Abstract:
For poor universal applicability of traditional coal-rock recognition methods whose application was limited by mining techniques and coal seam conditions, a coal-rock recognition method based on statistical modeling of steerable pyramidal decomposition coefficients was presented. Firstly, multi-scale steerable pyramidal decomposition was conducted on coal and rock images. Then, asymmetric generalized Gaussian distribution was adopted as a statistical model to fit coefficients of every steerable directional subband, and parameters of the asymmetric generalized Gaussian distribution were obtained by means of the maximum-likelihood estimation. Finally, symmetric relative entropy was employed as distance metrics to complete automatic identification of coal and rock images. The experimental results show that the method has high accuracy rate of coal-rock recognition than that of existing ones, which achieves 86.90%.