Ship surveillance is important in maritime management. Space-borne Synthetic Aperture Radar (SAR), High Frequency Surface Wave Radar (HFSWR) and the Automatic Identification System (AIS) are three main sensors for the ship surveillance of large maritime areas. Fusion of these sensors' measurements can produce an accurate ship image distribution in a surveillance area. Data association is fundamental to data fusion. A Maximum Likelihood (ML) association algorithm with multi-feature improvements is proposed to increase detection accuracy and reduce false alarms. The tested features are position, size, heading and velocity. First, the ship measurement model is established. Then, the problem of data association for SAR, HFSWR and AIS is formulated as a multi-dimensional assignment problem. In the data assignment process, Jonker-Volgenant-Castanon (JVC) and Lagrangian relaxation algorithms are applied. Simulation results show that the algorithm proposed here can improve the association accuracy compared with the Nearest Neighbour (NN) and the position-only ML algorithms, using the additional features of length and velocity. Real data experiments illustrate that the algorithm can enhance target identification and reduce false alarms.