講演抄録/キーワード |
講演名 |
0000-00-00 00:00
CNN-Based Radar-Video Fusion for Object Classification ○Chih-Hsuan Huang・Dan-Dan Yao・Kun-Shan Chen・Chiung-Shen Ku(Xuchang Univ.)・Zhao-Liang Li(CNRS)・Gen-yuan Du(Xuchang Univ.) |
抄録 |
(和) |
An object classification system based on sensor fusion with millimeter-wave (MMW) radar and camera is investigated. Recently, the importance of enhancing intelligent system performances has gained more attention. MMW radar and camera sensors play a complementary role in providing more object's information and feature. In the lab testing, both static and motion detections were conducted and the corner-shape targets were selected in three different types of material, which are wood, plastic, and metal within ten different colors. The fusion algorithms are consisting of convolutional neural network (CNN) by adding into color, range, radar cross-section and Doppler information to extract object feature. The data were taken 2000 samples by placing the targets within different position and aspect angle. By the proposed method, it not only improves the accuracy rate in classification and recognition but also serves as a good foundation of reference target to integrate the temporal component of videos into intelligent systems. |
(英) |
An object classification system based on sensor fusion with millimeter-wave (MMW) radar and camera is investigated. Recently, the importance of enhancing intelligent system performances has gained more attention. MMW radar and camera sensors play a complementary role in providing more object's information and feature. In the lab testing, both static and motion detections were conducted and the corner-shape targets were selected in three different types of material, which are wood, plastic, and metal within ten different colors. The fusion algorithms are consisting of convolutional neural network (CNN) by adding into color, range, radar cross-section and Doppler information to extract object feature. The data were taken 2000 samples by placing the targets within different position and aspect angle. By the proposed method, it not only improves the accuracy rate in classification and recognition but also serves as a good foundation of reference target to integrate the temporal component of videos into intelligent systems. |
キーワード |
(和) |
MMW / CNN / fusion / classification / / / / |
(英) |
MMW / CNN / fusion / classification / / / / |
文献情報 |
信学技報 |
資料番号 |
|
発行日 |
|
ISSN |
|
PDFダウンロード |
|
|