Quality approach for multi-parametric data fusion

Data fusion has gained widespread use in different areas of measurement in the last years, mainly where the sought output of an analysis cannot be measured directly—medical imaging, non-destructive testing, remote sensing, etc. The process of data fusion as such is not characterized today by any quality characteristics, which makes it impossible to estimate the effectiveness and reliability of the fused data. The authors propose a new approach to the process, free of this drawback. The introduced term of relevant information, with allowance for the first- and second-kind errors and their corresponding penalties, allows application of the Taguchi methodology via the loss function. The presented information scale, with the data-fusion types classified according to the information volume, ensures effectiveness of the planned process. The proposed approach is illustrated in examples of application.