基于双BP神经网络数据融合的水声定位研究

Acoustic Locating Based on Double-BP Neural Network Data Fusion

针对多传感器水声定位的数据融合问题,对目前常用的基于BP神经网络的数据融合算法进行改进,提出了基于双BP神经网络数据融合的水声定位算法。该算法首先将传感器采集回来的数据依次输入到异常数据识别网络,进行训练,剔除异常数据点;再将异常数据识别网络所训练出来的有效数据输入到有效点融合网络,得出最终的定位目标。仿真实验表明,与常规的基于多传感器的水声定位算法相比,该方法在速度和精度上都存在一定优势。

Aiming at the problem of multi-sensors acoustic positioning data fusion, the paper improved the conventional algorithms of data fusion based on BP neural network, and proposed an algorithm for acoustic locating based on double-BP neural network data fusion. First the algorithm sent the data, which were collected by sensors, into the recognition network to eliminate abnormal data. Then the data, which were obtained by the recognition network from training the network, were processed by fusion network about efficient points to gain the ultimate target position. Last Simulations are accomplished to prove this algorithm has certain advantages in both speed and accuracy compared with the conventional acoustic locating algorithm based on multi-sensors.