Coal and Gangue Recognition Based on Color Names and KSVM

Recognition of the gangue from the coal is an essential step in automatic coal production. To solve the indistinguishability of coal and gangue in the low-dimensional color space, a kernel support vector machine (KSVM) model based on color names (CN) is proposed in this paper. First, the CN is used to map the original RGB image into a high-dimensional color space. It provides eleven color channels for description of the image. By use of the principal component analysis (PCA), two fundamental dimensional elements are selected to construct the principal color model. Then, through the combination with the gray component, an expanded color model with three channels is generated for the coal and gauge image with a higher contrast between the coal and gangue than the low-dimensional color model. Second, a screening mapping matrix is used to extract the contributing components of the gangue image. The so-called Mask-CN color model is derived thereafter capable of determining the gangue candidate area. Finally, six contributing features based on gray color histogram and gray-level co-occurrence matrix are selected and the SVM algorithm is adopted to realize the coal and gangue recognition. The experimental results show that the proposed method has provided a satisfactory recognition accuracy up to 97.79%. It shows a strong robustness to diverse background conditions.