ログイン
言語:

WEKO3

  • トップ
  • ランキング
To
lat lon distance
To

Field does not validate



インデックスリンク

インデックスツリー

メールアドレスを入力してください。

WEKO

One fine body…

WEKO

One fine body…

アイテム

{"_buckets": {"deposit": "78b8f2c4-a2b1-4e27-abd7-d0996944be54"}, "_deposit": {"id": "9853", "owners": [], "pid": {"revision_id": 0, "type": "depid", "value": "9853"}, "status": "published"}, "_oai": {"id": "oai:nagoya.repo.nii.ac.jp:00009853", "sets": ["734"]}, "author_link": ["29682"], "item_12_biblio_info_6": {"attribute_name": "書誌情報", "attribute_value_mlt": [{"bibliographicIssueDates": {"bibliographicIssueDate": "2009-03-25", "bibliographicIssueDateType": "Issued"}}]}, "item_12_date_granted_64": {"attribute_name": "学位授与年月日", "attribute_value_mlt": [{"subitem_dategranted": "2009-03-25"}]}, "item_12_degree_grantor_62": {"attribute_name": "学位授与機関", "attribute_value_mlt": [{"subitem_degreegrantor": [{"subitem_degreegrantor_language": "ja", "subitem_degreegrantor_name": "名古屋大学"}, {"subitem_degreegrantor_language": "en", "subitem_degreegrantor_name": "Nagoya University"}], "subitem_degreegrantor_identifier": [{"subitem_degreegrantor_identifier_name": "13901", "subitem_degreegrantor_identifier_scheme": "kakenhi"}]}]}, "item_12_degree_name_61": {"attribute_name": "学位名", "attribute_value_mlt": [{"subitem_degreename": "博士(情報科学)", "subitem_degreename_language": "ja"}]}, "item_12_description_4": {"attribute_name": "抄録", "attribute_value_mlt": [{"subitem_description": "This thesis addresses the problem of recognizing 3D objects and human faces from still-images and video-sequences. Here, the recognition problem is formulated as an appearance matching process. In the appearance-based approach, an object is represented in the form of two-dimensional image sets. To gain efficiency, these images are then projected into a low dimensional space in which the images of each object are represented as Eigenpoints. However, since the appearance of an object in a two-dimensional image is influenced by several parameters, such as shape, pose, illumination, etc., it is important for the recognition system to capture these variations for gaining high recognition performance. A novel method which performs generalization of Eigenpoints through feature lines, called the Modified Nearest Feature Line (MNFL) method is proposed. The feature lines can be acquired by corresponding every pair of Eigenpoints in the same class and projecting every Eigenpoint to the constructed feature lines. Since the feature lines virtually provide an infinite number of Eigenpoints in each class, it expands the capacity of the available database and increases the system’s ability to capture object variations. Moreover, focusing on the problem of pose variability, an appearance manifold with View-dependent Covariance matrix (VC) method is proposed. Being different from the feature line scheme which forms a generalization of Eigenpoints regardless of their pose, the appearance manifold scheme constructs a continuous curve (appearance manifold) by linking two Eigenpoints of consecutive poses. The appearance manifold is constructed along with view-dependent covariance matrices, so that it could capture pose variability and also learn the samples’ distribution of each pose for gaining robustness to pose changes and also degradation effects. Finally, a new incremental unsupervised-learning framework of appearance manifolds is also proposed to present more realistic recognition applications. It is obvious that it is difficult to collect large amounts of training images which depict an object under all poses (from left sideview to right sideview). The incremental unsupervised-learning framework allows us to train the system with initial imagesequences, and later updates the existing categories incrementally every time an unlabeled image-sequence is input. The unlabeled images are first recognized based on the minimum distance of their projected Eigenpoints to the manifolds of objects and then the results are integrated to produce the final sequence’s decision. The performance of the proposed methods were studied through experiments and the results showed that the proposed methods could accurately recognize 3D objects and human faces from still-images and video-sequences under a wide variety of poses, expressions, and degradation effects.", "subitem_description_language": "en", "subitem_description_type": "Abstract"}]}, "item_12_description_5": {"attribute_name": "内容記述", "attribute_value_mlt": [{"subitem_description": "名古屋大学博士学位論文 学位の種類:博士(情報科学) (課程) 学位授与年月日:平成21年3月25日", "subitem_description_language": "ja", "subitem_description_type": "Other"}]}, "item_12_dissertation_number_65": {"attribute_name": "学位授与番号", "attribute_value_mlt": [{"subitem_dissertationnumber": "甲第8401号"}]}, "item_12_identifier_60": {"attribute_name": "URI", "attribute_value_mlt": [{"subitem_identifier_type": "HDL", "subitem_identifier_uri": "http://hdl.handle.net/2237/11662"}]}, "item_12_select_15": {"attribute_name": "著者版フラグ", "attribute_value_mlt": [{"subitem_select_item": "publisher"}]}, "item_12_text_14": {"attribute_name": "フォーマット", "attribute_value_mlt": [{"subitem_text_value": "application/pdf"}]}, "item_12_text_63": {"attribute_name": "学位授与年度", "attribute_value_mlt": [{"subitem_text_value": "2008"}]}, "item_access_right": {"attribute_name": "アクセス権", "attribute_value_mlt": [{"subitem_access_right": "open access", "subitem_access_right_uri": "http://purl.org/coar/access_right/c_abf2"}]}, "item_creator": {"attribute_name": "著者", "attribute_type": "creator", "attribute_value_mlt": [{"creatorNames": [{"creatorName": "Lina", "creatorNameLang": "en"}], "nameIdentifiers": [{"nameIdentifier": "29682", "nameIdentifierScheme": "WEKO"}]}]}, "item_files": {"attribute_name": "ファイル情報", "attribute_type": "file", "attribute_value_mlt": [{"accessrole": "open_date", "date": [{"dateType": "Available", "dateValue": "2018-02-20"}], "displaytype": "detail", "download_preview_message": "", "file_order": 0, "filename": "thesis_Lina2009.pdf", "filesize": [{"value": "3.1 MB"}], "format": "application/pdf", "future_date_message": "", "is_thumbnail": false, "licensetype": "license_note", "mimetype": "application/pdf", "size": 3100000.0, "url": {"label": "thesis_Lina2009.pdf", "objectType": "fulltext", "url": "https://nagoya.repo.nii.ac.jp/record/9853/files/thesis_Lina2009.pdf"}, "version_id": "f568f8d3-391e-4a52-be78-e015588b9a19"}]}, "item_language": {"attribute_name": "言語", "attribute_value_mlt": [{"subitem_language": "eng"}]}, "item_resource_type": {"attribute_name": "資源タイプ", "attribute_value_mlt": [{"resourcetype": "doctoral thesis", "resourceuri": "http://purl.org/coar/resource_type/c_db06"}]}, "item_title": "3D Object and Human Face Recognition using Appearance Manifold with View-dependent Covariance Matrix", "item_titles": {"attribute_name": "タイトル", "attribute_value_mlt": [{"subitem_title": "3D Object and Human Face Recognition using Appearance Manifold with View-dependent Covariance Matrix", "subitem_title_language": "en"}]}, "item_type_id": "12", "owner": "1", "path": ["734"], "permalink_uri": "http://hdl.handle.net/2237/11662", "pubdate": {"attribute_name": "PubDate", "attribute_value": "2009-05-01"}, "publish_date": "2009-05-01", "publish_status": "0", "recid": "9853", "relation": {}, "relation_version_is_last": true, "title": ["3D Object and Human Face Recognition using Appearance Manifold with View-dependent Covariance Matrix"], "weko_shared_id": -1}
  1. A500 情報学部/情報学研究科・情報文化学部・情報科学研究科
  2. A500d 学位論文
  3. 博士論文(情科博・論情科博)

3D Object and Human Face Recognition using Appearance Manifold with View-dependent Covariance Matrix

http://hdl.handle.net/2237/11662
http://hdl.handle.net/2237/11662
cfac9725-be73-4e9f-9398-cfa9d67240bf
名前 / ファイル ライセンス アクション
thesis_Lina2009.pdf thesis_Lina2009.pdf (3.1 MB)
Item type 学位論文 / Thesis or Dissertation(1)
公開日 2009-05-01
タイトル
タイトル 3D Object and Human Face Recognition using Appearance Manifold with View-dependent Covariance Matrix
言語 en
著者 Lina

× Lina

WEKO 29682

en Lina

Search repository
アクセス権
アクセス権 open access
アクセス権URI http://purl.org/coar/access_right/c_abf2
抄録
内容記述 This thesis addresses the problem of recognizing 3D objects and human faces from still-images and video-sequences. Here, the recognition problem is formulated as an appearance matching process. In the appearance-based approach, an object is represented in the form of two-dimensional image sets. To gain efficiency, these images are then projected into a low dimensional space in which the images of each object are represented as Eigenpoints. However, since the appearance of an object in a two-dimensional image is influenced by several parameters, such as shape, pose, illumination, etc., it is important for the recognition system to capture these variations for gaining high recognition performance. A novel method which performs generalization of Eigenpoints through feature lines, called the Modified Nearest Feature Line (MNFL) method is proposed. The feature lines can be acquired by corresponding every pair of Eigenpoints in the same class and projecting every Eigenpoint to the constructed feature lines. Since the feature lines virtually provide an infinite number of Eigenpoints in each class, it expands the capacity of the available database and increases the system’s ability to capture object variations. Moreover, focusing on the problem of pose variability, an appearance manifold with View-dependent Covariance matrix (VC) method is proposed. Being different from the feature line scheme which forms a generalization of Eigenpoints regardless of their pose, the appearance manifold scheme constructs a continuous curve (appearance manifold) by linking two Eigenpoints of consecutive poses. The appearance manifold is constructed along with view-dependent covariance matrices, so that it could capture pose variability and also learn the samples’ distribution of each pose for gaining robustness to pose changes and also degradation effects. Finally, a new incremental unsupervised-learning framework of appearance manifolds is also proposed to present more realistic recognition applications. It is obvious that it is difficult to collect large amounts of training images which depict an object under all poses (from left sideview to right sideview). The incremental unsupervised-learning framework allows us to train the system with initial imagesequences, and later updates the existing categories incrementally every time an unlabeled image-sequence is input. The unlabeled images are first recognized based on the minimum distance of their projected Eigenpoints to the manifolds of objects and then the results are integrated to produce the final sequence’s decision. The performance of the proposed methods were studied through experiments and the results showed that the proposed methods could accurately recognize 3D objects and human faces from still-images and video-sequences under a wide variety of poses, expressions, and degradation effects.
言語 en
内容記述タイプ Abstract
内容記述
内容記述 名古屋大学博士学位論文 学位の種類:博士(情報科学) (課程) 学位授与年月日:平成21年3月25日
言語 ja
内容記述タイプ Other
言語
言語 eng
資源タイプ
資源 http://purl.org/coar/resource_type/c_db06
タイプ doctoral thesis
書誌情報
発行日 2009-03-25
学位名
言語 ja
学位名 博士(情報科学)
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 13901
言語 ja
学位授与機関名 名古屋大学
言語 en
学位授与機関名 Nagoya University
学位授与年度
学位授与年度 2008
学位授与年月日
学位授与年月日 2009-03-25
学位授与番号
学位授与番号 甲第8401号
フォーマット
application/pdf
著者版フラグ
値 publisher
URI
識別子 http://hdl.handle.net/2237/11662
識別子タイプ HDL
戻る
0
views
See details
Views

Versions

Ver.1 2021-03-01 11:54:12.035749
Show All versions

Share

Mendeley Twitter Facebook Print Addthis

Cite as

エクスポート

OAI-PMH
  • OAI-PMH JPCOAR
  • OAI-PMH DublinCore
  • OAI-PMH DDI
Other Formats
  • JSON
  • BIBTEX

Confirm


Powered by WEKO3


Powered by WEKO3