Shape classification using a radial feature token

This paper presents a process for classifying shapes in digital images using a radial feature token. This shape classification process has been implemented as the ALISA/sup (R)/ shape module, the third module in the adaptive learning image and signal analysis (ALISA) system hierarchy, which also includes geometry and texture classification modules. Shape classification in images is a challenging problem because the basic shapes in an image can occur in any position, at any orientation, and at any scale. For this reason, translation, rotation, and scale invariance is a critical property of this or any general-purpose shape recognition system. The ALISA shape module learns to recognize shapes from a supervised set of training images. These learned shapes are stored as a set of vectors that are then used to classify shapes in test images. Results indicate that this process can learn to classify shapes from small training sets and then classify similar shapes despite extraneous edges or partially overlapping shapes. The radial feature token also enables the ALISA shape module to classify some shapes that are only partially visible or that have gaps in their edges.