PIERS ONLINE, VOL. 4, NO. 6, 2008 621 Classing and Extracting Information from Radar Images G. Angiulli 1 , V. Barrile 1 , G. M. Meduri 1 , R. Pucinotti 2 , and S. Tringali 1 1 DIMET, University Mediterranea, Reggio Calabria 89100, Italy 2 MECMAT, University Mediterranea, Reggio Calabria 89100, Italy Abstract— GPR techniques had been successfully using for many years to examine structures and materials from inside. They are based on the possibility of sending electromagnetic pulses and registering the echo time and the amplitude of the signals eventually back-scattered by the interface between propagation media of different dielectric constants. The result consists of some radar images (the so-called radargrammes), where incidental targets generate distinctive hyperbolic plots or discontinuities, as an indirect evidence of either changes in the properties, nature and composition of the medium or the presence of internal breaks. The present note aims at expounding some tests and preliminary results drawn by the employment of specific algorithms for automatically interpreting, classing and extracting explicit information from radargrammes, once these have been traced from raw radar data by an ad hoc application designed in the MatLab framework. 1. INTRODUCTION The advance following the introduction of new measurement instruments and methods and the development of more and more reliable techniques of data processing, has been allowing professional men and researchers to carry out accurate and fast investigations not even conceivable till some time ago. In the past decades non-invading analysis [1] has been used with growing spread, both for inquiring the physical and mechanical properties of any sort of structural elements composing a work of engineering, and for the entry of detailed information about the internal geometry and composition. Moreover, these have revealed themselves indispensable when dealing with complex works or investigations of a different nature would entail an unavoidable interruption of vehicular traffic in areas where the structure is indeed an essential connection. On the other hand, the usage of radar technology (GPR) in structural surveys [2, 3] comes from its basic applications to underground investigations (namely, Georadar) and nowadays is mostly concerned with the problem of locating bars of armor and detect cavities or breaks inside structures [4], in such a way that is not invasive or destroying. The present note deals with preliminary results coming from the application of a classification algorithm for raster images to radargrammes (in a raster format) obtained from processing raw GPR data (concerning investigations on the girder of a bridge in armoured concrete) by means of a simple application designed and developed in Matlab. This algorithm permits to extrapolate the contour lines due to structural discontinuities or to draw the classical hyperboles produced by a certain signal in presence of reinforcement bars. Hence it is useful for the automized location of such elements inside materials or buildings, once treated with suitable filtering routines (for example, able to remove noisy information), which we are still working on to optimize. 2. BASICS ON GPR IN CIVIL ENGINEERING Localizing bars of reinforcement in concrete is one of the most important applications of GPR to Civil Engineering. In fact for a few years many researchers and practicers have been making intensively use of GPR technology in common applications, especially when dealing with structural inspections, testing and control or with the spotting of reinforcements and holes internal to fer- roconcrete buildings. As known, the physical principle behind georadar techniques is founded on transmitting electromagnetic pulses and recording the time-stamp and the strength of eventually back-scattered signals produced at the interface between materials with different dielectric con- stants. This technology has turned out particularly effective in the surveys of non-electroconductive means and in reveling the presence of metallic objects inside the same materials. Due to relatively short survey times, this allows the survey of large structures. A GPR relief produces a radargram. These surveys are exploited on buildings for which it is practically difficult to get any information about internal structure. They allows to establish the number, the position and the diameter of all steel bars in the surveyed element. It is also possible to localize — whenever present — voids both