采用粗BP神经网络和D-S证据理论的目标识别

Target recognition method based on combination of rough BP neural network and D-S evidence theory

为克服传统的目标识别方法的不足,提高目标识别的实时性和准确性,提出将粗BP神经网络与D-S证据理论相结合的识别模型。在多传感器数据融合中利用粗集理论对大量的传感器数据进行处理,对输入信息进行约简,剔除冗余信息,简化了生成规则和BP神经网络模型结构,提高了网络训练速度和运行速度。以BP神经网络输出作为证据,后端对不同传感器的证据用D-S证据理论进行融合,得到待识别目标的识别概率。实验表明该模型减少了识别的主观因素,简化了BP神经网络结构,提高了运算速度和识别效果。该混合模型有比较好的应用前景。

In order to overcome the shortages of traditional target recognition method and improve the real-time performance and precision of target recognition, a target recognition model is put forward which combines the rough BP neural network with D-S evidence theory. Rough BP set theory is used to deal with the great deal of data from sensors in multi-sensor data fusion, predigest the input information and eliminate the redundant information, thus can simplify the BP neural network model structure and improve the network training speed and operation speed. Then, taking the output information of the BP nerve network as evidences, the D-S evidence theory is used to fuse the evidences of the different sensors, and recognition probability of the target is obtained. The experiment shows that this model reduces subjective factors in recognition, simplifies the BP neural network structure and improves speed of operation and recognition.