A statistical approach to multichannel blind signal detection for ground penetrating radar arrays

We address the problem of detecting buried mines or objects from multichannel sequentially collected ground penetrating radar (GPR) data for applications such as landmine and unexploded ordnance remediation, utility line mapping, and archaeology. Usually, the exact form of the target signal is not known. We develop and analyze a low-complexity, decision-theoretic, sequential detection strategy to solve this problem. At each stop of the array, a windowed analysis of variance (ANOVA) is designed to look for significant difference among the multichannel signals and suppress accumulated noise. A sequential estimator-detector then processes the ANOVA results to detect mines as the array proceeds down track. Its performance in terms of mean time between false alarms and probability of detection is analyzed. Results from applying the method on real-world data are presented