本論文使用遺傳演算法配合兩階段法來重建一均勻或多層介電物體的微波影像。比較電場的量測值和每一物體的電場計算值, 可以成功地逆推出介電物體的位置、大小及形狀。本方法比先前單獨使用的形狀突變法更快收斂, 且只要信號雜訊比小於-20分貝, 重建效果即相當良好。 A computational method combining the genetic algorithm(GA) and a two-phase technique is reported for electromagnetic imaging of a homogeneous and/or multicylinder dielectric object of arbitrary shape. By measuring the scattered field, the shape, location, size and permittivity of the object are retrieved quite successfully. It is found that the proposed two-phase technique converges faster than a previously reported method using shape mutation scheme does. Numerical simulation shows that good reconstruction can be obtained for a homogeneous and/or multiple-cylinder object as long as the noise level is .ltoreq.-20dB.
一九九八年全國電信研討會論文集=Proceedings of 1998 National Symposium on Telecommunications，頁459-464