Comparative Analysis of Flame Image Features for Combustion Analysis

Transcription

1 Ida Joural of Scece ad Techology, Vol 9(6), DOI: /jst/06/v96/79904, February 06 ISSN (Prt) : ISSN (Ole) : Comparatve Aalyss of Flame Image Features for Combusto Aalyss Ab P. Mathew *, A. Asoa, K. atr 3 ad D. Svaumar St. Joseph s College of Egeerg ad Techology, Pala, haraagaam Pravthaam Road, Choodacherry , Kerala, Ida; abmthw@gmal.com Departmet of Istrumetato, Aamala Uversty, Ruma ashmpathy Road, Egmore, Chea , Taml Nadu, Ida; asoa7@gmal.com, ds5@gmal.com 3 PSNA College of Egeerg ad Techology, Kothadarama Nagar, Natoal Hghway 09, Ddgul 646, Taml Nadu, Ida; rsha.batr@gmal.com Abstract acgroud/ Objectve: Ths artcle detfes the best feature of the flame vdeo, captured wth a camera wth frequecy respose vsble spectrum, from whch the flame temperature ca be estmated. Methods/Statstcal aalyss: The flame vdeos at dfferet ar ad fuel lets wth dfferet boler temperatures were recorded from a desel fred boler prototype. I the vdeo frames, the flame rego was localsed by testy based adaptve thresholdg. The correlato betwee boler temperature ad measures of cetral tedecy ad dsperso of dfferet colour chaels of the vdeo frames were vestgated. Fdgs: Amog the features of the flame vdeo, Stadard devato of blue chael grey levels above 3.95, varace greater tha 93 ad mea absolute devato (MAD) above could effcetly represet the rego of optmum combusto ar supply at whch boler temperature s maxmum above 684 degree Celsus. Rage of gree chael grey levels, terquartle mea, varace ad mea absolute devato of blue chael grey levels are the vdeo features exhbtg maxmum correlato (ρ>-0.96) wth boler temperature. Applcatos/Improvemets: The features of the flame vdeo whch are correlated wth ts temperature ca be utlsed to develop o-trusve methods of temperature measuremet. Ths wll eable effcet cotrol of combusto process. Keywords: Combusto, Flame Image Processg, Flame Temperature Measuremet, Image Features, Vdeo Processg. Itroducto Combusto motorg ad cotrol has bee vtal dustral bolers. Motorg combusto process has always bee a area of fascato for researchers over the decades. Image processg has bee the tool for aalyss of the flame most of the o-trusve approachees -. A attempt to study combusto process for dfferet fuels through aalyss of flame mage s made by. 3-D recostructo ad vsualzato of the flame structure are doe usg mages captured from multple drectos may of the studes -5. I 3 proposed a mult-camera based magg system for 3-D vsualzato ad charactersato of the flame frot structures of a turbulet gaseous flame. 3-D temperature proflg ad vsualzato 4,5 s a advacemet of the mere recostructo of the 3-D flame structure. I 6 developmet of mage based ole motorg ad charactersato of fossl fuel fred flames s detaled. I 7 the temperature profle ad emssvty mages are derved from a colour mage. Temperature estmato ad dstrbuto the flame feld s detaled 8. Flame mages tae by hgh dyamc rage cameras processed for recostructg the flame radace exstece feld s detaled 9. It s challegg to detfy the vdeo features through whch the flame temperature ca be measured or to detfy the rage of these features correspodg to optmum combusto. Usg a commercally avalable vdeo camera of vsble spectrum, the tas s challegg tha usg a IR camera. Aalyss of combusto from *Author for correspodece

2 Comparatve Aalyss of Flame Image Features for Combusto Aalyss vdeo usg IR camera had bee doe by 0. I a radato model was developed to relate the flame mages wth the -D temperature dstrbuto. It s worthy f a relato betwee boler temperature ad features extracted from the flame vdeo, captured wth a camera of vsble spectrum ca be establshed. The objectve of most of the lterature had bee aalyss of combusto status usg features extracted from the flame mage or vdeo wthout tag the combusto parameters to accout. Combusto parameters le ar let flow rate ad fuel flow rate has predomat fluece o boler temperature ad combusto status. Through ths expermetal research, a attempt has bee made to aalyse the features of flame systematcally at regular tervals of let ar varatos ad thereby at dfferet boler temperatures. The regular varato of let ar flow tur geerates dstct flame structures ad testy features. The vestgato s to detfy the vdeo features whch are correlated to the boler flame temperature ad the rage of these features whch characterse the optmum combusto.. Expermetal Cofgurato The expermet comprses capturg vdeo of turbulet boler flame, at ow combusto codtos ad boler temperature for extractg vdeo features. Expermets are doe a desel fred, col type fully automatc-stat steam geeratg boler whch s show Fgure. Ilet ar supply to the boler s suppled by a forced draft (FD) fa rotatg at 480 rpm wth let opeg cotrolled by a adjustable damper. The desel pressure at the outlet of fuel jecto pump s mataed costat at.5g/ cm durg the vdeo acqusto. At ths costat desel supply, the amout of combusto ar to the boler s vared gradually by chagg the percetage opeg of the damper. Smultaeously vdeo of the desel flame s captured. To correlate the flame vdeo features drectly to the boler temperature, stead of steam temperature, a K-type thermocouple s serted to the boler to measure the flame temperature. Vdeo of flame s captured at a frame rate of 30 frames per secod by a CCD camera through the sght glass of boler. y matag the amout of combusto ar at dfferet levels, the boler temperature s oted ad correspodg vdeo s captured for fve secods each. Each vdeo segmet correspodg to a partcular level of combusto ar comprses approxmately oe hudred ad ffty frames. Features from each frame are extracted ad cumulatve feature of the vdeo segmet s estmated from the meda, arthmetc mea ad quadratc mea of dvdual frame features, to elmate fluece of outlers. Schematc of expermetal setup s gve Fgure. The fgure explas the let parameters as ar let ad fuel let for combusto ad a CCD camera s utlsed for capturg vdeo sgals. The temperature measured s fed to the computg system for further processg. 3. Segmetato of ocal Flame Rego Istead of the global frame features, local features of the flame rego are extracted. The flame rego has bee segmeted from red, gree ad blue chaels of whole Fgure. Desel fred boler prototype. Fgure. Expermetal Cofgurato. Vol 9 (6) February 06 Ida Joural of Scece ad Techology

3 Ab P. Mathew, A. Asoa, K. atr ad D. Svaumar frames, by masg these chaels wth a multplcato mas. Ths s possble as the grey level testes at the flame rego s hgher tha the other mage regos. The bary multplcato mas s geerated by thresholdg the greyscale mage obtaed from the RG frame. I fact, the testy features of each frame are uque ad dfferet hece a costat threshold may ot be applcable to the etre frames the vdeo segmet. To determe the optmum threshold sutable for each frame, Otsu s method 3 s employed. The multplcato mas to solate R, G ad chael grey levels the local flame rego s geerated by thresholdg the grey scale verso of the vdeo frame usg the optmal Otsu s threshold. The mathematcal strategy adopted for estmatg optmum threshold s as, et s the maxmum possble grey level the greyscale verso of the vdeo frame. The ormalsed grey level hstogram of grey scale vdeo frame, P N P 0, P () where N s the total umber of pxels the greyscale frame, s the umber of occurrece of grey level ad p s ts dscrete probablty desty, gve, { 0,,,..}. If the grey levels are assumed to be of two classes C0 ad C, separated by a threshold,, so that, C0 {0,, } ad C { +..}. Ths meas that the testes preset the greyscale vdeo frame belogs to two classes or regos, local flame rego ad o-flame rego. The the probabltes of class occurrece ad the class mea levels, respectvely, are gve by, µ µ 0 ω ω Pr(C ) ω() () 0 0 K P Pr( C ) P ω( ) + µ P Pr( ) ( ) C ω ω( ) 0 0 µ µ p T ( ) Pr( ) C 0 ω ω( ) + + ω( ) P (3) (4) (5) (6) µ( ) p (7) Where w() ad m() are the zeroth ad frst order cumulatve momets of the hstogram up to the th level, respectvely. µ µ () T p (8) m T s the mea of the grey levels the grey scale verso of R,G, vdeo frame. ωµ + ωµ µ, ω + ω, (9) 0 0 T 0 The varaces of testes the flame ad o-flame regos respectvely are gve by, σ ( µ ) Pr( C ) ( µ ) P ω 0 (0) σ ( µ ) Pr( C ) µ ω ( ) P () + + To evaluate the performace of the threshold, dfferetatg the flame ad o-flame regos the greyscale frame, followg dscrmat crtero s used, where W T W W λ σ σ K σ σ η σ σ W 0 0 () σ ω σ + ω σ (3) σ ω ( µ µ ) + ω ( µ µ ) ω ω ( µ µ ) (4) 0 0 T T from (9) 0 0 σ T ( µ T ) P (5) where l, K ad h are wth class varace, betwee class varace, ad varace of total grey levels. At optmum threshold, the object fucto () maxmzes The dscrmat crtera maxmzg l, K, h respectvely, for are, however, equvalet to oe aother; e.g., K l + ad h l/(l + ) terms of l, because the followg basc relato always holds: W T σ + σ σ (6) It s otced that s ad w s are fuctos of threshold level, but s s depedet of. It s also oted that T s w Vol 9 (6) February 06 Ida Joural of Scece ad Techology 3

4 Comparatve Aalyss of Flame Image Features for Combusto Aalyss s based o the secod order statstcs whle s s based o the frst order statstcs. Therefore, h s the smplest measure wth respect to. Thus at the optmum threshold, the object fucto h maxmses. The optmal threshold whch maxmses h or equvaletly maxmses s, s selected through a sequetal search usg (6) ad (7), or explctly usg () (5): σ [ µ Tω ( ) µ ( )] () ω [ ω( )] ( ) (7) ad the optmal threshold * whch separates the flame ad o-flame regos the greyscale vdeo frame s σ * ( ) max σ ( ) (8) < The multplcato mas used to extract R G ad chael grey levels the local flame rego, ^ f fj (,) f( j,) 0 else (9) Where f(, j) s the grey scale verso of the vdeo frame. Fgure 4. Grey scale verso of Fgure 3. ^ Rj (,) Rj (,) ^ ^ Gj (,) Gj (,)*fj ( ^ j (,) j (,) (0) R^ ( j), G^ (, j) ad ^ (, j) are the grey levels of the R, G ad chaels the local flame rego. The orgal RG vdeo frame, ts grey scale verso, the bary multplcato mas geerated ad the segmeted flame rego s demostrated Fgure 3 to Fgure 6 cosecutvely. Fgure 5. Multplcato mas. Fgure 3. RG vdeo frame. Fgure 6. Segmeted flame rego. 4 Vol 9 (6) February 06 Ida Joural of Scece ad Techology

5 Ab P. Mathew, A. Asoa, K. atr ad D. Svaumar Segmetato of flame rego from the vdeo frame, extracto of local flame features ad the aalyss of the depedecy of vdeo features o the combusto ar supply ad boler temperature are performed Matlab. 4. Extracto of Features from Flame Vdeo The vdeo has bee captured at varous levels of combusto ad boler temperature by cotuously varyg the combusto ar flow. The flame rego s extracted from the vdeo frame. Itesty features ad hstogram features of the local flame rego of each frame s computed. As metoed earler features from dvdual frames each vdeo segmet s extracted. Features of dvdual frames a vdeo segmet are tegrated to reflect the features of the vdeo segmet. The term vdeo segmet refers to the flame vdeo captured for a partcular value of boler temperature ad combusto ar flow. Spread s a geometrcal feature extracted from the frame, whch represet the area of the flame. The spread of the flame each frame s computed by coutg the umber of s the correspodg bary multplcato mas. The wavelegth of the spectral emssos of the flame chages wth respect to the temperature. Hece, the testy or hstogram features ca better reflect the combusto effcecy rather tha textural or geometrcal features. The testy features extracted from the vdeo frames are the measures of dsperso ad cetral tedecy of grey levels R, G, ad chaels of the segmeted flame rego. These measures of cetral tedecy ad dsperso computed are mode of local testy, arthmetc mea, quadratc mea, root mea square testy, meda, geometrc mea, varace, stadard devato, MAD, rage, IQR ad terquartle mea (IQM). The hstogram features extracted are sewess ad urtoss. The meda ad trmmed mea are robust statstcs but arthmetc mea, geometrc mea ad harmoc mea are ot resstat to outlers. Geometrc mea ad harmoc mea are useful, whe the sample s dstrbuted log-ormal. The uderlyg dstrbuto of local-flame testy the vdeo frame remas uow ad perhaps, the dstrbuto may be dfferet for dvdual frames as the flame s turbulet. The probablty desty fucto of the varato of local flame features wth respect to the dex of the frames a vdeo segmet correspodg to certa ar-flow rate obvously may ot be ormally dstrbuted. The probablty of terferece of outlers s hgh because of the olear behavour of the turbulet desel flame ad heret desg aspects of desel fred boler. If the sample dstrbuto s ormal oe, the arthmetc mea may be optmal.e. mmum varace ubased estmator. The presece of outler dsplaces the average away from the cetre of the rest of the data by a arbtrarly large dstace. A small fracto of aomalous measuremets, wth abormally large devato from the cetre may chage the mea of the populato substatally. The trmmed mea s a famly of measures of cetral tedecy le arthmetc mea ad t s a robust statstcs. Trmmed or trucated mea s the mea of the populato dscardg gve parts of the probablty dstrbuto or sample at the hgh or low ed, typcally fve percetage, te percetage or twety-fve percetage based o the sample sze. Twety-fve percetage trmmed mea s ow as ter quartle mea. Meda ca be regarded as a fully trucated mea whch s most robust. Trmmed mea s sutable for mxed dstrbutos ad heavy taled dstrbutos le Cauchy dstrbuto. Amog the popular measures of dsperso, the stadard devato ad varace are optmal mmum varace ubased estmator for ormally dstrbuted data. These are heavly sestve to outlers whereas MAD, to a certa extet s mmue to outlers. IQR s aother measure of dsperso whch s robust to outlers ad sutable for uow probablty dstrbutos. The mode of local flame testy s the most frequet value or the value exhbtg maxmum probablty desty. Whe there are multple testes wth equal cdece, the mode returs the smallest of these testes. Root mea square or quadratc mea s a measure of cetral tedecy most sutable for a varyg quatty especally whe the varatos are susods. Sewess s a hstogram feature that may oscllate betwee postve ad egatve values based o whether the local flame testy hstogram of RG chaels are rght taled or left taled. Sewess s the thrd momet of dsperso ad expresses the lop-sdedess or symmetry of probablty desty fucto or testy hstogram. Kurtoss s the fourth momet of dsperso ad t s a measure of peaedess of the probablty desty fucto of R, G ad chael grey levels the segmeted flame rego. Rage, whch s aother measure of dsperso, s the dfferece betwee mmum ad maxmum of the samples ad t s proe to outlers. The most robust measure of dsperso, IQR otherwse called mdspread or mddle-ffty s the dfferece betwee upper ad Vol 9 (6) February 06 Ida Joural of Scece ad Techology 5

6 Comparatve Aalyss of Flame Image Features for Combusto Aalyss lower quartles of the R, G ad chael grey levels, the segmeted flame rego. MAD x µ () 3 4 IQM x + 4 () IQR ( Q3 Q ) (3) Where Q3 s the thrd Quartle ad Q s the secod Quartle Geometrc Mea m Harmoc Mea m Quadratc Mea m 3 x (4) x (5) x ( ) (6) Where Arthmetc Mea µ x, x s the grey levels of R, G ad chaels the segmeted flame rego of vdeo frame, {,,.. } where s the total umber of pxels the segmeted local flame rego σ µ ( x ) (7) σ µ ( x ) (8) Meda x of a sorted set wof samples x {x, x, x 3 x } The umber of pxels gettg dstrbuted to dfferet grey level bs of the R, G ad chaels chages respose to the varato furace temperature. Ths chages the symmetry ad peaedess of hstogram. The sewess s gve by, β 3 ( x µ ) ( x µ ) 3 (9) The urtoss s gve by, β 4 ( x µ ) ( x µ ) (30) To detfy the best of statstcal vdeo features whch are robust to outlers ad ca characterse the effcecy of combusto, Pearso correlato coeffcet betwee the vdeo features ad the boler temperature s estmated. The Pearso correlato coeffcet (r) betwee boler temperature ad vdeo features s gve by µ µ ρ ( x )(y ) (3) σσ x Where x seres s the samples of boler temperature ad y seres s samples of vdeo segmet features. The rages of these features correspodg to the optmum combusto are also estmated. 5. Expermetal Results As apparet Fgure 9 ad Table wth the crease combusto ar flow, followg a traset shoot-up, a gradual decay of boler temperature s observed. A comparatve evaluato of the relato betwee features of R, G ad chael grey levels ad boler temperature has bee performed. The blue chael grey-level varatos exhbt comparatvely close relato to the combusto effcecy, as obvous Fgure 7 to Fgure 8. The maxmum of blue chael grey levels ad the mea of blue chael greylevels fall abruptly wth crease temperature as vsble Fgure 7. Ths fall the average ad maxmum blue chael eergy may be accordace wth the crease IR spectral emsso. It has bee observed that the features vary sgfcatly amog dfferet frames of the same vdeo segmet. Perhaps these sgfcat varatos may be a reflecto of turbulet ature of the flame. Due to ths turbulet ature of the flame, some of the frames may have features far from majorty of the frames the same vdeo segmet. The quadratc mea, meda ad arthmetc mea tegrate the features of dvdual frames to form the cumulatve feature of vdeo segmet. The robust statstcs, quadratc mea ad meda elmate the probable y 6 Vol 9 (6) February 06 Ida Joural of Scece ad Techology

7 Ab P. Mathew, A. Asoa, K. atr ad D. Svaumar Table. Rage of vdeo segmet features ad correspodg temperature. Vdeo Segmet feature Rage of the feature Rage of boler temperature Mea testy of blue chael 77.76< >660 Geometrc mea of blue chael 68.9< >660 Harmoc mea of blue chael 57.95< >660 Quadratc mea of blue chael 84.76< >660 Meda of blue chael 77 < >660 Mode of red chael 55 >647 Mode of blue chael 78< >660 Mmum testy of gree chael Mmum testy of blue chael Maxmum Itesty of blue chael 88< >660 8< >660 68< >650 Rage of gree chael >63 >660 Rage of blue chael 57< >647 Iterquartle rage of blue chael Stadard devato of blue chael 60< > < >684 Varace of blue chael 93< >684 Mea absolute devato of blue chael 30.38< >684 Fgure 8. Mea Absolute devato of blue chael grey levels versus boler temperature. Fgure 9. Varace of blue chael grey levels versus boler temperature. Fgure 7. Maxmum testy blue chael grey levels versus boler temperature. Fgure 0. Stadard devato of blue chael grey levels versus boler temperature. Vol 9 (6) February 06 Ida Joural of Scece ad Techology 7

8 Comparatve Aalyss of Flame Image Features for Combusto Aalyss Fgure. Mmum testy of blue chael grey levels versus boler temperature. Fgure 4. Harmoc mea of blue chael grey levels versus boler temperature. Fgure. Mea testy of blue chael grey levels versus boler temperature. Fgure 5. Meda of blue chael grey levels versus boler temperature. Fgure 3. Quadratc mea of blue chael grey levels versus boler temperature. Fgure 6. Mode of red chael grey levels versus boler temperature. 8 Vol 9 (6) February 06 Ida Joural of Scece ad Techology

9 Ab P. Mathew, A. Asoa, K. atr ad D. Svaumar Fgure 7. Maxmum testy of blue chael grey levels versus boler temperature. caot be used to detfy the optmum combusto ar flow. ut the mode of the red chael reaches ts saturato oly at optmum arflow ad maxmum boler temperature. Mode of red chael grey levels s the grey level wth maxmum probablty desty the red chael. From the Fgure 6, t s perspectve; the feature, mode of red chael outperforms the rest of feature vectors effectvely portrayg the optmum combusto ar flow. From the Fgure9, t s evdet that the boler temperature falls after ad before 35% opeg of the ar let valve. Hece the optmum arflow ad optmum combusto s expected to be at 35% ar let valve opeg. The Table cota the rage of vdeo segmet features correspodg to the optmum boler temperature ad combusto ar flow. ut surprsgly eve for a vdeo of desel flame of extreme turbulece, arthmetc mea, meda ad quadratc mea of dvdual frames the vdeo segmet are roughly equal. Ths meas that the procedure tself s ot proe to outlers. The rage of features of the vdeo segmet correspodg to optmum boler temperature show Table s the arthmetc mea of dvdual frame features. The umercal values of stadard devato, varace ad MAD of the blue chael grey level features of vdeo segmets furshed Table ad vsual specto of the Fgure 8 to Fgure 0 reveals that the optmum boler temperature s mataed whe the stadard devato, varace ad MAD are below 3.95, 93 ad 30.38, respectvely. elow ths threshold, the measured boler temperature s above 684 degree cetgrade. The maxmum boler temperature reached s Fgure 8. Rage of gree chael grey levels versus boler temperature. fluece of outlers, caused by the flame turbulece. Eve though the fluece of outlers due to the turbulece of the flame s expected, quadrature, arthmetc mea ad meda of the dvdual frame features do ot exhbts sgfcat devatos. The measures of cetral tedecy ad dsperso of blue chael grey level decay sgfcatly wth crease combusto ar flow ad boler temperature. More terestgly the slope of decay creases substatally as the boler temperature reaches ts optmum value. Maxmum testy of the red chael reaches, ts saturato, 55, early before the optmum boler temperature. Hece maxmum testy of the red chael Fgure 9. Temperature versus valve opeg. Vol 9 (6) February 06 Ida Joural of Scece ad Techology 9

10 Comparatve Aalyss of Flame Image Features for Combusto Aalyss 6930C. The optmum boler temperature s a arbtrarly defed rage betwee 6840C ad the maxmum boler temperature. The spread s ot a apt feature to characterse the combusto effcecy desel fred boler ad the spread exhbt a ter frame dsperso wth a mea of 9986 pxels ad stadard devato of Fall stadard devato ad varace of grey levels the blue chael coform that the blue testes get coverged to the mmum rage as the emsso spectrum get shfted to IR. Pearso correlato coeffcets amog boler temperature ad flame features are exhbted Table ad Table 3. Table 3 comprses Pearso Correlato Coeffcet for vdeo features whch are less correlated wth boler temperature ad Table demostrates Pearso Correlato Coeffcet for hghly correlated vdeo features. The vdeo features exhbtg Pearso Correlato Coeffcet above 0.9 s referred as hghly correlated ths cotext. The vdeo features of blue chael grey levels are learly correlated to boler temperature. Rage of gree chael grey levels, IQR, varace ad MAD of blue chael grey levels exhbts correlato close to Table. Pearso correlato coeffcet betwee boler temperature ad hghly correlated flame features. Flame features Pearso correlato coeffcet Arthmetc mea of blue chael Geometrc mea of blue chael Harmoc mea of blue chael Quadratc mea of blue chael Meda of blue chael Meda of red chael Mode of red chael 0.98 Mode of blue chael 0.94 Mmum grey level of blue chael Maxmum grey level of blue chael Rage of gree chael Iter quartle rage of blue chael Stadard devato of blue chael Varace of blue chael Mea absolute devato of blue chael Table 3. Pearso correlato coeffcet betwee boler temperature ad less correlated flame features. Flame features Pearso correlato coeffcet Arthmetc mea of gree chael 0.77 Arthmetc mea of red chael Geometrc mea of gree chael Geometrc mea of red chael Harmoc mea of gree chael Harmoc mea of red chael Quadrature mea of gree chael 0.40 Meda of gree chael 0.83 Mode of gree chael Mmum grey level of gree chael Mmum grey level of red chael Maxmum grey level of gree chael Maxmum grey level of red chael Rage of blue chael Rage of red chael Iter quartle rage of gree chael Iter quartle rage of red chael Stadard devato of gree chael Stadard devato of red chael Cocluso The correlato betwee the features extracted from flame vdeo ad the boler temperature s statstcally evaluated. The measure of cetral tedecy ad dsperso of blue chael grey levels are correlated more learly to boler temperature tha red ad gree chael grey levels. However the rage of gree chael grey levels ad mode of red chael grey levels exhbted apprecable correlato, almost smlar to the blue chael features. The value of the Pearso correlato coeffcet esures that the vdeo features are learly related to the boler temperature. Hece these features ca be employed for flame temperature measuremet ad vdeo based combusto cotrol. 7. Future Scope The boler temperature depeds equally o combusto effcecy ad the rate at whch the heat s swept away from the boler. The latter s drectly related to ar flow rate. 0 Vol 9 (6) February 06 Ida Joural of Scece ad Techology

11 Ab P. Mathew, A. Asoa, K. atr ad D. Svaumar Hece, the boler temperature aloe, s ot a effectve dcatve for optmum combusto, rather, flue exhaust aalyss alog wth the boler temperature would be a apt soluto, to detfy the optmum arflow whch offers perfect combusto of ar-desel mxture ad maxmum flame temperature. The temperature profle of ay flame s heavly correlated to the IR spectrum, emtted from the flame tha the vsble eergy. The depedecy of blue chael grey level varatos over the IR spectral chages may also be evaluated further. A mathematcal relato ca be establshed betwee correlated vdeo features ad boler temperature. The relatoshp betwee features extracted from the vdeo of the flame ad boler temperature may be cofrmed at dfferet fuel let rate also. 7. Referece. Cecerrado G, Pellcer A P, Martíez G, ad Rubo A. Dgtal mage processg for aalyss of flame characterstcs ad stablty, [Ole]. Accessed o /03/0; Avalable from: uo Z, Zhou H C. A combusto-motorg system wth 3-D temperature recostructo based o flame-mage processg techque, IEEE Trasactos o Istrumetato ad Measuremet. 007; 56(5): heemul H C, u G, Ya Y. Dgtal magg-based threedmesoal characterzato of flame frot structures a turbulet flame, IEEE Trasactos o Istrumetato ad Measuremet, 005; 54( 3): rsley P M, u G, Ya Y, Corwell S. Three dmesoal temperature measuremet of combusto flames usg a sgle moochromatc CCD camera, Proc. st IEEE Istrumetato ad Measuremet Techology Coferece. 005; 54(4): Zhou H C, ou C, Cheg Q, Jag Z, He J, Huag, Pe Z, u C. Expermetal vestgatos o vsualzato of three-dmesoal temperature dstrbutos a largescale pulverzed-coal-fred boler furace, Proceedgs of the Combusto Isttute. 005; 30 (): u G, Ya Y, Colech M. A dgtal magg based multfuctoal flame motorg system, IEEE Trasactos o Istrumetato ad Measuremet. 004; 53(4): Jag ZW, uo ZX, Zhou HC. A smple measuremet method of temperature ad emssvty of coal-fred flames from vsble radato mage ad ts applcato a CF boler furace, Fuel. 009; 88(6): u G, Ya Y, Corwell S, Rley G. Temperature proflg of pulverzed coal flames usg mult-colour pyrometrc ad dgtal magg techques, IEEE Trasactos o Istrumetato Ad Measuremet. 006; 55(4): Wag X, Wu Z, Zhou Z, Wag Y, Wu W. Temperature feld recostructo of combusto flame based o hgh dyamc rage mages, Optcal Egeerg. 03; 5:4360:- 4360:0. 0. Sujatha K, Vemath M, Pappa N. Flame motorg power stato bolers usg mage, ICTACT Joural o Image ad Vdeo Processg. 0; 0(04): ou C, Zhou HC. Deducto of the two-dmesoal dstrbuto of temperature a cross secto of a boler furace from mages of flame radato, Combusto ad Flame. 005; 43: odhe RG, Nadurar YY, Aat SS, aar S, Performace aalyss of separatg ad throttlg calormeter to determe qualty of steam o thermax desel fred water tube boler, Iteratoal Joural of Research Egeerg ad Appled Sceces (IJREAS). 04; 0(0): Otsu N., A threshold selecto method from gray-level hstogram, IEEE Trasactos o Systems, Ma, ad Cyberetcs. 979; 9():6 66. Vol 9 (6) February 06 Ida Joural of Scece ad Techology