Professional Documents
Culture Documents
Ma2012 PDF
Ma2012 PDF
Ma2012 PDF
Abstract: In order to realize the fast detection of the printed In the study of PCB automatic test system, because
circuit board's quality, a set of PCB defect intelligent automatic foreign research institutions and companies started relatively
detection device based on machine vision is analysed and early, at present, the AOI system of PCB in our country
designed. The paper introduced the robot system market is basically foreign brands. Because the domestic
construction,machine system,electronic control system,the related technology still stay in the initial stage and
design and realization of vision imaging system. software used a exploratory research stage, domestic equipment performance
fast iterative algorithm for image segmentation based on 2D can't compared with the similar foreign equipment. These
maximum between-cluster variance, using the template cause it stays in a passive disadvantage status in the
matching method for PCB image of the automatic detection and competition.
recognition defects. Experiment results on model machine show As the domestic PCB manufacturer can't bear the great
that the design of the intelligent visual inspection robot can economic pressure caused by the expensive price of import
detect PCB defect exactly and effectively, it also can be used in equipment and domestic operators also isn't used to
the actual PCB online detection. operating these equipment. Combining with the actual
Keywords: machine vision; PCB; detection robot; defect reco situation of our country enterprise , the paper studies a PCB
gnition defect intelligent measuring robot based on machine vision.
I. INTRODUCTION The system uses machine vision instead of the human eye for
With the application scope of modern electronic on-line measuring and sorts out the conforming product and
equipment expanding, PCB (PCB:Printed Circuit Board) as nonconforming product through the intelligent machine arm,
the carrier of an integrated, installation of electronic then it accurately detects defects that may appear in PCB
components plays a more and more important role, which production process such as hiatus,spots, burr short circuit,
makes the PCB quality become the core problem in narrow line, open circuit pinhole and so on. The robot can
electronic manufacturing industry. The current domestic improve the speed and precision of PCB quality inspection,
PCB manufacturer still adopt the method of function test and reduce the rate of false detection and missed detection.
II. SYSTEM DESIGN
artificial visual inspection in the production line, this far
couldn't keep up with the pace of PCB manufacturing According to the investigation of actual production site,
technology. Relative to the limitation of the artificial there is a variety of reasons caused PCB defects. These
detection, AOI (AOI:Automatic Optic Inspection) system defects may exist in the process of every working procedure.
has obvious advantage in speed, accuracy, repeatability, According to artificial detection and semi-automatic
reliability and so on. So the AOI system is being applied detection system, the paper designed a PCB defect intelligent
more and more in all kinds of production line. visual inspection robot and developed special-purpose
L −1 L −1 L −1 L −1
1 1 (5)
μ 1 = ( μ 1i , μ 1 j ) T = [
w1
∑∑ iPij ,
w1
∑∑ jPij ]
Figure 4. Defect detection and recognition flow diagram i=t j=0 i=0 j=s
necessary to enhance image contrast. The article uses the In this paper, the trace t r δ b of discrete measurement
method of interception type linear transformation. matrix between target class and background class is used to
⎧ c ; f ( x, y ) < a their distance measurement function.
⎪ d −c
g ( x, y ) = ⎨c + [ f ( x, y ) − a ]; a ≤ f ( x, y ) < b
⎪ b−a trδb (t, s) = w0[(μ0i − μzi )2 + (μ0 j − μzj )2 ] + w1[(μ1i − μzi )2 + (μ1 j − μzj )2 ] (7)
⎩ d ; f ( x, y ) ≥ b
(1)
in formula (1), f ( x , y ) represents for the original image, The paper choose the corresponding threshold vector of
g ( x , y ) represents for the transformation image. According the maximum distance measurement function as the optimal
to transformation function formula (1), the gray level map threshold vector ( t * , s * ) of 2d Otsu method. Then we can
from [a, b] to [c, d]. get the best threshold vector from type (3) to type (7).
In addition, aimming at salt &pepper noise and gaussian 1 1 (8)
(t * , s * ) = ( ( μ 0 i + μ 1i ), ( μ 0 j + μ 1 j ) )
noise that polluted the image, this paper uses the 5 x 5 2 2
median filter to remove. Then it does the Laplace sharpness The initial values of t and s are setted to image gray
to the denoised image. level mean t 0 and neighbourhood average gray level mean s 0 .
B. Image segmentation Putting them into type (8) to iterate until both the results of t
It is suppose that M × N image gray level is 0, 1,... , L-1. and s are fixed values, namely t(i+1)=t(i),s(i+1)=s(i).
Let us use set Z to say L gray levels, then C. Defect detection and recognition
Defects detection is using already divided PCB image uses area threshold processing method to get the accurate
to match with the standard PCB image which does the same defects figure.
image processing, then judge them qualified or not. IV. THE EXPERIMENT RESULTS AND ANALYSIS
I ( x, y ) = f ( x, y ) ⊗ g ( x, y ) (9) The configuration of the main controlling computer
in formula (9), f ( x , y ) is the original image, g ( x , y ) is the system is Intel Core 2 Duo CPU, 2 GB memory, 2.20 GHz
transformation image, I ( x , y ) is the pixel with different frequency. Software development environment applies vc
grayscale value in the corresponding position of the two ++ 6.0. The different stages of the image processing results
images. Obviously, because of the limitation of manufacture in this algorithm is shown in fig.5.
process, it is sure to produce some noise.The paper would