基于DeeplabV3+网络和注意力机制的矿区地表隐患识别——以塔山煤矿30507工作面为例

Identification of Surface Hazards in Mining Areas Based on DeeplabV3+Network and Attention Mechanism:Taking 30507 Working Face of the Tashan Coal Mine as an Example

为解决矿区地表隐患监测问题,以塔山煤矿30507工作面为背景,提出一种快速监测和识别矿区地表隐患的方法。利用无人机搭载传感器和倾斜摄影测量技术,构建矿区地表三维实景模型,生成正射影像和空间点云。提出基于DeeplabV3+网络和注意力机制的裂隙识别方法,改进效果优于当前主流网络。提出基于三维空间点云的沉陷分析法,用密度噪声空间聚类算法去噪,通过基于KD Tree的K近邻算法和基于Delaunay三角网的点云表面重建技术进行沉陷分析,实现矿区地表数据监测。

In order to solve the problem of monitoring hidden dangers on the surface of the mine area,a method for rapid monitoring and identification of hidden dangers on the surface of the mine area is proposed with the background of the 30507 working face of the Tashan Coal Mine.Using UAV-mounted sensors and tilt photogrammetry technology,a three-dimensional live model of the mine surface is constructed,and orthophotos and spatial point clouds are generated,a fissure identification method based on the DeeplabV3+network and the attention mechanism is proposed,and the improvement effect is better than that of the current mainstream network,a subsidence analysis method based on the three-dimensional spatial point clouds is proposed,and the noise is removed with the density-noise spatial clustering algorithm,and the noise is removed through the KD Tree-based K-nearest neighbour algorithm based on KD Tree and point cloud surface reconstruction technology based on Delaunay triangulation network for subsidence analysis,to achieve the monitoring of surface data in the mining area.