Abstract: The particle-size distribution of ore aggregates is a key parameter for optimizing mineral processing techniques and a prerequisite for determining resource recovery rates and the utilization efficiency. At present, vision-based particle-size determination technologies are predominantly used during transport with minimal obstruction; however, their detection accuracy decreases significantly when applied to the flow of high-density, severely interlocked ore particles. To address these issues, this article proposes a rapid method for determining the particle-size distribution of ore aggregates using a flexible vibrating disk. First, orthogonal and single-factor experiments were conducted to systematically investigate the effects of vibration parameters on the dispersion behavior of different ore particle groups. Subsequently, an optimal particle dispersion state was established based on the derived principles. Building upon this optimized dispersion state, a U-Net network was used for the precise segmentation of particle boundaries in the images of dispersed ore particles. The results indicated that vibration frequency is the key parameter governing dispersion effectiveness and significantly reduces particle occlusion and agglomeration by inducing resonance within ore aggregates. Further, particle-size detection results obtained from the U-Net-segmented images of optimally dispersed ore particles exhibited a maximum error of <4.3% compared with those obtained using traditional screen sieving. Thus, the method achieved improved reliability of particle-size detection by mitigating particle stacking and occlusion at the physical source, stabilizing ore particle distribution prior to image detection and minimizing interference with segmentation models. The proposed method enhances the adaptability of segmentation models to complex transport conditions without requiring complex modifications to existing vision-based models and provides technical support for efficient, rapid, and offline detection of ore particle-size distribution.