A Linear Discriminant Analysis framework for bacterial type identification based on morphological parameters

The objective of the current work is to develop an automatic tool to identify microbiological data types using computer vision and pattern recognition. Current systems rely on the subjective reading of profiles by a human expert. This process is time-consuming and prone to errors. Bacteriophage (phage) typing & Fluorescent imaging methods are used to extract representative feature profiles and identify the bacterial types. For feature selection of Bacterial identification system, the most successful method seems to be the appearance-based approach, which generally operates directly on images or appearances of bacterial objects. The image segmentation, Linear Discriminant Analysis (LDA), Direct Fractional LDA (DFLDA) and Principal Component Analysis (PCA) are the powerful tools used for feature extraction. Then the principal components are analyzed by DFLDA and simple Nearest Neighbor Classifier technique is used to identify the type of bacteria. The trained feed forward back propagation neural network is used for validating and testing the bacterial images. The effectiveness of the proposed method has been verified through experimentation using fifty popular bacterial image databases.