Wide Angle Stereo Camera with Monocular Vision using Camera Posture Estimation and Correction
Wide Angle Stereo Camera with Monocular Vision using Camera Posture Estimation and Correction
- 提供方法
- 本サイト上にてダウンロード・閲覧可
- 形態
- 価格
- 一般価格(税込):¥1,100 会員価格(税込):¥880
- 文献番号
- 20226268
- 文献・情報種別
- 学術講演会予稿集(秋)
No.139-22
- 掲載ページ
- 1-5(Total 5 p)
- 発行年月
- 2022年 10月
- 出版社
- (公社)自動車技術会
- 言語
- 英語
- イベント
- 2022年秋季大会
書誌事項
著者(英) | 1) Felipe Gomez-Caballero, 2) Masayuki Kobayashi, 3) Kazuyoshi Yamazaki, 4) Takuma Osato, 5) Takeshi Endo, 6) Masayuki Takemura, 7) Takeshi Nagasaki, 8) Takeshi Shima |
---|---|
勤務先(英) | 1) Hitachi, 2) Hitachi, 3) Hitachi, 4) Hitachi, 5) Hitachi, 6) Hitachi, 7) Hitachi Astemo, 8) Hitachi Astemo |
抄録 | Stereo cameras have been used to effectively avoid traffic accidents, but peripheral vision of current systems may be insufficient for full safety coverage at road intersections, where most accidents involving pedestrians occur. Also, problems related to detection distance performance degradation and image distortion remain when simply increasing the system field of view (FOV). In response, we developed a stereo camera employing a multi-shift approach to preserve far object detection performance and obtain wider FOV without using expensive image sensors. We also developed an object detection system based on top view image subtraction method, enabling pedestrian and cyclist detection in the monocular region created at the outer region of the stereo camera FOV. Furthermore, we can achieve high accuracy for detection and distance measurement at the monocular region by using the disparity information from the central region to estimate the camera posture and correct it in real-time. |
抄録(英) | Stereo cameras have been used to effectively avoid traffic accidents, but peripheral vision of current systems may be insufficient for full safety coverage at road intersections, where most accidents involving pedestrians occur. Also, problems related to detection distance performance degradation and image distortion remain when simply increasing the system field of view (FOV). In response, we developed a stereo camera employing a multi-shift approach to preserve far object detection performance and obtain wider FOV without using expensive image sensors. We also developed an object detection system based on top view image subtraction method, enabling pedestrian and cyclist detection in the monocular region created at the outer region of the stereo camera FOV. Furthermore, we can achieve high accuracy for detection and distance measurement at the monocular region by using the disparity information from the central region to estimate the camera posture and correct it in real-time. 翻訳 |