Object Detection inside Virtual Cylindrical Screen using Optical Flow from Wide Angle Foveated Optical System.
Object Detection inside Virtual Cylindrical Screen using Optical Flow from Wide Angle Foveated Optical System.
复制标题
使用广角焦点光学系统的光流进行虚拟圆柱屏幕内的物体检测。
DOI：
10.1299/kikaic.64.979
复制
发表时间：
1998
期刊：
影响因子：
--
通讯作者：
H. Hosono
中科院分区：
文献类型：
--
作者：
S. Shimizu;Y. Suematsu;Jianming Yang;H. Hosono
为了实现光学上的凹陷视觉信息，研制了一种广角高失真透镜系统。该透镜系统在中心视觉和外围视觉的应用中具有几个优点，并且它能够基于来自单个相机的图像进行多用途使用。本文提出了虚拟柱面屏幕检测前场景目标的新概念，并提出了一种基于序列图像间光流的单目运动视觉方法来检测虚拟柱面屏幕内存在的静止目标。对于移动的机器人来说，快速识别前面的对象是很重要的。针对该方法的特点，比较了特殊镜头系统和针孔相机的性能。基本实验结果也证明了该方法的有效性。
A wide angie high distortion lens system has been developed to realize a foveated visual information optically. The lens system has several advantages in the application of the central vision and the peripheral vision, and it enables multi-purpose use based on the image from a single camera. This paper introduces a new concept of Virtual Cylindrical Screen to detect objects in the front scene, and proposes a method of monocular motion vision, based on the optical flow between the sequential images, to detect stationary objects existing inside the Virtual Cylindrical Screen. For a moving robot it is important to recognize quickly the object in the front. The comparisons between the special lens system and a pinhole camera are carried out with respect to the characteristics of the proposed method. The basic experimental results also prove that the proposed method is effective.