Data Reduction Algorithm Based on Planar Surface Fitting

Data reduction tools are developed and evaluated using a data analysis framework. Simple and intelligent thinning algorithms are applied to both synthetic and real data and the thinned datasets are ingested into an analysis system. A major problem of data reduction is that certain types of camera or scanner produce vast amounts of data, the processing of which presents serious problems. Rather than process all of this data at every stage of the representation process, an alternative is to use a strategy in which the data is initially reduced, then a preprocessing can be completed without consuming a lot of time. This paper presents an algorithm for managing the amount of point data acquired by laser scanner. The proposed algorithm includes a method based on computing the surface normal which is fundamental in the most of reverse engineering algorithms. The normal vectors are calculated by fitting the best fit plane to the neighborhood. A point is assigned to normal and the angle between an arbitrary direction and the normal is obtained. The point data is subdivided into cells based on the angles, while the non-uniform cells are obtained. Thus, the amount of points can be reduced by sampling the representative points for each cell. Experimental results show that the proposed method has good results and appears to be quite stable even for large scale data reduction.