Abstract: The problem of the optimisation in distributed systems had taken an important place for the estimation of the detection threshold. As a result, a big variety of mathematical methods are proposed in literature, in an attempt to achieve the optimum without any prior assumptions. Recently Genetic Algorithms (GAs) were proposed and processed as an optimization tool for a large variety of domains. We propose in this study, an optimization of the CFAR (OS-CFAR and CML-CFAR) threshold by an EMS-GA in non homogeneous backgrounds, for which the environment is characterized by the presence of interfering targets. The EMS-GA was applied to estimate the order statistic (K) and the multiplied factor (T) in a distributed system that contains more than one detector, then the performance of such method is analysed in different situations of multiple targets case. In spite of the efficiency and flexibility of the GA to resolve such problems the CML-CFAR system has given best results over the OS-CFAR system. In the other hand the overall probability of detection was largely influenced by the presence of interfering targets and the best results were found in the case of the OR fusion rule on a wide interval of SNR (Signal to Noise Ratio). The increase of the number of detectors in a distributed system improves its performance and the quality of the detection is affected in the sense of an increase of the detection probability in a critical situation with a presence of big number of interfering targets that saturate all the detectors.