Spectral Imaging for Material Detection and Classification

Wang,  Chao

WEKO3

lat lon distance

[[sub_check.contents]]

[[sub_radio.contents]]

Field does not validate

[[sub_attr.contents]]　

インデックスツリー

アイテム

マテリアルの検出と識別のためのスペクトルイメージング

https://doi.org/10.18997/00008634

名前 / ファイル	ライセンス	アクション
jou_k_360.pdf (8.7 MB)

アイテムタイプ

学位論文 = Thesis or Dissertation(1)

公開日

2021-12-06

資源タイプ

資源タイプ識別子

http://purl.org/coar/resource_type/c_db06

資源タイプ

doctoral thesis

タイトル

Spectral Imaging for Material Detection and Classification

言語

タイトル

マテリアルの検出と識別のためのスペクトルイメージング

言語

eng

著者

Wang, Chao

抄録

内容記述タイプ

Abstract

内容記述

Material refers to the substance that makes up an object. They come in various forms, for example, solids, gases, and liquids. There are many important applications for material detection and classification, such as visual inspection, robot path planning, and texture understanding, etc. Material detection and classification can be achieved through imaging technology. When light enters the surface of an object, various optical phenomena occur, and these optical phenomena are related to the object material and the wavelength of the incident light. However, due to the limited spectral sensitivities of RGB cameras, part of the information in the outgoing light is lost, and therefore different materials could appear the same color under the same observation condition, i.e. metamerism occurs. Therefore, the conventional imaging technique with usual RGB cameras makes the detection and classification of materials difficult. In this dissertation, we proposed novel methods for material detection and classification via spectral imaging. Specifically, our methods are based on the fact that optical phenomena such as reflection, absorption, and scattering depends on materials themselves as well as wavelengths of incident light. In material （liquid） detection, based on the optical phenomenon of absorption whose characteristic is described by spectral absorption coefficient, we proposed novel camera-based spectral imaging for water detection and then extended it to water and oil detection. First, in water detection, we propose an approach to per-pixel water detection on unknown surfaces with a hyperspectral image. Our proposed method is based on the water spectral characteristics: water is transparent for visible light but translucent/opaque for near-infrared （NIR） light and therefore the apparent near-infrared spectral reflectance of a surface is smaller than the original one when water is present on it. Specifically, we use a linear combination of a small number of basis vectors to approximate the spectral reflectance from visible to NIR wavelengths and estimate the original NIR reflectance from the visible reflectance （which does not depend on the presence or absence of water） to detect water. Second, in water and oil detection, we propose a novel per-pixel water and oil detection method based on the Lambert-Beer's law and the low-dimensional linear model for spectral reflectance. We show that our method enables us to pixelwisely detect water and oil on surfaces with unknown and spatially-varying reflectance at high accuracy by using a hyperspectral image. In material classification, because the appearance of material depends not only on the wavelength but also on the direction of incident light, we extend camera-based spectral imaging to illumination-based spectral imaging. We propose an approach to one-shot per-pixel classification of raw materials on the basis of spectral BRDFs; a surface of interest is illuminated by multispectral and multidirectional light sources at the same time. Specifically, we achieve two-class classification from a single color image; it directly finds the linear discriminant hyperplane with the maximal margin in the spectral BRDF feature space by jointly optimizing the non-negative coded illumination and the grayscale conversion. In addition, we extend our method to multiclass classification by exploiting the degree of freedom of the grayscale conversion. The experiments using an LED-based multispectral dome show that the performance of our proposed method with only a single image is better than or comparable to the state-of-the-art methods with multiple images. Furthermore, by imposing a non-negative value and sparse constraint on the light source intensity, our method achieves material classification with a single color image taken under a single color image with a small number of light sources.

内容記述タイプ

TableOfContents

内容記述

1 Introduction||2 Spectral Imaging||3 Material Detection||4 Material Classification||5 Conclusions

備考

内容記述タイプ

Other

内容記述

九州工業大学博士学位論文学位記番号：情工博甲第360号学位授与年月日：令和3年9月24日

キーワード

主題Scheme

Other

主題

Spectral Imaging

キーワード

主題Scheme

Other

主題

Material Detection

キーワード

主題Scheme

Other

主題

Material Classification

キーワード

主題Scheme

Other

主題

Computational Photography

キーワード

主題Scheme

Other

主題

Computer Vision

アドバイザー

岡部, 孝弘

学位授与番号

甲第360号

学位名

博士（情報工学）

学位授与年月日

2021-09-24

学位授与機関

学位授与機関識別子Scheme

kakenhi

学位授与機関識別子

17104

学位授与機関名

九州工業大学

学位授与年度

内容記述タイプ

Other

内容記述

令和3年度

出版タイプ

VoR

出版タイプResource

http://purl.org/coar/version/c_970fb48d4fbd8a85

アクセス権

open access

アクセス権URI

http://purl.org/coar/access_right/c_abf2

ID登録

10.18997/00008634

ID登録タイプ

JaLC

戻る

views

See details

	Views

Versions

Ver.1

2023-05-15 12:56:44.089160

Show All versions

Cite as

Other

エクスポート

OAI-PMH

JPCOAR 2.0
JPCOAR 1.0
DublinCore
DDI

Other Formats

インデックスリンク

インデックスツリー

アイテム

マテリアルの検出と識別のためのスペクトルイメージング

× Wang, Chao

Versions

Share

Cite as

Other

エクスポート

コミュニティ

メニューを最小化

インデックスリンク

インデックスツリー

アイテム

マテリアルの検出と識別のためのスペクトルイメージング

× Wang, Chao

Versions

Share

Cite as

Other

エクスポート

コミュニティ