3D Pollen particle recognition based on spatial geometric constraints histogram descriptors

This paper presents a new method for 3D pollen particle recognition based on spatial geometric constraints histogram descriptors (SGCHD). For reducing high dimensionality and noise disturbance, the surface curvature voxels are extracted as the primitive features instead of the original 3D pollen particles. The geometric constraints vectors are computed to describe the spatial correlations among the curvature voxels on the 3D pollen particle surface. The histogram algorithm is applied on the geometric constraints vectors to obtain the statistical histogram descriptors with fixed dimension. Experimental results verified the good invariance and robustness of our proposed descriptors on two pollen image databases.