Alicia Abella and John R. Kender. Qualitatively describing objects using spatial prepositions. In Eleventh National Conference on Artificial Intelligence, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....relations, such as near, far, and along, can be context dependent. What is far is one context may be near in another. Recently, there has been much work in the development of quantitative methods for evaluating the applicability of various topological relations within a given context [1, 29, 56, 65]. Directional spatial relations, such as left of and right of, require a frame of reference. Depending on the needs of the application, the frame of reference may be extrinsic (predesignated fixed frame of reference) intrinsic (based on the orientation of an object of interest, such as the ....

....inside the rectangular shape abstraction chosen for the triangular object, but does not actually touch the triangular object. If spatial relations are proven only with respect to the shape abstraction chosen for an object, it is possible to conclude the circular object 30 Figure 2. 2: Figure from [1] used by Abella and Kender to illustrate their model of near . is overlapped with or enclosed within the triangular object even though it is not true of the actual spatial scene. What Guarantees Can Shape Abstractions Provide Any spatial representation which relies on the use of shape ....

[Article contains additional citation context not shown here]

A. Abella and J. Kender. Qualitatively describing objects using spatial prepositions. In Proc. 11th National Conf. on Artificial Intelligence, Cambridge, MA, 1993. AAAI/MIT Press.

....this proximity measure has many advantages. It is easily constructed, relatively efficient, and captures the intuition that large elements (such as a large rectangles or long polylines) relate visually to many other elements. Similar approaches to rating nearness have been used successfully (e.g. Abella and Kender, 1993). Running the visual operations. Once the LLRD determines which elements are proximate, it looks for other visual relations between proximate elements, using a visual operation library. Each visual operation detects specific visual relations that are part of early vision. All visual operations ....

Abella, A., & Kender, J. R. (1993). Qualitatively describing objects using spatial prepositions, Proc. Eleventh National Conference on Artificial Intelligence.

....shape representations. Many have expressed the viewpoint that it is difficult to formulate a general purpose qualitative spatial model since it would require knowledge of exact shapes of objects. However, recent work in qualitative spatial reasoning has focused on approximating shapes of objects. AK93] presents a method for fuzzifying qualitative spatial prepositions such as near and along . The proposed method works for irregular shaped objects and uses a shape representation based on the center and elongation axes of the object. Raj94] discusses a model for integrating qualitative ....

Alicia Abella and John R. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In Proceedings of the Eleventh National Conference on Artificial Intelligence (AAAI-93), pages 536--540, Washington, DC, 1993.

.... They propose that these kinds of relations depend mainly on boundedness, surface, or volumetric nature of an object and its axial structure (see also Herskovits, 1986; Talmy, 1983) Hence, the process of establishing spatial relations involves only the essential shape properties of objects (cf. Abella Kender, 1993; Mukerjee Joe, 1990) Spatial relations are fuzzy (cf. Freeman, 1975) i.e. their applicability region s cannot be defined by sharp boundaries. Attempts by speakers to restrict this vagueness can be seen in the use of linguistic hedges like exactly in front of 1 or just behind and has ....

....try to keep the computational costs of spatial relations as low as possible so as to provide efficient reasoning calculus. The range stretches from simple point representations (e.g. Freksa, 1992) towards more sophisticated convex hulls, collision parallelograms, bounding boxes and circles (e.g. Abella Kender, 1993; Mukerjee Joe, 1990) The focus changes if we consider the domains of vision and natural language processing. The models developed in these areas have the requirement of being close to the human processing of spatial relationships in complex (real world) environments. Most of the algorithms ....

[Article contains additional citation context not shown here]

Abella, A., & Kender, J. R. (1993). Qualitatively describing objects using spatial prepositions. Proc. of AAAI-93, Washington, DC, 536--540.

....to specific subsets in advance. 2 A Computational Model of Spatial Relations Increasingly sophisticated approaches to the computation of spatial relations have been developed in the last couple of decades (e.g. Andr e et al. 1988; Egenhofer and Herring, 1990; Mukerjee and Joe, 1990; Abella and Kender, 1993; Rajagopalan, 1994; Schirra and Stopp, 1993; Olivier et al. 1994 ] In [ Gapp, 1994a ] the CSR 3D 1 system, a model for the computation of topological and projective relations (cf. Talmy, 1983; Herskovits, 1986 ] in 2D and 3D 1 The CSR 3D system (Computation of Spatial Relations in ....

A. Abella and J. R. Kender. Qualitatively describing objects using spatial prepositions. In Proc. of AAAI-93, pages 536--540, Washington, DC, 1993.

....reasoning about the properties of two and three dimensional objects, such as intersection and containment. In previous work, two and three dimensional spatial representations have been developed based on the use of convex hulls [Cui, Cohn, and Randell 92] inertia tensorbased bounding rectangles, Abella and Kender 93] bounding rectangles based on intrinsic front directions [Retz Schmidt 88, Mukerjee and Joe 90] and cuboids [Maass 94] as shape abstraction methods. These theories do not provide a qualitative model for the shape abstractions, and instead assume that they are provided as part of the input ....

Abella, A. and J. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In: Proceedings Eleventh National Conference on Artificial Intelligence (AAAI-93), 1993.

....but not rotate, and the bounding circle for problems where objects may both translate and rotate. The two abstractions are necessary to guarantee that the actual region of space occupied by an object is contained within the region of space defined by the abstraction used. Abella and Kender [1] describe a shape abstraction method based on rectangular bounding boxes that does not provide this guarantee. Thus, in their approach, it is possible to conclude, for example, that two objects are not near each other when they are actually intersecting To illustrate why rectangular bounding ....

A. Abella and J. Kender. Qualitatively describing objects using spatial prepositions. In Proc. 11th National Conf. on Artificial Intelligence, Cambridge, MA, 1993. AAAI/MIT Press.

....category words like collie or spaniel) are treated as modifiers of the basic concept. Thus, the dynamic predicate for Collie would become a move in the appearance space for dogs to a Collie model. Words associated 6 There is a large body of relevant work (Herskovits 1986, Herskovits 1985, Abella and Kender 1993, Andre et al. 1987, Olivier and Tsujii 1994, Yamada 1995, Yamada et al. 1992, Gapp and Maa 1994, Herzog 1995, Herzog and Wazinski 1994, Eschenbach et al. 1997, Vandeloise 1991) 7 This is the chief technical point that distinguishes the sense of dynamic used here from work in dynamic semantics ....

Abella, A., and J. R. Kender. 1993. Qualitatively describing objects using spatial prepositions.

....Obj a and the other objects is less than the the sum of the distances between any Obj b j and the other objects, then Obj a is the subject of the surrounded by predicate. The primitive between can be viewed as a ternary specialization of surrounded by. Near is often interpreted qualitatively [ Abella and Kender, 1993 ] In our domain, an object is near another object if there is no other object nearer the latter. 1 Verifying near in our domain entails computing relationships between all the objects. If we defined the set M = fAllObjects Gamma fObj a ; Obj b gg, then near(Obj a ; Obj b ; M ) TRUE if the ....

Alicia Abella and John R. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In Proceedings of AAAI-93, pages 536--540, 1993.

....for object pairs: The principles of the computation are demonstrated for 2D objects in 2D space. An object specific partitioning of 2D space into 12 acceptance volumes is chosen. As an example the generation of the relation behind is shown for the objects IO and RO w.r.t. the reference frame ref. (Abella Kender, 1993) as abstractions. Although this is a rough shape abstraction, it is sufficient to investigate the influence of the objects shape and extension to the use of projective prepositions. It must be pointed out that the principal computational structure of our approach is not bound to this sort of ....

Abella, A. & Kender, J. (1993). Qualitatively Describing Objects Using Spatial Prepositions. In Proc. of AAAI-93, pp.

.... local coordinate system is then mapped to a fixed evaluation function (e Gamma vh d 1 ) The idea of defining a local reference system was taken up in Wazinski s LOC SYS system [Wazinski 91] which provides 2D localizations for the knowledge based presentation system WIP [Wahlster et al. 93] Abella Kender 93] presented a framework for a system that describes qualitatively 2D objects using spatial prepositions. They define a preposition by a set of inequalities. A fuzzification procedure using Monte Carlo Simulation was used to account for the vagueness of prepositions. Some prerequisites appear to be ....

A. Abella and J. R. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In: Proc. of AAAI-93, pp. 536--540, Washington, DC, 1993.

....point for solving spatial reasoning problems is a representation which identifies the location, extent, and orientation of objects and spatial relations (topological and directional) between them. Several formalizations of topological relations (e.g. connected, equal, overlapped) are available [Abella and Kender 93, Cui, et. al 92, Egenhofer and Al Taha 92, Hadzilacos and Tryfona 92] Some of these include descriptions of how the topological relationship between two objects may change over time due to motion or changes in size or shape [Cui, et al. 92, Egenhofer and Al Taha 92] Formalizations of ....

....for which abstractions are used is the shape (spatial extent) of an object. Rectangular bounding boxes are a commonly used abstraction method, but the criteria for drawing the boxes may vary. Retz Schmidt [1988] and Mukerjee and Joe [1990] draw bounding boxes around the front of an object. Abella and Kender [1993] draw a bounding box based on a mathematical abstraction, inertia tensors. Cui, Cohn, and Randell [1992] use the convex hull as the abstraction method. We feel that the choice of the abstraction method for shapes is fundamental. The chosen method can affect the way in which spatial relations are ....

[Article contains additional citation context not shown here]

Abella, A. and J. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In: AAAI-93, Washington, D.C., 1993.

....recognition and 3 D scene reconstruction. The goal is to extract qualitative spatial information from real images. The geometric object models are more detailed than needed to determine spatial relations. Therefore, the objects are approximated by boxes. Similar to Abella and Kender s approach, Abella Kender 93 ] for the formulation of the 2D relation aligned, an object s bounding box BO is given as the bounding cuboid collinear to the object s principal axes. 2.2 Model based 3 D reconstruction The object recognition process starts as a knowledge based search in regions of color segmented images ....

.... in the reference object, the speaker or either some third object or person a reference frame is called intrinsic, deictic or extrinsic, respectively [ Retz Schmidt 88 ] Instead of directly calculating prepositions from 3 D descriptions of two objects and a given reference frame [ Gapp 94; Abella Kender 93 ] we first construct an object specific reference independent spatial representation of the object constellation. This is based on acceptance relations that are induced by acceptance volumes partitioning the 3 D space in an object specific way. In a second step, reference frames are used to ....

[Article contains additional citation context not shown here]

A. Abella, J.R. Kender, Qualitatively Describing Objects Using Spatial Prepositions. Proc. of AAAI-93, 1993, pp. 536--540.

....a combined semantic description of the stone as a calyceal stone. 2. The Semantic Representation of Spatial Prepositions At the heart of the computational understanding of spatial prepositions is their semantic representation, which captures the vital properties su#cient for their succinct use [1]. Objects are treated as blobs , and most of the properties can be encoded using simple geometric properties such as alignment and distance. Figure 3. An x ray image of the urinary system depicting a kidney stone in the right kidney, highlighted by the circle. 2.1. Notations and Definitions ....

.... it is important to fuzzify the ideal set, as defined through fuzzy set theory [11] In a fuzzy set, members of the set are assigned a membership grade value that indicates to what degree they belong to the set; here the fuzzified ideal set is defined through a membership function p i (a, b) [0, 1]. As a special case, define p i (a, b) p i (a, b) 0 or 1. 2.2. The Semantic Representation The semantic representation of prepositions is based here on object area, centers of mass, and elongation properties calculated from their moments. Each preposition is then defined through a set of ....

[Article contains additional citation context not shown here]

Alicia Abella and John R. Kender. Qualitatively describing objects using spatial prepositions. In Eleventh National Conference on Artificial Intelligence, 1993.

....is lost (error detection) but because of the particular local environment, cannot continue (no error correction) Park and Kender 1994] 6. 4 Natural Language Description of Images We have significant extensions to our previous work on qualitative description of scenes using spatial propositions [Abella and Kender 1993] . In the past year, we have built and demonstrated a system that integrates image processing and natural language processing for tasks that involve communicating visual spatial information [Abella and Kender 1994] The system determines information about the spatial relationship of objects in ....

A. Abella and J. R. Kender. Qualitatively describing objects using spatial prepositions. Proc. of AAAI, July 1993.

....yielded the preposition calibration data shown in figure 1. The preposition calibration data is then fitted to the parameters of the semantic representation. This calibration data allows the system to capture some of the vagueness associated with prepositions by a process we call fuzzification [ Abella and Kender 1993 ] It is the responsibility of the locative expression generator to limit the number of prepositions ultimately used to describe an object pair. The semantic representation module gives the LE generator information about how all the preposi tions are satisfied by all the object pairs. The LE ....

.... the spatial relationship of the object pair in the form of an English sentence [ Elhadad 1993 ] 3 The LE Generator In this paper we will focus on the methodologies employed by the LE generator [ Abella and Kender 1994 ] The description of the semantic representation module can be found in [ Abella and Kender 1993 ] and [ Abella forthcoming ] We have included a semantic representation of the following prepositions: P = fnear ; far ; above; below ; aligned ; next ; inside; left ; right ; betweeng We also found that a description can be enhanced by the use of object features such as size, and ....

Alicia Abella and John R. Kender. Qualitatively describing objects using spatial prepositions. Eleventh National Conference on Artificial Intelligence, 1993.

No context found.

Alicia Abella and John R. Kender. Qualitatively describing objects using spatial prepositions. In Proceedings of the Eleventh National Conference of Artificial Intelligence, 1993.

....prepositions is their semantic representation. A representation is necessary if the system is to be able to assign a preposition to an object pair. The semantic representation, to be discussed in this chapter, captures the vital properties sufficient for a succinct use of the chosen prepositions [Abella and Kender, 1993]. The spatial prepositions can be encoded fairly concisely because objects are treated as blobs and because most of the properties characterized by these prepositions can be encoded using geometric properties such as alignment and distance. The following sections will provide the details of the ....

Alicia Abella and John R. Kender. Qualitatively describing objects using spatial prepositions. In Eleventh National Conference on Artificial Intelligence, 1993.

No context found.

A. Abella and J. R. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In: Proc. of AAAI-93, pp. 536--540, Washington, DC, 1993.

No context found.

Abella, A. and J. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In: Proceedings Eleventh National Conference on Artificial Intelligence (AAAI-93), 1993.

No context found.

Abella, A. and J. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In: AAAI-93, Washington, DC, 1993.

No context found.

A. Abella and J. R. Kender. Qualitatively Describing Objects Using Spatial Prepositions. In: Proc. of AAAI-93, pp. 536--540, Washington, DC, 1993.

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