In this study, we introduced a new method to address the limitations of the traditional spherical harmonic (SH) analysis when reconstructing the morphology of elongated/flattened particles. The proposed method combines SH analysis with a stretching algorithm (SH-stretching) to enable the reconstruction of irregular particles with arbitrary values of elongation index (EI) and flatness index (FI). Based on reconstruction of real cobble particles, the SH-stretching algorithm demonstrates superior performance in reconstructing elongated/flattened particles, and can ensure convergence. The SH-stretching algorithm is further integrated with principal component analysis (PCA) and coefficient standardization to develop a particle generation technique. This technique can strictly control the EI and FI distributions of the generated particles and is verified to be efficient in generating virtual cobble particles with realistic shape characteristics. Furthermore, the proposed technique allows the generation of particle aggregates with specified wide range distributions of EI and FI, while preserving their polydisperse characteristics.