The use of the clonal selection algorithm for NQR signal detection optimization

Nuclear quadrupole resonance represents a method that provides good results in the detection of explosives substances, narcotics or drugs. It ensures multiple and selective detection with very a high identification accuracy. The phenomenon is well known, but the response signal is very weak and covered by noise, thus requiring algorithms with very good prediction probability. The devices based on this principle can be used in critical operational environments where the false alarm rate must be as close as possible to zero, such as airports and customs. This paper proposes a method for fusing the data from six algorithms for predicting the detected substances, each optimized for a specific scanning scenario. In order to reduce the number of false alarms, the algorithms' results are fused and the decision is taken by applying specific weights to each result. This study proposes a clonal selection algorithm for finding the optimal weights, that is based on a dataset composed of 6000 acquisitions in different conditions. Sodium nitrite is used as the test substance, because it has a piezoelectric response which makes the detection process more difficult in particular situations.