Fracture morphology has insightful information related to the residual effect of fractured structural glass, which is vital in assessing the post-fracture performance of glass members. This study experimentally characterized the fracture morphology of heat-treated glass and developed a novel method of morphology reconstruction, which aims to facilitate the numerical analysis of fractured structural glass. With the development of a computer-vision-based method for transparent objects, the morphology information from fragmentation tests was extracted and systematically investigated for monolithic heat-treated glass with various thicknesses, surface compressive stresses and fracture initiation locations, which are considered as the key influencing factors of heat-treated glass fracture. The geometrical features of fragments and their spatial distribution were quantitatively analysed, identifying their correlations with glass properties. The result indicates that the distribution of fragment centroids shows greater dispersion as the tempering level increases, and the fragments tend to be smaller and more rounded. The strain energy release at fracture was also assessed by fracture patterns, showing it presents high sensitivity to the glass thickness and surface compressive stress. Subsequently, a novel approach was proposed for the stochastic reconstruction of fracture morphology, combining feature points distribution and Voronoi tessellation concept. The control parameters are determined by data from the fragmentation tests and the influence of fracture load could be properly considered. The proposed method shows satisfactory outcomes and good agreement with the experimental records, which has further potential in developing refined numerical models by considering more realistic fracture morphology of glass members.