CN1260116C - Computer vision based test apparatus and method for micro electro-mechanical systems - Google Patents

Abstract

The invention discloses a computer vision-based micro-electro-mechanical system testing device and method. The device mainly includes an optical microscope, a strobe lighting device, a MEMS structure motion excitation device, and a CCD camera. High voltage op amp with multi-stage gain adjustment. The test method includes synchronous control of stroboscopic and drive signals, normalized evaluation of geometric parameters of plane images, motion estimation of fuzzy images, and motion estimation of image matching. The invention has the advantages of: adopting virtual instrument mode to realize function adjustment and expansion; under continuous light illumination, realize normalized evaluation of plane geometric quantity parameters, supplemented by analysis of fuzzy motion images, realize rapid measurement of motion characteristics, and The frequency is unlimited; under the strobe lighting, the image matching technology is used to achieve high-precision motion characteristic testing.

Description

Computer Vision-Based MEMS Testing Device and Method

Technical field

The invention relates to a device and a method for measuring geometric quantity parameters and motion characteristics of a planar structure in a micro-electromechanical system (MEMS) based on computer micro-vision technology. It belongs to the geometric quantity and mechanical quantity testing technology of the photoelectric non-contact method for micro-electromechanical systems.

Background technique

Testing technology has important practical significance in the research and development process and industrialization process of microelectromechanical systems (MEMS). At present, MEMS testing methods mainly rely on traditional IC testing tools and expensive microscopic testing equipment such as scanning electron microscopes and atomic force microscopes. These devices are expensive, have slow test speeds, small measurement ranges, and are demanding on the test environment. In recent years, as MEMS has gradually entered the industrialization stage from the research stage, the demand for test systems is more urgent. Because the measurement of geometric dimensions and motion characteristics can indirectly reflect the basic performance of MEMS devices, such as: three-dimensional micro-motion of MEMS microstructures, material properties and mechanical parameters, MEMS reliability and device failure modes, failure mechanisms, etc., Therefore, the importance of MEMS dynamic testing technology is more prominent.

The test of the MEMS structure is divided into the following parts: one is the measurement of the plane size of the structure; the other is the measurement of the plane motion of the structure; the third is the measurement of the longitudinal size of the structure; the fourth is the measurement of the longitudinal motion of the structure. Because the dynamic characteristics of MEMS structure determine the basic performance of MEMS, dynamic testing technology is of great significance in the process of MEMS research and development.

In recent years, many beneficial explorations have been made on the testing methods of the dynamic characteristics of MEMS devices at the micro scale, and some practical research results have been obtained, such as the application of stroboscopic imaging technology to collect the motion of MEMS moving parts with periodic high-speed motion Image, use digital image processing technology to analyze its dynamic characteristics; use phase shift interference technology in laser precision measurement and sub-pixel analysis technology in computer vision to improve measurement accuracy; use laser Doppler velocimetry technology to realize the transient state of MEMS devices Real-time measurement of motion, etc.

The measurement technology based on computer micro vision has been widely used in the detection of MEMS structural motion parameters, and foreign research institutions have reported some comprehensive application examples on the use of computer micro vision for MEMS dynamic testing. The main technical characteristics are: A momentary "freeze" of periodic motion is achieved using strobe lighting.

Through comprehensive analysis and comparison of technologies in this research direction, there are mainly problems in the following aspects: First, the test method is single, and it is difficult to meet the requirements of the diverse planar motion of MEMS structures. Second, the measuring device is not systematic and has poor expansion ability; third, the computer micro-vision algorithms used basically directly adopt static computer vision processing methods, and there are certain deviations when dealing with the fuzziness of dynamic images; Fourth, there is a lack of normalized systematic evaluation of the shape errors of MEMS structures, resulting in poor comparability of measured data.

Contents of Invention

The purpose of the present invention is to provide a device and method for measuring plane geometric parameters and dynamic characteristics of micro-electromechanical systems (MEMS) based on computer micro-vision technology, which has various measurement methods, strong expansion capabilities, wide measurement frequency range, and high measurement accuracy. features.

The present invention is achieved through the following technical solutions. Based on computer micro-vision technology, a device for testing the plane geometric parameters and dynamic characteristics of micro-electromechanical systems (MEMS), the device includes a three-dimensional micro-motion test bench, an optical microscope, a strobe lighting source, a MEMS structure motion excitation device, a CCD camera, and an image Acquisition card, GPIB control card, arbitrary waveform generator, measurement and control computer, as shown in Figure 2, is characterized in that: the light source adopts high-brightness light-emitting diode LED, supplemented by high-bandwidth constant current drive circuit and front Arbitrary waveform generator; MEMS motion drive uses a single-chip high-voltage operational amplifier as the core voltage drive circuit, with multi-stage gain adjustment, supplemented by a pre-arbitrary waveform generator.

The above-mentioned light source is a high-brightness light-emitting diode LuxeoTM Star, and the main chip of the driving circuit is EL6249C, which can generate a strobe signal of 50ns, and can realize the measurement of a motion frequency up to 1MHz.

The above-mentioned motion drive part uses the high-voltage operational amplifier PA85 as the core. The circuit is equipped with 10X, 20X and adjustable 3-level gain adjustment, which can amplify the ±10V voltage signal output by the pre-arbitrary waveform generator to ±10V～±200V, as The driving voltage of the MEMS device to be tested meets the driving voltage requirements of different MEMS devices.

Using the above-mentioned device, based on computer micro-vision technology, the measurement method of the plane geometric parameters and dynamic characteristics of the micro-electromechanical system (MEMS), the process includes the synchronous control of stroboscopic and driving signals, the acquisition of MEMS structure images, and the normalization of plane image features. Unified geometric parameter evaluation, planar motion estimation using blurred images, and planar motion estimation using image matching.

It is characterized in that, when measuring the plane geometric parameters of the stationary MEMS structure under continuous lighting conditions:

(1) Fix the MEMS device to be tested on the three-dimensional micro-motion test bench; set the lighting device to continuous lighting mode to illuminate the MEMS device;

(2) Utilize the standard PAL CCD camera to obtain the still image of the MEMS structure under test, and use image feature path tracking technology and sub-pixel analysis technology to perform normalized evaluation and measure the plane geometric parameters;

(3) The criterion for normalization evaluation is: take the gray midpoint of the sudden change area on the image as the contour point, and use the artificially defined measurement line as the optimal search path, and combine the sub-pixel analysis technology of cubic curve fitting to determine the corresponding Sub-pixel precision surface contour line; according to the surface contour line, the center line of the envelope straight line is determined as the measurement reference line for the evaluation of angle, straightness, parallelism and perpendicularity; the measurement of length is the two measurement references in the local area of the measurement line The distance of the line, rather than the distance between the pixels corresponding to two measurement points on the measurement line;

When using fuzzy image technology to measure planar motion parameters of moving MEMS structures under continuous lighting conditions:

(1) Apply a motion excitation signal to the electrode pin end of the three-dimensional micro-motion test bench, so that the MEMS structure under test can perform periodic planar motion at a certain frequency;

(2) Utilize the standard PAL system CCD camera to obtain the planar motion blur image of the measured structure of MEMS, and use image processing techniques such as edge detection and sub-pixels to obtain the size of the fuzzy band in the blur image, and obtain the planar motion range of the structure;

(3) Adjust the frequency of the driving signal with a certain step, so that the MEMS structure moves at different frequencies, and also obtain the fuzzy moving image at the corresponding frequency, that is, the motion amplitude of the structure at a series of driving frequencies, through data analysis Obtain the resonant frequency and quality factor of the structure;

When using image matching technology to measure planar motion parameters of moving MEMS structures under stroboscopic lighting conditions:

(1) Set the lighting device to stroboscopic lighting mode, and perform stroboscopic lighting on MEMS devices. The cycle of the stroboscopic signal is the same as that of the motion excitation signal, and a fixed delay time is maintained, so the movement of the MEMS structure is in the stroboscopic lighting mode. The following belongs to the "frozen" state;

(2) The number of times of stroboscopic flashes of the same phase is set, so that the standard PAL CCD camera is exposed multiple times, and the integral effect of the exposure can be obtained to obtain the moving position of the MEMS measured structure under the above-mentioned fixed phase;

3) By adjusting the phase difference between the stroboscopic signal and the motion excitation signal, the motion position image of the MEMS structure under test in different phases corresponding to the motion stage can be obtained. By using block matching and phase correlation technology comprehensive analysis on the motion position image sequence, a certain The motion status of the MEMS structure at different phases under the frequency;

(4) The combination of phase correlation and quadratic surface fitting can solve the problem of rapid detection of sub-pixel precision displacement under limited motion range, and at the same time, the exhaustive method of rotation angle is used to realize the measurement of rotation angle; block matching can realize large Under the range of motion, the rapid detection of rough motion displacement; under the comprehensive application of the two, the rapid detection of motion displacement and angle is realized;

(5) Adjust the frequency of the driving signal with a certain step, so that the MEMS structure moves at different frequencies, repeat steps (2) and (3), and the detailed characteristics of the structure's motion under a series of driving frequencies can be obtained, through Comprehensive analysis can not only obtain the resonant frequency and quality factor of the structure, but also obtain the whole process of the structure's motion state.

The advantages of the present invention are: adopting the method of virtual instrument to build internal functional modules, which facilitates the function adjustment and expansion of the system; under the continuous light illumination mode, adopts image surface feature tracking technology and sub-pixel analysis to realize the adjustment of plane geometric parameters Normalized evaluation to enhance the comparability of measurement parameters; under continuous light illumination, through the analysis of the fuzzy motion image of the moving structure in MEMS, the motion amplitude of the structure at a certain frequency can be determined, and the resonance of the moving structure can be obtained Parameters such as frequency and quality factor have the advantages of fast measurement speed, basically unlimited measurement frequency, and simple structure. At the same time, the use of wavelet transform-based plane motion edge point detection technology reduces the impact of image blur on motion estimation under stroboscopic lighting. ;In the way of strobe lighting, the standard PAL CCD camera is used to realize the instantaneous image acquisition of the high-speed periodic motion structure in MEMS, and the mutual cooperation of block matching and phase matching image processing technologies is used to realize fast and high-precision planar Motion characteristic detection.

Description of drawings

Figure 1 is a schematic diagram of the basic principle of dynamic characteristic measurement using strobe lighting;

Fig. 2 system block diagram of MEMS testing device based on computer micro-vision technology;

Figure 3 Schematic diagram of the strobe drive circuit;

Figure 4 The schematic diagram of the MEMS motion excitation circuit;

Figure 5. Static image of the MEMS resonator;

Fig. 6 is an example diagram of measuring geometric quantities by using image feature path tracking technology;

Figure 7 is a schematic diagram of the implementation of sub-pixel analysis technology;

Figure 8 Blurred image of MEMS resonator motion;

Fig. 9 realizes the measurement of the resonant frequency of the MEMS resonator by frequency scanning and amplitude measurement based on fuzzy images;

Figure 10 The instantaneous image of the MEMS resonator with a motion phase of 90°;

Figure 11 The instantaneous image of the MEMS resonator with a motion phase of 180°;

Figure 12 The instantaneous image of the MEMS resonator with a motion phase of 270°;

Figure 13 MEMS resonator periodic motion curve;

Figure 14. Image matching calculation flow chart.

Detailed ways

Example 1:

This embodiment mainly focuses on the measurement of the plane geometry parameters of the MEMS structure by using computer micro-vision technology under continuous lighting conditions.

The measurement and control computer controls the arbitrary waveform generator to output a DC voltage drive signal through the GPIB control card, so that the strobe lighting device works in a continuous lighting state. The MEMS structure is placed on the three-dimensional micro-motion test bench under the optical microscope, and the CCD is used to The camera captures the image of the MEMS planar structure, which is then digitized by the image acquisition card and stored in the computer and displayed on the computer screen. Fig. 5 is a static image of the planar structure of the MEMS resonator. By processing and analyzing the image in Fig. 5, the geometric parameters of the planar structure can be determined. In order to measure the geometric parameters of a specific planar structure, it is necessary to manually select a specific area: (1) Length measurement: as shown in Figure 6, the distance between the two edges of the structure is measured, first at the two Pull out a schematic line a and b at each edge, and then pull out a measurement line c between the two schematic lines. The length measurement does not directly obtain the pixels of the measurement line c between the schematic lines a and b, but based on The position of schematic lines a and b is located using image feature path tracking technology and sub-pixel analysis technology to extract the edge line of the structure, and use the least square method to fit the edge center line near a and b, and then calculate the center line The minimum distance near the measurement line; (2) Measurement of parallelism: measure the parallelism of the two edges of the structure as shown in Figure 6, mark 2, and first pull out a schematic line d and e, according to the positions of schematic lines d and e, use image feature path tracking technology and sub-pixel analysis technology to extract the edge line of the structure, and use the least square method to fit the edge centerline near d and e, and then calculate Parallelism of the center line; (3) Measurement of perpendicularity: measure the perpendicularity of the two edges of the structure as shown in Figure 6, mark 3, first pull out a schematic line f and g on each of the two edges, According to the positions of the schematic lines f and g, the image feature path tracking technology and sub-pixel analysis technology are used to extract the edge line of the structure, and the edge center line near f and g is fitted by the least square method, and then the center line is calculated verticality. The measurement of other geometric quantity parameters (such as: angle, straightness, etc.) can be implemented in a similar manner.

In the above steps, if the general image enhancement, binarization and edge extraction methods are directly used to obtain the contour line, then the resolution of the measurement can only reach the pixel level. In order to obtain the sub-pixel measurement resolution, the above steps comprehensively utilize the image feature path tracking technology and the sub-pixel analysis technology to extract the edge line of the structure. The specific implementation steps can be expressed as follows: (1) Since the geometric quantity measurement parameters are proposed based on the contour of the surface structure, and the contour of the surface structure will show certain fluctuations due to the influence of the processing technology, the general image enhancement, two-dimensional Value-based and edge extraction methods to obtain pixel-level contours, the contours are not used as the baseline for measurement, but as the preferred search path for the following steps; (2) In order to achieve sub-pixel accuracy, it is necessary to artificially A measurement line is defined, and the measurement line is required to be basically consistent with the outline of the measured structure; (3) Since there is generally an obvious gray scale change on both sides of the outline, the middle gray can be searched in the vertical direction of the measurement line. If the middle gray value cannot match a single pixel, then perform cubic curve fitting processing on the pixels on both sides of the middle gray value, that is, perform sub-pixel analysis, as shown in Figure 7. The determination of the structure contour points can reach the accuracy of the sub-pixel; (4) each point on the measurement line is analyzed in step (3) in turn, and a series of measured structure contour points can be determined, and the measurement accuracy reaches the sub-pixel (5) When determining each contour point, it is necessary to calculate the distance from the optimal search path obtained in step (1), and select the one with the shortest distance as the real structural contour point; (6) connect the above contour points to get Structural contour lines with sub-pixel precision, and then use the least squares fitting method to obtain the envelope straight line of the contour line, and finally take the center line of the two envelope straight lines as the reference line for measurement.

Example 2:

This embodiment mainly discusses the measurement of the plane motion parameters of the MEMS structure by using fuzzy image processing technology under continuous lighting conditions.

Under continuous lighting conditions, the MEMS structure-resonator is driven by a sine wave with a certain frequency, generally above 10kHz, to produce periodic planar reciprocating motion. The exposure time of the CCD camera is tens of milliseconds. Due to the integral effect of exposure, The image in the area where the resonator moves back and forth shows a fuzzy band, so it can be considered that this fuzzy band reflects the motion amplitude of the resonator at this driving frequency. Figure 8 is the fuzzy image of the MEMS resonator moving under the excitation of a sinusoidal drive signal with a frequency of 20kHz, a bias voltage of 20V, and a peak-to-peak value of 160V. The length of the fuzzy band can be measured by image processing technology. The measurement of the length of the fuzzy band is related to The measurement of the dimensions in Embodiment 1 is the same, so that the motion amplitude of the MEMS resonator at the driving frequency of 20 kHz is obtained. Combining with the sub-pixel analysis technology, the measurement resolution of the motion amplitude can reach the sub-pixel level.

The frequency of the motion-driven sine wave is adjusted by frequency sweep technology, and the motion images of the resonators at a series of driving frequencies are obtained, and their motion amplitudes are calculated by the same method. The amplitude-frequency characteristic curve of the resonator can be obtained by using the curve fitting method by using the measured motion amplitude under this series of driving frequencies, so as to obtain the resonant frequency of the resonator and the corresponding motion amplitude of this frequency. Figure 9 shows the resonant frequency measurements of the MEMS resonator. The frequency scanning is as follows: 20kHz as the starting frequency, 0.2kHz as the frequency scanning step, and 27kHz as the stop frequency; the bias voltage of the sinusoidal drive signal is 20V, and the peak-to-peak voltage is 160V; according to the measured maximum vibration amplitude Value, finally determine the resonant frequency of the MEMS resonator to be 23.4kHz.

Example 3:

This embodiment mainly discusses the measurement of the plane motion parameters of the MEMS structure by using block matching and phase correlation image processing technology under the condition of stroboscopic lighting.

In order to capture the instantaneous state of the periodic motion structure, the system adopts the method of stroboscopic lighting to realize the "freezing" of the motion state, so that the measurement of the high-speed motion state can be realized by using a general CCD camera. Figure 1 shows the basic principle of using strobe lighting to realize the measurement of motion characteristics. The measurement and control computer controls the arbitrary waveform generator through the GPIB control card to generate periodic sinusoidal motion excitation signals and fixed-phase stroboscopic signals (0° and 30° as shown in the figure), and the stroboscopic signals of a certain phase continue to appear multiple times , to ensure that the CCD camera has a sufficiently long effective exposure time; because it is a dark field when the stroboscopic pulse does not appear, the stroboscopic pulse with the same cycle as the motion excitation signal reflects the same state of the periodic motion structure, which is equivalent to the high-speed motion structure The state is "frozen", which is convenient for image acquisition with a common CCD camera, and its effective exposure time is the total width of multiple strobe pulses.

In this embodiment, the MEMS structure is excited by a 21kHz sine wave drive signal (bias voltage is 20V, peak-to-peak voltage is 160V), and reciprocating motion is generated in the Y direction, and one cycle of this sine wave is performed with a phase of 30° There are 12 phases in one cycle. In the initial state of moving image acquisition, the pulse width of the stroboscopic signal is 1.5 μs, and it is in the 0° phase of the sinusoidal driving signal. Before the stroboscopic pulse appears, the CCD camera is triggered to start exposure, and then every A stroboscopic exposure is performed on the 0° phase, and finally the CCD camera ends the exposure, and the collected images are transmitted to the computer. After the above process is completed, adjust the position of the pulse of the stroboscopic signal to 30°, repeat the above stroboscopic and exposure process, and obtain an image with a phase of 30°, and then adjust the phase of the stroboscopic pulse at intervals of 30°, which can be obtained in turn The images of the MEMS structure moving at different phases, Figures 10, 11 and 12 are the MEMS structures at 90°, 180°, and 270° phases respectively, and the position of the moving structure can be distinguished by the image in the area surrounded by the box in Figure 10 change. By performing image matching analysis on 12 images in one cycle, the motion position of the structure at different phases can be obtained, as shown in Figure 13, it can be seen that the motion of the structure basically conforms to the sinusoidal driving waveform.

In the image matching process, in order to achieve both measurement speed and measurement accuracy, a combination of block matching, phase correlation and quadratic surface fitting is used for processing. Block matching is the most commonly used motion estimation algorithm. It has a large search distance, can search in the entire image area, and is less sensitive to rotation; phase correlation is a matching algorithm that is less affected by geometric distortion of the image, and can be obtained in one calculation. The displacement of the two images, but the displacement difference between the two images is required not to exceed half of the width of the selected area, which is highly sensitive to rotation; quadratic surface fitting can obtain sub-pixel image displacement. The process of image matching analysis is as follows: (1) The user selects a certain area on the acquired initial position image, which should be on the moving part and has relatively obvious features; (2) Compare the image of the selected area with the subsequent Perform phase correlation calculation on the images at the corresponding positions in the image. If a higher correlation peak is obtained, it indicates that the displacement of the two images is less than half of the width of the selected area. The position of the correlation peak is the displacement of the two images, so skip directly to step ( 6) Carry out quadratic surface fitting analysis, otherwise it indicates that the displacement of the image is out of range or has a large rotation angle, and step (3) is required and performed; (3) Use block matching to perform rough measurement of large-scale motion displacement, The block matching criterion is the minimum mean absolute difference, and the search strategy is the logarithmic search method. In order to meet the requirements of real-time processing of the measurement image, the search stops when the approximate area is found. At this time, the displacement between this area and the original area can be obtained, and then execute Phase correlation in the next step; (4) Perform phase correlation processing on the image of the selected area and the image of the area searched by block matching in the subsequent image. Since the displacement of the feature structure on the two images meets the requirements of the algorithm, it can generally be obtained Higher correlation peak, that is, the displacement of the two related images, the sum of the displacement and the displacement of the search area obtained in step (3) is the total displacement of the movement, skip to step (6) for quadratic surface fitting Otherwise, it shows that there is a certain rotation angle between the two related images, so that the correlation peak is too small, and step (5) needs to be performed; (5) Use the dichotomy method to traverse and retrieve in a certain angle area, and rotate one of the images by a certain Then, the phase correlation of the two images is carried out to obtain the correlation peak, and the next rotation angle is determined according to the size of the correlation peak, and the successive approximation will obtain the largest correlation peak in a certain angle area. At this time, the corresponding rotation angle and correlation The position of the peak is the displacement and angular offset of the two images; (6) After the above steps, the translation and angular offset of the two images can be obtained, the offset is pixel-level precision, and sub-pixel analysis is required , the idea of using the surface fitting method at this time is: take the best matching point at the pixel level as the center, perform surface fitting according to the similarity measure, and then calculate the precise position of the extreme point through the corresponding mathematical method. The device adopts the correlation coefficient of phase correlation as the similarity measurement feature, selects the quadratic surface as the fitting function, and uses the multivariate least squares regression method to determine the precise position of the extremum point in the calculation. The above process is shown in Figure 14. Through further evaluation, the rotation detection accuracy of the measurement device can reach 0.1 degree, and the translation detection accuracy can reach 1/50 pixel.

The following formula is the basic formula of phase correlation operation.

$\frac{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&xi;</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&xi;</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&xi;</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&xi;</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&xi;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Among them, F ₁ and F ₂ are the results of Fourier transform of two images (images collected at different motion phases) respectively. From formula (1) and the theory of Fourier transform, it can be seen that the phase spectrum contains the position translation information of the two images, and it is a power spectrum whose spectral amplitude is 1 in the whole frequency domain. The inverse Fourier transform of (1) shows that the phase correlation function is a delta pulse function located at the position offset (x0, y0) of the two images, also known as the correlation peak. When the two images are completely similar, its value is 1, otherwise it is 0. Therefore, in the present invention, the phase correlation calculation results of the two images are used to determine the displacement offset of the images, so as to determine the motion status.

The invention adopts quadratic surface fitting to analyze sub-pixels. The idea of the surface fitting method is: take the best matching point at the pixel level as the center, perform surface fitting according to the similarity measure, and then calculate the precise position of the extreme point through the corresponding mathematical method. The device adopts the correlation coefficient of phase correlation as the similarity measurement feature, selects the quadratic surface as the fitting function, and uses the multivariate least squares regression method to determine the precise position of the extremum point in the calculation.

The formula used for the quadratic surface fitting function is:

PC(x, y) = ax ² +by ² +cxy+dx+ey+f

Wherein, PC(x, y) is the phase correlation value corresponding to the position (x, y). The above function can be written as follows:

AX=B

In the formula,

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccccc}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="C20031010713200141.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}& {\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="C20031010713200141.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ {{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>\\ <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>\\ <span\; class="notranslate">\; \&CenterDot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>& <span\; class="notranslate">\; \&Center\; Dot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>& <span\; class="notranslate">\; \&CenterDot;</span>\\ {{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; a</span>}\\ \mathrm{<span\; class="notranslate">\; b</span>}\\ \mathrm{<span\; class="notranslate">\; c</span>}\\ \mathrm{<span\; class="notranslate">\; d</span>}\\ \mathrm{<span\; class="notranslate">\; e</span>}\\ \mathrm{<span\; class="notranslate">\; f</span>}\end{array}\right]\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="0"\; label="PC"\; filenames="C20031010713200141.png"\; state="\{\{state\}\}">PC</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\\ {\mathrm{<span\; class="notranslate">\; PC</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ <span\; class="notranslate">\; \&CenterDot;</span>\\ <span\; class="notranslate">\; \&CenterDot;</span>\\ <span\; class="notranslate">\; \&CenterDot;</span>\\ {\mathrm{<span\; class="notranslate">\; PC</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}\end{array}\right]$

The present invention adopts multi-variable least square regression method in the fitting calculation, so that the calculation is simple and accurate. In the calculation process, the vector X is used as the regression coefficient, and it is assumed that the value of the random variable B depends on the independent variables in the matrix A, and the regression coefficient is the coefficient of the fitting function. After obtaining the coefficients of the fitting function, the precise position of the image offset with sub-pixel precision can be obtained by using the following formula.

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; db</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; ce</span>}}{{\mathrm{<span\; class="notranslate">\; c</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; ab</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; ae</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; dc</span>}}{{\mathrm{<span\; class="notranslate">\; c</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; ab</span>}}$

Through the above process, the motion curve of the MEMS structure at a certain driving frequency can be obtained. If the phase interval is reduced, the resulting motion curve will be more accurate, providing more reference information for the design of the MEMS structure; if the same driving frequency scan as in Embodiment 2 is used, a series of corresponding driving frequencies will be obtained. For the motion curve, by calculating the maximum amplitude, the amplitude-frequency curve shown in Figure 9 can also be drawn to determine the resonant frequency of the moving structure.

Claims (4)

1. A test device for micro-electro-mechanical systems based on computer vision, which includes a three-dimensional micro-motion test bench, an optical microscope, a strobe lighting source, a MEMS structure motion excitation device, a CCD camera, an image acquisition card, a GPIB control card, any Waveform generator, measurement and control computer, characterized in that: the light source adopts high-brightness light-emitting diode LED, supplemented by high-bandwidth constant current drive circuit and pre-arbitrary waveform generator; MEMS motion drive adopts single-chip high-voltage operation The voltage drive circuit with amplifier as the core has multi-stage gain adjustment, supplemented by a pre-arbitrary waveform generator. 2. The computer vision-based MEMS testing device according to claim 1, characterized in that: the light source is a high-brightness light-emitting diode LuxeoTM Star, and the main chip of the driving circuit is EL6249C, which can generate a strobe signal of 50 ns. Realize the measurement of motion frequency up to 1MHz. 3. The test device based on computer vision MEMS according to claim 1, characterized in that: the motion drive part uses the high-voltage operational amplifier PA85 as the core, and the circuit is provided with 10X, 20X and adjustable 3-speed gain adjustment, which can be The ±10V voltage signal output by the pre-arbitrary waveform generator is amplified to ±10V~±200V, which is used as the driving voltage of the MEMS device to be tested to meet the driving voltage requirements of different MEMS devices. 4, adopt the method for testing by the test device based on computer vision microelectromechanical system according to claim 1, its process comprises the synchronous control of stroboscopic and drive signal, the acquisition of MEMS structure image, the normalization of plane image feature Geometric parameter evaluation, planar motion estimation using blurred images, and planar motion estimation using image matching, characterized in that: When measuring planar geometric parameters of stationary MEMS structures under continuous illumination conditions: (1) Fix the MEMS device to be tested on the three-dimensional micro-motion test bench; set the lighting device to continuous lighting mode to illuminate the MEMS device; (2) Utilize the standard PAL CCD camera to obtain the static image of the MEMS structure under test, and use the image feature path tracking technology and sub-pixel analysis technology to perform normalized evaluation and measure the plane geometric parameters; (3) The criterion for normalization evaluation is: take the gray midpoint of the sudden change area on the image as the contour point, and use the artificially defined measurement line as the optimal search path, and combine the sub-pixel analysis technology of cubic curve fitting to determine the corresponding Sub-pixel precision surface contour line; according to the surface contour line, the center line of the envelope straight line is determined as the measurement reference line for the evaluation of angle, straightness, parallelism and perpendicularity; the measurement of length is the two measurement references in the local area of the measurement line The distance of the line, rather than the distance between the pixels corresponding to two measurement points on the measurement line; When using fuzzy image technology to measure planar motion parameters of moving MEMS structures under continuous lighting conditions: (1) Apply a motion excitation signal to the electrode pin end of the three-dimensional micro-motion test bench, so that the MEMS structure under test can perform periodic planar motion at a certain frequency; (2) Utilize the standard PAL system CCD camera to obtain the planar motion blur image of the measured structure of MEMS, and use image processing techniques such as edge detection and sub-pixels to obtain the size of the fuzzy band in the blur image, and obtain the planar motion range of the structure; (3) Adjust the frequency of the driving signal with a certain step, so that the MEMS structure moves at different frequencies, and also obtain the fuzzy moving image at the corresponding frequency, that is, the motion amplitude of the structure at a series of driving frequencies, through data analysis Obtain the resonant frequency and quality factor of the structure; When using image matching technology to measure planar motion parameters of moving MEMS structures under stroboscopic lighting conditions: (1) Set the lighting device to stroboscopic lighting mode, and perform stroboscopic lighting on MEMS devices. The cycle of the stroboscopic signal is the same as that of the motion excitation signal, and a fixed delay time is maintained, so the movement of the MEMS structure is in the stroboscopic lighting mode. The following belongs to the "frozen" state; (2) The number of times of stroboscopic flashes of the same phase is set, so that the standard PAL CCD camera is exposed multiple times, and the integral effect of the exposure can be obtained to obtain the moving position of the MEMS measured structure under the above-mentioned fixed phase; (3) By adjusting the phase difference between the stroboscopic signal and the motion excitation signal, the motion position image of the MEMS structure under test in different phases corresponding to the motion stage can be obtained, which can be obtained by comprehensive analysis of the motion position image sequence using block matching and phase correlation techniques The motion status of MEMS structure at different phases at a certain frequency; (4) The combination of phase correlation and quadratic surface fitting can solve the problem of rapid detection of sub-pixel precision displacement under limited motion range, and at the same time use the exhaustive method of rotation angle to realize the measurement of rotation angle; block matching can realize large Under the range of motion, the rapid detection of rough motion displacement; under the comprehensive application of the two, the rapid detection of motion displacement and angle is realized; (5) Adjust the frequency of the driving signal with a certain step, so that the MEMS structure moves at different frequencies, repeat steps (2) and (3), and the detailed characteristics of the structure's motion under a series of driving frequencies can be obtained, through Comprehensive analysis can not only obtain the resonant frequency and quality factor of the structure, but also obtain the whole process of the structure's motion state.

Images