We present a novel mathematical framework that encodes high-energy collision events as radial mapping images suitable for rigorous analysis and machine learning. Each detected particle is represented by a disk on a circular plot: its angular coordinate corresponds to azimuthal angle, its radial coordinate to pseudorapidity, and its radius to energy. This construction embeds each event into the space of compactly supported measures on the plane, enabling the application of measure-theoretic tools and geometric functional analysis. We train a lightweight convolutional neural network on these measure-based images using publicly available CERN Open Data and validate performance against Monte Carlo simulations, achieving classification accuracy above 90\%. We prove stability bounds showing that small perturbations in angular and energy parameters yield only small changes in the embedding, ensuring robustness. These results demonstrate how carefully designed geometric embeddings and supervised-learning techniques can unlock insights from massive scientific datasets, with potential applications across data-intensive domains.