A multivariable optical remote sensing image feature discretization method applied to marine vessel targets recognition

摘要: The effective extraction of continuous features in ocean optical remote sensing image is the key to achieve automatic detection and identification for marine vessel targets. Since many existing data mining algorithms can only deal with discrete attributes, it necessary transform into ones adapting these intelligent algorithms. However, most current discretization methods do not consider mutual exclusion within attribute set when selecting breakpoints, cannot guarantee that indiscernible relationship information system destroyed. Obviously, they are suitable processing multiple features. Aiming at this problem, a multivariable feature method applied targets recognition presented paper. Firstly, equivalent model established based on theories entropy rough set. Secondly, change extent before after evaluated. Thirdly, scans executed each band until termination condition satisfied generating optimal number intervals. Finally, we carry out simulation analysis high-resolution collected near coast South China Sea. In addition, also compare proposed mainstream Experiments validate has better comprehensive performance terms interval number, consistency, running time, prediction accuracy rate.