本論文研究以有限時域差分法(FDTD)為基礎，利用基因演算法(Genetic Algorithm)來重建埋藏柱體之電磁成像問題，並探討完全導體柱在半空間中的逆散射問題。 於半空間環境中，將一個未知形狀與位置的二維導體柱埋藏於下層介質中，並由第一層介質(空氣)入射電磁波(高斯脈波或雙極性脈波)照射掩埋在下層介質中的未知物體，於第一層利用十五組接收點量測其散射場，利用接收到的散射場再應用基因演算法，將逆散射問題轉化為求解最佳化問題。並探討當加入不同程度之雜訊於散射場中，對重建導體柱影像的影響。於正、逆散射中使用三次仿樣函數(Cubic spline)來描述未知物體之形狀，並利用次網格技術使柱體的形狀更為圓滑。 基因演算法是一種模擬自然界生物進化的最佳化搜尋方法，此法的優點可突破傳統最佳化法的解答只能收斂於局部極小值，而非總體最小值，就算最初的猜測值與實際值相距甚遠，但仍可以求出準確的數值解，成功的重建出柱體的位置與形狀。 This paper presents an FDTD-based time domain inverse scattering problem is investigated. The genetic algorithm is used to reconstruct the microwave image of perfect conductor cylindrical object, which are buried in a half-space stratified material medium. Assume a perfect conductor cylindrical object is buried in the second layer stratified material medium. The electromagnetic wave source (Gaussian pulse or bipolar pulse) located in the first medium (air) is used excite to illuminate the problem space, and then the scattered electric field are measured in the same medium. The genetic algorithm is used to covert the inverse scattering problem into an optimization problem. The measure scattering E fields are compared with the calculated E fields obtained by FDTD method. When the time domain waveforms of the two fields are getting closer, it means the reconstructed object is converging to the original one. In order to more effectively describe an unknown cylinder with arbitrary shape, used the spline function to describe the object shape and we also used the subgridding technique to model the cylinder shape more smoothly. The genetic algorithm (GA) is global optimization emulation the natural evolution procedure-survival of the fitness. The main advantage of the GA is to overcome the convergence to local minima as the traditional optimization schemes usually do. Therefore, even the initial guesses are far from the real solution the object properties such as position and shape can be reconstructed successfully.
