American Journal of Chemical Engineering

2013; 1(4): 70-78 Published online January 10, 2014 (http: !!!"s#ien#epublishin$$roup"#o% & a&#he) doi: 10"11'48 &"a&#he"20130104"12

Image analysis technique as a tool for extracting features from the copper surface froth in the flotation process
Nasser Saghatoleslami1, *, Hajir Karimi2
1 2

(epart%ent o) *he%i#al +n$ineerin$, ,erdo!si -ni.ersity o) /ashhad, /ashhad, 0ran (epart%ent o) *he%i#al +n$ineerin$, -ni.ersity o) 1asu&, 1asu&, 0ran

Email address:
sla%i2u%"a#"ir (3" 4a$hatolesla%i), ha5ar2%ail"yu"a#"ir (6" 7ari%i)

To cite this article:

3asser 4a$hatolesla%i, 6a&ir 7ari%i" 0%a$e 8nalysis 9e#hni:ue as a 9ool )or +;tra#tin$ ,eatures )ro% the *opper 4ur)a#e ,roth in the ,lotation Pro#ess" American Journal of Chemical Engineering. <ol" 1, 3o" 4, 2013, pp" 70-78" doi: 10"11'48 &"a&#he"20130104"12

Abstract: 9he )roth #an be adopted as an indi#ator o) the per)or%an#e o) )lotation pro#esses" 9he study o) )roth i%a$e stru#ture !ould enable us to establish a nu%ber o) para%eters )ro% !hi#h #ould #on.ey the )roth #hara#teristi#s" 9o %onitor the operatin$ per)or%an#e o) the )loatation #ell by %a#hine .ision syste%, it is #ru#ial to identi)y and e;tra#t those )eatures that are des#ripti.e o) the sur)a#e )roth" *onse:uently, it #an pro.ide interdependen#y bet!een the )roth #hara#teristi#s !ith the operatin$ #onditions on one hand (e"$", aeration rate, )roth depth, #he%i#al #o%pound and p6 .ariation) and the #ell para%eters per)or%an#e on the other hand, as !ell (e"$", #opper $rade, re#o.ery and solid #ontents)" 9he ai% o) the present study is to e;a%ine the #opper )roth #hara#teristi#s, by adoptin$ an i%a$e analysis te#hni:ue and hen#e e.aluatin$ )roth )eatures su#h as the a.era$e bubble si=e, bubbles distribution, bubble shape )eatures, bubble elon$ation )a#tor, i%a$e a.era$e #olor and the #olor distribution" >!in$ to the intri#a#y aspe#t o) the )roth stru#ture and in order to %at#h properly bet!een the real )roth and the se$%entation i%a$es, this al$orith% adopts )eatures si%ilar to proper )ilters in the pre-pro#essin$ sta$e, ed$e dete#tion )un#tions, threshold )un#tions and di))erent %athe%ati#al %orpholo$y %odels" 9he )indin$s o) this !or5 re.eal that the si=e and shape o) the )roth bubbles plays an i%portant role in #lassi)yin$ the )roth" 6en#e, it is possible to in#orporate su#h )eatures )or either e.aluatin$ the )lotation #ell per)or%an#e or adoptin$ it )or the auto%ati# on-line #ontrol o) the )lotation pro#ess" 9he )indin$s o) this resear#h #ould also be i%ple%ented to!ards the trainin$ o) the operators" Keywords: 0%a$e 8nalysis, *opper ?rade, ,lotation, ,roth *olor, @ubble 4i=e, @ubble (istribution

1. !trod"ctio!
0n the )lotation #ell, the per)or%an#e o) the #ell is hi$hly dependent on the %orpholo$y o) the sur)a#e )roth" *onse:uently, to #ontrol the %etallur$i#al per)or%an#e o) the #ell (i"e", #opper $rade, #on#entrate re#o.ery and %ass )lo! rate), #ontinuous obser.ation o) the )roth )eatures (i"e", te;ture and #olor )roth) by an operator is #ru#ial" 9o enhan#e the per)or%an#e o) the #ell, the operator %ust respond a##ordin$ to the .isual appearan#e o) the )roth" *o%ple; te;ture o) the )roth and di))erent operatin$ para%eter intera#tions, %a5e this type o) e.aluation not an e))i#ient %eans" 0n re#ent years, a nu%ber o) a#hie.e%ents ha.e been %ade in the )ield o) %a#hine .ision syste% )or enhan#e%ent o) the per)or%an#e o) the )loatation #ell" 9heir ob&e#ti.e !ere initially to a#hie.e a repla#e%ent )or the .isual obser.ation o) the operator and )inally to asso#iate the )roth te;ture #hara#teristi#s !ith the #ell per)or%an#e" 9o adopt this syste% )or #ontrollin$ the )loatation pro#ess, the analysis o) the )roth te;ture !as o) pri%e #on#ern" 9here)ore, )eature e;tra#tion %ethod and identi)i#ation o) the type o) )roth !ere essential throu$h the $eo%etri# )eatures" 0n the 4ar#hesh%eh *opper Plant (lo#ated in 4outh-+ast o) 0ran near the #ity o) 7er%an), a relationship bet!een the )roth te;ture and per)or%an#e o) the #ell !as )ounded by the authors and hen#e %odeled (4a$hatolesla%i et al", 2002) as are illustrated in ,i$ure 1 A1B" 3et!or5 de#ision %a5in$ ba#5 propa$ation al$orith% !ith %ultilayer per#eption stru#ture adopted by 4a$hatolesla%i et al", 2002 )or %odelin$ and #hara#teri=ation o) )lotation )roth #olor and te;ture )or assessin$ the )lotation per)or%an#e in a #opper plant are sho!n in ,i$ure 2 A1B" 0n this !or5, the test !as #arried out in a #ontinuous )lotation test ri$ !ith a rou$her #ell o) appro;i%ately 40 liters in .olu%e" Pi#tures !ere ta5en by a handy #a% under suitable li$ht intensity" 9he operatin$ #onditions and the %aterials !hi#h !ere used in the test are e;hibited in 9able 1"

8%eri#an Journal o) *he%i#al +n$ineerin$ 2013; 1(4): 70-78

71

Table 1. Operating conditions and the material used in the test 'arameters *olle#tors ( E111 F G407 2) ,rothers (8'D3 F /0@*4 ) 8eration rate (8,G) p6 8.era$e )eed )roth $rade &imits 10 - 8D 10 - H0 %!its $r ton $r ton <al.e openin$ C 11"8 0"H #emar$s 10C solution 10C solution

Figure 1. a) Relationship between the copper grade and H (hue) color component. b) Variation of copper grade with the surface froth bubble si e. c) Variation of solid mass flow rates with the surface froth bubble si e. d) Variation of solid mass flow rates with the degree of the sphericit! of the froth bubble si e

1 3

4odiu% isopropyl ;anthate, 2 4odiu% %er#aptoben=othia=ole and sodiu% dial5yl thiophosphate ?ly#ol polypropylene, 4 /ethyl isobutyl #arbonyl

72

3asser 4a$hatolesla%i and 6a&ir 7ari%i:

0%a$e 8nalysis 9e#hni:ue as a 9ool )or +;tra#tin$ ,eatures )ro% the *opper 4ur)a#e ,roth in the ,lotation Pro#ess

Figure 2. "etwor# decision ma#ing $% algorithm with &'% structure adopted b! (aghatoleslami et al.) *++* for modeling and characteri ation of flotation froth color and te,ture -./

9he appli#ation o) %a#hine learnin$ te#hni:ues to e;ploit in)or%ation )ro% di$ital i%a$es o) the )roth phase o) industrial )lotation plants has been analy=ed by ?rou!s and #o-!or5s A2B" 9hey in#luded both #onne#tionist te#hni:ues to identi)y #ontrol de#isions ne#essary to %aintain opti%al operation o) the plant, as !ell as sy%boli# %ethods, su#h as indu#tion te#hni:ues" @oth approa#hes !ere utili=ed to #lassi)y )roth stru#tures based on statisti#al )eatures deri.ed )ro% di$iti=ed i%a$es o) the )roth sur)a#es" 9hey also )ound that ba#5 propa$ation al$orith%s per)or% si$ni)i#antly better than either non-in#re%ental or in#re%ental indu#tion A2B" Pelletier and #o-!or5ers has also analy=ed the )roth by the use o) the %ulti.ariate i%a$e analysis (/08) %ethod to e;tra#t the spe#tral .ariation #ontained in a G@? i%a$e" 9he partial least s:uares (PI4) %ethod !as applied to de.elop an e%piri#al %odel )or the predi#tion o) )roth $rade" 9hey also studied the )easibility o) i%ple%entin$ a hybrid #ontrol strate$y #onsistin$ in a )roth stru#ture super.ision #ontroller based on the )roth #on#entrate $rade esti%ated by /08 and re#o.ery in)erred by i%a$e analysis A3B" 9he robustness o) 833 to interpret i%a$es !ith illu%ination perturbations, produ#ed by li$ht proble%s or dirt atta#hed to the !indo! o) the .ideo

#a%era !as also e.aluated" 0t !as #on#luded that 833 are reliable )or learnin$ and produ#in$ $enerali=ed predi#tions o) the )roth %ean bubble dia%eter and bubble si=e distribution, !hen the %odel is trained usin$ a database that #ontains in)or%ation on the illu%ination intensity A3B" 9he sele#tion o) the proper )rother, spe#i)i#ally )or a $i.en ore, is essential in )lotation" 9he e.aluation o) di))erent )rothers, in ter%s o) %etallur$i#al per)or%an#e, is #o%pli#ated at plant s#ale, sin#e it %ay lead to periods o) %al)un#tionin$ o) the pro#ess and una##eptable de#reases in re#o.ery" >n the other hand, e.aluation based on )rother #he%istry only, is o)ten unsatis)a#tory, be#ause the )rothin$ response is stron$ly ore dependent" ?rano in 200' in.esti$ated the )roth stability a)ter introdu#in$ the pulp sa%ples into a )oa%in$ #olu%n, !here di))erent #onditions in ter%s o) )rother addition, a$itation and aeration #ould be %aintained A4B" +strada in 200H analy=ed a ne! approa#h to esti%ates bubble si=e distribution o) )roth sur)a#es usin$ arti)i#ial neural net!or5s ADB" *ontrol o) )lotation pro#esses is %ostly %ana$ed by plant operators, !ho assess the per)or%an#e o) the plant based on their o!n e;perien#e and other heuristi# rules" 9hese rules tend to be sub&e#ti.e or ill-de)ined, sin#e %ost o) the% are #on#erned !ith the

8%eri#an Journal o) *he%i#al +n$ineerin$ 2013; 1(4): 70-78

73

stru#ture o) the )lotation )roth, su#h as #olor, bubble si=e and shape distributions, )roth %obility and )roth stability" 9hese pheno%ena are .ery di))i#ult to :uanti)y ob&e#ti.ely, and ine;perien#e on the part o) the operators and hu%an error !hi#h #ould lead to si$ni)i#ant ine))i#ien#ies in plant operation" 0n 200H a )u==y syste% !as de.elop%ent in order to support the #ontrol de#isions o) plant operators, !hi#h !ould leads to si$ni)i#antly s%oother #ontrol a#tion and %ore stable plant operation than #ould be obtained !ith #risp sets o) rules or %anual #ontrol strate$ies A'B"

(b) Figure 3. 0roth of a) shallow depth with spherical t!pe of bubbles of low froth grade and b) high depth with bubbles of pol!hedral configuration and of high froth grade

2. Sig!i(ica!ce o( the )roth *haracteristics i! )lotatio! 'rocess

0n the past, the relation bet!een the )roth #hara#teristi#s and per)or%an#e o) the #ell !as not o) the pri%e #on#ern" 9hese )eatures are related to the )roth te;ture and #olor" >nly in re#ent years, #hara#teristi#s o) )roth !ith the aid o) i%a$e analysis ha.e attra#ted the interest o) so%e resear#hers A7-HB" 9he %ost si$ni)i#ant para%eters in the )roth )eatures are the a.era$e bubble si=e, bubble shape, )roth #olor, )roth stability and )roth %obility" 9he resear#h sho!s that the per)or%an#e o) )lotation #ell is hi$hly dependent on bubble si=e in the pulp and there e;ist a relationship bet!een the bubble si=e in the pulp and the )roth phase A10B" 9here)ore, the bubble si=e and its distribution in the )roth are o) si$ni)i#an#e" 9he resear#h also re.eals that the )roth stru#ture #an be #lassi)ied a##ordin$ to the bubble shape and its distribution in di))erent type o) )roth" 6en#e, the per)or%an#e o) the )lotation pro#ess #an be #ontrolled in an e))e#ti.e %anner" ,urther%ore, the distribution o) bubble si=e is a po!er)ul indi#ator o) other para%eters su#h as aeration rate, p6 .ariation and the a%ount substan#e that atta#hes to the bubbles !hi#h are hi$hly depends on the type o) #olle#tor and )rother A11B" 0n $eneral, )roth !ith s%all a%ount o) !ater #ontent has bubbles o) polyhedral #on)i$uration" 9his type o) te;tures has a hi$h )roth $rade" 4pheri#al types o) bubbles are the indi#ation o) diluted )roth o) shallo! depth and lo! )roth $rade" >n the other hand, a bubble o) an ellipti#al #on)i$uration si$ni)ies the sti))ness o) the )roth and the dire#tion o) !hi#h the )or#es are a#tin$ alon$ the )lo!" ,i$ures 3 and 4 de%onstrate the )orth te;tures !ith bubbles o) spheri#al, polyhedral and ellipti#al #on)i$urations"

Gesear#h also re.eals that the )roth #olor (or li$ht intensity) #on.eys in)or%ation about the )roth sur)a#e $rade and there e;ist a #lose relation bet!een the )roth #olor and the inor$ani# #o%positions present in the #on#entrate A12B" 0n the #opper )lotation pro#ess, the hi$h intensity o) yello! #olor in the sur)a#e )roth is an indi#ation o) the hi$h #on#entration o) #al#opyrite and pyrite in the #on#entrate" >n the other hand, the )roth o) dar5ish yello! #olor si$ni)ies the presen#e o) #al#osite in the #on#entrate" ,urther%ore, )roth o) !hitish #olor #an %ani)est that the #on#entration o) the #opper is lo!" ,i$ure 4 sho!s a )roth o) di))erent #olors" ,i$ure Da de%onstrates )roth o) hi$h intensity and o) a uni)or% #olor" >n the other hand, ,i$ure Db represents )roth o) lo! intensity and o) a non-uni)or% #olor" 9here)ore, it has a lo! solid #ontents and a lo! #on#entration o) #opper !ith a lar$e a%ount o) !ater"

(a)

(b) Figure 4. 0roth of a) high depth with elliptical t!pe of bubbles which e,hibit the stiffness bubbles and their resistance against the flow and also carr! particles of specific si es and b) irregular in nature) high depth with bubbles of elliptical configuration and high stabilit!

(a)

74

3asser 4a$hatolesla%i and 6a&ir 7ari%i:

0%a$e 8nalysis 9e#hni:ue as a 9ool )or +;tra#tin$ ,eatures )ro% the *opper 4ur)a#e ,roth in the ,lotation Pro#ess

(a)

Figure 6. (chematic flow diagram of an image processing s!stem

(b) Figure 5. 0roth of a) high intensit!) uniform in color and high froth grade and b) low intensit! and non1uniform in color and low froth grade

,. )roth mage A!alysis Tech!i-"e

9he )unda%ental sta$es o) an i%a$e analyses %ethod in#ludes: i%a$e pre-pro#essin$, i%a$e se$%entation !hi#h in#orporates ed$e dete#tion, spe#i)i#ation o) the )eatures in the pi#ture and )inally e.aluation o) the $eo%etri#al )eatures" 4.1. Image Pre-Processing 0n this sta$e, a spe#i)ied =one o) the pi#ture ou$ht to be sele#ted (i"e", re$ion o) interest or in brie) G>0)" 9his re$ion should ha.e a re#tan$ular #on)i$uration that #ontains the %ain i%a$e and !here the al$orith% should be #arried out" 9he %ain pi#tures are #olored and re#orded as '80J480 #% or 11D2J840 #% )ra%e" 0n $eneral, pi#tures o) '40J480 #% are e%ployed and prepared )or the analysis" 9he %ain ai% o) this !or5 !as to redu#e the pi#ture noise, to uni)or%ly distribute li$ht intensity on the bubble sur)a#e, to pre.ent the brea5s up o) a bubble to %ore than one and to ha.e %ore #larity at the bubble ed$es" 9o obtain these ob&e#ti.es, s%oothin$-a.era$in$ )ilters, s%oothin$ lo!-pass and ?aussian )un#tion !ere adopted" /ost )ilters are %athe%ati#al )un#tions in !hi#h alter the intensity o) a $rey pi;el in a##ordan#e !ith the intensity o) nei$hborin$ pi;els" 0n these types o) )ilters, the %a$nitude )or ea#h pi;el are #o%puted and #orre#ted !ith respe#t to the %a$nitudes o) the nei$hborin$ pi;els and as a result, a #oe))i#ient !as allo#ated )or the 5ernel )ilter" 0) the %a$nitude or #oe))i#ients !hi#h !as allo#ated )or the 5ernel .aries, the per)or%an#e o) the )ilter on the allo#ated .alue to the #entral pi;el !ill also alters" 0n $eneral, the #oe))i#ients !hi#h !ere de)ined on the 5ernel !ould de%onstrate the #ontribution o) nei$hborin$ pi;els on the allo#ated .alue )or the #entral pi;el" 9he si=e o) ea#h 5ernel is an indi#ation that ho! %any o) the nei$hborin$ pi;els #an e))e#t and alter the allo#ated .alue to the #entral pi;el" ,i$ure 7 de%onstrates a typi#al per)or%an#e o) this #ate$ory o) )ilters !hi#h also in#orporates a 5ernel o) the si=e 3J3 that has been de)ined )or it"

4tability o) the )roth is an indi#ation o) the de$ree o) the separation o) inor$ani# #o%pounds and sele#ti.ity o) pre#ious substan#es" >n the other hand, the instability o) the )roth #auses the bubbles to rupture, to loss solid #ontents and the re-e%er$en#e o) the )roth in the pulp phase" 6o!e.er, e;#essi.e stability #auses the sele#ti.ity o) pre#ious substan#es to be redu#ed (i"e", a )roth o) lo! $rade)" 0n )lotation pro#ess, %obility and .is#osity o) )roth are also o) $reat #on#ern" 9he inor$ani# hydrophobi# #ontents o) )roth are also depends lar$ely to the .is#osity o) the )roth" ,roth o) hi$h .is#osity #an delay the dis#har$e o) the hydrophobi# substan#es and hen#e #ould ha.e a ne$ati.e e))e#t on the per)or%an#e o) the pro#ess A13B" ,roth bubbles o) polyhedral and ellipti#al #on)i$urations ha.e a hi$h .is#osity and lo! %obility" 9here)ore, these types o) )roth ha.e hi$her )roth $rade and lo!er re#o.ery"

+. )"!dame!tal stages o( mage 'rocessi!g

8 typi#al i%a$e pro#essin$ )lo!#hart is sho!n in ,i$ure '" ,unda%entally, i%a$e analysis %ethod is di.ided into the )ollo!in$ #ate$ories: pro#essin$ o) the pri%ary sur)a#e (!hi#h in#ludes the stored i%a$es and pre-pro#essin$), pro#essin$ o) the inter%ediate sur)a#e (!hi#h in#ludes i%a$e se$%entation) and re#o$nition and e;tra#tion o) the essential )eatures )ro% the ori$inal i%a$e" Pro#essin$ o) the abo.e sur)a#e in#ludes the #lassi)i#ation and des#ription o) the i%a$e )eatures" Pi#tures #an be ta5en )ro% the )roth phase by a handy #a%" 0nstallation o) a )ra%e $rabber on the #o%puter #ould trans)or% an analo$ si$nal to di$ital one" ,urther%ore, i%a$e analyses and pro#essin$ #ould also be #arried out on the sa%e #o%puter"

8%eri#an Journal o) *he%i#al +n$ineerin$ 2013; 1(4): 70-78

7D

7ernel

,ilterin$ )un#tion

*entral pi;el 3ei$hbors

Figure 7. 0iltering mechanism

%odi)ied i%a$e, boundaries !ould appear in a %il5y #olor !ith a bla#5 ba#5$round" 0n the se#ond #ase, )ilter !ould #ause #han$es si%ilar to the abo.e one" 6o!e.er, in the pre#edin$ #ase the bubble !ould appear in %il5y #olor !ith a bla#5 boundary" @y %eans o) the #o%binin$ the %odi)ied i%a$es throu$h the aid o) the operators su#h as >G or 83(, one #ould obtain an enhan#ed results" 4.3. Frot" Feature Identi#ication and $%aluation ,roth )eature is a part o) an i%a$e that #an be identi)ied in it and then be #ounted or %easured" 0n a )roth i%a$e, ea#h bubble is #onsidered as a )eature" ,eature identi)i#ation stands )or identi)i#ation o) #ertain )eatures in an i%a$e" @y adoptin$ proper thresholin$ )un#tion, one #ould di.ide an i%a$e into t!o parts: one is the )eature =one and the other the distin#tion =one" 9o isolate the i%a$e )eatures, one #an adopt a thresholin$ )un#tion" 0n $eneral, thresholin$ )un#tion is e%ployed )ollo!in$ the utili=ation o) ed$e )un#tion" 9his pro#edure enables )eatures to be identi)ied )ro% the ba#5$round and to enhan#e the per)or%an#e" 9he )un#tion !hi#h is e%ployed is #alled %etri# thresholin$ )un#tion" 0n #ontrast to %anual %ethod, the use o) auto%ati# thresholin$ )un#tion does not re:uire the ad&ust%ent o) %a;i%u% or %ini%u% o) the thresholin$ )un#tion" 8s sho!n in ,i$ure 10, an auto%ati# al$orith% %ethod utili=es i%a$e histo$ra% )or the assess%ent o) the threshold .alue" 0n this )un#tion, the .alue )or the threshold is #alled 5, and )ro% that, the )ollo!in$ #orrelation #ould be %ini%i=ed:
h(i)

4.2. Bubbles Boundaries Considerations +stablishin$ bubble boundaries in an i%a$e is one o) the %ost i%portant and also #o%ple; sta$es in an i%a$e analysis pra#ti#e" 9his is due to the #o%ple; nature o) )roth te;ture and the une.en re)le#tion o) the li$ht on the bubble sur)a#e" ,ollo!in$ the pre-pro#essin$ o) the %ain i%a$e, the ed$e dete#tion sta$e has to be #arried out" 9his is a#hie.ed by adoptin$ Iapla#ian ed$e dete#tion )un#tion" ,i$ure 8 illustrate the e))e#t o) this )un#tion on the i%a$e a)ter bein$ pro#essed"

Figure . 2he effect of edge detection function on the froth image

9his )un#tion is a )ilter o) linear type !ith a hi$h pass #ate$ory" 9his type o) )ilter atte%pts to e;tra#t out the bubble boundary (!ith respe#t to the li$ht intensity $radient bet!een the bubble ed$e boundary and bubble itsel)) and hen#e to bri$hten up the i%a$e" Iapla#ian )ilter e%ploys an operator o) a se#ond order di))erential" 0n $eneral, they utili=e 5ernels o) di))erent resolutions (e"$", 3J3, DJD or 7J7) !ith di))erent ele%ents" 8 typi#al 5ernel o) resolution 3J3 is sho!n in ,i$ure H, !here a, b, # and d are inte$er nu%bers"

*lass 0 5
?ray Ie.el <alue

*lass 1 i

Figure 1&. A histogram picture

a b c

d , d

c b a

h(i) (i 1) =
i=0

i=#

i="1 i=#+1

h(i) (i )
2

(1)

0n this e:uation, 1 is the a.era$e li$ht intensity on all pi;els in the ran$e o) 0 to 5 and 2 the a.era$e .alue o) all pi;els in the ran$e o) 5F1 to 2DD" ,urther%ore, h (i ) is the nu%ber o) pi;els ha.in$ the intensity o) i and 5 desi$nates the .alue o) the $ray le.el !hi#h ha.e been allo#ated )ro% the threshold )un#tion" 4.4. Binar' (or)"olog' 0n spite o) the )a#t that in binary se$%entation, the bubbles appear to be deta#hed )ro% the ba#5$round" 6o!e.er, due to the #o%ple; nature o) )roth te;ture the probability o) the #o%plete separation o) bubbles in pre.ious

Figure !. A t!pical #ernel of resolution 343

(ependin$ on !hether the #entral pi;el ; is e:ual or $reater than absolute .alue o) the s:uare o) the nei$hborin$ pi;els #oe))i#ient, )ilter !ould ha.e t!o di))erent e))e#ts" 0n the )irst #ase, )ilters #auses pi;el re#o.ery o) those in !hi#h possesses %ore li$ht intensity .ariations" 9hese type .ariations #an be ori$inated due to the sharp ed$es, boundaries bet!een the bubbles and the noises" 0n the

7'

3asser 4a$hatolesla%i and 6a&ir 7ari%i:

0%a$e 8nalysis 9e#hni:ue as a 9ool )or +;tra#tin$ ,eatures )ro% the *opper 4ur)a#e ,roth in the ,lotation Pro#ess

sta$es are i%possible" ,urther%ore, o!in$ to the dissi%ilar e))e#ts o) pre.ious al$orith%s on the i%a$e stru#ture, undesired )eatures su#h as noises, dendrites and #onta#t o) bubbles on the i%a$e ed$e !ould be $enerated" /orpholo$i#al )un#tions #ould ha.e an i%%ense si$ni)i#an#e on the redu#tion or eli%ination o) these undesired )eatures on the i%a$e and hen#e %odi)yin$ the i%a$es and the si=e o) bubbles" 9he pro#ess o) binary %orpholo$y #an in#lude the ele%entary pro#ess o) pi;el erosion and dilation or a #o%bination o) an ad.an#ed pro#ess" 9hese )un#tions #ould be %anipulated and e;e#uted !ith the aid o) stru#turin$ ele%ents on the i%a$e and bubble ed$es !hi#h are allo#ated )or these )ilters" @y %odi)yin$ the si=e o) stru#turin$ %atri#es and allo#ation o) appropriate .alues to the stru#turin$ ele%ents o) these %atri#es, the #on)i$urations o) the bubbles and the !ay in !hi#h they #ould be separated %i$ht be %anipulated and %odi)ied" 0n ea#h o) the stru#turin$ ele%ents, three para%eters are des#ribed !hi#h #ould identi)y ho! a pi;el in the operation o) %orpholo$y is pro#essed" 9hese para%eters are: the si=e o) the stru#turin$ ele%ents o) the %atri#es, the %a$nitude o) these %atri#es and the !ay in !hi#h the nei$hborin$ pi;els to the #entral one ha.e been pro#essed are sele#ted" 8 typi#al o) this is a 3J3 %atri#es o) %a$nitude one" (ue to the li$ht intensity o) these pi;els, the sele#tion o) the stru#ture )or the nei$hborin$ pi;els has an i%%ense e))e#t on the per)or%an#e o) the )un#tions on the #entral one" 9he per)or%an#e o) a 3J3 %atri#es on the i%a$e (by #onsiderin$ the type o) stru#tures) are sho!n in ,i$ure 11" 0n so%e o) these )un#tions, it is also re:uired that the para%eter o) #onne#ti.ity to be #onsidered" 9he sele#tion o) this ite% #ould help the operator to identi)y a #ertain nu%ber o) the nei$hborin$ pi;els and be #lassi)ied as a sin$le i%a$e" 0n ,i$ure 12, the per)or%an#e o) this )un#tion on the i%a$e pro#essin$ is illustrated"
4 tru#tur in$ + le% ent 0 1 0 1 1 1 0 1 0 0% a$e

(a)

(b) Figure 12. 2he effect of a) elementar! and b) ad5anced morpholog! functions on the image processing

4.5. *etermination o# +eometric Parameters #or Bubbles 9he %ost i%portant para%eters that e;tra#ted )ro% the bubble i%a$e $eo%etry are: the nu%ber o) bubbles in the spe#i)ied =one, the a.era$e bubble si=e (i"e", e:ui.alent dia%eter), the distribution o) the bubble si=es (i"e", the standard de.iation the bubble si=es), #ir#ularity )a#tor, bubbles elon$ation (i"e", e:ui.alent ellipse dia%eter) and the orientation o) bubbles !ith respe#t to the #ell ed$e" 9able 2 illustrates the $eo%etri# para%eters o) typi#al i%a$e )roth"
Table 2. 'ist of parameters and their 5alues according to a t!pical image +:ui.alent dia%eter 2"H1D %% 1"1H1 4"DD2 %%

p1

p2 p0 p7

p3 pD p8 p 0 = 2 ( p0 , p 2 , p 4 , pD , p 7 ) 4:u are Pi; el ,ra%e

p4 p' (a)

*ir#ularity )a#tor +llipse %a&or a;is

4tru#tu rin $ +le% ent 0 1 0 1 1 1 0 1 0

0% a$e

p1

p2 p0 p' 6 e; a$ on al Pi;el ,ra%e p4 = 2 ( p 0 , p2 , p3 , p4 , p' ) p0

49(+< o) the bubble si=es /ean 6ue #olor #o%ponent 49(+< bubble #olors

2"23H %%

p3 pD

2H"008

(b)

21"1H8

Figure 11. 2he operation of transforming the function with the help of structuring elements and the structure that ha5e been chosen for the image for a) s6uare and b) he,agonal elements

4.6. Color ,nal'sis *lassi)i#ation o) a #olor #an be #hara#teri=ed by G?@ (i"e", red, $reen and blue), 64I (i"e", hue, saturation and

8%eri#an Journal o) *he%i#al +n$ineerin$ 2013; 1(4): 70-78

77

lu%inan#e), 640 (i"e", hue, saturation and intensity) or 64< (i"e", hue, saturation and .ariety)" 0n G?@ #oordinate, all the three #olors are re:uired si%ultaneously" 6o!e.er, in the 64I #oordinates the portrayal o) #olor is only done !ith the 6" >n the other hands, G?@ #oordinates ha.e no intrinsi# superiority o.er the 64I" 6o!e.er, the spe#ial )eatures o) 64I #oordinate ha.e #aused it to be #hosen )or the %a&ority o) i%a$e pro#essin$ appli#ations" 0n the pi;els #olor en.iron%ent, 6, 4 and I are )eatures that distin$uish one #olor )ro% another" 6ue #oordinate #orresponds to !a.elen$th ran$e )or a #olor, there)ore it si$ni)ies a #olor" 9he 4 #o%ponent si$ni)ies the relati.e saturation o) a #olor" 8 #olor other than !hite has one hundred per#ent saturation" I #o%ponent si$ni)ies the rate o) lu%inan#e o) the i%a$e" 9he #olor o) an i%a$e depends on the #o%bination o) 6 #o%ponent and de$ree o) its i%purity" 9he relation bet!een these t!o si$ni)ies the de$ree o) the bri$htness o) the #olor" 0n analy=in$ the #olors, initially the histo$ra% o) i%a$e %ust be #onstru#ted in 64I #oordinates" 0n these $raphs, the .alues o) #olor pi;el ele%ent are plotted a$ainst the a.ailable pi;els in the i%a$e" Kith the help o) these $raphs, statisti#al )a#ts about the #olors as a set o) pi;els #ould be e;tra#ted (as de%onstrated in 9able 2)" 9here)ore, the utili=ation o) #olor analysis #ould be e%ployed as si%ple te#hni:ue in di))erent )lotation pro#esses and #onse:uently )or the e.aluation o) di))erent ores in the )roth"

0n order to analy=e the #olors, histo$ra%s in 64I #oordinates should be prepared" 4ubse:uently by adoptin$ this type o) histo$ra%, the a.era$e .alue o) #olor #o%ponent and its standard de.iation #ould be deter%ined" 0t !as also )ounded that )ro% the 5no!led$e o) the )roth i%a$e analyses, .arious types o) )roth #ould be #ate$ori=ed both :uantitati.ely and :ualitati.ely" 6en#e, )ro% the abo.e relation, the operatin$ #onditions and the per)or%an#e o) the #ell #ould be studied" 9he )indin$s o) this resear#h #ould also be i%ple%ented to!ards the trainin$ o) the operators"

Ac$!owledgeme!ts
Ke $rate)ully a#5no!led$e the support o) the depart%ent o) resear#h and de.elop%ent o) 4ar#hesh%eh *opper Plant )or this resear#h and (r" Gahi%i, (r" 4hira=i and (r" 8%iri )or their #ooperation and help)ul su$$estions"

Nome!clat"re
@P

h (i )

@a#5 propa$ation 3u%ber o) pi;els ha.in$ the intensity o) i 6ue, saturation and intensity 6ue, saturation and lu%inan#e 6ue, saturation and .ariety <alue o) the $ray le.el /ultilayer per#eptron 9otal data input to &th neuron )ro% the lth layer Ged, $reen and blue Ge$ion o) interest 4u% o) s:uare errors <e#tor related to #onne#tion !ei$hts o) neurons in the ith layer

604 64I 64< 5 /IP G?@ G>0 44+

.. *o!cl"sio!s
0n this resear#h di))erent type o) )ilters, ed$e )un#tions, thresholdin$ )un#tions and %athe%ati#al %orpholo$y ha.e been e%ployed )or a series o) i%a$es" ,urther%ore, the proper al$orith% )or these #ases !as e.aluated" 0t !as also atte%pted to utili=e the proper )un#tions in ea#h sta$e" ,urther%ore, by #onsiderin$ the )roth te;ture, the proper type o) )ilters )or the ele%entary and )inal i%a$e pro#essin$ sta$es has been e%ployed" 9he )indin$s o) this !or5 re.eal that by applyin$ an appropriate )ilter on the i%a$e in the ele%entary pro#essin$ sta$es, a better out#o%e )or the ed$e dete#tion sta$es #ould be resulted" >n the other hands, thresholdin$ )un#tions are .ery sensiti.e" 6en#e, by s#rutini=in$ the per)or%an#e o) the pre.ious sta$es, one #ould establish and i%ple%ent a proper )un#tion" 9he sele#tion and hen#e #o%putation o) the upper and lo!er .alue )or threshold li%it #ould be #arried out by hand" 6o!e.er, this is not )easible on the #ontinuous i%a$e pro#essin$ sta$es" 9here)ore, )or the pre#edin$ #ase, it is re#o%%ended to adopt auto%ati# threshold )un#tions" ?enerally spea5in$, %orpholo$y )un#tions #ould be e%ployed )or the )ollo!in$ #ases: i%pro.in$ pi#tures and bubble ed$es, separation o) #oales#ed bubbles, )ilterin$ noises and unne#essary )eatures o) the i%a$es" 0n all sta$es o) i%a$e pro#essin$, the )roth te;ture #ould appre#iably a))e#t the e%ploy%ents o) the appropriate 5ernels in )ilters, the stru#turin$ ele%ents and the type o) pi;els #onne#ti.ity in %orpholo$y )un#tions"

/ree$ &etters
1 2
8.era$e li$ht intensity on all pi;els in the ran$e o) 0 to 5 8.era$e .alue o) all pi;els in the ran$e o) +rror ratio o) to &th neuron )ro% the lth layer 5F1 to 2DD

#e(ere!ces
A1B 3" 4a$hatolesla%i, 6" 7ari%i, G" Gahi%i and 6"6"8" 4hira=i, L/odelin$ o) 9e;ture and *olor ,roth *hara#teristi#s )or +.aluation o) ,lotation Per)or%an#e in 4ar#hesh%eh *opper Plant, -sin$ 0%a$e 8nalysis and 3eural 3et!or5s,M 0nt" J" o) +n$ineerin$, pp" 1-10, 2002" ,"4" ?ou!s, *" 8ldri#h and ?"P"J" 4#h%it=, L9he -se o) /a#hine Iearnin$ 9e#hni:ues )or the +;tra#tion o) Pro#ess 7no!led$e )ro% 0ndustrial ,lotation Plants,M 2'th Pro#eedin$s o) the 8ppli#ation o) *o%puters and >perations Gesear#h in the /ineral 0ndustry, p" 7, 1HH'"

A2B

78

3asser 4a$hatolesla%i and 6a&ir 7ari%i:

0%a$e 8nalysis 9e#hni:ue as a 9ool )or +;tra#tin$ ,eatures )ro% the *opper 4ur)a#e ,roth in the ,lotation Pro#ess A8B ?" @oni)a=i, P" /assas##i, 8" /eloni, L3( ,roth /odelin$ by 0%a$e 8nalysis,M Pro#" NN0 0nt" /in" Pro#" *on$", @8a-178-@8a-18H, 2001" 6"6"8" 4hira=i, L8ppli#ation o) )roth i%a$e analysis in %ineral industry,M Ph( 9hesis, 1HH7"

A3B

P" Pelletier, J" 9essier and *" (u#hesne, L8ppli#ation o) 3u%eri#al 0%a$e 8nalysis to Pro#ess (ia$nosis and Physi#al Para%eter /easure%ent in /ineral Pro#esses Part 0: ,lotation *ontrol @ased on ,roth 9e;tural *hara#teristi#s,M *entenary o) ,lotation 4y%posiu%, p" 12, 200D" 4"G" ?rano, L4ele#tin$ ,rothers )or the ,lotation o) 4pe#i)i# >res by /eans o) @at#h 4#ale ,oa%in$ 9ests,0 /etallur$i#al Plant (esi$n and >peratin$ 4trate$ies , 8-40//, 200'" G"6" +strada-Gui=, L3eural 3et!or5s to +sti%ate @ubble (ia%eter and @ubble 4i=e (istribution o) ,lotation ,roth 4ur)a#es - 4ynopsisM 9he Journal o) 9he 4outhern 8)ri#an 0nstitute o) /inin$ and /etallur$y, .ol" H, p" H, 200H" *" 8ldri#h, L(e.elop%ent o) ,u==y Gule-@ased 4yste%s )or 0ndustrial ,lotation Plants by -se o) 0ndu#ti.e 9e#hni:ues and ?eneti# 8l$orith%s,M J480//, p" ', 200H" +"9" Koodburn, J"@" 4to#5ton and ("J" Gobbins,M <ision-@ased *hara#teri=ation o) 9hree Phase ,roths,M 0nternational *ollo:uiu% (e.elop%ents in ,roth ,lotation, .ol" 1, pp" 1-30, 1H8H"

AHB

A4B

A10B <"8" ?le%bots5i, L,lotation Pri%ary 4our#es,M 3e! 1or5, 1H72" A11B ("K" /ool%an, J"J" +5esteen, *" 8ldri#h and J"4"J" .an (e.enter, L9he 4i$ni)i#an#e o) ,lotation ,roth 8ppearan#e )or /a#hine <ision *ontrol,M 0nt" J" /iner" Pro#ess, .ol" 48, pp" 13D-1D8, 1HH'" A12B J"/" 6ar$ra.e, 3"J" /iles and 4"9" 6all, L9he -se o) ?ray Ie.el /easure%ent in Predi#tin$ *oal ,lotation Per)or%an#e,M /ineral +n$", .ol" H, pp" ''7-'74, 1HH'" A13B @"/" /oud$il, L*orrelation bet!een ,roth <is#osity and ,lotation +))i#ien#y,M /iner-/etallur$ Pro#ess, .ol" 10, 3o" 2, pp" 100-101, 1HH3"

ADB

A'B

A7B