3-D Shape Reconstruction from Camera Motion with Inexact Motion Parameters

When a sequence of images are obtained by a moving camera, if the exact camera velocity and corresponding points on the images are determined, the 3-D shape of the object can be reconstructed using stereo method. However, in general, it is difficult to determine both of them exactly. In this paper, the reconstruction of 3-D shape, using the sequence of images of the object is described, for the case when the camera velocity cannot be known exactly. First, we calibrate the camera position and camera velocity at a time from image points whose 3-D positions are already calculated exactly using previous images in the sequence, then, using this camera velocity and the positions of corresponding points on the image whose 3-D positions are not determined yet, we calculate their 3-D exact positions. Repeating these two phases on the sequence of images, we reconstruct the 3-D position of the object. In the later phase, to reduce the effect of quantization errors, Kalman filtering method is int...

• Type: Conference
• Year: 1990
• Where: MVA
• Authors: Keisuke Kinoshita, Koichiro Deguchi
• Keywords: 3-D Shape, Camera Velocity, Image, MVA 1990, MVA 2007