S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....undergoing special motion, most notably pure transla tion [10] Affine structure is distorted away from the true Euclidean structure of the scene by unknown scaling and skew along three axes. Despite these shortcomings, affine structure has been proven useful for tasks as var ied as tracking [3], navigation [2] symmetry detection [13] and object recognition [16] For most of these tasks, however, can the knowledge of the full Euclidean structure be of considerable benefit. To upgrade an affine structure to Euclidean means providing additional information allowing the recovery of the ....

S. FairIcy, I. Reid, and D. Murray. Transfer of fixa- tion for an active stereo platform via aftinc structure recovery. In Proc. ICCV'95 [5], pages 1100 1105.

....frame) This is a way of calculating a stable fixation point on an object from whatever corners are detected in each frame. Since corner features can also be matched between images obtained from two cameras, this is also one of the pieces of work so far which has been fully implemented in stereo [28]. The large size of the first Yorick was chosen to give good performance when recovering the depth of scene features using stereo in the kind of scenes likely to be encountered (see Section 4.2.2 for a discussion of this) Geared (as opposed to direct drive) motors were used throughout, as it was ....

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105, 1995.

....transformation from each set of two three consecutive images, using the measurement matrix factorization algorithm described in [18] The motion parameters were then used to transfer a chosen fixation point from the previous image(s) to the latest. This algorithm was extended to stereo cameras in [5]. While we have a simpler problem, with a fixed camera and a 2D scene representation, similar principles apply. The fixation point here refers to the point chosen as representative of the vehicle for the purpose of estimating its position, and hence its range. 2D fixation transfer may be ....

....data and combines data from multiple images in a statistically near optimal manner. z z M t M t r z t 2D affine transformations planar scene X image planes =M X =M z X t Figure 1: Stereo affine 2D projection from planar scene to image planes. In comparison with the previous system [5] we have improved the robustness of the feature tracker to the extent that outliers are much less likely to occur, because the set of feature matches is forced to be globally consistent [19, 10] By integrating the VSDF reconstruction algorithm we now have a stable method of transferring the ....

[Article contains additional citation context not shown here]

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.

....transformation from each set of two three consecutive images, using the measurement matrix factorization algorithm described in [26] The motion parameters were then used to transfer a chosen fixation point from the previous image(s) to the latest. This algorithm was extended to stereo cameras in [6]. While we have a simpler problem, with a fixed camera and a 2D scene representation, similar principles apply. The fixation point here refers to the point chosen as representative of the vehicle for the purpose of estimating its position, and hence its range. 2D fixation transfer may be ....

....Filter (VSDF) algorithm [16] specialized to 2D affine scene reconstruction. The VSDF is a general algorithm for visual reconstruction that deals naturally with fragmentary data and combines data from multiple images in a statistically near optimal manner. In comparison with the previous system [6] we have improved the robustness of the feature tracker to the extent that outliers are much less likely to occur, because the set of feature matches is forced to be globally consistent [27, 12] By integrating the VSDF reconstruction algorithm we now have a stable method of transferring the ....

[Article contains additional citation context not shown here]

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.

....handled using a well understood combination of odometry and estimation [11, 12] In a purely sensor based approach, prediction must be performed solely on image information. This problem is closely related to the image transfer problem discussed in the area of projective geometry applied to vision [2, 5, 8, 15]. In this paper, we describe specializations of image transfer methods for the two commonly used image features, point features and line features, for a mobile system operating on a planar surface. 2 Problem Formulation In an image or appearance based approach to navigation, the robot can be ....

.... projective invariance, methods for performing image transfer have been developed for point and line features under a variety of assumptions about the configurations of the features, the camera model (orthographic projection, affine, perspective, projective) and availability of camera calibration [2, 5, 8]. For point features, Barrett et al. show that linear methods can be used to transfer points from two map images to a third image if eight additional points are observed in all three images [2] Hartley s methods for projective reconstruction of lines can be used for line transfer, and it ....

S. Fairley, I. Reid, and D. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Int. Conf. Computer Vision, pages 1100-- 1105, 1995.

....the fixation point is tracked using its structure; all the scene points are tracked by individual constant velocity Kalman filters. The use of the fixation point and the features structure overcomes the temporal instability of corner detection. Their work was later extended to use a stereo head [30]. A hybrid correlation and feature based tracking system has been demonstrated recently by McLauchlan [66] which is similar to [89] in that only planar structure is recovered 5 . In a recent work [62] Manku et al. demonstrate a very similar tracker which switches between tracking feature based ....

....of techniques used in ESA RADAR tracking to the computer vision community. The tracker differs from RADAR, and indeed most visual, trackers in being more interested in the motion of ensembles of features, rather than the life of any particular feature; in this respect the most similar work is [30, 62, 89] which develop other trackers intended for motion recovery. The tracker described here manages a collection of tracks, and uses them as sensors with which to sample the image. The tracker uses the data provided by the tracks to determine the egomotion, and to segment any independently moving ....

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.

....enough to cover all the environment. Therefore any potential intrusion is more easily detected since no scanning is required. Systems based on active cameras usually employ longer focal length cameras and therefore provide better resolution images. Some of the systems are active and binocular [20]. These enable the recovery of 3D trajectories by tracking stereoscopically. Proprioceptive data from camera platform can be used to recover depth by triangulation. Trajectories in 3D can also be recovered monocularly by imposing the scene constraint that motion occurs in a plane, tipically the ....

Fairley, S., Reid, I., Murray, D.: Transfer the Fixation for an Active Stereo Platform via Affine Structure Recovery, In Proc. ICCV95-I, (1995), 100-105.

No context found.

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.

No context found.

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.

.... in the scene both capture saccades when the resultant projected image motion is largely translational, and panic saccades when the image motion is predominantly divergent, indicating looming [6] The more purposive behaviours include tracking for surveillance using both feature clusters [7, 8] and optical flow [9] line following [10] and fixation for navigation [11] This paper describes the vision system designed to support our research into active processes. The need to implement a variety of visual behaviours whose exact nature was unknown a priori precluded the design of a fixed ....

....up to a 3D global affine transformation, and sufficient to compute images from arbitrary novel viewpoints. This last process, known as transfer, is used to give a stable object based position for fixation. The theory of monocular affine transfer was detailed in [7] and that for stereo transfer in [8]. For brevity and clarity here we will outline the method using the minimal number of point correspondences, namely four points in three images. 4.1 Theory review Where scene relief is small in comparison with depth, it is valid to assume that images are formed under an affine camera projection ....

[Article contains additional citation context not shown here]

Fairley, S.M., Reid, I.D. & Murray, I.D. (1995) Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int. Conf. on Computer Vision, Boston, pp 1100--1105. IEEE Computer Society Press.

....method when corners fall off or enter the sides of image during zooming in and out, and appear or disappear because of scale effects. Recent work has demonstrated the active tracking of clusters of corner features using the method of affine transfer, both monocularly [11, 12] and stereoscopically [4]. The method finesses the difficulty caused by the temporal instability of a single corner by, in the simplest case of 3D transfer, replacing the requirement to track one corner through all image frames by the less demanding requirement to track any four points across three successive frames. ....

....giving angular fields of view from 50 10 ffi . The motorized zoom was operated open loop, whilst viewing a moving object. We note that for the 3D method to function, sufficiently different views must be obtained so that the structure can be recovered. Whilst this is achieved easily using stereo [4], using a single fixed camera requires object rotation (or camera rotation if visually servoing) There is of course no such requirement in the 2D algorithm used in the real time work. 3.2 Results We show results from an experiment where the subject rotates his head while the camera zooms in. ....

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.

No context found.

S.M. Fairley, I.D. Reid, and D.W. Murray. Transfer of fixation for an active stereo platform via affine structure recovery. In Proc. 5th Int'l Conf. on Computer Vision, Boston, pages 1100--1105. IEEE Computer Society Press, 1995.

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