基于遗传神经网络的伪装效能评估模型

Evaluation Model for Engineer Camouflage Effectiveness Based on Genetic BP Neural Network

在量化样本光学特征指标并划分伪装效能等级基础上,采用基于遗传算法的神经网络建立伪装效能评估模型。其步骤包括确定GA算子及相关参数,初始化网络连接权值和阈值向量,计算各个体适应度函数并将其排序,执行遗传操作,最后用神经网络进行二次训练。将样本光学特征指标量化值作为神经网络输入值,量化后的样本等级作为神经网络教师值进行评估。仿真表明该混合算法收敛速度快,能有效避免局部极值问题。

The genetic algorithm neural network model for evaluating camouflage effectiveness is created based on evaluated optical characteristics index values of stylebooks and dividing camouflage effectiveness classification. The steps includes determining GA operators and relevant parameters, initializing weights and thresholds vectors of network, calculating and ranking the fitness values of each individual, executing genetic operation and finally retraining the network. The optical characteristics index values are treated as the input of network, while the camouflage effectiveness classification values are regarded as the output. Simulation result proves that the hybrid algorithm has fast convergence and the partial extremum problem can be effectively avoided.