非線形距離指標学習によるカメラ間人物照合

IR IR (HANDLE) Web Site Web Site Web Site 2 Citations 82 References Open Access

Bibliographic Information

Other Title
  • 非線形距離指標学習によるカメラ間人物照合(一般セッション,医用画像処理分野における計測・認識・理解)
  • ヒセンケイ キョリ シヒョウ ガクシュウ ニ ヨル カメラ カン ジンブツ ショウゴウ
  • Human Re-identification by Non-linear Distance Metric Learning

Search this article

Description

To do tracking of a specific person, since a field of view in a camera is limited, the use of multiple surveillance cameras is essential. On the other hand, even if the number of these cameras is increased, in many practical scenarios it is difficult to cover a large area without any blind spot. Therefore, human tracking across cameras with non-overlapping field of views have been studied. In previous work, clothing color matching is the most popular method, assuming that clothing colors are not changed in short time. However, matching with histograms via different color calibrations affects adversely accuracy. Therefore in this paper we propose to use distance metric learning with nonlinear kernels to do robust matching in spite of camera differences. The effectiveness of the proposed method is proved through experimental studies.

Journal

Citations (2)*help

See more

References(82)*help

See more

Keywords

Report a problem

Back to top