Abstract: An approach to the detection of underwater remote target by estimating the back-scattering coefficient of target is proposed. The estimation of the sequence of back-scattering coefficients of target and its nearby scatters is transformed into the deconvolution of system with multiplicative noise (ocean channel) in state space model, and the optimum deconvolution algorithm is presented. An adaptive state filter is built to alternatively estimate the state and the time-varying noise statistics of SMN with unknown a priori statistics of additive noise and multiplicative noise, and the adaptive deconvolution algorithm is presented. Then a weak signal with SNR equal to-6.1 dB is simulated and proceeded by means of above-mentioned two deconvolution algorithms respectively, and the final results prove the effectiveness of proposed approach.