基于多层感知器和运动轨迹的海上目标类型识别

Marine Target Type Recognition Based on MLP and Motion Trajectory

提出了一种基于多层感知器和运动轨迹的海上目标类型识别方法,设计了基于海上目标运动轨迹监测数据的实验方案,对原始监测数据进行数据清洗并人工对少量数据设置标签。建立了多层感知器网络结构模型,优化了网络参数。先使用少量带标签的样本进行实验,结果表明,多层感知器可有效进行目标分类。将训练好的网络用于多数样本分类并设置标签,之后对多数样本的数据集进行实验。实验表明,多层感知器识别速度快且正确率很高,为基于多层感知器和运动轨迹的海上目标类型识别提供理论依据。

A method of marine target type recognition based on multi-layer perceptron and moving trajectory is proposed.And an experimental scheme based on marine target trajectory monitoring data is designed.The original monitoring data are cleaned and manually labeled with a small amount of data.The MLP network structure model is established and the network parameters are optimized.Firstly,experiments with a small number of labeled samples show that the MLP network can effectively classify the target.The trained network is used to classify and label most samples,and then the data set of most samples is experimented.Experiments show that the MLP network has high recognition speed and accuracy,which provides a theoretical basis for marine target type recognition based on MLP and motion trajectory.