Volker Gass

Kernel Low-Rank and Sparse Graph for Unsupervised and Semi-Supervised Classification of Hyperspectral Images

In this paper, we present a graph representation that is based on the assumption that data live on a union of manifolds. Such a representation is based on sample proximities in reproducing kernel Hilbert spaces and is thus linear in the feature space and nonlinear in the original space. Moreover, it also expresses sample relationships under sparse and low-rank constraints, meaning that the resulting graph will have limited connectivity (sparseness) and that samples belonging to the same group will be likely to be connected together and not with those from other groups (low rankness). We present this graph representation as a general representation that can be then applied to any graph-based method. In the experiments, we consider the clustering of hyperspectral images and semi-supervised classification (one class and multiclass).

Cluster validity measure and merging system for hierarchical clustering considering outliers

Clustering algorithms have evolved to handle more and more complex structures. However, the measures that allow to qualify the quality of such clustering partitions are rare and have been developed only for specific algorithms. In this work, we propose a new cluster validity measure (CVM) to quantify the clustering performance of hierarchical algorithms that handle overlapping clusters of any shape and in the presence of outliers. This work also introduces a cluster merging system (CMS) to group clusters that share outliers. When located in regions of cluster overlap, these outliers may be issued by a mixture of nearby cores. The proposed CVM and CMS are applied to hierarchical extensions of the Support Vector and Gaussian Process Clustering algorithms both in synthetic and real experiments. These results show that the proposed metrics help to select the appropriate level of hierarchy and the appropriate hyperparameters. HighlightsCluster validity measure for arbitrary shaped clusters with outliers.Cluster merging system grouping cluster cores based on the outliers� structure.Truly hierarchical variants of support vector and Gaussian process clustering.Benefits for unsupervised change detection applications are presented.

Unsupervised change detection via hierarchical support vector clustering

When dealing with change detection problems, information about the nature of the changes is often unavailable. In this paper we propose a solution to perform unsupervised change detection based on nonlinear support vector clustering. We build a series of nested hierarchical support vector clustering descriptions, select the appropriate one using a cluster validity measure and finally merge the clusters into two classes, corresponding to changed and unchanged areas. Experiments on two multispectral datasets confirm the power and appropriateness of the proposed system.

Semi-supervised and unsupervised novelty detection using nested support vector machines

Very often in change detection only few labels or even none are available. In order to perform change detection in these extreme scenarios, they can be considered as novelty detection problems, semi-supervised (SSND) if some labels are available otherwise unsupervised (UND). SSND can be seen as an unbalanced classification between labeled and unlabeled samples using the Cost-Sensitive Support Vector Machine (CS-SVM). UND assumes novelties in low density regions and can be approached using the One-Class SVM (OC-SVM). We propose here to use nested entire solution path algorithms for the OC-SVM and CS-SVM in order to accelerate the parameter selection and alleviate the dependency to labeled “changed” samples. Experiments are performed on two multitemporal change detection datasets (flood and fire detection) and the performance of the two methods proposed compared.

Time and Covariance Threshold Triggered Optimal Uncooperative Rendezvous Using Angles-Only Navigation

A time and covariance threshold triggered optimal maneuver planning method is proposed for orbital rendezvous using angles-only navigation (AON). In the context of Yamanaka-Ankersen orbital relative motion equations, the square root unscented Kalman filter (SRUKF) AON algorithm is developed to compute the relative state estimations from a low-volume/mass, power saving, and low-cost optical/infrared camera’s observations. Multi-impulsive Hill guidance law is employed in closed-loop linear covariance analysis model, based on which the quantitative relative position robustness and relative velocity robustness index are defined. By balancing fuel consumption, relative position robustness, and relative velocity robustness, we developed a time and covariance threshold triggered two-level optimal maneuver planning method, showing how these results correlate to past methods and missions and how they could potentially influence future ones. Numerical simulation proved that it is feasible to control the spacecraft with a two-line element- (TLE-) level uncertain, 34.6% of range, initial relative state to a 100 m v-bar relative station keeping point, at where the trajectory dispersion reduces to 3.5% of range, under a 30% data gap per revolution on account of the eclipse. Comparing with the traditional time triggered maneuver planning method, the final relative position accuracy is improved by one order and the relative trajectory robustness and collision probability are obviously improved and reduced, respectively.

Robust Phase-Correlation Based Registration of Airborne Videos Using Motion Estimation

This paper presents a robust algorithm for the registration of airborne video sequences with reference images from a different source (airborne or satellite), based on phase-correlation. Phase-correlations using Fourier-Melin Invariant (FMI) descriptors allow to retrieve the rigid transformation parameters in a fast and non-iterative way. The robustness to multi-sources images is improved by an enhanced image representation based on the gradient norm and the extrapolation of registration parameters between frames by motion estimation. A phase-correlation score, indicator of the registration quality, is introduced to regulate between motion estimation only and frame-toreference image registration. Our Robust Phase-Correlation registration algorithm using Motion Estimation (RPCME) is compared with state-of-the-art Mutual Information (MI) algorithm on two different airborne videos. RPCME algorithm registered most of the frames accurately, retrieving much better orientation than MI. Our algorithm shows robustness and good accuracy to multisource images with the advantage of being a direct (non-iterative) method.