A ship detection algorithm based on human vision perception is proposed. The key idea is using designed Gabor filters to simulate human vision perception for dealing with visible images, in order to reduce the negative influence of complex sea surface background and enhance the dark ship target. This algorithm includes several models to realize the separating of land and sea areas, the wiping of the cloud and the reduction of false alarms ratio using color, texture and morphologic characters. Experiment results show that the algorithm can effectively detect the ship targets in the complex sea environment, especially for dark ships, and the missed alarms ratio and false alarms ratio are very low, the ship target detection probability can be as high as 85%, higher than the 51.3% in traditional algorithm.