Analysis of Image Clusterization Methods for Oceanographical Equipment

The method of visual analysis, which provides detection of contours and clusterization of objects of interest optimized for real-time operation is offered in this paper. The method is intended for use in oceanographic equipment to estimate the abundance and dynamics of plankton spatial distribution in seas and oceans in situ. The structural elements of the monitoring system are given and their mathematical model is described. The comparison of the recursive clusterization method and the method of K-means clustering by the criteria of correct and erroneous classification coefficients was made based on plankton video analyzer using OpenCV framework. The results of practical tests of the suggested method show significant reduction in nuisance tripping factor at the expense of insignificant increase in the processing time and the volume of memory used. These aspects meet the requirements for work in high-load Big Data processing systems, standalone image processing systems designed for long-term deployment.