Anomaly-aware leveling algorithm for airborne gamma-ray spectroscopy data

Airborne gamma-ray spectrometry (AGS) data often exhibit directional anomalies termed “stripe noise”, which obscure true subsurface distributions and hinder mineral exploration. To address this without tie-line assistance, this paper proposes an anomaly-aware AGS data leveling algorithm. Original data are clustered using an improved directional neighborhood search strategy with entropy and deviation factors. Based on noise characteristics, a dual-dimensional background estimation technique is designed, while an enhanced sample entropy method accurately delineates anomalous regions to avoid masking crucial mineralization features. Compared to traditional micro-leveling, the proposed approach yields a larger reduction in the Root Mean Square Error (RMSE) across survey lines, improving data consistency by approximately 25\% while successfully preserving roughly 92\% of genuine ore-caused anomalies. Power spectral density and profile analyses confirm that post-leveling data exhibit smoother trends, effectively filtering out high-frequency noise. Specifically, the coefficient of variation decreases by an average of 26\% across 13 out of 14 survey lines using an optimal 3 × 9 filter window. These quantitative results demonstrate that our algorithm achieves excellent stripe suppression with minimal data distortion, providing an efficient, robust solution for high-fidelity AGS data processing.