A machine vision system using circular autoregressive models for rapid recognition of salmonella typhimurium

Abstract

The objective of this research was to develop a machine vision system using image processing and statistical modeling techniques to identify and enumerate bacteria on slides containing Salmonella typhimurium. Pictures of bacterial cells were acquired with a CCD camera attached to a motorized fluorescence microscope. A shape boundary modeling technique, based on the use of circular autoregressive model parameters, was used. A feature weighting classifier was trained with ten images belonging to each shape class (rod shape and circle shape). In order to enhance the discrimination of circular shapes, a size range was added to the recognition algorithm. Experimental results showed that the model parameters could be used as descriptors of shape boundaries detected in digitized binary images of bacterial cells. The introduction of the rotated coordinate method and the circular size restriction, reduced the differences between automated and manual recognition/enumeration from 7% to less than 1%. The computer analyzed each image in approximately 5 s (a total of 2 h including sample preparation), while the bacteriologist spent an average of 1 min for each image.