基于决策树的海上搜救目标检测算法

Marine rescue target detection algorithm based on decision tree

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.

In order to overcome the shortcomings of the traditional target recognition method in dealing with the data of highly complex spatial characteristics distribution,a multi-feature detection algorithm based on decision tree theory is proposed and applied to the marine rescue target detection based on video.In the algorithm,color,intensity and other information of the images are extracted firstly,then the deci-sion tree is built by calculating the information gain of each feature,thus the detection of the rescue target is transformed into the classification process of three-layer decision tree.Experiments indicate that the al-gorithm can improve the efficiency of multi-feature target detection,and it works well in the rescue target detection such as lifeboat and life raft.