The space-based automatic identification system (AIS) has attracted much attention because the system can receive AIS signals from ships on the high seas. As its serious problem, multiple AIS packets can collide at one satellite because it has a wide coverage area. Parallel interference cancellation (PIC) based on Viterbi algorithm (VA) is very effective in separating the collided packets and shows an excellent bit error rate (BER) performance under the condition that Doppler frequencies, channel impulse responses, and propagation delay time are ideally estimated. When the estimation of those channel parameters is not sufficiently accurate with desired to undesired signal power ratio (DUR) being low, however, the BER of PIC can be seriously degraded. In this paper, we propose a novel iterative channel estimation scheme for each channel parameter to improve the estimation accuracy even under the low DUR condition. The least squares and quasi-Newton methods are applied to the estimation of the channel impulse responses and Doppler frequencies, respectively, besides coarse estimation using correlators. The correlation-based estimation of the propagation delay time is repeated. Computer simulations demonstrate that the proposed scheme can estimate the channel parameters more accurately, which leads to remarkable improvement in the BER performance.