In the paper, the multi-node underwater acoustic localization problem is considered. For the problem, traditional localization schemes first process the received data of the nodes to estimate the time-of-arrival parameters of the received signals, and then use the estimated time-of-arrival of the signals to solve for the source location. Since each node achieves the estimation of time of arrival independently, the fact that the signals of each node correspond to the source location is ignored. To the end, a maximum likelihood localization method that uses the output of the matched filter as input to determine the source location directly is proposed. A likelihood function is designed based on the relationship between the source location and the output of the matched filter, which considers the fact that the signals at each node all correspond to the source location in order to fully capture the processing gain of multiple nodes. The sound source location is determined based on the designed likelihood function and maximum likelihood estimation theory. Finally, the localization precision of the proposed method is theoretically analyzed through an accuracy analysis to verify the effectiveness of the proposed method.