Particulate product manufacturing covers a wide range of industries including, for example, pharmaceutical and agrochemicals. The size distribution and the shapes of these products are important quality properties because they have a strong impact on the performance of the products and their further processing, transportation, and utilisation. Despite the existence of several commercial systems to characterise the shapes of large samples by image analysis methods, one of the major challenges is the availability of methods to effectively classify different particle shapes and identify appropriate shape descriptors to distinguish between good quality from faulty samples or batches. In practice, samples contain particles with multiple shapes often very difficult to characterise by physical shape factors. This work will compare various descriptors and also presents novel techniques for particle shape characterisation and automated classification of shapes of practical particulate samples. The techniques are based on the application of physical descriptors, such as aspect ratios, as well as latent shape descriptors, based on Fourier analysis, in combination with principal component analysis and automated classification methods. Results of the techniques will be presented with real samples of different particulate compounds obtained from laboratory and industry.