Automatic fruit inspection and classification in real time

In automatic fruit inspection and classification in real time modelling and segmenting the blemish from images and image sequences is an imperative pre-curser to a number of high-level computational tasks in computer vision such as fruit segmentation, identification and classification. This thesis investigates modelling methods and contributes new algorithms in this area. The research investigates advanced computer vision algorithms for fruit inspection and classification in real time. A new system architecture using advanced computer vision algorithms is introduced. The new methods with improve parameters of the existing industrial fruit sorters has increased their inspection and classification performance and ability. In general the fruits are classified for size, shape and skin blemishes. On the computer vision side the related methods include shape, surface and texture reconstruction form multiple views. On the pattern recognition side the methods applied include variety of neural networks including radial-basis functions and support vector machines. The Support Vector Machine (SVM) is a new and a capable classification method can be seen as substitute training technique for Polynomial, Radial Basis Function and Multi-Layer Perceptions classifiers. Support vector learning machines achieved high simplification ability by minimizing a bound on the expected test error; this new method has incorporated all known attributes, parameters, colour models, features and transformation of this invariance by applying transformations to support the vectors is achieved and tested for real-time application to allow fruit classification. This research work was supported in collaboration with the Industry Partner, Colour Vision Systems (CVS) and was funded and supported by Australian Research Council's (ARC) Discovery Projects funding scheme (project LP0560847). All the fruit image data, company’s research facility was used and at our discretion to modify and calibrate and setup a test bed at the CVS company’s research centre. [Author edited abstract].