Luminescent nanoparticles have shown wide applications ranging from lighting, display, sensors, and biomedical diagnostics and imaging. Among these, fluorescent nanodiamonds (FNDs) containing nitrogen-vacancy (NV) color centers are posed as emerging materials particularly in biomedical and biological imaging applications due to their room-temperature emission, excellent photo- and chemical- stability, high bio-compatibility, and versatile functionalization potentials. The shape variation of nanoparticles has a decisive influence on their fluorescence. However, current relative studies are limited by the lack of reliable statistical analysis of nanoparticle shape and the difficulty of achieving a precise correlation between shape/structure and optical measurements of large numbers of individual nanoparticles. Therefore, new methods are urgently needed to overcome these challenges to assist in nanoparticle synthesis control and fluorescence performance optimization. In this thesis a new correlative TEM and photoluminescence (PL) microscopy (TEMPL) method has been developed that combines the measurements of the optical properties and the materials structure at the exact same particle and sample area, so that accurate correlation can be established to statistically study the FND morphology/structure and PL properties, at the single nanoparticle level. Moreover, machine learning based methods have been developed for categorizing the 2D and 3D shapes of a large number of nanoparticles generated in TEMPL method. This ML-assisted TEMPL method has been applied to understand the PL correlation with the size and shape of FNDs at the single particle level. In this thesis, a strong correlation between particle morphology and NV fluorescence in FND particles has been revealed: thin, flake-like particles produce enhanced fluorescence. The robustness of this trend is proven in FND with different surface oxidation treatments. This finding offers guidance for fluorescence-optimized sensing applications of FND, by controlling the shape of the particles in fabrication. Overall the TEMPL methodology developed in the thesis provides a versatile and general way to study the shape and fluorescence relationship of various nanoparticles and opens up the possibility of correlation methods between other characterisation techniques.