Mobile crushers produce, depending on the used crusher settings, different particle size distributions. Information concerning the particle size of the produced aggregate is necessary for optimisation and quality control of the crusher setup. Today there are no systems in use in mobile application to deliver information about the produced grading curve. Therefore, a research has been executed regarding the best available technology for mobile applications. Based on this research a system for delivering information about the particle size distribution of a monolayer of processed aggregate was developed for RUBBLE MASTER HMH GmbH in cooperation with the Software Competence Center Hagenberg. The stereo camera Intel RealSense D435 was used for image acquisition and the provided depth data of it was passed to two algorithms developed in cooperation with the Software Competence Center Hagenberg. One of the algorithms is based on a classic vision approach as described in the best available technology. The second algorithm consists of a trained deep convolutional neural network which can estimate the particle size distribution directly from the depth images. Evaluations of the algorithms showed that the classic vision approach worked best on single stone scenarios. When acquiring depth data of more than one stone of a monolayer the deep neural network outperformed the classic vision approach. Regarding the passing sum of the particle size distribution the classic vision system resulted in a root mean square error in total of 0,138 and the deep neural network reached 0,093.