The Research for A Kind of Information Fusion Model Based on BP Neural Network with Multi Position Sources and Big Data Selection

摘要

Positioning accuracy is a strong support for autonomous driving and intelligent transportation. In this paper, multi-position sources and single target positioning model based on big data selection and BP neural network data fusion is proposed. The model can reasonably and efficiently fuse multi-position sources information through the synergies of preprocessing, fusion and correction, to improve the positioning accuracy. Firstly, a specific error information elimination algorithm is proposed in the preprocessing stage to filter the data before positioning information fusion. Secondly, image positioning, which can provide accurate and reliable positioning information, is applied to the fusion stage and the post-fusion position correction stage. The fusion stage serves as the expected value of network training, and the correction stage uses the extracted image information such as angle and displacement to supervise the fusion data. The simulation results in Python3.6 show that the maximum position error of the model can be reduced by half than before and the model is more stable in the whole positioning process.