A Key Point Method for Data Registration for MultiSensor Fusion

It has been recognized that multi-sensor data fusion can provide a more holistic, accurate and reliable information of the measured surface. Data registration, which is used to align data into one coordinate system, is a key step of data fusion. Widely used feature-based methods find correspondence between features, and then a geometrical transformation is determined to map the target data to the reference data. Reliable and accurate feature selection is thus very important for data registration. In this research, a reliable key point method called Scale Invariant Feature Method (SIFM) for data registration is investigated. By using this method, for each data, one can build a set of feature descriptors of the defined key points, which have the scale/shift/rotation invariant properties. Then the correspondence of two data and geometrical transformation can be achieved by finding the matching of two feature descriptors through closeness measurement. Initial tests on freeform and structured surfaces have proven the effectiveness and efficiency of the method.