病态条件下的逆卷积问题及其多参数最佳化解

On Deconvolution Effective for Detectionof Returned Signals under IllPosed Conditions

给出一个多参数最佳化解决方法，该方法采用模拟韧化最佳化技术，将病态逆卷积问题分解为多参数最佳估计问题，成功地解决了相应的病态条件问题，最后用一个简单的计算机仿真例子说明了该方法。

In many applications, including radar, returned signals from target need to be deconvoluted for ascertaining its characteristics. Frequently illposed conditions make deconvolution extremely difficult. We present a simulated annealing algorithm that was found to be much better than existing algorithms. The central idea for new algorithm was reached only after repeated analysis; it is that we replaced the traditional singleoperationstructure simulated annealing algorithm with a multioperationstructure one. Multioperationstructure meant that for each of several parameters we selected suitable corresponding annealing procedure, pertubation rule, and acceptance probability. Specifically, we constructed cost function to be eq.(8), and selected perturbation rule to be eq.(9), and acceptance probability to be eq.(10). With new multioperationstructure algorithm, we obtained very good deconvolution results for two different examples as shown in Figs.2 and 4 respectively. They are significantly better than corresponding results obtained with CIF method (Fig.1) and reblurring method (Fig.3).