Object Classification in a Distributed Environment

This chapter provides a limited overview of object classification approaches and explains a summary of classification architectures. It addresses distributed object classification issues, discusses classifier fusion, and also provides an extended discussion on distributed Bayesian classification and performance evaluation, as the Bayes formalism remains a central methodology for many classification problems. The chapter offers some perspectives on the structural design of an object classification process as determined by multiple observational data. In a broad sense, the basic processing steps for classification involve sensor-dependent preprocessing that in the multisensor fusion case includes common referencing or alignment, features and attributes (FandA) extraction, FandA association, and class-estimation. The measurement-based approach ideally would use combined raw sensor data at the measurement level to form information-rich features and attributes that would then provide the evidential foundation for a classification/recognition/identification algorithm.