Reservoir stochastic simulation usually can be divided into two categories: object-based modeling method and pixel-based modeling method. For fluvial reservoirs, the complex geometry of sedimentary microfacies, such as the morphology of channels and natural levees, can be reproduced better by using the object-based methods. But these object-based approaches also have difficulty in conditioning models to dense well data. Most of previous conditional methods are iterative optimization algorithms, which take a long time to reach convergence when there are many conditional data, and they generally obtain a low degree of conditionalization. Therefore, we develop a new conditional methodology. The process of conditioning to well data is as follows. First, the classical object-based method is used to establish the channel reservoir model. Based on this, the distance parameter field (D1) related to the center lines of channels is established. Next, we convert the information of channel and nonchannel wells into the distance from the center lines and then set up the distance parameter field (D2) by kriging and other methods. Finally, D1 is modified by using the logical relationship between D2 and D1, and then D1 is truncated to obtain the river channel model. The analysis of the results of three examples indicates that this new approach can improve the conditional level greatly.