, pp.42-46 http://dx.doi.org/10.14257/stl.2016.137.08 Infrred Imge Edge Detection bsed on Morphology- Cnny Fusion Algorithm Tng Qingju 1, Bu Chiwu 2, Liu Yunlin 1, Zng Jinsuo 1, Li Dyong 1 1 School of Mechnicl Engineering, Heilongjing University of Science nd Technology, Hrbin, 150022, P.R. Chin 2 College of Light Industry, Hrbin University of Commerce, Hrbin, 150028, P.R.Chin tngqingju@126.com Abstrct. Effectively extrcting the contour edge of the defect in the infrred imge cn relize the recognition of the geometricl fetures of the defect. In view of the trditionl Cnny edge detection lgorithm in the Guss filter vrince nd high nd low threshold selection need rtificil intervention, does not hve the dptive bility, nd its defects in the grdient clcultion, proposed method bsed on the improved Cnny opertor nd imge morphologicl fusion edge detection method. Using the improved Cnny opertor nd imge morphology to edge detection, the simultion results show tht the lgorithm hs good nti-noise bility, effectively improve the ccurcy nd integrity of the edge detection in the infrred imge, nd chieve better extrction of geometric fetures of the edge of the defect. Keywords: Morphology; Cnny opertor; edge detection; infrred imge. 1 Introduction Edge detection is the bsis of imge informtion extrction nd pttern recognition, which is n importnt content in the field of imge processing. Imge edge detection results directly ffect the effect of further imge processing, pttern recognition. Effectively extrcting the contour edge of the defect in the infrred imge cn relize the recognition of the geometricl fetures of the defect. In recent decdes, imge edge detection technology hs become one of the importnt reserch topics of digitl imge processing technology. Edge hs importnt informtion chrcteristics, ccurte nd relible edge detection method will be helpful for imge feture description, imge enhncement, imge segmenttion nd pttern recognition nd so on. Becuse of the noise nd the edge of the imge re high frequency signl, the generl method is very difficult to effectively seprte the two regions, good edge detection method is ble to filter out the noise while the edge is lso to be ccurte nd cler. So, the ccurcy nd integrity of edge detection is direct impct on the selection of the whole imge, nd the reserch on it hs become one of the hot spots in imge nlysis nd processing technology [1-3]. ISSN: 2287-1233 ASTL Copyright 2016 SERSC
2 Bsic Theory of Mthemticl Morphology nd Edge Detection Method Mthemticl morphology is mthemticl method, which is bsed on morphologicl structure elements to nlyze nd describe the geometry nd structurl properties. It is lso kind of nonliner processing system bsed on geometric lgebr nd set theory. The bsic ide of mthemticl morphology is to mesure the effectiveness of the method of the structurl elements of certin shpe to mesure the effectiveness of the method nd the method of filling in the trget imge. Mthemticl morphology cn eliminte the morphologicl nd structurl properties, which is not relted to the trget imge, while preserving the bsic nture of the shpe nd structure, nd chieve the purpose of simplifying the trget imge dt. The most bsic morphologicl opertors include corrosion, expnsion, open opertion nd close opertion [4-6]. 2.1 Morphology Expnsion A new form of structure element B, which hs certin form of structurl element B shifts distnce, is obtined. If A nd B re in the sme A, the B is expressed s set of elements tht stisfy the condition: ( ) { } D A = B? A A? B (1) 2.2 Morphology Erosion A new form of structure element B is obtined by B shifts distnce, which is composed of A, nd ll the collection of elements tht stisfy this condition is clled A is eroded by B. The formul is expressed s ( ) { } E A = B? A AQ B (2) 2.3 Open Opertion Using the sme structurl elements of the imge first erosion opertion, nd then the results of the method of expnsion is clled open opertion, the formul is expressed s A B = ( AQ B)? B (3) Copyright 2016 SERSC 43
Open opertion cn smooth the imge contour, weken the nrrow prt, removing the burr nd the isolted spots of the long nd thin protruding edges, disconnect between the trget nd so on., its min effect on corrosion is similr. 2.4 Closed Opertion Using the sme structurl elements to expnd the imge first, nd then the results of the method of erosion opertion is clled closed opertion, the formul is expressed s ( ) A B = A휬 B B (4) Close opertion cn lso be smoothed imge of the contour. Compred with open opertion, closed opertion is generlly used to fill the smll hole nd crck in the trget. The min function of the connection is similr to the expnsion effect, but it is lso the sme s the size of the trget. 2.5 Morphologicl Edge Detection Opertor The bsic ide of the trditionl morphologicl edge detection opertor is to do morphologicl grdient processing of the originl imge, so tht the gry level of the input imge is more cute, nd then the imge edge is detected. With the help of the bsic opertions of ll kinds of morphologicl opertions, the morphologicl grdient MG is usully expressed in the following forms: 1 2 3 ( ) = (? ) ( ) = - ( Q ) ( ) = (? ) ( Q ) MG A A B A MG A A A B MG A A B A B { } ( ) = 殞 (? ) 殞 - ( Q ) MG4 A min 薏 A B A, 薏 A A B (5) 3 Experiment nd Anlysis In this pper, the defects edge of infrred therml imge is extrcted. Fig.1 shows therml imge cptured by n infrred cmer SC7000. Fig.2 shows the imge fter gry level trnsformtion. Fig.3shows the edge detection result by Cnny opertor. Fig.4 shows the imge segmenttion result by morphology lgorithm, fter which crried out edge extrction by Cnny opertor, nd the result is shown in Fig.5. 44 Copyright 2016 SERSC
Fig. 1. The input infrred imge Fig. 3. Edge detection by Cnny opertor Fig. 4. Imge segmenttion by morphology lgorithm morphology - Cnny opertor Fig. 5. Edge detection by By compring Figure 3 nd Figure 5, it is known tht the trditionl Cnny opertor is sensitive to noise, nd the morphology-cnny opertor hs strong noise suppression bility, nd the Cnny opertor hs cler nd coherent edge, which cn improve the performnce of smooth noise nd suppress flse edges. And the morphology-cnny opertor fusion lgorithm enriches the locl edge detils. 4 Conclusion In this pper, we identify the edge of infrred imge bsed on morphology- Cnny opertor. The simultion experiments show tht the proposed lgorithm cn detect the edge detils nd more complete contour informtion, improve the ccurcy nd ccurcy of edge detection, which is vlid edge detection method. Copyright 2016 SERSC 45
Acknowledgments. This project is supported by Coopertion Project between Heilongjing Province nd Chinese Acdemy of Science (Grnt No. YS15A10), Youth Innovtion Tlent Trining Progrm of Heilongjing Province Regulr Institutions of Higher Eduction Study on CFRP lminte defects detection using infrred therml wve nondestructive testing technology under modulted lser exittion, nd Hrbin Innovtion Fund Project for Tlent Youth (Grnt No. 2013RFQXJ089). References 1. Lopez, Molin C., De Bets B., Bustince H.: Multiscle edge detection bsed on gussin smoothing nd edge trcking,j. Knowledge Bsed Systems. 44,101-111 (2013) 2. Tbor, JJ., Slis, HM., Simpson, ZB.: A synthetic genetic edge detection progrm, J. Cell. 137,1272-1281 (2009) 3. Vitulno, S., Di Ruberto, C., Nppi, M.: Edge Detection, J. Interntionl Journl of Pttern Recognition & Artificil Intelligence.2,130 (2011) 4. Beern Kutty S., Sidin, S., Megt Yunus P.N.A.: Evlution of cnny nd sobel opertor for logo edge detection. Technology Mngement nd Emerging Technologies (ISTMET), (2014), October 153-156; Wshington, Americ. 5. Lng, B.H., Shen, L.Y., Hn, T.L.: An Adptive Edge Detection Method Bsed on Cnny Opertor, J. Advnced Mterils Reserch. 255-260:2037-2041 (2011) 6. Mlik, J., Sinrynn, G., Dhiy, R.: Personl uthentiction using plmprint with sobel code, cnny edge nd phse congruency feture extrction method [J]. Intct Journl on Imge & Video Processing. 2 (2012) 46 Copyright 2016 SERSC