http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
A Fusion Approach Based ColorSiFT and Dense Trajectory Features for Video Event Recognition
Nguyen Thanh Dieu,Tran Anh Tuan,Pham The Bao,Bui Nogc Nam,Jin Young Kim 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.3
Event recognition plays an essential role to conduct a high-performing content based video retrieval system. Recently, Dense Trajectory Features has attracted much attention due to its effective encoding for motion features. On the other hand, beside the Gradient Image, ColorSIFT provide an option to extract the apparent model of interest targets. In order to develop a high-performing system, in this paper, we approach the fusion methods of appearance feature ColorSIFT and motion feature MBH, wich are following Dense Trajectory. After that, the model BoW (Bag of Words) is utilized to aggregate a feature set into a single vector. And then, SVM classifier is used to recognize them. Our approach gives the performance of early fusion classification of 95.1% that outperforms other conventional features under the UCF11 dataset.