S. Chandrashekhar and R. Chellappa. Passive navigation in a partially known environment. Proc. IEEE Workshop on Visual Motion, pages 2--7, 1991.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....gravity, inertia and initial velocities. We present an optimization framework to identify these physical parameters from video. Our dynamic model captures the true rotational physics of a tumbling rigid body. This aspect distinguishes our work from prior work in motion tracking and analysis [4,7,11,5], where the focus is on identifying object kinematics, i.e. motion trajectories. Moreover, A. Heyden et al. Eds. ECCV 2002, LNCS 2350, pp. 551 565, 2002. c Springer Verlag Berlin Heidelberg 2002 552 K.S. Bhat et al. Fig. 1. Four frames of a tumbling object superimposed (left) and a ....

....the inertial parameters. The problem of simultaneously recovering the physical parameters of the object, camera, and environment from a single camera has not been previously addressed. 554 K.S. Bhat et al. Our work is closely related to prior work on model based tracking in computer vision [11,5,21,4,7,24,17,16]. However, the notion of a dynamic model in the tracking literature is different from the one presented here. We use ordinary differential equations to model the non linear rotational dynamics of tumbling rigid bodies, and extract its parameters from video. These parameters include initial ....

[Article contains additional citation context not shown here]

S. Chandrashekhar and R. Chellappa. Passive navigation in a partially known environment. Proc. IEEE Workshop on Visual Motion, pages 2--7, 1991.

....than in [1] except it assumed that the centre of rotation is always visible, which appears difficult in practice to achieve. Silv en and Repo [9] have recently developed an integrated monocular visual tracking system, with an emphasis on real time operation. Finally, Chandrashekhar and Chellappa [10] use known navigational landmarks in their state estimation formulation, and interleave motion estimation and feature correspondence. Salient features of these recursive algorithms are compared in Table I. In contrast, optical flow based motion and structure estimation requires computation of an ....

S. Chandrashekhar and R. Chellappa, "Passive Navigation in a Partially Known Environment", in Proceedings of the IEEE Workshop on Visual Motion. Institute of Electrical and Electronic Engineers, 1991, pp. 2--7.

....= jj U jj 2 jj P jj 2 = jj Ujj2 X3 jj Y jj 2 is the depth scaled observer speed at Y at time t. We refer to as relative depth in this paper. 1. 2 Literature Survey In recent years there has been considerable interest in using long image sequences to recover motion and structure [1, 9, 13, 19, 21, 23, 15, 20, 24, 6, 5, 18, 7]. Most of the algorithms use monocular image sequences but a few use binocular image sequences [16, 25] A few of the algorithms are batch (processing all the data at once) 21, 6, 4, 20] but most use some type of recursive estimation method (for example, an extended Kalman filter) 4, 5, 10] or ....

.... 24, 6, 5, 18, 7] Most of the algorithms use monocular image sequences but a few use binocular image sequences [16, 25] A few of the algorithms are batch (processing all the data at once) 21, 6, 4, 20] but most use some type of recursive estimation method (for example, an extended Kalman filter) [4, 5, 10] or (ofter nonlinear) minimization or factorization methods [6, 20] Batch methods are conceptually less complicated but recursive estimation methods yield a best answer at any given time and may be computationally more feasible when using real time acquired data. Most methods are based on point ....

[Article contains additional citation context not shown here]

S. Chandrashekhar and R. Chellappa. Passive navigation in a partially known environment. In IEEE Workshop on Visual Motion, pages 2--7, 1991.

....Kalman filter to integrate the computed binocular motion parameters over time, thus providing a best estimate of the parameter values at each time. 2. LITERATURE SURVEY There is an extensive body of research on recovering monocular motion and structure parameters from long image sequences [6, 8, 23, 10, 29, 16, 25, 31, 5, 22, 7]. Typically these use pixel, point or feature based correspondence as the input and compute motion and structure parameters for monocular sequences. Due to depth speed ambiguity, 3D depth and 3D translational speed cannot be recovered, only their ratio. A few algorithms assume stereo image ....

....translational speed cannot be recovered, only their ratio. A few algorithms assume stereo image sequences [17, 33] but usually for restricted circumstances (for example, pure translation) Some of the algorithms use some type of recursive estimation method (for example, an extended Kalman filter) [30, 4, 5, 8, 9]. Only a few of the methods involve Kalman filtering of motion parameters that are derived from optical flow. In particular, the work proposed by De Micheli, Torre and Uras [18] uses the optical flow method of [26] to compute flow in long image sequences and then computes time to collision and ....

S. Chandrashekhar and R. Chellappa. Passive navigation in a partially known environment. In IEEE Workshop on Visual Motion, pages 2--7, 1991.

....image plane they can robustly solve for u, then and finally, given the computed u and , they can solve for at each image point where there is an image velocity measurement. In recent years there has been considerable interest in using long image sequences to recover motion and structure [15, 18, 42, 28, 49, 25, 24, 58, 36, 48, 60, 14, 13, 41, 16]. Most of the algorithms use monocular image sequences but a few use image binocular sequences [37, 61] A few of the algorithms are batch (processing all the data at once) 49, 14, 10, 48] but most use some type of recursive estimation method (for example, an extended Kalman filter) 59, 10, 13, ....

.... 13, 41, 16] Most of the algorithms use monocular image sequences but a few use image binocular sequences [37, 61] A few of the algorithms are batch (processing all the data at once) 49, 14, 10, 48] but most use some type of recursive estimation method (for example, an extended Kalman filter) [59, 10, 13, 18, 20] or (often non linear) minimization or factorization methods [14, 48, 42] Batch methods may be more accurate [48] but they are a form of post hoc analysis. Recursive estimation methods can yield a best answer at any given time and may be computationally more feasible when using real time ....

[Article contains additional citation context not shown here]

S. Chandrashekhar and R. Chellappa. Passive navigation in a partially known environment. In IEEE Workshop on Visual Motion, pages 2--7, 1991.

....This is formulated as a recursive tracking problem, with the dual objective of estimating the motion of the camera, and tracking feature points in the image sequence. The method used for feature point matching is discussed in Section 3.3.1. The motion estimation aspects are discussed in detail in [35], and are summarized in Section 3.3.2. The problem of motion correspondence is somewhat similar to the face recognition problem in the sense that both require a correspondence between distinct features in two or more images, or between stored patterns and a test pattern. In both cases, labeled ....

....neighborhood being proportional to the scale. 3.3.2 Interleaving Matching and Motion Estimation The matching process for the motion correspondence problem is interleaved with the recursive estimation of 3 D motion parameters. Details of the recursive estimation technique used may be found in [35]. They are briefly summarized here. The motion parameters consist of 3 D feature point positions, camera velocities and camera pose parameters. These are contained in a state vector . The recursive estimator used is the extended Kalman filter, which operates in two steps, a time update or ....

S. Chandrashekhar and R. Chellappa, "Passive navigation in a partially known environment," in IEEE Workshop on Visual Motion, (Princeton, NJ), pp. 2--7, October 1991. 18

No context found.

S. Chandrashekhar and R. Chellappa. Passive navigation in a partially known environment. Proc. IEEE Workshop on Visual Motion, pages 2--7, 1991.

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