keylei/White_Light_Triangulation_Sensor_for_Flexible_Inspection_System_test

text stringlengths 0 1.24k
2015 Ninth International Conference on Sensing Technology

==title==
White Light Triangulation Sensor for Flexible Inspection System

==authors==

Abraham Mario Tapilouw*, Yi-Wei Chang, Hsiao-Wei Liu, Hau-Wei Wang, and Hung-Ming Tai
Center for Measurement Standards,
Industrial Technology Research Institute,
Hsinchu City, Taiwan
*email: mario@itri.org.tw

==Abstract==
A triangulation sensor containing two light sources with different wavelength and individual slit has been developed. Differential technique is applied to improve spatial resolution as performing height reconstruction of a specimen. The system can measure up to 7 mm with accuracy up to 1.144 um. Robotic arm with multiple degrees of freedom can be employed as the motion platform to increase flexibility of inspection.

==Keywords==
triangulation; white light; flexible inspection system; sensor; robotic arms

==INTRODUCTION==
Laser based optical triangulation system has been widely applied in industry for measuring distance and also surface profile [1, 2]. One of the drawbacks of laser triangulation is it produces speckle and affects measurement uncertainty [3]. Triangular method using low-coherence white light and a slit would reduce speckle. For laser or white light, width of the
projected line profile and intensity of the light determines the resolution and accuracy of the measurement. Producing a narrow line source with high intensity requires a strong light
source. However, efficiency of the light source is very low because most of the light is blocked by the narrow slit.

In this paper, a white light triangulation sensor is presented.
The study aims to develop a system with multiple color LED
light sources achieving higher depth resolution without
sacrificing light intensity. A differential technique is applied to
calculate the height of the sample. Measurement range of the
sample is 7 mm and the accuracy of the measurement is up to
1.144 um.

==SYSTEM DESCRIPTION==
Fig. 1 shows the optical system design of the developed system. The system consists of light source module and color imaging device for capturing images.

Lateral shift in the image plane is determined by calculating the local peak of the line profile. Therefore the resolution of the peak detection is determined by the width of the line profile; the narrower and sharper the line profile the higher the resolution of peak detection is. In the developed system, a narrower and sharper line profile can be obtained by subtraction of the two line profiles with different colors. Proper adjustment, alignment, and calibration of the light sources are needed to ensure that the two line profiles are
overlapping. The receiving angle of the camera also has to be adjusted to the enable capturing image of the wholemeasurement range without obtaining defocused images due to limitation of lens’ Depth of Field. The differential technique applied for calculating the peak of the line profile is shown in Fig. 2. A captured image is first separated into red and blue channel to obtain line profile from red and blue LEDs. Red and blue LEDs are chosen to match spectral response of the employed camera to avoid spectrum overlap. Intensity of the red light is then subtracted with the blue light to obtain the differential profile. Positive value of differential intensity is then normalized to generate two zero crossing points which allow distinctive separation of the region for peak calculation from other region (shown by the blue line on Figure 2). Position of the peak is calculated from the differential profile. By applying this technique the Full Width Half Maximum (FWHM) of the profile is smaller than
the original red line profile.


==CALIBRATION PROCESS==
Eq. 1 shows the relation between image coordinate ( ∆x') and actual height ( ∆z ) in triangulation measurement. Therefore calibration is first performed to generate calibration curve between pixel position and actual sample height. Calibrated gauge block samples are employed as reference. After the calibration curve is obtained, the triangulation sensor can be used for measurement. Figure 3 (a) and (b) show the image captured by camera and the detected peaks after applying differential calculation technique, respectively. Resolution of the image is estimated by using USAF 1951 standard target and it is obtained that each pixel is 12.72 um.

In Figure 3, the distance between peaks represents the height difference between two gauge blocks. The smallest height difference that can be made by the gauge blocks is 10 um.Therefore, three height differences are tested, 10 um for the range of 1.0-1.1 mm, 100 um for the range of 1.1-2 mm, and 1 mm for the range of 2-5 mm. Figure 4 shows the calibration curve obtained from the measurement of gauge blocks. Mean Absolute Error (MAE) of the calibration curve for the whole calibrated range is 10.83 um.

After the system has been calibrated, it is put to repeatability test. Each gauge block with height ranging from 1.1 – 1.9 mm is measured five times to estimate the repeatability of the system. Table 1 shows the static repeatability measurement results. Repeatability of the measurement ranges from 0.3 – 1.2 um.

Static and dynamic repeatability of the sensor is also tested by measuring a flat mirror with accuracy of 1/10λ mounted on a linear stage. The linear stage is equipped with a linear encoder, which is used as position indication. Stage is moved at a total distance of 7 mm and the image is acquired at every 50 um. For each position, the acquisition is repeated 35 times.

Repeatability of the static measurement is 2.06 um (1 standard deviation) and. Fig. 5 shows the measured relative displacement at each sampling point. It is shown that the measurement fluctuates more at the end of the measurement range. The standard deviation is 3.41 um and the range of the data is 14.18 um. Accuracy of the peak detection for the whole measurement range is 1.144 um (0.07 pixels).

==EXPERIMENT RESULTS==
Aspect ratio is one of important parameters of measurement system. It is tested by measuring gap built by two gauge blocks. Figure 6 shows the sample used in the measurement. Thicknesses of the gauge blocks are 5 mm and the gap between two gauge blocks is 1 mm. Aspect ratio of the sample is 5:1.

Figure 7(a) shows the image captured by the camera and Figure 7(b) shows the peak detection result.

Figure 8 shows the measured profile of the sample. The red region shows the profile for calculating the depth. The measured depth of the gap is 4.980 mm. This result shows the capability of measuring sample with aspect ratio of 5:1.

To validate the performance of the system in measuring real workpiece, this study uses the developed system is to measure the depths of the piston oil ring grooves as shown in Fig. 9. There are several ring grooves to be measured for the inspection. Measuring piston oil ring grooves by a side-view camera had been found to yield unstable accuracy due to alignment errors and image distortion. Instead, pistons then were measured by using the coordinate measuring machine before the new triangular sensor was introduced.

In the test, we focus on one piston groove at a time to get an image. Figure 10 (a) shows the portion of the piston to be measured and Figure 10 (b) shows the captured image in section 1-3.Measured profile is shown in Figure 11. The portion for calculating depth is marked by red dots.

The measurement is performed 10 times and the results are compared to measurement with Coordinate Measuring Machine (CMM) as reference instrument for measurement mechanical components as shown in Figure 9. Table 2 summarizes the comparison of the measurement results with the reference instrument.

The developed system is mounted on robot arm as its motion platform. Table 2 shows the measurement when position of the developed sensor is not changed during repetitive measurement. Another measurement is performed by moving the robot arm into other position and return to the measurement position to test the repeatability of the positioning combined with the repeatability of the measurement system. In the same condition and same position, the measurement is performed 10 times and the result is the repeatability of measurement is 0.0077 and 0.0178 mm for height measurement of section 1-2 and section 2-3 respectively. Parts of measurement difference are contributed by the positioning error of the robot arm. Ambient light also contributes to measurement error by increasing the background intensity because the light source is in visible spectrum. However, due to small receiving angle of the camera and implementation of normalization technique, the effect of background intensity can be neglected. After comparing measurements results performed with and without ambient lighting, the average difference between two measurements is 0.2 um. Therefore, it’s proven that the effect of ambient lighting to measurement errors is negligible.

After the system is calibrated, it is used to measure a sample that is hold by other robot arm. Figure 12 shows the setup of the developed system mounted on the robotic arm (left hand side) and another robotic arm holding the piston to be measured (right hand side). Using this approach, the system can be applied for measuring free form shapes and surfaces. The measurement head can be interchangeable as one would like to use other heads to measure different parameters.

==CONCLUSIONS==
A white light triangulation sensor employing multiple light sources has been developed. The system is capable of generating narrow line profile without sacrificing light intensity. The accuracy of the measurement is up to 1.144 um for the whole measurement range. Repeatability of the measurement can be between 0.2~1.2 um for measured height range of 1.1-1.9 mm. The system can measure sample with up to 5:1 aspect ratio.Repeatability of the system for measuring the piston groove is up to 6.9 um in static measurement and up to 8.2 um in dynamic measurement. The repeatability of dynamic measurement is affected by positioning accuracy of the robotic arm

==REFERENCES==
[1] Costa, M.F.M. Surface inspection by an optical triangulation method. Opt. Eng. 1996, 35, 2743–2747.
[2] Zeng, L.J.; Yuan, F.; Song, D.Q.; Zhang, R. A two-beam laser triangulation for measuring the position of a moving object. Opt. Lasers Eng. 1999, 31, 445–453.
[3] R. Dorsch, G. Häusler, and J. Herrmann, "Laser triangulation: fundamental uncertainty in distance measurement," Appl. Opt. 1994, 33, 1306-1314
2015 Ninth International Conference on Sensing Technology

==title==
White Light Triangulation Sensor for Flexible Inspection System

==authors==

Abraham Mario Tapilouw*, Yi-Wei Chang, Hsiao-Wei Liu, Hau-Wei Wang, and Hung-Ming Tai
Center for Measurement Standards,
Industrial Technology Research Institute,
Hsinchu City, Taiwan
*email: mario@itri.org.tw

==Abstract==
A triangulation sensor containing two light sources with different wavelength and individual slit has been developed. Differential technique is applied to improve spatial resolution as performing height reconstruction of a specimen. The system can measure up to 7 mm with accuracy up to 1.144 um. Robotic arm with multiple degrees of freedom can be employed as the motion platform to increase flexibility of inspection.

==Keywords==
triangulation; white light; flexible inspection system; sensor; robotic arms

==INTRODUCTION==
Laser based optical triangulation system has been widely applied in industry for measuring distance and also surface profile [1, 2]. One of the drawbacks of laser triangulation is it produces speckle and affects measurement uncertainty [3]. Triangular method using low-coherence white light and a slit would reduce speckle. For laser or white light, width of the
projected line profile and intensity of the light determines the resolution and accuracy of the measurement. Producing a narrow line source with high intensity requires a strong light
source. However, efficiency of the light source is very low because most of the light is blocked by the narrow slit.

In this paper, a white light triangulation sensor is presented.

2015 Ninth International Conference on Sensing Technology

==title==

White Light Triangulation Sensor for Flexible Inspection System

==authors==

Abraham Mario Tapilouw*, Yi-Wei Chang, Hsiao-Wei Liu, Hau-Wei Wang, and Hung-Ming Tai

Center for Measurement Standards,

Industrial Technology Research Institute,

Hsinchu City, Taiwan

*email: mario@itri.org.tw

==Abstract==

A triangulation sensor containing two light sources with different wavelength and individual slit has been developed. Differential technique is applied to improve spatial resolution as performing height reconstruction of a specimen. The system can measure up to 7 mm with accuracy up to 1.144 um. Robotic arm with multiple degrees of freedom can be employed as the motion platform to increase flexibility of inspection.

==Keywords==

triangulation; white light; flexible inspection system; sensor; robotic arms

==INTRODUCTION==

Laser based optical triangulation system has been widely applied in industry for measuring distance and also surface profile [1, 2]. One of the drawbacks of laser triangulation is it produces speckle and affects measurement uncertainty [3]. Triangular method using low-coherence white light and a slit would reduce speckle. For laser or white light, width of the

projected line profile and intensity of the light determines the resolution and accuracy of the measurement. Producing a narrow line source with high intensity requires a strong light

source. However, efficiency of the light source is very low because most of the light is blocked by the narrow slit.

In this paper, a white light triangulation sensor is presented.

The study aims to develop a system with multiple color LED

light sources achieving higher depth resolution without

sacrificing light intensity. A differential technique is applied to

calculate the height of the sample. Measurement range of the

sample is 7 mm and the accuracy of the measurement is up to

1.144 um.

==SYSTEM DESCRIPTION==

Fig. 1 shows the optical system design of the developed system. The system consists of light source module and color imaging device for capturing images.

Lateral shift in the image plane is determined by calculating the local peak of the line profile. Therefore the resolution of the peak detection is determined by the width of the line profile; the narrower and sharper the line profile the higher the resolution of peak detection is. In the developed system, a narrower and sharper line profile can be obtained by subtraction of the two line profiles with different colors. Proper adjustment, alignment, and calibration of the light sources are needed to ensure that the two line profiles are

overlapping. The receiving angle of the camera also has to be adjusted to the enable capturing image of the wholemeasurement range without obtaining defocused images due to limitation of lens’ Depth of Field. The differential technique applied for calculating the peak of the line profile is shown in Fig. 2. A captured image is first separated into red and blue channel to obtain line profile from red and blue LEDs. Red and blue LEDs are chosen to match spectral response of the employed camera to avoid spectrum overlap. Intensity of the red light is then subtracted with the blue light to obtain the differential profile. Positive value of differential intensity is then normalized to generate two zero crossing points which allow distinctive separation of the region for peak calculation from other region (shown by the blue line on Figure 2). Position of the peak is calculated from the differential profile. By applying this technique the Full Width Half Maximum (FWHM) of the profile is smaller than

the original red line profile.

==CALIBRATION PROCESS==

Eq. 1 shows the relation between image coordinate ( ∆x') and actual height ( ∆z ) in triangulation measurement. Therefore calibration is first performed to generate calibration curve between pixel position and actual sample height. Calibrated gauge block samples are employed as reference. After the calibration curve is obtained, the triangulation sensor can be used for measurement. Figure 3 (a) and (b) show the image captured by camera and the detected peaks after applying differential calculation technique, respectively. Resolution of the image is estimated by using USAF 1951 standard target and it is obtained that each pixel is 12.72 um.

In Figure 3, the distance between peaks represents the height difference between two gauge blocks. The smallest height difference that can be made by the gauge blocks is 10 um.Therefore, three height differences are tested, 10 um for the range of 1.0-1.1 mm, 100 um for the range of 1.1-2 mm, and 1 mm for the range of 2-5 mm. Figure 4 shows the calibration curve obtained from the measurement of gauge blocks. Mean Absolute Error (MAE) of the calibration curve for the whole calibrated range is 10.83 um.

After the system has been calibrated, it is put to repeatability test. Each gauge block with height ranging from 1.1 – 1.9 mm is measured five times to estimate the repeatability of the system. Table 1 shows the static repeatability measurement results. Repeatability of the measurement ranges from 0.3 – 1.2 um.

Static and dynamic repeatability of the sensor is also tested by measuring a flat mirror with accuracy of 1/10λ mounted on a linear stage. The linear stage is equipped with a linear encoder, which is used as position indication. Stage is moved at a total distance of 7 mm and the image is acquired at every 50 um. For each position, the acquisition is repeated 35 times.

Repeatability of the static measurement is 2.06 um (1 standard deviation) and. Fig. 5 shows the measured relative displacement at each sampling point. It is shown that the measurement fluctuates more at the end of the measurement range. The standard deviation is 3.41 um and the range of the data is 14.18 um. Accuracy of the peak detection for the whole measurement range is 1.144 um (0.07 pixels).

==EXPERIMENT RESULTS==

Aspect ratio is one of important parameters of measurement system. It is tested by measuring gap built by two gauge blocks. Figure 6 shows the sample used in the measurement. Thicknesses of the gauge blocks are 5 mm and the gap between two gauge blocks is 1 mm. Aspect ratio of the sample is 5:1.

Figure 7(a) shows the image captured by the camera and Figure 7(b) shows the peak detection result.

Figure 8 shows the measured profile of the sample. The red region shows the profile for calculating the depth. The measured depth of the gap is 4.980 mm. This result shows the capability of measuring sample with aspect ratio of 5:1.

To validate the performance of the system in measuring real workpiece, this study uses the developed system is to measure the depths of the piston oil ring grooves as shown in Fig. 9. There are several ring grooves to be measured for the inspection. Measuring piston oil ring grooves by a side-view camera had been found to yield unstable accuracy due to alignment errors and image distortion. Instead, pistons then were measured by using the coordinate measuring machine before the new triangular sensor was introduced.

In the test, we focus on one piston groove at a time to get an image. Figure 10 (a) shows the portion of the piston to be measured and Figure 10 (b) shows the captured image in section 1-3.Measured profile is shown in Figure 11. The portion for calculating depth is marked by red dots.

The measurement is performed 10 times and the results are compared to measurement with Coordinate Measuring Machine (CMM) as reference instrument for measurement mechanical components as shown in Figure 9. Table 2 summarizes the comparison of the measurement results with the reference instrument.

The developed system is mounted on robot arm as its motion platform. Table 2 shows the measurement when position of the developed sensor is not changed during repetitive measurement. Another measurement is performed by moving the robot arm into other position and return to the measurement position to test the repeatability of the positioning combined with the repeatability of the measurement system. In the same condition and same position, the measurement is performed 10 times and the result is the repeatability of measurement is 0.0077 and 0.0178 mm for height measurement of section 1-2 and section 2-3 respectively. Parts of measurement difference are contributed by the positioning error of the robot arm. Ambient light also contributes to measurement error by increasing the background intensity because the light source is in visible spectrum. However, due to small receiving angle of the camera and implementation of normalization technique, the effect of background intensity can be neglected. After comparing measurements results performed with and without ambient lighting, the average difference between two measurements is 0.2 um. Therefore, it’s proven that the effect of ambient lighting to measurement errors is negligible.

After the system is calibrated, it is used to measure a sample that is hold by other robot arm. Figure 12 shows the setup of the developed system mounted on the robotic arm (left hand side) and another robotic arm holding the piston to be measured (right hand side). Using this approach, the system can be applied for measuring free form shapes and surfaces. The measurement head can be interchangeable as one would like to use other heads to measure different parameters.

==CONCLUSIONS==

A white light triangulation sensor employing multiple light sources has been developed. The system is capable of generating narrow line profile without sacrificing light intensity. The accuracy of the measurement is up to 1.144 um for the whole measurement range. Repeatability of the measurement can be between 0.2~1.2 um for measured height range of 1.1-1.9 mm. The system can measure sample with up to 5:1 aspect ratio.Repeatability of the system for measuring the piston groove is up to 6.9 um in static measurement and up to 8.2 um in dynamic measurement. The repeatability of dynamic measurement is affected by positioning accuracy of the robotic arm

==REFERENCES==

[1] Costa, M.F.M. Surface inspection by an optical triangulation method. Opt. Eng. 1996, 35, 2743–2747.

[2] Zeng, L.J.; Yuan, F.; Song, D.Q.; Zhang, R. A two-beam laser triangulation for measuring the position of a moving object. Opt. Lasers Eng. 1999, 31, 445–453.

[3] R. Dorsch, G. Häusler, and J. Herrmann, "Laser triangulation: fundamental uncertainty in distance measurement," Appl. Opt. 1994, 33, 1306-1314