A New High Reliability and Dual Measure Method for Multi-system/sensor Remote-sensing Decision Fusion

In this paper we will introduce a new high reliability multi-system/sensor decision fusion scheme based on dual measure calculations and formulations. The data are collected from remote sensing of the ground targets in different spectral bands including visible, near infrared (NIR), IR, thermal, and microwave by multi-system/sensor systems. At first, we will review the decision fusion methods such as voting methods, rank based algorithm, Bayesian inference, and Dempster-Shafer combination scheme. We show that the essential and common weaknesses of these formal methods are ignoring the class correlation of local classification results and classification error distributions for all classes at different pixels. Then by establishing the commission and omission errors distribution vectors and matrixes, we will formulate and introduce a new dual measure decision fusion (DMDF) algorithm. Formulation the similarity and correlation of local classification results and errors for different classes and need to hard decisions, can be considered as the main features of DMDF. The assumption of uncorrelated errors is not necessary for DMDF, because an optimal class selector always selects the most appropriate class for each pixel. Finally, we deploy these methods for fusion of local classification results, obtained from remote sensing in 12 different spectral bands. In commission and omission errors viewpoints, we will obviously show that the DMDF method is more accurate and reliable than other methods.