The invention discloses a rough-set-based data fusion method for a wireless multimedia sensor network. The wireless multimedia sensor network can be flexibly deployed in an area of interest of a user to sense richer multimedia information such as audios and videos. However, power is supplied to nodes of the wireless multimedia sensor network by batteries, and the hardware power consumption of the nodes for acquiring and transmitting the multimedia information is far higher than that of the conventional nodes, so that how to save the energy of the nodes to maximally prolong the life cycle of the network becomes one of main difficulties and challenges for the design and the implementation of the wireless multimedia sensor network. For such a problem, the invention provides a rough-set-based data fusion scheme for the wireless multimedia sensor network. According to the scheme, an optimal network topology structure is constructed, cluster head nodes are selected according to certain rules, and a data fusion tree is constructed among the cluster head nodes, so that acquired information can be forwarded along an optimal fusion path to reduce the energy consumption of the nodes in data transmission; and redundancy is eliminated by using an indiscernibility relation of a rough set theory to obtain reduced information, the reduced information is more accurate, and the data volume of the reduced information is greatly reduced compared with a raw data volume, so that the energy of the nodes is further saved, and the life cycle of the whole network is prolonged.