2013 Fiscal Year Final Research Report
Performance Improvement for Classifiers by Optimizing Training Samples
| Project/Area Number |
22500172
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | National Institute of Advanced Industrial Science and Technology |
|---|
Principal Investigator |
NISHIDA Kenji 独立行政法人産業技術総合研究所, ヒューマンライフテクノロジー研究部門, 主任研究員 (50344148)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KURITA Takio 広島大学, 大学院工学研究院情報部門, 教授 (10356941)
|
|---|
| Project Period (FY) |
2010-04-01 – 2014-03-31
|
| Keywords | パターン認識 / 画像認識 / データマイニング / ITS |
|---|
| Research Abstract |
A precise vehicle detection-and-tracking algorithm that is robust for the appearance change of the objects has been determined using an optimization of the training sample-set.
An algorithm for obtaining the Random-Subset SVM ensemble is determined. The algorithm combines some weak-classifiers that are trained by small samples extracted from whole training sample set. The required training time for the ensemble classifier is considered to be smaller than the training a classifier using full-set of training samples, while attaining better generalization performance.
|
