针对径向基神经网络在激光图像分类识别中识别率低及训练时间长的问题，提出粗糙集与神经网络相结合的方法，将粗糙集算法简约后的样本特征作为神经网络的前置输入。首先建立不同视点的激光主动成像三维仿真图像，然后提取17个目标特征，并采用粗糙集算法选择分类的属性，从17个特征中筛选出5个影响决策的特征属性，最后选用4层径向基神经网络作为基本的网络结构，并采用在各层节点上与粗糙集相结合方法识别目标。仿真结果表明，结合粗糙集的集成神经网络方法识别正确率保持在80％以上，与未结合粗糙集的神经网络相当，但训练与识别时间缩短10倍以上。%In order to improve the low recognition rate and long training time of radial basis neural network in laser im-age identification and classification,a new method based on rough set and neural network is applied,which uses the feature simplified by rough set as pre-input of neural network. Firstly,the laser active imaging three-dimensional sim-ulation images of the different viewpoints are established. And then 17 target features are extracted,and 5 character-istic attributes that affect decisions are selected using rough sets algorithm from 17 features. Finally,the 4 layer RBF neural network is used as the basic network structure and is combined with the rough set method to identify the target on each layer nodes. The simulation results show that correct recognition rate of this method reaches more 80%,which is comparable to neural network without combining rough set,but the training and recognition time is shortened by more than 10 times.
展开▼
