Granulometry is essential for understanding geological granular flows, sediment transport, and deposition processes. Traditionally, sieving has been the standard particle-size analysis method; however, it presents limitations in accuracy and field applicability, particularly in inaccessible environments. This perspective article reviews emerging technologies for granulometric analysis in geosciences, covering studies published between 1967 and 2025 from Web of Science and Scopus. Over the last three decades, digital image analysis, Structure-from-Motion (SfM) photogrammetry, and Light Detection and Ranging (LiDAR) have enabled remote, high-resolution particle characterization. More recently, artificial intelligence and deep learning have accelerated this progress. Applications based on Convolutional Neural Networks (CNNs), transformer architectures, Graph Neural Networks (GNNs), and LiDAR-derived 3D point clouds are discussed. A comparative analysis of methods is presented considering spatial resolution, accuracy, particle-size range, processing time, cost, and Technology Readiness Level (TRL). Results indicate that hybrid approaches generally outperform single-method solutions. Future developments point toward hybrid systems, Bayesian uncertainty quantification, and autonomous field operations supported by edge computing.