基于三维点云分割的煤堆体积计算方法研究

Research on the calculation method of coal pile volume based on 3D point cloud segmentation

我国是一个极度依赖煤炭的国家,对于煤炭的数量统计对我国发展具有重大现实意义。但是大型煤炭矿区的煤堆体积庞大,数量惊人,很难用传统的人工测量方法进行统计,为解决这一难题,本文提出了一种利用激光雷达并结合神经网络的体积测量方法。首先,利用激光雷达扫描煤堆表面绘制其三维点云图像,并结合Meshlab软件和Cloud Compare软件对点云进行降噪和补全,形成较为完整简洁的点云模型,并建立数据集;其次,训练Point-Net神经网络,利用训练好的网络模型对点云数据进行分割;最后,利用Matlab软件进行体积计算。通过模拟实验可知,相较于传统测量,该方法具有精度高、速度快、无需进行接触测量等优点,能够满足大型煤炭矿区煤堆的体积测量需求;相较于真实体积,该方法计算结果偏小,这是由于Point-Net神经网络未能识别出全部的煤堆点云数据,可以通过对神经网络进行后续训练来加以改进。

China is a country heavily dependent on coal, it is of great significance for coal quantitative statistics.But the coal heaps in large coal mining areas are so large and numerous that it is difficult to count them by traditional manual measurement.To solve the difficult problem of volume measurement of coal heap in large coal mine area, a volume measurement method based on Lidar and neural network is proposed in this paper.Firstly, 3 D point cloud image of coal heap surface is drawn by means of Lidar scanning, then noise reduction and completeness of point cloud are carried out by combining Meshlab and Cloud Compare software to form a more complete and concise point cloud model, and data set is established.Subsequently, Point-Net neural network is trained to segment the point cloud data by using the trained network model, and finally volume calculation is carried out by using Matlab software.The simulation experiment show that this method has the advantages of high precision, fast speed and no contact measurement compared with traditional measurement, and can meet the volume measurement requirements of coal heap in large coal mine area.Compared with the true volume, the calculation result of this method is smaller because Point-Net neural network can not recognize all the coal pile point cloud data and can be improved by subsequent training of the neural network.