一种障碍物重构的混合方法

A Hybrid Method to Recover the Scattering Obstacles

针对不可穿透障碍物的重构问题,提出了基于朴素贝叶斯和宽度学习系统的混合方法。该方法首先利用朴素贝叶斯对训练集按照形状参数分类,将分类后的训练集作为模型输入,利用激活函数作用得到特征节点,再将所有特征节点和随机生成权重的增强节点作为整体,通过线性映射连接到输出端,然后求解伪逆得到网络参数,最后对基于朴素贝叶斯和宽度学习的混合方法进行测试,实验结果表明,该方法可以同时重构散射体形状和位置,并且具有收敛性和鲁棒性。

We propose a hybrid method for the reconstruction of impenetrable obstacles based on Na?ve Bayes and Broad Learning System.Firstly,we classify the training data through the different shape parameters by Na?ve Bayes.Secondly,the training set after classification is used as the input of the model,and the feature nodes are generated after activation function.All feature nodes and enhancement nodes which has random weights are connected to the output by a linear mapping.Finally,the pseudo-inverse is solved to obtain the network parameters,and the hybrid method is tested.Some experiments fully show that this method can reconstruct the shape and position of scatterers simultaneously,and has convergence and robustness.