Probabilistic Pose Recovery Using Learned Hierarchical Object Models
نویسندگان
چکیده
This paper presents a probabilistic representation for 3D objects, and details the mechanism of inferring the pose of real-world objects from vision. Our object model has the form of a hierarchy of increasingly expressive 3D features, and probabilistically represents 3D relations between these. Features at the bottom of the hierarchy are bound to local perceptions; while we currently only use visual features, our method can in principle incorporate features from diverse modalities within a coherent framework. Model instances are detected using a Nonparametric Belief Propagation algorithm which propagates evidence through the hierarchy to infer globally consistent poses for every feature of the model. Belief updates are managed by an importance-sampling mechanism that is critical for efficient and precise propagation. We conclude with a series of pose estimation experiments on real objects, along with quantitative performance evaluation. RÉSUMÉ. Ce texte présente une représentation probabiliste pour objets 3D, et détaille le mécanisme d’inférence de pose d’objets réels à partir d’observations visuelles. Notre modèle d’objet se présente comme une hiérarchie de caractéristiques 3D de plus en plus expressives, et représente de manière probabiliste les relations 3D entre ces dernières. Les caractéristiques au niveau inférieur de la hiérarchie sont associées à des perceptions locales. Nous limitions l’usage actuel aux caractéristiques visuelles ; cependant, le principe de notre méthode permet intrinsèquement d’incorporer des caractéristiques de diverses modalités au sein d’une même représentation. Les instances d’un modèle sont détectées à l’aide d’un algorithme de propagation de croyances non-paramétrique, qui propage l’information observée au travers de la hiérarchie pour inférer une pose globalement cohérente pour chaque caractéristique du modèle. Nous présentons un mécanisme de mise à jour de croyance par échantillonage par importance ; nous 1re soumission à International Cognitive Vision Workshop (ICVW) 2008, le 10/02/2008 2 1re soumission à International Cognitive Vision Workshop (ICVW) 2008 présentons également une série d’expérience d’estimation de pose d’objets réels, ainsi qu’une évaluation quantitative des performances atteintes.
منابع مشابه
Hierarchical Integration of Local 3D Features for Probabilistic Pose Recovery
This paper presents a 3D object representation framework. We develop a hierarchical model based on probabilistic correspondences and probabilistic relations between 3D visual features. Features at the bottom of the hierarchy are bound to local observations. Pairs of features that present strong geometric correlation are iteratively grouped into higher-level meta-features that encode probabilist...
متن کاملProbabilistic Object Models for Pose Estimation in 2D Images
We present a novel way of performing pose estimation of known objects in 2D images. We follow a probabilistic approach for modeling objects and representing the observations. These object models are suited to various types of observable visual features, and are demonstrated here with edge segments. Even imperfect models, learned from single stereo views of objects, can be used to infer the maxi...
متن کاملGeneralized 3D Object Representation using Bayesian Eigenobjects
Current methods for representing 3D objects for robotic interaction have significant limitations. They do not allow knowledge transfer from previously encountered objects to similar but novel objects, they construct large object databases that do not scale and are expensive to query, or they require hand tuned object models. We propose the use of Variational Bayesian Principal Component Analysi...
متن کاملGesture recognition using a probabilistic framework for pose matching
This paper presents an approach for view-based recognition of gestures. The approach is based on representing each gesture as a sequence of learned body poses. The gestures are recognized through a probabilistic framework for matching these body poses and for imposing temporal constrains between different poses. Matching individual poses to image data is performed using a probabilistic formulat...
متن کاملProbabilistic Models of Object Geometry with Application to Grasping
Robot manipulators typically rely on complete knowledge of object geometry in order to plan motions and compute grasps. But when an object is not fully in view it can be difficult to form an accurate estimate of the object’s shape and pose, particularly when the object deforms. In this paper we describe a generative model of object geometry based on Mardia and Dryden’s “Probabilistic Procrustea...
متن کامل