作成者 |
|
|
|
本文言語 |
|
出版者 |
|
|
発行日 |
|
収録物名 |
|
巻 |
|
出版タイプ |
|
アクセス権 |
|
関連DOI |
|
|
関連URI |
|
|
関連情報 |
|
|
概要 |
This paper proposes a method for recognizing moving objects, which is based on a wavelet decomposition technique and learning of eigenspaces. High frequency vectors for reference image sequences are c...onstructed using a wavelet decomposition formula. These high frequency vectors express the characteristics of moving objects in the image sequences. From them, a covariance matrix is made and eigenvectors of this matrix are computed. Some parametric eigenspaces are learnt by these eigenvectors. Next, orthogonal projections of the reference image sequences on the parametric eigenspaces are produced. To recognize a moving object included in an unknown image sequence, a high frequency vector and its orthogonal projections on the parametric eigenspaces are constructed. The recognition is carried out by computing the distance between the projection for the unknown image sequence and that for the reference image sequences.続きを見る
|