Automatic Detecting Outliers in Multibeam Sonar Based on Density of Points

Because of device noises, bad sea state or incorrect ship parameter, multibeam bathymetry data easily conceal many outliers. In order to process such large amount of data, we must research an automatic and rapid and robust approach. We present an automatic algorithm for detecting outliers based on density of points. Firstly, each swath data are projected along orthogonal and side direction respectively. On each plane, an initial point can be determined according to the corresponding maximum density. Then a big region will be searched by the connected neighboring points on each plane. Then we adopt erosion and dilation algorithms to eliminate a few outliers which connected with the big region. Afterward we obtain the edge of region by edge tracing. All data beyond of the region will be considered as outliers and deleted. Finally a local window filter is used to delete some outliers which conceal in the scope of real depth. The algorithm is verified by real data. It is a kind of rapid, robust algorithm.