Multi-target tracking based on vessel image sequences monitored on waterway is the premise of carrying out the vessel active warning and improving bridge-area ship navigation safety. Accurate tracking results are unable to be obtained by traditional Camshift tracking only based on color histogram under complex weather conditions. For this reason, a Camshift ship tracking algorithm based on color, shape and texture multi-feature adaptive fusion is proposed. In this method, the target model is realized by multi-feature adaptive fusion, which increases the reliability of observation and improves the robustness of observation model. When tracking targets, the target model based on fusion is included in the Camshift tracking algorithm. The experimental results show that the proposed tracking algorithm can provide higher correctness and stability than traditional Camshift algorithm.