This paper presents a new action recognition approach based on local spatio-temporal features. The main contributions of our approach are twofold. First, a new local spatio-temporal feature is proposed to represent the cuboids detected in video sequences. Specifically, the descriptor utilizes the covariance matrix to capture the self-correlation information of the low-level features within each...

Many models for sparse regression typically assume that the covariates are known completely, and without noise. Particularly in high-dimensional applications, this is often not the case. Worse yet, even estimating statistics of the noise (the noise covariance) can be a central challenge. In this paper we develop a simple variant of orthogonal matching pursuit (OMP) for precisely this setting. W...

An approach for identifying single-input single-output discrete-time dynamic nonlinear errors-invariables systems is presented where the system model can be linearized such that it is expressed as a linear combination of polynomials of input and output observations. We assume white Gaussian noise on both input and output, characterized by a noise magnitude and a normalized noise covariance stru...

This paper introduces a “weighted” matching algorithm to estimate a robot’s planar displacement by matching dense twodimensional range scans. Based on models of expected sensor uncertainty, our algorithm weights the contribution of each scan point to the overall matching error according to its uncertainty. A general maximum likelihood formulation is used to optimally estimate the displacement b...

Considering the mismatch and the high computational complexity of video mosaicing, a fast stitching method based on the normalized covariance is proposed. Firstly, the image corner features are extracted using the SURF algorithm and the corresponding matching points are found using the nearest neighbor method. The least square method is used to filter the matching points and accurately match th...

A covariance matrix is a tool that expresses the odometry uncertainty of mobile robots. The covariance matrix is a key factor in various localization algorithms such as the Kalman filter or topological matching. However, it is not easy to acquire an accurate covariance matrix because the real states of robots are not known. Till now, few results on estimating the covariance matrix have been rep...

This chapter addresses the problem of appearance matching, while employing the covariance descriptor. We tackle the extremely challenging case in which the same non-rigid object has to be matched across disjoint camera views. Covariance statistics averaged over a Riemannian manifold are fundamental for designing appearance models invariant to camera changes. We discuss different ways of extract...

Class specialization is the most original feature of object orientation, but identifying it to subtyping leads to the well known covariance-contravariance controversy. Type safety requires contravariance while specialization needs covariance. This paper aims to precisely analyse this problem, to show how irreductible it is and the need for type errors. We show that many alternatives as multiple...

We first describe in a unified way how to compute the covariance matrix from the gray levels of the image. We then experimentally investigate whether or not the computed covariance matrix actually reflects the accuracy of the feature position by doing subpixel correction using variable template matching. We also test if the accuracy of the homography and the fundamental matrix can really be imp...