In the actual detection, due to bandwidth, channel fading and channel noise, the ideal channel can not be assured from the local detector to fusion centre. In this paper, the global optimal detection algorithm is studied based on the non-ideal channels. In addition, in order to improve the detection performance and use communicated resource rationally, the soft decision is applied in local detector. According to Neyman - Pearson rule the optimal decision forms of local detectors and the fusion centre are derived. Finally, simulations of two detectors show that by using soft decision the detection performance is improved, and the worsening of the non-ideal channel absolutely depresses the detection performance of the system.