Spatial omics has rapidly expanded with increasingly diverse imaging modalities, molecular targets, and chip sizes. However, no general framework currently exists to construct cell level matrices that are robust across platforms and omics types. Here we present CellBin, a universal and scalable frame- work that unifies image stitching, cell segmentation, and spot-to-cell mapping for multiple spatial omics technologies. CellBin integrates a multi-field weighted stitching algorithm for large-area images, a fam- ily of U-Net–based models trained across diverse staining modalities, and an optimized computational architecture for high-throughput processing. Across five technological platforms and three omics data types, CellBin achieves robust segmentation and accurate single-cell matrix construction, consistently outperforming seven state-of-the-art methods in F1-score, cell size precision, and annotation accuracy. By providing a generalizable, cross-platform solution, CellBin bridges multiple spatial omics, enabling unified, high-resolution cell level analyses across technologies.