In order to present the reasonable crown of hot rolling mills, the bend force and shift position should be pre-computed in a reasonable manner, while maintaining a good profile. How to search the optimal crown trajectory in ratio crown range is the key problem. Crown and flatness allocation strategy is researched based on dynamic programming method which has been improved and optimized by the algorithm. The key parameters have been calculated by finite element method and regression calculation method, which can regress the equation of the mechanical crown. Comparing with a real production hot rolling mills, the result of state transition equation shows a perfect availability. Additionally, the parallel computing model and look-up table method have been utilized in dynamic programming calculation program. The results show that the bend force normalization population standard deviation fell from 0.2 to 0.06 and the continuously variable crown (CVC) roll shift position normalization population standard deviation fell from 0.3 to 0.15. The consequences show that the crown and flatness allocation strategy has a better application effect and the computing time is only 16 ms, which proves that the model can be applied online.