为克服传统目标识别方法在处理空间特征分布极为复杂的数据时的缺点，提出1 种基于决策树的多特征检测算法，并将其应用到基于视频的海上搜救目标检测中. 该算法首先提取图像中的颜色、亮度等信息，通过计算各特征的信息增益建立决策树，将搜救目标检测问题分解成3 层决策树分类问题. 实验表明，该算法能够提高多特征目标检测的效率，在救生艇、筏等海上搜救目标检测的应用中取得较好的结果.-Characteristics to overcome the traditional target recognition methods in dealing with space distribution of the shortcomings of the highly complex data, proposed a decision tree-based multi-feature detection algorithm, and apply it to video-based maritime search and rescue target detection. The algorithm first extracts the image in the color, brightness, and other information, decision trees, established by the information gain to calculate the characteristics of the rescue target detection problem is decomposed into three layers of the decision tree classification. experiments show that the algorithm can improve the efficiency of the multi-feature object detection in the lifeboat, to obtain better results in the raft and maritime search and rescue of target detection applications.