= Special TALK／IW-FCV2017 in Seoul ==
KIZKI-algorithm as
OMNIPOTENT Image Defect
Inspection and Its
Applications
Hiroyasu KOSHIMIZU, Kimiya AOKI,
Takuma FUNAHASHI (Chukyo University),
Yasuhiko MIWATA (RIKEN) and
Hiroyuki ISHII (TOYOTA)
key OMNIPOTENT solution to the instrumentation of
automated inspection processes, the ultimate problem
of the full automation of production.
<REFERENCES>
[1] H.Koshimizu，K.Aoki，T.Aoki，Y.Miwata，H.Ishii：Modeling
of Human Inspection Mechanism for Instrumentation of
Visual Inspection in Production Line ― Development of
“KIZKI” Consciousness Algorithm for Defects ―，Proc. JSPE，
Vol.83，No.2（2017.2）
[2] K.Aoki, T.Funahasi, H.Koshimizu, Y.Miwata, :”KIZUKI”
Processing for Visual Inspection – A Smart Pattern Pop-out
Algorithm based on Human Visual Architecture-", 22nd
International Conference on Pattern Recognition (ICPR2014),
(2014-8), pp.2317-2322, Stockholm-Sweden (2014.8)
Key words: appearance inspection, visual inspection,
inspection mechanism, consciousness of attention
<OVERVIEW>
Since the inspection is fatally indispensable process for
every kinds of productions, the number of image inspection
algorithms becomes just the same as the number of the
products. Thus several kinds of individually customized
image processing systems have been valid for the
individual inspection tasks so far. Because we do not
have yet an idea for establishing general purpose or
OMNIPOTENT method for coping with the diversity of
defects in the industrial fields.
Ultimate goal to do now is to establish ‘OMNIPOTENT’
method. The KIZKI consciousness algorithm for
defects is a smart image processing technology
initiated by human visual inspection mechanism.
In this proposed KIZKI algorithm (Japanese Patent
No.5821708), we developed a simple iterative and
simultaneous scheme of multiple image processing
processes realized by both the coarse to fine spatial
resolutions processing similar to the human periphery
vision and the spatially tremor phase processing
similar to the human micro saccadic vision.
(a) Input image (b) KIZKI map (c) Pop-out image
Fig.1 EXAMPLES: Same KIZKI procedure applied to
Different defects
<OMNIPOTENT algorithm & experiments>
The backgrounds of KIZKI algorithm were
characterized by both the eye-camera investigation of
human inspectors and the knowledge of ocular
physiology. It was already experimentally certified that
the KIZKI algorithm can be applicable to several
different kinds of defects such as cracks and dents on
mechanical and electronic parts, just by tuning a few
parameters, as shown in Fig.1. And Fig.2 shows the
flowof an implemented KIZKI image processing.
Fig.2 implemented KIZKI image processing
<SUCCESS in car production line>
And the proposed KIZKI technology was already
fabricated as shown in Fig.3 in the real production line
of the motor car electronic devices, and was currently
started to apply also to the mechanical parts in the car
production line. Thus it is industrially significant that
the proposed KIZKI technology could provide a unique
Fig.3 KIZKI technology in Car production line for
inspecting electronic devices (JSPE Technology
Awards 2016)