基于卷积神经网络的煤矸石识别算法研究  

Research on coal gangue recognition algorithm based on convolution neural network

作　　者：倪云峰[1] 封子杰 郭苹 王静[1] NI Yunfeng;FENG Zijie;GUO Ping;WANG Jing(College of Communication and Information Engineering,Xi’an University of Science and Technology,Xi’an 710054,China)

机构地区：[1]西安科技大学通信与信息工程学院,陕西西安710054

出　　处：《现代电子技术》2022年第10期57-62,共6页Modern Electronics Technique

基　　金：国家自然科学基金项目(61701393)。

摘　　要：煤矸石自动分选是充分利用能源和减轻环境污染的有效途径,然而传统的煤矸石分选方法效率低下、污染环境、成本过高,已不能满足当今智慧矿山的发展需求。为了提高煤矸石自动化分选的效率及识别精度,文中提出一种基于卷积神经网络的煤矸石智能识别算法。首先利用同态滤波增强煤和矸石的对比度;然后通过颜色空间转换得到HSV色彩,并利用K-means++聚类算法在HSV颜色空间对其进行图像分割,从而获得煤和矸石目标图像;最后构建无参数辨识的卷积神经网络模型,以解决人工选择煤矸石图像特征参数所导致的精度低、耗时长等问题,从而实现对煤和矸石图像的自动分类识别功能。实验结果表明:文中改进算法能够较好地对煤矸石样本图像进行分类识别,识别准确率达到93.3%;与多种先进的机器学习和深度学习煤矸石分选算法相比,改进算法的煤矸石分选识别准确率有明显提高。Automatic sorting of coal gangue is an effective way to make full use of energy resources and reduce environmental pollution. However,the traditional coal gangue sorting method has the problems of low efficiency,environment pollution and high cost,and cannot meet the development needs of smart mines nowadays. In order to improve the efficiency and recognition accuracy of the automatic sorting,a coal gangue intelligent recognition algorithm based on convolutional neural network is proposed. The homomorphic filtering is used to enhance the contrast between coal and gangue. The HSV color information is obtained by means of the color space converting,and the image is segmented in HSV color space by means of Kmeans ++ clustering algorithm,so as to obtain the target images of coal and gangue. The convolutional neural network model without parameter identification is constructed to solve the problems of low precision and long time-consumption caused by the manual selection of coal gangue image characteristic parameters,so as to realize the automatic classification and recognition functions of coal and gangue images. The experimental results show that the improved algorithm can better classify and recognize the coal gangue sample images,and the recognition accuracy is 93.3%. In comparison with many advanced machine learning and deep learning coal gangue sorting algorithms,the accuracy of coal gangue sorting recognition of the improved algorithm is significantly improved.

关 键 词：煤矸石分选 卷积神经网络 图像识别 图像增强 图像分割 自动分类 

分 类 号：TN711-34[电子电信—电路与系统] TP391.4[自动化与计算机技术—计算机应用技术]