2P1-G04 Adaptive K-Nearest Neighbor for Pose-Invariant Object Recognition

スポンサーリンク

概要

• 論文の詳細を見る

• In order to make mobile robots manipulate correct objects, firstly robots need ability to recognize and identify object precisely. Linear Discriminant Analysis (LDA) with static K-Nearest Neighbor has been widely used to solve pose-invariant object recognition problem. However it is difficult to select the best static K value for all unknown objects. To solve this problem, we introduce an improved algorithm called Adaptive K-Nearest Neighbor (AK-NN) that allows object recognition system uses an automatic adaptive K value to improve the accuracy of classification. We experimentally compare its accuracy and efficiency with Dynamic K-Nearest Neighbor (DK-NN), Static K-Nearest Neighbor (K-NN) and Nearest Neighbor (1-NN).

• 一般社団法人日本機械学会の論文
• 2009-05-25

著者

• Dung Le

  Shibaura Institute Of Technology

• KANLAYA Wittayathawon

  Shibaura Institute of Technology

• MIZUKAWA Makoto

  Shibaura Institute of Technology

関連論文

• 2P1-G04 Adaptive K-Nearest Neighbor for Pose-Invariant Object Recognition
• 2P1-G03 Hand feature extraction based on distance transformation and Hough transformation
• The Position-Adjusting System for an Autonomous Mobile Robot using Cylinders as Landmarks in the Environment of "Micro-Clipper Robot Contest"(Vision and Recognition 2,Session: MP1-D)

スポンサーリンク