Real-Time NDT-NDE Through an Innovative Adaptive Partial Least Squares SVR Inversion Approach

Salucci, Marco; Anselmi, Nicola; Oliveri, Giacomo; Calmon, Pierre; Miorelli, Roberto; Reboud, Christophe; Massa, Andrea

doi:10.1109/tgrs.2016.2591439

Cited by 74 publications

(51 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Initialization ( i = 0) ‐ Sample the ( σ L , σ R )‐space with the Latin hypercube sampling ( LHS ) method to collect T 0 samples

({};,, (σ_{L}^{t}, σ_{R}^{t}) t = 1 \dots T_{0})

. Use (1) to compute the EIT data for each t ‐th ( σ L , σ R )‐sample,

{\underset{false¯}{\underset{Z}{true}}}_{0} = ({};,, {\underset{false¯}{Z}}^{t} = normalΨ ((),, σ_{L}^{t}, σ_{R}^{t}) t = 1 \dots T_{0}) \in R^{T_{0} \times M}

.…”

Section: Inverse Problem Solution Approachmentioning

confidence: 99%

“…Use (1) to compute the EIT data for each t ‐th ( σ L , σ R )‐sample,

{\underset{false¯}{\underset{Z}{true}}}_{0} = ({};,, {\underset{false¯}{Z}}^{t} = normalΨ ((),, σ_{L}^{t}, σ_{R}^{t}) t = 1 \dots T_{0}) \in R^{T_{0} \times M}

. Apply the PLS to extract from

{\underset{false¯}{Z}}^{t}

( t = 1, …, T 0 ) the vector

{\underset{false¯}{U}}_{0}^{t} \in R^{J}

containing the J < < M most informative features:

{\underset{false¯}{U}}_{0}^{t} = {\underset{false¯}{Z}}^{t} \underline{\underset{false¯}{W}}

\underline{\underset{false¯}{W}} \in R^{M \times J}

being the PLS weight matrix found through the SIMPLS algorithm . Build the initial training set of cardinality T 0 as follows

D_{0} = ({};,, [{\underset{false¯}{U}}_{0}^{t}; ((),, σ_{L}^{t}, σ_{R}^{t})] t = 1 \dots T_{0});

…”

Section: Inverse Problem Solution Approachmentioning

confidence: 99%

“…The inversion of EIT data for recovering the conductivity distribution within the human chest (directly related to the air content of the lungs) is a nonlinear and ill‐posed problem needing proper regularization schemes and suitable inversion strategies . Recently, learning‐by‐examples (LBE) approaches have emerged as promising alternatives to standard inverse scattering techniques since they proved to yield accurate estimations of the DUT with quasi real‐time performance starting from the knowledge of an input‐output ( I/O ) database (ie, the training set) . In such a framework, this work presents an innovative LBE method for monitoring the human lungs.…”

Section: Introductionmentioning

confidence: 99%

“…In such a framework, this work presents an innovative LBE method for monitoring the human lungs. The training set is built by integrating the partial least squares (PLS) feature extraction with an adaptive sampling scheme to (i) collect the largest amount of information on the I/O relation, while (ii) keeping as low as possible the number of required training samples. Robust predictions are then yielded by means of a support vector regressor (SVR).…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Robust real‐time inversion of electrical impedance tomography data for human lung ventilation monitoring

Salucci

Oliveri

2018

Micro & Optical Tech Letters

Self Cite

View full text Add to dashboard Cite

This work presents an innovative strategy for the real‐time monitoring of the human lungs from electrical impedance tomography (EIT) data. The inverse problem at hand is solved within the learning‐by‐examples (LBE) framework by formulating the estimation of the lungs conductivity as a regression problem. Toward this end, the partial least squares (PLS) feature extraction technique is integrated within an adaptive sampling scheme to generate optimal (ie, highly‐informative and low cardinality) training sets for training a support vector regressor (SVR). Accurate predictions are then performed in the on‐line testing phase with remarkable robustness to the noise.

show abstract

“…Initialization ( i = 0) ‐ Sample the ( σ L , σ R )‐space with the Latin hypercube sampling ( LHS ) method to collect T 0 samples

({};,, (σ_{L}^{t}, σ_{R}^{t}) t = 1 \dots T_{0})

. Use (1) to compute the EIT data for each t ‐th ( σ L , σ R )‐sample,

{\underset{false¯}{\underset{Z}{true}}}_{0} = ({};,, {\underset{false¯}{Z}}^{t} = normalΨ ((),, σ_{L}^{t}, σ_{R}^{t}) t = 1 \dots T_{0}) \in R^{T_{0} \times M}

.…”

Section: Inverse Problem Solution Approachmentioning

confidence: 99%

“…Use (1) to compute the EIT data for each t ‐th ( σ L , σ R )‐sample,

{\underset{false¯}{\underset{Z}{true}}}_{0} = ({};,, {\underset{false¯}{Z}}^{t} = normalΨ ((),, σ_{L}^{t}, σ_{R}^{t}) t = 1 \dots T_{0}) \in R^{T_{0} \times M}

. Apply the PLS to extract from

{\underset{false¯}{Z}}^{t}

( t = 1, …, T 0 ) the vector

{\underset{false¯}{U}}_{0}^{t} \in R^{J}

containing the J < < M most informative features:

{\underset{false¯}{U}}_{0}^{t} = {\underset{false¯}{Z}}^{t} \underline{\underset{false¯}{W}}

\underline{\underset{false¯}{W}} \in R^{M \times J}

being the PLS weight matrix found through the SIMPLS algorithm . Build the initial training set of cardinality T 0 as follows

D_{0} = ({};,, [{\underset{false¯}{U}}_{0}^{t}; ((),, σ_{L}^{t}, σ_{R}^{t})] t = 1 \dots T_{0});

…”

Section: Inverse Problem Solution Approachmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Robust real‐time inversion of electrical impedance tomography data for human lung ventilation monitoring

Salucci

Oliveri

2018

Micro & Optical Tech Letters

Self Cite

View full text Add to dashboard Cite

show abstract

“…Based on efficiency, BP algorithm requires 40 s to 60 s on the computer with Intel Core i5 CPU. After obtaining characteristic data, prediction process of SVM based on BP algorithm requires only 1 s. Therefore, the method not only can recognize the location and shape of the tumors, but also can meet requirements of real-time detection [37]. It can be employed for tumor positioning and help clinicians make decisions.…”

Section: Introductionmentioning

confidence: 99%

Stomach Tumor Localization Method of a Support Vector Machine Based on Capsule Endoscopy

Gong¹,

Zhang²

2017

PIER B

View full text Add to dashboard Cite

Abstract-This study proposes a real-time method to solve the electromagnetic inverse scattering problem. This technique converts this problem into a regression problem using a support vector machine (SVM). The SVM-based solution successfully deals with the nonlinearity and ill-posedness inherent in this problem. Simulation results show the feasibility and effectiveness of the proposed method. The method can effectively locate the tumor target of the stomach regardless of the presence of noise. The positioning effect of the method improves as SNR increases. When the SNR is higher than 50 dB, noise minimally affects the results. Finally, the SVM prediction model is utilized to study the effect of the number of sampling locations on the prediction results. The results show that the more sampling locations, the better the prediction results.

show abstract

Automatic Detection and Classification of Buried Objects Using Ground‐Penetrating Radar for Counter‐Improvised Explosive Devices

et al. 2018

View full text Add to dashboard Cite

In this paper, a technique that can automatically detect and classify objects buried under the ground is proposed. The technique employs a ground‐penetrating radar that transmits electromagnetic waves in order to strike the objects and then receives the backscattering electromagnetic wave to perform signal processing. This signal processing is divided into four main steps as follows. First, preprocessing is used to reduce the clutter due to the effect of the media layer interface. Second, the late time of the scattering signal is estimated using a simple cross correlation. Third, a few successive poles are extracted from the scattering response at the estimated late time by using the short‐time matrix pencil method. Finally, the extracted poles are fed for object classification with different constitutions and/or shapes using a support vector machine. Simulations according to the practical situation in three southern provinces of Thailand to counter the improvised explosive devices were set up. The performance of the proposed technique was evaluated. The simulation results showed that the proposed technique can efficiently detect and classify buried objects for counter‐improvised explosive device operations in the military.

show abstract

Real-Time NDT-NDE Through an Innovative Adaptive Partial Least Squares SVR Inversion Approach

Abstract: International audienc

Cited by 74 publications

References 45 publications

Robust real‐time inversion of electrical impedance tomography data for human lung ventilation monitoring

Robust real‐time inversion of electrical impedance tomography data for human lung ventilation monitoring

Stomach Tumor Localization Method of a Support Vector Machine Based on Capsule Endoscopy

Automatic Detection and Classification of Buried Objects Using Ground‐Penetrating Radar for Counter‐Improvised Explosive Devices

Contact Info

Product

Resources

About