Radiation anomaly detection using continuous wavelet transform and Zernike moments

Detecting illicit radioactive materials has been always one of the significant tasks for homeland security due to the radiation emissions of such materials that may endanger the citizens. One of the major challenges in radiation detection is to distinguish between normal or background radiation and anomalous source radiation at low source to background radiation ratios. In this paper, an effective anomaly detection method is proposed to detect anomalous Gamma-ray radiation spectra. The method is based on the continuous wavelet transform to exploit the energy-frequency characteristics of the radiation energy spectrum. A Scalogram image is generated from the transform coefficients, where image processing can be applied for extracting effective feature values. The Zernike moments of the Scalogram image are selected to represent the feature vector of the corresponding radiation spectrum. Finally, classification of the given radiation spectrum is carried out using the Mahalanobis distance that represents the anomaly score. The performance is evaluated using real measurements that are collected from a network of radiation detectors. To present the effectiveness of the proposed work, a comparison is conducted with both benchmark and recent methods in terms of the anomaly detection capabilities. Experimental results show that the proposed method achieves better performance compared to the other methods even at low source to background radiation ratios.