A Visual Inspection System for Surface Mounted Components Based on
Color Features
Hui-Hui Wu, Xian-Min Zhang, Shi-Liang Hong
Abstract—The object of this study is to develop a reliable
and fast visual inspection system for surface mounted
components after they have been placed in wet solder paste on
a printed circuit board (PCB). In order to reach this goal,
firstly, the features of the components under three colors (red,
greed, blue) structure light source are analyzed. Then, a two-
stages inspection process is presented. In the first stage, the
horizontal and vertical integral projections of the component
electrode were obtained after segmentation in the red sub-
image, based on the projections, the location of component
was gotten by the sliding location window algorithm, then, the
defects such as missing component, wrong component, shift
and rotation were detected. In the second stage, the three color
features were extracted from the component body and the
Bayes classifier was used to inspect another wrong component
class. The proposed inspection system have been implemented
and tested with various types of components collected from
production line. The experiment results have verified the
validity of this scheme in terns of recognition rate and speed.
Index Terms
—Surface mounted components; electrodes;
color features; integral projection; Bayes classifier
INTRODUCTION
ITH
the development of surface mounting technology
(SMT), the current trends toward miniaturization of
components, denser packing of boards and highly automated
assembly equipment make the task of detecting these
defects on the PCBs more critical and more difficult, several
visual inspection systems have been reported with different
classification techniques[1-10].Nevertheless, most attention
were paid to the solder joints after reflow [1-4],however, the
later a defect is detected, the more expensive it is to be
repaired, thus, early detection (i.e., after surface mounting)
is inherently necessary [7].
MARK et al. developed an image vector technique for
components inspection [9]. The approach was fast, but
results in many false alarms since it ignored the features
such as color and light. Gallegos etal used a template
matching and wavelet decomposition to detect component
absence, which was sensitive to the component rotation
Manuscript received January 15,2009;This work was supported in part by
the Foundation of National Outstand Young Science
(50825504)
;Guangdong and Hong Kong Technology Cooperation Funding
(Dongguan Project:200816822)
H.-H. Wu is with the School of Mechanical and Automotive Engineering of
South China University of Technology, Guangzhou, China. (e-mail:
wuhuihui215@sohu.com).
X.-M. Zhang is with the School of Mechanical and Automotive
Engineering of South China University of Technology, Guangzhou, China.
(e-mail: zhangxm@scut.edu.cn).
S.-L. Hong is with the School of Mechanical and Automotive Engineering
of South China University of Technology, Guangzhou, China.
[10]. The other decision makers, suck as fuzzy systems,
neural networks with high performances were also applied
to detect the components [4], however, the slow operation
made it rarely used in real time system.
There are different types of components in the SMT,
among these, rectangular components with two electrodes
extending beyond the body of the components are the most
common components such as resistor, capacitor etc, thus, in
this paper, we focus our attention on visually inspecting the
placement quality of surface mounted rectangular
components. In order to develop a strategy to inspect the
components fast and reliable, the proposed system and
methodology is presented in the following sections. In the
next subsections the features of components under three
colors (red, greed, blue) structure light is analyzed. In
Section , a two-stages inspection process combined with
the segmentation, integral projection and Bayes classifier
was presented. Experimental results that highlight the
potential of the developed algorithm are given in Section IV
and the paper concludes in Section V.
COMPONENT FEATURE ANALYSE
It was reported that the feature distance of the
component should be enlarged under the structure light
source[12], so a lighting system is developed as shown in
Fig.1.Three layers (red, green, blue) of ring-shaped LEDs
project structure light from various position and orientation
onto the mounted component, a top-view CCD camera is
used to acquire images of components. It is found by
experimentation that color features of components are
enrich and reliable since the illumination of the lighting
system is uniform.
CCD
Red ring LEDs
Green ring LEDs
Blue ring LEDs
Component
Solder paste
PCB
z
Fig.1 The image acquisition system
In this study, the qualities of mounted component are
divided into five classes: good component (GC), missing
component (MC), component rotation (CR), component
shift (CS), wrong component (WC).Before inspection, the
W
978-1-4244-3608-8/09/$25.00 © 2009 IEEE.
571
Proceedings of the 2009 IEEE
International Conference on Information and Automation
June 22 -25, 2009, Zhuhai/Macau, China