Development a zooplankton recognition method for dark field image

In this paper, an automated method for detection and recognition of marine plankton in the dark field images is proposed and evaluated. The features extracted with shape, invariant moment and texture information are used for the support vector machine (SVM) classifier. Three species of planktons were used for the performance evaluation and 86% classification accuracy was achieved. It is shown that the recognition accuracy based on watershed algorithm is better than that method based on adaptive threshold segmentation algorithm. The result shows this is a promising method for real-time application.