利用LVQ神经网络识别放射性核素

Using LVQ Neural Network to Identify Radionuclides

设计了一种新的核素识别算法,利用小波滤波对原始能谱进行去噪,将神经网络模型嵌入该算法中,利用MC法得到大量的能谱数据作为LVQ神经网络模型的训练样本,同时利用放射性核素分支比这一指纹,最后,用MATLAB编写仿真代码,得到了该模型的识别结果,并给出了该模型与传统的核素识别算法,BP神经网络模型之间结果的比较,该算法不仅提高了识别的正确率,同时缩短了识别的时间。

A new radionuclides identification algorithm is designed. Wavelet filtering is used to reduce the noise of the original spectrum. The neural network model is embedded in the algorithm. MC method is used to obtain a large number of energy spectrum data as a training sample of LVQ neural network model.At the same time,the algorithm utilizes the radionuclide branch ratio as a fingerprint. Finally,the simulation code is written with MATLAB,and the identification result of the model is obtained. The comparison between the model and the traditional nuclide identification algorithm and BP neural network model is given. This algorithm not only improves the identification accuracy,but also shortens the time.