基于多准则神经网络优化的Compton背散射成像  

Multicriteria neural network approach to Compton backscatter imaging

作　　者：徐茂林[1] 于建华[2] 汪元美[1] 

机构地区：[1]浙江大学生物医学工程系,杭州310027 [2]北京科技大学数学力学系,北京100083

出　　处：《核技术》2005年第1期63-68,共6页Nuclear Techniques

摘　　要：详细阐述了基于散射能量谱的行扫描Compton背散射成像原理,将成像方程归结为一个大型非线性方程组,然后用基于多目标优化的神经网络来求出重建问题的解。数据模拟结果表明,所构造的神经网络对Compton背散射多准则成像是可行的,适于处理不适定重建问题。A principle of the line-scan Compton backscatter imaging based on the scattered photon energy spectra is presented. This imaging equation is concluded as a large-scale system of nonlinear equations. In view of the neural network with multi-criteria optimization, a solution of image reconstruction is worked out accordingly. Results of data simulations illustrate the fact that it is feasible to adopt the method of constructing neural network for the sake of multi-criteria Compton backscatter imaging, which is applicable to solving the ill-posed reconstruction problems.

关 键 词：Compton背散射 神经网络 多准则优化 

分 类 号：TN911.73[电子电信—通信与信息系统]