基于小波域非对称广义高斯模型的煤岩识别算法

针对采煤工作面无人开采、煤与矸石自动分离等工程实际需求，研究了基于计算机视觉的煤岩识别技术。提出了一种有效的基于小波域统计建模的煤岩识别算法。通过小波变换对煤岩图像进行多级分解；提出表达煤岩纹理细节特征的高频子带系数统计分布符合非对称广义高斯模型的假设，并通过最大似然估计方法确定其模型参数；利用改进的相对熵相似性测度实现煤岩图像的分类判别。结果表明:在小波域中，非对称广义高斯模型能够有力地刻画煤岩图像的纹理特征，与现有的其他算法相比较，所提出算法具有更高的正确识别率，其平均识别率达到了87.77%，为进一步研究煤岩界面的自动检测提供了参考。

A coal-rock recognition algorithm using wavelet-domain asymmetric generalized Gaussian models

To address some practical engineering issues such as unmanned coal mining and automatic separation of coal and gangue，a computer vision based coal-rock recognition technology was investigated.An effective algorithm for coal-rock recognition through statistical modeling in wavelet domain was proposed.Firstly，coal or rock images were decomposed via multi-level wavelet transformation.Then，an assumption that high frequency subband coefficients representing texture details obeyed asymmetric generalized Gaussian distributions was made，and their parameters were estimated using maximum-likelihood estimator.Finally，a modified similarity measurement with respect to relative entropy was employed to distinguish between coal images and rock images.Experimental results demonstrate that the asymmetric generalized Gaussian models have strong discrimination power in terms of the representation of coal and rock texture features，and the proposed algorithm achieves higher correct recognition rates compared with several other existing algorithms.With the method proposed，the average correct recognition rate reaches 87.77%.The study could provide a new reference for further study on coal-rock interface automatic detection.