中文摘要：

全球造山带及中国大陆中西部普遍具有强烈起伏的地形条件．复杂地形条件下的地壳结构成像问题像一面旗帜引领了当前矿产资源勘探和地球动力学研究的一个重要方向．深地震测深记录中反射波的有效探测深度可达全地壳乃至上地幔顶部，而初至波通常仅能探测上地壳浅部．为克服和弥补初至波探测深度的不足，本文基于前人对复杂地形条件下初至波成像的已有研究成果，采用数学变换手段将笛卡尔坐标系的不规则模型映射到曲线坐标系的规则模型，并将快速扫描方法与分区多步技术相结合，发展了反射波走时计算和射线追踪的方法．进而利用反射波走时反演，实现起伏地形下高精度的速度结构成像，从而为起伏地形下利用反射波数据高精度重建全地壳速度结构提供了一种全新方案．数值算例从正演计算精度、反演中初始模型依赖性、反演精度、纵横向分辨率以及抗噪性等方面验证了算法的正确性和可靠性．

英文摘要：

Global orogenic belts and the regions of middle and western China have strongly varied topography. Crustal structure imaging in such condition is of great interest in both the mineral resource exploration and the geodynamies study. The probing depth of seismic reflection waves in deep seismic sounding can reach up to the whole crust and even the top of the upper mantle of the earth, which is an effective supplement to the first-arrival refraction waves traveling in shallower depth. Current methods to deal with irregular surface usually take the step shape approximation or use a low velocity layer to overlay the surface. However, these methods may cause accuracy loss or even imaging distortion. This paper aims to introduce a grid-based method for traveltime tomography using the reflection traveltimes, to invert for velocity structure of the whole crust with an irregular surface. To deal with the irregular surface problem, this study is based on an irregular surface flattening scheme. The irregular surface flattening, which involves the transformation between the curvilinear and the Cartesian coordinate systems, can describe rugged terrain with no accuracy loss. Wave-front traveltime is calculated using the fast-sweeping method （FSM） to solve the topography-dependent Eikonal equation, and a multistage technique is then applied to calculate the reflection waves by reinitializing FSM in the incident layer. Ray paths are found by following the steepest traveltime gradient from the receiver to the interface and then to the source. Using the back-projection algorithm, the slowness perturbations are obtained, which have a relationship with the already existing traveltime. The final velocity model is obtained after a few iterations during which the slowness is updated, until the fit of the reflection traveltimes is satisfactory. We take four numerical examples to verify our method. First, the perfect match of traveltimes and ray paths obtained by our method and the shortest path method verifies that our forward mode