基于距离度量学习的煤岩识别方法

Coal-rock recognition method based on distance metric learning

摘要: 提出了一种基于距离度量学习的煤岩识别方法。该方法首先从煤岩图像训练集中提取煤岩图像特征；然后学习到特定的距离度量，使得煤样本特征间、岩石样本特征间距离变小，煤样本特征与岩石样本特征间距离变大，以提高分类识别效果；最后采用分类器进行煤岩识别。实验结果表明，对于煤岩样本图像的LBP特征、HOG特征、GLCM特征，与基于欧式距离、LDA、ITML的煤岩识别方法相比，该方法具有更高的煤岩识别率。

Abstract: A coal-rock recognition method based on distance metric learning was proposed. In the method, features of coal-rock images are extracted firstly from training sets of coal-rock images. Then a fit distance metric is gotten, which can make distance between any two features of coal samples or the ones of rock samples smaller and distance between features of coal samples and rock samples bigger, so as to improve classification and recognition effect. Finally, classifier is used to recognize coal-rock. The experimental results show when extracted coal-rock features are LBP, HOG or GLCM features, the method has higher coal-rock recognition rate than coal-rock recognition methods based on Euclidean distance, LDA or ITML.