摘要
针对不同定位体制的多系统协同定位能力不足的问题,提出一种基于深度学习的数据级多源融合定位增强方法。该方法从多源定位数据出发,以空间关联行为更加丰富的二阶特征矩阵作为网络输入,通过设计的5L-CNN多源融合定位增强网络一体化完成数据特征自主提取和融合预测,实现多源信息融合的目标定位精度提升。仿真结果表明该算法至少可对两种以上不同定位体制的多源定位数据进行融合增强,且具备实时融合定位能力。
Aiming at the problem of insufficient multi-system co-location capability of different positioning systems,a data-level multi-source fusion positioning enhancement method based on deep learning is proposed.This method starts from multi-source positioning data,and uses the second-order feature matrix with richer spatial correlation behavior as the network input.Through the designed 5L-CNN multi-source fusion positioning enhancement network,the autonomous extraction of data features and the fusion prediction are integrated to achieve the improvement of target positioning accuracy based on multi-source positioning data.Simulation results show that the algorithm can at least fuse multi-source positioning data from two or more different positioning systems and has real-time fusion positioning capabilities.
作者
丛迅超
Cong Xun-chao(Southwest China Institute of Electronic Technology,Sichuan Chengdu 610036)
出处
《电子质量》
2020年第4期13-16,共4页
Electronics Quality
关键词
深度学习
数据级
多源融合定位增强
卷积神经网络(CNN)
deep learning
data-level
multi-source Fusion positioning enhancement
convolutional neural network(CNN)
