SENSOR CALIBRATION USING NONPARAMETRIC STATISTICAL CHARACTERIZATION OF ERROR MODELS

Calibration is the process of identifying and correcting for thesystematic bias component of the error in sensor measurements.Traditionally, calibration has usually been conducted byconsidering a set of measurements in a single time frame andrestricted to linear systems with the assumption of equal-qualitysensors and single modality.The basis for the new calibration procedure is to construct astatistical error model that captures the characteristics of themeasurement errors. Such an error model can be constructedeither off-line or on-line. It is derived using the nonparametrickernel density estimation techniques. We propose fouralternatives to make the transition from the constructed errormodel to the calibration model, which is represented by piecewisepolynomials. In addition, statistical validation andevaluation methods such as resubstitution, is used in order toestablish the interval of confidence for both the error model andthe calibration model. Traces of the distance rangingmeasurements recorded by in-field deployed sensors are used asour demonstrative example.