本論文就柱型導體逆散射問題做了三個面向的探討。 第一個部分，考慮在不同環境下一個未知形狀及表面可變導電係數 的非完全導體，對於非完全導體之邊界條件，可藉由表面阻抗的概念配 合導體表面感應電流的觀念，可導出非線性積分方程式，繼而利用動差 法求得正散射公式。利用正散射公式，我們可以得到散射場的相關資料。 對於逆散射部分，我們引進了基因法則(genetic algorithm)。利用基因 法則時，我們適當地選取參數形式，同時結合所求的散射公式，由此即 可求出散射場的相關資料，藉以求得柱體的形狀函數與導電係數。 第二個部分，藉由傳統的Fourier series 以及cubic-spline 描述 物體外觀形狀，探討在基因法則下，對於不同環境逆散射問題的形狀重 建的適用性，實驗顯示，利用Fourier series 描述複雜形狀往往造成無 法得到收斂解，cubic-spline 的描述方式則可在較少的變數個數，獲得 良好的形狀描述結果。 第三個部分，藉由使用改良的基因法則(NU-SSGA)與傳統GA 比較， NU-SSGA 可藉由大幅減少計算正散射的次數，獲取遠優於傳統GA 的效 能，以往利用全域解演算法解逆散射問題最令人詬病的收斂時間，實驗 結果顯示，利用NU-SSGA 求解可獲得大幅改善。 The thesis presents three related aspects of computational approach to the imaging of a conducting cylinder. In the first one, an imperfect conducting cylinder of unknown shape and variable conductivity is considered. Two different cases of inverse problem in free space and half space have done respectively. In the second one, cubic-spline method and trigonometric series for shape description are used and compared in several different situations (half space, partially immersed, slab medium, and periodic conductor in free space). In the third one, the inverse scattering problem is addressed to discuss the CPU time for reconstructing a perfectly conducting cylinder for two different cases (half space and slab medium). It is solved by the improved Steady State Genetic Algorithm (SSGA) and Simple Genetic Algorithm (SGA) and the consuming time in finding out the global extreme solution of the objective function is compared. Based on the boundary conditions and the measured scattered field, a set of nonlinear integral equations is derived and the imaging problem is reformulated into an III optimization problem. In the first one, the genetic algorithm is employed to find out the global extreme solution of the objective function. Numerical results demonstrate that, even when the initial guess is far away from the exact one, a good reconstruction has been obtained. In the second one, the shape of the scatterer described by using cubic-spline method can be reconstructed. In such case, Fourier series expansion will fail. Numerical results show that the shape description by using cubic-spline method is much better than that Fourier series. In the third one, it is found that the searching ability of SSGA is much powerful than that of the SGA. The consuming time for converging to a global extreme solution by using SSGA is much less than that SGA. Numerical results show that the image reconstruction problem by using SSGA is much better than by SGA in time consuming.