Introduction To A Model Based Inversion Algorithm For Gpr Signal Processing

This paper describes a non-intrusive buried object classifier for a ground penetration<br>radar (GPR) system. Various GPR data sets and the implemented processing are described. A<br>model based inversion algorithm that utilizes correlation methodology for target classification is<br>introduced. Real data was collected with a continuous wave GPR. Synthetic data was generated<br>with a new software package that implements mathematical models to predict the<br>electromagnetic returns from an underground object. Sample targets and geometries were<br>chosen to produce two experimental scenarios.<br>Each of-the real measurements and their matching simulated data set were imaged with<br>the same signal processing algorithms. The imaged results were correlated amongst each other<br>to produce a performance measurement for each combination. Thus producing a confusion<br>matrix from which the real data can be analytically compared to the simulated. This final result<br>was used to determine the effectiveness of this technique to determine the real object’s identity.<br>The synthetic data images exhibited similar traits as present in the real data, however, good<br>correlation results were not observed.