The Advancing Layer Method (ALM) remains one of classical approaches for mesh generation in Computational Fluid Dynamics (CFD), yet it encounters significant challenges when processing complex geometric walls with pronounced concave-convex features. Specifically, it is difficult to maintain grid continuity and orthogonality due to cell distortion and overlapping in these special geometric regions. To address these limitations of the conventional ALM, this study introduces an improved ALM with geometric anisotropy adaptation for two-dimensional (2D) complex configurations. The proposed method comprises two components: a potential field drives the front advancement to satisfy gird orthogonality, while geometric anisotropy adaptation dynamically adjusts grid density distribution to satisfy grid continuity. This method achieves growth-type adaptive mesh generation without requiring domain partitioning. Ultimately produce all-quadrilateral meshes with superior orthogonality. Validation through diverse 2D test cases, including the F-shaped wall, canard missile and airfoil with grooves, confirms the method's effectiveness. Notably, the approach exhibits high automation in handling localized complex aircraft geometries, along with good robustness across varying geometric configurations.