基于KNN分类算法的n-γ脉冲信号甄别仿真研究

利用脉冲形状甄别（PSD）法区分中子和γ射线脉冲信号是核探测过程中一项重要的任务。本文基于Labview平台实现了n/γ脉冲信号的仿真及信号预处理过程,分别利用传统的甄别方法电荷比较法、脉冲梯度分析(PGA)法及上升时间法对所产生的n/γ脉冲信号进行甄别,筛选出以上三种甄别方法结果一致的中子和γ射线混合脉冲信号作为KNN分类算法的训练集。通过训练样本构建KNN分类模型,使得能够通过该模型实现中子和γ射线脉冲信号的分类。结果表明,基于KNN分类算法的中子和γ射线脉冲信号甄别准确率高达99.58%,与电荷比较法,上升时间法和PGA方法相比,甄别错误率显著降低。并且KNN分类算法原理简单,易于实现,因此可应用于实际混合场中的n-γ脉冲甄别。     

Calculation of Response Function for Bonner Sphere Spectrometer Based on Geant4

In order to make further studies on fusion neutron diagnosis on HL-2A /HL-2M,we have developed and succeeded in the calculation of the Response Function for a Bonner sphere spectrometer,which consists of eight polyethylene spheres with ³He proportional counters inside.The response function of the Bonner spectrometer to neutrons is of fundamental importance for its neutron spectrum unfolding procedure and is directly related to the quality of the unfolded spectrum.In this paper,we calculated the response function to neutrons from 10-9 MeV to100 MeV by Geant4.In order to test the accuracy of the Geant4 simulation,we apply it to measure an ²⁴¹Am-Be neutron source,and the measured neutron counts of the spectrometer and simulated counts are found to be highly consistent,with a relative error up to 9.3%.This has proven the calculation of the neutron response of the Bonner sphere spectrometer by Geant4 to be quite accurate.     

Tokamak fusion neutron spectrometer based on PXI bus

In order to realize on-line real-time measurement of dynamic and time-sharing neutron spectrum of HL-2A,a tokamak fusion neutron spectrometer based on PXI bus was developed.It consists of electronics system and eight thermal neutron detectors,namely SP9 3He proportional counter,embedded in eight polyethylene spheres in different diameters.Response function of the eight polyethylene spheres was the key to calculate the neutron spectrum accurately.In this paper,response function of the eight polyethylene spheres is simulated by adopting Geant4 code,and neutron counts from an 241Am-Be neutron source are measured by the eight detectors.The calculated spectrum of the Am-Be neutron is accurate in 0-2 MeV region,and is similar to the theoretical spectrum.The tokamak fusion neutron spectrometer was used in HL-2A device to monitor the dynamic neutron spectrum of HL-2A on-line and real-time.     

改进Fast ACE结合边缘检测的预处理算法研究

室内建图时,因采集的图像纹理性稀缺且噪声大,导致其匹配精确度较低。鉴于此,提出一种图像预处理方法。首先,在快速自动色彩均衡(Fast ACE)算法中引入Laplace Operator,得到一种改进Fast ACE算法,将其应用在原图像中得到增强图像。然后,使用Canny算法提取增强图像的边缘图像,并采用线性混合数学模型融合原图像与边缘图像。最后,利用AKAZE与ORB算法提取融合图像的特征点进行暴力匹配,使用随机抽样一致算法(RANSAC)进行误匹配筛选。实验表明,预处理后的图像进行匹配时,精确度明显提高。