In the intelligent greenhouse environment control system using wireless sensor networks, the crops growing is usually affected by various factors from the environment. The greenhouse control system makes inappropriate decisions based on the received measurements of wireless sensor networks, which will lead to go against the growth of crops. In view of this, a novel decisionmaking framework based on Dempster-Shafer (D-S) evidence theory was established to meet practical requirements of the greenhouse environment control system. Moreover, two approaches of the data preprocessing and the decision fusion were proposed, respectively. Firstly, the measuring outliers data were detected by using the box-plot, and then an effective approach was proposed to correct them adaptively, which overcame the disadvantage of directly removing the outlier data in existing approaches. The corrected measuring data were also clustered by using the weighted average distance. Finally, a novel basic probability assignment approach was proposed to make correct decisions for the control of greenhouse environment based on the D-S evidence theory. Experimental results demonstrated that the outliers detection rate of the box-plot was more accurate than that of the Dixon criterion (nearly 19.2%). Compared with existing approaches, the fusion performance of the uncertainty was reduced by 1~2 order-of-magnitude in the proposed basic probability assignment approach, which not only increased precision of environmental indicators for the greenhouse control, but also accelerated the convergence process and reduced the risk of decision-makings effectively.