Combined high-dimensional analysis of variance (HANOVA) and sequential probability ratio test (SPRT) to detect buried mines

Statistical method to detect subsurface objects using array ground-penetrating radar data

“…6(a). Because the SPRT in (17) has the form of a modified "integrator," a typical time series for the SPRT statistic takes a step-like form. The larger and sharper the step, the more likely it is that a target is present.…”

“…The most often used background removal methods include casual methods, such as subtraction of a moving average from the observation [17], [18], and noncausal methods, such as subtraction of an ensemble average from the observation [19], [20]. Causal methods use data from previous and present collection, noncausal methods use data from previous and future collection.…”

“…It turns out that we only need to estimate , and can be found from the following relation: (16) At the th stop, we estimate the mean of by its maximumlikelihood estimator [26]. The SPRT statistic is a cumulative sum, changing with the acquisition of each new (17) Because subsurface object detection is a binary hypothesis testing problem, e.g., we are only interested in knowing whether there is a buried object, the SPRT statistic is bounded from lower bound, zero. When is negative, is reset to zero.…”