VISION INTELLIGENCE AND IMAGE FEATURE EXTRACTION FOR AGRICULTURAL PRODUCTS CLASSIFICATION

Abstract

Intelligent vision systems are the next generation in machine vision (MV). The addition of intelligence into vision systems requires an understanding and structuring of human visual techniques. Image analysis in conjunction with statistical pattern recognition• was used to discriminate defected fruit from good . ones. Multivariate discriminate . analysis was used to develop classification rules to aid in the identification of defects. Twelve parameters describing the shape, color distribution and size of the defect were considered as discriminators. This approach has been developed to simulate manual fruit classification by human sorters.

This study presents an approach to automate fruit sorting . usmg information that is acquired from a selected camera, which measure and quantifY parameters that are indicators of fruit quality (defects). The information is used as input to a role base detector that fmds different d fects in fruits surface. This approach achieved excellent results in the. defect detection of oranges.

. The application of digital imaging beyond the present day color sorting to include different defect detection has been demonstrated. This makes it possible to apply a uniform standard for orange quality determination free of human error in application of standards to the inspection process.