N.Sumpter, R.D. Boyle, and R.D. Tillett. Modelling collective animal behaviour using extended point distribution models. pages 242--251, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....When combining information for statistical analysis via PCA it is important that constituent features (V H V C V S ) are scaled to ensure that any particular feature does not dominate the principal axis . This can be done by minimising the eigen entrophy as proposed by Sumpter, Boyle and Tillett [11]. However as all three components exist within the same co ordinate frame and are directly linked, and thus such scaling is unnecessary. Figure 1 (a) Position of head and hands V H (b) Body Contour V C (c) Corresponding 3D model V S 2.1 Hand and Head Position Estimation Colour information is ....

Sumpter, N., Boyle, R. D., Tillett, R. D., Modelling Collective Animal Behaviour using Extended Point Distribution Models, In Proc. British Machine Vision Conference, Vol 1, pp242-251, 1997.

....When combining information for statistical analysis via PCA, it is important that constituent features (V H ,V C ,V S ) are scaled to ensure that any particular feature does not dominate the principal axis of S. This can be done by minimising the eigen entropy as proposed by Sumpter et al. [11]. However, as all three components exist within the same co ordinate frame and are directly linked, such scaling is unnecessary. 3.1. Hand and head position estimation Colour information is an important attribute within an image, which is often discarded. McKenna, Gong and Raja have demonstrated ....

N. Sumpter, R.D. Boyle, R.D. Tillett, Modelling collective animal behaviour using extended point distribution models, in: A.F. Clark (Ed.), British Machine Vision Conference 1997, British Machine Vision Association, Essex, UK, 1997, pp. 242--251.

....the novelty of these experiments is in the animal robot interaction and not in the image processing. A description of the system and techniques used is provided here for completeness. Neil Sumpter has produced novel and interesting machine vision research in a separate stream of the project; see [Sumpter et al. 1997, Sumpter and Bulpitt, 1998, Sumpter et al. 1998] 3.5.1 Requirements The vision system is required to locate the position and heading of the robot and ock within the arena, feeding this back to the controller at as high a frequency as possible. At the start of development, the goal update ....

....constraint to the rest of the system; it would have to work without knowing the positions of individual birds, but only with a centre position, size and shape. Neil Sumpter developed a tracker which found the ock shape using a combination of image subtraction and hysteresis region growing [Sumpter et al. 1997]. This worked well on test sequences, but unfortunately proved to be slow and unreliable when integrated with the rest of the robot system. A good feature of this tracker was that if it lost the ock it could be manually reset by the operator using the mouse pointer on a window showing the arena ....

Sumpter, N., Boyle, R., and Tillett, R. (1997). Modelling collective animal behaviour using extended point distribution models. In Proc. British Machine Vision Conference, pages 242-251.

....the novelty of these experiments is in the animal robot interaction and not in the image processing. A description of the system and techniques used is provided here for completeness. Neil Sumpter has produced novel and interesting machine vision research in a separate stream of the project; see [Sumpter et al. 1997, Sumpter and Bulpitt, 1998, Sumpter et al. 1998] Figure 3.10: Example images from the overhead camera input to the vision system. 3.5.1 Requirements The vision system is required to locate the position and heading of the robot and ock within the arena, feeding this back to the controller ....

....to work without knowing the positions of individual birds, but only with a centre position, size and shape. Figure 3. 11: The tracker correctly locates the ock Neil Sumpter developed a tracker which found the ock shape using a combination of image subtraction and hysteresis region growing [Sumpter et al. 1997]. This worked well on test sequences, but unfortunately proved to be slow and unreliable when integrated with the rest of the robot system. A good feature of this tracker was that if it lost the ock it could be manually reset by the operator using the mouse pointer on a window showing the arena ....

Sumpter, N., Boyle, R., and Tillett, R. (1997). Modelling collective animal behaviour using extended point distribution models. In Proc. British Machine Vision Conference, pages 242-251.

....conversion of landmark p into polar coordinates, a constant scale factor R p is used. R p is calculated as being the average distance of landmark p from its centre landmark over the training set; this ensures that every training example is scaled by the same factor. Recent work by Sumpter et al. [70] describes a more general approach to finding suitable linear scalings for non comparable parameters prior to performing a PCA. Therein, the information content of the model is maximised via the calculation of eigen entropy. 3.2.2 Coping with the Angle Discontinuity Problem Another problem ....

N. Sumpter, R.D. Boyle, and R.D. Tillett. Modelling collective animal behaviour using extended Point Distribution Models. In Proc. BMVC, Colchester, UK, 1997. BMVA Press.

....To obtain comparable variance, we scale each shape vector x i by the maximum norm. A more considered approach to achieve comparable 2 The same training images are used for constructing the eigenspace. variance between the appearance (image) and shape vectors in eigenspace can be adopted [7]. It is worth pointing out though that better results were actually achieved with modelling the appearance and shape vectors in independent eigenspaces rather than with the concatenated eigenspace. Once the landmarks have been learned from the training set, we can recall the landmarks during the ....

N. Sumpter, R. Boyle, and R. Tillett, "Modelling collective animal behaviour using extended point distribution models," in BMVC, Colchester, September 1997, pp. 242--251.

....birds due to the poor resolution of the image sequences. The flock and robot move within a constrained environment an arena, eight metres in diameter. The flock is identified using background subtraction, and the outline of the resulting mass is used as the basis for a PDM, as described in [15]. Typically five shape parameters of the PDM are sufficient to describe at least 90 of the flock shape variation. The robot is segmented using a high contrast black and white motif placed on top of the robot, and this is located within the image frame. This motif allows the robot s orientation to ....

N. Sumpter, R.D. Boyle, and R.D. Tillett. Modelling collective animal behaviour using extended point distribution models. In Proc. British Machine Vision Conference, pages 242--251, 1997. 12

....in which the groups of animals to be studied are active, sometimes for protracted periods. One application is a laboratory maze in which time and route of an animal are of interest, and this problem has been successfully solved [12] another, more challenging, is the Silsoe Robotic Sheepdog [15] which requires a group of animals to be tracked at high frame rate. In a more restricted domain, success has been exhibited in the use of visual tracking to assist in automatic milking of dairy cattle [5] We consider here a different problem which has particular difficulties associated with it ....

N. Sumpter, R. D. Boyle, and R. D. Tillet. Modelling collective animal behaviour using extended point distribution models. In A. F. Clark, editor, British Machine Vision Conference, 1:242--251, Sept. 1997. 20

....the animal directly in the development process. The RSP is a collaborative, multidisciplinary project covering robot building, machine vision, behavioural modelling and ethological experiments. For an overview of the project as a whole, see [VHS97] Work on machine vision has been reported as [SBT97] and [SB98] The sheepdog s gather and fetch task was chosen because of its familiarity and the strong interaction between the dog, shepherd and flock animals. Using ducks instead of sheep allows us to experiment on a conveniently small scale, in a controlled indoor environment. Duck flocking ....

N. Sumpter, R.D. Boyle, and R.D. Tillett. Modelling collective animal behaviour using extended point distribution models. In Proceedings of the British Machine Vision Conference, 1997.

....to the poor resolution of the image sequences (see Figure 3) The flock and robot move within a constrained environment an arena, eight metres in diameter. The flock is identified using background subtraction, and the outline of the resulting mass is used as the basis for a PDM, as described in [15]. Typically five shape parameters of the PDM are sufficient to describe at least ### of the flock shape variation. The robot is segmented using a high contrast black and white motif placed on top of the robot, and this is located within the image frame. This motif allows the robot s orientation to ....

N. Sumpter, R.D. Boyle, and R.D. Tillett. Modelling collective animal behaviour using extended point distribution models. In Proc. British Machine Vision Conference, pages 242--251, 1997.

....building, machine vision, behavioural modeling and ethological experiments. An early overview of the project as a whole is given in [11] and results from our robot experiments have been presented at several conferences [12 14] Our work on image processing and machine vision has been reported as [8 10]. 1 now at the Robotics Research Laboratory, University Of Southern California, Los Angeles, CA 90089 0781, USA. Email: vaughan robotics.usc.edu Preprint submitted to Elsevier Preprint 16 July Fig. 1. A young sheepdog in training with a group of ducks. Photographed by the rst author in ....

N. Sumpter, R.D. Boyle, and R.D. Tillett. Modelling collective animal behaviour using extended point distribution models. In Proceedings of the British Machine Vision Conference, 1997.

....prompted by two independent applications. 3. 1 The Robotic Sheepdog project The robotic sheepdog project is constructing an autonomous robot that has the intelligence to herd living creatures (for the purposes of demonstration, ducks) The project and its motivation are described elsewhere [7, 8]. The robot receives its information about its own and the ducks parameters (position and velocity) via a camera monitoring the scene. The ducks behave in a flocking manner, with the coarse shape of the flock being a good indicator of their behaviour it is thus natural to model the flock ....

N Sumpter, R D Boyle, and R Tillet. Modelling collective animal behaviour using extended PDMs. In A Clark, editor, Proceedings of the British Machine Vision Conference, 1997, Colchester, UK, 1997. To appear.

No context found.

N.Sumpter, R.D. Boyle, and R.D. Tillett. Modelling collective animal behaviour using extended point distribution models. pages 242--251, 1997.

No context found.

N. Sumpter, R. D. Boyle and R. D. Tillett, Modelling Collective Animal Behaviour Using Extended Point Distribution Models, British Machine Vision Conference BMVC97, 1997, 242251.

No context found.

N. Sumpter, R. D. Boyle and R. D. Tillett, Modelling Collective Animal Behaviour Using Extended Point Distribution Models, British Machine Vision Conference BMVC97, 1997, 242-251.

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