A filtering method based SVM in the processing of multibeam data

Bathymetric data collection accompanied by all sorts of interference, the depth of these disturbances will bring the quality of the data and must be removed. In this paper, the effectiveness of an algorithm based SVM in identifying the potential outliers in multibeam data is proposed. The normal data detected according to 3σ rule are used as training sample. The erroneous data have been reconstructed by the SVM method. The SVM algorithm performance based different kernel function is discussed by experiment. The algorithm is tested and evaluated using synthetic as well as real field multibeam data. The results obtained show that the algorithm that detects most of the outliers with minimal data degradation is feasible and effective.