随着探测对象日趋复杂，适应于起伏界面、复杂结构的逆时偏移成像算法得到了不断发展，此成像方法可高精度重构无法观测区域的物质形态，然而由于此算法存在计算效率低和低频噪声的问题而降低了其实用性。因此本文主要针对这两个问题进行研究。

针对逆时偏移算法存在计算效率低的问题，本文对有限时域差分算法（FDTD）和伪谱时域算法（PSTD）接收波形进行仿真对比，表明前者达到数值稳定性的要求为一个波长内至少解剖12个网格密度、而后者只需要2个网格即可，表明伪谱法可以大网格剖分，从而降低了空间存储量和计算量，计算效率得到大幅度提高；此外，在满足上述两种算法各自精度要求下，计算出它们接收波形之间的最大相对误差为1.13%，平均误差为0.61%，因此用伪谱法代替差分法进行数值模拟是一种有效的方法。针对高电导率介质中的电磁波传播衰减问题，本文将改变电导率正负号的衰减补偿方法应用到逆时偏移成像中，保证电磁波进行逆时传播时的波动方程对称性，仿真波形表明衰减补偿逆时偏移方法能够很好地恢复由电导率造成的信号衰减，从而提高了成像的分辨率。针对逆时偏移算法成像中存在低频噪声现象，本文从噪声产生的机制出发，首先分析了基于时间窗互相关成像条件的逆时偏移算法，通过与传统互相关成像结果对比，说明了此算法有较好的去噪效果；接着从传播过程中去噪采用了波场分离的方法和从成像后滤波采用了拉普拉斯滤波的方法，并将这两种去噪方法进行结合，通过对比仿真结果表明波场分离结合拉普拉斯滤波是一种相对较好的去噪方法。

最后本文将改进的逆时偏移成像算法应用到隧道病体探测成像中。通过对三种隧道病体结构进行仿真成像，观察到此算法可以对隧道病体结构进行清晰成像，表明改进的逆时偏移成像算法可以为隧道病体成像研究提供理论支持。

 As the detection object becomes more and more complex, the reverse time migration(RTM) adapted to the undulating interface and complex structure has been continuously developed. This imaging method can reconstruct the material morphology of the unobservable area with high precision. However, this algorithm has low computational efficiency and low wavenumber atifact, which reduces its practicality. Therefore, This thesis  focuses on these two issues.

To solve the problem of low computational efficiency of the RTM, This thesis simulates and compares the received waveforms of the finite difference time domain(FDTD) method and the pseudospectral time-domain(PSTD) method, indicating that the former requires numerical stability to dissect at least 12 grid densities within a wavelength,while the latter requires only 2 grids. It shows that the PSTD can be divided into large grids, thereby reducing the space storage and calculation amount, and the calculation efficiency is greatly improved; In addition, under the respective accuracy requirements of the above two algorithms, the maximum relative error between the received waveforms is calculated to be 1.13%, and the average error is 0.61%,so it is effective to use PSTD instead of FDTD for numerical simulation method. Aiming at the problem of electromagnetic wave propagation attenuation in high-conductivity media, This thesis applies the attenuation compensated method that changes the sign of conductivity to RTM to ensure the symmetry of the wave equation when the electromagnetic wave propagates in reverse time. The simulation waveform shows that the attenuation compensated RTM can restore the signal attenuation caused by the conductivity well, thereby improving the imaging resolution.In view of the phenomenon of low-frequency noise in the imaging of the RTM, starting from the mechanism of noise generation, This thesis first analyzes the RTM based on the time window cross-correlation imaging conditions. The comparison with the traditional cross-correlation imaging results shows the algorithm has better denoising effect, and then the wave-field decomposition method is applied from the denoising during the propagation process and the Laplace filter method is applied from the post-imaging filtering,and the two denoising methods are combined. The comparison of the simulation results shows that the wave-field decomposition combined with the Laplace filter is the best denoising method.

Finally, This thesis applies the RTM of improved to tunnel disease detection imaging. By imaging three types of tunnel disease structures, it is observed that this algorithm can clearly image tunnel disease structures, indicating that the RTM of improved can provide theoretical support for tunnel disease imaging research.