Research on terrain suitability of iterative closest contour point algorithm for underwater navigation

An alternative approach to underwater navigation - terrain-aided navigation (TAN) - is described in this paper. The focus is not on the TAN algorithm per se but in deriving a quantitative relationship of the matching errors of a TAN algorithm with terrain factors. The TAN algorithm used in this paper is known as the "iterative closest contour point" (ICCP) algorithm. Firstly, the ICCP algorithm is run 10 times along each path across 20 regions to collect 207 matching points. 182 points are used in the analysis and the other 25 points are used for verification. Then 6 fctors, among the total 17 factors, are identified as being most correlated with the accuracy of the ICCP algorithm by cluster and correlation analysis. Finally, three statistical methods, including multiple regression, logistics analysis and discriminant analysis, are applied to derive three different types of relationship between the terrain factors and the matching errors. Each formula uses no more than 3 terrain factors to fit the matching errors. The verification with the 25 points demonstrates the effectiveness of the formulas. The relationships derived herein are more practical, due to their simplification and effectiveness, in applications such as trajectory planning.