DISTRIBUTED SENSOR NETWORKS AND NEURAL TREES FOR MULTISENSOR DATA FUSION IN COMPUTER VISION

摘要

This chapter describes a Distributed Sensor Network (DSN) that is applied to integrating data coming from multiple sensors in the context of important computer vision tasks such as object recognition and scene understanding. The concept of multisensor data fusion is not new, and the advantage of using several sensors has been demonstrated in several application domains. Different sensors are sensitive to different properties of the environment, each one of which can contribute significantly in interpreting the environment itself. Humans and animals have evolved the capability to use multiple senses to improve their ability to survive. The human brain is an excellent example of a data fusion system that performs extremely well. For example, sight, sound, smell, taste and touch data are combined to determine whether a food item is rotten. Sight and sound data are integrated to make decisions regarding safety in an unknown environment. Multisensor data fusion can be defined formally as a complex process which refers to acquisition, processing and combination of information gathered by various knowledge sources and sensors to provide a better understanding of the phenomenon under consideration. In recent years, thanks to the development of new sensors and processing hardware which make possible the real-time fusion of data, multisensor data fusion has been an area of intense research activity with applications in numerous fields. The applications can be classified into two groups, namely military and nonmilitary applications. Military applications includes: automated recognition of military entities such as ships, aircraft, weapons and missiles, guidance of autonomous military vehicles, battlefield surveillance, air-to-air and surface-to-air systems, and so on. Nonmilitary applications includes air traffic control, law enforcement, robotics, manufacturing, medical diagnosis and remote sensing.