The Infona portal uses cookies, i.e. strings of text saved by a browser on the user's device. The portal can access those files and use them to remember the user's data, such as their chosen settings (screen view, interface language, etc.), or their login data. By using the Infona portal the user accepts automatic saving and using this information for portal operation purposes. More information on the subject can be found in the Privacy Policy and Terms of Service. By closing this window the user confirms that they have read the information on cookie usage, and they accept the privacy policy and the way cookies are used by the portal. You can change the cookie settings in your browser.
Detecting incidental scene text is a challenging task because of multi-orientation, perspective distortion, and variation of text size, color and scale. Retrospective research has only focused on using rectangular bounding box or horizontal sliding window to localize text, which may result in redundant background noise, unnecessary overlap or even information loss. To address these issues, we propose...
We present a theoretically grounded approach to train deep neural networks, including recurrent networks, subject to class-dependent label noise. We propose two procedures for loss correction that are agnostic to both application domain and network architecture. They simply amount to at most a matrix inversion and multiplication, provided that we know the probability of each class being corrupted...
Depth from focus (DfF) is a method of estimating depth of a scene by using the information acquired through the change of the focus of a camera. Within the framework of DfF, the focus measure (FM) forms the foundation on which the accuracy of the output is determined. With the result from the FM, the role of a DfF pipeline is to determine and recalculate unreliable measurements while enhancing those...
Over the past few years, softmax and SGD have become a commonly used component and the default training strategy in CNN frameworks, respectively. However, when optimizing CNNs with SGD, the saturation behavior behind softmax always gives us an illusion of training well and then is omitted. In this paper, we first emphasize that the early saturation behavior of softmax will impede the exploration of...
We present an approach to effectively use millions of images with noisy annotations in conjunction with a small subset of cleanly-annotated images to learn powerful image representations. One common approach to combine clean and noisy data is to first pre-train a network using the large noisy dataset and then fine-tune with the clean dataset. We show this approach does not fully leverage the information...
Compressive Sensing is practical and implemented into many areas. For these applications the conventional sensing noise (e.g. AWGN) with low energy on measurements could be reduced by the robustness of Compressive Sensing. Parallel, there exist some errors, which would strongly noise or remove some parts of measurements. In this case, the recovered information would contain much noise, since the robustness...
This paper proposes a robust fault detection observer based on zonotopes for discrete-time uncertain systems with sensor faults and unknown but bounded uncertainties. The main advantage of this method is that the observer gain of the robust zonotopic observer is designed to be robust against bounded uncertainties while being sensitive to faults. In order to detect sensor faults with low magnitudes,...
This paper aims at non-asymptotically estimating the fractional integral and derivative of the output for a class of fractional order linear systems in noisy environment with unknown initial conditions. For this purpose, the considered system modeled by the pseudo-state space representation is firstly transformed into a fractional order differential equation. Secondly, based on the obtained equation...
This paper investigates the distributed localization problem for sensor networks with noisy distance measurements. A distributed iterative algorithm called ECHO-MN is presented based on the signed barycentric coordinate representation, which can be calculated by relative distance measurements. The measurement noise model is presented followed by an unbiased distance estimator which utilizes the past...
For multi-model multisensor system with uncertain variance linearly correlated white noises, the problems of designing robust weighted fusion Kalman estimators (predictor, filter, smoother) are addressed. According to the minimax robust estimation principle, applying Lyapunov equation approach, a unified design approach to obtain the local and three weighted fusion robust Kalman estimators of the...
Estimating input signal of a system is called deconvolution or input estimation. The white noise deconvolution has important applications in oil seismic exploration, communications, and signal processing. This paper addresses the problem of designing robust white noise deconvolution estimators for a class of uncertain systems with missing measurements, uncertain noise variances and linearly correlated...
The paper addresses the problem of distributed sensor fusion in the framework of random finite set. The Generalized Covariance Intersection (GCI) rule of multi-target densities is extensively used in multi-target Bayesian filtering scheme. But there are two problems in GCI which are unreasonable design of fusion weight and unable to tackle informative differentiation. In order to get rid of the bad...
In this paper, we attack the estimation problem in Kalman filtering when the measurements are contaminated by outliers. We employ the Laplace distribution to model the underlying non-Gaussian measurement process. The maximum posterior estimation is solved by the majorization minimization (MM) approach. This yields an MM based robust filter, where the intractable ℓ1 norm problem is converted into an...
This paper investigates the robust weighted fusion estimation problem of multi-sensor systems with mixed uncertainties, including stochastic parameter uncertainties, missing measurements and uncertain noise variances. The stochastic parameter uncertainties are described by multiplicative noise. Especially, the variances of both the multiplicative and additive noises are uncertain. By introducing two...
This paper addresses the design problem of robust weighted fusion white noise deconvolution estimators for a class of uncertain multisensor systems with missing measurements, uncertain noise variances and linearly correlated white noises. By introducing the fictitious noise, the considered system is converted into one with only uncertain noise variances. According to the minimax robust estimation...
When impulsive noise occurs in a general power line communication (PLC) channel, it is known that the Orthogonal Frequency Division Multiplexing (OFDM) systems are relatively robust to such noise. In this paper, we provide a closed-form analysis of a communication channel where Impulsive Noise (IN) and Additive White Gaussian Noise (AWGN) coexist. Our analysis is used to describe a “noise power spread...
Variable structure filter (VSF) is a robust filter for linear uncertain system. In order to remove the chattering caused by high-frequency gain switching of VSF, the smoothing boundary layer (SBL) has been introduced. And similar to Kalman filter, the optimal state estimation of VSF can be got at the optimal smoothing boundary layer (OSBL). However, in practical applications, the statistical characteristics...
For multi-model multisensor systems with uncertain-covariance multiplicative and additive white noises, a universal fictitious noise-based Lyapunov equation approach is presented, by which the original system can be converted into one with only uncertain additive noise variances. According to the minimax robust estimation principle, based on the worst-case system with conservative upper bounds of...
This paper is concerned with the guaranteed cost robust weighted fusion prediction problem for discrete-time systems with multiplicative noises, colored measurements noises and uncertain noise variances. Applying the augmented state approach and a fictitious noise technique, the original system is converted into a system only with uncertain noise variances. Two classes of guaranteed cost robust weighted...
In this work, the modulating functions method is proposed for estimating coefficients in higher-order nonlinear partial differential equation which is the fifth order Kortewegde Vries (KdV) equation. The proposed method transforms the problem into a system of linear algebraic equations of the unknowns. The statistical properties of the modulating functions solution are described in this paper. In...
Set the date range to filter the displayed results. You can set a starting date, ending date or both. You can enter the dates manually or choose them from the calendar.