Multi-Sensor Data Fusion: An Overview of Concepts, Methods, and Classifications

In recent years, the advancement of sensor-based systems has increasingly relied on real-time heterogeneous data from multiple sources. Therefore research on data fusion has been a key to enable the interpretation of large amounts of data in various scientific and engineering fields. It involves integrating multiple data sources to produce more reliable, accurate, and meaningful information than any single source would provide independently. The primary objective of this paper is to explore key concepts and classification schemes in data fusion while also examining traditional methods and recent advancements. It presents a systematic literature review of recent data fusion methods and highlights key standards, tools, and metrics associated with their use. It explores multi-sensor data fusion across diverse application domains, including Road Condition Monitoring (RCM), Intelligent Transportation Systems (ITS), among others. Additionally, it presents several data fusion techniques based on whether the fusion of the data is operated at the sensor, feature, or decision level. The paper also shows how the performance of these techniques was assessed on several data sets based on relevant metrics. A last focus is given to identifying relevant research that can propose solutions to road condition monitoring, to establish a solid foundation for future works in this area.