Multi-Sensor Data Fusion: An Introduction

by H.B. Mitchell

This textbook provides a comprehensive introduction to the theories and techniques of multi-sensor data fusion. It is aimed at advanced undergraduate and first-year graduate students in electrical engineering and computer science, as well as researchers and professional engineers. The book is intended to be self-contained. No previous knowledge of multi-sensor data fusion is assumed, although some familiarity with the basic tools of linear algebra, calculus and simple probability theory is recommended.

282 pages, Paperback

First published January 1, 2007

Reviews

[Behrooz Parhami · Rating 4 out of 5]

Before the age of data science and data-directed everything, the discipline of data fusion brought together scientists and engineers who were concerned with combining data from multiple, diverse sources into a coherent whole to improve our understanding of a domain or system of interest (e.g., factory, air traffic, battlefield). This graduate-level textbook designed for a one-semester course summarizes the state of the field as of two decades ago. The book is composed of five parts.

Part I, Basics (Chs. 1-3): Introduction; Sensors; Architectures

Part II, Representation (Chs. 4-7): Common representational format; Spatial alignment; Temporal alignment; Sensor value normalization

Part III, Data Fusion (Chs. 8-13): Bayesian inference; Parameter estimation; Robust statistics; Sequential Bayesian inference; Bayesian decision theory; Ensemble learning

Part IV, Sensor Management (Ch. 14): Sensor management

Part V, Appendices: Software sources; Background material; References (360 citations); Index

A modern application of data fusion is seen in self-driving vehicles, which use data from a vast array of sources (maps, cameras, microphones, radar, and so on) to make navigational decisions. Combined results from multiple knowledge sources and real-time sensors can provide a more-accurate picture of the environment, leading to better decisions.

Data fusion is a specialized subfield of data science. While data science deals with the use of statistics, machine learning, data visualization, and algorithms/software to extract actionable knowledge from large, complex data sets using end-to-end analysis or predictive modeling, data fusion provides a useful toolbox for combining disparate, multi-modal data streams (text, video, sensor outputs) towards improving the accuracy of models.

Bayesian inference is a key element of data fusion, because starting from a state of prior knowledge (say, map, in the case of a self-driving vehicle), each data stream provides additional knowledge that allows for the refinement of the decision-making model.

Data fusion also has a significant overlap with the field of fault-tolerant computing (one of my research specialties), because of the inherent redundancy of the data coming from multiple sources. In fact, data diversity is a cherished method for ensuring system robustness and fault tolerance.

I end my review by referring the reader to a couple of newer books on data fusion:

- H. B. Mitchell's Data Fusion: Concepts, Ideas and Deep Learning (3rd ed., 2026) unifies classical fusion methods with modern deep learning and statistical estimation using real-life examples.

- Felix Govaers' Theory and Methods for Distributed Data Fusion Applications (2023) explores the mathematical models and applications for connecting multiple sensors and nodes in decentralized networks.