Abstract: In view of the low recognition rate of the existing coal-rock interface recognition, an information extraction method of the coal-rock interface is proposed based on machine vision in this paper. In the case that there is little difference in physical mechanics properties and great difference in gray level between coal and rock, one image of the coal face is shot by a camera mounted on a shearer at a fixed time. The curve of the coal-rock interface is obtained by preprocessing and feature extracting, then the pixel coordinates in the image are gotten. After the camera calibration, the relationship between the pixel coordinates and the actual imaging angle is calculated based on the projection model of the shearer and the coal face. Then the height information of the coal-rock interface is calculated according to this relationship. Ultimately, according to the height information?obtained in each piece of the images, the entire coal-rock interface of the coal face is fitted out, which provides a basis for the attitude control of the shearer.

Key words: machine vision, feature extraction, coal-rock interface recognition, monocular vision measurement

摘要： 针对现有煤岩识别方法存在识别率低的问题，提出了一种基于机器视觉的煤岩界面信息提取方法。在煤和岩石物理力学性质差别不大，但灰度差异较大的情况下，利用采煤机机身上安装的摄像头拍摄采煤面，定时提取一帧图像，经预处理和特征提取获得煤岩界面曲线，进而获得其在图像中的像素坐标；标定出摄像头参数后，结合采煤机和采煤面特定的投影模型，求解出像素坐标与实际成像角度之间的关系，然后根据此关系求解煤岩界面的高度信息。根据每一幅图像中求得的高度信息，最终拟合出采煤面的完整煤岩界面，为采煤机姿态控制提供依据。

关键词: 机器视觉, 特征提取, 煤岩界面识别, 单目视觉测量