基于随机森林动态集成的水下尺度目标识别

Underwater Target Recognition Based on Dynamic Ensemble of Random Forest

正确识别目标是水下声自导武器攻击敌方目标的关键。文中提出一种基于动态选择集成技术的水下声自导武器实时目标识别方法。利用水下声自导武器主动宽带探测波形照射下目标不同的反射特性,从目标宽带相关检测输出提取了能量分布和空间分布统计特征,并构建了基于随机森林的动态选择集成模型,利用海试数据集进行训练与测试。仿真分析表明,文中所提出的动态集成模型识别效果优于其他分类算法,可以较好地应用于水下声自导武器目标识别中。

Accurate recognition of the target is the key to attacking enemy for underwater acoustic homing weapon.A real-time target recognition method for underwater acoustic homing weapon was proposed based on dynamic ensemble selection technology.The statistical features of energy distribution and spatial distribution were extracted from the output of target wideband correlation detection by using the different reflection characteristics of the target irradiated by the active wideband detection waveform of the underwater acoustic homing weapon.In addition,a dynamic ensemble model based on a random forest was constructed,and it was trained and tested on the marine dataset.The simulation analysis shows that the dynamic ensemble model proposed in this paper has better recognition effects than other classification models and can be applied to target recognition by underwater acoustic homing weapon.