講演抄録/キーワード |
講演名 |
2020-05-28 11:35
Real-time object detection by point cloud information from a TOF camera with an SOC-FPGA ○Zijian HU・Yoshiki Yamaguchi(Univ.of Tsukuba) RECONF2020-3 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
A Time of Flight (ToF) camera operates by sending a modulated infrared light and receiving the objects reflected light. Time cost between the sending and receiving is measured and calculated to distance. ToF technology is revolutionizing the machine vision industry by providing point cloud information. This paper proposes two implementations of SoC-FPGA based real-time object detection with the depth image captured by ToF camera. The one named HDD (Histogram of Depth Difference), which is a HOG-like feature extractor, it could robustly calculate the feature histogram of local pixel area and simultaneously provide it for SVM classifier. Another one named flood-fill that is a classical recursive algorithm, which usually used in the bucket fill tool of paint programs to fill connected and similarly colored areas with a different color. The implementation of this algorithm could also color fill different objects in the depth image depends on the boundary information of themselves. A series of experiments have been carried out both indoor and outdoor under different illumination conditions with the frame rate of 30fps and VGA resolution. Finally, the results have been evaluated and discussed which proves this research might be useful in the coming future. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Real-time objects detection / SoC-FPGA / ToF camera / Self-driving / / / / |
文献情報 |
信学技報, vol. 120, no. 36, RECONF2020-3, pp. 13-18, 2020年5月. |
資料番号 |
RECONF2020-3 |
発行日 |
2020-05-21 (RECONF) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
査読に ついて |
本技術報告は査読を経ていない技術報告であり,推敲を加えられていずれかの場に発表されることがあります. |
PDFダウンロード |
RECONF2020-3 |