Segmentation of GPR images in the case of presence of the error in measured time of signal arrival

This paper deals with GPR data segmentation. A new method is developed for situation when time of arrival is measured with an error. In this case segmentation is considered as the detection of the random signal in the noise. Main statistics in the detection rule is likelihood ratio. This statistics is described by a system of differential equations. In the paper a solution of this system is suggested and a decision rule with its characteristics is described.