基于深度学习和单目视觉的采煤机滚筒测距算法

Ranging Algorithm of Shearer Drum Based on Deep Learning and Monocular Vision

提出了一种基于目标检测的采煤机滚筒位置测算方法,并进行了实验室验证。首先应用基于YOLOV4的目标检测算法训练出用以识别滚筒目标的模型。然后使用自动跟机技术保证滚筒位置在视频的中央部分来提高准确度,再将识别出的滚筒模型数据以及摄像头相关数据进行预处理。最后,通过单目视觉计算图像中滚筒与摄像仪的距离,推测采煤机位置。该方法已通过实验室验证,结果可满足生产需要,简单、高效,可移植性高,硬件需求低,实时性好。

A method for measuring the position of shearer drum based on object detection was proposed and was verified in the laboratory.Firstly,the target detection algorithm based on YOLOV4 was applied to train the model used to identify the drum target.Then used the automatic machine following technology to ensure that the drum position is in the center of the video to improve the accuracy,and then preprocessed the identified drum model data and camera related data.Finally,the distance between the drum and camera in the image is calculated by the monocular vision to conjecture the position of shearer.The method has passed the laboratory validation,and the results can meet the production needs.The method is simple and efficient,with high portability,weak hardware requirements and good real-time performance.