Location Technology Of Cutting Unit And Underground Boom-type Roadheader

Boom-type roadheader with compact structure and strong adaptability has become the main mining equipment in China’s coal mines,but the current coal mine comprehensive mining operation surface intelligence is low,the risk of manual mining,roadheader mining problems are prominent,and it is difficult to ensure the quality of roadway tunneling forming,so how to achieve automatic measurement of boom-type roadheader position posture is the key to solve the above problems.In this paper,we propose a boom-type roadheader position positioning scheme based on NIR binocular stereo vision and 1D convolutional neural network magnetic field positioning.The scheme consists of a NIR binocular stereo vision localization system and a magnetic field-assisted localization system based on a 1D convolutional neural network.The main idea of this system is to construct NIR vision extractable feature points on the body and arm of the cantilevered roadheader,i.e.,by installing NIR LED targets on the body and arm of the roadheader,and using dispersion and Euclidean distance methods to identify,extract and track the LED targets.By using binocular stereo vision measurement primitive to solve the 3D information of LED targets,the spatial point set of roadheader body and arm is established.Then,by establishing the mathematical model of boom-type roadheader attitude solution,the position of boom-type roadheader is solved.In order to solve the problem of failure of binocular stereo vision image acquisition,this system adopts a magnetic field-assisted positioning system based on one-dimensional convolutional neural network for the cutter section of the roadheader.The system is based on convolutional neural network technology,and a one-dimensional convolutional neural network position prediction model is constructed.The magnetic field data collected by the magnetic field sensors installed on both sides of the roadheader and the relevant position data of the roadheader measured by the near-infrared binocular stereo vision system are used as the training set and test set for deep learning to realize the position prediction of the boom-type roadheader.An experimental platform is built to verify the theory and scheme proposed in this paper.The experimental results show that the spatial position error of the cutter section of the tunneling machine based on the NIR binocular stereo vision measurement scheme is within 50 mm,and the absolute error of the position is within 1°.In case of failure of binocular stereo vision measurement,the magnetic field-assisted localization scheme based on one-dimensional convolutional neural network predicts the trajectory of the cutter section better,and the prediction accuracy of pitch angle and yaw angle is over99%.Meet the requirements of underground coal mine tunneling machine operation.