Proceedngs of the World Congress on Engneerng 2008 Vol I Influence of Ambent Lght on the Qualty of Laser Dgtzed Surfaces D. Blanco, P. Fernández, E. Cuesta and C. M. Suárez Abstract Laser trangulaton systems allow for non-contact, accurate and fast dgtzng of surfaces. The qualty of the dgtzed surface depends on many factors. Some, le laser emsson wavelength, are characterstcs of the system. Other factors, le part surface qualty are external to the system. Laser trangulaton systems project a laser beam onto the surface of the worpece, so that an mage of ths projecton s captured n a photo-sensor. Ths mage s processed usng trangulaton technques to calculate the spatal poston of every pont on the projecton. The sensor does not only capture the lght projected by the laser, but also captures the ambent lght emtted n the same wavelength of the laser beam. Snce calculaton of ponts postons depends on the mage characterstcs, ambent lght becomes a potental error source. The present wor evaluates the nfluence of ambent lght on the results of the scannng process. A methodology for testng dfferent lght sources under the same dgtzng condtons has been developed. Three dfferent crtera have been used to analyze the qualty of the pont cloud: the number of captured ponts, the average dsperson of the test pont cloud wth respect to a reference pont cloud, and the dstrbuton of such geometrc dsperson across the whole surface. Index Terms Laser trangulaton, dgtzng, nfluence of ambent lght. I. INTRODUCTION Laser trangulaton (LT) systems are used for nspecton and reverse engneerng of surfaces. Ther characterstcs mae them sutable when accurate and fast non-contact scannng s needed. In most LT systems, the projecton of a laser beam onto a surface s captured as an mage n a charged coupled devce (CCD). Applyng mage processng technques and the trangulaton prncple, 3D coordnates of the surface ponts are acqured (Fg.1). If the dstance between a partcular pont P and the CCD matches exactly the value of the reference dstance (stand-off), ts mage n the CCD wll be placed exactly n a reference pont P. Otherwse, f the pont onto Manuscrpt receved March 18, 2008. Ths wor s part of the results obtaned n a research project supported by the Spansh Educaton and Scence Mnstry (MEC-04-DPI2004-03517) and FEDER. D. Blanco s Lecturer of the Manufacturng Engneerng Department n the Unversty of Ovedo, Campus of Gjón, Span (phone: 985-182-444; fax: 985-182-016; e-mal author: dbf@unov.es). E. Cuesta and C.M. Suárez are Senor Lecturers of the Manufacturng Engneerng Department n the Unversty of Ovedo, Campus of Gjón, Span (phone: 985-182-016; fax: 985-182-016; e-mal authors: ecuesta@unov.es; csuarez@unov.es ). P. Fernández has collaborated wth the Manufacturng Engneerng Department n the Unversty of Ovedo for developng the research project mentoned above (e-mal author: pedrofa@unov.es) the surface were further away a dstance H n the drecton of the laser beam, ts mage on the CCD wll be placed a dstance h from the reference pont. As the geometry of the laser system s completely nown, t s possble to determne the spatal poston of every sngle pont from ts mage poston on the sensor [1]. In order to dgtze a part, a relatve movement between the laser system and the part surface s needed, so that the laser beam projecton sweeps the target surface. Ths results on a set of dgtzed ponts (pont cloud) that represents the surface of the part. Usually, LT systems are nstalled on a coordnate measured machne (CMM) that provdes precse and controlled dsplacements along ts axes. Usng motorzed heads on the CMM also allows for rotatng the LT system to obtan a sutable orentaton for the scannng process. Accurate calculaton of the spatal poston for each pont of the laser strpe depends on the accurate calculaton of the centrod of ts lght dstrbuton n the sensor [1], [2]. If the ntensty of the lght dstrbuton captured n the sensor s too wea, the system can not properly calculate the poston of the ponts. Otherwse, f laser ntensty s too hgh, the sensor wll turn nto saturaton, so that the system could not calculate the poston of ponts. For ntermedate stuatons, the lght dstrbuton s analysed to determne ts centrod poston, whch corresponds to dstance h measured from reference pont. Consequently, the lght dstrbuton affects the accuracy of dstance H calculaton (Fg.1). Reference Dstance (Stand off) +H H Laser Dode P Lens CCD Sensor h +h P Fg. 1. Scheme of the laser trangulaton prncple
Proceedngs of the World Congress on Engneerng 2008 Vol I φ=0 Dstance δ Lattude φ Azmut θ θ=0 Fg. 2. Sphercal coordnate system used to orentate the lght source The result of the scannng process depends on the LT system characterstcs, the geometry and qualty of the surface [3] and the envronmental condtons. These elements determne the shape and contrast of the laser strpe onto the surface and the mage captured by the sensor. Snce surface qualty s an mportant nfluence factor, most LT systems allow for adjustng laser ntensty accordng to surface colour and roughness requrements. The objectve s to acheve an mprovement n the sharpness of the laser beam projecton. The ambent lght present at the scannng process s one of the possble envronmental nfluences [4]. Usually, LT systems ncorporate optcal flters to reduce or elmnate the nfluence of the ambent lght. These flters accept only those wavelengths n the laser emsson band. Commercal lght sources emt lght n a wde spectrum of frequences. Snce ambent lght emtted n the laser emsson band wll not be fltered, t wll become part of the nformaton captured by the sensor and wll be used n the calculaton of pont poston. II. OBJECTIVES In ths wor, ambent lght nfluence on the qualty of dgtzed pont clouds s evaluated. Dgtzng tests have been carred out to compare results under dfferent ambent lght condtons. Although there s a wde range of commercal lght sources, the present wor deals wth the most commonly used lamps. Tests have been carred out under laboratory condtons, where llumnaton for each experment s reduced to a sngle lght source. For each test confguraton, nature and mportance of the uncertanty ntroduced by ambent lght have been establshed. Results for each lght source have been afterwards compared to elaborate usage suggestons. III. CONFIGURATION OF THE TESTS Tests have been carred out usng a LT strpe commercal system from Metrs (model Metrs LC50) whch has been mounted on a Brown & Sharpe Global CMM (model Image). The LC50 uses a laser beam emttng n the red vsble spectrum wth a wavelength emsson band between 635 nm and 650 nm. The maxmum pea power s 1 mw. A reflectance standard from Labsphere has been used as the surface to be dgtzed [5]. Ths surface (a..a. reference surface) s 99% reflectance certfed. Ths means that ts surface s quas-lambertan, so 99% of the receved energy s reflected. Ths reflexon s deally dffused, as the energy s reflected n a unform way n all spatal drectons. Orentaton of the lght source wth respect to the reference surface and the LT system has been selected so that the lght drecton theoretcally maes an angle of 45º (ϕ = 45º) wth the standard surface. Moreover, theoretcal drecton of the ncdent lght s orthogonal to the sweep drecton. For every pont, the system calculates the z coordnate value tang nto account dstance H (Fg.1) and sensor poston and orentaton. Therefore, nfluence of ambent lght affects the calculated vertcal poston of the ponts. Fg. 2 shows the sphercal coordnate system used for the lght sources orentaton. The orgn of ths coordnate system s the centre of the reference surface. In order to ncorporate the nfluence of lght source ntensty to ths wor, tests have been carred out wth two dfferent postons for the lght sources: 200 mm (δ1) or 400 mm (δ2) from the orgn of the coordnate system. Lght sources and reference surface are mounted n a test-bench desgned ad hoc. Ths test-bench provdes a proper poston and orentaton of the lght source accordng to the sphercal coordnate system (Fg. 3). Ths mountng allows for comparng pont clouds obtaned under dfferent test condtons. The set formed by the test-bench, the lght source and the reference surface has been nstalled on the CMM table. The lght sources used for ths wor are among the most usual types on a metrologcal laboratory or a worshop. Ths way, three types of ncandescent lamps (clear, tnted n blue and halogen), a fluorescent lamp, a low pressure sodum lamp and a mercury vapour lamp consttute the fnal selecton. Although there s a wde varety of lght sources for each class that can be tested (attendng to power or shape), the selected lamps have smlar values for the lumnous flux (lumens). Ths selecton crteron s based on fndng alternatves that offer the operator a smlar vsual comfort, when performng long-tme runnng dgtzng processes. Commercal references for the lght sources used n ths wor are n Table I. Lght Source Metrs LC 50 Reflectance Standards Fg. 3. Scheme of the man elements used n the test
Proceedngs of the World Congress on Engneerng 2008 Vol I TABLE I. DIFFERENT LAMPS USED IN THE EXPERIMENTS K Model Manufacturer Lamp Type lumens 1 CLAS A CL OSRAM Clear ncandescent 1360 2 DECOR A OSRAM Blue ncandescent - 3 64476 BT OSRAM Halogen 1600 4 PLE-T PHILIPS Fluorescent 1500 5 HQL 50 SDL OSRAM Mercury vapour 1600 6 SOX 18 OSRAM Low pressure sodum 1800 IV. EXPERIMENTAL PROCEDURE Test where the only llumnaton comes from the laser tself wll not be altered by any external energy. Assumng ths, a pont cloud that has been dgtzed n the absence of lght wll suffer no dstortons. Hence, dgtzng n the dar appears to be the most approprated way for scannng surfaces, although worng n the absence of lght s an unapproachable stuaton for human operators. Nevertheless, a pont cloud obtaned n the absence of lght can be used as a reference when evaluatng the qualty of pont clouds dgtzed under normal ambent lghtng. Expermental procedure used n ths wor allows for comparng the results obtaned when dgtzng under partcular lght sources wth the results obtaned n the absence of ambent lght. Although a sngle reference pont cloud for all test may seem a sutable opton, n practce, ths approach s not recommended. Senstvty of the nternal geometry of the sensor to thermal varatons (related to the tme the laser remans swtched on) must be taen nto account. The fact that the manufacturer of the sensor recommends a mnmum 40 mnutes warm-up perod snce swtchng on the laser untl a proper stablty s reached, confrms the mportance of ths effect. Therefore, nstead of usng a sngle reference cloud for all tests, specfc reference clouds have been used n each test comparson. These reference clouds must be dgtzed mmedately after the capture of each sngle test cloud. Ths procedure wll mnmze the possble alteraton of the sensor nternal geometry due to thermal drft. Thus, the frst cloud N δ s obtaned by dgtzng the reference surface under a partcular type of lght source (), placed at a gven dstance from the orgn of the coordnate system (δ). Immedately after the frst one, a second pont cloud, nown as reference pont cloud P δ, s obtaned by dgtzng the same surface n the absence of ambent lght. Comparson between these two clouds requres each pont n the test cloud to have ts equvalent n the reference cloud. To ensure ths relatonshp, a computer applcaton has been mplemented, capable of selectng and classfyng a group of 400 ponts (20 x 20) n the central area of the reference surface. Matrx constructed n ths way allows for the comparson between each pont and ts equvalent. LT system test parameters (as laser lght ntensty) have been adjusted to avod loos of ponts due to saturaton n the reference cloud. Dstance between dgtzed ponts has been set to be 1.5 mm n both X and Y drectons of the CMM coordnate system. V. ANALYSIS CRITERIA Three crtera have been used for comparng the qualty of pont clouds obtaned under dfferent sources of lght. The frst crteron evaluates the nfluence of lghtng on the number of ponts captured n the test cloud. As dscussed prevously, an excessve nput of lght energy turns the sensor to saturaton. Therefore, t becomes mpossble to calculate a proper value for the z coordnate of the saturated ponts. The saturated ponts are not ncluded n the cloud N δ as the system rejects them. The parameter used to characterze ths crteron s the number of vald ponts (n δ ) on the cloud. The second crteron evaluates the nfluence of lghtng n the proper calculaton of z coordnate value for each pont. Improper values wll cause the ponts of the test cloud to appear n a hgher or lower place than they really are. The absolute dfference between the z values for each equvalent par of vald ponts n both the test cloud N δ and ts reference cloud P δ s calculated (1). d = z z (1) δ Nδ Pδ The standard devaton σ δ of the calculated dfferences d δ has been used as the characterstc parameter for ths second crteron (3). μ σ δ 1 n δ = d (2) n = 1 δ 1 n δ δ = 1 ( d ) 2 μ = (3) n The last crteron used n ths wor s qualtatve. It conssts on a graphcal representaton of devatons d δ for each dgtzed pont cloud. Ths representaton shows how the ambent lght modfes the poston of each sngle pont. It allows for determnng whether lght nfluence s equal across the whole surface or not. VI. RESULTS DISCUSSION Attendng to frst crteron (n δ ), the results of the tests n Table II, llustrate how certan types of lght sources cause a hgh percentage of ponts to become saturated. Thus, n three of the tests (N 11, N 12 y N 31 ), no pont has been captured due to saturaton caused by the great amount of energy n the laser wavelength band. The sensor can not obtan properly the z coordnate value of these ponts, therefore the pont cloud s empty. A partal loose of ponts has only occurred n one case (test N 21 ). However, the same lght source placed on a further poston (test N 22 ) provdes a complete pont cloud. The rest of the tests provde complete pont clouds, so that proper nformaton from surface geometry can be easly obtaned. An order of preference between dfferent lght sources can be establshed by usng the second crteron (explaned n δ prevous secton) referrng to the standard devaton ( σ ). From these tests t can be concluded that the best results for both testng postons are obtaned for the low pressure
Proceedngs of the World Congress on Engneerng 2008 Vol I sodum lamp. Ths lght source causes a lower dstorton for the test cloud over the reference cloud. Comparson between pars of clouds, all obtaned n the absence of lght, provdes a medum value of 1.2 µm as the systematc error [6] attrbutable to the system tself. Then, the result obtaned for δ the sodum lamp and the furthest poston ( σ =1.13 µm) ndcates a moderate nfluence of ths lght source on ponts postons. TABLE II. RESULTS OF THE TESTS FOR THE NUMBER OF VALID POINTS AND THEIR STANDARD DEVIATION WITH RESPECT TO THE REFERENCE CLOUD K δ N Kδ n Kδ σ Kδ d mn d max [µm] [µm] [µm] 1 1 N 11 0 - - - 2 N 12 0 - - - 2 1 N 21 44 8.36-29.24 17.33 2 N 22 400 4.64-21.49 21.12 3 1 N 31 0 - - - 2 N 32 400 4.83-19.4 25.15 4 1 N 41 400 4.33-7.33 21.97 2 N 42 400 2.14-8.97 9.89 5 1 N 51 400 2.15-10.75 9.21 2 N 52 400 1.88-7.75 9.89 6 1 N 61 400 1.99-18.19 13.79 2 N 62 400 1.13-5.92 5.07 The behavour of the fluorescent lamp s clearly worse than the mercury vapour one for the closest poston, whle for the furthest one the dfference s not so evdent. For the tnted blue ncandescent lamp and the furthest poston, the standard devaton s approxmately 4.1 tmes greater than the value calculated for the sodum lamp. For the closest poston, ts standard devaton s extremely hgh, but t must be remared that ths value has been calculated consderng a small number of vald ponts, as most of the theoretcal ponts n the cloud have not been captured due to saturaton. In the case of the halogen lamp, the results for devatonσ δ are the worst of all. At the furthest poston, the devaton s the maxmal observed, approxmately 4.3 tmes greater than the sodum lamp. Ths result was predctable as ths lamp causes all the ponts to become saturated for the dstance of 200 mm. Fnally, the thrd crteron (graphc representaton of the parameter d δ along the whole area) shows how the dstrbuton of these values depends on the type of lght source and s clearly non-unform (Fg. 4 and Fg. 5). Thus, when testng the blue ncandescent lamp at a dstance of 200 mm from the orgn of the coordnate system (Fg.4), the graph shows how the lac of ponts due to saturaton s produced n ponts of the reference surface that are close to the lght source. On the other hand, vald ponts are only regstered n a narrow strp placed n the area where the ponts of the surface are further from the lght source. In certan areas of each test cloud, some of the ponts are located n an upper vertcal poston from ther equvalent ones n the reference cloud, whereas ponts n other areas are located n a lower poston. However, ths dstorton n the test clouds does not seem to be related to the proxmty of ponts to the lght source, even when such a relatonshp can be establshed for the saturaton of ponts. Fg. 4. Dstrbuton of d δ for a dstance δ = 200 mm By contrast, the dstrbuton of peas and valleys n Fg. 4 and Fg. 5 shows a parallel orentaton to the laser strpe projecton onto the surface. Ths result does not ft wth any of the prevous assumptons. The effect may be related to local rregulartes on the reference surface propertes. Ths should be confrmed by later wor. The appearance of the dfferences plotted n Fg. 4 confrms the conclusons obtaned for the second crteron. Snce sodum lamp generates less dstorton n the cloud, t provdes the best performance among the tested lamps. Ths dstorton s ncreased for mercury lamp and for fluorescent lamp. On the other hand, the blue ncandescent lamp causes a huge dstorton effect. When testng the lghts on the furthest poston (δ = 400 mm), fve completely-full clouds have been obtaned (Fg. 5). As t was set for the closest dstance, dstrbuton of d δ shows a non-unform behavour. Furthermore, the parallelsm between the preferental drecton of peas and valleys n the graphs and the orentaton of the laser strpe can be notced as n prevous Fg.4. Moreover, the result n terms of the level of dstorton for the test clouds follows the same prevously establshed pattern accordng to the type of lght source. The sodum lamp s agan the source that has a lower nfluence n the dstorton ntroduced n the test cloud. For ths poston, the dstorton ntroduced by mercury and fluorescent lamps s very smlar, when a better behavour of the mercury one has been establshed for the closest dstance.
Proceedngs of the World Congress on Engneerng 2008 Vol I Each test s carred out after rotatng the reference surface 90º clocwse around ts geometrc centre. In second test, the surface has been rotated 90º clocwse. In thrd test the rotaton reaches 180º. Fg. 6 shows the results of these three tests and how rregulartes n the form of peas and valleys rotate wth the surface, even when the rest of the elements of the test have remaned unchanged. Therefore, t can be concluded that the assumpton of a lac of unformty n terms of the reference surface propertes s true. Fg. 5. Dstrbuton of d δ for a dstance δ = 400 mm The results for the ncandescent lamps (both the blue one and the halogen) are the worst among the tested lamps. The dfferences (both postve and negatve) are sgnfcantly hgher than for the rest of lamps. Ths result matches prevous conclusons. Specfc test has been carred out to clarfy the causes of the non-unform dstrbuton of peas and valleys. The new tests provde three clouds dgtzed usng the blue ncandescent lamp n poston δ2. VII. CONCLUSION Ths wor has establshed the nfluence of ambent lght on the qualty of laser trangulaton dgtzed surfaces. Results of the tests show how the nfluence of the lght sources affects the dgtzng n dfferent ways. Sources that ntroduce a huge amount of energy on the laser wavelength band wll cause some of the ponts to saturate. In severe cases ths affects the whole pont cloud and no nformaton wll be obtaned. It has been also demonstrated that dfferent types of sources cause dfferent results when calculatng vertcal poston for every pont of the cloud. In addton, t has been verfed that ths nfluence s not unform. Thus, dependng on the type of lght source used, some of the ponts of the test cloud are located n an upper vertcal poston from ther equvalent ponts n the reference cloud, whereas other ponts on the same test cloud are located n a lower poston. Ths effect s more evdent for sources that apparently have a greater nfluence on clouds, such as the ncandescent ones. The expermentaton carred out confrms that laser dgtzng of surfaces n complete absence of external sources of lght provdes the best results. In the usual case that ths requrement can not be satsfed, results lead to recommend usng those sources of lght that cause less dstorton of the pont cloud. Accordng to ths, the recommended sources shall be the low pressure sodum lamp and the mercury vapour lamp. Sodum lamps emt n the orange range (589 nm) of the vsble spectrum, whch s especally annoyng when worng for long tme perods, as t dsables the operator for dstngushng dfferent colors. Ths leads to recommend mercury vapour lamps as the most approprate electon. Future wor wll deal wth the evaluaton of the nfluence of lght source orentaton upon the qualty of the dgtzed pont clouds. Fg. 6. Pont clouds obtaned under the same condtons when rotatng the reference surface REFERENCES [1] D. Hüser-Teuchert, E. Trapet, A. Garces, F. Torres-Leza, T. Pfefer, P. Scharsch. Performance test procedures for optcal coordnate measurng probes fnal project report. European Communtes. 1995. [2] D. Hüsser, H.Rothe. Robust averagng of sgnals for trangulaton sensors. Measurement Scence and Technology, vol. 9. 1998, pp.1017-1023. [3] Boehler W, M Bordas Vcent, A Marbs (2003). Investgatng laser accuracy. In: CIPA XIXth. Int. Symposum, 30 Sept.-4 Oct., Antalya, Turey, pp 696-701. [4] Blas, F. "A Revew of 20 Years of Ranges Sensor Development," SPIE Proceedngs, Electronc Imagng, Vdeometrcs VII. Santa Clara, Calforna, USA. Vol. 5013, 2003. pp 62-76.
Proceedngs of the World Congress on Engneerng 2008 Vol I [5] J.Forest, J.Salv, E. Cabruja, C.Pous. Laser strpe pea detector for 3D scanners. A FIR flter approach. Proceedngs of the 17th Internatonal Conference on Pattern Recognton, 2004 Vol.3 pp.646-649 [6] Feng H-Y., Lu Y. X F. Analyss of dgtzng errors of a laser scannng system. Journal of the Internatonal Socetes for Precson Engneerng and Nanotechnology, vol. 25, 2001, pp 185-191.