基于深度视觉原理的工作面液压支架支撑高度与顶梁姿态角测量方法研究

Measurement method of support height and roof beam posture angles for working face hydraulic support based on depth vision

针对工作面液压支架姿态传统接触式测量方法存在的传感器数量多、故障率高、解算误差大、可视性差等问题,建立了一种基于支架运动机理的液压支架姿态视觉测量模型,提出了一种基于深度视觉原理的液压支架支撑高度与顶梁姿态角测量方法,并搭建了测量系统。该测量方法通过在液压支架顶梁上布设RGB-D相机,利用深度视觉技术获取液压支架底座彩色信息和深度信息并实现二者数据流的同步,通过ORB算法和FLANN算法实现底座特征点的快速提取与匹配,基于空间3D-3D的ICP模型估计相邻时刻相机位姿变化,结合液压支架姿态视觉测量模型设计姿态解算算法完成支撑高度和顶梁姿态角参数解算。测量实验结果表明:该方法液压支架姿态参数解算速度快,稳定性好,实现了液压支架内部姿态的非接触实时性监测,同时该姿态解算算法可保证液压支架支撑高度测量绝对误差在10 mm左右,顶梁姿态角测量绝对误差在2°以内,精度上能够满足井下液压支架姿态监测精度需求;此外,该方法减少了支架上传感器的数量,降低了安装与维修成本,为实现工作面综采设备运行位姿状态信息的精准感知与动态监测,以及智能化工作面的精准实时控制提供了技术支撑。

In view of large number of sensors,high failure rate,large calculation error,and poor visibility of traditional contact measurement method of hydraulic support posture on working face,a visual measurement model for posture of hydraulic support based on movement mechanism was established,a depth vision measurement method for support height and roof beam posture angles based on visual measurement model of hydraulic support was proposed,and a corresponding measurement system was built.An RGB-D camera was installed on roof beam to obtain color and depth information and realize synchronization of the two data streams.Then,ORB algorithm and FLANN algorithm were employed to achieve rapid extraction and matching of base feature points.ICP model based on spatial 3 D-3 D was employed to estimate camera pose changes at adjacent moments.Algorithm was designed to complete calculation of support height and roof beam posture angles combined with visual measurement model of hydraulic support.The experimental results have shown that this measurement system has fast calculation speed and good stability,and it realizes non-contact real-time monitoring of support internal posture to ensure that support height measurement absolute error is about 10 mm,and that roof beam posture angles measurement absolute error is within 2°,which could satisfy requirements of monitoring accuracy.In addition,this method reduces the number of sensors on support and cost of installation or maintenance,which provides technical support for realizing accurate perception and dynamic monitoring of fully mechanized mining equipment operation information and controlling intelligent working face effectively.