Automatic detection and classification of buried interferences with GPR using Artificial Neural Networks (ANNs): Study on IAG/USP test site.

• Grant number: 09/05882-6
• Support type: Scholarships in Brazil - Doctorate
• Effective date (Start): July 01, 2009
• Effective date (End): March 31, 2014
• Field of knowledge: Physical Sciences and Mathematics - Geosciences - Geophysics
• Principal researcher: Jorge Luís Porsani
• Grantee: Vinicius Rafael Neris dos Santos
• Home Institution: Instituto de Astronomia, Geofísica e Ciências Atmosféricas (IAG). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

This PhD project is a continuation of the research that is being development under the research project, already completed, and supported by FAPESP entitled "Geophysical characterization of shallow targets with applications in urban planning, environment and archeology: study on IAG/USP test site (grant no. 02/07509-1). The objective of this research consist to improve the detection of buried targets, determining the type of material aiming to reduce the ambiguities in the interpretation of two important families of targets found in the urban environment: the resistive targets (concrete tubes and plastic drums) and conductive targets (steel drums and metal pipes). These targets will be studied in a controlled site (SCGR - IAG/USP), using the GPR - Ground Penetrating Radar profiles by using 2D and quasi-3D acquisition. On the data will be applied an algorithm that use artificial neural networks (ANNs) that will go automatically detect the targets according to the types of materials found in the subsurface. First, to recognition of hyperbolic diffraction will be used the results of numerical modeling of the targets studied, generating a pattern of synthetic reflection, which serves as a parameter of input files of the ANNs algorithm. Once recognized patterns of hyperbolae, they are included in the characteristics of the diffractions regarding to targets and thus classify them as plastic, metal or concrete. The training algorithm is based on supervised learning for correction of errors or backpropagation. After verification of the results under controlled conditions, the routine of ANNs will be used to recognize and classify hyperbolae in a real situation in the field, in places that there are targets of plastic, metal or concrete. The priori these studies will be done within of the USP campus in São Paulo city. The results obtained in controlled and uncontrolled conditions will improve the interpretation of geophysical data and will have applications in the areas of urban planning, infrastructure and environmental studies in urban areas.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

DOS SANTOS, VINICIUS RAFAEL N.; AL-NUAIMY, WALEED; PORSANI, JORGE LUIS; TOMITA HIRATA, NINA S.; ALZUBI, HAMZAH S. Spectral analysis of ground penetrating radar signals in concrete, metallic and plastic targets. JOURNAL OF APPLIED GEOPHYSICS, v. 100, p. 32-43, JAN 2014. Web of Science Citations: 7.

NERIS DOS SANTOS, VINICIUS RAFAEL; PORSANI, JORGE LUIS. Comparing performance of instrumental drift correction by linear and quadratic adjusting in inductive electromagnetic data. JOURNAL OF APPLIED GEOPHYSICS, v. 73, n. 1, p. 1-7, JAN 2011. Web of Science Citations: 7.