Contents lists available at ScienceDirect
Optics Communications
journal homepage: www.elsevier.com/locate/optcom
Weak scratch detection and defect classification methods for a large-
aperture optical element
Xian Tao, De Xu, Zheng-Tao Zhang
⁎
, Feng Zhang, Xi-Long Liu, Da-Peng Zhang
Research Center of Precision Sensing and Control, Institute of Automation, Chinese Academy of Sciences, Beijing, China
ARTICLE INFO
Keywords:
Optical inspection
Weak scratches
Surface defects classification
Large-aperture optical element
ABSTRACT
Surface defects on optics cause optic failure and heavy loss to the optical system. Therefore, surface defects on
optics must be carefully inspected. This paper proposes a coarse-to-fine detection strategy of weak scratches in
complicated dark-field images. First, all possible scratches are detected based on bionic vision. Then, each
possible scratch is precisely positioned and connected to a complete scratch by the LSD and a priori knowledge.
Finally, multiple scratches with various types can be detected in dark-field images. To classify defects and
pollutants, a classification method based on GIST features is proposed. This paper uses many real dark-field
images as experimental images. The results show that this method can detect multiple types of weak scratches in
complex images and that the defects can be correctly distinguished with interference. This method satisfies the
real-time and accurate detection requirements of surface defects.
1. Introduction
Large-aperture optics are indispensable in inertial confinement
fusion (ICF) high-power laser systems, such as National Ignition
Facility (NIF) [1] in the United States and Laser MegaJoule (LMJ)
Facility [2] in France, and have the dual function of guiding and
amplifying laser beams. The surface defects of large-aperture optics,
such as scratches, digs, and bubbles, may harm the safe operation of
high-power laser system. Therefore, detecting surface defects has an
important practical significance. Currently, automated visual inspec-
tion based on the dark-field scattering principle should be a rapid and
effective method to detect defects of large-aperture optical elements.
There are a number of methods to detect surface defects. The French
ICF project LMJ uses a high-resolution 45-Mpixel CCD camera with an
LED array edge-illuminating frame [3] to detect defects [4]. proposes a
defect detection method using dark-field back illumination and a
camera to obtain the flaw information for full sized optics at one time,
but the accuracy of these two methods is at most 110 µm [5,6]. design a
surface defect evaluation system (SDES) with an area CCD and a
precise two-dimensional motion platform. The detection diameter of
the SDES is up to 810×460 mm
2
with a resolution of 0.5 µm, but the
image acquisition time in sub-aperture scanning is longer than half an
hour. To achieve fast and high-precision detection of large-aperture
optical elements, a new and effective inspection instrument [7,8] with
two imaging systems for large-aperture optical elements was proposed.
It combines a dark-field imaging system (DFIS) with a line-scan
camera at a 10 µm resolution and a bright-field imaging system
(BFIS) with a microscopic camera at a 0.85 µm resolution. The DFIS
is developed to scan the entire optical element to quickly obtain all of
the distributions of surface flaws for coarse detection. BFIS is used to
measure the size of the flaws and recognize the flaw type for fine
detection.
The original system works well, but defect detection encounters
problems in some special conditions. On the one hand, scratches on the
optics, such as Nd-doped glass, are usually shallow. In [9] Fig. 1, the
depth of the scratch after burnishing is only approximately 200–
300 nm, as obtained by an interference microscope. The detected
scratches in our previous paper [8] are deep and obvious and can be
effectively detected by conventional binarization methods. In some
related works [10–12], the detected flaws are also often notably deep or
have a large area. However, these detection methods can lead to partial
detection and lead to undetectable problems when they encounter
these weak scratches. Meanwhile, there are surface defects in the
optical element compared to pollutants, such as dust, fibres, dirt, bright
spot, fingerprint, and so on. These disturbances are inevitably involved
in the inspection and can affect the detection results [13]. realizes
automated discrimination between digs and dust particles on optical
surfaces using dark-field scattering microscopy. In [14], the machine
learning method is applied to classify linear defects and circular flaws.
In current detection systems, such as those described in [7,8], these
disturbances are removed by manual marking before defect detection
begins. This method significantly reduces the detection efficiency. To
http://dx.doi.org/10.1016/j.optcom.2016.10.062
Received 11 July 2016; Received in revised form 26 October 2016; Accepted 28 October 2016
⁎
Corresponding author.
E-mail address: zhengtao.zhang@ia.ac.cn (Z.-T. Zhang).
Optics Communications 387 (2017) 390–400
Available online 10 November 2016
0030-4018/ © 2016 Elsevier B.V. All rights reserved.
MARK