Plant phenotyping has evolved from manual observation to intelligent analysis across four distinct eras. Here we conceptualize this roadmap as Phenotyping 1.0 to 4.0. While the high-throughput advantages of Phenotyping 3.0 have led to a "data tsunami" and "knowledge islands", Phenotyping 4.0, which is driven by artificial intelligence (AI) foundation models, intelligent swarms and digital twins represents the core trend in precision breeding. We further identify three core bottlenecks hindering Phenotyping 4.0: inadequate data standardization, limited model generalization, and low-throughput molecular phenotyping with insufficient multi-omics integration. To address these, we propose targeted strategies, including establishing full-process data standardization, developing cross-crop and cross-environment generalized models, and upgrading molecular phenotyping platforms integrated with mechanism-guided hybrid models. Together, these advances will redefine the operational architecture of Phenotyping 4.0 and facilitate its functional integration with Breeding 4.0, thus accelerating the shift from agricultural digitization to data-driven intelligence.