Most algorithms based on Constant False Alarm Rate (CFAR) which are used to detect ship targets often produce problems such as time waste,lower efficiency,missing detection and so on.To solve the problems caused by slipping windows,an algorithm based on CFAR and adaptive windows is proposed.Firstly,global saliency is measured to find the target potential regions.According to adaptive windows setting condition,target regions are divided into inside regions and edge regions: background parameters after getting rid of target pixels are used to detect inside regions;targets detected and global background parameters are united to detect edge regions.Final result is achieved by getting the two parts together to subsequent processing.The experiments show that the algorithm proposed in this paper can locate regions of targets fast and truly,depict the shape of the regions accurately and perform well for ship detection.