J.W. Roach and J.K. Aggarwall. Determining the movement of objects from a sequence of images. IEEE Transactions on Patterm Analysis and Machine Intelligence, 2(6):554-562, 1980.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....in which the objects in the recorded scene are dicult to distinguish, either because of poor recordings, poor recording conditions, restricted recording devices media, or because the objects appear identical anyway. The research elds concerned with these issues are among others object tracking [25], feature or token tracking [3] 13] 27] 38] and optical ow or motion estimation [12] 19] Applications range from surveillance [20] 30] 36] motion analysis, and structure from motion [31] 32] 35] 37] to (multi )target tracking [11] 21] 24] Here, we restrict ourselves to the case ....

J.W. Roach and J.K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(6):554-562, 1980.

....stations [4] Faig s iterative approach incorporating radial symmetric and decentering lens distortion [7] and Thompson s method on linearizing the nonlinear equations and iterating the process [35] This latter method obtains a rotation matrix guaranteed to be orthonormal. Roach and Aggarwal [32] later developed a nonlinear algorithm and dealt with noisy data. Their results show that accuracy can be improved by increasing the number of corresponding point pairs. Common problems with the non linear algorithms in self calibration are the multiplicity of solutions and the selection of the ....

Roach, J. W., and Aggarwal, J. K. Determining the Movement of Objects from a Sequence of Images. IEEE Trans. on Pattern Analysis and Machine Intelligence, 6 (1980), 554--562.

....absence of a target to minimize the computational cost of evaluating the candidate trajectory sets. Feature Detection and Tracking A large number of algorithms, which compute the three dimensional motion of a rigid object from a set of n feature correspondences over m frames, have been proposed [56,58,94 97]. However, relatively few researchers have addressed the problem of automating the establishment of feature correspondences in a robust and computationally efficient manner. The establishment of feature correspondences is confounded by problems in feature detection which can lead to the ....

J.W. Roach and J.K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2:554--562, 1980.

....Since the time between images can be measured, the result is an approximation of translational and rotational velocities. Much existing work begins by assuming that the problems of extracting feature points and establishing the required correspondences have been accomplished. Roach and Aggarwal [6] develop a system of nonlinear equations that relate the measured image plane coordinates, the (x; y; z) coordinates on the object and the camera position parameters. Numerical techniques for solving the nonlinear equations are discussed, as well as methods for choosing good initial conditions and ....

....two images in the sequence is also made, by averaging the results of each of the successive pairs. This is shown to give no significant improvement. It is widely held that most existing schemes for motion estimation perform very poorly when the data (image coordinates of match points) are noisy [6, 9, 10, 11]. In [6, 7, 9] smoothing is achieved by using a larger number of match points in each image. This gives some improvement [9, 10] but as discussed in [12] using additional match points in each frame introduces additional unknown parameters, which limits the amount of new information that can be ....

[Article contains additional citation context not shown here]

J.W. Roach and J.K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Trans. on Patt. Anal. Mach. Intell., PAMI-2:554--562, November 1980.

....missing and noisy data. 27 Since then, most work in the analysis of image sequences of moving objects has been directed to the analysis of the two dimensional movement of objects (Martin Aggarwal 1978) Finally, in 1980, an attempt to recover 3 D information in object tracking was made. Roach and Aggarwal (1980) experimented on finding the three dimensional model of points on an object s surface as well as its movement (up to a scale factor) from a sequence of images from multiple views. A technique for solving for viewpoint and model parameters was independently developed by Lowe (1980, cited by Lowe ....

Roach, J. W. and Aggarwal, J. K. (1980). Determining the movement of objects from a sequence of images, IEEE Transactions on Pattern Analysis and Machine Intelligence PAMI-2(6): 554-562.

....images has been studied for a while in the computer vision community. We can trace it back to the late seventies: Ullman [19] assumed a orthographic camera projection model and showed that three views are necessary to recover the motion and structure from point correspondences; Roach and Aggarwal [15] used a full perspective projection model and thus two views are sufficient from point correspondences. Since then, many approaches have been proposed to solve the problem using either linear or nonlinear methods. The reader is referred to [11] for a complete review. Essentially, two types of ....

J.W. Roach and J.K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Trans. PAMI, 2(6):554--562, 1980.

....the image motion is related to the 3D motion and structure in a non linear way. It is this non linearity that has generated a large amount of research in the area of structure from motion. Initial attempts at addressing this problem started with the observation that five point correspondences [Roach and Aggarwal, 1980] are needed, since any point correspondence gives two equations and the number of unknowns is equal to nine (four scaled depth values, two translational and three rotational motion parameters) Such approaches suggested obtaining the structure and motion parameters through iterative methods ....

J.W. Roach and J.K. Aggarwal. Determining the movements of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2:554--562, 1980.

....images has been studied for a while in the computer vision community. We can trace it back to the late seventies: Ullman [22] assumed a orthographic camera projection model and showed that three views are necessary to recover the motion and structure from point correspondences; Roach and Aggarwal [18] used a full perspective projection model and thus two views are suOEcient from point correspondences. Since then, many approaches have been proposed to solve the problem using either linear or nonlinear methods. The reader is referred to [1, 12] for a complete review, and to [16] for a ....

J.W. Roach and J.K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Trans. PAMI, 2(6):554562, 1980.

....is then established between these features frame by frame. Finally, motion is computed based on a series of correspondences. Three approaches to estimate motion via feature based analysis are along Aggarwal and Nandhakumar s review [Agga88] the direct formulation of rigid body motion [Ullm79a, Ullm79b, Roac79, Roac80, Tsai84, Faug87, Webb82], the approach of the explicit use of rigidity [Miti85, Miti86] and the methods using a long sequence of monocular images [Ullm84, Broi86a, Broi86b] Methods based on both the parallel projection and the perspective transformation, and on both 3 D points and 3 D lines as features have been ....

Roach, J.W., Aggarwal, J.K., "Determining the Movement of Objects from a Sequence of Images," IEEE Trans. on Pattern Analysis and Machine Vision, vol. 2, no. 6, 1980, pp. 554--562.

....the scene undergoing motion in successive frames, and from these descriptions deduce the motion parameters [8, 9] A feature based approach is, in general, a three stage process: 1. Features are detected in two consecutive frames. Typical features in use include lines and edges [10] corner points [11], line intersections, points of local maximum curvature on contour lines [12] etc. 2. According to some searching scheme, features in one frame are matched with the features of the adjacent frame. 3. The spatial mapping between the corresponding features is calculated, hence providing an estimate ....

J. W. Roach and J. K. Aggarwal, "Determining the Movement of Objects From a Sequence of Images," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 2, pp. 554--562, Nov. 1980.

....that the correspondences are already known. Point correspondences were used by Ullman [138] who formulated a method for the recovery of the 3 D motion and position of 4 points with known correspondences, given three projections of those points. A similar method was proposed by Roach and Aggarwal [119] using 8 points in two consecutive frames under the assumption of a static scene and a moving camera. Their approach involved the solution of a set of 20 equations in 2.2 A Review of Motion Estimation Techniques 27 27 unknowns, where assumptions were made about the projection geometry such that ....

J. W. Roach and J. K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(6):554--562, November 1980.

....and motion of the body [140] The parallel projection is adequate only in some situations because most realworld applications require the use of the perspective projection. The use of the perspective projection is much more complex than the use of the parallel one. Roach and Aggarwal [114] [115] were among the first to study the computation of structure and motion from images using imaging transformation. The relationship between the three dimensional coordinates of a point (X; Y; Z) and its image plane coordinates (x; y) is x = F a 11 (X Gamma X 0 ) a 12 (Y Gamma Y 0 ) a 13 (Z ....

....system. 14 Roach and Aggarwal made interesting considerations. Two views of six points or three views of four points are needed to provide an overdetermined set of equations when the images are noisy. To recover parameters of equations in Equation (8) five points in two views are needed [114] [115]. They related the number of points and the number of equations available for the solution of 3 D coordinates and motion parameters. First, they had 27 unknowns and 20 equations. Finally, choosing some parameters a priori they had 18 projection equations and 18 unknowns. Since the equations of the ....

Roach, J.W., Aggarwal, J.K., "Determining the Movement of Objects from a Sequence of Images", IEEE Transactions on Pattern Analysis and Machine Vision, vol. 2, no.6, 1980, pp. 554-562.

....of point or feature correspondences between two or more perspective views for determining structure 1 This research has been supported by the Defense Advanced Reseach Projects Agency under Army ETL contract DACA7689 C 0017 and the National Science Foundation CER grant number NSF CDA 8922572. [5, 6, 7, 8, 9, 10, 11, 12, 13]. These correspondence based approaches take advantage of the image displacements induced by egomotion. Most such methods match a large number of points or features in two temporally separated images and quantitatively measure the image displacements. A consistent set of motion parameters is ....

J. W. Roach and J. K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 554--562, November 1980.

No context found.

J.W. Roach and J.K. Aggarwall. Determining the movement of objects from a sequence of images. IEEE Transactions on Patterm Analysis and Machine Intelligence, 2(6):554-562, 1980.

No context found.

J. W. Roach and J. K. Aggarwal. Determining the move- ment of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 554-562, November 1980.

No context found.

J. W. Roach and J. K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Patter Analysis and Machine Intelligence, pages 554 562 November 1980.

No context found.

J.Roach and J.K.Aggarwal. Determining the movement of objects from a sequence of images. IEEE Trans. Pattern Anal. Machine Intell., 2(6):554--562, November 1980. 20

No context found.

J. W. Roach and J. K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(6):554--562, 1980.

No context found.

J. W. Roach and J. K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(6):554--562, 1980.

No context found.

J. W. Roach and J. K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2(6), November 1980.

No context found.

J. W. Roach and J. K. Aggarwal. Determining the movement of objects from a sequence of images. IEEE Transactions on Pattern Analysis and Machine Intelligence, pages 554--562, November 1980.

No context found.

Roach, J. and J. Aggarwal (1980). Determining the movement of objects from a sequence of images. IEEE Trans. PAMI 2 (6), 554--561.

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