A multilayer detection/tracking algorithm based on Hidden Markov Model

Ground-penetrating radar (GPR) is a nondestructive geophysical method that uses radar pulses to image the subsurface and can be widely used for archaeology, geology exploration, military survey, structure survey and so on. Layer-tracking, which extracts the information of underground layer, is important for location layer underground. A multilayer detection/tracking (D/T) algorithm based on Hidden Markov Model is proposed in this paper. A self-adapting and sliding time window, which adjusts itself according to the position of the tracked layer, acts on the GPR data to ensure the current processed data being corresponding to the current tracked layer without introducing an extra interface, which is often caused by data resetting. Moreover, a Gauss windowing carried out on the prior probability density function of the layer position is used to force the tracking upon the current layer and avoid the target skip during the multilayer D/T. In addition, the filtering of the tracked layer points increases the accuracy of the algorithm. The experimental results show that the algorithm performs well and has obvious advantage comparing with the conventional algorithm based on Hidden Markov Model.