Phytoplankton feature extraction from microscopic images based on SURF-PCA

Phytoplankton are photosynthesizing microscopic organisms that inhabit the upper sunlit layer of almost all oceans and bodies of fresh water. Its abundance and taxonomic composition have impact on marine ecosystem dynamic and global environment change. Therefore, many researchers pay attention to effectively monitoring phytoplankton quantity and species composition. In order to realize this purpose, phytoplankton recognition using computer vision is now arousing more and more interest. In this paper, we combine two feature extraction methods, including SURF-PCA and LBP-PCA, to identify phytoplankton. SURF has rotational invariance and LBP is a very efficient texture operator. So combining the two algorithms can take advantage of the complementary properties to extract features, and the classification result shows that this approach is fast and can get a high recognition rate. In order to reduce the computational complexity and obtain a higher recognition rate, we use Principal Component Analysis algorithm for the feature dimension reduction. The experiment results demonstrate that our method can reduce computational complexity without loss of accuracy.