Object Recognition from gpr images
Download 47.38 Kb. Pdf просмотр

 Навигация по данной странице:
 Novelties
 Timescale
 Description of the project
 Signal pre processing and clutter suppression
 Nonfocused image segmentation
 Focused image segmentation
 Object feature extraction
 Expertise of the applicants in the area of the project
Object Recognition from GPR images Project TUDelft/IRCTR en MIPT, printed: 20/03/2006, page: 1 of 7
The goal of this project consists in designing a system of automatic recognition of objects in images obtained with an impulse GPR. Development of such a system suggests the solution of the following problems: 1.
development of methods for automatic object selection in images; 2.
development of methods for object description by a system of features; 3.
estimation of possibility of recognising given classes of objects from the obtained feature vector; 4.
development of efficient recognition algorithms. The choice of methods for solving the first and second problems essentially depends on signal pre processing used in the GPR and methods for constructing GPR images. We are aimed to develop two groups of methods for object selection in an image. The methods of the first group take into account specific features of object patterns on the GPR images constructed without aperture synthesis and are based on waveletanalysis. The methods of the second group are independent of specific features of GPR images and include image filtration and segmentation. Methods of object description by a feature vector also depend on the procedure of constructing radar images. In the context of the first group of methods for object selection, the object description is based on the waveletexpansion of the radar image. For the second group, an object is described by the feature set able to characterize the object shape and texture. For object recognition, we suggest the use of treelike statistical classificators. The advantage of such an approach consists in the fact that such classificators allow for estimating classification quality if the alphabet of classes to be recognized in known.
1.
The use of the twodimensional wavelet specially designed for selecting and describing an object in the radar image. 2.
3.
Description of object shape with the specially designed technique that includes expansion of the object contour in a set of orthogonal functions (in particular, these functions can be wavelets) and construction of a metric that offers a possibility of comparing arbitrarily shaped objects independently of object translations, rotations, and reflections. 4.
Derivation of new treelike statistical classificators (classification algorithms).
Timescale
Object Recognition from GPR images Project TUDelft/IRCTR en MIPT, printed: 20/03/2006, page: 2 of 7 % of activity of scientists Names Year 1
Year 2 Vladimir Sazonov 15 15
Aleksandr Vinogradov 20
20 Natalya Verdenskaya 20 20
Irina Ivanova 20
20 Vladimir Chepelev 35 35
Dmitrii Chikin 40
40 Andrei Tereshin 40 40
Description of the project (a)
Technical/scientific description The general scheme of designing the recognition system is the classical scheme shown in Fig. 1:
Signal preprocessing and clutter suppression Data focusing
Segmentation of the notfocused and focused images Feature extraction
Classification Fig. 1. Either initial or preprocessed data can be used when solving the problems of object detection and object selection. Both approaches have their advantages and limitations. Indeed, dealing with the initial data, we loose no informative data; however, the object pattern and feature set describing the object will essentially depend on the procedure of getting the data. If the data are the preprocessed data (for example, the data is the SAR image constructed from received signals), the information can be partially loosed. However, one can face with the opposite situation in which the image will have greater resolution and object description will be universal. Both approaches of working with data are suggested for the consideration.
In line with the commonly used methods of suppressing clutters from the ground surface (such as subtraction of the signal averaged over certain window), we plan to analyse the usability of other filters. In addition we will attempt to construct the model of the desired signal and the model of environmental Object Recognition from GPR images Project TUDelft/IRCTR en MIPT, printed: 20/03/2006, page: 3 of 7 clutters. With these models, we hope to be able of suggesting the procedure for selecting the desired signal against the background of environmental clutters. Data focusing We suggest to use data preprocessing according to the following scheme: 
quadrature signals, we hope to form the second quadrature signal by the Hilbert transform of received signal, 

aperture synthesis. Nonfocused image segmentation Depending on size, shape, and reflectivity (refractivity), an object form in the GPR image a pattern of several perfect or distorted hyperbolas or ellipses in time or space slices. We suggest the use of certain special wavelet expansion for selecting such patterns. With this goal in view, we will attempt to construct a twodimensional wavelet function with backbone/valley of hyperbolic and elliptic forms. The scale of such a wavelet is governed by two curve parameters. The rate of backbone decay will form an additional parameter of the wavelet. The expansion of a GPR image in wavelet series with this base wavelet will offer a possibility of selecting a family of hyperbolas related to the same object. The set of parameters corresponding to this family, in turn, will offer object description by the set of features.
This processing uses the standard filtrationsegmentationselection scheme. The first step— filtration—is assumed to be performed according to standard procedures. For the second step— segmentation (it is meant as selecting image segments homogeneous in the sense of texture)—, we suggest to use statistical algorithms of segmentation, which are modified algorithms for mixture separation. The third step—object selection—assumes selecting connection regions in the image, and these regions form the object pattern. Both of these procedures (filtration and segmentation) are based on pixel (image dots) probability distribution features. For this reason, these procedures allow multidimensional extensions simply by changing the dimension of the likelihood function. Here we mean 2D and 3D realvalued images. Processing of the complexvalued image require certain transformation of the above algorithms, because the likelihood function changes the number of parameters in this case.
For object description, we suggest to form the feature vector from an analysis of object shape, shapes of object constituents, and texture parameters of these elements. Despite vast literature about the describing and comparing shapes of arbitrary closed curves, this topic is far from being positively solved. However, this topic is urgent in recognising objects in an image. Object Recognition from GPR images Project TUDelft/IRCTR en MIPT, printed: 20/03/2006, page: 4 of 7 The solution that we suggest to consider in this project is as follows. An arbitrary closed contour is represented as a periodic complex function of real argument. This function is expanded in the series in certain system of orthogonal functions (in particular, in wavelets). Then, the function is approximated to a given error by a finite sum of expansion terms, so that the finite set of coefficients used in the finite sum describe the function to a given error. Such curve representations are compared using either the specially constructed metric invariant relative motions in the plane, or the standard Euclidean metric. In the case of the Euclidean metric, a specially transformed set of coefficients is used as curve description instead of the initial one. The shape of a threedimensional object can be described using algorithms of object formation from sections or from expansions in a series in the multidimensional systems of functions. In addition, we plan to consider the possibility of constructing the complex object description that will include the images obtained for both polarisations. Standard procedures of decreasing dimension, such as the KarhunenLoeve expansion, can be used to decrease the dimension of the feature vector.
At the first stage, we suggest to consider the problem of recognition with training. In this case, an additional step appears in the scheme—construction of training samples from experimental data. These samples are used for the formation of the alphabet of classes to be recognised in the developed system of features with and without using the corresponding procedures of decreasing dimension. The same training samples are used for estimating the possibility of recognising a given set of object classes. We suggest to design classification algorithms using statistical treelike classificators. For estimating the possibility of classifying the given set of object classes, we suggest to use the procedures of automatic classification (such as Forel) that allow estimating the quality of resulting division. The final choice of the algorithms will be done after a half year of working with the IRCTR data and will be agreed with IRCTR. The possibility of realtime processing is mainly limited by the speed of algorithms. The speed of every algorithm can be estimated in terms of the number of required operations. In this case, the main portion of calculations is evaluated as the body of data to be processed multiplied by of the number of operations required to process a single dot of the image. The quality of fast algorithms using a small number of operations per one dot is lower than the quality of more complicated procedures, which are, naturally, slower. The choice of specific algorithm will be done at the stage of testing with actual data obtained from GPR.
Thus, the speed of the processing procedure and, consequently, the possibility of its realtime implementation depend on the following basic factors: 

body of data to be processed; Object Recognition from GPR images Project TUDelft/IRCTR en MIPT, printed: 20/03/2006, page: 5 of 7 
required quality; 
the speed of the computer used. It is assumed that processing algorithms will use the digital GPR signal as the initial data. All the algorithms will have a set of parameters to be tuned. The parameters will be tuned during testing the algorithms with data obtained from the IRCTR video impulse radar with consideration for basic GPR performances, such as radar bandwidth, sampling time, and acquisition parameters. The quality of algorithms will be examined using model received signals formed from the model of object reflectivity and the model of surrounding medium. With these models, we assume to form test images that will be used for estimating the quality of processing algorithms. Detection and false alarm probabilities will be estimated using actual samplings obtained from the IRCTR radar. The algorithms to be developed must have characteristics better that the characteristics of the energy projection detector. All algorithms will be implemented as the dynamically linked library (DLL). We assume to develop the algorithms using Delphi 7.0 as the main development medium. We deliver the description of the algorithms, text of routines, description of the library and the examples of MATLAB calls of the main functions from the library.
(b) Utilization As the main utilisation of the results of this project, we see the automation of searching for objects of relatively small size buried for relatively small depths, such as •
•
foreign objects of other origin and, in particular, organic inclusions; •
airfilled cavities of small size. The main purpose of this project consists in automation of deciding the availability of an object, which will offer a possibility of making the process faster and exploring large areas, possibly, without direct participation of personnel. (c) Applicants
No Problems Applicants 1 Data preprocessing (clutter suppressing, designing of models for clutter and desired signal) Aleksandr Vinogradov, Andrei Tereshin 2 Image construction by the synthetic aperture (SA) method Vladimir Sazonov, Dmitrii Chikin 3 2D and 3D segmentation Natalya Verdenskaya 4 Feature extraction Irina Ivanova Object Recognition from GPR images Project TUDelft/IRCTR en MIPT, printed: 20/03/2006, page: 6 of 7 5
Natalya Verdenskaya Irina Ivanova Vladimir Chepelev
Aleksandr Vinogradov Graduated from Moscow Institute of Physics and Technology (MIPT), Department of General and Applied Physics (1970). Degree of Candidate of Sciences in Physics and Mathematics (1974). Thesis: "Effect of Backscattering Enchancement". Scientific title: Senior Scientific worker in Radiophysics. Knowledge: radiophysics, mathematics, wave propagation in random media, radar engineering. Experience: with Mints RTI from 1970, participated in many projects on designing radar systems as theoretician, one of the main designers of automatic system of blood cell recognition and counting. With MIPT from 1990 (assistant professor). Lectures: "Theory of diffraction" (19902000), "Wave propagation" (2000present). Supervision. Publications: more than a hundred papers and scientific reports.
Natalya Verdenskaya Graduated from Moscow State University, Faculty of Mechanics and Mathematics (1985). Degree of Candidate of Sciences in Physics and Mathematics (2002). Thesis: "Image segmentation algorithms and their application to designing automatic recognition systems in medicine and radar engineering". Knowledge: mathematical statistics, image analysis and image processing, pattern recognition. Experience: with Mints RTI from 1987, participated in some projects on designing radar systems as theoretician, one of the main designers of automatic system of blood cell recognition and counting. With MIPT from 2003 (assistant professor). Lectures: "Theory of image processing". Supervision. Publications: more than a fifty papers and scientific reports.
Irina Ivanova Graduated from Moscow State University, Faculty of Computational Mathematics and Cybernetics (1987).
Knowledge: mathematical statistics, image analysis and image processing, pattern recognition. Experience: with Mints RTI from 1987, participated in some projects on designing radar systems as theoretician, one of the main designers of automatic system of blood cell recognition and counting.
Object Recognition from GPR images Project TUDelft/IRCTR en MIPT, printed: 20/03/2006, page: 7 of 7 With MIPT from 2003 (assistant professor). Lectures: "Theory of image processing". Supervision. Publications: more than a forty papers and scientific reports.
Vladimir Chepelev Graduated from Moscow Institute of Physics and Technology (MIPT), Department of General and Applied Physics (2000). Knowledge: radiophysics, radar engineering, pattern recognition. Experience: postgraduate student MIPT. Publications: 7 papers.
Dmitrii Chikin Student, MIPT, IV year.
Andrei Tereshin Student, MIPT, V year. Publication: 2 papers. Каталог: fileadmin fileadmin > Haqiqat izlab fileadmin > Driving the Best Science to Meet Global Health Challenge s fileadmin > Db regiobus Stuttgart, Niederlassung 74080 Heilbronn fileadmin > Busverkehr Gültig ab 11. 12. 2016 b 646 fileadmin > 174 Reisebüro Gross, 74388 Talheim fileadmin > Mazungumzo na adam shafi juu ya uandishi wake wa riwaya fileadmin > 3 mfh / 3 efh lodiker fileadmin > Tistics handbook european athletics championships fileadmin > 48 Kassel Ihringshäuser Straße Fuldatal Vellmar Ahnatal Reised ienst Bonte Do'stlaringiz bilan baham: 
ma'muriyatiga murojaat qiling