LINKING RINGS INTO COMPLEX STRUCTURES

Funder

Organizational Unit

Publications

GPR target detection using a neural network classifier designed by a multi-objective genetic algorithm

Publication . Harkat, Houda; Ruano, Antonio; Graca Ruano, Maria; Bennani, S. D.

Ground Penetrating Radar (GPR) is an electromagnetic sensing technology employed for localization of underground utilities, pipes, and other types of objects. The radargrams typically obtained have a high dimensionality, containing a number of signatures with hyperbolic pattern shapes, and can be processed to retrieve information about the target's locations, depths and material type of underground soil. The classical Hough Transform approach used to reconstruct these hyperbola shapes is computationally expensive, given the large dimensionality of the radargrams. In literature, several approaches propose to first approximate the location of hyperbolas to small segments through a classification stage, before applying the Hough transform over these segments. However, the published classifiers designed for this task present a relatively complex architecture. Aiming at an improved target localization, we propose an alternative classification methodology. The goal is to classify windows of GPR radargrams into two classes (with or without target) using a neural network radial basis function (RBF), designed via a multi-objective genetic algorithm (MOGA). To capture samples' fine details, high order statistic cumulant features (HOS) were used. Feature selection was performed by MOGA, with an optional prior reduction using a mutual information (MIFS) approach. The obtained results demonstrate improvement of the classification performance when compared with other models designed with the same data and are among the best results available in the literature, albeit the large reduction in classifier complexity. (C) 2019 Elsevier B.V. All rights reserved.

2019-06Journal article

Restricted access

A method for sub-sample computation of time displacements between discrete signals based only on discrete correlation sequences

Publication . Teixeira, Cesar A.; Mendes, Luis; Graca Ruano, Maria; Pereira, Wagner C. A.

In this paper, we propose a new method for sub-sample computation of time displacements between two sampled signals. The new algorithm is based on sampled auto- and cross-correlation sequences and takes into account only the sampled signals without the need for the customary interpolation and fitting procedures. The proposed method was evaluated and compared with other methods, in simulated and real signals. Four other methods were used for comparison: two based on cross-correlation plus fitting, one method based on spline fitting over the input signals, and another based on phase demodulation. With simulated signals, the proposed approach presented similar or better performance, concerning bias and variance, in almost all the tested conditions. The exception was signals with very low SNRs (<10 dB), for which the methods based on phase demodulation and spline fitting presented lower variances. Considering only the two methods based on cross-correlation, our approach presented improved results with signals with high and moderate noise levels. The proposed approach and other three out of the four methods used for comparison are robust in real data. The exception is the phase demodulation method, which may fail when applied to signals collected from real-world scenarios because it is very sensitive to phase changes caused by other oscillations not related to the main echoes. This paper introduced a new class of methods, demonstrating that it is possible to estimate sub-sample delay, based on discrete cross-correlations sequences without the need for interpolation or fitting over the original sampled signals. The proposed approach was robust when applied to real-world signals and presented a moderated computational complexity when compared to the other tested algorithms. Although the new method was tested using ultrasound signals, it can be applied to any time-series with observable events. (C) 2016 Elsevier Ltd. All rights reserved.

2017-01Journal article

Restricted access