Visual detection of tobacco packaging film based on apparent features
The main purpose of this article is to study the detection of transparent film on the surface of tobacco packs. Tobacco production line needs an industrial robot to remove the transparent film in the process of unpacking. Therefore, after the industrial robot removes the transparent film, it is nece...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
SAGE Publishing
2021-06-01
|
Series: | International Journal of Advanced Robotic Systems |
Online Access: | https://doi.org/10.1177/17298814211024839 |
id |
doaj-b42f58e904a6442a8cab08b839794a15 |
---|---|
record_format |
Article |
spelling |
doaj-b42f58e904a6442a8cab08b839794a152021-06-29T23:03:46ZengSAGE PublishingInternational Journal of Advanced Robotic Systems1729-88142021-06-011810.1177/17298814211024839Visual detection of tobacco packaging film based on apparent featuresZhenxun Jin0Fengyan Zhong1Qiang Zhang2Weisong Wang3Xuanyin Wang4 China Tobacco Zhejiang Industrial Co., Ltd, Hangzhou, China Zhejiang University, Hangzhou, China China Tobacco Zhejiang Industrial Co., Ltd, Hangzhou, China China Tobacco Zhejiang Industrial Co., Ltd, Hangzhou, China Zhejiang University, Hangzhou, ChinaThe main purpose of this article is to study the detection of transparent film on the surface of tobacco packs. Tobacco production line needs an industrial robot to remove the transparent film in the process of unpacking. Therefore, after the industrial robot removes the transparent film, it is necessary to use machine vision technology to determine whether there is transparent film residue on the surface of tobacco packaging. In this article, based on the study of the optical features of semitransparent objects, an algorithm for detecting the residue of transparent film in tobacco packs based on surface features is proposed. According to the difference of surface features between tobacco and film, a probability distribution model considering highlights, saturation, and texture density is designed. Because the probability distribution model integrates many features of tobacco and film, it is more reasonable to distinguish the tobacco film regions. In this article, an appropriate foreground box with a trapezoidal mask and image segmentation algorithm GrabCut is used to segment the foreground area of tobacco pack more accurately, and the possible film area is obtained by image differential and morphological processing. Finally, on the basis of comparing the effect of various machine learning algorithms on the image classification of possible film regions, support vector machine based on color features is used to judge the possible film region. Application results of the system show that the method proposed in this article can effectively detect whether there is film residue on the surface of tobacco pack.https://doi.org/10.1177/17298814211024839 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Zhenxun Jin Fengyan Zhong Qiang Zhang Weisong Wang Xuanyin Wang |
spellingShingle |
Zhenxun Jin Fengyan Zhong Qiang Zhang Weisong Wang Xuanyin Wang Visual detection of tobacco packaging film based on apparent features International Journal of Advanced Robotic Systems |
author_facet |
Zhenxun Jin Fengyan Zhong Qiang Zhang Weisong Wang Xuanyin Wang |
author_sort |
Zhenxun Jin |
title |
Visual detection of tobacco packaging film based on apparent features |
title_short |
Visual detection of tobacco packaging film based on apparent features |
title_full |
Visual detection of tobacco packaging film based on apparent features |
title_fullStr |
Visual detection of tobacco packaging film based on apparent features |
title_full_unstemmed |
Visual detection of tobacco packaging film based on apparent features |
title_sort |
visual detection of tobacco packaging film based on apparent features |
publisher |
SAGE Publishing |
series |
International Journal of Advanced Robotic Systems |
issn |
1729-8814 |
publishDate |
2021-06-01 |
description |
The main purpose of this article is to study the detection of transparent film on the surface of tobacco packs. Tobacco production line needs an industrial robot to remove the transparent film in the process of unpacking. Therefore, after the industrial robot removes the transparent film, it is necessary to use machine vision technology to determine whether there is transparent film residue on the surface of tobacco packaging. In this article, based on the study of the optical features of semitransparent objects, an algorithm for detecting the residue of transparent film in tobacco packs based on surface features is proposed. According to the difference of surface features between tobacco and film, a probability distribution model considering highlights, saturation, and texture density is designed. Because the probability distribution model integrates many features of tobacco and film, it is more reasonable to distinguish the tobacco film regions. In this article, an appropriate foreground box with a trapezoidal mask and image segmentation algorithm GrabCut is used to segment the foreground area of tobacco pack more accurately, and the possible film area is obtained by image differential and morphological processing. Finally, on the basis of comparing the effect of various machine learning algorithms on the image classification of possible film regions, support vector machine based on color features is used to judge the possible film region. Application results of the system show that the method proposed in this article can effectively detect whether there is film residue on the surface of tobacco pack. |
url |
https://doi.org/10.1177/17298814211024839 |
work_keys_str_mv |
AT zhenxunjin visualdetectionoftobaccopackagingfilmbasedonapparentfeatures AT fengyanzhong visualdetectionoftobaccopackagingfilmbasedonapparentfeatures AT qiangzhang visualdetectionoftobaccopackagingfilmbasedonapparentfeatures AT weisongwang visualdetectionoftobaccopackagingfilmbasedonapparentfeatures AT xuanyinwang visualdetectionoftobaccopackagingfilmbasedonapparentfeatures |
_version_ |
1721354250886316032 |