DETECTING DEFECTS IN CHERRIES USING MACHINE VISION
Abstract
This thesis describes machine vision procedures which are able to classify defective cherries from non-defective cherries. Defects can be further divided into bruises, dry cracks, and wet cracks. First, bandpass filters that enhance the intensity contrast between bruised and unbruised cherries are determined. An optimum combination of two wavelengths is identified at 750 nm in the infrared range and 500 nm in the green range. An optimum single wavelength band is identified to be at 750 nm. The image acquisition using these filters is described. The gray level of a bruised surface (tissue) in the image is considerably lower than that of an unbruised surface. Four classification methods using single view infrared cherry images are studied, namely
The error estimates of the four classification methods are presented and methods (3) and (4) are shown to outperform methods (1) and (2). Methods (3) and (4) perform well in distinguishing bruised cherries and wet cracked cherries from non-defective cherries. However, they do not perform well in classifying dry cracked cherries from non-defective cherries. Another classification method is described, namely green image edge detection. This method classifies cherries based on images taken in the green range. This method performs well in classifying dry cracked cherries from non-defective cherries. However, it performs poorly in classifying bruised and wet cracked cherries from non-defective cherries. The infrared image edge detection and green image edge detection are used in combination to perform the classification on the entire surface of cherries. Two images, infrared and green, are taken from each of 6 orthogonal directions from the cherries. The integrated classifier misclassified 13% of non-defective cherries, 16% of bruised cherries, 0% of cherries with wet cracks, and 10% of cherries with dry cracks.
The thesis can be downloaded here. (msthesis.ps.gz, 6 Mbytes)