An Embodied Intelligence Interaction System for the Blind or Low-Vision Populations Based on Large Multimodal Models.