Automated Detection and Recognition of Specimens in Ichthyoplankton Images

This toolbox implements a set of methodologies for detecting plankton specimens in cluttered images and recognizing their taxa. The functionalities of the developed system are divided into three major procedures [6]: 1) First, a novel image segmentation algorithm is applied to extract salient objects from a set of input cluttered images. The employed methodology, based on a novel nonparametric Bayesian approach to hidden Markov random field models [7], the infinite hidden Markov random field (iHMRF) model, is designed to effectively handle images containing multiple objects as well as significant levels of noise. 2) Subsequently, a set of low-level image descriptors (features) are computed, so that each extracted blob is represented as a feature vector of characteristic image features. Here, the extracted feature vectors include shape histograms, blob solidity, Hu moments up to third order, Fourier descriptors, and the circular projection descriptors defined in [3, 5]. 3) Finally, a set of advanced machine learning methodologies are used to select those of the extracted ogbjects that correspond to plankton images, and determine their most likely types, choosing from a set of predefined plankton categories previously learned from the system.