APPLICATION OF DIGITAL IMAGE PROCESSING TECHNOLOGY IN DYNAMIC TRIAXIAL TEST OF SOIL MECHANICS

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 APPLICATION OF DIGITAL IMAGE PROCESSING TECHNOLOGY IN DYNAMIC TRIAXIAL TEST OF SOIL MECHANICS 1 Shao Longtan *, 2 Yang Song, 3 Sang Yong, 4 Huang Chuan, 5 Guo Xaoxa 1 Prof., Department of Engneerng Mechancs, Dalan Unversty of Technology, 116024, Dalan, Chna 2 Assoc. Prof., Department of Engneerng Mechancs, Dalan Unversty of Technology, 116024, Dalan, Chna 3 Assoc. Prof., Department of Mechancal Engneerng, Dalan Unversty of Technology, 116024, Dalan, Chna 4 Assoc. Prof., Department of Computer Scence, Wuhan Polytechnc Unversty, 430022, Wuhan, Chna 5 Assoc. Prof., Department of Mechancal Engneerng, Dalan Unversty of Technology, 116024, Dalan, Chna E-mal: 1 ocean_epoch@163.com, 3 shaolt@dlut.edu.cn, 3 sang110@163.com, 4 mbonehc@sna.com 5 hanyuer@dlut.edu.cn ABSTRACT Sol stablty problems caused by vbratory loadng, such as that of earthquakes and wnd, have been ncreasngly observed. Desgns and developments, whch nclude a sol dynamc traxal testng apparatus and a traxal sample surface dynamc deformaton measurement method based on the dgtal mage processng technology, have been proposed to study the characterstcs of sol under dynamc loads. The sub-pxel coordnates of corner ponts of squares on rubber membrane wrappng sol sample can be extracted usng a hgh-speed camera that synchronously shoots mages of sol surface dynamc deformaton combned wth the Canny operator and the Harrs method. In accordance wth the concept of the four-node soparametrc element n a fnte element analyss, the dynamc deformaton of sol sample can be calculated. The test results show that the dynamc deformaton measurement method nvolvng the use of dgtal mage processng technology s effectve n measurng the characterstcs of the axal and radal deformaton of the stran feld, thereby provdes a strong expermental foundaton to study the mechancal characterstcs of sol samples. Keywords: Dgtal mage Processng, Geotechncal Test, Dynamc Traxal, Transformaton Measurement 1. INTRODUCTION The stablty and deformaton problems of sols under dynamc loads are becomng ncreasngly sgnfcant, especally those problems concernng the damage, nstablty, and deformaton of foundaton caused by vbratory loads such as earthquakes, wnd, and waves are ssues that need to be solved as soon as possble. Current research on these complex sol dynamc mechancs ssues often rely on smulaton tests conducted n laboratores. Dynamc traxal tests are conducted to study parameters such as sol lquefacton resstance, dynamc elastc modulus, dynamc dampng rato, and so on under dfferent vbraton stresses. These tests analyze the deformaton characterstcs of sol durng and after lquefacton, thereby provdng a parameter bass for numercal calculaton [1]. The dgtal mage processng technology, whch s economcal, practcal, and smple to operate, s used as the research method n tests performed n geotechncal engneerng. The technology has a wde applcaton n the feld of geotechncal expermental mechancs and has been acknowledged by scholars locally and abroad. Whte [2], Alshbl [3] used dgtal mage processng technology to study ssues such as the deformaton process n sand ple dvng and the thckness of the shear zone. Lu [4], Sh [5],and Zhou [6],and others use dgtal mage processng technology to study the sol mcrocosmc structure. Zhao [7], L [8], Le [9]et al., appled dgtal speckle mage to ther studes on rock damage and deformaton measurement. Wang [10], Lu [11], Shao [12]-[13] et al. used dgtal mage measurng technology to study the surface deformaton of traxal samples. The use of dgtal mage processng technology to measure the deformaton of traxal samples has some specfc advantages [14] that tradtonal measurement methods cannot provde. These advantages nclude a hgher deformaton measurement accuracy, no-contact deformaton measurement wthout dsturbng the sample and ts 1358

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 deformaton, and applcaton n large deformaton measurements of traxal samples. Any measurng secton can be chosen to measure the axal and radal deformaton of the traxal sample drectly, and the entre process of the traxal test durng the deformaton measurement can be photographed to analyze and study the sample deformaton after the test. However, untl now, the dgtal mage processng technology s mostly appled n statc loadng ssues, whereas dynamc loadng n traxal testng s complex and has numerous mpact factors, whch explans why dgtal mage processng technology s rarely appled n dynamc tests locally or abroad. A dgtal mage measurement system that nvolves the use of the dgtal mage processng technology was establshed to measure the surface deformaton of sol samples and to satsfy the need for a groundbreakng study of sol tests based on a set of self-developed hydraulc dynamc deformaton measurement hardware. We used ths measurng system to obtan the corner pont coordnates, and the surface deformaton feld of the sol sample was then calculated. At the end of ths paper, a traxal test was conducted under dynamc loads on specfc talngs, and the results were used as an example to verfy the feasblty and relablty of ths system n practcal applcaton. 2. GEOTECHNICAL DYNAMIC TRIAXIAL APPARATUS BASED ON DIGITAL IMAGES Fgure 1: Framework Of Dynamc Traxal Testng Apparatus The geotechncal dynamc traxal test apparatus developed n ths paper s manly composed of a hydraulc cycle loadng system, a pressure test rg, a specally desgned pressure chamber, a dgtal control system, an mage measurement system, a hgh-frequency dynamc montorng and control system, and so on. The entre framework and appearance are shown n Fgure.1 and Fgure.2, respectvely. Fgure 2: Dynamc Traxal Testng Apparatus 2.1 Workng Prncple for Dynamc Traxal Apparatus A hydraulc cyclc loadng system s the core module of the traxal test apparatus, whch can obtan the statc and dynamc loadng of force and dsplacement. We decded to use the hydraulc servo for loadng to ensure that the loadng servo system has good dynamc response and steady-state accuracy. The hydraulc cyclc loadng system conssted of hydraulc ol source (hydraulc staton, 21 mpa, 30 L/mn), servo amplfer, servo cylnder, servo valves, and accumulators, whch can obtan hgh-frequency response for dynamc loadng. Ths system has good dynamc performance, hgh relablty, large power to weght rato, flexble confguraton, convenent power transmsson and control, and large loadng rgdty. A pressure rg connects the servo cylnder to a specal pressure chamber, whch s desgned as a gate-lke structure wth a beam that moves up or down dependng on the sze of the pressure chamber. The pressure rg can also adjust ts perpendcularty durng loadng to ensure an accurate, relable, and homogeneous loadng force on the axal drecton of the sample. The pressure chamber s desgned as a specal structure wth a flat glass to facltate the mage measurement of sol deformaton, to ensure the qualty of measurement, and to acheve the basc functons of the fxed sample. Ths specally desgned pressure chamber has a pressure of 1,000 kpa. The dgtal control system can dsplay real-tme data durng tests, such as pressure, temperature and dsplacement. The hgh-frequency dynamc montorng and control system manly conssts of hgh-precson tenson and compresson sensors, hgh-precson dsplacement sensors, a hgh-speed AD/DA card (16 bt), and a computer. Among these 1359

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 components, the computer s manly responsble for the operaton of varous control algorthms, hghfrequency loadng waveform generaton, data dsplay, data reportng, and storage. The bult-n force sensors have hgh-accuracy and precson. Its force measurement accuracy can reach up to 0.3 N, ts control accuracy up to 1 N, and ts dsplacement measurement accuracy up to 0.002 mm. The test curves have a number of samplng ponts. Each oscllator has 500 samplng ponts, whch can truly reflect the entre process of sol stress, deformaton, and pore pressure as they vary over tme, thereby demonstratng the entre transent process better. The dynamc traxal test apparatus nvolves the followng workflow: a hgh-frequency dynamc montorng and control system generates a hghfrequency dynamc waveform accordng to the test requrements and calculates the output value of the controller. The hydraulc cyclc loadng system receves the nformaton and starts to complete the hgh-frequency dynamc loadng whle trggerng the mage measurement system to work and record the surface deformaton data of the sample under that dynamc loadng process. In addton, the dgtal control system dsplays real-tme data durng the test. 2.2 Dgtal Image Measurement System The ordnary vdeo camera s slow and cannot capture subtle changes n the sol surface under dynamc loadng. Therefore, our system used an ndustral hgh-speed camera (Germany AVT Pke F-100C hgh-speed camera, whch has the fastestmage acquston speed of 60 frames at a resoluton 1000*1000 pxels; 1394 B nterface, hgh data transfer rate, and s programmable to multple trgger mode) to shoot the real-tme deformaton mages of the sample surface under dynamc loadng. Crcular LED lghts were nstalled n the pressure chamber, whch provded an even and stable lght source to ensure the qualty of captured mages. Gven the lmtaton of computer processng capacty and data capacty, the extracton of corner pont coordnates was not accomplshed n real-tme. Therefore, we saved the mages durng the test and processed these mages afterwards. A latex membrane was wrapped around a cylndrcal talngs sol sample, and the membrane was regularly prnted wth squares havng a sde length of 12 mm and a spacng of 10 mm between each other. The membrane was made black and the squares were made whte to mprove recognton accuracy of corner ponts. Durng the test, we synchronzed the vbraton of traxal dynamc system and the camera shootng. The dynamc traxal test apparatus had a samplng frequency of 250 khz, whch records 500 dsplacement data n 1 second. The camera was set to capture 50 mages per second, whch means that t shot a pcture at every 10 samplng nterval, to acheve the contrast of the measurng data. The traxal hgh-frequency dynamc montorng and control system, as the major synchronzaton controller, perodcally sent a 5-V TTL pulse sgnal to a hgh-speed camera at the same tme that the vbraton began. The camera mmedately shot pctures upon recevng the external pulse. When the test ended, the montorng and control software stopped sendng pulse sgnals, and the camera stopped recordng. The aforementoned synchronous operaton effectvely guaranteed a match between the dsplacement data collected by the sensors and the dsplacement data n subsequent calculatons. The mplementaton process of the dgtal mage processng system s descrbed as follows: the hghspeed camera was nstalled on a sold trpod, arranged at approxmately 0.6 meters away from the pressure chamber. Its lens had the same heght as the mddle of the pressure chamber. The camera was focused so that the sample clearly appeared n the vew feld of shootng. When the camera started shootng, t located corner ponts of whte squares on the latex membrane as trackng targets. Once the test began, the camera started to work under synchronous nstructon from a the hgh-frequency montorng and control software perodcally shot surface mages of the sample, and sorted the mages by shootng sequence, whch were then stored. When the test was fnshed, a post-processng program was employed to read mages n sequence, track the coordnates of target corner ponts by usng dgtal mage measurement algorthm n the followng context, and calculate the surface deformaton of the sample based on the changes of coordnates. 3. DIGITAL IMAGE MEASUREMENT METHOD FOR SURFACE DEFORMATION OF GEOTECHNICAL SAMPLE 3.1 Dgtal Image Measurement Method The surface mages of the sol sample taken by a hgh-speed CCD camera requred several treatments to obtan the deformaton nformaton of the sample. The man process s llustrated n Fgure. 3. 1360

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 Fgure 3: The Flow Chart Of The Proposed Algorthm The mage enhancement algorthm was used to mprove the contrast of the black rubber membrane and the whte squares, whch resulted n better edge extracton. The captured mages nevtably contaned nose. In the process, a 5 * 5 Gaussan pyramd [15] was used n mage flterng to elmnate the mpact of nose effectvely, whch s shown n Fgure. 4 The Canny operator [16] was employed to extract the contour of whte squares. The hgh threshold and low threshold of the Canny can detect strong and weak contour, respectvely. The method s not susceptble to nose nterference and s capable of detectng true weak edges. The Canny operator uses Gaussan smoothng functon: 2 2 1 X + Y G( XY, ) = exp (1) 2 2 2pσ 2σ where G x and G are the dervatves of Gxy (, ) y to X and Y 2 2, respectvely. A= G + G s the edge strength. After the non-maxma suppresson of the edge strength, the edge ponts were determned, as shown n Fgure. 6. The complete outlne of the grd can be obtaned n the case of deal llumnaton and mage qualty. However, the grd contour extracton s often ncomplete because of the lghtng qualty n practce. In ths paper, we used a mathematcal morphology 3 * 3 square structure element to conduct dlaton and eroson operatons on the mages, and we connected the broken edges to close the rectangles. Although the rectangles could be closed after the morphologcal operatons, ther borders became wder. Therefore, the nsect followng [17] algorthm was used to extract a sngle pxel from the outer boundary of the contours. The Douglas-Peucker [18] algorthm was used to ft the obtaned polygons. The number of the edges of the polygon were consdered, and f t s not equal to four (that s quadrlateral), the polygon was then dscard to complete the flterng of the non-target nformaton, as shown n Fgure.7. For the extracton of the rectangles vertex poston, the hghly accurate Harrs [19] algorthm was employed to obtan the pxel level coordnates of the corner ponts and to exclude corner ponts far from the four vertces of the small square, as shown n Fgure. 8. The sub-pxel corner detecton algorthm proposed by Chstoph [20] was used to x y obtan more precse coordnates of the corner ponts and ther number, and to sort the coordnates, as shown n Fgure. 9. The aforementoned algorthms were employed to process the captured mages, calculate the transformaton of specfc postons, and obtan the deformaton nformaton based on the calculatng prncple of the deformaton feld. Fgure 4: The Image Of Sol Sample Fgure 5: The Image Of Gray-Scale Enhancement And Flterng Fgure 6: The Image Of Canny Edge Detecton Fgure 7: The Image Of Non-Object Flterng 1361

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 Fgure 8: The Image Of Harrs Corners Extracton Fgure 9: The Image Of Sub-Pxel Corners Wth The Order Number 3.2. Method for Calculatng Stran Feld of the Sample The dstance of each object must be normalzed and converted nto a two-dmensonal coordnate system (unt: pxels) of same object dstance, that s, the data must be normalzed based on the same pxel equvalent to take full advantage of the measurement data on the corner ponts on the sample surface n calculatng the axal and radal stran. Combned wth the concept of the four-node soparametrc element n the fnte element analyss, the nterpolaton calculaton was conducted on the normalzed coordnates of corner ponts so that the coordnates of the plane at any pont could be obtaned. 4 x= NX (2) = 1 4 = 1 y= NY (3) Where X and Y are corner pont coordnates, 0 0 N s shape functon. Let X and Y be ntal corner pont coordnates, then dsplacement of each corner pont can be got from these ntal coordnates: 0 u = X X (4) 0 v = Y Y (5) Then, dsplacement of any pont can be represented as below: 4 u= Nu (6) = 1 4 = 1 v= Nv (7) Then, stran feld can be expressed as below: ε x u = { B} (8) ε y v where ε x denotes radal stran of sample, ε y denotes axal stran of sample. 4. APPLICATION 4.1. Dynamc Traxal Test on Modulus and Dampng The data obtaned by the mage measurement system are compared wth those by the sensor, to verfy the relablty of the traxal mage measurement system under dynamc loads. The deformaton values obtaned by the mage measurement system are smaller than those by the sensor, as a result of the end restrant. Ths s because the gap between the sample and the sample cap, base, porous stone and flter paper wll be compressed under the loads, whch makes the axal stran obtaned by the external sensor larger than the real deformaton of the sample. The dfference n dynamc dsplacement and n dynamc stran between the two methods s analyzed wth dfferent vbraton frequences. As shown n Fgure.10, under the same vbraton frequency, the dfference n dynamc dsplacement between the two methods ncreases (but not lnearly) wth the ncrease of dynamc stress. In addton, as vbraton frequency ncreases, the dfference between the two methods s greater. It can be magned that n the dynamc damage tests wth a small-value but hgh-frequency load, the deformaton value obtaned by the sensor s much hgher than that by the mage measurement system. In ths case, the sensor data have already reached the damage level, but the data from mage measurement system have not yet reached the damage level. In other words, the tradtonal measurement method would probably underestmate the dynamc strength of constructons. The relatonshp between the dfference n dynamc dsplacement-stran obtaned by dfferent methods (mage measurement system and sensor) and the vbraton frequency s also nvestgated. As shown n Fgure.11, under a certan dynamc stress, the dfference n dynamc dsplacement-stran by the two methods dsplays a lnear varaton as the 1362

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 frequency ncreases. The dfference between the two methods ncreases as dynamc stress ncreases. Dynamc Dsplace Error(mm) 0.6 0.5 0.4 0.3 0.2 0.1 Vbraton Frequency=3 Vbraton Frequency=8 Vbraton Frequency=14 0.0 5 10 15 20 25 30 35 Dynamc Stress(kPa) (A) Curve Of Dynamc Dsplacement Error And Dynamc Stress Dynamc Stran Error 0.005 0.004 0.003 0.002 0.001 Vbraton Frequency=3 Vbraton Frequency=8 Vbraton Frequency=14 0.000 5 10 15 20 25 30 35 Dynamc Stress(kPa) Dynamc Stran Error 0.005 0.004 0.003 0.002 0.001 Dynamc Stress=13Kpa Dynamc Stress=23Kpa Dynamc Stress=33Kpa 0.000 2 4 6 8 10 12 14 16 Vbraton Frequency (B) Relaton Between Dynamc Stran Error And Vbraton Frequency Fgure 11: Relaton Of The Error Of Dynamc Dsplacement And Dynamc Stran To Vbraton Frequency 4.2. Dynamc Deformaton and Damage Test One of the salent advantages of dgtal mage measurement s that t can drectly obtan the axal and radal stran felds of the sample at any moment under a dynamc load, as shown n Fgure. 12 and Fgure. 13. From the surface deformaton felds at dfferent moments, the occurrence and expandng of sol deformaton and the ultmate damage of the sol can be observed under a dynamc load. The deformaton morphology can also be captured. (B) Curve Of Dynamc Stran Error And Dynamc Stress Fgure 10: Relaton Of The Error Of Dynamc Dsplacement And Dynamc Stran To Dynamc Stress Dynamc Dsplacement Error(mm) 0.6 0.5 0.4 0.3 0.2 0.1 Dynamc Stress=13Kpa Dynamc Stress=23Kpa Dynamc Stress=33Kpa 0.0 2 4 6 8 10 12 14 16 Vbraton Frequency (A) Radal Stran Feld At The Begnnng (B) Axal Stran Feld At The Begnnng Fgure 12: Stran Feld At The Begnnng (A) Relaton Between Dynamc Dsplacement Error And Vbraton Frequency (A) Radal Stran Feld At The End (B) Axal Stran Feld At The End Fgure 13: Stran Feld At The End 1363

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 5. CONCLUSIONS We have developed a geotechncal dynamc traxal test loadng system that uses the hydraulc servo technology and a bult-n hgh-precson force sensor for effectve stress collecton, whch can accurately reflect the deformaton of sol samples under dynamc loads, and can provde a prerequste for geotextle dynamc traxal tests. The dgtal mage processng method was performed on the geotextle dynamc load tests, a dynamc deformaton measurement method was establshed based on mage processng technology, and radal and axal deformaton measurements were mplemented. In addton, ths method can also detect the dynamc axal stran and dynamc radal stran feld of the samples at any moment, reflect the local deformaton characterstcs of sample more comprehensvely, and enrch the deformaton nformaton on samples for measurement. ACKNOWLEDGEMENT The work descrbed was completed n the Geotechncal laboratory at the Faculty of Engneerng Mechancs, Dalan Unversty of Technology. It was partally supported by Natonal Natural Scence Foundaton of Chna (Grant No.50905022), Natonal H-tech Research and Development Program of Chna (973 Program, Grant No.2010CB7315022), State Key Laboratory of Structural Analyss for Industral Equpment specal funds (Grant No.S09104), and the Fundamental Research Funds for the Central Unverstes (Grant No.DUT12LK21).The system n focus was produced at the Dalan unversty of Technology, Dalan, by Mr. Yong Sang and hs coworkers. These contrbutons are gratefully acknowledged. REFERENCES: [1] H. L. Lu, Y. D. Zhou. and Y. F. Gao., Study on the Behavor of Barge Ground Dsplacement of Sand due to Sesmc Lquefacton, Chnese Journal of Geotechncal Engneerng, Vol. 24, No.3, 2002, pp. 142-146. [2] Whte. D. J., Take. W. A., Bolton. M. D., Measurng Sol Deformaton n Geotechncal usng Dgtal Images and PIV Analyss, 10th Internatonal Conference on Computer Methods and Advances n Geotechancs, Tucson, 2001, pp. 997-1002. [3] Alshbl. K. A.,Sture.S., Sand Shear Band Thckness Measurement by Dgtal Imagng Technque, Journal of Computng n Cvl Engneerng, Vol. 13, No. 2, 1999, pp. 103-109. [4] J H Lu, B. N. Hong, and H. B. Zhang, A New Expermental Method of Sol Mcrostructure Changng Process, Rock and Sol Mechancs, Vol. 24, No. 5, 2003, pp.744-747. [5] B. Sh, H. T. Jang, Research on the Analyss Technques for Clayey Sol Mcrostructure, Chnese Journal of Rock Mechancs and Engneerng, Vol. 20, No 6, 2001, pp. 864-870. [6] J. Zhou, D. D. Sh, F Wu et al., Vsualzed Cyclc Traxal Tests on Sand Lquefacton usng Dgtal Imagmg Technque, Chnese Journal of Geotechncal Engneerng, Vol. 33, No. 1, 2011, pp. 81-86. [7] Y. H. Zhao, H. H. Lang, C Y Xong et al., Deformaton Measurement of Rock Damage by Dgtal Image Correlaton Method, Chnese Journal of Rock Mechancs and Engneerng, Vol. 21, No. 1, 2002, pp. 73-76. [8] Y. H. L, H. W. Jng., Software Development of a Dgtal Speckle Correlaton Method and ts Applcaton, Journal of Chna Unversty of Mnng & Technology, Vol. 37, No. 5, 2008, pp. 635-640. [9] D. Le, P. Z. Qao, A L et al., Hgh-Speed Dgtal Image Correlaton Studes for Dynamc Damage of Rock, Journal of Chna Coal Socety, Vol. 36, No. 2, 2011, pp. 274-277. [10] W. B. Lu, Z. Y. Chen, L. J Su., Study on the Traxal Text of Renforced Sol based on Dgtal Image Processng Technque, Chna Cvl Engneerng Journal, Vol. 44, 2011, pp. 41-45. [11] J. T. Wang, H. L. Cao, M. Y. Dng et al. Dgtal Image Measurement of Specmen Deformaton n Traxal Test based on Dgtal Camera, Journal of HUST, Vol. 6, 2004, pp.1-3,15. [12] L T Shao, Z P Wang, Y L Lu. Dgtal Image Processng Technque for Measurement of the local Deformaton of sol Specmen n Traxal Test, Chnese Journal of Geotechncal Engneerng, Vol. 24, No. 2, pp. 159-163. [13] L. T. Shao, Y. Zheng, Wang Zhupn, et al. Applcaton of dgtal mage processng technque to traxal test n sol mechancs, Roch and Sol Mechancs, Vol. 27, No 1.2006, pp. 129-34. 1364

Journal of Theoretcal and Appled Informaton Technology 2005-2013 JATIT & LLS. All rghts reserved. ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 [14] J. Wang and Y. Guo. Deformaton Study on Traxal Specmen usng the Dgtal Image Measurng System, Northwestern Sesmologcal Journal, Vol. 33, 2011, pp. 175-180. [15] N. Ln, We C Yu, James S. Duncan., Combnatve mult-scale level set framework for echocardographc mage segmentaton, Medcal Image Analyss, Vol. 7, No. 4, 2003, pp. 529 537. [16] J. Canny, A Computatonal Approach to Edge Detecton, IEEE TransPAMI-8, Vol. 6, 1986, pp. 679-698. [17] L F Zhou, Z X Hu, Y M Xao et al. A New Algorthm on Objects Boundary Trackng of Bnary Image, Mcro-Computer Informaton, Vol. 23, No.(2-3), 2007, pp.259-261. [18]Douglas D H,Peucker T K. Algorthm for the Reducton of the Number of Ponts Requred to Represent a Dgtzed Lne or ts Carcature, The Canadan Cartographer, Vol.10, No.2, 2001, pp. 112-122. [19] Harrs C, Satephoens M J. A Combned Corner and Edge Detector, In Alvey Vson Conference, Manchester, 1988, pp. 147-152. [20] Chstoph Stock, Ulrch Muhlmann. Subpxel Corner Detecton for Trackng Applcatons usng CMOS Camera Technology, 26th Workshop of the Austran Assocaton for Pattern Recognton, Graz, Austra, 2002, pp. 191-199. 1365