光学学报, 2020, 40 (24): 2411001, 网络出版: 2020-11-23
基于多光谱成像和改进YOLO v4的煤矸石检测 下载: 1689次
Coal Gangue Detection Based on Multi-Spectral Imaging and Improved YOLO v4
光谱学 多光谱成像 波段选择 目标检测 煤矸石智能分离 spectroscopy multispectral imaging bands selection object detection coal gangue intelligent separation
摘要
煤矸石分离对环境保护和资源高效利用具有重要意义,因此,提出了一种基于多光谱成像技术和目标检测的煤矸石智能分离方法。首先,在实验室搭建了煤矸石多光谱采集系统,共采集850组多光谱数据;其次,研究了多光谱中各波段煤矸石的识别率及相关性,从25个波段中选出3个波段构成伪RGB(Red,Green,Blue)图像;最后,用改进的目标检测模型YOLO v4.1检测煤矸石。实验结果表明,YOLO v4.1在测试集上检测煤和煤矸石的平均精度均值为98.26%,检测时间约为4.18 s。该方法不仅能准确识别出煤和煤矸石,还能获取两者的相对位置和大小,对煤矸石的分离操作具有重要意义。
Abstract
The separation of coal gangue from coal is of great significance for environmental protection and resource-saving. Therefore, this article proposes an intelligent separation method for coal gangue based on multi-spectral imaging technology and object detection. First, a multi-spectral data acquisition system for coal and coal gangue is set up in the laboratory, and 850 groups of multispectral data are collected. Second, by studying the coal gangue recognition rate and the correlation of each band of multi-spectral data, three bands from 25 bands are selected to form a pseudo-RGB (Red, Green, and Blue) image. Finally, the improved object detection model YOLO v4.1 is used to detect coal gangue. Experimental results show that the the mean average precision of YOLO v4.1 for coal and coal gangue detection on the test set is 98.26%, and the detection time is about 4.18 s. The method can not only precisely identify coal and coal gangue, but also obtain their relative position and size, which is important for the seperation operation of coal gangue.
来文豪, 周孟然, 胡锋, 卞凯, 宋红萍. 基于多光谱成像和改进YOLO v4的煤矸石检测[J]. 光学学报, 2020, 40(24): 2411001. Wenhao Lai, Mengran Zhou, Feng Hu, Kai Bian, Hongping Song. Coal Gangue Detection Based on Multi-Spectral Imaging and Improved YOLO v4[J]. Acta Optica Sinica, 2020, 40(24): 2411001.