基于草图输入的煤矿巷道建模方法的研究.pdfVIP

摘 要 近年来，矿山的数字化建设提高了矿产的使用水平和社会经济效益。矿山信 息化过程的主要工作是建立数字矿山系统，而煤矿巷道则是矿山的重要组成部分。 简单，快速的巷道建模方法将会降低数字矿山工程的开发难度，缩短数字矿山工 程的开发周期，并降低矿山信息化工程的成本。 在以往研究的基础上，本文对基于手绘草图的煤矿巷道建模进行了研究，包 括以下主要任务： 1、研究了可变形卷积核，分析其在煤矿巷道横截面识别中的优缺点，将 ResNet18 中的固定卷积核替换为可变形卷积核，增强模型对于手绘笔画的建模 能力，并通过对比实验加以证明。 2 、研究了去耳法三角剖分，分析其在巷道横截面生成过程中的优缺点，将去 耳思想用于巷道建模的三角剖分方法中，提高三角剖分效率，并且通过对比实验 加以证明。 3、设计了一个基于草图输入的煤矿巷道模型生成系统并且予以实现，该系统 具有根据用户横截面轮廓输入和扫描线输入生成煤矿巷道模型的功能。对该系统 进行了测试，分析研究中可能的不足之处并且在后续研究中进行改进和完善。 关键词：卷积神经网络；草图；煤矿巷道；草图识别；三维建模 I Abstract In recent years, the digital construction of mines has greatly accelerated social and economic development. The main work of the mine informatization process is to establish a digital mine system, and the coal mine tunnel is an important part of the mine system. Simple and fast tunnel modeling methods will make the development of digital mine engineering easier, the development cycle of digital mine engineering quicker, and the cost of mine information engineering lower. Based on the previous research, this paper studies the modeling of coal mine tunnels based on hand-drawn sketches, including the following main tasks: 1. It studied the deformable convolution kernel, including its advantages and disadvantages in coal mine tunnel cross-section identification. The fixed convolution kernel in ResNet18 is replaced with a deformable convolution kernel to enhance the modeling ability of the model for hand-drawn strokes. The advantage is proved through comparative experiments. 2. It studied the de-earing triangulation, including its advantages and disadvantages in the process of roadway cross-section generation. The de-earing idea is used in the triangulation method of tunnel modeling to improve the triangulation efficie