-
[show abstract]
[hide abstract]
ABSTRACT: Object recognition is a complex and challenging problem. It involves examining many different hypothesis in terms of the object
class, position, scale, pose, etc., but the main trend in computer vision systems is to lazily rely on the brute force capacity
of computers, that is to explore every possibilities indifferently. Sadly, in many case this scheme is way too slow for real-time
or even practical applications. By incorporating salience in the recognition process, several approaches have shown that it
is possible to get several orders of speed-up. In this chapter, we demonstrate the link between salience and cascaded processes
and show why and how those ones should be constructed. We illustrate the benefits that it provides, in terms of detection
speed, accuracy and robustness, and how it eases the combination of heterogeneous feature types (i.e. dense and sparse features)
by some innovating strategies from the state-of-the-art and a practical application.
Keywordstask-specific salience–cascades–feature combination–optimization
02/2011: pages 59-85;
-
20th International Conference on Pattern Recognition, ICPR 2010, Istanbul, Turkey, 23-26 August 2010; 01/2010
-
Proceedings of the 9th ACM International Conference on Image and Video Retrieval, CIVR 2010, Xi'an, China, July 5-7, 2010; 01/2010
-
Proceedings of the 6th ACM International Conference on Image and Video Retrieval, CIVR 2007, Amsterdam, The Netherlands, July 9-11, 2007; 01/2007
-
[show abstract]
[hide abstract]
ABSTRACT: Résumé Nous présentons une approche rapide et robuste d'appa-riement de graphes destinée à la reconnaissance d'objets spécifiques 2D dans les images. A partir d'un petit nombre d'images d'apprentissage, un graphe modèle de l'objet à apprendre est automatiquement construit. Il contient ses caractéristiques locales (de différents types : points d'inté-rêts, contours et textures) ainsi que leurs relations de proxi-mité spatiale. La détection est basée sur une pré-sélection des sous-graphes les plus performants avec l'information mutuelle puis sur la programmation dynamique avec un treillis. Les expériences démontrent que la méthode pro-posée surpasse les détecteurs d'objets spécifiques de l'état de l'art dans des conditions réalistes de bruit grâce à l'uti-lisation des textures et des contours. Abstract We present a fast and robust graph matching method spe-cially designed for 2D object recognition. Using a small set of training images, a model graph of the object is automati-cally built, containing its different local features (from dif-ferent types : keypoints, lines, textures) together with their proximity spatial relationships. A selection of the most si-gnificant model subgraphs using mutual information, a de-tection lattice and an efficient indexing of the image fea-tures enable a fast detection. Experiments demonstrate that the proposed method outperforms the state-of-the-art spe-cific object detectors in realistic noise conditions thanks to the texture and contour features.
