Deakin University
Browse

File(s) under permanent embargo

An automatic visual inspection method based on statistical approach for defect detection of ship hull surfaces

Version 2 2024-06-12, 18:40
Version 1 2019-04-30, 09:54
conference contribution
posted on 2024-06-12, 18:40 authored by A Jalalian, WF Lu, FS Wong, SM Ahmed, CM Chew
Robotized blasting of ship hull surfaces requires an accurate identification of defective regions of the hull to maximize the blasting efficiency. Accurate surface defect detection may not be achieved by current manual procedures, as its success is highly vulnerable to the human operators' experience and their subjective judgements. Therefore, there is a need for a more accurate and non-subjective method for defect detection. This paper proposes a computer vision based method for detection of ship hull defects. The method utilizes the histogram of hue and entropy data of the hue to identify the defects in two steps. Step 1 is an automatic circular thresholding based on the histogram of hue to distinguish the defects whose hue is different from the defect-free regions. A wrapped Gaussian mixture model is utilized to estimate the circular hue histograms, and maximum likelihood criterion is adopted to set the thresholds. Step 2 uses the probability distribution of the entropy for each segment identified in the first step to decide whether the segments are either defective, defect-free or a mixture of both. For the mixed regions, a Gaussian mixture model is fitted to the probability distribution of the entropy. The maximum likelihood criterion is utilized to segment these regions so as to discriminate their defective and defect-free parts. The high accuracy (F-measure=0.89) and short execution time (3.5 s) of the proposed method show that it is a good starting point for an automatic defect detection for a fully autonomous ship hull blasting.

History

Volume

2018-August

Pagination

445-450

Location

Singapore

Start date

2018-08-21

End date

2018-08-23

ISSN

2161-8070

eISSN

2161-8089

ISBN-13

9781538635933

Language

eng

Publication classification

E Conference publication, E1.1 Full written paper - refereed

Copyright notice

2018, IEEE

Title of proceedings

CASE 2018 : 2018 IEEE International Conference on Automation Science and Engineering

Event

Automation Science and Engineering. Conference (14th : 2018 : Singapore)

Publisher

IEEE

Place of publication

Piscataway, N. J.

Usage metrics

    Research Publications

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC