KR101517554B1 - Control method for color lighting of vision system by conjugate gradient algorithm - Google Patents

Abstract

The present invention relates to a color lighting control method of a vision system using a conjugate gradient algorithm, and more specifically, to a method for automatically setting best optimal lighting conditions in the shortest time by analyzing a camera image in a gray level and applying a differential-based conjugate gradient algorithm to a spot, which has a maximum sharpness. According to an embodiment of the present invention, a color lighting control method of a vision system by using a conjugate gradient algorithm comprises the steps of: acquiring an inspection target image focused on a camera and converting the inspection target image to a gray level image; calculating a sharpness of the gray level image; defining the relationship between sharpness and a cost function by using the relationship between the sharpness and a drive voltage; a conjugate gradient calculation step of calculating a next drive voltage of the lighting by using a differential of the cost function and calculating a convergence coefficient required for the drive voltage calculation and calculating a calculation error; determining boundary conditions for the drive voltage; and repeating aforementioned processes until completion conditions are satisfied.

Description

TECHNICAL FIELD [0001] The present invention relates to a color illumination control method for a vision system using a conjugate gradient algorithm,

The present invention relates to a color illumination control method of a vision system using a conjugate gradient algorithm, and more particularly, to a color illumination control method of a vision system by analyzing a camera image at a gray level to obtain a point having a maximum sharpness by a conjugate gradient algorithm To a method for automatically setting an optimum illumination condition in the shortest time.

Generally, a vision system is provided in an inspection apparatus developed for automatically and quickly performing various visual inspections on the appearance of an object to be inspected, which has relied on human vision, to take a digital image of an object to be inspected, Function to the processing system having the function.

The vision system includes an illuminating device for illuminating a predetermined object with illumination, and a camera for photographing the object to be inspected to generate a digital image. The camera includes a monochromatic camera (monochrome camera, camera) is used.

The illumination device may apply a predetermined single-color illumination. In recent years, however, it has been difficult to apply a controllable color illumination to more effectively read surface defects of an object to be inspected, such as an external appearance inspection device for LCD disclosed in Korean Patent Publication No. 2006-0027225 Vision systems are on the rise.

However, the quality of a monochromatic image photographed by a monochromatic camera and the color illumination condition are ambiguous, so that an operator has to manually find and set an optimal color illumination condition manually.

This method is not only inconvenient but also troublesome because it is necessary to perform the troublesome setting operation again every time the object to be inspected is changed as it is determined depending on the viewpoint of the operator whether or not the found color illumination condition is optimum.

In case of inspecting various kinds of inspection objects due to such inconvenience, a separate vision system is provided for each of the inspection objects, which causes an increase in the cost of the inspection equipment.

Therefore, in the case of a vision system using color illumination, a processing system that makes a judgment on the object to be inspected through the images collected through the vision system can read the image obtained by photographing with a single color camera, The most important problem is the optimal color lighting control method that maximizes the image quality.

In the vision system, if the standard deviation (σ) of the gray level (I) is large, it means that the photographed image shows different contrast differences over the whole, so that it is preferable that the standard deviation value is large for the optimum image acquisition.

On the other hand, since the gray level I varies according to the voltage V of the illumination, the standard deviation? Or sharpness varies according to the voltage V of the illumination, It is required to quickly search for a voltage value at which the maximum value of the voltage value becomes maximum.

Conventionally, as shown in FIG. 1, which shows a method for controlling an equally spaced color light using a single illumination, a standard deviation (?) According to each voltage value that is varied every time while continuously increasing a voltage value by a predetermined magnitude And then a voltage value corresponding to the highest standard deviation (?) Value among them is searched.

However, in such an equal search method, it takes a long time to perform measurement and calculation for every range of the voltage value. In particular, in the case of color illumination, .

For example, to search an optimum voltage value when adjusting the voltage value of the n steps for each one trillion people of m n ^m It is necessary to repeat the adjustment of the voltage value and the calculation.

Therefore, assuming that 0 to 5V is adjusted by 0.05V for each of the three RGB light sources, since the optimum condition is obtained through one million times of voltage adjustment, it takes a great deal of time and becomes more serious.

Japanese Patent Application Laid-Open No. 10-2006-0027225 (Mar. 27, 2006)

SUMMARY OF THE INVENTION The present invention has been conceived in order to solve the above problems, and it is an object of the present invention to provide a method and apparatus for analyzing a camera image at a gray level and finding a maximum sharpness by using a conjugate gradient based on a differential, And a method of controlling color illumination of a vision system using a conjugate gradient algorithm.

According to an aspect of the present invention, there is provided a method of controlling a color vision of a vision system using a conjugate gradient algorithm, the method comprising the steps of: A color lighting control method of a vision system for finding a driving voltage,

(a) acquiring a digital image from a scanned object image incident on a camera and converting the digital image into a gray level image;

(b) calculating a sharpness (?) of the gray level image;

(c) Define the relationship between sharpness (ρ) and cost function (h (V)) as ρ = h (V) = - g (V) using the sharpness and drive voltage relation (σ = g ;

(d) calculating a next driving voltage V ^{(k + 1)} of the illumination using the derivative of the cost function and calculating a convergence coefficient? _k necessary for calculating the driving voltage V ^{(k + 1)} , And a computation error (E);

(e) determining a boundary condition for the driving voltage V ^{(k + 1)} ; And

(f) repeating steps (a) through (e) until the end condition │E│ <ε is satisfied for the end constant ε; .

According to the solution of the above-mentioned problem, the optimal illumination condition can be found in the shortest time by analyzing the camera image at the gray level and using the derivative-based conjugate gradient algorithm to find the maximum sharpness.

As a result, the inspection speed of the product is dramatically improved, thereby reducing the time required for inspection of the product, thereby improving the process yield.

The present invention has the effect of promptly searching for a voltage value for optimal image acquisition more quickly, especially when the number of illumination increases.

1 is a diagram showing a color illumination control method for optimum image acquisition in the prior art.
FIG. 2 is a diagram illustrating an RGB illumination example to illustrate a vision system capable of implementing a color illumination control method of a vision system using a conjugate gradient algorithm according to the present invention.
3 is a flowchart illustrating a color lighting control method of a vision system using a conjugate gradient algorithm according to the present invention.
4 is a conceptual diagram of a search time reduction method according to the conjugate gradient of the present invention.
Figs. 5A to 5C are photographs showing a dark image and a bright image, which are sample images, in comparison with an image optimized by the present invention. Fig.
6 is a graph comparing the distribution of the pixels optimized by the sample and the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

It is to be noted that the same components of the drawings are denoted by the same reference numerals and symbols as possible even if they are shown in different drawings.

In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

Also, when a part is referred to as "including " an element, it does not exclude other elements unless specifically stated otherwise.

FIG. 2 is a diagram illustrating an RGB illumination example for explaining a vision system capable of implementing a color illumination control method of a vision system using a conjugate gradient algorithm according to the present invention. FIG. FIG. 4 is a flowchart illustrating a color lighting control method of a vision system using an algorithm.

Fig. 2 is actually applicable to color illumination with any number of light sources.

The vision system collects a digital image of an object to be inspected, and transmits the digital image to a processing system having a function of determining the quality of the object. The vision system is divided into an illumination apparatus 100, a camera 200, and a control system 300.

The control system 300 is provided with a PC and a frame grabber, and digital image acquisition is performed by a frame grabber and image analysis is performed by a PC.

The illumination apparatus 100 includes a plurality of illumination units 110 for irradiating the object to be inspected 10 with light having different wavelengths so that the camera 200 can obtain an image of the object to be inspected 10, An optical fiber 130, a mixer unit 140, and a connection unit 150.

The vision system to which the present invention is applied is implemented by an LED lighting device in which a plurality of illumination units 110 emit red (R), green (G), and blue (B) The LED lighting device may include two LED lighting devices for irradiating lights of different wavelengths. The LED lighting device may further include an LED for emitting white light, and may include four LED lighting devices such as LED lighting Equipment may be made.

The amplifier 120 is provided between the control system 300 for controlling the input to the illumination unit 110 and the plurality of illumination units 110 to amplify signals of the control system 300, The light emitted from the plurality of illumination units 110 or the light synthesized by the mixer unit 140 is transmitted.

The mixer unit 140 is a chamber in which light emitted from the plurality of illumination units 110 is gathered and synthesized. The connection unit 150 transmits the light synthesized by the mixer unit 140 to the camera 200 The optical fiber 130 connected to the mixer 140 and the camera 200 is connected.

The camera 200 is generally a monochrome camera.

However, it is needless to say that the detailed configuration of the illumination device 100 as described above can be implemented as an example.

The light generated in each illumination unit 110 is transmitted to the mixer unit 140 through the optical fiber 130 and the light mixed in the mixer unit 140 is transmitted through the optical fiber 130 to the object 10 , And the wafer is irradiated.

The reflected light reflected by the inspection object 10 is incident on the camera 200 through coaxial illumination.

The image captured by the camera 200 is converted into a digital image by the frame grabber of the control system 300 and analyzed by the PC.

To this end, the control system 300 is equipped with software for calculating according to a formula described later. The voltage (for example, V _R , V _B , and V _G ) The maximum value of the sharpness is searched in a manner of changing by the conjugate gradient method described later, and the optimum illumination condition is found in a short time.

As described above, the control system 300 acquires a digital image from the image of the inspection object 10 irradiated to the camera 200 (S30) and converts it into a gray-level image (S32).

&Lt; Illumination driving voltage and sharpness >

Assuming that the voltage driving the illumination with n different light sources is V, it can be defined as follows.

_{V = (V 1, V 2} , ..., V n) T

The following relationship is established between the gray level I (x, y) of each pixel and the illumination drive voltage V applied to the illumination unit 110 in the image taken by the camera 200. [

I (x, y) = f (V)

(Hereinafter referred to as sharpness ()) as an index (scale) showing the sharpness of the digital image, which is obtained by dividing each pixel I (x, y) in the image into an evaluation expression for the entire image .

That is, the sharpness (σ) of the gray level in a digital image composed of m × n pixels can be obtained by any one of standard deviation, entropy, absolute difference, and auto correlation (S34).

First, the sharpness (?) Through the standard deviation method can be obtained by the following equation (1).

는 이미지 전체의 그레이 레벨의 평균치이다.Where m is the number of horizontal axis pixels of the image, n is the number of pixels in the vertical axis of the image, I (x, y) is the gray level value in the pixel corresponding to the x,

Is the average gray level of the entire image.

The sharpness (?) Through the entropy method can be obtained by the following equation (2).

Where p _i is the normalized value on the image histogram, h (i) is the number of pixels, and HW is the height and width of the image.

The sharpness (?) Through the absolute difference method can be obtained by any one of the following equations (3) and (4).

The sharpness (?) Through the automatic correction method can be obtained by the following equation (5).

As described above, since the sharpness (?) Is related to the driving voltage (V) of the illumination, it can be described by the following relationship.

? = g (V)

Alternatively, the cost function h (V) can be defined as a negative sharpness (rho) as follows.

ρ = h (V) = - g (V)

As a result, optimal illumination is defined as an optimization problem that finds the maximum or minimum value of the following function, since ρ is located at the minimum point.

In the present invention, the negative sharpness (rho) is used for the following reason.

In order to find the optimal illumination voltage condition in the vision system, a voltage value having the maximum sharpness (σ) of the gray level is found. Since the optimization algorithms are configured to find the minimum value, the sharpness (σ) This is because the negative sharpness (rho) is more suitable.

<Differential operation>

을 사용하여 최소값을 찾는다.The optimization problem minimizes the amount of computation and the computation time. In the present invention, the differential value of the cost function

To find the minimum value.

In general, illumination control for machine vision involves digital computation, so that the derivative values can be obtained by pre-differential or post-derivative methods.

In the conjugate gradient method described below, the synthesized light is generated by adjusting the illumination input (driving) voltage (V) according to the differential of the sharpness, and the maximum value of the sharpness is quickly searched.

<Conjugate gradient algorithm>

(S36) of calculating the next drive voltage V ^{(k + 1)} by differential will be described as follows.

If the initial voltage is V ⁽⁰⁾ in the first step and the next drive voltage calculated by the optimization (conjugate gradient) algorithm is V ⁽¹⁾ , the next drive voltage V ⁽¹⁾ The same relationship can be calculated.

Where α ₀ is the convergence coefficient, d ⁽⁰⁾ is the derivative, and -∇ (h) is the derivative of the gradient.

In the second step, the next driving voltage (V ⁽²⁾ ) can be calculated by the following relationship.

If this step is repeated k times, the next drive voltage V ^{(k + 1)} can be calculated by the following equation (6).

<Decision of convergence factor>

을 최소화하는 지점,

에서 결정되나, α_k를 구하기 위해 조명 장치를 구동해야 하는데, 반응 속도가 수십~수백ms이므로 이 이론 수식을 적용할 경우 조명 장치를 다수 구동해야 하므로 시간이 많이 소요된다.In Equation (6),? _K is the convergence coefficient, which is theoretically

To minimize the point,

However, since the lighting system needs to be driven to obtain α _k , the reaction speed is several tens to several hundreds of ms. Therefore, when the theoretical formula is applied, it takes a long time to operate a large number of lighting devices.

Therefore, in the present invention,? _K is calculated by any one of the theoretical formulas satisfying the expression (7) and the alternative expressions (8) to (11).

Where η is a convergent constant.

Where τ is a vector quantity size limiting constant.

Here,? E is a computation error change value, and? Is another convergence constant.

In the above Equations 10 and 11, the calculation error E is calculated by one of the following three equations.

<Boundary condition discrimination>

Then, whether or not the driving voltage V ^{(k + 1)} adjusted by the conjugate gradient algorithm satisfies a boundary condition, that is, the maximum voltage value (for example, V _max : 5 V) example V _min: to determine if the between 0V) ignores (S38), the maximum voltage value and the drive voltage ^{(V (k + 1} if the outside of the minimum voltage ^value)) and the maximum voltage value and the minimum voltage value .

This can be expressed as follows.

을 만족할 때까지 반복하여 상기 종료 조건을 만족하지 못하는 경우(S40), 구동전압(V^(k+1))으로 조정하여 조명부(110)에 공급한(S42) 후 상기 S30단계로 되돌아가 수행한다.
Then, the above-described steps S30 to S38 are repeated for an extremely small ending constant &lt; RTI ID = 0.0 &gt;

^{(K + 1} ) and supplies the driving voltage V ^{(k + 1)} to the lighting unit 110 (S42), and then returns to the step S30 .

FIG. 4 is a conceptual diagram of a search time reduction method according to the conjugate gradient according to the present invention. The calculation time can be shortened by reducing the amount of computation as compared with the conventional uniform search of FIG.

In the case of the actual interval search, it is necessary to repeat 125,000 times in the range of 0 to 5 V for the RGB tri-color light source. However, according to the conjugate gradient of the present invention, the RGB three-color light source is repeated 51 times in the range of 0 to 5 V To find the optimum condition illumination.

At this time, the voltage difference of the optimal condition by equidistant search and conjugate gradient is 0.019V, which is negligible noise level, and the difference (Δu, Δv) = (0.0022, 0.0012) on the color coordinate is low.

Figs. 5A to 5C are photographs showing a dark image and a bright image, which are sample images, in comparison with an image optimized by the present invention. Fig.

6 is a graph comparing the distribution of the pixels optimized by the sample and the present invention.

In FIG. 6, the abscissa is gray level and the ordinate is the logarithmic interval of the number of pixels, and low intensity and high intensity are examples of samples compared with the R, G, B optimization according to the present invention.

It can be seen that the pattern and the background are clearly distinguished from the sample image, and the image under the light source optimized for the pixel distribution has a relatively even distribution than the sample.

Also, the peak of the RGB optimization result is formed at a lower level than a single light source, which means that the difference between the background and the pattern becomes clear and the image becomes clear.

In the above description, RGB three kinds of synthesized lights have been described in relation to color illumination, but it is needless to say that the present invention can also be applied to color lights having an arbitrary number of light sources.

Although the preferred embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, without departing from the scope and spirit of the invention as disclosed in the accompanying claims. In addition, it is a matter of course that various modifications and variations are possible without departing from the scope of the technical idea of the present invention by anyone having ordinary skill in the art.

100: illumination device 110: illuminating part
120: amplifier 130: optical fiber
140: mixer section 150: connection section
200: camera 210: translucent mirror
220: optical instrument 300: control system
10: object to be inspected

Claims (10)

A method for controlling a color vision of a vision system for finding a driving voltage of an illumination for optimal image acquisition in a vision system having a plurality of lights having different wavelengths,
(a) acquiring a digital image from a scanned object image incident on a camera and converting the digital image into a gray level image;
(b) calculating a sharpness (?) of the gray level image;
(c) Define the relationship between sharpness (ρ) and cost function (h (V)) as ρ = h (V) = - g (V) using the sharpness and drive voltage relation (σ = g ;
(d) calculating a next driving voltage V ^{(k + 1)} of the illumination using the derivative of the cost function and calculating a convergence coefficient? _k necessary for calculating the driving voltage V ^{(k + 1)} , And a computation error (E);
(e) determining a boundary condition for the driving voltage V ^{(k + 1)} ; And
(f) repeating steps (a) through (e) until the end condition │E│ <ε is satisfied for the end constant ε; A method for controlling color illumination of a vision system using a conjugate gradient algorithm. The method according to claim 1,
Wherein the sharpness (?) In the step (b) is calculated by the following equation using a standard deviation method.

Where m is the number of horizontal axis pixels of the image, n is the number of pixels in the vertical axis of the image, I (x, y) is the gray level value in the pixel corresponding to the x,

Is the average gray level of the entire image. The method according to claim 1,
Wherein the sharpness (?) In the step (b) is calculated by the following equation using an entropy method: a method for controlling a color vision of a vision system using a conjugate gradient algorithm;

Where p _i is the normalized value on the image histogram, h (i) is the number of pixels, and HW is the height and width of the image. The method according to claim 1,
Wherein the sharpness? In the step (b) is calculated by any one of the following two equations using an absolute difference method.

or

The method according to claim 1,
Wherein the sharpness (?) In the step (b) is calculated by the following equation using an automatic correction method.

The method according to claim 1,
Wherein the next driving voltage V ^{(k + 1)} is calculated by the following equation in the step (d).

Where d ^(k) is the current derivative, c ^(k) is ∇h (V ^(k) ), and ∇h is the derivative of the gradient. 7. The method according to claim 1 or 6,
The convergence factor (α _k) in step (d)

Of the color space of the vision system using the conjugate gradient algorithm. 7. The method according to claim 1 or 6,
Wherein the convergence coefficient? _K in the step (d) is calculated by one of the following four equations.

Where η is a convergent constant, τ is a vector quantity size limiting constant, and β is another convergent constant. The method according to claim 1,
Wherein the calculation error (E) in the step (d) is calculated by one of the following three equations.

The method according to claim 1,
Wherein the step (e) is represented by the following equation: &lt; EMI ID = 17.0 &gt;

Images