Apple Shape Inspection with Computer Vision

Two shape separation methods were compared for shape grading of 'Golden Delicious' apples using machine vision. The first one was based on moments and other geometric parameters while the second implied the computation of Fourier descriptors.Both methods required that one stem view and six cheek views of the fruit were presented to the camera. The Fourier descriptors were found the most efficient method since they allowed an accuracy in classification reaching 96% and can be made invariant to translation, rotation and scale. Furthermore, they are obtained from boundary information and do not need to examine the whole area of the apple. They permit the reconstruction of the apple shape from the harmonics.