Abstract: The process of rice cultivation can be treated as a tracing system and divided into five stages: namely, seedling, early tillering, late tiilerin, booting and yield format ion.Five control factors( basal fertilizer, seedling density, tiller fertilizer, panicle fertilizer and grain fertilizer),two ecological factors( illumination and temperature ) and the effects of which on the. rice population, characteristics (leaf color, plant height, number of tillers) have been studied in this paper.Based on the above factors,13 regression models were established and dynamic programming has been applied to optimize the agronomic strategies for rice cultivation according to the different ecological conditions and two indicator functions (net income and yield).The results showed that, the leaf color, plant height and number of tillers could reflect the dynamic criteria of the population quality of the rice. The dynamic programming model based on the regression modes of rice growth possessed a tracing control property and had better effect on the optimization of agronomic strategies of rice production. The optimal agronomic strategies proposed by the dynamic model under normal ecological condition for rice variety"Ya-ng Geng No.2"were:Basal fertilizer 62 kg/ha; Seedlinh density 188 ×104/haTiller fertilizer 92 kg/has Panicle fertilizer 60kg/ha.Grain fertilizer 45 kg/ha.The expected yield be 8042 kg/ha with a net income 3562 yuan/ha.