CN100501386C - Digital image detection and analysis system for micro mechanical structure - Google Patents

Abstract

The invention relates to a micromachine construction digital picture detection and analysis system, which belongs to optical electromechanical integration field.The characteristics of this system are as follows: light source supplies brightness needed by the detected object; CCD lens ingests visual signal of the detected object picture information through microscope 2; picture gathering card transforming digital signal D/A; then picture digital signal entering detection and analysis system. The said detection and analysis system comprises three-dimensional physical dimension measurement modular, surface defects measurement modular and surface roughness, skewness degree and flattening measurement modular.

Description

The micro mechanical structure digital picture detects and analytic system

Technical field

The present invention relates to a kind of micro mechanical structure digital picture and detect and analytic system, is to utilize the ray machine electricity synthetic image of digital image processing techniques to detect and analytic system, belongs to the light mechanical and electrical integration field.

Background technology:

At present, known measurement and the analytic system of utilizing digital image processing techniques, the SimplePCI of U.S. Compix company for example, the Image-Pro of U.S. MediaCybernetics company etc. have functions such as Image Acquisition, Flame Image Process, image measurement analysis, image data base.They mainly are widely used in fields such as biology, medical science, scientific research by finishing work such as particle counting and quantification, area morphological analysis; Simultaneously, they also can carry out the physical dimension measurement, thereby finish the measurement and the analysis of some parameter of micro mechanical structure.But, these existing systems are except costing an arm and a leg, be not suitable for outside the China's national situation, they do not make comprehensive measurement and analysis by digital image processing techniques at the micro mechanical structure three-dimensional surface, thereby can not fully finish the measurement and the analysis task on micro mechanical structure surface.

Summary of the invention

The technical matters that solves

Measure the shortcoming that to finish comprehensive measurement of micro mechanical structure three-dimensional surface and analysis task with analytic system in order to overcome conventional images, the present invention sets up a kind of image measurement and analytic system from measuring hardware and process software two aspects, by obtaining the three-dimensional geometry size, residual defect area in surface and residual defect ratio, surface topography parameters such as surfaceness, measure of skewness and flatness are finished micro mechanical structure three-dimensional surface measurement and analytic function comprehensively.

Technical scheme

Technical characterictic of the present invention is: light source 4 provides object to be detected needed luminance brightness, CCD camera lens 3 is by the vision signal of microscope 2 picked-up objects to be detected 1 image information, after image pick-up card 5 carried out picture signal D/A conversion, the digital signal of image entered and detects and analytic system 6; Described detection and analytic system 6 comprise: three-dimensional geometry dimensional measurement module 8, surperficial residual defect measurement module 9, surfaceness, measure of skewness and flatness measurement module 10.

Described three-dimensional geometry dimensional measurement 8 is to utilize edge detecting technology, the theoretical profile recognition technology of tracking based on wavelet analysis to realize, is divided into planar dimension and depth dimensions and measures.

Described planar dimension measurement is: whole measuring process is promptly set up the geometrical correspondence between the corresponding point on object plane and the image planes according to the lens imaging principle according to how much image-forming principles.For under identical conditions, recording image, form with pixel shows on computers, because unit planar dimension (length or area) is fixed deserved number of pixels, be that Kd=L (allocating the face size)/Pn (pixel) is a constant, the measurement of the calibration element that passes through, can obtain Kd, can obtain pixel count Pn when measuring, tested length is exactly L=Pn * Kd.Because actual camera lens and theoretic perfect lens have than big difference, its object-image relation neither with the perfect lens imaging formula describe so simple, so the method that has adopted test to demarcate in this measuring system is promptly by establishing the dimension scale relation between image to the experiment with measuring of standard component.This system has comprised the autonomous Measurement Algorithm of matrix and two kinds of typical geometry of circle.At first obtain the number of pixels Pn of one group of standard component: in the measuring process of described circle, earlier to carried out the Filtering Processing image 12 that is enhanced by altimetric image 11, carry out rim detection then and obtain edge image 15, perhaps enhancing image 12 is carried out binary conversion treatment and obtain bianry image 13, retouch operator by square and obtain central coordinate of circle and radius 14, and then obtain round edge image 15, edge image 15 is fitted handle the measurement result 16 that obtains circle.In the measuring process of described rectangle, earlier to carried out the Filtering Processing image 12 that is enhanced by altimetric image 11, carry out rim detection then and obtain edge image 15, carry out scan process and obtain refinement edge 17, obtain the edge matrix 18 of rectangle by edge analysis, carry out obtaining vector graphics 19 after figure fits, carry out Filtering Processing again and obtain measurement result 16.Divided by measurement result Pn, data are fitted the COEFFICIENT K d that obtains curve L=Kd * Pn with the physical size L of standard component, the number of pixels Pn ' that records during with this COEFFICIENT K d and actual measurement multiplies each other and obtains actual planar dimension L ', i.e. L '=Pn ' * Kd.

Described depth dimensions measurement is: the present invention is used for the depth survey field to the irradiance theory, set up the mathematical relation of surface element illuminance and the variation of respective regions gray level according to the irradiance theory, and further release the linear relationship of case depth variation and shade of gray.For by altimetric image, different aspects on its vertical direction show as the difference of gray scale on computers, for the image that records under identical conditions, because the gray scale difference of certain altitude correspondence is fixed, promptly Kg=Δ H (fixed length)/Δ Gn (gray scale difference) is a constant.At first, obtain the gray scale difference Δ Gn of known altitude difference H correspondence, fit the COEFFICIENT K g among data acquisition curve Δ H=Δ Gn * Kg by the measurement of one group of calibration element.When measuring, obtain the gray scale difference Δ Gn ' between measured surface and the reference field C, so tested height, be depth dimensions H '=Δ Gn ' * Kg+C.

The residual defect in described surface measures 9: on computers, certain zone of image shows as certain number of pixels, so the quantity of pixel and image area (measured piece area) are one to one, promptly there are much areas how many pixels are just arranged, the work that will do during measurement is exactly the quantity of measuring tested area pixel, scale-up factor Kp=Pn/Sa between measured piece pixel quantity Pn and its corresponding region area Sa is a constant, so Kp=Pnz (the total pixel of measured piece)/Sz (the measured piece total area) is arranged, Kp=Pnc (residual defect pixel)/Sc (residual defect area).In this measurement module, also the method for service test demarcation promptly concerns by the area ratio that the experiment with measuring of standard component is established between image, thereby reduces the error of bringing owing to the difference of real lens and theoretic perfect lens.The demarcation detailed process is: the number of pixels Pnz that measures the original image of an area known standard spare ₁, use the pixel number divided by real area Sz ₁Obtain COEFFICIENT K p.The number of pixels Pnz that records when real area Sz equals actual measurement is divided by the COEFFICIENT K p of correspondence,

Because:

Sz＝Pnz/Kp………………………………………………………………(1)

Sc＝Pnc/Kp………………………………………………………………(2)

Residual defect is than Cn=Sc/Sz., and with (1), (2) formula substitution gets:

So residual defect is than Cn=Pnc (residual defect pixel)/Pnz (the total pixel of measuring piece).

In the measuring process of residual defect, the a certain surface image 26 that has residual defect is carried out the Filtering Processing image 12 that is enhanced, carry out binary conversion treatment then and obtain bianry image 13, adopt border following algorithm to obtain residual defect edge 20, adopt seed fill algorithm to obtain the residual defect image 21 of this inside, edge, read out residual defect pixel number Pnc22, read the pixel number Pnz24 on whole surface again, what just obtain with Pnc/Pnz is that the surperficial residual defect of measured piece compares Cn.Residual defect area Sc23 equals Pnc/Kp.

Described surfaceness, measure of skewness and flatness measure 10, and its roughness concentration is earlier sampled point to be carried out on one group of standard component surface to choose, and each standard component all obtains one group of gray-scale value x ₁, x ₂..., x _n, calculate the mean value of each standard component $\stackrel{\‾}{x}=\frac{1}{n}(x_1+x_2+\·\·\·+x_n),$ And the gray scale mean square deviation deviation of each standard component: $\mathrm{\δ}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(\stackrel{\‾}{x}-x_i)}^2}.$ According to the known roughness value of this group standard component and the gray scale mean square deviation deviation of acquisition, by cubic spline interpolation relational expression Ra=A δ ³+ B δ ²+ C δ+D simulates curve, promptly obtains each rank coefficient A of splines, B, C, D.When measuring, obtain the δ ' of measured piece then, the surfaceness that obtains measured piece is Ra '=A δ ' ³+ B δ ' ²+ C δ '+D.According to measure of skewness and flatness computing formula, have the interpolation relation between 3 powers of the mean square deviation of the gray-scale value of these two parameters and image and 4 powers equally, can obtain the measure of skewness S of measured piece _SkWith flatness S _Ku

Beneficial effect

The invention has the beneficial effects as follows, that the performance photoelectric image is measured is contactless, high precision, advantage of wide range of application, digital image processing techniques are used for the micro mechanical structure surface measurement, obtain the accurately measurement and the analysis result of comprehensively sufficient micro mechanical structure three-dimensional surface shape.

Description of drawings

The present invention will be further described below in conjunction with drawings and Examples.

Fig. 1: system framework figure

Fig. 2: system module composition diagram

Fig. 3: the measurement process flow diagram of circle

Fig. 4: rectangle Measurement Algorithm process flow diagram

Fig. 5: residual defect measurement module process flow diagram

Fig. 6: micro-cantilever figure

Fig. 7: plane surveying processing result image synoptic diagram

Fig. 8: the survey sheet of embodiment table 1

Fig. 9: roughness concentration interface and three-dimensionalreconstruction figure thereof

Figure 10: measure of skewness and flatness are measured interface and three-dimensionalreconstruction figure thereof

Figure 11: residual defect is than measuring the interface legend

1-measured piece monitor station, 2-microscope, 3-CCD camera lens, 4-light source, 5-image pick-up card 6-detection and analysis

7-output, 8-three-dimensional geometry dimensional measurement 9-surperficial residual defect is measured

10-surfaceness, measure of skewness and flatness are measured

Embodiment

We have carried out concrete test and application for this total system.What adopt during enforcement is that (20W～50W), microcobjective (Na0.5)/common object lens (f6mm) and Pentium II 300MHz processor utilize these software systems to carry out data processing and demonstration for 753 x, 582 pixel Array CCD Camera, fluorescence light source (35W)/source of parallel light.

1. depth survey and three-dimensionalreconstruction example

Utilize this system that this microstructure of micro-cantilever has been carried out the measurement of depth parameter, and it is carried out three-dimensionalreconstruction.Micro cantilever structure figure as shown in Figure 6.

Demarcate with the measuring condition match condition under, we are to the experimentize least square linear fit of test and data of five groups of samples, and employing light cross-section method (precision 1 μ m) is carried out system calibrating.Table 1. has provided experimental calibration and measurement result.

In digitized process, the capture card sampling precision is a little more than CCD resolution, and quantified precision is 8bits, adopts double precision

Data type; For Δ g, itself there is not any unit when initial, through demarcating, Δ g is between 0 and 255 in theory.Under table 1. condition, measuring maximum absolute error is 2 μ m, and maximum relative error is about 1.1%, and systemic resolution is 1 μ m/ lattice, and measurement result is stable.

Table 1. degree of depth is demarcated and measurement result

2. horizontal edge is handled and plane geometry dimensional measurement example

Under object distance 30mm condition, obtain system proportional coefficient K such as table 2.Be to improve system accuracy, experiment draws system's average proportions COEFFICIENT K through repeatedly demarcating and the gross error of interpolation removal K; Make an uproar after filtration, edge thinning and data fitting, obtain in the micro-cantilever target parts of images result to be measured as shown in Figure 7;

Experiment adopts the image-forming objective lens of focal length 6mm and light cross-section method to demarcate, and utilizes K and edge span, tries to achieve two rectilineal intervals promptly

The semi-girder deck-siding is 2.07mm (theoretical value 2.03mm), and the measurement repeatable accuracy that obtains satisfies actual needs less than 0.01mm.

Table 2. system proportional coefficient is demarcated

3. surface finish measurement example

We obtain interface as shown in Figure 8 when carrying out the measurement of surfaceness example.The roughness value that can obtain quantizing.

4. measure of skewness and flatness practical measuring examples and result

When this system of application carried out measure of skewness and flatness measurement, measurement result and the following Fig. 9 of three-dimensionalreconstruction pattern showed.Measure of skewness that can obtain quantizing and flatness value.

5. surperficial residual defect is than practical measuring examples and result

When this system of application carried out residual defect than measurement, measurement result was illustrated in fig. 10 shown below.

When measuring, getting the gray scale threshold values is 120, residual defect is partly carried out the area tracking algorithm obtain edge image, and this edge image is carried out the fitted figure picture that seed fill algorithm obtains all be presented on the interface.Residual defect area value that can obtain quantizing at last and residual defect ratio.

Claims (4)

1. a micro mechanical structure digital picture detects and analytic system, it is characterized in that: light source (4) provides object to be detected needed luminance brightness, CCD camera lens (3) is by the vision signal of microscope (2) picked-up object to be detected image information, after carrying out picture signal D/A conversion by image pick-up card (5), enter and detect and analytic system (6); Described detection and analytic system (6) comprise three-dimensional geometry dimensional measurement module (8), surperficial residual defect measurement module (9), surfaceness, measure of skewness and flatness measurement module (10).

2. micro mechanical structure digital picture according to claim 1 detects and analytic system, it is characterized in that: described three-dimensional geometry dimensional measurement module (8) is divided into planar dimension and depth dimensions and measures; Described planar dimension measurement is: at first the original image to one group of standard component carries out image filtering, edge detection process, geometric configuration according to measurand is carried out match and measurement to the edge image that obtains, the measurement result of the standard component that obtains is number of pixels Pn, with the physical size L of standard component divided by measurement result Pn, data are fitted the COEFFICIENT K d that obtains curve L=Kd * Pn, the number of pixels Pn ' that records during with this COEFFICIENT K d and actual measurement multiplies each other and obtains actual planar dimension L ', i.e. L '=Pn ' * Kd; Described depth dimensions measurement is: by the measurement of one group of calibration element, obtain the gray scale difference Δ Gn of known altitude difference Δ H correspondence, fit the COEFFICIENT K g among data acquisition curve Δ H=Δ Gn * Kg; When measuring, obtain the gray scale difference Δ Gn ' between measured surface and the reference field C, so tested height, be depth dimensions H '=Δ Gn ' * Kg+C.

3. micro mechanical structure digital picture according to claim 1 detects and analytic system, it is characterized in that: the residual defect measurement module in described surface (9) is: a certain surface image that has residual defect is carried out filtering, after the edge detection process, the a certain gray scale threshold values of setting, adopt border following algorithm to obtain to surpass the part that this sets threshold values, it is the edge of residual defect part, adopt seed fill algorithm to obtain the residual defect pixel number Pnc of this inside, edge, the pixel number Pnz on whole surface reentries, with Pnc/Pnz just obtain be the surperficial residual defect of measured piece than Cn, residual defect area Sc=Kd * Pnc.

4. micro mechanical structure digital picture according to claim 1 detects and analytic system, it is characterized in that: its roughness concentration of described surface finish measurement module (10) is earlier sampled point to be carried out on one group of standard component surface to choose, and each standard component all obtains one group of gray-scale value x ₁, x ₂..., x _n, calculate the mean value of each standard component $\stackrel{\‾}{x}=\frac{1}{n}(x_1+x_2+\·\·\·+x_n),$ And the gray scale mean square deviation deviation of each standard component: $\mathrm{\δ}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(\stackrel{\‾}{x}-x_i)}^2};$ According to the known roughness value of this group standard component and the gray scale mean square deviation deviation of acquisition, by cubic spline interpolation relational expression Ra=A δ ³+ B δ ²+ C δ+D simulates curve, promptly obtains each rank coefficient A of splines, B, C, D; When measuring, obtain the δ ' of measured piece then, the surfaceness that obtains measured piece is Ra '=A δ ' ³+ B δ ' ²+ C δ '+D.

Images