【摘要】 为了解决动态红外点目标多光谱模式识别问题,提出了一种利用神经网络并行子网作为前级处理,证据理论于后级融合的多周期模式识别推理模型。由于并行子网的引入,该模型避开了识别过程中采用单一神经网络所带来的大样本训练问题,用带有加性噪声的点目标红外光谱作为识别模型的目标数据源进行了算法验证,计算结果表明该算法对多周期不确定性证据有很强的证据聚焦能力。更多还原

【Abstract】 In order to solve dynamic small weak infrared object multi spectrum data fusion problem, a multi period pattern recognition model has been established by using ANN parallel nets and D S theory. This model solves the training time problem for single net with large numbers of samples because of importing of the parallel nets. The model was validated using dot infrared spectrum data with additive noise as target data source and the result of calculation indicate the recognition model has very strong multi period uncertainty evidence focusing capability.更多还原