Research on Probability Statistics Method for Multi-sensor Data Fusion

Abstract

In multi-sensor systems, data fusion is one of the key technologies for solving information diversification in wireless sensor networks. Data fusion is a process of information processing to automatically analyze and synthesize data collected by multiple sensors under certain rules to complete the required decisions or tasks, including information fusion, feature fusion, relationship fusion and decision fusion. It extends the lifespan of wireless sensor networks and improves data accuracy. It is generally considered that data fusion is an integrated process of information processing. It is generally considered that data fusion is a process of information synthesis and processing, making various information and data detected, correlated, estimated, and synthesized at multiple levels and from many aspects to obtain accurate and complete information. There are many methods for sensor data fusion, such as Bayesian method, D-S method, neural network, fuzzy reasoning, genetic algorithm, deep learning, etc. This article focuses on the application, analysis and comparison of probabilistic statistical methods in multi-sensor data fusion. The data fusion methods of probability statistics are divided into three categories: data fusion method based on estimation theory, data fusion method based on regression theory, and data fusion method based on information theory. This article just has a simple analysis on the three types from the perspective of theory and has a detailed analysis on the core Bayesian fusion in probability statistics.