Variety identification by computer vision

The objective of the paper reported was the development of a machine vision system and method to identify varieties, especially pear cultivars on the basis of shape and color characteristics. According to the results both the FFT and the PQS shape analysis methods were found to be suitable for the purpose. However the introduction of the ground color analysis improved the goodness of the results and the combination of the shape and ground color analysis was found to be suitable for both the identification and the selection of the five pear cultivars tested. However the use of the three mentioned parameters by the means of the discriminant analysis provided with a better efficiency for either identification, or for selection, than either the shape, or the color analysis.