基于字典学习的煤与矸石图像特征识别方法

Feature Recognition Method of Coal and Gangue Image Based on Dictionary Learning

为了提升采煤过程中煤与矸石图像的自动识别效率,解决现有图像特征提取方式的瓶颈,根据相应的机器视觉原理,提出了利用字典学习算法对煤与矸石的图像进行特征提取的方法。字典学习算法经过煤与矸石图像的预处理之后,在字典初始化与更新时采用随机选择的方式,调整到最优的字典列数、稀疏误差和稀疏度等参数,最后利用支持向量机进行识别。实验表明:利用字典学习算法可以有效地表达煤与矸石的图像特征,自动识别效果较好,为采矿过程中煤与矸石的自动分选提供了一种新的思路和方法。

In order to improve the efficiency of automatic recognition of coal and gangue image and solve the bottleneck of image feature extraction,the dictionary learning algorithm was proposed to apply to the extraction and recognition of coal and gangue image feature by using the principle of machine vision.Firstly,the dictionary learning algorithm was used to preprocess the image of coal and gangue.Then parameters such as optimal number of dictionaries,sparse errors and sparsity were adjusted during initializing and updating the dictionary by using the method of random selection.Finally,support vector machine was used for recognition.Experimental results show that the dictionary learning algorithm,with effective expression of coal and gangue image features and better automatic recognition effect,provides a new idea and method for the automatic separation of coal and gangue in coal mining.