Based on the auto-focus vision of MEMS motion measurement technology (1)

Abstract: in order to seal electrical and mechanical systems (MEMS) dynamic performance of the mechanical test-station auto-focusing, machine display logo visual and strobe illumination, and other technology, designed based on the auto-focus vision of MEMS dynamic test systems, MEMS Planar motion and longitudinal movement of measurement.

The article describes the system components and key technologies and systems made a confirmatory test. Experimental results show that plane sub-pixel displacement algorithm matching accuracy up to l/50 pixels in system 25 x magnification, plane rigid body motion measurement resolution 7.2 nm8.3 nm; auto-focusing process quickly, focal plane precise positioning, longitudinal movement measurement resolution reached 0 £ ® 1 μ m.

Keywords: micro electro mechanical systems (MEMS); vision; auto-focus; stroboscopic Imaging; subpixel algorithms

With the micro-motor system (MEMS) technology for the rapid development of MEMS device properties, particularly for its mechanical and dynamic characteristics of test, is becoming an increasingly compelling topics.

MEMS device itself small size and high-frequency characteristics, determining the traditional piezoelectric, strain measurement methods such as contactless measurement cannot be competent.

And scanning electron microscopy and atomic force microscopy and other expensive micro-testing device cannot achieve the requirements of the dynamic test. In view of the dynamic characteristic testing of MEMS complexity and specificity, the development of new optical-based non-contact measurement instrument has become increasingly important. With stroboscopic vision measurement and stroboscopic interference measurement represents the most advanced MEMS technology dynamic test. University of California Berkeley's sensor and actuator Center developed the stroboscopic microscopy Interferometer system, using stroboscopic imaging and intervention phase shift's technology for nano-precision planar and longitudinal movement measurement. United States MIT also developed based on machine vision and interference measurement test systems, also developed the computer micro vision system, and both are compared, their system is capable of measuring cycles repeat the campaign process, planar motion measurement accuracy better than 5 nm. China, Tianjin University developed the computer vision-based MEMS test system by fuzzy image synthesis technology, and introduce Mirau Interferometer for MEMS devices measurement. Huazhong University of science and Mechanical College Center for micro system has developed three dimensional static and dynamic MEMS test system that integrates a stroboscopic imaging and microscopy Interferometry for MEMS three-dimensional measurement of static and dynamic characteristics.

This article is based on the auto-focus vision of MEMS and dynamic monitoring system, through the acquisition of MEMS devices, micro-vision images using a plane subpixel movement algorithms and focal plane orientation to be logging MEMS plane and longitudinal movement of the test.

This article describes a vision based on automatic focusing of MEMS dynamic test system components and critical measurement technology and data processing algorithms, and system verification experiment conducted an analysis of the data.

1 system design and composition

System by optical microscopy, auto-focus system and three-dimensional precision table, MEMS devices incentive, stroboscopic lighting imaging system and the computer control and integration testing software components, Figure 1 is a system map.

　　

　　

The entire system is placed in the floatation isolation platform, to isolate the outside interference vibration on the test.

In the Planar motion measurement, MEMS devices placed on the 3-d precision table, through the computer control of stepper motor with auto-focus, making the measured MEMS devices near the imaging to the focal plane, at which point it gets a CCD camera MEMS device via microscope Zoom image after a clear vision. Analog output card output cycle signal power amplification excitation MEMS devices that periodically. While the analog output card output frequency of the pulse signal driving frequency Flash source, using stroboscopic imaging camera MEMS devices in high-frequency motions "freezing" of the image. By adjusting the stroboscopic drive signal and MEMS devices the difference between the excitation signal, shooting sports cycle different times in different locations of the image. On the collection of photos like using subpixel movement algorithms calculate the different moments of relative displacement for the measured MEMS devices in motor cycle different moments of plane displacement. Motion measurement from the surface, using three-dimensional precision table z to move towards automatic focusing on MEMS motion cycles each moment of focal plane location for accurate positioning, focal plane where the difference between the different moments of MEMS devices away from the plane moving relative displacement. Through the Planar motion and longitudinal movement of description, eventually gets measured MEMS three-dimensional dynamic parameters of the machinery.

MEMS dynamic test system is a typical optical, mechanical and electronic integration systems, to ensure the system's flexibility and expandability to meet the requirements of automated tests, generally using virtual instruments and structures, using LABVIEW to schema software platform.

The main features of the software platform for: set system parameters (excitation frequency, output phase shift, the synchronization parameters, etc.), implementation of auto-focusing, acquisition MEMS device images, image preprocessing (noise reduction, correction), image analysis and processing three-dimensional motion characteristics, acquired error analysis and generate test reports and graphic interface displays.

2 key technology and data processing algorithms

2.1 stroboscopic Imaging

In MEMS dynamic testing process, as a result of MEMS devices movement frequency are relatively high, usually in the 1 kHz to l MHz.

In this system introduces stroboscopic imaging technology for testing purposes.

Stroboscopic imaging technology is in a certain frequency fast flashing light lighting under observation for high speed rotating or moving object, when the frequency of flashing frequency of the Flash source and the measured object is strictly of rotation or movement speed equal to or is it an integer x, see object is relatively static.

This Visual phenomenon, known as interning "strobe" effect. Stroboscopic effect to observing high speed moving objects directly on the health, some not visiblePhenomenon of "visible".

Figure 2 shows the stroboscopic Imaging principle in the application of this system, that is, to a certain frequency signal incentive MEMS devices, and use the same frequency small dutyfactor frequency pulse driven, Flash source in the CCD camera's exposure time, the device is lighting time is very short, leaving most of the time devices have no lighting, dark field status, camera photographic surface only in the light of that time produces optical integration.

When the lighting time short enough, you can view captured is MEMS devices during this time was "frozen" image, movement restrictions to a small extent, or even that basic no movement, have a high speed movement devices phase on the image. By adjusting the frequency and the Flash source drive signal and MEMS devices the difference between the excitation signal, you can obtain the measured MEMS devices in each movement of the image at the location. According to the principle, stroboscopic imaging needs of MEMS devices for periodic motion or repeatable instantaneous movement; and in order to accurately describe the movement of the device, you need precise control of MEMS stimulus signals and strobe synchronization between the drive signal and the phase shift.

　　

　　

2.2 plane movement algorithms

In order to MEMS device Visual image plane rigid body motion in the estimate, you need to use a digital image processing technology to extract its movements.

In practice, the entire pixel displacement is very easy to obtain. But the actual displacement values generally do not happen to whole pixels, in order to enhance the digital image correlation method of measuring accuracy, in this system, using digital image correlation to pixel-level shift then the correlation coefficient for quadratic surface fitting methods for sub-pixel displacement, have anti-noise ability, calculation of quantity of small, high precision.

Digital image correlation is done on moving image sequences related to movements.

As shown in Figure 3, select the image 1 (MN) size of the template A, in the image 2 templates location (x, y) in the surrounding setting of Windows Mobile, and some related functions and template covered area to calculate the correlation coefficient, look for a template matching, correlation coefficient values for the maximum position. Here choose effects better standardization of covariance, correlation function

Its range is [-1, 1].

Of which: A (m, n), B (m + n + I, j) for image 1 is selected in the gray-scale distribution of template and image 2 are moved by the template to the regional distribution of grayscale, A, B as its average grayscale values. Calculated in image 2, locate the correlation coefficient R the greatest location (x + dx, y + dy), dx, dy for pixel-level movements.

　　

　　

For digital image acquisition of pixel-level offset through correlation coefficient and surface fitting way to strike a sub-pixel displacement.

This article uses the following binary quadratic polynomials to fit the correlation function of the surface. Search for pixel-level shift to location (x + dx, y + dy) and its surrounding neighborhood of eight points with the following binary quadratic function to represent:

Here a total of ao, ...

The a5 6 pending coefficient, and 33 of fitting window 9 equation, so you can use the least-squares method to solve. Function f (x, y) in the fitting surface of extreme points should meet the following equations:

2.3 automatic focusing

On MEMS devices for longitudinal movement tests, you need at each movement of the "freeze" on the use of automatic focusing technique to obtain the exact location of the FPA in order to determine their relative movement.

System based on digital image processing method for automatic focusing on acquisition to do noise reduction of image pre-processing filter to determine a reasonable reliable evaluation function, according to the evaluation function to determine whether the sample is focusing and determines the direction from focus to electromechanical [1] [2]