In this research, the main purpose is the application of an automatic classification method for six kinds of 3D Drosophila Calyx Images. We have six different kinds of image data. We will use extracted features describing the spatial dispersion and connectivity of 3D olfactory neuron
pathway for classification. It is worth noting that much of the image data contain redundant information so we determine the essentialness of these features by cross validation accuracy. In the leave-one-out cross validation analysis, a six-category SVM classifier is three times better than random guess. Besides, there is no evidence of over-fitting, because compared to 3D spatial RST-invariant feature set alone, the 64-view Rotational Skeleton Endpoint feature set together with it raises the accuracy rate.
