Abstract—Machine vision detection and recognition
technology has a wide range of advantages such as non-contact,
real-time online, fast, strong anti-interference ability. This
technology has been used to achieve zero inferior parts
production of mechanical parts, to meet the requirements of
modern manufacturing progress and development. In the
actual production it shows a broad application prospects. In
this paper, we apply the Fully Convolutional Networks (FCN˅
method to the robot vision system for mechanical part
inspection. We found that the use of FCN can make the
classification accuracy quickly reach a stable value, thus
greatly reducing the training time. The proposed method is
based on (1)our own database is collected on an industrial
conveyor belt, which contains 2248 images of various types of
parts with different sizes, (2) Fully Convolutional
Networks(FCN) is trained by using back-propagation
algorithm to extract valuable features from input training
data ,and realized pixel-level semantic classification on the
validation images.
Index Terms
—Fully Convolutional Networks
mechanical
components
semantic segmentation, deep learning.
I. I
NTRODUCTION
ACHINE vision is known as the automation of the
eyes. In the national economy, scientific research and
national defense construction and other fields have a wide
range of applications. The biggest advantage of machine
vision is that there is no contact with the object which being
obs erved, s o the obs ervation and the obs erver will not have
any damage, very safe and reliable, this is the other way of
feeling cannot match. In addition, the machine vision can
detect the object is very broad, it can be said that most of the
things can be detected. In theory, machine vision can also
observe the scope that the human eye cannot observe
For
example, humans cannot observe the infrared, microwave,
ultrasonic, but machine vision can use some sensitive
Manuscript received April 15th, 2017. This work was supported by
the National Science Foundation of China (NSFC) under Grants
61461022 and 61302173.
The authors are with the Department of Mechanical Engineering,
Faculty of Mechanical and Electrical Engineering,
Kunming University of Science and Technology,No. 727 Jingming Sou
th Road, Chenggong, Kunming, Yunnan, China 650500.
* Corresponding author: 412467892@qq.com.
devices for these to form infrared, microwave, ultrasound and
other images. So it can be said that machine vision extends
the human visual range. In addition, people cannot observe
the object for a long time, computer vision is not tired,
always observing, so, the machine vision can be widely used
for a long time harsh working environment.
With the continuous development of manufacturing,
advanced manufacturing technology research and
application more and more widely, advanced manufacturing
technology, automated manufacturing systems and
advanced production models to require the use of advanced
detection and identification means to adapt to.
Machine vision is a kind of detection method suitable for
the development of modern manufacturing technology. First
of all, the machine vision can achieve non-contact online
detection, to test all parts of the production line, to meet the
automated manufacturing system in the process of testing
and process testing requirements. Second, the machine
vision detection is through the computer or digital signal
processor in the program to process the image information
obtained by the measurement results, so the machine vision
detection has a certain intelligence and flexibility.
Third, as long as selecting high-precision lens and image
sensor, machine vision detection technology can achieve
higher detection accuracy; finally, the machine vision is easy
to achieve information integration and management, to
achieve the computer integrated manufacturing technology
to provide the necessary support.
In China, the application of machine vision technology
began in the 1990s , becaus e the industry is a new area, and
machine vision product technology is not enough popularity,
leading in a lot of industry applications are basically in a
blank state, even if there are only simple Applications. As
the supply, the core technology, the prices and many other
factors, the domestic market has been in a slow state of
development.
China is a manufacturing country, and the current product
tes ting is s till by artificial means , it s erious ly res tricting the
development of China's manufacturing industry. Mechanical
parts and components processing have been achieved with
mechanization, automation, but mechanical parts of the
quality of testing, mainly rely on artificial to complete,
usually visual inspection method, instrument measurement
method. In the detection process, the human factor and the
Semantic segmentation of mechanical parts
based on Fully Convolutional Network
Yuqi Wu, Yinhui Zhang*, Chunquan Zhang, Zifen He, Yue Zhang
The 9th International Conference on Modelling, Identification and Control
(ICMIC 2017), Kunmin
, China, Jul
10-12, 2017
612