The architecture of cooperative spectrum sensing in cognitive satellite network based on CSLNet. The architecture consists of two subnetwork models. The front subnetwork is SCLNet, which is used to extract the sensing soft features (SSFs) of the data from each SU satellite and sends the SSFs to the PU satellite for feature fusion. The rear subnetwork is Soft-FusionNet, which is responsible for fuzing the SSFs from various SU satellites and outputs the final spectrum sensing results.