Early diagnosis of crop growth at various growth stages will help to make an optimum fertilization. If we can diagnose crop growth at around the time of topdressing of N fertilizer, N fertilization can be made based on crop growth and target crop yield, which may provide economic and environmental benefits as compared with fixed rate fertilization. In this study we devised methods to diagnose rice growth non-destructively at panicle initiation stage and to determine N topdressing rate. SPAD-502, Field Scout CM1000 and Green Seeker GNDVI were used to diagnose the growth status of rice grown at different soil N fertilities. The values measured by the diagnostic equipments at rice panicle initiation stage were then regressed to rice grain yield. It was found that CM1000 and GNDVI were more efficient than SPAD to diagnose rice growth. Therefore, a multivariate model with CM1000 and GNDVI values was developed to make a decision of N fertilization at rice panicle initiation stage. In a subsequent field study, N fertilization determined by non-destructive growth diagnose by CM1000 and GNDVI, and the multivariate model could minimize N fertilizer use to achieve our target yield, resulted in significant reduction of N fertilizer as compared with fixed rate N fertilization.