F. Stein and G. Medioni. Structural indexing: Efficient 3d object recognition. Transaction on Pattern Analysis and machine Vision (PAMI), 14:125 -- 145, February 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the efficiency of object recognition. Our application domain uses the recognition of 3D polyhedral objects via in terpretation tree search [5] This approach served as a basis for several systems which adjusted some aspects, e.g. ordering of features [1] or creating derived features [4] [10]. We show how to obtain attributed features with context and ordering information (section 2) Furthermore, we use this information to formulate new constraints as needed for the interpretation tree approach to recognize objects (section 3) In section 4 we present exemplary recognition results ....

Fridtjof Stein and Grard Medioni. Structural Indexing: Efficient 3D Object Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14(2):125-145, February 1992.

....recognition is an active research area in the field of computer vision, 1] 5] 8] for example. Most of the work on object recognition is based on matching sets of geometric image features (e.g. edges, lines and corners) to 3 D object models and significant progress has been achieved, 5] [6], for example. However, most of the geometry based schemes can only handle simple man made objects due to the fact that geometric features are not always adequate for discriminatory and robust 3 D object recognition. Color provides powerful information for object recognition due to the 3 fold ....

Stein, F. and Medioni, G., Structural Indexing: Efficient 2-D Object Recognition, IEEE PAMI, Vol. 14, pp. 1198-1204, 1992.

....and occlusion in the scene. Many local representations are primitive based. In [29] model surfaces are approximated by linear primitives such as points, lines and planes. The recognition of objects is carried out by attempting to locate the objects through a hypothesize and test process. In [78], super segments and splashes are proposed to represent 3D curves and surface patches with significant structural changes. A splash is a local Gaussian map describing the distribution of surface normals along a geodesic circle. Since a splash can be represented as a 3D curve, it is approximated by ....

F. Stein and G. Medioni, Structural indexing: efficient 3-D object recognition, IEEE Transaction Pattern on Pattern Analysis and Machine Intelligence, 14(2): pp. 125-145, 1992.

.... The interpretation of geospatial images is a complex task due to its complexity and diversity of objects within it involving object recognition and scene analysis tasks [DH73] which have been pursued by different approaches, namely pattern recognition techniques based on texture and shape analysis [SM92], and statistical classification algorithms based on spectral similarity analysis [Ric86] Research in the field of spatial analysis are mainly related to the clustering problem, being the statistical spatial association measures [Ric86] GW92] Ans93] 1 [Cre92] and artificial intelligence ....

F. Stein and G. Medioni. Structural indexing: Efficient 2-d object recognition. IEEE Transactions on Pattern Recognition and Machine Intelligence, 14(12):1198--1204, 1992.

.... with respect to the number of known instances in the database has been a difficult problem and has been tackled from the perspectives of both searching strategies [41] and invariance indexing [89] Techniques such as hashing and indexing have also been developed in order to solve this problem [12, 89, 95, 103]. A major problem with indexing is the many to one mapping from object representation to indexed space, i.e. the representation can not be recovered from the index itself. Under this condition, the probability that multiple objects or noise resolve to the same index is high. Consequently, ....

F. Stein and G. Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Trans. Patt. Anal. Machine Intell., 14(2), 1992.

...., can then be defined as: G s = f ae s ae max ; s max g: 9) where ae max and max are the maximum elongation and compactness values of a shape and must be determined experimentally. 3. 2 Local features Many approaches have been developed to extract local shape features from visual data [6, 32, 16, 37, 21, 34]. Due to the computationally intensive nature of the image processing, it is not practical to use all the contour points to describe the shape of an object. Furthermore, not all the contour points have significant contribution to the shape identification. We first use eight directional onepixel ....

F. Stein and G. Medioni. Structural Indexing: Efficient 2-D Object Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14(12):1198--1204, December 1992.

.... subsets are typically of the minimal size that is needed to fully determine all pose parameters (in the 3 D case, there are six parameters, three translational and three rotational) The generated hypotheses are then verified by performing exhaustive comparison between scene and model features [8, 9, 10, 11, 12, 13]. 3. Vote Accumulation: Hypotheses are generated by selecting scene feature subsets that cover all scene features, and aligning them with model subsets. These hypotheses are then clustered, and the one corresponding to the cluster of largest size is selected [14, 15, 16] Choice of the appropriate ....

F. Stein and G. Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Trans. on Pattern Anal. and Mach. Intell., 14(2):125--145, 1992.

....and pose estimation algorithms are based on the Feature Space Trajectory [5] FST) representation for different distorted views (perspective distortion) of an object. The features are global ones not local geometric primitives such as edges, corners, or surfaces typically used in other methods [6]. In this paper we consider global features derived from the Karhunen Lo eve (KL) transform [7] also known as principle component analysis) since this reduces the dimensionality of image data using eigenanalysis. Eigenfeatures are attractive as they compress data and are easily updated when new ....

F. Stein and G. Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Trans. PAMI, 14(2):125-- 144, February 1992.

....and free form surfaces interchangeably. Since there is no other restriction on O, it is not constrained to be polyhedral, piecewisequadric or superquadric; the shape of the object can be arbitrary. Recent approaches using algebraic polynomials [10, 16] splash and super (polygonal) segments [17], simplex angle image [4] 2D silhouettes with internal edges [3] and point sets based registration [2] have specifically sought to address the issue of representing complex curved free form surfaces. They suffer from one or more limitations relating to segmentation issues, surface fitting ....

Fridtjof Stein and G'erard Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14(2):125--145, 1992.

....as free form has proved to be a formidable research problem, because of the lack of simple and segmentable descriptions of object geometry in models and images of those models. While general principles for free form recognition have been proposed [3] and a few prototype systems have been developed [22, 18, 8], little sustained work on this problem has yet appeared in the literature. The lack of progress is particularly vexing given recent advances in computer aided design technology, which has made free form design more intuitive and rapid than in the early days of CAD. The development of computer ....

....models is the viewpoint independent representation. Features in this sort of model are designed to be invariant to translation and rotation; hence, the matching of these features will only be affected by occlusion. An example of this method is Stein and Medioni s work based on splash features [18]. These local features encode the difference between the surface normal at the point of interest and the normal on a contour of constant geodesic radius away. This information is transformed into a periodic curve and encoded into a discriminatory model through a polygonal approximation. The ....

F. Stein and G. Medioni. Structural indexing: Efficient 3-d object recognition. IEEE Trans. Pattern Anal. Machine Intell., 14(2):125--145, February 1992.

.... algorithms which record the variations in the methods contains the efforts of Ayache and Faugeras [1] Pollard, et al. 9] Grimson and Lozano Perez [6] Bolles and Horaud [2] Faugeras and Hebert [5] Thompson and Mundy [14] Huttenlocher and Ullman [7] Lamdan and Wolfson [8] Stein and Medioni [13], and Califano and Mohan [3] In this article we study how verification based on understanding model and image topology leads to better verification results. We use the term topology to represent the connectivity relationships between features. The importance of the use of topology in vision is ....

F. Stein and G. Medioni. Structural Indexing: Efficient 3-D Object Recognition. PAMI, 14(2):125--145, 1992.

....robust object localization can be achieved. The problem of 3D object recognition 1 has received significant attention during the last two decades (e.g. see surveys [1, 2, 3, 4] Most 3D object recognition systems rely on a single type of sensory data such as vision [5, 6, 7, 8, 9, 10] range [1, 11, 12, 13], or touch [14, 15, 16] Thus, they are unsuitable for utilizing various types of sensory data which may be readily available in some tasks, such as ours, in order to improve efficiency and robustness. There have been few efforts for integrating visual and tactile data in the context of 3D object ....

....is presented in the Appendix. 3. 2 Constraints on Transformation Invariant Model Attributes It is a common practice in many 3D object recognition systems to utilize transformationinvariant attributes of model feature sets, in order to reduce the number of scene model matching processes (e.g. [24, 25, 26, 16, 13]) Given a scene feature set, a number of bounds on invariant model attributes are computed. A model set can be consistent with the scene set, only if its attributes lie within the respective scene bounds. Thus, a large number of incompatible model sets can be eliminated from matching using very ....

F. Stein and G. Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Trans. on Pattern Anal. and Mach. Intell., 14(2):125--145, 1992.

.... more efficient by applying geometric constraints [18, 19] or by using local feature focus methods [6, 7, 15] In addition, researchers have studied indexing schemes for quickly recovering correspondence hypotheses without resorting to comparison of all pairs of model image feature sets [5, 8, 13, 26, 32, 33]. The problem of pose estimation from feature correspondences has also been studied independent of other stages [12, 25, 30, 35] Our current work addresses remaining problems that need to be overcome in order to construct more general machine vision systems: first, we need to handle free form ....

....let oe(C) denote this shape property of a part C which can provide some discriminating power for part recognition. We encode each part C using the quantized attributes. All encoded parts serve as indexing primitives to search a hash table for model hypotheses (similar to the hashing scheme in [33]) Invariant attributes of parts index to model aspects. The variant attributes of parts can be used to group consistent model aspect hypotheses into clusters and also to provide an approximate alignment between a sensed part and a model part. Thus, variant attributes provide an initial pose for ....

F. Stein and G. Medioni. Structural indexing: Efficient 2-D object recognition. IEEE Trans. on Pattern Analysis and Machine Intelligence, 14(12):1198--1204, 1992.

....for matching are generated. These can then be verified using a relatively more complex procedure [24] The hypothesis generation can be made very efficient if it is formulated as an indexing problem where model data is stored in tables that are indexed by some function of the observed image data. [8, 17, 23, 34, 37] Since index tables are by definition discrete, the discrete nature of the combinatorial structure representation is ideally suited for this technique. In this work we will be primarily concerned with the hypothesis generation part using the combinatorial structure of groups of line segments for ....

Stein F., and Medioni G., Structural Indexing: Efficient 2-D object Recognition, IEEE Trans. on Pattern Analysis and Machine Intelligence, 14, pp. 1198 - 1204, (1992)

....well defined and continuous almost everywhere, except at vertices, edges, and cusps [1] Figure 1 shows a set of range images of 3D objects with free form surfaces. With a growing interest in automated manufacturing and inspection, representation of sculpted objects is gaining a lot of attention [1, 10, 2]. Previous approaches to 3D object representation can be categorized as either viewpoint independent (objectcentered) or viewpoint dependent (viewer centered) A viewpoint independent representation attaches a coordinate system to an object; all points or object features are specified with respect ....

Fridtjof Stein and G'erard Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 14(2):125--145, 1992.

....database of 3,200 views of 10 objects that view aspects can be determined for sculpted objects easily and effectively. 1. INTRODUCTION With increased interest in automated manufacturing and inspection, representation and recognition of sculpted (freeform) 3D objects is gaining a lot of attention [1, 2, 3]. A free form surface is defined as a smooth surface on which the surface normal is well defined and continuous almost everywhere, except at vertices, edges and cusps [1] Figure 1 shows a set of range images of 3D objects with free form surfaces. One of the important design issues in building an ....

Fridtjof Stein and G'erard Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Trans. on Pattern Anal. and Machine Intell., 14(2):125--145, 1992.

....years. We briefly discuss some of these methods, and then discuss the relationship between our method and those other methods that are more closely related to it. 11.1 A review of labelling methods The matching problem has been approached in many different ways in the computer vision literature [2, 3, 6, 9, 10, 11, 20, 23, 24, 27, 28, 49, 52, 54, 58, 60, 63]. The early attempts, still widely popular, are based on graph search methods. These techniques generally rely on heuristic measures to reduce the complexity of the inherently NP complete search problem to a more manageable level. More recent are the efforts based on energy minimisation using ....

F. Stein and G. Medioni. Structural indexing: efficient 3-D object recognition. IEEE Trans. Pattern Analysis and Machine Intelligence, 14:125--145, 1992.

....and Engineering Research Council of Canada under Grant No. OGP0042194. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of NSERC. about the identity of the scene object and its 3 D pose (e.g. [3, 8, 11, 14, 20, 22]) Each generated hypothesis is then verified, and possibly refined, by comparing the hypothesized model object with the rest of the scene features (e.g. 3, 4, 5, 9, 11, 13] This approach has the advantage of having a polynomial time complexity. However, using a small set of features usually ....

F. Stein and G. Medioni. Structural indexing: Efficient 3-D object recognition. IEEE Trans. on Pattern Anal. and Mach. Intell., 14(2):125--145, 1992.

.... the sensor independence of our algorithm has been demonstrated using several different 3 D sensing modalities [11] Our recognition technique was developed from a combination of basis geometric hashing proposed by Lamdan and Wolfson [14] and structural indexing proposed by Stein and Medioni [17]. Because we use information from the entire surface of the object in our representation, instead of a curve or surface patch in the vicinity of the point, our representation is more discriminating than the curves used to date in structural indexing. Furthermore, because bases are computed from ....

F. Stein and G. Medioni. Structural indexing: efficient 3-D object recognition. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 14, no. 2, pp. 125-145, 1992.

No context found.

F. Stein and G. Medioni. Structural indexing: Efficient 3d object recognition. Transaction on Pattern Analysis and machine Vision (PAMI), 14:125 -- 145, February 1992.

No context found.

F. Stein and G. Medioni. Structural indexing: Efficient 3d object recognition. IEEE Trans. on PAMI, 14(2), 1992.

No context found.

F. Stein and G. Medioni. Structural indexing: efficient 3D object recognition. IEEE Trans. on Pattern Analysis and Mach. Int. (PAMI), 14(2):125--45, Feb. 1992.

No context found.

F. Stein and G. Medioni. Structural indexing: effi- cient 3-D object recognition. IEEE Trans. Pattern Analysis andMachine Intelligence, 14(2): 125-145, 1992.

No context found.

Stein, F. and Medioni, G. Structural Indexing: Efficient 3-D Object Recognition. IEEE Transactions on Pattern aAnaluysis and Machine Intelligence, 14(2): 125---145, 1992.

No context found.

F. Stein and G. Medioni. Structural Indexing: efficient 3-D object recognition. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 14, no. 2, pp. 125-145, 1992.

First 50 documents

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