLID MATERIALS OR FLUIDS; SEPARATION BY HIGH-VOLTAGE ELECTRIC FIELDS
    • B03D—FLOTATION; DIFFERENTIAL SEDIMENTATION
    • B03D1/00—Flotation
    • B03D1/02—Froth-flotation processes
    • B03D1/028—Control and monitoring of flotation processes; computer models therefor

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a sulfur flotation liquid level measuring method based on a foam image characteristic and an air volume. The sulfur flotation liquid level measuring method comprises the following steps of: on the basis of using a foam image obtained by an industrial camera mounted in a sulfur flotation field, extracting a speed characteristic of the foam image by using a method based on macro block tracking; carrying out statistics on the pixels of a differential image so as to acquire a stability characteristic; and combining the characteristics with an air volume data collected in situ as a model to be input, and establishing a liquid level measurement model of a relevance vector machine (RVM) after removing abnormal data by using a Pauta standard. The liquid level measuring method provided by the invention has the characteristics of simplicity, convenience and rapidness, is applicable to liquid level real-time measurement in a sulfur flotation production process, effectively solves the problem that the liquid level measurement in a sulfur flotation field is inaccurate, and optimizes the production and operation of sulfur flotation, so that the liquid level measuring method provided by the invention has an important meaning for improvement of the grade of sulfur concentrate.

Description

A kind of sulphur flotation level measuring method based on froth images characteristic and ventilation
Technical field
The invention belongs to the froth images feature extraction and the level gauging technical field of floatation process, be specially a kind of sulphur flotation level measuring method based on froth images characteristic and ventilation.
Background technology
Flotation is to use the most a kind of beneficiation method in the mineral processing, is usually directed to complex physicochemical process, its objective is to obtain high-grade concentrate product.Because the strong-hydrophobicity of sulfide; The sulphur floatation process need not to add any medicament; Only relating to physical process, mainly is to make the strong sulfide mineral of hydrophobicity from useless gangue, separate through regulating liquid level (froth bed thickness) and ventilation, obtains high-grade sulphur concentrate.
Liquid level directly has influence on the height of sulphur concentrate grade as a key operation amount of sulphur floatation process.Flotation cell is prone to turn over ore pulp when liquid level is higher, and flotation cell does not have overflow usually when liquid level is on the low side, causes concentrate grade low.Therefore, the accurate measurement of liquid level is the basis of realizing Optimization for liquid level control, researches and develops a kind of effective level measuring method and has great importance for improving the sulphur concentrate grade.
The on-the-spot float-type fluid level transmitter of sulphur flotation is solidified by ore pulp easily and is stuck in flotation cell bottom or ore pulp layer middle part; Cause that level gauging is inaccurate; Cause the operator that the mistake of liquid level is regulated; When serious even cause flotation cell emptying, do not have overflow, turn over unusual service condition such as ore pulp, have a strong impact on the sulphur concentrate grade.The height of liquid level directly influences flotation top layer foam flooding velocity and foam stabilization degree, and ventilation also can have influence on the height of liquid level, therefore; Research is based on the level measuring method of machine vision; Adopt a kind of method that liquid level is carried out non-contact measurement, for operating personnel provide level gauging value accurately, for realizing the control of sulphur floatation process Optimization for liquid level based on froth images characteristic and ventilation; Improve concentrate grade, the market competitiveness that improves enterprise has great importance.
Summary of the invention
The objective of the invention is to solve the problem of sulphur floatation process level gauging, proposed a kind of sulphur flotation level measuring method based on froth images characteristic and ventilation.Main contents of the present invention are following:
A kind of sulphur flotation level measuring method based on froth images characteristic and ventilation is characterized in that, may further comprise the steps:
The first step; Under normal production conditions; The froth images that obtains with the field erected industrial camera of sulphur flotation is the basis, adopts the method for following the tracks of based on macro block to extract the velocity characteristic of froth images, and utilizes velocity information statistical difference partial image pixel to obtain the degree of stability characteristic; Ventilation data and liquid level data in conjunction with collection in worksite make up sample set, and adopt La Yida criterion rejecting abnormalities data.
Step 1: gather the froth images that video camera obtains under the normal production conditions, extract foam speed and degree of stability characteristic;
The velocity characteristic method for distilling that employing is followed the tracks of based on macro block; The calculating principle is following: select a sub-piece (macro block) as template from a certain position of the former frame of image sequence consecutive frame; In present frame search best match position; The criterion that macro block is followed the tracks of adopts Normalized Cross Correlation Function, and its computing method are following:
Figure 201210120613X100002DEST_PATH_IMAGE001
Wherein
Figure 391836DEST_PATH_IMAGE002
and
Figure 201210120613X100002DEST_PATH_IMAGE003
distinguishes representation module image and image subblock zone to be searched; The gray average of , difference representation module image and subgraph to be searched; The size of
Figure 217896DEST_PATH_IMAGE006
, representation template,
Figure 97865DEST_PATH_IMAGE008
,
Figure 201210120613X100002DEST_PATH_IMAGE009
represent displacement.
Making the maximum position of cross correlation function is best match position; Utilize this position and previous frame to obtain position poor of template; And confirm time interval of two continuous frames according to frame per second, obtain the speed parameter
Figure 127132DEST_PATH_IMAGE010
of this moment Pixel-level.
Utilize the foam velocity information, a back two field picture of two continuous frames image is transformed to the same position of former frame image, calculate the difference of first two field picture and changing image then, the number of pixels of difference image surpasses given threshold value and then calculates degree of stability.
Foam can be expressed as with mathematical expression:
Figure 201210120613X100002DEST_PATH_IMAGE011
Figure 383539DEST_PATH_IMAGE012
Wherein:
Figure 201210120613X100002DEST_PATH_IMAGE013
,
Figure 798340DEST_PATH_IMAGE014
Represent respectively the two continuous frames image ( i, j) point grey scale pixel value.
Figure 553937DEST_PATH_IMAGE016
expression froth images degree of stability threshold value.The total pixel number of
Figure 201210120613X100002DEST_PATH_IMAGE017
expression froth images processing region.
Step 2: according to foam speed and the degree of stability characteristic calculated; On-the-spot ventilation data and corresponding level value make up four-dimensional sample set
Figure 54189DEST_PATH_IMAGE018
: ;
Figure 786390DEST_PATH_IMAGE020
is the foam velocity characteristic;
Figure 201210120613X100002DEST_PATH_IMAGE021
is foam stabilization degree characteristic;
Figure 134326DEST_PATH_IMAGE022
is on-the-spot ventilation data;
Figure 201210120613X100002DEST_PATH_IMAGE023
is corresponding level value, and
Figure 44513DEST_PATH_IMAGE024
is number of samples;
Step 3: sample set data
Figure 969744DEST_PATH_IMAGE018
are adopted La Yida criterion rejecting abnormalities data;
if
Figure 201210120613X100002DEST_PATH_IMAGE025
= =
Figure 201210120613X100002DEST_PATH_IMAGE027
(
Figure DEST_PATH_IMAGE029
=1;,
Figure 770395DEST_PATH_IMAGE024
;
Figure 54746DEST_PATH_IMAGE030
=1; 2; 3); Calculate the mean value
Figure DEST_PATH_IMAGE031
of each dimension data respectively; Record the standard error
Figure 864612DEST_PATH_IMAGE032
of each dimension value by the Bezier formula; If certain measured value
Figure 201210120613X100002DEST_PATH_IMAGE033
satisfies ; Think that then corresponding
Figure 683980DEST_PATH_IMAGE025
is exceptional value, the data of deletion correspondence in
Figure 139233DEST_PATH_IMAGE018
.
In second step, because same liquid level can correspondingly be organized froth images and air quantity parameters more, the liquid level model comprises necessary probabilistic information, and associated vector machine (RVM) utilizes Bayesian frame to make up learning machine, and the result of output has probability density distribution.Therefore adopt the RVM method, import as model, set up the level gauging model with characteristics of image and ventilation.
Step 1: the sample set according to the first step obtains is set up the RVM regression model;
Input, output collection for training are (M is the number of samples after the rejecting abnormalities data), establish the model that objective function
Figure 979067DEST_PATH_IMAGE036
carries noise:
Where noise
Figure 828206DEST_PATH_IMAGE038
with mean zero and variance
Figure 201210120613X100002DEST_PATH_IMAGE039
Gaussian distribution.Wherein
Figure 828261DEST_PATH_IMAGE040
;
Figure DEST_PATH_IMAGE041
is kernel function, and is weight vector.The likelihood function of training sample set is:
Figure 201210120613X100002DEST_PATH_IMAGE043
Wherein
Figure 194968DEST_PATH_IMAGE044
,
Figure DEST_PATH_IMAGE045
Step 2: parameter reasoning;
The prior distribution of the weight in the definition step 1 is for depending on the Gaussian distribution of ultra parameter
Figure 725044DEST_PATH_IMAGE046
, promptly
Figure 201210120613X100002DEST_PATH_IMAGE047
Figure 584416DEST_PATH_IMAGE048
is the ultra parameter of decision weights
Figure 201210120613X100002DEST_PATH_IMAGE049
prior distribution in the formula, the sparse characteristic of its final decision model.According to bayesian criterion, the posteriority likelihood that can obtain weight vectors
Figure 68618DEST_PATH_IMAGE049
is distributed as:
Figure 242110DEST_PATH_IMAGE050
Figure 201210120613X100002DEST_PATH_IMAGE051
Wherein the posteriority covariance is:
Figure 689010DEST_PATH_IMAGE052
In the formula
Figure 201210120613X100002DEST_PATH_IMAGE053
.
Posterior Mean is:
Figure 30867DEST_PATH_IMAGE054
The likelihood function formula of training sample set can be carried out integration through the weight variable, and the edge likelihood that can be depended on
Figure 126999DEST_PATH_IMAGE046
and is distributed as:
In the formula
Figure 579157DEST_PATH_IMAGE056
.
Step 3: ultra parameter optimization;
Owing to can not obtain to make the edge likelihood in the step 3 to distribute maximum
Figure 780331DEST_PATH_IMAGE046
and
Figure 675344DEST_PATH_IMAGE039
with analytical form, so use the estimation technique that iterates.To following formula about
Figure 987377DEST_PATH_IMAGE046
differentiate; Making it is zero, can get
Wherein
Figure 785700DEST_PATH_IMAGE058
; is i Posterior Mean,
Figure 157775DEST_PATH_IMAGE060
be that current and
Figure 593490DEST_PATH_IMAGE039
is according to i diagonal entry in the posteriority weight covariance matrix
Figure 201210120613X100002DEST_PATH_IMAGE061
that calculates gained in the step 2.
To noise ; Utilize the said method differentiate, obtain upgrading formula:
Figure 789297DEST_PATH_IMAGE062
After obtaining parameter
Figure 201210120613X100002DEST_PATH_IMAGE063
and
Figure 455639DEST_PATH_IMAGE064
, reappraise the Posterior Mean and the variance of weight.In the iteration estimation procedure; Most
Figure 201210120613X100002DEST_PATH_IMAGE065
value is for more and more approaching infinity; Promptly corresponding
Figure 375053DEST_PATH_IMAGE066
is 0; Its corresponding basis function can be deleted, thereby reaches sparse property.Other
Figure 882389DEST_PATH_IMAGE065
can stablize the convergence finite value, and corresponding with it promptly is called associated vector.
Step 4: according to the RVM level gauging model that obtains, be the model input, measure the real-time level value with on-site real-time froth images characteristic and ventilation value.
The present invention proposes to measure sulphur flotation liquid level with froth images characteristic and ventilation; The method that employing is followed the tracks of based on macro block is extracted the velocity characteristic of froth images, and utilizes velocity information statistical difference partial image pixel to obtain the degree of stability characteristic, in conjunction with the ventilation data and the liquid level data structure sample set of collection in worksite; And employing La Yida criterion rejecting abnormalities data; And based on the RVM method, import as model with characteristics of image and ventilation, set up the level gauging model; Can avoid float type level gauge to be solidified the inaccurate problem of level gauging that causes, realize the accurate measurement of sulphur flotation liquid level by ore pulp.
The present invention has the characteristics simply and easily of calculating; Be suitable for the sulphur floatation process; Have stronger practicality, make the site operation personnel can quick and precisely measure sulphur floatation process liquid level, can effectively reduce sulphur flotation liquid level mistuning joint; Optimize the sulphur floating operation, make monthly sulphur concentrate grade improve 11.17%.
Description of drawings
Fig. 1 sulphur flotation liquid level measurement result.
Embodiment
Embodiment 1 technical scheme for a better understanding of the present invention is that example further describes embodiment of the present invention with certain plumbous zinc factory zinc hydrometallurgy method sulphur floatation process.
The first step; Under normal production conditions; The froth images that obtains with the field erected industrial camera of sulphur flotation is the basis, adopts the method for following the tracks of based on macro block to extract the velocity characteristic of froth images, and utilizes velocity information statistical difference partial image pixel to obtain the degree of stability characteristic; Ventilation data and liquid level data in conjunction with collection in worksite make up sample set, and adopt La Yida criterion rejecting abnormalities data.
Step 1: gather the froth images that video camera obtains under the normal production conditions, extract foam speed and degree of stability characteristic;
The velocity characteristic method for distilling that employing is followed the tracks of based on macro block; The calculating principle is following: select a sub-piece (macro block) as template from a certain position of the former frame of image sequence consecutive frame; In present frame search best match position; The criterion that macro block is followed the tracks of adopts Normalized Cross Correlation Function, and its computing method are following:
Figure 641584DEST_PATH_IMAGE001
（1）
Wherein
Figure 348377DEST_PATH_IMAGE002
and
Figure 959487DEST_PATH_IMAGE003
distinguishes representation module image and image subblock zone to be searched; The gray average of
Figure 860578DEST_PATH_IMAGE004
,
Figure 189929DEST_PATH_IMAGE005
difference representation module image and subgraph to be searched; The size of
Figure 434834DEST_PATH_IMAGE006
,
Figure 166030DEST_PATH_IMAGE007
representation template; According to the Feature Selection of actual sulphur flotation froth,
Figure 424973DEST_PATH_IMAGE008
,
Figure 726772DEST_PATH_IMAGE009
represent displacement here.
Making the maximum position of cross correlation function is best match position; Utilize this position and previous frame to obtain position poor of template; And confirm time interval of two continuous frames according to frame per second, obtain the speed parameter
Figure 791680DEST_PATH_IMAGE010
of this moment Pixel-level.
Utilize the foam velocity information, a back two field picture of two continuous frames image is transformed to the same position of former frame image, calculate the difference of first two field picture and changing image then, the number of pixels of difference image surpasses given threshold value and then calculates degree of stability.
The foam stabilization degree is relevant with factors such as mineral species and aeration quantitys, the break degree of difficulty or ease of expression flotation froth.Can be expressed as with mathematical expression:
Figure 315065DEST_PATH_IMAGE011
Figure 790915DEST_PATH_IMAGE012
?（2）
Wherein:
Figure 32540DEST_PATH_IMAGE013
,
Figure 635560DEST_PATH_IMAGE014
Represent respectively the two continuous frames image ( i, j) point grey scale pixel value. expression froth images degree of stability threshold value (here get global threshold 20% ~ 25%).The total pixel number of
Figure 427247DEST_PATH_IMAGE017
expression froth images processing region is
Figure 201210120613X100002DEST_PATH_IMAGE067
here. sThe difference of respective pixel gray-scale value of representing adjacent two frame froth images is less than the ratio of pixel number and total pixel number of the threshold value of regulation.This ratio is big more, and foam stability is good more.
Step 2: according to foam speed and the degree of stability characteristic calculated; On-the-spot ventilation data and corresponding level value make up four-dimensional sample set
Figure 264490DEST_PATH_IMAGE018
:
Figure 671201DEST_PATH_IMAGE068
;
Figure 169178DEST_PATH_IMAGE020
is the foam velocity characteristic;
Figure 488295DEST_PATH_IMAGE021
is foam stabilization degree characteristic;
Figure 766830DEST_PATH_IMAGE022
is on-the-spot ventilation data;
Figure 649335DEST_PATH_IMAGE036
is corresponding level value, and number of samples is 316;
Step 3: sample set data
Figure 313404DEST_PATH_IMAGE018
are adopted La Yida criterion rejecting abnormalities data;
if
Figure 52690DEST_PATH_IMAGE025
=
Figure 303674DEST_PATH_IMAGE026
=
Figure 52187DEST_PATH_IMAGE027
(
Figure 201210120613X100002DEST_PATH_IMAGE069
=1;, 316;
Figure 570762DEST_PATH_IMAGE030
=1; 2; 3); Calculate the mean value
Figure 746528DEST_PATH_IMAGE031
of each dimension data respectively; Record the standard error of each dimension value by the Bezier formula; If certain measured value
Figure 709116DEST_PATH_IMAGE033
satisfies
Figure 98509DEST_PATH_IMAGE034
; Think that then corresponding
Figure 717882DEST_PATH_IMAGE025
is exceptional value, the data of deletion correspondence in
Figure 864829DEST_PATH_IMAGE018
.
In second step, because same liquid level can correspondingly be organized froth images and air quantity parameters more, the liquid level model comprises necessary probabilistic information, and associated vector machine (RVM) utilizes Bayesian frame to make up learning machine, and the result of output has probability density distribution.Therefore adopt the RVM method, import as model, set up the level gauging model with characteristics of image and ventilation.
Step 1: the sample set
Figure 955145DEST_PATH_IMAGE018
according to the first step obtains is set up the RVM regression model;
Input, output collection for training are (300 is the number of samples after the rejecting abnormalities data), establish the model that objective function
Figure 12094DEST_PATH_IMAGE036
carries noise:
Figure 778931DEST_PATH_IMAGE037
（3）
Where noise
Figure 678753DEST_PATH_IMAGE038
with mean zero and variance
Figure 307181DEST_PATH_IMAGE039
Gaussian distribution.Wherein
Figure 156319DEST_PATH_IMAGE072
;
Figure 782473DEST_PATH_IMAGE041
is kernel function;
Figure 201210120613X100002DEST_PATH_IMAGE073
is weight vector, and the likelihood function of training sample set is:
Figure 277914DEST_PATH_IMAGE074
（4）
Wherein ,
Step 2: parameter reasoning;
The prior distribution of the weight in the definition step 1 is for depending on the Gaussian distribution of ultra parameter
Figure 866207DEST_PATH_IMAGE046
, promptly
Figure 201210120613X100002DEST_PATH_IMAGE077
（5）
is the ultra parameter of decision weights
Figure 147464DEST_PATH_IMAGE049
prior distribution in the formula, the sparse characteristic of its final decision model.According to bayesian criterion, the posteriority likelihood that can obtain weight vectors
Figure 366961DEST_PATH_IMAGE049
is distributed as:
Figure 439960DEST_PATH_IMAGE050
Figure 201210120613X100002DEST_PATH_IMAGE079
（6）
Wherein the posteriority covariance is:
Figure 220965DEST_PATH_IMAGE052
（7）
In the formula .
Posterior Mean is:
Figure 91018DEST_PATH_IMAGE054
（8）
The likelihood function formula of training sample set can be carried out integration through the weight variable, and the edge likelihood that can be depended on
Figure 533369DEST_PATH_IMAGE046
and
Figure 734544DEST_PATH_IMAGE039
is distributed as:
Figure 201210120613X100002DEST_PATH_IMAGE081
（9）
In the formula .
Step 3: ultra parameter optimization;
Owing to can not obtain to make the edge likelihood in the step 3 to distribute maximum
Figure 754638DEST_PATH_IMAGE046
and
Figure 739912DEST_PATH_IMAGE039
with analytical form, so use the estimation technique that iterates.
Figure 111987DEST_PATH_IMAGE046
vectorial initial value all is taken as 0.01 in the iterative process; Examining wide is 0.5; To following formula about
Figure 792498DEST_PATH_IMAGE046
differentiate; Making it is zero, can get
Figure 845905DEST_PATH_IMAGE057
（10）
Wherein
Figure 997270DEST_PATH_IMAGE058
;
Figure 540246DEST_PATH_IMAGE059
is i Posterior Mean, be that current
Figure 502834DEST_PATH_IMAGE046
and
Figure 525017DEST_PATH_IMAGE039
is according to i diagonal entry in the posteriority weight covariance matrix
Figure 488163DEST_PATH_IMAGE061
that calculates gained in the step 2.
To noise
Figure 330217DEST_PATH_IMAGE039
; Utilize the said method differentiate, obtain upgrading formula:
Figure 725426DEST_PATH_IMAGE062
（11）
After obtaining parameter
Figure 352848DEST_PATH_IMAGE063
and , reappraise the Posterior Mean and the variance of weight.In the iteration estimation procedure; Most
Figure 504660DEST_PATH_IMAGE065
value is for more and more approaching infinity; Promptly corresponding
Figure 15145DEST_PATH_IMAGE066
is 0; Its corresponding basis function can be deleted, thereby reaches sparse property.Other can stablize the convergence finite value, and corresponding with it
Figure 739704DEST_PATH_IMAGE025
promptly is called associated vector.
Step 4: according to the RVM level gauging model that obtains, be the model input, measure the real-time level value with on-site real-time froth images characteristic and ventilation value.According to on-site real-time froth images speed, degree of stability feature extraction; And collect corresponding ventilation data, the real-time measurement result of sulphur flotation liquid level that obtains is as shown in Figure 1, and (wherein, 1 is actual value; 2 is the RVM measured value), the measured value error analysis is shown in subordinate list 1.Can know that from subordinate list 1 the on-line measurement accuracy is high, error is little, can satisfy the production demand.
Sulphur flotation production scene is by a day chemical examination sulphur concentrate grade value, and the on-the-spot sulphur concentrate grade before and after this level measuring method is used is to such as shown in the subordinate list 2.Can know the level measuring method of use from subordinate list 2 based on froth images characteristic and ventilation; Can accurately survey real-time amount sulphur flotation liquid level; For important basis has been established in the accurate adjusting of sulphur floatation process liquid level, and make sulphur concentrate grade maximal value bring up to 81.78% from 72.77%.
Its result shows; Method proposed by the invention can be avoided the on-the-spot float-type fluid level transmitter of sulphur flotation to be solidified by ore pulp causing that level gauging forbidden problem; Reduced sulphur flotation liquid level mistuning joint, realized the Optimizing operation of sulphur floatation process, made monthly sulphur concentrate grade improve 11.17%.
Subordinate list 1:
Project Maximum relative error Average relative error Minimum relative error
Sulphur flotation level gauging 18.13% 6.46% 0.96%
Subordinate list 2:
Project Monthly sulphur concentrate grade Sulphur concentrate grade minimum value Sulphur concentrate grade maximal value
Before the use 61.72% 44.13% 72.77%
After the use 72.89% 60.45% 81.78%

Claims (1)

1. sulphur flotation level measuring method based on froth images characteristic and ventilation is characterized in that may further comprise the steps:
The first step; The froth images that obtains with the field erected industrial camera of sulphur flotation is the basis; The method that employing is followed the tracks of based on macro block is extracted the velocity characteristic of froth images, and utilizes velocity information statistical difference partial image pixel to obtain the degree of stability characteristic, in conjunction with the ventilation data and the liquid level data structure sample set of collection in worksite; And adopt La Yida criterion rejecting abnormalities data, be specially:
Step 1: the froth images that acquisition camera is obtained, extract foam speed and degree of stability characteristic;
The velocity characteristic method for distilling that employing is followed the tracks of based on macro block, select from a certain position of the former frame of image sequence consecutive frame piece be macro block as template, in present frame search best match position, the criterion that macro block is followed the tracks of adopts Normalized Cross Correlation Function:
Figure 434523DEST_PATH_IMAGE001
Wherein and
Figure 325173DEST_PATH_IMAGE003
distinguishes representation module image and image subblock zone to be searched; The gray average of
Figure 968644DEST_PATH_IMAGE004
, difference representation module image and subgraph to be searched; The size of
Figure 970415DEST_PATH_IMAGE006
, representation template,
Figure 422574DEST_PATH_IMAGE008
, represent displacement;
Making the maximum position of cross correlation function is best match position; Utilize this position and previous frame to obtain position poor of template; And confirm time interval of two continuous frames according to frame per second, obtain the speed parameter
Figure 581077DEST_PATH_IMAGE010
of this moment Pixel-level;
Utilize the foam velocity information, a back two field picture of two continuous frames image is transformed to the same position of former frame image, calculate the difference of first two field picture and changing image then, the number of pixels of difference image surpasses given threshold value and then calculates degree of stability;
The foam stabilization degree can be expressed as with mathematical expression:
Figure 463583DEST_PATH_IMAGE011
Figure 753750DEST_PATH_IMAGE012
Wherein:
Figure 633981DEST_PATH_IMAGE013
,
Figure 71915DEST_PATH_IMAGE014
Represent respectively the two continuous frames image ( i, j) point grey scale pixel value,
Figure 201210120613X100001DEST_PATH_IMAGE015
Expression froth images degree of stability threshold value,
Figure 866434DEST_PATH_IMAGE016
The total pixel number of expression froth images processing region;
Step 2: according to foam speed and the degree of stability characteristic calculated; On-the-spot ventilation data and corresponding level value make up four-dimensional sample set : ;
Figure 253050DEST_PATH_IMAGE019
is the foam velocity characteristic;
Figure 477358DEST_PATH_IMAGE020
is foam stabilization degree characteristic;
Figure 476538DEST_PATH_IMAGE021
is on-the-spot ventilation data;
Figure 197107DEST_PATH_IMAGE022
is corresponding level value, and
Figure 547317DEST_PATH_IMAGE023
is number of samples;
Step 3: data sample set
Figure 575316DEST_PATH_IMAGE017
is adopted La Yida criterion rejecting abnormalities data;
if
Figure 694581DEST_PATH_IMAGE024
=
Figure 884254DEST_PATH_IMAGE025
=
Figure 721760DEST_PATH_IMAGE026
;
Figure 724089DEST_PATH_IMAGE027
=1 wherein;
Figure 25757DEST_PATH_IMAGE023
;
Figure 324015DEST_PATH_IMAGE028
=1; 2; 3; Calculate the mean value
Figure 711134DEST_PATH_IMAGE029
of each dimension data respectively; Record the standard error
Figure 18618DEST_PATH_IMAGE030
of each dimension value by the Bezier formula; If certain measured value
Figure 846897DEST_PATH_IMAGE031
satisfies
Figure 643952DEST_PATH_IMAGE032
; Think that then corresponding is exceptional value, the data of deletion correspondence in
Figure 800181DEST_PATH_IMAGE017
;
In second step, because same liquid level can correspondingly be organized froth images and air quantity parameters more, the liquid level model comprises necessary probabilistic information, the associated vector machine, and promptly RVM utilizes Bayesian frame to make up learning machine, and the result of output has probability density distribution; Therefore adopt the RVM method, import as model, set up the level gauging model with characteristics of image and ventilation;
Step 1: the sample set
Figure 810863DEST_PATH_IMAGE017
according to the first step obtains is set up the RVM regression model;
Collection is
Figure 716502DEST_PATH_IMAGE033
for the input of training, output; is the number of samples after the rejecting abnormalities data, establishes the model that objective function
Figure 461921DEST_PATH_IMAGE022
carries noise:
Figure 763327DEST_PATH_IMAGE035
Noise in the formula
Figure 902184DEST_PATH_IMAGE036
is obeyed the Gaussian distribution that average is zero, variance is ; Wherein
Figure 610694DEST_PATH_IMAGE038
;
Figure 595968DEST_PATH_IMAGE039
is kernel function;
Figure 843410DEST_PATH_IMAGE040
is weight vector, and the likelihood function of training sample set is:
Wherein
Figure 341442DEST_PATH_IMAGE042
,
Figure 181222DEST_PATH_IMAGE043
Step 2: parameter reasoning;
The prior distribution of the weight in the definition step 1 is for depending on the Gaussian distribution of ultra parameter
Figure 599565DEST_PATH_IMAGE044
, promptly
Figure 686787DEST_PATH_IMAGE046
is the ultra parameter of decision weights
Figure 817291DEST_PATH_IMAGE047
prior distribution in the formula; The sparse characteristic of its final decision model
According to bayesian criterion, the posteriority likelihood that can obtain weight vectors is distributed as:
Figure 451852DEST_PATH_IMAGE048
Figure 784745DEST_PATH_IMAGE049
Wherein
Figure 271221DEST_PATH_IMAGE050
for the posteriority covariance does,
Figure 359262DEST_PATH_IMAGE051
is Posterior Mean;
The likelihood function formula of training sample set can be carried out integration through the weight variable, and the edge likelihood that can be depended on and
Figure 199097DEST_PATH_IMAGE037
is distributed as:
Figure 602397DEST_PATH_IMAGE052
In the formula
Figure 799023DEST_PATH_IMAGE053
;
Step 3: ultra parameter optimization;
Owing to can not obtain to make maximum and
Figure 228047DEST_PATH_IMAGE037
of edge likelihood distribution in the step 2 with analytical form; So use the estimation technique that iterates; To following formula about differentiate; Making it is zero, can get
Figure 617495DEST_PATH_IMAGE054
Wherein
Figure 531224DEST_PATH_IMAGE055
;
Figure 9610DEST_PATH_IMAGE056
is i Posterior Mean, be that current
Figure 925931DEST_PATH_IMAGE044
and is according to i diagonal entry in the posteriority weight covariance matrix
Figure 169885DEST_PATH_IMAGE050
that calculates gained in the step 2;
To noise
Figure 605545DEST_PATH_IMAGE037
; Utilize the said method differentiate, obtain upgrading formula:
Figure 49296DEST_PATH_IMAGE058
After obtaining parameter
Figure 265514DEST_PATH_IMAGE059
and
Figure 85702DEST_PATH_IMAGE060
; Reappraise the Posterior Mean and the variance of weight; In the iteration estimation procedure; Most value is for more and more approaching infinity; Promptly corresponding
Figure 840390DEST_PATH_IMAGE062
is 0; With its corresponding basis function deletion; Thereby reach sparse property; Other
Figure 216008DEST_PATH_IMAGE061
can stablize the convergence finite value, and corresponding with it
Figure 574308DEST_PATH_IMAGE024
promptly is called associated vector;
Step 4: the RVM level gauging model that adopts preceding 3 steps to obtain is the model input with on-site real-time froth images characteristic and ventilation value, measures the real-time level value.
CN201210120613.XA 2012-04-24 2012-04-24 Sulfur flotation liquid level measuring method based on foam image characteristic and air volume Expired - Fee Related CN102680050B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201210120613.XA CN102680050B (en) 2012-04-24 2012-04-24 Sulfur flotation liquid level measuring method based on foam image characteristic and air volume

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201210120613.XA CN102680050B (en) 2012-04-24 2012-04-24 Sulfur flotation liquid level measuring method based on foam image characteristic and air volume

Publications (2)

Publication Number Publication Date
CN102680050A true CN102680050A (en) 2012-09-19
CN102680050B CN102680050B (en) 2014-03-05

Family

ID=46812324

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201210120613.XA Expired - Fee Related CN102680050B (en) 2012-04-24 2012-04-24 Sulfur flotation liquid level measuring method based on foam image characteristic and air volume

Country Status (1)

Country Link
CN (1) CN102680050B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104784978A (en) * 2015-04-14 2015-07-22 冶金自动化研究设计院 Liquid level defoaming device for flotation liquid level image recognition equipment
CN107389139A (en) * 2017-08-03 2017-11-24 尤立荣 Micrometeor vision measurement device and vision measuring method
CN107478287A (en) * 2017-08-29 2017-12-15 北矿机电科技有限责任公司 Detection method for determining optimal flotation machine inflation recovery factor beta
CN108844955A (en) * 2018-04-28 2018-11-20 大唐环境产业集团股份有限公司 A kind of Desulphurization for Coal-fired Power Plant absorption tower Slurry bubble quantization device
CN109272548A (en) * 2018-09-28 2019-01-25 北京拓金科技有限公司 A kind of measurement method of floatation process bubble diameter
CN109772593A (en) * 2019-01-25 2019-05-21 东北大学 A kind of mineral pulp level prediction technique based on flotation froth behavioral characteristics
CN112330588A (en) * 2020-08-07 2021-02-05 辽宁中新自动控制集团股份有限公司 Flotation froth image classification method
CN113570636A (en) * 2021-06-16 2021-10-29 北京农业信息技术研究中心 Draught fan ventilation amount detection method and device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN101315669A (en) * 2008-07-15 2008-12-03 北京石油化工学院 Floatation foam image processing method and device
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

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN101315669A (en) * 2008-07-15 2008-12-03 北京石油化工学院 Floatation foam image processing method and device
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

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
何桂春等: "浮选泡沫图像处理技术研究现状与进展", 《有色金属科学与工程》, vol. 2, no. 2, 30 April 2011 (2011-04-30) *
郝元宏等: "一种新的浮选泡沫图像识别方法", 《西安交通大学学报》, vol. 45, no. 4, 30 April 2011 (2011-04-30) *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104784978B (en) * 2015-04-14 2016-08-17 冶金自动化研究设计院 Flotation liquid level image recognition apparatus liquid level defoaming device
CN104784978A (en) * 2015-04-14 2015-07-22 冶金自动化研究设计院 Liquid level defoaming device for flotation liquid level image recognition equipment
CN107389139B (en) * 2017-08-03 2023-01-24 尤立荣 Micro-flow vision measuring device and vision measuring method
CN107389139A (en) * 2017-08-03 2017-11-24 尤立荣 Micrometeor vision measurement device and vision measuring method
CN107478287A (en) * 2017-08-29 2017-12-15 北矿机电科技有限责任公司 Detection method for determining optimal flotation machine inflation recovery factor beta
CN107478287B (en) * 2017-08-29 2019-10-29 北矿机电科技有限责任公司 Detection method for determining optimal flotation machine inflation recovery factor beta
CN108844955A (en) * 2018-04-28 2018-11-20 大唐环境产业集团股份有限公司 A kind of Desulphurization for Coal-fired Power Plant absorption tower Slurry bubble quantization device
CN109272548A (en) * 2018-09-28 2019-01-25 北京拓金科技有限公司 A kind of measurement method of floatation process bubble diameter
CN109772593A (en) * 2019-01-25 2019-05-21 东北大学 A kind of mineral pulp level prediction technique based on flotation froth behavioral characteristics
CN112330588A (en) * 2020-08-07 2021-02-05 辽宁中新自动控制集团股份有限公司 Flotation froth image classification method
CN112330588B (en) * 2020-08-07 2023-09-12 辽宁中新自动控制集团股份有限公司 Classification method for flotation foam images
CN113570636A (en) * 2021-06-16 2021-10-29 北京农业信息技术研究中心 Draught fan ventilation amount detection method and device
CN113570636B (en) * 2021-06-16 2024-05-10 北京农业信息技术研究中心 Method and device for detecting ventilation quantity of fan

Also Published As

Publication number Publication date
CN102680050B (en) 2014-03-05

Similar Documents

Publication Publication Date Title
CN102680050B (en) Sulfur flotation liquid level measuring method based on foam image characteristic and air volume
Cao et al. Integrated prediction model of bauxite concentrate grade based on distributed machine vision
CN103544483B (en) A kind of joint objective method for tracing based on local rarefaction representation and system thereof
CN104156734B (en) A kind of complete autonomous on-line study method based on random fern grader
CN105869178A (en) Method for unsupervised segmentation of complex targets from dynamic scene based on multi-scale combination feature convex optimization
CN104408724A (en) Depth information method and system for monitoring liquid level and recognizing working condition of foam flotation
CN102646279A (en) Anti-shielding tracking method based on moving prediction and multi-sub-block template matching combination
CN101719278B (en) Automatic tracking method for video microimage cells based on KHM algorithm
CN101615183B (en) System and method for analyzing spatial image information and GIS based river time sequence
CN103324936A (en) Vehicle lower boundary detection method based on multi-sensor fusion
CN102063727B (en) Covariance matching-based active contour tracking method
CN102902974A (en) Image based method for identifying railway overhead-contact system bolt support identifying information
CN104063713A (en) Semi-autonomous on-line studying method based on random fern classifier
CN102693216A (en) Method for tracking point feature based on fractional-order differentiation
CN115578732B (en) Label identification method for fertilizer production line
CN108931621B (en) Zinc ore grade soft measurement method based on process texture characteristics
CN108647722B (en) Zinc ore grade soft measurement method based on process size characteristics
CN115131561A (en) Potassium salt flotation froth image segmentation method based on multi-scale feature extraction and fusion
CN114639064B (en) Water level identification method and device
CN104537686A (en) Tracing method and device based on target space and time consistency and local sparse representation
CN115761513A (en) Intelligent remote sensing identification method for mountain large landslide based on semi-supervised deep learning
CN105405149A (en) Composite texture feature extraction method for flotation froth image
CN109993772B (en) Example level feature aggregation method based on space-time sampling
CN106127813B (en) The monitor video motion segments dividing method of view-based access control model energy sensing
CN102136060A (en) Method for detecting population density

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20140305

Termination date: 20150424

EXPY Termination of patent right or utility model