P. McLauchlan. A batch/recursive algorithm for 3D scene reconstruction. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, Hilton Head, South Carolina, 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... scene structure to a common reference frame defined by the initial camera pose, as in the work in robotics known as simultaneous localization and mapping (SLAM) 9, 10, 11, 8] using laser range scanners [10, 12, 11, 13] Some vision researchers have pursued similar approaches for limited scenes [5, 14]. 2. THE ALGORITHM VPs intersection line tracks clusters rotation updates translation updates edges 3D lines Fig. 1. Data Flow Graph. Figure 1 summarizes the data flow in our system. Given a sequence of omni directional images and detected linear features, our task is to estimate the 3D ....

McLauchlan, P.F.: A batch/recursive algorithm for 3d scene reconstruction. In: Int. Conf. Computer Vision and Pattern Recognition. Volume 2., Hilton Head, SC, USA (2000) 738--743

....how to move are made continuously, and an updated SFM solution is needed for each decision. Solving SFM has been approached in a number of ways, from non linear least squares bundle adjustment [10, 12] using Levenburg Marquadt (LM) to recursive methods using the extended Kalman filtering (EKF) [6, 11, 1, 3, 5] to linear methods on a few frames [4, 2, 9] Bundle adjustment is the process of iteratively adjusting the camera poses and point positions in order to move toward the optimal least squares answer. As with all gradient descent methods, there is the possibility of getting stuck in local minima. If ....

....adjustment. The accuracy gained by increasing the size of the moving window past some point should become small and not be worth the extra computational cost of increasing it. 3. Recursive Methods Kalman filtering and other recursive techniques such as the Variable State Dimension Filter (VSDF) [6] make first order assumptions about the non linear error function as a way to speed up processing. In this section, we look at where these approximations are valid and cases where they are not. The EKF has been implemented in a couple of different ways [11, 1, 3, 5] EKF methods typically have a ....

[Article contains additional citation context not shown here]

P. McLauchlan. A batch/recursive algorithm for 3d scene reconstruction. In CVPR00, pages II:738--743, 2000.

....for the highly non linear structure and motion problem. Kalman filtering and batch methods for calibrated image sequences using no regularity constraints on the motion were analyzed in [25] and they concluded that the performance difference between these two types of methods is quite large. In [14] a recursive method is proposed that closely approximates a batch process, using a variable state dimension filter, but still the camera motion is not modelled. 2. The Camera Model The following notations will be used: X denotes object points in homogeneous coordinates, x denotes image points in ....

P. McLauchlan. A batch/recursive algorithm for 3d scene reconstruction. In Conf. Computer Vision and Pattern Recognition, volume 2, pages 738--743, Hilton Head SC, USA, 2000.

No context found.

P. McLauchlan. A batch/recursive algorithm for 3D scene reconstruction. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, Hilton Head, South Carolina, 2000.

No context found.

P. McLauchlan. A batch/recursive algorithm for 3D scene reconstruction. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, Hilton Head, South Carolina, 2000.

No context found.

P. McLauchlan, A batch/recursive algorithm for 3D scene reconstruction, in: Proc. CVPR#00, vol. 2, 2000, pp. 738--743.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC