煤炭工程 ›› 2022, Vol. 54 ›› Issue (7): 159-163.doi: 10.11799/ce202207028

• 研究探讨 • 上一篇    下一篇

基于光学图像的煤矸石识别方法综述

张红,李晨阳   

  1. 西安科技大学
  • 收稿日期:2021-08-04 修回日期:2021-09-22 出版日期:2022-07-15 发布日期:2022-07-20
  • 通讯作者: 李晨阳 E-mail:19207205077@stu.xust.edu.cn

A review of coal gangue identification methods based on optical images

  • Received:2021-08-04 Revised:2021-09-22 Online:2022-07-15 Published:2022-07-20

摘要: 基于光学图像的煤矸石识别方法具有设备简单、易实现、绿色环保等优势,是实现智能化煤矸石分选的重要途径。该类方法分为两种研究路径,一种是需要人为提取特征进行识别的路径,一般包括煤矸图像数据采集、图像预处理、特征选择与提取和煤矸识别|另一种是利用深度学习神经网络进行自主提取特征识别的路径。文章对这两种研究路径的各类方法进行了总结,指出现有识别方法存在煤矸图像数据集不完备不充分、特征理解不全面不深入、识别方法无法兼顾高效与实时性等缺点,给出进行高效煤矸石识别的建议。

关键词: 煤矸石识别, 图像识别, 特征识别, 机器学习, 深度学习

Abstract: The coal gangue identification method based on optical image has the advantages of simple equipment, easy realization, greenness and environmental protection, and it is an important way to realize intelligent coal gangue separation.There are two research ways. One way requiresextracting artificial features for recognition, which generally includes four steps: coal and gangue image data set collection, image preprocessing, feature extraction and selection, and coal and gangue recognition. The other way uses deep learning neural network to independently extractfeatures.This paper summarizes the different methods in the two research ways, and points out that the existing identification methods have shortcomings such as incomplete coal gangue image data set, incomplete feature understanding, and failure to give consideration to both efficiency and real-time performance. Suggestions for efficient coal gangue identification are put forward.

中图分类号: