Today, the importance of electronic warfare in the field of defense is not hidden. Among the various techniques of electronic warfare, chaff can play an effective role, which we discussed in this research. The chaff has many capabilities. Among these capabilities, it can be mentioned that it is ease to make and use, low cost jamming and wide band frequency coverage. It can also refer to deception in angle and distance, angle tracking error, and effect on the victim's radar in time, frequency, and polarization domain. Many features can be used for discrimination of chaff from target (ship). In this research, among the various features to discrimination it, the fluctuation feature, which is the variation of radar cross section, has been used. Also, in this research, in order to achieve better discrimination rate and less error, different deep learning algorithms and networks have been used. In addition, while we were able to discrimination the target of chaff based on deep learning, we achieved better recognition rate than the works that have been published so far. For example, in this study, we achieved a 15% improvement in performance compared to the resent published articles in the signal to noise ratio power of 6 dB and in the signal to noise ratio of 12 dB, the discrimination rate is 99.5%. Furthermore, we proposed the most suitable structure among the various deep learning structures in this application.