Measurement 58 (2014) 317–324

Contents lists available at ScienceDirect

Measurement

journal homepage: www.elsevier .com/ locate/measurement

A denoising method based on multiple-stage acousto-optictunable filter in the green laser measurement system

http://dx.doi.org/10.1016/j.measurement.2014.09.0070263-2241/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: School of Electrical Engineering, YanshanUniversity, Qinhuangdao City, Hebei Province, China.

E-mail address: binxf2013@126.com (X. Fu).

Yucun Zhang, HaiBin Song, Xianbin Fu ⇑Yanshan University, Qinhuangdao City 066004, China

a r t i c l e i n f o a b s t r a c t

Article history:Received 4 April 2014Received in revised form 23 August 2014Accepted 4 September 2014Available online 18 September 2014

Keywords:The green laser stripeCCDMultiple-stage acousto-optic tunable filterIntelligent filter

The charge handling capacity of CCD camera is limited, so image distortion is caused bybright light generated by the hot workpiece. The green laser stripe captured by CCD camerais fuzzy in the green laser measurement system. In order to make the image of the stripeclearer, the filter system is designed based on multiple-stage acousto-optic tunable filter.In the filter system, a computational model is built by combining the visible radiation char-acter of hot workpiece and the threshold power density of CCD saturation. The curve of thefiltering rate of bright light can be obtained using the computational model. In order torealize the goal of intelligent filter, each stage of the acousto-optic tunable filter is adjustedaccording to the curve of the filtering rate. Finally, the image of green laser stripe on hotworkpiece is clearer. The filter system designed in the paper is viable according to theexperiment results. And the accuracy of the green laser measurement system is improved.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction spectrum selective method is proposed by Jia Zhen-yuan.

For the green laser measurement system, the cleargreen laser stripe is necessary for improving the measure-ment accuracy. In the past decades, a lot of researches havebeen conducted by domestic and foreign scholars. Imagesof hot workpiece are dealt with by Zhao Zhuan-pingthrough the image processing technology, such as selectivefilter, image enhancement and smoothing. Then the influ-ence of part of ambient light on the image clarity of thegreen laser stripe is removed [1–3]. The influence of redthermal radiation on the image clarity is overcome by LiuGui-hua using the combination of digital filter technologyand physical filter technology [4,5]. The edge values ofthe object are obtained by Dr. Shiuh-Jer Huang using theimage processing procedures [6–8]. These methods areproposed from the point of image. The following methodsare from the point of the course of image acquisition. The

The self-emitted radiation character and reflection charac-ter of hot workpiece are analyzed. The light intensity issupplemented by the illumination lamp to acquire imagesof hot workpiece. Using this method, radiation light cap-tured by CCD is removed [9–11]. A free-electron model ofmetal is adopted by Ke Wei-na. By using the model, themetal surface irradiated with green laser is simulated.The impact of angle of incidence on metal absorption rateis presented [12]. Bulygin and Kovalev measure the qualityof laser beams by methods of fourier optics. The quality oflaser beams is improved by a Fourier transform [13–16].The influence of angle of incidence, object color and mea-suring distance on the laser scanning process in computerare analyzed by Nikola Vukašinovic and Drago Bracun [17].Thereby it is necessary to find a more effective solution tothe fuzzy green laser stripe for the green laser measure-ment system.

The filter system is designed to make the image of thegreen laser stripe clearer on hot workpiece. In the filtersystem, the computational model is built by combiningthe visible radiation character of hot workpiece and the

Rf Power Driver

The control computer

The laser device

AOTF1 AOTF2 AOTF3 CCD

The temperature measuring equipment

Fig. 2. The schematic of the filter system.

318 Y. Zhang et al. / Measurement 58 (2014) 317–324

threshold power density of CCD saturation. The curve of fil-tering rate of bright light is obtained by using the compu-tational model. According to the curve of the filtering rate,each stage of the acousto-optic tunable filter is adjusted bychanging the radio frequency (RF) signal and the ultrasonicpower. Finally, the image of green laser stripe on hot work-piece is made clearer. And the accuracy of the green lasermeasurement system is improved.

2. The green laser measurement system

The diagram of the measuring principle is shown inFig. 1. The green laser device is installed on the linear guiderail, and the device is moved by the servo motor. Firstly,the green laser device is moved to the left edge of thehot workpiece, and the location of the device is recordedby the control computer. Then the green laser device ismoved to the right edge of the hot workpiece, and the loca-tion of the device is recorded by the control computer too.Finally, the measurement data of the length of the hotworkpiece is displayed through the control computer pro-cessing. But the image of green laser stripe on hot work-piece is fuzzy. The fuzzy stripe results in the lowmeasurement accuracy. In order to improve the measure-ment accuracy, the filter system is designed.

The filter system is mainly composed of five parts suchas the temperature measuring device, the laser device, themultiple-stage acousto-optic tunable filter, CCD and thecontrol computer. The filter system is shown in Fig. 2.

Firstly, the temperature of the hot workpiece is mea-sured by the temperature measuring device and is receivedby the control computer. The temperature value is input tothe computational model. Then the curve of filtering rate ofbright light is obtained. Each stage of the acousto-optictunable filter is adjusted according to the curve. Finally,the light is received by CCD camera, after it goes throughthe filter system. And the clear image of green laser stripeon hot workpiece is obtained.

3. The system principle of multiple-stage acousto-optictunable filter

3.1. The way of obtaining the curve of filtering rate

The curve of filtering rate of bright light is the precondi-tion of working well for the filter system. The curve of

The length

Fig. 1. Schematic of the green laser measurement system.

filtering rate is obtained by combining the visible radiationcharacter of hot workpiece and the threshold power den-sity of CCD saturation. According to the curve of filteringrate, the visible radiation that causes image distortion isremoved.

3.1.1. The visible radiation character of hot workpieceWhile the hot workpiece is captured by CCD camera,

the temperature of the hot workpiece can reach 1200 �C.The visible radiation of hot workpiece is very strong. Theimage of the green laser stripe is fuzzy. It is necessary toobtain the visible radiation character of hot workpiecefor making the image of the stripe clearer.

According to Planck radiation law, the visible radiationcharacter of hot workpiece is connected with temperatureof hot workpiece. Eq. (1) is Planck radiation law:

Mkðk; TÞ ¼ c1 �k�5

expðc2=kTÞ � 1ð1Þ

where c1 is the first radiation constant, and c2 is the secondradiation constant. c1 and c2 can be written as:

c1 ¼ 2pc2h ¼ ð3:741832� 0:000020Þ � 10�10 W m2 ð2Þc2 ¼ ch=k ¼ ð1:438786� 0:000045Þ � 10�2 m K ð3Þ

where h is the Planck Constant, k is the boltzmann con-stant, and c is the velocity of light. h, k and c can be writtenas:

h ¼ ð6:626176� 0:000036Þ � 10�34 J=K

k ¼ ð1:380662� 0:000044Þ � 10�23 J=K

c ¼ 3� 108 m=s

While the hot workpiece is forged, the temperaturerange of the hot workpiece is 700–1200 �C. So there are 3temperature values (800 �C, 1000 �C and 1200 �C) of hotworkpiece chosen to be researched. The values of the tem-perature are respectively substituted into Eq. (1). Then thecurves of the visible radiation character under differenttemperature are obtained, and the curves are shown inFig. 3.

Y. Zhang et al. / Measurement 58 (2014) 317–324 319

As shown in Fig. 3, when the temperature of hotworkpiece is 1200 �C, the center wavelength of radiationis 1.967 lm. When the temperature of hot workpieceis 800 �C or 1000 �C, the center wavelength of radiation isabove 2 lm. All center wavelengths of radiation are above600 nm. And the visible radiation light is mainly red light.The visible radiation is very strong. With the temperatureof hot workpiece becoming low, the visible radiation isdecreased obviously.

3.1.2. The threshold power density of CCD saturationThe limited power density of the incident light is deter-

mined by the threshold power density of CCD saturation. Itis necessary to obtain the threshold power density of CCDsaturation at different wavelengths for making the imageof green laser stripe clearer.

Electric charge is the internal signal of CCD. The basicfunction of CCD is the charge storage and the charge trans-fer. The depth of the potential well of CCD is limited, andthe potential well of CCD cannot hold electrical chargeunlimitedly. Therefore, bright light causes imagedistortion.

Inside the CCD camera, for the MOS capacitor in deepdepletion condition, the voltage relation is shown asfollows:

VG þ Vs0 ¼ VCi þ Vs ð4Þ

where VG is the grid voltage, Vs0 is the surface potentials ofthe CCD unit which the grid voltage is not applied to, VCi isthe electric potential difference of insulating layer, and Vs

is the electric potential difference between semiconductorand the surface of semiconductor. Vs, VCi and Vs0 can bewritten as:

Vs ¼1

2e0eieNAx2

d ð5Þ

VCi ¼eS NAxd þ Nexd0ð Þ þ Q s

Cið6Þ

Vs0 ¼ Q fc=Ci � Vms ð7Þ

where NA is the acceptor impurity concentration of semi-conductor, Ne is electron concentration of semiconductorin thermal equilibrium, Qs is the signal quantity of electric-ity, S is the sectional area of deep depletion layer, xd0 is the

Fig. 3. The curves of the visible radiation character of hot workpieceunder different temperature.

depth of depletion layer before the semiconductor is irradi-ated, xd is the depth of depletion layer after the semicon-ductor is irradiated, Vms is the contact potentialdifference between metal and semiconductor, Qfc is thequantity of electricity in the oxide layer, and Ci is thecapacitance of MOS capacitor. Ci can be written as:

Ci ¼ e0eiS=d ð8Þ

where ei is the dielectric constant of insulating layer, and dis the thickness of insulating layer.

Substituting Eqs. (5)–(7) into Eq. (4), we obtain:

xdðQ sÞ ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffid2 þ 2d

eSNAðQt � QsÞ

s� d ð9Þ

Z Qs

0

dQs

1� e�axd¼ eSg0qt0 ð10Þ

According to Eqs. (9) and (10), Eq. (11) can be obtained:

Q s ¼ esgpt ð11Þ

With the signal charge increasing, the electric potentialdifference Vs is decreased. When the electric potentialdifference Vs is equal to Vs0 near the MOS capacitance,the signal charges are diffused into the potential well nearthem. Therefore, the normal image signal cannot beobtained.

Substituting Qs0 ¼ VGc1 � sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1

4p eeNAVs0

qinto Eq. (11), we

obtain:

p ¼eVG �

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi8pd2eeNAVs0

q4pdeg tmin

ð12Þ

where e is the relative dielectric constant, and tmin is theshutter time.

According to Eq. (12), the limited power density of theincident light can be obtained. This is the threshold powerdensity of CCD saturation.

Combining Eqs. (1) and (12), the diagram is obtained byMatlab, and shown in Fig. 4. In the diagram, the powerdensity of the visible radiation of hot workpiece is againstthe threshold power density of CCD saturation under dif-ferent temperature. The normalized data processing isused in the diagram.

As is shown in Fig. 4, when the temperature of hotworkpiece is 1200 �C, the power density of the visible radi-ation exceeds the threshold power density of CCD satura-tion between 650 nm and 1050 nm. So the visibleradiation between 650 nm and 1050 nm causes image dis-tortion. When the temperature of hot workpiece is 1000 �C,image distortion is caused by the visible radiation inbetween 760 nm and 920 nm. When the temperature ofhot workpiece is 800 �C, the power density of the visibleradiation does almost not exceed the threshold power den-sity of CCD saturation.

According to Fig. 4, the curve of filtering rate of brightlight under different temperature is obtained.

The curve is shown in Fig. 5.As is shown in Fig. 5, the filtering rate of bright light is

different at different wavelengths. According to the curveof the filtering rate, each stage of the acousto-optic tunablefilter is adjusted.

Fig. 4. The diagram where the power density of the visible radiation ofhot workpiece is against the threshold power density of CCD saturationunder different temperature.

320 Y. Zhang et al. / Measurement 58 (2014) 317–324

3.2. The principle of multiple-stage acousto- optic tunablefilter

After the light from hot workpiece goes through thefirst stage of the acousto-optic tunable filter, the light isdivided into ±1 order diffraction light and zero-orderdiffraction light. The zero-order diffraction light isnot diffracted. Then the zero-order diffraction light goesthrough the second stage of the acousto-optic tunable fil-ter. So after the function of one stage of the acousto-optictunable filter, two light beams of different wavelengthare removed. The wavelength of the diffracted light andthe acousto-optic diffraction efficiency are adjustedaccording to the curve of filtering rate of bright light torealize the goal of intelligent filter. The schematic of multi-ple-stage acousto-optic tunable filter is shown in Fig. 6.

The RF signal of the fixed frequency is applied to theultrasonic transducer, and it is coupled to birefringentcrystal. The refractive index of the crystal is changed peri-odically. For the fixed acousto-optic medium and the fixeddirection of propagation, the tuning relationship of acou-sto-optic tunable filter can be written as:

k ¼ DnVf

sin2 2hi þ sin4 hi

h i1=2ð13Þ

where k is the wavelength of the light diffracted, V is theultrasonic velocity, hi is the incident polar angle of light

Fig. 5. The curve of filtering rate of bright light under differenttemperature.

wave, f is the frequency of the RF signal, and Dn is therefractive index difference.

According to Eq. (13), the wavelength of the light dif-fracted is relevant to the frequency of the RF signal. Sothe wavelength of the light diffracted is adjusted by chang-ing the frequency of the RF signal.

3.2.1. The acousto-optic diffraction efficiency of each stage ofthe acousto-optic tunable filter

The spectral penetration function of acousto-optic tun-able filter is the ratio between the diffracted light intensityand the intensity of incident light. The spectral penetrationfunction can be written as:

h ¼ C2L2 sin C2 C2 þ DK2

4

!1=2

� L

24

35 ð14Þ

where L is the length of interaction between ultrasoundand light, DK is the momentum mismatch, and C is:

C ¼ n6p2p2Pa=ð2k20qV3AÞ ð15Þ

where q is the density of the medium, n is the refractiveindex of the medium, p is the photoelastic constant ofthe medium, Pa is acoustical power, and A is the sectionalarea of acoustic beam and light beam.

DK is shown as follows:

DK ¼ sin2 hi

k0

b � Dkk0þ p � Dn Fh � a2 þ Ghb

2� �� �ð16Þ

where k0 is the wavelength of the light in a vacuum, Dk isthe wavelength difference, Dn is the birefringence of themedium, b is the dispersion constant of the medium, hi isthe incident polar angle of light wave, a is polar angle ofincident light relative to condition of momentum match,and b is the deviation of azimuth angle. Fh = 2ctg2hi � 1,Gh = cos2hi � 1. When the incident light is parallel light,Eq. (16) can be simplified as:

DK ¼ b � sin2 hi � Dk0=k0 ð17Þ

Substituting Eq. (17) into Eq. (14), we can obtain:

h ¼ C2L2 sin C2 C2 þ b2 sin4 hi

4k0� Dk2

0

!1=2

� L

24

35 ð18Þ

According to Eq. (18), after setting the parameters ofacousto-optic tunable filter, the spectral penetration func-tion of acousto-optic tunable filter is adjusted by changingultrasonic power Pa. The spectral penetration function ofacousto-optic tunable filter is relevant to the acousto-opticdiffraction efficiency of acousto-optic tunable filter. Thenthe acousto-optic diffraction efficiency of acousto-optictunable filter is adjusted by changing ultrasonic power Pa.

3.2.2. The relationship between +1 order diffraction lightand �1 order diffraction light

When the frequency of RF signal is input in the RFpower drive, two light beams of different wavelength arediffracted. Two light beams of different wavelength are+1 order diffraction light and �1 order diffraction light.

80 90 100 110 120 130 140

0.7

0.8

0.9

1

1.1

Ultrasonic frequency/MHz

Wav

elen

gth/

µm

the diffracted light+1

the diffracted light -1

Fig. 7. The theoretical curve of the relationship between +1 orderdiffraction light and �1 order diffraction light.

7 8 9 10 11

x 10-7

0

0.2

0.4

0.6

0.8

1

Wavelength/m

Diff

ract

ion

effic

ienc

y

1231200

Fig. 8. The curve of the simulation of filter under 1200 �C.

Fig. 6. The schematic of multiple-stage acousto-optic tunable filter.

Y. Zhang et al. / Measurement 58 (2014) 317–324 321

The relationship between +1 order diffraction light and �1order diffraction light can be written as:

f acoðhi; kÞ ¼ Vak n2

ie hi; kð Þ þ n2doðkÞ

��2nie hi; kð ÞndoðkÞ cos hdoðhi; kÞ � hi½ �g1=2

f aco hi; kð Þ ¼ Vak n2

ioðhi; kÞ þ n2deðkÞ

��2nio hi; kð ÞndeðkÞ cos hdeðhi; kÞ � hi½ �g1=2

8>>>><>>>>:

ð19Þ

where faco is the ultrasonic frequency, Va is the ultrasonicvelocity, nie is the refractive index of the medium whenincident light is +1 order diffraction light, ndo is the refrac-tive index of the medium when the light is �1 order dif-fraction light, k is the wavelength of the light in avacuum, hdo is the angle between �1 order diffraction lightand the optical axis of crystal, and hde is the angle between+1 order diffraction light and the optical axis of crystal.

According to Eq. (19), the relationship between +1 orderdiffraction light and �1 order diffraction light is simulatedby Matlab. The theoretical curve of the relationshipbetween +1 order diffraction light and �1 order diffractionlight is shown in Fig. 7.

As shown in Fig. 7, the center wavelengths of +1 orderdiffraction light and �1 order diffraction light is differentunder circumstances of the same ultrasonic frequency.The gap between the two wavelengths is 30–40 nm.

4. The simulation of filter

The multiple-stage acousto-optic tunable filter is theimportant part of the green laser measurement system.The quality of the filter system has direct impact on themeasurement accuracy. Then it is necessary to conductthe simulation of filter for preferences. The curve of thesimulation of filter under 1200 �C is shown in Fig. 8.

The wavelength of the light diffracted is adjusted bychanging the frequency of the RF signal. The acousto-opticdiffraction efficiency of each stage of the acousto-optictunable filter is adjusted by changing ultrasonic powerPa. The wavelength of the diffracted light and the acou-sto-optic diffraction efficiency are adjusted in real timeaccording to the curve of filtering rate. The center wave-length of the first stage of the acousto-optic tunable filteris 800 nm. The center wavelength of the second stage ofthe acousto-optic tunable filter is 880 nm. The center

wavelength of the third stage of the acousto-optic tunablefilter is 940 nm. The acousto-optic diffraction efficiency ofeach stage of the acousto-optic tunable filter is respectivelyset to 83%, 72% and 49%. What is more,+1 order diffractionlight is removed with �1 order diffraction light. Finally, thecurve of filtering rate of bright light is met. The curve of thesimulation of filter under 1000 �C is shown in Fig. 9.

In Fig. 9, because the curve of filtering rate of brightlight under 1000 �C is narrow, two stage of the acousto-optic tunable filter can meet the requirement for the curveof filtering rate of bright light. The center wavelength of

Fig. 10. The image of the heating furnace.

322 Y. Zhang et al. / Measurement 58 (2014) 317–324

the first stage of the acousto-optic tunable filter is 810 nm.The center wavelength of the second stage of the acousto-optic tunable filter is 870 nm. The acousto-optic diffractionefficiency of each stage of the acousto-optic tunable filter isrespectively set to 60% and 45%.

5. Testing experiment

In order to verify the validity of the filter system, anexperiment platform is designed by our research group.Equipments of the experiment platform are as follows:

(1) The heating furnace used is model C19-P1600B2.The heating temperature range of the heating fur-nace is 500–1550 �C. The image of the heating fur-nace is shown in Fig. 10.

(2) Industrial CCD camera used is of model MV-VE078SM/SC. The maximum resolution is 1024� 768, and the size of a pixel on the CCD array is4.65 lm � 4.65 lm. The focal length of the opticallens is 12 mm, and its back focal length is 9.7 mm.The relative aperture (F) is 1:1.4. The field of viewis 2/300. Working distance is 4 m.

(3) The line laser projector is of MGL-III type. Wave-length is 532 nm. The aperture angle is 30�. Servomotors and drives are of MR-J2S-10A/B type. Themaximum error of the drive is ±10 turns.

(4) High-precision guide used is BGXS45BE-type highprecision linear guide rail, and the walking error isless than 10 lm.

(5) The filter system based on multiple-stage acousto-optic tunable filter is designed by us.

The experimental environment is no wind with dust.The ambient light is lamplight. The indoor temperature is300 K and the humidity is 60%.

The procedure of this experiment is given as follows:

(1) The 45#steel workpiece is heated to 800 �C by theheating furnace. In order to overcome temperaturevariations and surface parameters variations, as faras possible, the measurement time should be cut.

(2) The hot workpiece is placed at a distance of fourmeters from the laser transmitter.

7 8 9 10 11

x 10-7

0

0.2

0.4

0.6

0.8

1

Wavelength/m

Diff

ract

ion

effic

ienc

y 121000

Fig. 9. The curve of the simulation of filter under 1000 �C.

(3) The filter system based on multiple-stageacousto-optic tunable filter is installed. The mea-surement system is launched. The green laser stripeis moved to the left edge of the hot workpiece by thecontrol system. Then the green laser stripe is movedto the right edge of the hot workpiece. Themeasurement data is received and displayed.The measurement process is repeated 10 times andthe measurement data are recorded.

(4) The filter system based on multiple-stageacousto-optic tunable filter is uninstalled. Themeasurement process is repeated 10 times andthe measurement data are recorded.

(5) The above measurement process is repeated under1000 �C and 1200 �C. The measurement data arerecorded.

The error under 800 �C is shown in Fig. 11. After the fil-ter system is installed, the error is reduced weakly. Becausethe curve of filtering rate of bright light indicates that thevisible radiation that causes image distortion is weakunder 800 �C. The effect of the filter system is small.

The error under 1000 �C and 1200 �C are shown in Figs.12 and 13. After the filter system is installed, the error isreduced greatly. Because the curve of filtering rate ofbright light indicates that the visible radiation that causesimage distortion is intense under 1000 �C and 1200 �C.

When the filter system is uninstalled under 1200 �C, theimage is shown in Fig. 14(a). When the filter system isinstalled under 1200 �C, the image is shown in Fig. 14(b).The laser stripe on hot workpiece in Fig. 14(b) is clearerthan Fig. 14(a).

The mean square error of the measurement data isshown in Table 1. According to Table 1, when the filter sys-tem is applied to the measurement system, the meansquare error is reduced under different temperature. Butthe decreasing amplitude is different under different tem-perature. The decreasing amplitude under 1000 �C and1200 �C is bigger than that under 800 �C. Because the curveof filtering rate of bright light indicates that the visibleradiation that causes image distortion is intense under1000 �C and 1200 �C.

Table 1The mean square error under different temperature.

Temperature(�C)

The mean square errorwhen the filter system isuninstalled

The mean square errorwhen the filter systemis installed

800 1.581 1.1861000 4.778 1.8451200 6.598 1.098

Fig. 14. The images received by the control computer.

2 4 6 8 100

1

2

3

4

Sequence Number

Erro

r/cm

The filter system is uninstalledThe filter system is installed

Fig. 13. The error under 1200 �C.

2 4 6 8 100

0.5

1

1.5

2

2.5

3

Sequence Number

Erro

r/cm

The filter system is uninstalledThe filter system is installed

Fig. 11. The error under 800 �C.

2 4 6 8 100

1

2

3

4

Sequence Number

Erro

r/cm

The filter system is uninstalledThe filter system is installed

Fig. 12. The error under 1000 �C.

Y. Zhang et al. / Measurement 58 (2014) 317–324 323

The accuracy is shown as follows:d = (Dmax)/(Amax) � 100%, where d is the accuracy, Dmax isthe maximum error of the measurement, and Amax is theinstrument range. The green laser device is installed onthe linear guide rail. The instrument range is determinedby the length of the linear guide rail. The length of the lin-ear guide rail used is 360.0 cm. According to the measure-ment data, when the filter system is uninstalled under800 �C, the accuracy is: d = 1.6/360.0 � 100% = 0.44%.When the filter system is installed under 800 �C, the accu-racy is: d = 1.5/360.0 � 100% = 0.41%. When the filter sys-tem is uninstalled under 1000 �C, the accuracy is: d = 2.9/360.0 � 100% = 0.81%. When the filter system is installed

under 1000 �C, the accuracy is: d = 1.7/360.0 � 100%= 0.47%. When the filter system is uninstalled under1200 �C, the accuracy is: d = 3.3/360.0 � 100% = 0.92%.When the filter system is installed under 1200 �C, the accu-racy is: d = 1.8/360.0 � 100% = 0.50%. The accuracy of thegreen laser measurement system under different tempera-ture is shown in Table 2. According to Table 2, the accuracyof the green laser measurement system under differenttemperature is improved while the filter system isinstalled. It can meet the requirement of accuracy.

In order to verify that the denoising method in thispaper is better, the measuring principle should be thesame, but the denoising method is different. Comparedwith other methods [18], the accuracy of the measurementsystem is improved by the denoising method in this paper.

Table 2The accuracies of the green laser measurement system under differenttemperature.

Temperature(�C)

The accuracy when thefilter system isuninstalled (%)

The accuracy when thefilter system isinstalled (%)

800 0.44 0.411000 0.81 0.471200 0.92 0.50

324 Y. Zhang et al. / Measurement 58 (2014) 317–324

6. Conclusions

In this paper, the filter system is designed based onmultiple-stage acousto-optic tunable filter. When the filtersystem is applied to the measurement system, the accu-racy of measurement is improved from 0.81% to 0.47%under 1000 �C and the accuracy of measurement isimproved from 0.92% to 0.50% under 1200 �C. The rangeof the improved accuracy under 1000 �C and 1200 �C isgreat. Because the curve of filtering rate of bright light indi-cates that the visible radiation that causes image distortionis intense. The visible radiation that causes image distor-tion is removed by the multiple-stage acousto-optic tun-able filter. The tuning range of the acousto-optic tunablefilter is large, and the adjusting speed of the acousto-optictunable filter is high. The image of green laser stripe on hotworkpiece is made clearer. The dimension measurement ofhot workpiece online is realized. The validity of the filtersystem is verified through the simulation of filter and theexperiment.

Acknowledgement

This study is supported by the Natural Science Founda-tion of Hebei Province, China (Grant No.: E2014203070).

References

[1] L. Li, Z.P. Zhao, Research on dimensional measurement of heavycylindrical forging at high temperature based on binocular stereovision, Transducer Microsyst. Technol. 29 (4) (2010) 49–51.

[2] H. Zhang, L.Y. Zhang, Z.P. Zhao, Field calibration of binocular stereosystem based on planar template and free snapping, Acta Aeronaut.Astronaut. Sinica 28 (3) (2007) 695–701.

[3] Y. Xue, Z.P. Zhao, Measurement system of tracking contour alongedge based on visible technology, Mach. Des. Manuf. 7 (2009)209–210.

[4] G.H. Liu, X.Y. Liu, Online measurement techniques of large forgingsin thermal state, Opto-Electron. Eng. 37 (9) (2009) 91–97.

[5] L.J. Wu, G.H. Liu, Shiny surfaces measurement based on structuredlight system, Tool Eng. 45 (3) (2011) 88–92.

[6] D.S. Huang, Y.W. Lin, A prototype system of three-dimensional non-contact measurement, Int. J. Adv. Manuf. Technol. 11 (1996)336–342.

[7] D.S. Huang, Y.W. Lin, A three-dimensional non-contact measurementsystem, Int. J. Adv. Manuf. Technol. 13 (1997) 419–425.

[8] X.B. Fu, Y.C. Zhang, An optical non-contact measurement method forhot-state size of cylindrical shell forging, Measurement 45 (6) (2012)1343–1349.

[9] Z.Y. Jia, B.G. Wang, W. Liu, An improved image acquiring method formachine vision measurement of hot formed parts, J. Mater. Process.Technol. 210 (2010) (2009) 267–271.

[10] W. Liu, S.J. Liu, Z.Y. Jia, Study on online measurement system ofdimension for hot heavy forging, Piezoelect. Acoustoopt. 33 (5)(2011) 831–836.

[11] W. Liu, Z.Y. Jia, An improved online dimensional measurementmethod of large hot cylindrical forging, Measurement 45 (8) (2012)2041–2051.

[12] W.N. Ke, D.Q. Zhu, G.B. Cai, Simulation and analysis of spectralemissivity of metals, Acta Aeronaut. Astronaut. Sinica 31 (11) (2010)2139–2145.

[13] F.V. Bulygin, A.A. Kovalev, K.D. Maramzin, Measurement of thequality of laser beams by methods of fourier optics, Measur. Tech. 49(10) (2006) 1015–1020.

[14] M. Jouaneh, R.L. Lemaster, D.A. Dornfeld, Measuring workpiecedimensions using a non-contact laser detector system, Int. J. Adv.Manuf. Technol. 2 (1) (1987) 59–74.

[15] C. Ferri, J. Faraway, E. Brousseau, Calibration of a white lightinterferometer for the measurement of micro-scaledimensions, Int. J. Adv. Manuf. Technol. 47 (1–4) (2010)125–135.

[16] Y. Hojjat, M.R. Karafi, M. Ghanbari, M. Lotfi, Development of aninductive encoder for simultaneous measurement of twodimensional displacement, Int. J. Adv. Manuf. Technol. 5–8 (2011)681–688.

[17] N. Vukašinovic, D. Bracun, J. Mozina, J. Duhovnik, The influence ofincident angle, object colour and distance on CNC laser scanning, Int.J. Adv. Manuf. Technol. 50 (2010) 265–274.

[18] J.S. Pan, Y.M. Liu, The experimental research of large forgingsdimension measuring system, J. Northeast Heavy Mach. Inst. 19 (4)(2005) 303–307.