CN102401638A - Parallel precise measuring system for measuring surface area and perimeter of tiny object - Google Patents

Abstract

The invention discloses a parallel precise measuring system for measuring the surface area and the perimeter of a tiny object. In the invention, geodesic active contour line is applied to the parallel measurement problem of irregular-shaped image multi-objective area/perimeter, a quick and precise measuring method is developed and is applied to a two-dimensional image measuring machine, so as to realize the quick and precise measurement of surface area/perimeter of multiple tiny objects. Besides, a novel centroid self-calibration technology is used, the calibration of pixel size equivalent can be finished accurately without any standard test pieces, finally, the actual surface area and perimeter of the tiny object can be obtained. A guide rail is moved during calibration, and the displacement of the guide rail is precisely measured by a double-frequency laser.

Description

The small items surface area girth precision measurement system that walks abreast

1. technical field

Image measurement technology in the machine vision is the novel measuring technique that forms in field of measurement over nearly 20 years; It is used means or the carrier of image as the detection and the information of transmission, accomplishes multiple measuring task such as geometric sense dimensional measurement, biological products detection, airborne remote sensing measurements.This technology can adapt to the measurement requirement of many new manufacturing process, for example to the measurement of yielding accessory size, microsize, curved surface profile and hole in piece part distance.This patent is developed a kind of parallel measuring system efficiently of area/perimeter of small items to the area/perimeter problems of measurement in the geometric measurement field.

2. background technology

In the image measurement field, it is a great problem that the elemental area of image object (referring to the number of pixels that image object comprises) is measured, and it can be divided into 2 types: the target and any erose object pixel area computational problem that promptly have regular geometric shapes.For the former, generally adopt integral method perhaps to calculate through the detection and the application area formula on angle point/summit from the target center of gravity to point.For the latter, adopt the interior pixels that curve is surrounded to count and find the solution, the following formula of general employing:

$A=\underset{{j=y}_{\min }}{\overset{y_{\max }}{\mathrm{\Σ}}}(x_{j\max }-x_{j\min })---\left(1\right)$

In following formula, x _JmaxFor being the x coordinate maximal value of j corresponding to the y coordinate figure on the object boundary; x _JminFor being the x coordinate minimum value of j corresponding to the y coordinate figure on the object boundary.About the pixel circumferential measurements problem of image object, classical way is divided into two big types, and first kind method is the border reconstruction that application curves contour reconstruction algorithm carries out least error, the length and the summation of calculating each segment of curve then respectively.Second class methods are to use the chain code method to carry out length computation, utilize the in addition level and smooth correction at edge of angle point information then.Sum up the area/perimeter measuring method of above-mentioned image object, can find that they can only measure the single body target,, need carry out repeatedly duplicate measurements so and cause lower efficiency of measurement if there are a plurality of targets in the image.

3. summary of the invention

This patent is applied to geodesic line driving wheel profile in the erose image multiple goal area/perimeter problems of measurement; Develop a kind of quick accurate measurement method; And this method is applied to the two dimensional image measuring machine, has realized the measurement quick and precisely of a plurality of small items surface area/girths.Like this, the measurement function of expansion two dimensional image measuring machine has solved it and can't accurately measure the restriction of small items surface area/girth fast.In addition, used a kind of centre of form self-calibration technology of novelty, it just can accurately accomplish the demarcation of Pixel Dimensions equivalent without any need for standard specimen.

4. description of drawings

Fig. 1 two dimensional image measuring machine overall construction drawing, the 1. vertical reading CCD.3. of optical scale .2. horizontal read number CCD.4. aiming CCD.5. measured object 6. worktable .7. guide rail .8. measured angular cone prism 9. spectroscopes and reference mirror .10. two-frequency lasers.

Fig. 2 image-type automatic focusing sighting system structure diagram 1. light source .2. aperture diaphragm .3. worktable .4. focusing. device 5. aiming CCD 6. measured objects.

850 results of Fig. 3 aircraft multiple goal convergence a initialization driving wheel profile 200 c convergences as a result of b convergence

Fig. 4 sub-pix area calculates

5. embodiment

5.1 two dimensional image measuring machine

The two dimensional image measuring machine is mainly by the image-type displacement measurement system, and image-type automatic focusing sighting system and corresponding physical construction are formed, and basic structure is as shown in Figure 1.Its basic functional principle is: use two ccd video cameras to pick up in length and breadth the positional information to optical scale respectively, and utilize image card to link to each other with computing machine, confirm the coordinate figure of both direction through image processing techniques.Sighting system adopts a ccd video camera to receive the measured piece profile through optical system imaging, also is to link to each other with computing machine through image card.Like this, can extract the out of focus information of workpiece on the one hand, utilize stepper motor and certain gear train driving optical system and video camera to move up and down, realize the automatic focusing of sighting system through image processing techniques; On the other hand, the positional information of relevant workpiece profile has also been sent into computing machine, just can confirm to be aimed at the position of profile again through certain algorithm, and comprehensive displacement measurement result and aiming result can obtain measured size or coordinate.

In the two dimensional image measuring machine, sighting system adopts ccd video camera to replace human eye to aim at, and specific practice is the image planes place in the main microscope system, settles a ccd video camera, makes its light-sensitive surface perpendicular to optical axis.Contour of object just is imaged on the CCD light-sensitive surface like this, utilizes image processing techniques to confirm the position of object edge again, just can reach the purpose to the measured piece aiming.The one-piece construction sketch of image sighting system is as shown in Figure 2.In Fig. 2, the light that light source 1 sends projects on the profile of measured object 6 through collimated light path, and tested profile is imaged onto area array CCD through object lens, through after the light-to-current inversion, through interface circuit signal is sent into computing machine.Image to gathering moves up and down the realization automatic focusing through focusing function and step motor drive CCD5.Then the measured object contour images is handled, realized the profile aiming.

Here the real surface that has used the image sighting system directly to measure a plurality of small items amasss/girth, and this needs 2 steps.At first, use the geodesic line active contour model and try to achieve image multiple goal (contour images of measured object) elemental area and girth; Secondly, use centre of form self-calibration technology again and try to achieve the Pixel Dimensions equivalent, try to achieve the long-pending and girth of real surface of measured object at last.The timing signal moving guide rail, and accurately measure its displacement with two-frequency laser.

5.2 image multiple goal elemental area/girth parallel measurement

In Level Set Method; Geodesic line driving wheel profile is to express with the zero level collection of level set function , and the zero level set representations here is that the evolution equation of level set function

is following:

The F here is called velocity function.In the evolutionary process of geodesic line driving wheel profile, need constantly reinitialize level set function

and make the stable convergence of driving wheel profile.In practical application; Reinitialize

and have some problems, for example when carry out initialization and how

is initialized as the symbolic distance function also do not have perfect answer.Therefore; Li Chunming et al. has proposed a kind of geodesic line active contour model based on variation level set (Variational LevelSet), and it no longer need reinitialize and have evolutionary rate faster.This paper has used this geodesic line driving wheel profile, and has carried out respective extension and be applied to multiple goal area circumferential measurements.As an example, Fig. 3 has demonstrated the situation that the geodesic line active contour model converges to a plurality of aircraft image objects, and its iterations is 850 times.

Consider the multi-target detection ability of geodesic line active contour model, we are applied to the measurement of image multiple goal elemental area and girth, have designed a kind of sub-pix multiple goal area and girth while measuring method of novelty.Concrete performing step is following:

1) for image I to be detected, use variation type geodesic line driving wheel profile and carry out multi-target detection, make the driving wheel profile converge to image object contour edge separately.The background pixel of statistical picture I (non-object pixel); Use following method: if (x, level set function value y)

then this pixel is a background pixel to current processed pixels p.

2) for separately image object; The inner number of pixels of difference statistical objects; Use following method: if current processed pixels p (x; All less than zero, this pixel is the image object interior pixels to the level set function value of the territory of the neighbours up and down pixel of level set function value y)

and this pixel so.This pixel will join in the area of this image object as 1 square measure like this.

3) for separately image object; The number of pixels on difference statistical objects border; Use following method: if current processed pixels p (x; The level set function value that has a pixel in the territory of the neighbours up and down pixel of level set function value y)

and this pixel at least is greater than zero, and this pixel is the image object boundary pixel so.

4) for the boundary pixel p of image object separately ₁(x ₁, y ₁), find background pixel p in the neighbours territory up and down in this pixel ₂(x ₂, y ₂), carry out linear interpolation and calculate the level set function value of sening as an envoy to

Sub-pix coordinate p ₀(x ₀, y ₀), carry out corresponding sub-pix area on this basis and calculate.Illustrate below, establish boundary pixel p ₁(x ₁, y ₁) top pixels be background pixel p ₂(x ₂, y ₁), (here pixel being made as has long and wide rectangular area) as shown in Figure 4.We calculate the level set function value of sening as an envoy to through linear interpolation

Sub-pix coordinate p ₀(x ₀, y ₁).According to this coordinate p ₀(x ₀, y ₁) calculate corresponding sub-pix area (dash area among Fig. 4).If background pixel is positioned at bottom, left part or the right part of boundary pixel, corresponding calculated and top situation are similar, omit here.

5) after the sub-pix area of interior pixels number that calculates image object separately respectively and object boundary, the two addition is just obtained the elemental area of image object separately.

6) suppose that the image object number is N, the boundary coordinate that utilizes the sub-pix method to calculate each image object does

q _i=[x _i, y _i] ^T, i=1 ... N _j, j=1 ... N.The respective target pixel image using the following formula to calculate the circumference (

):

$S_p^j=\underset{i=1}{\overset{N_j}{\mathrm{\Σ}}}\sqrt{{\left[(x_{i+1}-x_i)\right]}^2+{\left[(y_{i+1}-y_i)\right]}^2}---\left(3\right)$

5.3 centre of form self-calibration technology

It is an important step of measuring system of picture that pixel equivalent is demarcated, and a measuring system of picture will provide the physical size of measured workpiece, just must demarcate.The accuracy of demarcating directly has influence on the precision of image measurement.This patent has proposed a kind of centre of form self-calibration technology of novelty, and it has overcome the shortcoming that adopts the standard specimen timing signal to have.At first, it is without any need for standard specimen, and only the centre of form to the image object surface aims at the location, has eliminated the different error of focusing that bring with the thickness of testee of thickness of standard specimen.Secondly, because making the centre of form more be prone to accurate location than common object edge, the geometrical property of the centre of form is not subject to noise.Therefore, the centre of form self-calibration technology of this novelty is specially adapted to lack the Pixel Dimensions equivalent that has the small items image at complicated shape and edge under the situation of standard specimen and demarcates, and provides concrete steps below:

1) to the target image of initial position, initialization geodesic line driving wheel profile, the using iterative algorithm makes it converge to each objective contour edge again.

2) calculate the centre of form coordinate C of each image object of initial position respectively _i=(x _i, y _i), then with the zero clearing of two-frequency laser reading, the i here is the image object index.

3) along the x direction of principal axis guide rail is moved certain distance, each image object profile still is in the visual field.Through two-frequency laser read the actual translation of the centre of form apart from l.

4) to the target image of current location, reinitialize the driving wheel profile, the using iterative algorithm makes it converge to each objective contour edge again.

5) calculate the centre of form coordinate C ' of each image object of current location respectively _i=(x ' _i, y ' _i), calculate distance (is unit with the pixel) n that the centre of form moves along the x direction of principal axis _i=x ' _i-x _iThe x direction of principal axis size equivalent of pixel is

6) repeat the y direction of principal axis size equivalent that above-mentioned steps is calculated pixel.

Measure accuracy, the size equivalent average of available repeatedly measured value calculating pixel for improving.In addition, should carefully adjust the attitude of ccd video camera before demarcating, make the coordinate axis of image coordinate system and coordinate system of machine parallel to each other.

Claims (1)

1. 2 of this patent utilization and Harbin Institute of Technology's photoelectricity test technology and equipment research institute cooperation research and development tie up the image measurement machines; Geodesic line driving wheel profile (Geodesic active contour) model is applied to a plurality of small items surface areas and precision Circumference Measurement; Develop the parallel precision measurement system of a plurality of small items surface areas and girth; This system has fully used the topology independence characteristic of geodesic line driving wheel profile, this model convergence good stability and can while/parallel detection image in each measured target; Simultaneously; Used a kind of centre of form self-calibration technology of novelty, under the situation that does not need standard specimen, just can accomplish pixel equivalent and demarcate, and have higher demarcation degree of accuracy; Therefore; The parallel precision measurement of a plurality of small irregularity body surface areas and girth when this measuring method is specially adapted to lack standard specimen, the geometric sense that can be widely used in miniature industrial products detects the perhaps miniature defective detection by quantitative of various industrial products, and corresponding measuring repeatability error is less than 0.5%.

Images