C. Ferm uller and Y. Aloimonos. Observability of 3D motion. International Journal of Computer Vision, 37:43--63, 2000.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....which are independent of the structure of the scene in view, encodes the underlying 3D motion parameters, enabling stabilization through the interpolation of the UGV s intended motion in the temporal evolution of the measured motion. Figure 14 shows experiments using the approach developed in [4] with real data collected from the vehicle. A recent technical development related to the perception of the UGV s environment is the concept of iso distortion surfaces, a framework for studying the relationship between the computation of 3D motion and depth from a sequence of images [3] The ....

....the countryside, thus undergoing continuously changing rigid motion. a) shows one frame of the sequence with the normal flow field overlaid. b) d) and (f) show the positive (light color) and negative (dark color) vectors of the longitudinal patterns corresponding to the x , y and z axes (see [4]) c) e) and (g) show the corresponding fitted patterns. h) shows, superimposed on the image, the boundaries of the patterns whose intersections provide the FOE and the AOR (the point where the rotation axis pierces the image plane) i) Measurements are not everywhere available (strong ....

C. Ferm uller and Y. Aloimonos. Qualitative egomotion. International Journal of Computer Vision, 15:7--29, 1995.

....for a full field of view camera. For a planar camera, which by construction has a limited field of view, the problem is nonlinear and ill posed. If, however, the field of view approaches 360 # , that is, the pencil of light rays is cut by a sphere, then the problem becomes well posed and stable [7], although still nonlinear. The basic understanding of the influence of the field of view has attracted a few investigators over the years. In this paper we will not study this question in more detail and only refer to the literature for more information [3, 6, 10] Thus, in conclusion, there are ....

C. Ferm uller and Y. Aloimonos. Observability of 3D motion. International Journal of Computer Vision, 37:43--63, 2000.

....only a few investigators over the years [3, 8, 13, 15] Having in mind the design of an optimal sensor, we are interested in how the stability of the estimation of motion changes with the field of view. In particular, we have compared the planar small field of view camera with a spherical camera [10]. Since motion estimation amounts to solving some minimization function, we analyzed the minimization functions corresponding to the different constraints described above. To be more precise, we performed a geometric statistical analysis; we compared the expected values of the different functions ....

C. Ferm uller and Y. Aloimonos. Observability of 3D motion. International Journal of Computer Vision, 37:43--63, 2000.

....a few investigators over the years [3, 11, 12, 19, 21] Having in mind the design of an optimal sensor, we are interested in how the stability of the estimation of motion changes with the field of view. In particular, we have compared the planar small field of view camera with a spherical camera [14]. Since motion estimation amounts to solving some minimization function, we analyzed the minimization functions corresponding to the different constraints described above. To be more precise, we performed a geometric statistical analysis; we compared the expected values of the different functions ....

....sequence we worked on. Often we found the valley to be rather wide, but in many cases it was close in position to the predicted one. If we increase the field of view of a sensor to 360 # , proofs in the literature show that we should be able to accurately recover 3D motion and subsequently shape [14]. Catadioptric sensors can provide the field of view, but the low and non uniform resolution makes it difficult to perform the signal processing that is necessary to recover shape models. Thus, we built the Argus eye [5] a construction consisting of six cameras pointing outward (as in Fig. 2) ....

C. Ferm uller and Y. Aloimonos. Observability of 3D motion. International Journal of Computer Vision, 37:43--63, 2000.

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