Showing papers by "Jana Kosecka published in 2006"

Image Based Localization in Urban Environments

Abstract: In this paper we present a prototype system for image based localization in urban environments. Given a database of views of city street scenes tagged by GPS locations, the system computes the GPS location of a novel query view. We first use a wide-baseline matching technique based on SIFT features to select the closest views in the database. Often due to a large change of viewpoint and presence of repetitive structures, a large percentage of matches (> 50%) are not correct correspondences. The subsequent motion estimation between the query view and the reference view, is then handled by a novel and efficient robust estimation technique capable of dealing with large percentage of outliers. This stage is also accompanied by a model selection step among the fundamental matrix and the homography. Once the motion between the closest reference views is estimated, the location of the query view is then obtained by triangulation of translation directions. Approximate solutions for cases when triangulation cannot be obtained reliably are also described. The presented system is tested on the dataset used in ICCV 2005 Computer Vision Contest and is shown to have higher accuracy than previous reported results.

Nonparametric estimation of multiple structures with outliers

Abstract: Common problem encountered in the analysis of dynamic scene is the problem of simultaneous estimation of the number of models and their parameters. This problem becomes difficult as the measurement noise in the data increases and the data are further corrupted by outliers. This is especially the case in a variety of motion estimation problems, where the displacement between the views is large and the process of establishing correspondences is difficult. In this paper we propose a novel nonparametric sampling based method for estimating the number of models and their parameters. The main novelty of the proposed method lies in the analysis of the distribution of residuals of individual data points with respect to the set of hypotheses, generated by a RANSAC-like sampling process. We will show that the modes of the residual distributions directly reveal the presence of multiple models and facilitate the recovery of the individual models, without making any assumptions about the distribution of the outliers or the noise process. The proposed approach is capable of handling data with a large fraction of outliers. Experiments with both synthetic data and image pairs related by different motion models are presented to demonstrate the effectiveness of the proposed approach.

Probabilistic location recognition using reduced feature set

Abstract: The localization capability is central to basic navigation tasks and motivates development of various visual navigation systems. In this paper we describe a two stage approach for localization in indoor environments. In the first stage, the environment is partitioned into several locations, each characterized by a set of scale-invariant keypoints and their associated descriptors. In the second stage the keypoints of the query view are integrated probabilistically yielding an estimate of most likely location. The novelty of our approach is in the selection of discriminative features, best suited for characterizing individual locations. We demonstrate that high location recognition rate is maintained with only 10% of the originally detected features, yielding a substantial speedup in recognition and capability of handling larger environments. The ambiguities due to the self-similarity and dynamic changes in the environment are resolved by exploiting spatial relationships between locations captured by hidden Markov model

Generalized RANSAC Framework for Relaxed Correspondence Problems

Abstract: Finding correspondences between two (widely) separated views is essential for several computer vision tasks, such as structure and motion estimation and object recognition. In the wide-baseline matching using scale and/or affine invariant features the search for correspondences typically proceeds in two stages. In the first stage a putative set of correspondences is obtained based on distances between feature descriptors. In the second stage the matches are refined by imposing global geometric constraints by means of robust estimation of the epipolar geometry and the incorrect matches are rejected as outliers. For a feature in one view, usually only one "best" feature (the nearest neighbor) in the other view is chosen as corresponding feature, despite the fact that several match candidates exist. In this paper, we will consider multiple candidate matches for each feature, and integrate this choice with the robust estimation stage, thus avoiding the early commitment to the "best" one. This yields a generalized RANSAC framework for identifying the true correspondences among sets of matches. We examine the effectiveness of different sampling strategies for sets of correspondences and test the approach extensively using real examples of hard correspondence problems caused by a large motion between views and/or ambiguities due to repetitive scene structures.

A new inlier identification scheme for robust estimation problems

Abstract: Common goal of many computer vision and robotics algorithms is to extract geometric information from the sensory data Due to noisy measurements and errors in matching or segmentation, the available data are often corrupted with outliers In such instances robust estimation methods are employed for the problem of parametric model estimation In the presence of a large fraction of outliers sampling based methods are often the preferred choice Traditionally used RANSAC algorithm however requires a large number of samples, prior knowledge of the outlier ratio and an additional, difficult to obtain, inlier threshold for hypothesis evaluation To tackle these problems we propose a novel efficient sampling based method for the robust estimation of model parameters The method is based on the observation that for each data point, the properties of the residual distribution with respect to the generated hypotheses reveal whether the point is an outlier or an inlier The problem of inlier/outlier identification can then be formulated as a classification problem The proposed method is demonstrated on motion estimation problems from image correspondences with a large percentage of outliers (70%) on both synthetic and real data and estimation of planar models from range data The method is shown to be of an order of magnitude more efficient than currently existing methods and does not require a prior knowledge of the outlier ratio and the inlier threshold

Strangeness Based Feature Selection for Part Based Recognition

Abstract: Motivated by recent approaches to object recognition, where objects are represented in terms of parts, we propose a new algorithm for selecting discriminative features based on strangeness measure. We will show that k-nearest neighbour strangeness can be used to measure the uncertainty of individual features with respect to the class labels and forms piecewise constant decision boundary. We study its properties and generalization capability by comparing it with optimal decision boundary and boundary obtained by k-nearest-neighbor methods. The proposed feature selection algorithm is tested both in simulation and real experiments, demonstrating that meaningful discriminative local features are selected despite the presence of large numbers of distractors. In the second stage we demonstrate how to integrate the local evidence provided by the selected features in the boosting framework in order to obtain the final strong classifier. The performance of the feature selection algorithm and the classifier is evaluated on the Caltech five object category database, achieving superior results in comparison with existing approaches at lower computational cost.

Ensemble Method for Robust Motion Estimation

Abstract: The core of the traditional RANSAC algorithm and its more recent efficient counterparts is the hypothesis evaluation stage, with the focus on finding the best, outlier free hypothesis. Motivated by a non-parametric ensemble techniques, we demonstrate that it proves advantageous to use the entire set of hypotheses generated in the sampling stage. We show that by studying the residual distribution of each data point with respect to the entire set of hypotheses, the problem of inlier/ outlier identification can be formulated as a classification problem. We present extensive simulations of the approach, which in the presence of a large percentage (> 50%) of outliers, provides a repeatable and, an order of magnitude more efficient method compared to the currently existing techniques. Results on widebaseline matching and fundamental matrix estimation are presented.

Object recognition using strangeness and transduction

Abstract: Object recognition is one of the most essential functionalities of human vision. It is of fundamental importance for machines to be able to learn and recognize objects and to provide a rejection option when the unknown probe has no mates in the set of known objects. This thesis presents novel recognition algorithms using strangeness for generic part-based object recognition and open set recognition, respectively. For part-based object recognition, objects are assumed to be represented in terms of features characterized by descriptors which are invariant to variation of object appearance. This thesis presents a simple and efficient feature selection algorithm to deal with irrelevant and background features and a new non-parametric weak leaner employed in the boosting framework. Specifically a k-Nearest Neighbor strangeness measure is defined to quantify the uncertainty of features with respect to the class labels and used as the criterion to select the discriminative features from the initial feature set to reduce the complexity of learning stage. The boosting learning algorithm is further used to build the final classifier with strangeness as the non-parametric weak learner to characterize the discriminative evidence of each part. This learning approach is able to handle changes in viewpoint, partial occlusion, local deformations, varying illumination and background clutter. We apply and validate the approach on location recognition, part-based face recognition and weakly supervised object category recognition problems. For the latter task, we propose a two-stage learning strategy to distinguish the object from both background and other objects, with the performance and efficiency superior to the state of the art methods. Finally, in the context of face recognition problem, we present the Open Set Transductive Confidence Machine(TCM)-k Nearest Neighbor (kNN) algorithm for open set recognition using strangeness. The algorithm provides a priori availability of a reject option to answer "none of the above" without modelling the distribution of any object and using the knowledge of unknown classes. It provides a productive solution for the open set recognition problem with a small number of training examples and a large number of classes.