CN109685733A - A kind of lead zinc floatation foam image space-time joint denoising method based on bubble motion stability analysis - Google Patents
A kind of lead zinc floatation foam image space-time joint denoising method based on bubble motion stability analysis Download PDFInfo
- Publication number
- CN109685733A CN109685733A CN201811563903.5A CN201811563903A CN109685733A CN 109685733 A CN109685733 A CN 109685733A CN 201811563903 A CN201811563903 A CN 201811563903A CN 109685733 A CN109685733 A CN 109685733A
- Authority
- CN
- China
- Prior art keywords
- image
- block
- denoising
- frame
- sub
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 78
- JQJCSZOEVBFDKO-UHFFFAOYSA-N lead zinc Chemical compound [Zn].[Pb] JQJCSZOEVBFDKO-UHFFFAOYSA-N 0.000 title claims abstract description 45
- 230000033001 locomotion Effects 0.000 title claims abstract description 40
- 239000006260 foam Substances 0.000 title claims abstract description 36
- 238000004458 analytical method Methods 0.000 title claims abstract description 9
- 238000001914 filtration Methods 0.000 claims abstract description 22
- 238000012545 processing Methods 0.000 claims description 25
- 230000009466 transformation Effects 0.000 claims description 12
- 239000000203 mixture Substances 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 claims description 3
- 238000004091 panning Methods 0.000 claims description 3
- 238000003672 processing method Methods 0.000 claims description 3
- 238000005188 flotation Methods 0.000 abstract description 24
- 230000008569 process Effects 0.000 abstract description 21
- 238000012544 monitoring process Methods 0.000 abstract description 15
- 230000000007 visual effect Effects 0.000 abstract description 4
- 238000000605 extraction Methods 0.000 abstract description 3
- 238000012795 verification Methods 0.000 abstract 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 10
- 239000011707 mineral Substances 0.000 description 10
- 239000011701 zinc Substances 0.000 description 9
- 229910052725 zinc Inorganic materials 0.000 description 8
- HCHKCACWOHOZIP-UHFFFAOYSA-N Zinc Chemical compound [Zn] HCHKCACWOHOZIP-UHFFFAOYSA-N 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000007667 floating Methods 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 229910020218 Pb—Zn Inorganic materials 0.000 description 2
- 238000006073 displacement reaction Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000009291 froth flotation Methods 0.000 description 2
- 238000004801 process automation Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003116 impacting effect Effects 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 239000003595 mist Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000003756 stirring Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 229910001656 zinc mineral Inorganic materials 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/10—Image enhancement or restoration by non-spatial domain filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Abstract
The present invention discloses a kind of lead zinc floatation foam image space-time joint denoising method based on bubble motion stability analysis, kinetic characteristic of this method based on bubble, accurate recognition goes out the steady state of motion (SMS) and unsteady motion state (UMS) of foam image subblock, to the sub-block with SMS characteristic using the time-domain filtering denoising based on motion compensation, sub-block with UMS characteristic is denoised using the airspace filter method based on local spatial correlation, and according to the related coefficient of bubble sub-block, joint time-domain filtering result and airspace filter result obtain the space-time joint denoising result of lead zinc floatation foam image to be processed.Experimental verification has been carried out in lead zinc floatation process vision monitoring, the result shows that, this method can obtain the froth images of high s/n ratio, and denoising result structural similarity is strong, lay a good foundation for the accurate extraction and the effective monitoring of floatation process of lead zinc flotation froth visual signature.
Description
Technical field
The present invention relates to commercial lead zinc floatation process automatic monitoring fields, and in particular to arrives lead zinc floatation foam image
Process field.
Background technique
Lead zinc is widely used in electrical industry, mechanical industry, war industry, metallurgical industry, chemical industry, light industry and medicine
The fields such as industry.Tcrude ore, which needs to handle to obtain by froth flotation, has more high-grade Pb-Zn deposits, has subsequent metal
The value of mineral processing.Since lead zinc belongs to non-renewable resources, although lead and zinc mineral resources distribution in China's is wide, concentration degree is high,
But lean ore is more, and rich ore is few, and ore type is complicated, and Associated Constituents are more altogether, therefore present lead Zn Cr coating increasingly payes attention to lead zinc flotation
The performance of process.
Pb-Zn deposits object froth flotation is the difference using mineral particle Surface Physical Chemistry property (mainly hydrophobicity), is made
Mineral grain is selectively realized the beneficiation method that different minerals efficiently separate by bubble adhesion.Due to flotation process
It is long, influence factor is more, sub-loop is complicated and couple seriously, key technique index can not the reasons such as on-line checking, floatation process
Automatic monitoring is difficult, seriously constrains the stable optimal operation of flotation production and the promotion of mineral processing automation level, floatation process
Automatically-monitored is generally acknowledged one of the problem of academia and engineering circles.In order to realize the sustainable development of mineral resources, mineral
The automatic monitoring level of floatation process is in urgent need to be improved.
In recent years, some studies have shown that flotation froth surface visual information such as foam color, size, flow velocity, texture etc.
Feature includes information largely closely related with flotation production status, it might even be possible to be used as the instruction of flotation technology index
Device.It is experienced that factory staff is selected also to be exactly based on observation flotation froth surface visual signature combination flotation production technology data
Manual evaluation is carried out to flotation operating condition, flotation production is carried out according to the recognition result that flotation production run operating condition changes by rule of thumb and is adjusted
Section, to stablize flotation production procedure.Therefore, the Mineral Floating Process monitoring based on machine vision is considered as the following realization flotation
The essential tool of process automation monitoring.
In the lead zinc floatation process monitoring based on machine vision, being effectively treated for lead zinc floatation foam image is steeped with lead zinc
The accurate extraction of foam Image Visual Feature is the key that realize lead zinc floatation process Intellectualized monitoring.However, due to lead zinc flotation
Industry spot image capture environment is often relatively more severe, for example uneven illumination, live dust are more, water mist weight, and along with other
The electromagnetic interference of equipment, picture signal is in collection and transmission inevitably by the interference of noise.The presence of noise
The processing result for significantly impacting lead zinc froth images has seriously affected monitoring system to the standard of lead zinc floatation process work condition judging
The validity of true property and process automation monitoring, to affect machine vision monitoring in lead zinc floatation process Intellectualized monitoring
Middle further genralrlization application.Therefore, effective image denoising processing method is to ensure the lead zinc floatation process based on machine vision
The automatically-monitored key for capableing of successful implementation.
Although have a large amount of image de-noising method at present, in ore dressing process, because of the stirring of flotation cell, scraper plate
Inevitably there is deformation (rotation, size contracting during flowing to concentrate foam recovery slot in scrolling collecting action, flotation froth
Put), collapse, annex, rupture etc. dynamic variation characteristics, common denoising method difficulty be effectively treated these dynamic changes, fragile
Broken froth images.The present invention is directed to lead zinc froth images, and old bubble easily occurs in Mineral Floating Process and collapses, annex, is broken, revolving
Turn and the kinetic characteristic of the continuous dynamic change along with new bubble floating, targetedly proposes a kind of based on bubble motion
The lead zinc floatation foam image space-time of stability analysis combines denoising method, and this method passes through identification lead zinc flotation bubble movement
Stability uses different denoising methods to the foam sub-block with different stability, does not exist together finally, using to these
The froth images sub-block of reason method is weighted processing, obtains final lead zinc floatation foam image joint denoising result.
Summary of the invention
(1) the technical issues of solving
In view of the complicated unsteady motion characteristic of lead zinc flotation froth, slide window processing pixel-by-pixel is carried out to image to be processed and is obtained
It takes to denoising image subblock, by estimation, to kinetic stability of each bubble sub-block in all reference image frames
It is analyzed, effectively picks out steady state of motion (SMS) foam sub-block and unsteady motion state (UMS) foam sub-block.It is right
The time-domain filtering denoising method based on motion compensation is used in the image subblock of SMS, the image subblock in UMS is adopted
It is denoised with the airspace filter method based on local spatial correlation.For same sub-block, because of the fortune in different reference frames
The denoising result that dynamic stability is different and distinct methods is used to obtain, is weighted place using the Weighted Rule based on block similarity
The space-time that reason obtains each sub-block combines denoising result.Finally, the space-time of each froth images sub-block of sliding window covering is combined
Result of making an uproar is integrated, and the space-time of final entire image combines denoising result.This method has fully taken into account lead zinc flotation froth
Kinetic characteristic, greatly maintain image detail while improving image denoising effect, can obtain high quality when
Sky joint denoising picture signal, lays a good foundation for the lead zinc floatation process based on machine vision is automatically-monitored.
(2) technical solution
S1: prepare the reference froth images of lead zinc floatation foam image and Time Continuous to be processed.Specific step is as follows:
Based on currently pending noisy lead zinc floatation foam image, be set as kth frame, kth frame froth images it
It is preceding and collect M frame image respectively later, one image sequence X={ X of total 2M+1 frame image composition1,gk,X2}.Wherein, gkIt represents
Kth frame froth images, that is, currently to denoising image, X1={ gk-M,gk-(M-1),…,gk-1Indicate close to kth frame figure
The set of M frame (having carried out denoising) froth images composition before picture, X2={ gk+1,gk+2,…,gk+MIndicate close to kth
The set of M frame froth images composition after frame image, if the Pixel Dimensions of each frame image are R × C；Wherein M can take 0 to 4
Between natural number；
S2: from froth images g to be processedkThe most upper left corner start, extract a froth images sub-block, according to the sub-block with
The motion relevance of 2M frame reference picture carries out space-time joint denoising.Specific step is as follows:
WithCentered on, it chooses the image subblock progress space-time that a size is S × S and combines
It makes an uproar processing, the equal odd numbers (S≤min (R, C)) for being no more than picture size of S desirable 9,15,33, the space-time joint denoising
Specific step is as follows:
S21: g is setkIn the 1st sub-block beIt is detected by estimationIn X1∪X2In respectively refer to froth images frame
Position and deformation coefficient (rotation angle and size scaling coefficient).IfIn reference picture gj(j=k ± 1 ..., k ± M ∧ gj
∈{X1∪X2) in (center) position beOpposite gjIn the rotation angle of registration block beSize scaling
Coefficient isSpecifically kinetic characteristic detecting step includes:
S211: in reference picture gjIn same position p1Select an equal amount of sub-blockIt is equivalent toHair
It is after having given birth to rotation, zooming and panning as a result, soIt can indicate are as follows:
S212: by image subblockWithLog-polar system is transformed to, is obtainedWithAccording toWithMotion relevance, have
S213: willWithIt is transformed into Fourier transform domain, using phase correlation method
Obtain image subblockRelative to reference image frame gjIn sub-blockRotation angleAnd zoom factorIt is described
Phase correlation method carry out image subblock kinetic characteristic estimation method step are as follows:
Normalized coherence spectra is first calculated,
WhereinFourier transformation is represented,Represent complex conjugate.
Then right againFourier inversion is carried out, the corresponding coordinate position of search maximum is
That is the logarithm of required selected angle and corresponding zoom factor；
S214:, will according to obtained rotation and zoom factorContravariant be changed toWith identical rotation angle and
Expression under zoom scaleIt is obtained again using phase correlation methodRelative toMotion vector, and then obtainIn reference image frame gjIn center position coordinates
S22: according toIn reference picture gjIn matching position and deformation coefficient, with matching positionFor in
The heart, in gjIn takeThe sub-block of size carries out deformation inverse transformation to the sub-block that is intercepted, and by inverse transformation
As a result edge is cut based on center, intercept for currently processed pieceAn equal amount of S × S sub-block, is denoted asAsIn reference frame gjIn form matching and correlation block, useMeasure image subblockWithDissimilar degree, whereinIt indicatesIn reference image frame gjIn matching residual error sub-block, L be lead zinc bubble
Foam image pixel gray level number of stages；
S23: if image subblockWithDissimilar degreeNo more than predetermined threshold τ, then it is assumed thatRelative to reference image frame gjIn sub-blockFor steady state of motion (SMS) foam sub-block, using being based on motion compensation
Time-domain filtering denoising method pairIt is filtered (denoising) processing, is obtainedRelative to reference image frame gjDenoising
As a resultOtherwise, then it is assumed thatFor unsteady motion state (UMS) foam sub-block, using the sky of local spatial correlation
Domain filtering method pairDenoising is carried out, is obtainedRelative to processing image g itselfkDenoising result
Wherein, the specific steps of the time-domain filtering denoising based on motion compensation are as follows:
For SMS sub-block,The time-domain filtering denoising result based on motion compensation are as follows:WhereinFor currently processed sub-blockIn reference picture gjIn corresponding matching and correlation
Block,To represent for weightWithA kind of similarity measurement, useIt is calculated.
Airspace filter method pair based on local spatial correlationCarry out the specific steps of denoising are as follows:
For UMS sub-block, it can only consider that the autocorrelation for handling image itself denoises image.Relative to not
Same reference frame, it is understood that there may be multiframe belongs to UMS, although there is different reference frames at this time,Local spatial correlation sky
Filtering method denoising in domain only needs to calculate primary, circular are as follows:
ForIn any pixel (x, y), 1≤x≤S, 1≤y≤S, the denoising result of the point are as follows:
Wherein w (x, y, u, v) is representedMiddle pixel (x, y) and (2R+1) × (2R+1) size centered on (x, y)
A kind of similarity measurement of (u, v) pixel in search window can be calculated using Euclidean distance:
Wherein, σ represents Gaussian kernel standard deviation, and h is a filtering parameter relevant to σ, can control declining for exponential function
Subtract and change the weight of Euclidean distance；d2(x, y, u, v) represents the size centered on pixel (x, y) as (2f+1) × (2f+1)
Image subblock with pixel (u, v) be in size be (2f+1) × (2f+1) image subblock Euclidean distance:
S24: it is based onIn X1∪X2In each reference image frame in estimation and deformation parameter estimated result, weight
Step S22 and step S23 is executed again, is obtainedWith X1∪X2Denoising result collection based on middle difference reference frameAnd it records simultaneouslyBelong to SMS in 2M reference frame (using the time domain filter based on motion compensation
Wave denoising) number D, and then obtainSpace-time in all 2M reference frames combines denoising resultTool
Body calculates as follows:
If D < 2M (show in 2M frame reference image frame,Existing SMS also has UMS), then
Wherein jForWith gjThe related coefficient of middle match block.
Otherwise, D=2M (show in 2M frame reference image frame,All show as SMS), it is first empty using part at this time
The airspace filter method of domain correlation denoises, and obtainsAbout denoising image g itselfkA kind of denoising resultAnd then it obtains
The space-time obtained finally combines denoising result
Wherein, wjForWith the related coefficient of match block in gj, C is normalized weight coefficient,
S3: in froth images g to be processedkIn, (i.e. from left to right, from top to bottom pixel-by-pixel using sliding window processing mode
Sliding), the image subblock that size is S × S is successively chosen, using method described in step S21 to step S24, to each height
Block individually carries out denoising.Specific steps include:
S31: by top left co-ordinate pointTo lower right corner coordinate pointsInstitute's structure
At rectangular area in, slide, can choose altogether pixel-by-pixelImage that block size is S × S
Block collection
S32: using method described in step S21 to step S24, to selected each image subblock, it is carried out at denoising
Reason, obtains the space-time joint denoising result of relatively all 2M frame reference pictures of each sub-block
S33: all 2M frame reference pictures, froth images g are based onkSpace-time combine denoising resultIt calculates as follows:
For gkIn the calculation method of final denoising result of any pixel p=(x, y) be, according to selectedA image subblock is in image g to be processedkIn positional relationship, record p is by each selected image
The number Q of block covering and the Time-space processing in correspondence image sub-block are as a result, the space-time joint denoising of so pixel is tied
Fruit carries out processing resultWherein (qx,qy) representative image gkMiddle world coordinates
Point p is in image subblockIn local coordinate, (qx,qy) and the corresponding relationship of (x, y) be
(3) beneficial effect
A kind of lead zinc floatation foam image space-time based on bubble motion stability analysis according to the present invention is combined
Method for de-noising, the advantage with the following aspects:
1, the present invention has supplied the lead zinc floatation foam image space-time based on bubble motion stability analysis to combine denoising method,
The correlation and interframe movement characteristic in lead zinc floatation foam image frame are fully considered, using the phase correlation method of extension to lead
The movement of zinc froth images sub-block and deformation parameter (rotation angle and zoom scale) are quickly and effectively estimated, are estimated based on movement
Meter is as a result, the kinetic stability with the foam sub-block to be processed under identical rotation angle and scaled size can be analyzed effectively.
2, the present invention carries out the denoising of space-time joint to lead zinc froth images using the processing mode of sliding window, sufficiently examines
Dynamic motion characteristic of each local block of lead zinc froth images in flotation cell is considered, based on foam sub-block relative to reference picture
The difference of the kinetic stability of frame targetedly selects different denoising methods, effectively increases collapsing, broken, rotation
Turning grade has the space-time joint denoising performance of foam sub-block of a variety of deformation behaviors.Experiment shows going for the method for the present invention
Result of making an uproar has preferable structural similarity, preferably can keep the details of image while removing noise, be subsequent lead zinc
Lead zinc floatation process monitoring of the froth images feature extraction with analysis and based on machine vision is laid a good foundation, while the present invention
Method can also be generalized in other Mineral Floating Process froth images denoisings.
Detailed description of the invention
Illustrate present invention implementation or technical solution in the prior art to become apparent from, it below will be to embodiment or existing skill
Attached drawing needed in art description is briefly described, and described attached drawing is only some embodiments of my invention,
For ordinary people in the field, without creative efforts, it is attached other can also to be obtained according to these figures
Figure.
Fig. 1 is flow chart of the present invention
Fig. 2 is present invention denoising effect picture
Specific embodiment
Below with reference to the attached drawing in present invention implementation, technical solution in the embodiment of the present invention carries out clear, complete
Description, described embodiment is only a part of the invention, based on the embodiments of the present invention, ordinary skill people
Member's all other embodiment obtained without making creative work, belongs to the scope of the present invention.
As shown in Figure 1, the flow chart that the present invention is embodied, specific steps include:
S1: prepare the reference froth images of lead zinc floatation foam image and Time Continuous to be processed.Detailed step is as follows:
Based on currently pending noisy lead zinc floatation foam image, be set as kth frame, kth frame froth images it
It is preceding and collect M frame image respectively later, one image sequence X={ X of total 2M+1 frame image composition1,gk,X2}.Wherein, gkIt represents
Kth frame froth images, that is, currently to denoising image；X1={ gk-M,gk-(M-1),…,gk-1Indicate close to kth frame figure
The set of M frame (having carried out denoising) froth images composition before picture；X2={ gk+1,gk+2,…,gk+MIndicate close to kth
The set of M frame froth images composition after frame image；If the Pixel Dimensions of each frame image are R × C；Wherein M can take 0 to 4
Between natural number.
S2: from froth images g to be processedkThe most upper left corner start, extract a froth images sub-block, according to the sub-block with
The motion relevance of 2M frame reference picture carries out space-time joint denoising.Specific step is as follows:
WithCentered on, it chooses the image subblock progress space-time that a size is S × S and combines
It makes an uproar processing, the equal odd numbers (S≤min (R, C)) for being no more than picture size of S desirable 9,15,33, the space-time joint denoising
Specific step is as follows:
S21: g is setkIn the 1st sub-block beIt is detected by estimationIn X1∪X2In respectively refer to froth images frame
Position and deformation coefficient (rotation angle and size scaling coefficient).IfIn reference picture gj(j=k ± 1 ..., k ± M ∧ gj
∈{X1∪X2) in (center) position beOpposite gjIn the rotation angle of registration block beSize scaling
Coefficient isSpecifically kinetic characteristic detecting step includes:
S211: in reference picture gjIn same position p1Select an equal amount of sub-blockIt is equivalent toHair
It is after having given birth to rotation, zooming and panning as a result, soIt can indicate are as follows:
S212: by image subblockWithLog-polar system is transformed to, is obtainedWithAccording toWithMotion relevance, have
S213: willWithIt is transformed into Fourier transform domain, using phase correlation method
Obtain image subblockRelative to reference image frame gjIn sub-blockRotation angleAnd zoom factorIt is described
Phase correlation method carry out image subblock kinetic characteristic estimation method step are as follows:
First calculate normalized coherence spectra:
WhereinFourier transformation is represented,Represent complex conjugate.
Then right againFourier inversion is carried out,In into
Row peak value searching,Indicate Fourier inversion, corresponding peak coordinate position abscissa is selected angle value, ordinate
For the logarithm of corresponding zoom factor, i.e.,
S213:, will according to obtained rotation and zoom factorContravariant be changed toWith identical rotation angle and
Expression under zoom scaleBy imageCarry out shape inverse transformation obtains imageCoordinate transform formula are as follows:
Wherein, (x, y)TRepresentative image sub-blockOriginal coordinates, (x1,y1)TRepresentative pairCarry out image inverse transformation
Coordinate afterwards.When carrying out image inverse transformation, after coordinate transform, non integer value can be generated, is inserted at this time using bilinearity
It is worth and interpolation processing is carried out to the pixel value of non-integer coordinates, obtains smooth image transformation results.
S214: it is obtained again using phase correlation methodRelative toMotion vector, and then obtainIt is referring to
Picture frame gjIn center position coordinatesSpecific steps include:
It isIn image gjIn matched corresponding sub-block, and corresponding rotation and scaling contravariant has been carried out
It changes, thenWithRelationship can be expressed as,Basis at this timeWithDisplacement relation, willWithFourier transform domain is transformed to, normalized coherence spectra is then calculated again and obtains
WillWithNormalized coherence spectra carry out inversefouriertransform, the displacement between two sub-blocks can be obtained
Vector is can obtainIn picture frame gjIn center are as follows:
Wherein (pos_x (p1),pos_x(p1)) representative image sub-blockCenter.
S22: according toIn reference picture gjIn matching position and deformation coefficient, with matching positionFor in
The heart, in gjIn takeThe sub-block of size, to the sub-block intercepted using coordinate transform described in formula (5)
Method carries out image deformation inverse transformation, and inverse transformation result cuts edge based on center, intercept for it is current
Process blockAn equal amount of S × S sub-block, is denoted asAsIn reference frame gjIn form matching and correlation block, use
Following formula measures image subblockWithDissimilar degree
WhereinIt indicatesIn reference image frame gjIn matching residual error sub-block, L be lead zinc froth images picture
Plain number of grayscale levels.
S23: if image subblockWithDissimilar degreeNo more than predetermined threshold τ, then it is assumed thatRelative to reference image frame gjIn sub-blockFor steady state of motion (SMS) foam sub-block, using being based on motion compensation
Time-domain filtering denoising method pairIt is filtered (denoising) processing, is obtainedRelative to reference image frame gjDenoising
As a resultOtherwise, then it is assumed thatFor unsteady motion state (UMS) foam sub-block, using the airspace of local spatial correlation
Filtering method pairDenoising is carried out, is obtainedRelative to processing image g itselfkDenoising result
S231: where the specific steps of the time-domain filtering denoising based on motion compensation are as follows:
For SMS sub-block, the time-domain filtering denoising result based on motion compensation are as follows:
WhereinFor currently processed sub-blockIn reference picture gjIn corresponding matching and correlation block,For for weight, generation
TableWithA kind of similarity measurement, useIt is calculated；
S232: the airspace filter method pair based on local spatial correlationCarry out the specific steps of denoising are as follows:
For UMS sub-block, it can only consider that the autocorrelation for handling image itself denoises image.Relative to not
Same reference frame, it is understood that there may be multiframe belongs to UMS, although there is different reference frames at this time,Local spatial correlation sky
Filtering method denoising in domain only needs to calculate primary, circular are as follows:
ForIn any pixel (x, y), 1≤x≤S, 1≤y≤S, the denoising result of the point are as follows:
Wherein w (x, y, u, v) is representedMiddle pixel (x, y) and (2R+1) × (2R+1) size centered on (x, y)
A kind of similarity measurement of (u, v) pixel in search window can be calculated using Euclidean distance:
Wherein, σ represents Gaussian kernel standard deviation, and h is a filtering parameter relevant to σ, can control declining for exponential function
Subtract and change the weight of Euclidean distance；d2(x, y, u, v) represents the size centered on pixel (x, y) as (2f+1) × (2f+1)
Image subblock with pixel (u, v) be in size be (2f+1) × (2f+1) image subblock Euclidean distance:
S24: it is based onIn X1∪X2In each reference image frame in estimation and deformation parameter estimated result, weight
Step S22 and step S23 is executed again, is obtainedWith X1∪X2Denoising result collection based on middle difference reference frameAnd it records simultaneouslyBelong to SMS in 2M reference frame (using the time domain filter based on motion compensation
Wave denoising) number D, and then obtainSpace-time in all 2M reference frames combines denoising resultTool
Body calculates as follows:
If D < 2M (show in 2M frame reference image frame,Existing SMS also has UMS), then
WhereinwjForWith gjThe related coefficient of middle match block.
Otherwise, D=2M (show in 2M frame reference image frame,All show as SMS), it is first empty using part at this time
The airspace filter denoising method of domain correlation obtainsAbout denoising image g itselfkA kind of denoising resultAnd then it obtains
The space-time obtained finally combines denoising result
Wherein, wjForWith gjThe related coefficient of middle match block, C are normalized weight coefficient,
S3: in froth images g to be processedkIn, (i.e. from left to right, from top to bottom pixel-by-pixel using sliding window processing mode
Sliding), the image subblock that size is S × S is successively chosen, using method described in step S21 to step S24, to each height
Block individually carries out denoising.Specific steps include:
S31: by top left co-ordinate pointTo lower right corner coordinate pointsInstitute's structure
At rectangular area in, slide, can choose altogether pixel-by-pixelImage that block size is S × S
Block collection
S32: using method described in step S21 to step S24, to selected each image subblock, it is carried out at denoising
Reason, obtains the space-time joint denoising result of relatively all 2M frame reference pictures of each sub-block
S33: all 2M frame reference pictures, froth images g are based onkSpace-time combine denoising resultIt calculates as follows:
For gkIn the calculation method of final denoising result of any pixel p=(x, y) be, according to selectedA image subblock is in image g to be processedkIn positional relationship, record p is by each selected image
The number Q of block covering and the Time-space processing in correspondence image sub-block are as a result, the space-time joint denoising of so pixel is tied
Fruit carries out processing result are as follows:
Wherein (qx,qy) representative image gkMiddle world coordinates point p is in image subblockIn local coordinate, (qx,qy) and
The corresponding relationship of (x, y) are as follows:
S4: preparing new picture frame to be processed, and enabling currently processed picture frame is k=k+1, to other frames in video, weight
Step S1 to S4 is executed again, lead zinc floatation foam image sequence space-time is obtained and combines denoising result, until monitoring stops.
In order to verify the denoising effect of inventive algorithm, zinc is thick, zinc is smart, zinc sweeps three groups in selection lead zinc flotation flowsheet respectively
First slot in flotation cell obtains different image sequences as Image Acquisition point, and respectively zinc is thick, zinc is smart, zinc sweeps three groups
The white Gaussian noise of different standard deviations is added in image sequence, chooses Y-PSNR (PSNR) and structural similarity (SSIM)
Evaluation index as algorithm.
PSNR is that a kind of image based on error between corresponding pixel points objectively evaluates index.It is expressed as follows:
In formula (18) and formula (19), MSE represents the mean square error of present image X and reference picture Y, and H, W are respectively
The height and width of image, n represent the bit number of each pixel, and general gray level image takes 8, i.e. the gray number of pixel takes 256.
The bigger expression image fault of SSIM value is smaller, the value range [0,1] of SSIM, respectively from brightness, contrast, structure
Three aspects measure image similarity, are expressed as follows:
SSIM (X, Y)=l (X, Y) * c (X, Y) * s (X, Y) (20)
In formula (20)~(23), μX、μYRespectively indicate the mean value of image X and Y, σX、σYRespectively indicate the side of image X and Y
Difference, σXYIt is expressed as the covariance of image X and Y, C1、C2、C3For constant.
Its result is as shown in the table:
The different denoising method performances of table 1 compare
It can be seen from the data in Table 1 that the lead zinc flotation froth provided by the invention based on bubble motion stability analysis
Image space-time joint denoising method improves the denoising performance of time-domain filtering, and PSNR with higher can successfully manage bubble
The various dynamic changes such as collapsing, rupture, merger, rotation, are conducive to industrial application.
The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although with reference to the foregoing embodiments
Invention is explained in detail, those skilled in the art should understand that；It still can be to aforementioned each implementation
Technical solution documented by example is modified or equivalent replacement of some of the technical features；And these modification or
Replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (1)
1. a kind of lead zinc floatation foam image space-time based on bubble motion stability analysis combines denoising method, feature exists
In comprising the steps of:
S1: based on currently pending noisy lead zinc floatation foam image, it is set as kth frame, before kth frame froth images
Collect M frame image respectively later, total 2M+1 frame image forms an image sequence X={ X1,gk,X2}.Wherein, gkRepresent kth
Frame froth images, that is, currently to denoising image, X1={ gk-M,gk-(M-1),…,gk-1Indicate close to kth frame image it
The set of preceding M frame (having carried out denoising) froth images composition, X2={ gk+1,gk+2,…,gk+MIndicate close to kth frame figure
The set of M frame froth images composition as after.If the Pixel Dimensions of each frame image are R × C, wherein M can be taken between 0 to 4
Natural number；
S2: in froth images g to be processedkThe most upper left corner, withCentered on, choosing a size is
The image subblock of S × S carries out the denoising of space-time joint, and S desirable 9,15,33 etc. is no more than odd number (S≤min of picture size
(R, C)), the specific steps are as follows:
S21: g is setkIn the 1st sub-block beIt is detected by estimationIn X1∪X2In respectively refer to froth images frame position
It sets and deformation coefficient (rotation angle and size scaling coefficient).IfIn reference picture gj(j=k ± 1 ..., k ± M ∧ gj∈
{X1∪X2) in (center) position beOpposite gjIn the rotation angle of registration block beSize scaling system
Number isSpecifically kinetic characteristic detecting step includes:
S211: in reference picture gjIn same position p1Select an equal amount of sub-block It is equivalent toIt has occurred
Rotation, after zooming and panning as a result, soIt can indicate are as follows:
S212: by image subblockWithLog-polar system is transformed to, is obtainedWith
S213: willWithIt is transformed into Fourier transform domain, is obtained and is schemed using phase correlation method
As sub-blockRelative to reference image frame gjIn sub-blockRotation angleAnd zoom factor
S214:, will according to obtained rotation and zoom factorContravariant be changed toWith identical rotation angle and scaling
Expression under scaleIt is obtained again using phase correlation methodRelative toMotion vector, and then obtain
In reference image frame gjIn center position coordinates
S22: according toIn reference picture gjIn matching position and deformation coefficient, with matching positionCentered on,
In gjIn takeThe sub-block of size carries out deformation inverse transformation to the sub-block that is intercepted, and by inverse transformation result
Edge is cut based on center, intercept for currently processed pieceAn equal amount of S × S sub-block, is denoted asMake
ForIn reference frame gjIn form matching and correlation block, useMeasure image subblockWithDissimilar degree, whereinIt indicatesIn reference image frame gjIn matching residual error sub-block, L be lead zinc bubble
Foam image pixel gray level number of stages；
S23: if image subblockWithDissimilar degreeNo more than predetermined threshold τ, then it is assumed thatPhase
For reference image frame gjIn sub-blockFor steady state of motion (SMS) foam sub-block, using the time domain based on motion compensation
Filtering and noise reduction processing method pairIt is filtered (denoising) processing, is obtainedRelative to reference image frame gjDenoising resultOtherwise, then it is assumed thatFor unsteady motion state (UMS) foam sub-block, using the sky based on local spatial correlation
Domain filtering and noise reduction processing method pairDenoising is carried out, is obtainedRelative to processing image g itselfkDenoising result
Wherein, the specific steps of the time-domain filtering denoising based on motion compensation are as follows:
For SMS sub-block,The time-domain filtering denoising result based on motion compensation are as follows:WhereinFor currently processed sub-blockIn reference picture gjIn corresponding matching and correlation
Block,To represent for weightWithA kind of similarity measurement, useIt is calculated.
The specific steps of airspace filter denoising based on local spatial correlation are as follows:
For UMS sub-block, it can only consider that the autocorrelation for handling image itself denoises image.Relative to different
Reference frame, it is understood that there may be multiframe belongs to UMS, although there is different reference frames at this time,Airspace denoising only need technology one
It is secondary, circular are as follows:
ForIn any pixel (x, y), 1≤x≤S, 1≤y≤S, the denoising result of the point are as follows:
Wherein w (x, y, u, v) is representedMiddle pixel (x, y) is searched for (2R+1) × (2R+1) size centered on (x, y)
A kind of similarity measurement of (u, v) pixel in window can be calculated using Euclidean distance:
Wherein, σ represents Gaussian kernel standard deviation, and h is a filtering parameter relevant to σ, can control the decaying of exponential function and
Change the weight of Euclidean distance；d2(x, y, u, v) represents the size centered on pixel (x, y) as the figure of (2f+1) × (2f+1)
As sub-block with pixel (u, v) be in size be (2f+1) × (2f+1) image subblock Euclidean distance:
S24: it is based onIn X1∪X2In each reference image frame in estimation and deformation parameter estimated result, repetition hold
Row step S22 and step S23 is obtainedWith X1∪X2Denoising result collection based on middle difference reference frameAnd it records simultaneouslyBelong to SMS in 2M reference frame and (uses the time domain based on motion compensation
Filtering and noise reduction processing) number D, and then obtainSpace-time in all 2M reference frames combines denoising resultIt is specific
Step are as follows:
If D < 2M (show in 2M frame reference image frame,Existing SMS also has UMS), then
WhereinωjForWith gjThe related coefficient of middle match block.
Otherwise, D=2M (show in 2M frame reference image frame,All show as SMS), at this time first using empty based on part
The airspace filter denoising method of domain correlation obtainsAbout denoising image g itselfkA kind of denoising result
And then obtain final space-time joint denoising result
Wherein, ωjForWith gjThe related coefficient of middle match block, C are normalized weight coefficient,
S3: in froth images g to be processedkIn, it (i.e. from left to right, is slided pixel-by-pixel from top to bottom) using sliding window processing mode,
The image subblock that size is S × S is successively chosen, it is independent to each sub-block using method described in step S21 to step S24
Carry out denoising.Specific steps include:
S31: by top left co-ordinate pointTo lower right corner coordinate pointsIt is constituted
It in rectangular area, slides, can choose altogether pixel-by-pixelThe image subblock collection that block size is S × S,
I.e.
S32: using method described in step S21 to step S24, to selected each image subblock, it carries out denoising,
Obtain the space-time joint denoising result of relatively all 2M frame reference pictures of each sub-block
S33: all 2M frame reference pictures, froth images g are based onkSpace-time combine denoising resultIt calculates as follows:
For gkIn the calculation method of final denoising result of any pixel p=(x, y) be, according to selectedA image subblock is in image g to be processedkIn positional relationship, record p by each selected image
The number Q of sub-block covering and the Time-space processing in correspondence image sub-block are as a result, the space-time of so pixel combines denoising
As a result carrying out processing result isWherein (qx,qy) representative image gkMiddle global seat
Punctuate p is in image subblockIn local coordinate, (qx,qy) and the corresponding relationship of (x, y) be
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811563903.5A CN109685733B (en) | 2018-12-20 | 2018-12-20 | Lead-zinc flotation foam image space-time combined denoising method based on bubble motion stability analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811563903.5A CN109685733B (en) | 2018-12-20 | 2018-12-20 | Lead-zinc flotation foam image space-time combined denoising method based on bubble motion stability analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109685733A true CN109685733A (en) | 2019-04-26 |
| CN109685733B CN109685733B (en) | 2023-06-23 |
Family
ID=66187997
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811563903.5A Active CN109685733B (en) | 2018-12-20 | 2018-12-20 | Lead-zinc flotation foam image space-time combined denoising method based on bubble motion stability analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109685733B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110109446A (en) * | 2019-05-28 | 2019-08-09 | 中南大学 | A kind of zinc floatation process Fuzzy Fault Diagnosis based on time series feature |
| CN112926606A (en) * | 2021-04-25 | 2021-06-08 | 南京工程学院 | Heterogeneous fusion bubble plume feature extraction method based on deep learning |
| CN113591654A (en) * | 2021-07-22 | 2021-11-02 | 中南大学 | Zinc flotation working condition identification method based on long-term depth characteristics |
| CN116912102A (en) * | 2023-05-11 | 2023-10-20 | 上海宇勘科技有限公司 | Edge-preserving image denoising method and system based on non-local mean value |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101315669A (en) * | 2008-07-15 | 2008-12-03 | 北京石油化工学院 | Floatation foam image processing method and device |
| US8629937B1 (en) * | 2012-07-25 | 2014-01-14 | Vixs Systems, Inc | Motion adaptive filter and deinterlacer and methods for use therewith |
| CN103530653A (en) * | 2013-10-28 | 2014-01-22 | 中国矿业大学(北京) | Flotation froth tracking method and device |
| CN104050687A (en) * | 2014-06-26 | 2014-09-17 | 中国矿业大学(北京) | Analyzing and processing method for flotation bubble motion pattern |
| CN104657949A (en) * | 2015-02-12 | 2015-05-27 | 太原理工大学 | Method for optimizing structural elements during denoising of coal slime flotation froth image |
| CN105991900A (en) * | 2015-02-05 | 2016-10-05 | 扬智科技股份有限公司 | Noise detection method and denoising method |
| CN107705283A (en) * | 2017-06-14 | 2018-02-16 | 华北理工大学 | Particle and bubble hit detection method based on Otsu image segmentation |
-
2018
- 2018-12-20 CN CN201811563903.5A patent/CN109685733B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101315669A (en) * | 2008-07-15 | 2008-12-03 | 北京石油化工学院 | Floatation foam image processing method and device |
| US8629937B1 (en) * | 2012-07-25 | 2014-01-14 | Vixs Systems, Inc | Motion adaptive filter and deinterlacer and methods for use therewith |
| CN103530653A (en) * | 2013-10-28 | 2014-01-22 | 中国矿业大学(北京) | Flotation froth tracking method and device |
| CN104050687A (en) * | 2014-06-26 | 2014-09-17 | 中国矿业大学(北京) | Analyzing and processing method for flotation bubble motion pattern |
| CN105991900A (en) * | 2015-02-05 | 2016-10-05 | 扬智科技股份有限公司 | Noise detection method and denoising method |
| CN104657949A (en) * | 2015-02-12 | 2015-05-27 | 太原理工大学 | Method for optimizing structural elements during denoising of coal slime flotation froth image |
| CN107705283A (en) * | 2017-06-14 | 2018-02-16 | 华北理工大学 | Particle and bubble hit detection method based on Otsu image segmentation |
Non-Patent Citations (4)
| Title |
|---|
| JINLING YANG ET AL: "Denoising of Coal Flotation Froth Image Using Opening and Closing Filters with Area Reconstruction and Alternating Order Filtering", 《ARTIFICIAL INTELLIGENCE AND COMPUTATIONAL INTELLIGENCE》 * |
| JINPING LIU ET AL: "Spatial-temporal Fusion for Flotation Froth Image Denoising Based on BLS-GSM Method in Curvelet Domain", 《IEEJ TRANSACTIONS ON ELECTRICAL AND ELECTRONIC ENGINEERING》 * |
| 刘金平 等: "基于 Curvelet 变换的浮选泡沫图像序列时空联合去噪", 《控制与决策》 * |
| 李建奇 等: "基于提升改进方向波变换的浮选泡沫图像降噪方法", 《中国有色金属学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110109446A (en) * | 2019-05-28 | 2019-08-09 | 中南大学 | A kind of zinc floatation process Fuzzy Fault Diagnosis based on time series feature |
| CN110109446B (en) * | 2019-05-28 | 2020-08-25 | 中南大学 | Zinc flotation process fuzzy fault diagnosis method based on time series characteristics |
| CN112926606A (en) * | 2021-04-25 | 2021-06-08 | 南京工程学院 | Heterogeneous fusion bubble plume feature extraction method based on deep learning |
| CN113591654A (en) * | 2021-07-22 | 2021-11-02 | 中南大学 | Zinc flotation working condition identification method based on long-term depth characteristics |
| CN113591654B (en) * | 2021-07-22 | 2023-09-01 | 中南大学 | Zinc flotation working condition identification method based on long-time-range depth characteristics |
| CN116912102A (en) * | 2023-05-11 | 2023-10-20 | 上海宇勘科技有限公司 | Edge-preserving image denoising method and system based on non-local mean value |
| CN116912102B (en) * | 2023-05-11 | 2024-04-09 | 上海宇勘科技有限公司 | Edge-preserving image denoising method and system based on non-local mean value |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109685733B (en) | 2023-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109685733A (en) | A kind of lead zinc floatation foam image space-time joint denoising method based on bubble motion stability analysis | |
| CN106056053B (en) | The human posture's recognition methods extracted based on skeleton character point | |
| US8300949B2 (en) | Edge detection technique having improved feature visibility | |
| CN110837768B (en) | Online detection and identification method for rare animal protection | |
| CN109285179A (en) | A kind of motion target tracking method based on multi-feature fusion | |
| CN102075686B (en) | Robust real-time on-line camera tracking method | |
| CN103077531B (en) | Based on the gray scale Automatic Target Tracking method of marginal information | |
| CN103440644B (en) | A kind of multi-scale image weak edge detection method based on minimum description length | |
| CN104102904B (en) | A kind of static gesture identification method | |
| CN109903313A (en) | Real-time pose tracking method based on target three-dimensional model | |
| CN108537212A (en) | Students ' behavior detection method based on estimation | |
| CN105787950B (en) | A kind of method being detected to infrared image sea horizon that added up based on row gradient | |
| CN103020918A (en) | Shape-adaptive neighborhood mean value based non-local mean value denoising method | |
| CN108830831B (en) | Zinc flotation foam natural speed feature extraction method based on improved SURF matching | |
| CN106682689A (en) | Image matching method based on multiscale Fourier-Mellin transform | |
| CN109410248B (en) | Flotation froth motion characteristic extraction method based on r-K algorithm | |
| CN109389165A (en) | Oil level gauge for transformer recognition methods based on crusing robot | |
| CN108182704A (en) | Localization method based on Shape context feature | |
| CN105657432A (en) | Video image stabilizing method for micro unmanned aerial vehicle | |
| CN102737230B (en) | Non-local mean filtering method based on direction field estimation | |
| CN109886989A (en) | A kind of automatic tracing of horizons method of Ground Penetrating Radar based on Canny operator | |
| CN104166977B (en) | A kind of Image Matching Similarity Measurement Method and its image matching method | |
| CN106355576B (en) | SAR image registration method based on MRF image segmentation algorithm | |
| CN104616035A (en) | Visual Map rapid matching method based on global image feature and SURF algorithm | |
| CN105825514B (en) | Method for detecting image edge based on shearing wave coefficient probability density distribution |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |